Conceptions of Play and the Evolution of Animal Minds

Contents

Hubert Hendrichs 98 Accessing Animal Minds: Epistemological and Empirical Problems

Gordon G. Brittan, Jr. 105 The Mental Life of Other Animals

Gordon M. Burghardt 114 Conceptions of Play and the Evolution of Animal Minds

Colin Allen 123 Animal Concepts Revisited

Marc Bekoff 128 Social Cognition: Exchanging and Sharing Information on the Run

Gerhard Roth 137 The Neurobiological Basis of Consciousness in Man and Animals

Franz Seitelberger 149 Neurobiological Comments on Empirical Consciousness

Gunther S. Stent 156 Are Animals Persons?

Holk Cruse 162 Feeling Our Body—The Basis of Cognition?

Hans-Johann Glock 174 Animal Minds: Conceptual Problems

Rolf Lachmann 189 Animal Perception as Value Perception

Hans Lenk 198 Humans as Metasymbolic and Superinterpreting Beings

Theodore R. Schatzki 204 Humanism on the Ropes?

Evolution and Cognition: ISSN: 0938-2623 Published by: Konrad Lorenz Institut für Evolutions- und Kognitionsforschung, Adolf-Lorenz-Gasse 2, A-3422 Altenberg/Donau. Impressum Tel.: 0043-2242-32390; Fax: 0043-2242-323904; e-mail: [email protected]; World Wide Web: http://www.univie.ac.at/evolution/kli/ Editors: Rupert Riedl, Manfred Wimmer Layout: Alexander Riegler Aim and Scope: “Evolution and Cognition” is an interdisciplinary forum devoted to all aspects of research on cognition in animals and humans. The major emphasis of the journal is on evolutionary approaches to cognition, reflecting the fact that the cognitive capacities of organisms result from biological evolution. Empirical and theoretical work from both fields, evolutionary and cognitive science, is accepted, but particular attention is paid to interdisciplinary perspectives on the mutual relationship between evolutionary and cognitive processes. Submissions dealing with the significance of cognitive research for the theories of biological and sociocultural evolution are also welcome. “Evolution and Cognition” publishes both original papers and review arti- cles. Period of Publication: Semi-annual Price: Annuals subscription rate (2 issues): ATS 500; DEM 70, US$ 50; SFr 60; GBP 25. Annual subscriptions are assumed to be continued automatically unless subscription orders are cancelled by written information. Single issue price: ATS 300; DEM 43; US$ 30; SFr 36; GBP 15 Publishing House: WUV-Universitäts- verlag/Vienna University Press, Berggasse 5, A-1090 Wien, Tel.: 0043/1/3105356-0, Fax: 0043/1/3197050 Bank: Erste österreichische Spar-Casse, Acct.No. 073-08191 (Bank Code 20111) Advertising: Vienna University Press, Berggasse 5, A-1090 Wien. Supported by Cultural Office of the City of Vienna, Austrian Federal Ministry of Science/ Transportation and the Section Culture and Science of the Lower Austrian State Government.

Hubert Hendrichs

Accessing Animal Minds: Epistemological and Empirical Problems

OME 20 YEARS AGO, AT Abstract 1984; SEARLE 1983, 1992; S the 15th Interna- FODOR 1987; DENNETT tional Ethological Con- The paper outlines the origin and aims of an interdis- 1987; DRETSKE 1988; DREY- ference held in Bielefeld, ciplinary project on the problem of animal minds, in- FUS/DREYFUS 1988; PEN- a senior ethologist, Don volving the cooperation of ethology, neurobiology, ROSE 1989; RADNER/RAD- GRIFFIN, had proposed a robotics, and philosophy. The contributions, al- NER 1989; KIM 1996), round table on cognitive though heterogeneous and reflecting different posi- providing valuable con- ethology, on awareness tions in these disciplines, nevertheless indicate that ceptual and epistemologi- and intentionality in ani- there is a common ground to be considered in future cal tools. But in spite of mals.1 The faculty of biol- theoretical and empirical studies. The behavior of or- this great interdisciplinary ogy of the recently ganisms is oriented and organized by a compound of enthusiasm and effort3 founded university had highly integrated components of different dimension some appearingly simple only started work one and rationality. In human beings some of these com- (ontological) questions year before with a very ponents are named cognition, emotion, and con- asked by Don GRIFFIN in small number of scien- sciousness. Their evolutionary development can only the handout for his round tists. I was given the task be understood as a multilevel phenomenon, and by table 20 years ago have re- to organize the work- considering the heterogeneities in dimension and ra- mained unanswered: shops and round tables at tionality, in structure and processing, and their in- B Do animals know what the conference. I remem- creasing integration. they are doing? ber that this proposal B Can communicative caused irritation among Key words behavior be used to ask the organizers. No sober questions? biologist was—at that Animal minds, evolution of human mentality, cogni- B What hypotheses can time—expected to com- tion, emotion, consciousness. be formulated concern- mit himself to such ing animal awareness and unscientific topics, if he intention? wanted to keep his scientific reputation. That atti- B What observations or experiments could show, tude has changed considerably—not completely— whether animals have internal images? in the meantime. At an annual meeting of the Deut- This indicates that something is seriously wrong sche Zoologische Gesellschaft in Oldenburg a round or missing with the categories and conceptualiza- table was held, entitled “Toward a biological con- tions used in the various disciplines concerned with cept of consciousness” (cf. PFANNENSTIEL 1996). systems—living and nonliving, human and nonhu- Without doubt interest has grown and great man—which generate functionally adequate ac- progress has been achieved in the last years, espe- tions in situationally changing contexts, responding cially in neurobiology, investigating and document- to various forms of stimulation. To the unanswered ing correlations between brain activity and behav- questions one may add: ioral states2, and in primatology, investigating B To what extent is observable behavior influenced communication and cheating in apes and other pri- by processes involving feelings, cognitive represen- mates (cf. PREMACK/PREMACK 1983; WHITTEN/BYRNE tations, beliefs, and 1988; CHENEY/SEYFARTH 1990). Important contribu- B what types of intentional states are required, tions could in addition be recruited from cognitive B do they have to be individually acquired or pro- psychology (cf. BIRBAUMER 1996; MARKOWITSCH 1992; duced? BAARS 1997) and philosophy (AGAZZI 1981; PYLYSHYN B What are the established empirical findings?

Evolution and Cognition ❘ 98 ❘ 1999, Vol. 5, No. 2

Accessing Animal Minds: Epistemological and Empirical Problems

B How can these processes be conceptualized, and impeded by a number of unsolved problems which B what are the criteria that allow us to ascribe repre- have to be coped with first. They include conceptual sentations and intentional states, and methodological, that is, epistemic and empiri- B what are the necessary differentiations? cal, problems as addressed in the contributed papers. These questions concern a wide range of different Although this faraway goal will not yet be a central types of scientific approaches, from the evaluation of issue of this volume, it is nevertheless an important empirical data to the epistemological analysis of ar- background, providing an orienting horizon for its gumentations, but they center around an area in be- main task: discussing, evaluating, and connecting tween—and outside—the established scientific and the different concepts proposed by different disci- philosophical disciplines: an area in which concep- plines and possibly reflecting different positions tualizations have to be developed that can be con- within one discipline. Thus, the aim here is not yet nected with specific concepts of several disciplines to find answers to unanswered questions or solu- and can thereby form a bridge between these disci- tions for unsolved problems but, much more mod- plines. Such bridges are required and have to be de- estly, a contribution to the following three veloped to find answers to the above questions. interrelated tasks: These conceptualizations have to be built with scien- B to gain a better transparency of the basic concepts tifically and philosophically acceptable tools and underlying the various conceptualizations of the dif- building blocks (although outside the discipline spe- ferent disciplines, and cific tools; this is a problem: what may be acceptable B to look for possibilities of adequately associating to one discipline may not to an other, and vice versa). these different conceptualizations, finding and de- They are concerned with the performance of struc- scribing possible links or bridges, and tures and processes which are functional and goal- B thereby mapping out a field in which new ques- directed, while their steering and their intentionality tions may be asked and possibly answered. and cognitive orientation may be very different. Although relatively modest, this still is a daring These kinds of differences pose unsolved problems, enterprise. for example the difference between artificial and liv- Thus, the distant goal is to establish a solid base for ing intelligence, between machines and living sys- a careful ontological characterization of living systems, between the rights of animals and humans, and tems. This will be approached by describing and eval- between animals at different levels of complexity. uating epistemological and empirical requirements The ultimate aim of such efforts will be something and problems. One central aim is to draw attention like an ontological evaluation of states, processes, to differences between animals of lesser and greater and structures in systems, which may have a poten- mental complexity, of fewer or more levels in the tial for producing states of consciousness and suffer- organization of orientation and intentionality. These ing, and which are relevant when considering the differences are considered to be related to those be- status of these systems—morally, legally, and other- tween organisms and machines, between cognitive wise. But this is not the foremost topic of our present processes in real and in virtual spaces, and those in volume. Questions concerning ontological—and the degrees of freedom, individuality, personality, ethical—states, such as: and dignity of objects. The careful semantic analysis B what does it mean to be a living system, able to of concepts and argumentations by cognitive philos- intend and to suffer, ophy is essential, but not sufficient. The neurobiolog- B do these qualities require a specific kind of treat- ical analysis of physiological and structural founda- ment of the system, tions is very important, but not sufficient either. The B how do specific requirements of such systems careful and highly differentiating ethological analy- have to be considered in case of conflicting inter- sis of a wide spectrum of animal behaviors is essen- ests? tial, but likewise not sufficient with respect to an as- are not immediately relevant at this time. They are sessment of the intentional states involved. It is obviously very important, and it is a serious problem necessary to combine these different efforts, and for that there are no generally accepted answers to these such a combination between the disciplines, model- questions, and that (just or fair) solutions have to be ings are required that (a) are compatible with the confound without knowing the ontological status of the ceptualizations in the different disciplines, and systems involved. But this fundamental knowledge (b) can be related to empirical research. With careful is a goal that obviously lies ahead, albeit in the empirical analysis, it may then be possible to distin- future, and the first steps on the road to this goal are guish different types and levels of organization in

Evolution and Cognition ❘ 99 ❘ 1999, Vol. 5, No. 2 Hubert Hendrichs

such a way that it becomes possible to distinguish DRECKMANN, Hubert HENDRICHS, Joëlle PROUST, ontologically between objects of these different types Daisie RADNER, and Achim STEPHAN, while the and levels. Then it may become possible to connect present volume combines papers of Colin ALLEN (the with these objects a specific ontological status, to be same paper as in the Erkenntnis volume), Marc taken into account by moral and legal regulations. BEKOFF (the same paper as in the Erkenntnis volume), Finding an orientation in this difficult field re- Gordon BRITTAN, Gordon BURGHARDT, Holk CRUSE, quires the combination of two distinctive and heter- Hans-Johann GLOCK, Hubert HENDRICHS, Rolf LACH- ogeneous approaches: (a) a careful epistemological MANN, Hans LENK, Gerhard ROTH, Ted SCHATZKI, analysis of argumentations, and (b) the modeling of Franz SEITELBERGER, and Gunther STENT. Several of properties that can be ascribed empirically to objects the contributions of this volume address possibilities of observation. The combination of these two ap- of modeling for empirical investigations: from the proaches is so difficult that one of them is usually side of biology those of BEKOFF, BURGHARDT and abandoned to improve effectiveness of the other. In CRUSE, from the side of philosophy those of ALLEN, an attempt to draw attention to this problem, and BRITTAN, GLOCK, LACHMANN, and SCHATZKI. Some pa- find a way to overcome it, a cooperation or at least pers emphasize important ontological distinctions: discussion between philosophers, ethologists and SEITELBERGER and STENT as neurobiologists, LENK and neurobiologists has been initiated. The various pa- SCHATZKI in philosophy; some mainly focus on con- pers were discussed at a workshop held in September tinuities: ROTH in neurobiology, CRUSE in biological 1997 at the newly founded Hanse-Wissenschaftskol- cybernetics, GLOCK in analytical, and LACHMANN in leg in Delmenhorst near Bremen. The contributions phenomenological philosophy. have not been integrated, and do not use the same BRITTAN draws attention to some of the central terminology. Even within one discipline different questions and specifies some of the problems. While theoretical positions are presented: in ethology, neu- behavior of machines and animals and that of ani- robiology, and philosophy. They have primarily mals with a different complexity of mental organiza- been chosen for their suitability as potential building tion are often indistinguishable for an observer, per- blocks for a bridge to be built between these disci- formances can be distinguished on the basis of the plines. They may or may not represent the latest de- type of programming and the kind of learning in- velopments in their disciplines, but the authors, who volved, that is, referring to differences in the life his- where chosen for their competence in their respec- tory of the performing objects, the different ways in tive fields, have developed aspects of their work that which they acquired and use their capacities. He might be suitable building blocks. With this future stresses that using representations is different from goal in mind, behaviors of plants, animals, and ma- being aware of these. In addition he states reasons for chines are described and analyzed with respect to the his belief that the attention placed by some authors levels of intentionality involved. Phylogenetic levels on language is misleading as other expressions of of brain organization are described by neurobiolo- mind are important, and that animals are continu- gists, and phylogenetic levels of complex behavior by ously evaluating and making judgments. ethologists. In a number of contributions, the rele- BURGHARDT investigates the evolution of play, be- vance of such investigations for ontological and eth- ing an important type of behavior in animals with ical questions is pointed out. Each of the papers is an higher levels of mental organization. Its generation interdisciplinary attempt addressing several of these combines cognitive and emotional components. Al- tasks and including results of several disciplines. though not necessarily linked to intelligence, play In view of a growing number of contributions and constitutes a fertile ground for the joint development the wide range of topics they cover, we decided to of perception, emotion, cognition, and conscious- divide the papers up into two blocks, each with a ness. Like cognition and consciousness play provides different focus, which will appear as special issues in an example (a) of fundamental difficulties of defin- two scientific journals: the more epistemologically ing mental properties in relation to behavioral per- oriented papers in “Erkenntnis” (STEPHAN/HEN- formances, and (b) of an emergence of new qualities DRICHS/DRECKMANN 1999) and the more general pain ontogeny and phylogeny. pers, including ontological considerations, in “Evo- ALLEN distinguishes between social and individual lution and Cognition”, respectively. The concepts and is particularly interested in the relation- “Erkenntnis” volume combines papers (in alphabet- ship between concepts and perception. He gives rea- ical order) by: Colin ALLEN, Marc BEKOFF, Gordon sons not to abandon the notion of concepts and sug- BRITTAN, Donald DAVIDSON, Fred DRETSKE, Frank gestions for empirical investigations of concepts in

Evolution and Cognition ❘ 100 ❘ 1999, Vol. 5, No. 2 Accessing Animal Minds: Epistemological and Empirical Problems nonhuman animals. He proposes clauses to be used, and idealistic, monistic and dualistic positions, but to illustrating his arguments with examples from own approach their integration by distinguishing ontolog- studies with domestic pigs. ically and epistemologically between several levels of BEKOFF is interested in the use of the notion of (mental) organization which differ in number, form, intentionality in biology. He addresses the question and connectedness in different living and non living how individuals negotiate cooperative agreements “behaving” objects (cf. HENDRICHS 1994). and exchange information concerning their beliefs CRUSE investigates possible similarities and differ- and goals in different systematic taxa. Drawing on ences in cognition, awareness and subjective feelings his own empirical research he uses examples from between humans and highly developed future ro- social play in mammals and antipredator behavior bots. He points out transitions between external and of birds. He defends the use of the notion of inten- internal perspectives and, in internal experience, dis- tionality when investigating the behavior of animals tinguishes different degrees of connectedness with and the use of classical ethological techniques for symbols, finally leading to manipulable internal this investigations. models. For artificial neural networks, he describes ROTH gives a general overview over brains in the possible states of relaxation, with a high “harmony animal kingdom, especially characterizing the verte- value” when the positions of variables match, and of brate brain, and then concentrates on the brain of tension, when the harmony value is low; when two mammals, differentiating between primates and neural layers both represent the body, their represen- nonprimates, humans and apes. He considers differ- tations may (a) coincide, resulting in a relaxed state, ent states of awareness and relates them to different or (b) differ, resulting in an activated state, possibly brain areas, which contribute in different ways to the inducing the network to change into a new state. It different types of awareness. He distinguishes several is proposed that this second state (b), when experi- parameters of brain developments and investigates enced, constitutes a first elementary condition for their potential for possible distinctions. Clear results subjective internal perspectives. in the field relating cognitive abilities to brain activ- LACHMANN, from a phenomenological perspec- ity are still outstanding. Human consciousness in- tive, addresses the question of animal awareness cludes various diverse phenomena produced by dif- starting from conceptions developed by Alfred North ferent brain areas. Most of these are present in all WHITEHEAD, Arnold GEHLEN, and Maurice MERLAU- tetrapods (amphibians, reptiles, birds, mammals). PONTY. Referring also to the studies of Susanne K. The relatively large human isocortex is shared with LANGER he points to a pragmatic value perception in most primates, the large prefrontal cortex with apes. animals and discusses the possibility of gaining ac- SEITELBERGER also states that, despite remarkable cess to components of animal awareness by analyz- progress in neurosciences and bioinformatics, our ing basic components of human consciousness. knowledge of (higher) brain performances is still LENK, starting from anthropological propositions, scant and peripheral. After distinguishing between especially those of Ernst CASSIRER and Susanne LANGER, states and contents of consciousness, he points out specifies the “second nature” of human beings, char- difficulties in describing and investigating brain per- acterizing human culture as constituted by several lev- formances. Based on his experiences as a neurobiol- els of symbolization. Symbolic meta-representations ogist he describes the construction of mental experi- of symbolic representations are constructed in a “sym- ences in man. Information processing does not bolic ascent”. The “animal metasymbolicum” has to produce images, but compatible abstract reconstruc- integrate several levels of interpretation and can tions that are not conscious. Only some selected thereby be distinguished from animals with highest transcoded results, with autonomous ways of self- mental capacities, as apes for example. construction, do become conscious in humans. He GLOCK, from an analytical perspective, addresses notes new levels of complexity not reached in other the question to what extent thoughts and concepts primates and sees the human brain as an essential are dependent on verbal language. Avoiding ex- innovation with new structural completions and a treme positions he develops an intermediate posi- total cortalization of inputs. tion allowing for simpler kinds of thoughts and STENT, as a neurobiologist with extensive interdis- concepts in living beings without verbal language. ciplinary experience, enters a very difficult philosoph- He proposes, that (a) the possession of concepts ical domain, addressing fundamental philosophical does not require verbal language, and (b) animals and legal questions from a dualistic position. It is not can use thoughts of a simple kind, that do not re- the intention of this volume to oppose materialistic quire concepts.

Evolution and Cognition ❘ 101 ❘ 1999, Vol. 5, No. 2 Hubert Hendrichs

SCHATZKI analyzes important recent contributions that have to be bridged by the behaving or acting proposed for modeling agency, and combines them organism with specific decisions, combining con- into a promising new type of actor-network theory. scious and subconscious, cognitive and emotional He distinguishes two types of networks maintaining components. (The transformation of information different functional relations to the activity of an ac- and the generation of decision in future computing tor: composing networks and embedding networks, robots on one side and in mammals on the other will both effecting agency in a different way. He sees an thus possibly continue to rely on different processes, actor as an association of heterogeneous types of phe- with different causalities and different rationalities.) nomena in relation to a multilevel and intercon- Motivations, behavior, social relations, and nected array of networks. He applies discourse mul- knowledge are domains of different dimensions and tiplicity to this network multiplicity and describes involved to different extents in the production of de- three ways in which agency is effected by embedding cisions in mental and intentional acts. An experi- networks. He points out significant differences be- enced observer, knowing thoroughly the observed tween these compositional and embedding net- animals and their behavior, can in suitable situations works, and thereby develops a tool to distinguish be- on the basis of a careful, that is attentive, critical, and tween different types of actors: not only between patient registration of behavioral details, recognize animals, machines, and tornados, but also between with some degree of reliability the different processes scallops, dogs, chimpanzees, and humans. combining in the production of specific “decisions” In the following some brief remarks refer to my in the observed animals. He or she can distinguish own position.4 Human behavior can be influenced by between different types of awareness in the behaving beliefs, but it is not necessarily caused by them. There individual and between programmed and inten- can be a “gap” (SEARLE 1999) in the cognitive causa- tional behavior. He or she does not ascribe intention- tion of acts, leaving space for other components of ality when the behavior only “looks as if” it were orientation and organization. Already in higher intentionally brought about, not even when the be- mammals open areas in the programming of their havior in “shorthand” is described in intentional behavior can be relevant for the integration of com- terms. The experienced observer knows that behav- peting influences of different dimensions (cf. HEN- ior produced without any intention on the part of the DRICHS 1997). However, most of these openings, behaving individual may strike an inexperienced ob- which exist at birth in the programmed mind, be- server as “obviously” intentional. That of course is come closed during ontogeny by a development of easier to realize when observing a deer rather than an habits and customs, that is learned patterns of behav- ape. But this aspect is also very important when ob- ior. In higher mammals, these patterns cease to be serving apes, even here many actions that appear fixed, but remain open for an individual orientation “obviously” intentional can be produced without a in fields of learned “rules”. These rules, and the pro- specific awareness on part of the behaving animal. In cesses of orientation that make use of them, mostly particular, the experienced observer realizes that ob- are not consciously known by the acting individual. ject-related behavior of an animal does not require a In the domain of orientation and of organizing recognition of this object by the behaving animal, decisions, the mammalian brain has undergone dras- not to speak of a concept of this class of objects. Like- tic changes in comparison to the earlier reptile and wise the correcting of a behavior after “recognizing” bird-like brains. It could be described as a brain work- a mistake in a specific situation (e.g., mistaking a ing with “gaps”: sensual inputs are not transported higher-ranking group member for a lower-ranking in a linear fashion to central areas, but become trans- one) does not imply a concept of situation or mistake formed and connected to each other in highly com- (and is not dependent on verbal language). plex associations. Thereby the possibilities of reactive Propositional contents of beliefs and desires in behavior become enlarged, the field of possible reac- human orientation in most cases remain closely tions is opened up, and requires specific orientations connected to subconscious processes and thus re- and decisions in each situation. This appears to have lated to non-mental organismic and environmen- occurred at different levels of mental processing (and tal processes. They directly and indirectly are in similarly in the human analysis of these processes). contact with structures and processes of an ob- In between the programmed structures, “hard-” and server-independent reality. The contacted outside “soft-wired” instructions for organization, orienta- structures and processes are of different dimen- tion, and action, there remain “gaps” which are con- sions—physical, chemical, social, mental—and of stitutive for efficient action: uninstructed open areas different organizations. An orientation in these

Evolution and Cognition ❘ 102 ❘ 1999, Vol. 5, No. 2 Accessing Animal Minds: Epistemological and Empirical Problems different domains has to combine different forms consequences. The highest form of this of logic and to integrate them into one rationality. conscientiousness is only found in humans, but The rationality of decisions, moral and otherwise, there are many dispositions and tendencies in therefore also depends on the consistency of the higher animals that can be considered roots of this required integrations, and not only on calculations highest human quality (HENDRICHS 1999). They in “formal logic” (i.e., this rationality is possibly develop in socially transmitted fields of orientation not of a digital/binary type as commonly used). and consist of rules and taboos in networks of Looking back upon the heterogeneous papers relations and contexts. They are required for from various disciplines it is apparent that the differ- example for the organization of life in a social ent contributions, though they did not arise from a context without excessive aggressions, that is, for common ground, can be related to a common cen- creating a social situation in which the young can ter. This can be characterized by the following, rela- develop without excessive strain and thereby tively simple, but nevertheless important finding: develop their social and mental competence to the There is no simple dichotomy between phenomena full species specific level (HENDRICHS 1997). In a and underlying reality, but a multilayered structure world of increasing virtuality, it is becoming of different types of observable phenomena in one increasingly difficult not to loose sight of those common reality. The layers may have different di- differences between dogs, guinea pigs, and mensions, and their analysis may require different tamagotchis that are essential for their adequate rationalities. The access to a multilayered complex of treatment. To consider the cohesive evolution of different components—like “consciousness”—de- emotion, cognition, and conscientiousness will, in pends on specific properties of the approach. It has this situation, be essential for any attempt to to be able to integrate different dimensions and ra- contribute to a task indispensable to a future global tionalities in its analysis. In our case: for an analysis society: finding acceptable and sustainable solutions and an understanding of human consciousness and for agreements between humans of different conscience, and their evolution, it is not sufficient to cultures and for the fair treatment of living beings consider cognitive propensities and performances. with different levels of mental complexity. Other components of different dimension and rationality may have to be included, at least that of emo- Acknowledgments tionality. The evolution of cognition is closely connected to I thank Gordon BURGHARDT, Frank DRECKMANN, the evolution of emotionality (WIMMER/CIOMPI Ursula HENDRICHS, Rolf LACHMANN, and Achim 1996; cf. LEWIS 1997; LABAN HINTON 1999). And the STEPHAN for their critical comments on earlier ver- evolution of both these sions of this paper, Gerhard faculties is required for the ROTH for providing, at the phylogenetic and ontogenetic Author’s address Hanse-Wissenschaftskolleg, development of morality, the possibilities for the discussion ability and the disposition to Hubert Hendrichs, University of Bielefeld, of various papers, and Adel- Department of Ethology, Morgenbreede 45, take responsibility for heid E. BAKER for language individual behavior and its D-33615 Bielefeld, Germany. assistance.

Notes the problems addressed in our volume, although they have only partly been taken up again in the more recent discus- 1 Don GRIFFIN later documented his efforts: GRIFFIN (1981; sions. Cf. BROAD (1925), UEXKÜLL/KRISZAT (1932), HEBB 1992), cf. RISTAU (1991). (1949), GREY WALTER (1953), BOULDING (1956), ZIFF (1959), 2 Cf. CHANGEUX/KONISHI (1987), OESER/SEITELBERGER (1988), LORENZ (1963/65), (1981), POLANYI (1966), LANGER (1967; EDELMAN (1989), ROTH (1995); for cognitive ethology cf. 1972), PIAGET (1971), ECCLES (1973), LURIJA (1973), CAMP- RIEDL (1980), GOULD/GOULD (1994), ALLEN/BEKOFF (1997). BELL (1974), PUTNAM (1975). 3 Some of the older contributions to this field contain valu- 4 For details cf. HENDRICHS (1999). able proposals for interdisciplinary conceptualizations of

Evolution and Cognition ❘ 103 ❘ 1999, Vol. 5, No. 2 Hubert Hendrichs

References The Johns Hopkins University Press: Baltimore, London. Langer, S. K. (1972) Mind: An essay on human feeling, vol. II. Agazzi, E. (1981) Intentionality and artificial intelligence. The Johns Hopkins University Press: Baltimore, London. Epistemologia 4: 195–228. Lewis, M. (1997) The self in self-conscious emotions. Annals Allen, C./Bekoff, M. (1997) Species of mind. The philosophy of the New York Academy of Sciences 818: 119–142. and biology of cognitive ethology. MIT Press: Cambridge Lorenz, K. (1963/65) Haben Tiere ein subjektives Erleben? In: MA. Lorenz, K. (1965) Über tierisches und menschliches Verh- Baars, B. J. (1997) In the theater of consciousness. The work- alten, vol. II. Piper: München, pp. 359–374. space of mind. Oxford University Press: Oxford. Lorenz, K. (1981) The foundations of ethology. Springer: Birbaumer, N. (1996) Selbstregulation des Gehirns und Verh- New York. alten. Verhandlungen der Gesellschaft Deutscher Natur- Lurija, A. R. (1973) The working brain. Harmondsworth. forscher und Ärzte 119: 297–311. (Deutsch: Das Gehirn in Aktion. Einführung in die Neu- Boulding, K. E. (1956) The image. Michigan University Press: ropsychologie. Rowohlt: Hamburg 1992). Ann Arbor. Markowitsch, H. J. (1992) Neuropsychologie des Gedächtnis- Broad, C. D. (1925) The mind and its place in nature. Rout- ses. Hogrefe: Göttingen. ledge and Kegan, London. Oeser, E./Seitelberger, F. (1988) Gehirn, Bewußtsein und Campbell, D. T. (1974) Evolutionary epistemology. In: Schilpp, Erkenntnis. Wissenschaftliche Buchgesellschaft: Darms- P. (ed) The philosophy of Karl Popper, vol. 1. Open Court: La tadt. Salle, pp. 413–463. Penrose, R. (1989) The emperor’s new mind. Concerning Changeux, J. P./Konishi, M. (1987) The neural and molecular computers, minds, and the laws of physics. Oxford Univer- basis of learning. Wiley: Chichester. sity Press: Oxford. Cheney, D. L./Seyfarth, R. M. (1990) How monkeys see the Pfannenstiel, H.-D. (ed) (1996) Verhandlungen der Deut- world. Inside the mind of another species. University of schen Zoologischen Gesellschaft. 89. Jahresversammlung, Chicago Press: Chicago. G. Fischer: Stuttgart, New York. Dennett, D. C. (1987) The intentional stance. MIT Press: Piaget, J. (1971) Biology and knowledge. An essay on the rela- Cambridge MA. tions between organic regulations and cognitive processes. Dretske, F. (1988) Reasons in a world of causes. MIT Press: University of Chicago Press: Chicago. Cambridge MA. Polanyi, M. (1966) The tacit dimension. Doubleday: New Dreyfus, H. L./Dreyfus S. E. (1988) Mind over machine. The York. power of human intuition and expertise in the era of the Premack, D./Premack, A. (1983) The mind of an ape. W. W. computer. Free Press: New York. Norton: New York. Eccles, J. C. (1973) The understanding of the brain. McGraw- Putnam, H. (1975) Mind, language, and reality. Philosophical Hill: New York. papers, vol. II. Cambridge University Press: Cambridge. Edelman, G. M. (1989) The remembered past. A biological Pylyshyn, Z. (1984) Computation and cognition. Toward a theory of consciousness. Basic Books: New York. foundation for cognitive science. MIT Press: Cambridge Fodor, J. A. (1987) Psychosemantics. The problem of mean- MA. ing in the philosophy of mind. MIT Press: Cambridge MA. Radner, D./Radner, M. (1989) Animal consciousness. Prome- Gould, J. L./Gould, C. G. (1994) The animal mind. Freeman: theus: Buffalo. New York. Riedl, R. (1980) Biologie der Erkenntnis. Die stammesge- Grey-Walter, W. (1953) The living brain. Duckworth: Lon- schichtlichen Grundlagen der Vernunft. Parey: Berlin, don. Hamburg. Griffin, D. R. (1981) The qestion of animal awareness. Rock- Ristau, C. A. (1991) Cognitive ethology. The minds of other efeller Univeristy Press: New York. animals. Essays in honor of Donald R. Griffin. Erlbaum: Griffin, D. R. (1992) Animal minds. University of Chicago Hillsdale, London. Press: Chicago. Roth, G. (1995) Das Gehirn und seine Wirklichkeit. Suhr- Hebb, D. O. (1949) The organization of behavior. A neuro- kamp: Frankfurt. physiological theory. Wiley: New York. Searle, J. R. (1983) Intentionality. An essay in the philosophy Hendrichs, H. (1994) Tier—Mensch—Maschine: Übereinstim- of mind. Cambridge University Press: Cambridge. mungen und Verschiedenartigkeiten in der Regulierung Searle, J. R. (1992) The rediscovery of the mind. MIT Press: ihres Verhaltens. In: Letzgus, K. et al. (eds) Für Recht und Cambridge MA. Staat. Festschrift für Herbert Helmrich. Beck: München, pp. Searle, J. R. (1999) Rationality in action. Opening lecture of a 1107–1124. symposium “Speech acts, minds, and social reality. Discus- Hendrichs, H. (1997) On the development of psychosocial in- sions with John R. Searle” held at the Zentrum für interdis- dividualities in group living mammals. In: Weingart, P. et ziplinäre Forschung, Bielefeld, June 21–23. al. (eds) Human by nature. Between biology and the social Stephan, A./Hendrichs, H./Dreckmann, F. (eds) (1999) Ani- sciences. Erlbaum: Mahwah NJ, pp. 400–407. mal Mind. Erkenntnis 51. Hendrichs, H. (1999) Different roots of human intentionality Uexküll, J. v./Kriszat, G. (1932) Streifzüge durch die Umwel- in mammalian mentality. In: Stephan, A. et al. (eds) Ani- ten von Tieren und Menschen. Fischer: Frankurt. mal mind. Erkenntnis 51: 145–164. Wimmer, M./Ciompi, L. (1996) Evolutionary aspects of affec- Kim, J. (1996) Philosophy of mind. Westview Press: Boulder tive–cognitive interactions in the light of Ciompi’s con- Col. cept of “affect logic”. Evolution and Cognition 2: 37–58. Laban Hinton, A. (ed) (1999) Biocultural approaches to the Whitten, A./Byrne, R. (1988) Tactical deception in primates. emotions. Cambridge Univeristy Press: Cambridge. Behavioral and Brain Sciences 11: 233–273. Langer, S. K. (1967) Mind: An essay on human feeling, vol. I. Ziff, P. (1959) The feelings of robots. Analysis 19: 64–68.

Evolution and Cognition ❘ 104 ❘ 1999, Vol. 5, No. 2 Gordon G. Brittan, Jr.

The Mental Life of Other Animals

1. Most major twentieth Abstract rots… I add also that these century philosophers words or signs must not reject DESCARTES’ meta- express any passion, to DESCARTES claims that it is in principle possible to de- physical position, in par- sign and construct machines whose behavior is indis- rule out not only cries of ticular his dualism of tinguishable from that of any nonhuman animal. If joy or sadness and the mind and body. So it is the behavior is indistinguishable, then it follows that like, but also whatever perhaps surprising that we have to attribute no more capacity to the animal can be taught by training they also tend to accept than to the machine. Since the machine has no more to animals. If you teach a the rather narrow distinc- than a “mechanical capacity,” we need to assume no magpie to say good-day to tion that he makes more than a “mechanical capacity” (what DESCARTES its mistress, when it sees between human beings calls a “disposition of the organs”) on the part of the her approach, this can and other animals, espe- animal. But this argument turns on a non-sequitur. only be by making the ut- cially since he uses that Machines are programmed, animals are conditioned, terance of this word the distinction in turn as an and each acquires its capacity in a very different way. expression of one of its important support of his The fact that the behavior is indistinguishable does passions. For instance, it dualism. I won’t try to not imply that the capacity is indistinguishable. Al- will be an expression of unravel this apparent though disagreeing with him in a variety of ways, the hope of eating, if it inconsistency here. In my many philosophers (including KANT and DAVIDSON) has always been given a own view, an adequately have followed DESCARTES in stressing the important of tidbit when it says it… it naturalistic account of linguistic competence in making a distinction be- seems to me that the use human beings must tween humans, on the one hand, and animals and of words, so defined, is embrace their continuity machines, on the other. But a second non-sequitur is something peculiar to hu- with other animals. involved. Even supposing that other animals are not man beings… there has competent linguistically does not imply that they do never been an animal so not enjoy a mental life, in particular that they do not 2. DESCARTES’ argument perfect as to use a sign to for the distinction have beliefs or desires, or are not self-conscious. make other animals un- between human beings derstand something and other animals, set out Key words which expressed no pas- in detail in the fifth part sions and there is no hu- DESCARTES, behaviorism, linguistic competence, ani- of the Discourse on Method man being so imperfect as mal mental life. and the letter to the Mar- not to do so, since even quess of Newcastle of deaf-mutes invent special November 23, 1646, turns for the most part on signs to express their thoughts”. (KENNY 1970, questions of linguistic competence. The passage pp206–207) from the Discourse is already very well known. Here is what is said in the letter to the Marquess of New- 3. The main line of the argument can be put as fol- castle: lows: “…none of our external actions can show anyone (i) Any (relatively complex and uncoerced) behav- who examines them that our body is not a self-mov- ior which can be explained in terms of the “pas- ing machine but contains a soul with thoughts, with sions” does not require thought, any behavior which the exception of words, or other signs that are rele- cannot be so explained does. vant to particular topics without expressing any pas- (ii) All animal behavior, including apparently lin- sion. I say words or other signs, because deaf-mutes guistic behavior (the purring of cats, the barking of use signs as we use spoken words; and I say that these dogs, the chattering of magpies) can be explained in signs must be relevant, to exclude the speech of par- terms of the “passions”.

Evolution and Cognition ❘ 105 ❘ 1999, Vol. 5, No. 2 Gordon G. Brittan, Jr.

(iii) Therefore, no animal behavior requires There are at least two crucial differences between thought. conditioning and programming. One is that condi- (iv) But at least some human behavior, in partic- tioning is in a certain sense after the fact. An animal ular human linguistic behavior, cannot be explained is rewarded or punished for behavior that has al- in terms of the “passions”. ready occurred. Such reinforcement accordingly (v) Therefore, at least some human behavior re- raises or lowers the probability that the animal will quires thought. engage in the same behavior on the next occasion (vi) Therefore, human beings are to be sharply dis- when it is presented with approximately the same tinguished from other animals; they alone have conditions. A machine that has not already been minds. programmed (in some wide sense of the word that includes hydraulic automata as well as homing rock- 4. The argument is evidently valid. We need to look ets) will simply not do anything. In this sense, pro- more closely at its premises. Let me begin with the gramming is “before the fact”. A second difference second, that all animal behavior, including appar- between conditioning and programming is aligned ently linguistic behavior, can be explained in terms with the first. Systems of reward and punishment of the “passions”. Whatever the precise meaning of work only if the “reward” is accompanied by feeling the word “passion” in its seventeenth century con- of pleasure and the “punishment” by feelings of text, DESCARTES’ point is clear. All apparently linguis- pain. but the kinds of machines we are considering tic behavior on the part of animals can be explained do not have pleasure and pain receptors, hence they either as a relatively spontaneous outpouring (the cannot be conditioned. On the other hand, some- purring of a contented cat) or an instinctive reflex (a thing can be programmed only if it is capable (con- dog barking at the approach of strangers) or as a consciously or not) of following a set of instructions. ditioned response (the magpie taught to say good- It follows, as against DESCARTES, that animals are day to its mistress). In none of these cases, then, does not machines (at least not the kinds of machines the behavior require any thought. Spontaneity, in- DESCARTES had in mind). Animals, but not machines, stinct, and conditioning suffice to explain1 the pro- can be conditioned; it would be inappropriate, to say duction of what, in the Discourse, are called “natural the least, to raise a chess-playing computer’s voltage sounds”. after a successful match. If animals can be pro- We will not worry for the moment whether or not grammed, by way of a particular genetic endowment this claim is true, although it is worth remarking that for example, then the behavior which such program- in making it DESCARTES downplays the importance ming explains (much of what we otherwise term “in- of the traditional distinction between what is stinctive”) is not the same sort of (“learned”) behav- “learned” (DESCARTES says that the magpie has been ior which conditioning explains. Is a machine that “taught”) and what is merely “instinctive”. What could be conditioned, and in this sense that learns, concerns me now are the implications he takes the conceivable? I don’t see why not. The point is that it claim to have. One has already been mentioned; it is would be very much more like human beings than that if spontaneity, instinct, or conditioning suffice DESCARTES was willing to allow. to explain behavior, then its further explanation in Now the distinction between animals and ma- terms of “thought” is precluded. The other implica- chines has important corollaries. One of these corol- tion is that animals cannot be distinguished from laries has to do with the way in which DESCARTES sets machines. The first implication is problematic. The up his argument. He claims that it is in principle second is clearly false. possible to design and construct machines whose be- For the purposes of argument, we need to consider havior is indistinguishable from that of any animal machines whose behavior is (or can be imagined to in particular. If the behavior is indistinguishable, be) of the same order of complexity as, for example, then it follows that we have to attribute no more that of cats, dogs, and magpies, not only the hydrau- capacity to the animal than to the machine. Since lic automata in the gardens of the king whose move- the machine has no more than what might be called, ments so fascinated DESCARTES, but also so-called globally, a “mechanical capacity,” to be explained in “servo-mechanisms” like homing rockets. The be- terms of the motion and interaction of its parts, we havior of these latter, unlike the animals DESCARTES need to assume no more than a “mechanical capac- mentions, is not to be explained in terms of sponta- ity” on the part of the animal, to be explained at least neity, instinct, or (especially) conditioning, but in large part in terms of what DESCARTES calls the rather in terms of programming. “disposition of its organs”..

Evolution and Cognition ❘ 106 ❘ 1999, Vol. 5, No. 2 The Mental Life of Other Animals

But we granted that the behavior of animals and The point needs to be elaborated. First, in saying machines was indistinguishable and yet argued that an animal has desires, we are, at least initially, that they were to be distinguished nevertheless, in saying no more than that its behavior is (con- terms of the fact that the former but not the latter sciously or unconsciously) goal-directed. That such was to be accounted for in terms of conditioning. goal-directedness is not simply a matter of respond- In addition to behavior, we must also take into con- ing in a particular way to a given stimulus is clear sideration the way in which the capacity to behave when we consider cases in which the animal is frus- was acquired, that is, we must also take into account trated in achieving its goal. In such cases, there is the histories of the animals and machines involved. no option to saying that the animal is trying to reach However indistinguishable their behavior might its goal, that is, desires to reach its goal but is not be, if they have different histories of how their be- able to. Second, in attributing desires to animals we havioral capacities came to be acquired, then they eventually go beyond describing their behavior to are to be distinguished tout court. Human beings positing the existence of something “going on in- have the same sort of natural, if not also cultural, side their heads,” internal representations of what histories that many other animals do. From this is not now present, the goal-state to be achieved. perspective, human beings are in a crucial, al- Third, in explaining the behavior of machines we though not decisive, respect like animals, both of need posit no such internal representations. It is them to be distinguished sharply from machines. enough, at least in principle, to refer to their pro- The point needs to be stressed. DESCARTES as- grams and to their design. If there is a sense in sumes that behavior provides the only evidence we which a machine such as a homing rocket can be have for hypotheses concerning the minds of oth- said to have goals, the destruction of a plane or ers. The behavior of animals and machines, it missile target for example, and in this same sense might be granted, is indistinguishable. Therefore, to have desires, then these goals and desires are de- one has no less and no more “mind” than the other. rivative of the goals and desires of the person who But machines don’t have “minds”. Therefore, ani- designed and built the machine in the first place. mals don’t either. Human verbal behavior, on the Animals, of course, are not in the same way de- other hand, differs in kind from both animal and signed and built with a certain end in view. They machine behavior; it is creative, adaptable, and all evolved. the rest. Therefore, human beings do have minds. These conclusions follow, however, only if we as- 5. I have so far said nothing about the crux of DES- sume that behavior provides the only evidence we CARTES’ argument, the issue of linguistic compe- have for hypotheses concerning the minds of oth- tence. To this point, I have only wanted to maintain ers. It doesn’t. Among other things, we also have that animals are to be distinguished from machines the histories of those involved. Behaviorism, even and that the grounds of this distinction do not have in the weak way in which DESCARTES understands to do with behavioral differences (for the sake of it, as limiting the evidence on the basis of which argument we can assume that there are none), but our inferences to the minds of others are made, is with the different ways in which behavioral capac- false. ities are acquired. Animals acquire at least some of The other corollary of the distinction between their behavioral capacities through conditioning, animals and machines is this. In saying that ani- machines do not. Moreover, it is in virtue of their mals are conditioned, we presupposed that they are respective histories that it is appropriate to attribute “motivated”. DESCARTES insists on this point. We desires to animals, but not to machines. To this ex- teach the magpie to say good-day to its mistress on tent, that animals have desires (that they are “mo- her approach by giving it a tidbit only because it tivated” to behave in certain ways), they have a wants the tidbit.2 If the prospect were not motivat- mental life. To this same extent, DESCARTES would ing, then offering it after an appropriate “good- (perhaps not quite consistently, given his animal- day” would not raise the probability of more “good- machine hypothesis) agree. He too attributes to days” in the same circumstances. That is, in ex- them such otherwise “human” motivations as fear, plaining how conditioning works, it is inevitable hope, and joy. But, he claims (more in my words that we attribute certain desires, among them the than in his), we cannot attribute beliefs to them. “hope of eating,” to the animal. In the case of ma- For the attribution of belief is somehow bound up chine behavior, on the other hand, the attribution with linguistic competence, and animals are not of desire would be inappropriate. competent linguistically.

Evolution and Cognition ❘ 107 ❘ 1999, Vol. 5, No. 2 Gordon G. Brittan, Jr.

6. The apparent importance of language to the mensions. Apparently he does not think that analogies tal life of other animals can be brought out by con- between our own behavior and that of other animals sidering the views of Immanuel KANT. Although he establish that they, too, are self-conscious in the way puts the point in very different ways and with differ- that analogies establish that they are conscious. I ent emphases than I have, KANT too argues for attrib- take it, moreover, that for KANT no analogies be- uting desires (and in the process, representations) to tween their behavior and ours could establish that other animals, distinguishing them from machines other animals were self-conscious, since all we can and assimilating their behavior to ours. Perhaps the infer by analogy on the basis of behavior is that the most important of the several passages in which he behaver has “representations,” and not that it is fur- does so is in the Critique of Judgment (in a footnote to ther aware of these representations. But that we can- AK.5: 464).3 not establish self-consciousness by analogy on the What we can quite correctly infer by analogy, basis of behavior does not prove that other animals from the similarity between animal behavior (whose do not possess it, especially since, as we have already basis we cannot perceive directly) and men’s behav- seen, there are other factors which must be taken ior (of whose basis we are conscious directly), is that into account in attributing a mental life to others. animals too act according to representations [Vor- This is as true in the case of other human beings as stellungen] (rather than being machines, as DES- it is in the case of other animals. CARTES would have it), and that regardless of the dif- KANT also says that animals do not possess self- ference in specific kind between them and men, they consciousness since they are incapable of making are still of the same general kind (namely, as living judgments, which among other things involves beings). comparing representations. This time the link be- As his examples (beaver building lodges, and so on) tween the making of judgments and the possession make clear, KANT takes at least some animal behavior of self-consciousness is by way of the fact that the to be intentional or purposive, and thus to require the possession of self-consciousness involves, at a mini- attribution of desires (a type of internal representa- mum, distinguishing between how things stand tion in his framework) in its explanation. In this re- with the subject and how things stand with the spect, it is very much unlike machine behavior. world, and this distinguishing involves the making At the same time, KANT thinks that there is an of judgments. But again, no reason has been given important difference to be made between human be- why other animals cannot make judgments (or pos- ings and other animals. The former, but not the lat- sess the appropriate concepts) and thus why they ter, are self-conscious. In this respect, of course, he cannot distinguish between how things stand with sides with DESCARTES, for whom the essential thing them, and with the world. about animals (and also machines) is that they don’t Of course, KANT would have an effective argument know what they are doing. If their behavior is goal- in both cases if he added another premise. In the first directed, it is not consciously so, and in this fact lies case, the premise is that another animal is self-con- the principal difference between us and them. scious, e.g., possesses an “inner life” of its own, only KANT has a number of reasons for wanting to deny if it could tell us in some detail about this life. Since that animals are self-conscious. Many of them have it cannot tell us in any detail about this life, it is not to do with the fact that humans, unlike other ani- self-conscious. In the second case, the premise is that mals, are moral agents and can be held responsible one can make a judgment only if one possesses a for their actions. The conceptual tie between moral language in which to make it. That is, a judgment is agency and self-consciousness is supplied by the fact a propositional attitude and propositions are linguis- that although animals have desires, they are not tic in character. Since other animals lack language, aware that they do, and thus cannot consciously they cannot make judgments and are hence not self- choose to act on them or not, as we do. Moral agency conscious. presupposes freedom, and animals, lacking self-consciousness, are not free. 7. The sort of argument that KANT, and ultimately But these reasons for wanting to deny that ani- DESCARTES, hint at has recently been supplied in mals are self-conscious do not amount to a proof that some detail by Donald DAVIDSON in his essay, “Ratio- they are not, short of begging important questions. nal Animals” (DAVIDSON 1982). It has two main pre- In fact, so far as I can tell, KANT has no argument for mises: the claim that animals are not self-conscious, al- (i) In order to have a belief, it is necessary to have though he repeats this claim on a number of occa- the concept of belief.

Evolution and Cognition ❘ 108 ❘ 1999, Vol. 5, No. 2 The Mental Life of Other Animals

(ii) In order to have the concept of belief, one must plex and apparently “intelligent” as the behavior of have language. the former. In order to reach the conclusion that other ani- Third, to attribute language to other species, given mals do not have beliefs, and thus do not in any rich what we now know, would be to beg the most im- sense have a mental life, we must, of course, add the portant questions, for to attribute language to them premise that they do not have a language, a premise would be to endow their gestures, signs, and sounds which DAVIDSON accepts but does not argue for. The with meaning, and the only way in which they could argument in general, then, is that be said to have meaning would be as the expression (i) animals have a mental life only if they are com- of beliefs. But whether or not they can be said to have petent linguistically (this follows from DAVIDSON’s beliefs is precisely what is in question. two premises); Fourth, and most important, we do not need to (ii) animals are not linguistically competent; attribute linguistic competence to other animals in (iii) therefore, animals do not have a mental life. order to defend the claim that they have a mental In fact, most people who want to deny the con- life. That is, the problem lies with the first premise, clusion of this argument follow MONTAIGNE in deny- that animals have a mental life only if they are coming its second premise. I think it is dangerous to do petent linguistically. so, although not for the reason that DESCARTES gives, DAVIDSON claims, to begin with, that in order to that if other animals had a language they would long have a belief, it is necessary to have the concept of since have succeeded in communicating with us and belief. Why should this be? We need to focus on a with each other. Failure to understand another does particular case. not entail that he or she is not linguistically compe- Grizzlies, apparently unique among bears in this tent, witness our failure to understand the language respect, routinely cache food. We explain an indi- of the ancient Mayans. Four other sorts of consider- vidual’s returning to its cache and eating what was ation must be kept in mind. stored there by saying that it was hungry, desired First, there is preliminary evidence that linguistic food, and believed, on the basis of what it remem- competence is very closely tied to evolutionary de- bered, that it would find it at that spot. There is, in velopments in the cerebral cortex unique to human fact, no currently available option to anticipating its beings. There are other aspects in which our species behavior (or the behavior of a great variety of other is unique biologically, and it might very well be the animals) than to provide it with this sort of inten- case that a series of mutations occurred after we left tional explanation. The explanation is appropriate the savannah which made speech possible. Some when, among other things, the behavior is not co- venturesome linguists have already begun to suggest erced or prompted, not accidental (the bear did not dates when speech emerged.4 stumble across the cache), not automatic (the bear Second, it is at the very least controversial does not return to this particular place at regular in- whether the apparent competence of nonhuman tervals, whether or not food is there). Moreover, we primates in manipulating conventional signs can tell, in a perfectly objective way, whether or not (American Sign Language, for example) is not sim- the behavior is coerced or prompted, that is, we can ply to be accounted for, as DESCARTES insisted, in tell whether or not the bear was pushed or cued to terms of subtle cuing and conditioning. Since we do that spot, just as we can tell whether it returns merely not think that human language can be accounted coincidentally or as a function of habit. Occasionally for in terms of cuing and conditioning, no matter determining whether these conditions are satisfied how subtle, there is reason to deny that these ani- requires that we set up experiments. If they are sat- mals are “speaking” in the same way that we do. This isfied, then we are justified in taking what Daniel is not to deny that other primates, and for that mat- DENNETT calls the “intentional stance” with respect ter dolphins and whales as well, manipulate signs or to the animal’s behavior (DENNETT 1988). produce a rich variety of sounds. It is to suggest that Now it should be clear that taking the “inten- at this point we cannot make a case for animal lan- tional stance” presupposes that the bear’s behavior guage on the basis of what we know about them. It is adaptable to changed circumstances, that is, that might be added, again at this point in time, that it re-adjusts its behavior in the light of new informa- there is scarcely more reason to attribute a mental tion which it receives. On the assumption that its life to those species that manipulate conventional desire remains the same, to get something to eat, the signs than to those species that do not, for the be- re-adjustment of behavior can only result from a re- havior of these latter is in many cases just as com- adjustment of beliefs. But to say that the bear re-

Evolution and Cognition ❘ 109 ❘ 1999, Vol. 5, No. 2 Gordon G. Brittan, Jr. adjusts its beliefs is to say that, in the light of the new sary condition of having a concept, it is not also suf- information, it gives up some beliefs for others. This ficient. For to have a concept, once again of a cow entails that it has a belief about other beliefs, say, one must also be able to identify and individuate namely, that they are false. But to have a belief about cows, and this requires something like a language. others of its beliefs, what we might call a second- We can illustrate the point with reference to our order belief, it must have the concept of belief. grizzly bear case. Although the bear can distinguish Talk here of “concepts” is tendentious, for it al- between food and non-food, at least up to a point, it ready tips the scales in favor of the importance of does not really “have the concept” of food. “Having language to the having of beliefs. How, after all, the concept” involves having a number of beliefs could one have concepts without also having a lan- about food, that it is nutritious, that the body re- guage in which those concepts were inter-related quires it on a regular basis, and so on, in terms of and applied? But I see no problem in attributing a which we can identify food as food, in the process belief about beliefs, a second-order belief, to the griz- distinguishing it from other objects which are not zly, even in the absence of linguistic “concepts”. nutritious, which the body does not require on a Suppose that the bear has cached its food. When regular basis, and so on. The only way in which these the bear has left the site, we crawl in and remove it. various beliefs may themselves be distinguished, Later the bear returns to the cache site, looking for their particular content identified, is with respect to food. We know that it expects to find food there5, the sentences which express them. But I have admit- not simply on the basis of its “searching” behavior, ted that the bear is not capable of sentences. In but also because we are familiar with its history, which case it does not really have the concept of among other things with the facts that it hid food food. In which case it does not really have any beliefs there in the first place and that when it hides food it about food, in particular that some awaits it at the customarily finds it. How else are we to describe the cache site. Although, faute de mieux, we might at- situation than to say that the grizzly is surprised to tribute “beliefs” to it in order to explain and antici- find that the food is gone? But if this is the correct pate its behavior, this is no more than a facon de parler description (and it is certainly the natural, pre-theo- and should not be taken literally. retical description), then, if it eventually leaves the This line of argument is not persuasive. For one site without coercion or prompting and is still hun- thing, it does not recognize that the identification of gry, it must also be the case that it has given up the objects as such is a matter of degree, and often a belief that food was there. That is, it must have a matter of context. The child, the chef, the glutton all belief about one of its beliefs, namely, that it is false. have a number of beliefs about food. These beliefs DAVIDSON’s claim that “surprise about some are presumably not the same. Does it follow that things is a necessary and sufficient condition of they have different concepts of food, to the point thought in general” (DAVIDSON 1982, p326) seems where we can say confidently that their beliefs about much too strong. For surely surprise is not a neces- food are not really about food? But if it is a matter of sary condition of thought in general. But it is plau- degree, and there is no principled way in which to sibly a sufficient condition. And the fact is that some list a body of beliefs as necessary and sufficient for animals are from time to time clearly surprised. one to have the concept of food (Wittgenstein taught But we have not yet come to the crux of DAVID- us to say that with respect to concepts like this, dif- SON’s position, that in order to have the concept of ferent usages are tied together by no more than fam- belief one must have language. Two related lines of ily resemblances), then there would seem to be no thought are developed to reach this conclusion. reason why the bear couldn’t have some concept of Both depend on granting that having a belief about food (although we would have to identify that con- a belief presupposes having the “concept” of belief. cept in our language), and to that extent beliefs The first line of thought has to do with the indi- about it. That the word “belief” is correctly used here viduation of beliefs. Behaviorists sometimes say that is underlined by the fact that a bear could believe to have a concept of something, say of a cow, it is that food was at the cache site when we had, in fact, enough to be able to discriminate that thing from removed it; that is, the belief in this sort of case other things, to distinguish between cows and non- clearly has the “aboutness” characteristic of inten- cows. Clearly many animals have this power of dis- tionality. crimination with respect to a variety of objects; their For another thing, although we identify beliefs by survival depends on it. But, it might be argued, al- way of sentences, there is something highly prob- though the ability to discriminate objects is a neces- lematic about using this fact to argue that insofar as

Evolution and Cognition ❘ 110 ❘ 1999, Vol. 5, No. 2 The Mental Life of Other Animals an animal is incapable of sentences, it is incapable of self-consciousness, and only beings in possession of beliefs. a language are capable of making judgments (about DESCARTES seems to assume that we understand what is the case). another when he or she speaks to us, and thus that I agree that if an animal is not self-conscious, then in uttering a sentence he or she at the same time it is inappropriate to attribute beliefs to it on any- identifies the content of a particular belief. But as- thing more than an “intentional stance”. The ques- sume that we do not, at least initially, understand tion is: does an animal that lacks language necessar- what he is saying. How do we come to understand ily lack self-consciousness? Let me begin my answer him? To put a very complicated story very briefly, we to this question by making a very general point, and come to understand him only by interpreting his then turn more directly to the DESCARTES–KANT– behavior, including what might be called his “ver- DAVIDSON line of argument. bal” (written, spoken, and “body language”) behav- The very general point is that all of the attention ior. But this interpretation requires, on DAVIDSON’s placed on language in connection with the minds of argument, that we attribute any number of other others, animal or human, is misleading in at least beliefs to him before we can attribute the possession two closely related ways. One, this attention sug- of any concepts to him and thus attach meanings to gests that the use of language simply “reports” what his behavior. That is, we must ascribe any number of is going on inside the mind of others, as though the beliefs (inevitably, our own) to him before we can others were “phoning out” their thoughts and feel- ascribe to him a language in terms of which these ings, with no way of determining whether what is beliefs are to be identified. And what justifies such reported is true. On the other, attention to language ascription in the case of human beings, for example suggests that the mental states ostensibly reported an imagined primitive people whose beliefs are few are in every respect as determinate as the sentences in number and relatively simple (“food” for them reporting them. But both suggestions are mislead- might indicate anything in the environment that ing: the first, if taken seriously, leads to scepticism was green, that is, edible or digestible) and not in the about other minds, the second conflicts with the fact case of at least some other animals if it is not a prej- that our mental states are often unclear and indis- udice in favor of human beings from the outset, that tinct (until, after the fact, we reconstruct them in they have a mental life simply in virtue of sharing a precise verbal form). common physiology with us? DESCARTES says that These problems surface in the “argument by anal- deaf-mutes will always make themselves understood ogy” which KANT takes to be our way “into” other to us. Perhaps (in fact, I am not so sure). But if they minds. It is not, largely because it presupposes a mis- do, it is only because we have knowledge both of the taken view of the way in which behavior stands to context in which they find themselves and of what mind. On this “argument,” to put it briefly, analo- typically motivates their behavior, the same sort of gies or similarities between our behavior, as effect, knowledge we bring to bear when we say that the and that of other animals allows us to infer an anal- dog, scratching at the door, is “trying to tell us some- ogy or similarity between the causes of the behavior, thing”. Difficulties about the minds of animals inev- i.e., an inference to a mind more or less like ours. The itably become difficulties about the minds of other closer the analogies between the behavioral effects, human beings. the more tenable the inference to their mental The other line of thought behind the conclusion causes. that in order to have the concept of belief (and thus There are a variety of problems with this sort of to have beliefs) one must have language has to do “inference”. Unlike other arguments by analogy, to with self-consciousness. For in order to have the con- the existence and structure of atoms for example, it cept of belief, I must (as indicated earlier) have the is impossible in principle to verify its soundness; that concept of a true belief, and this in turn requires that is, there is no way in which we can ever verify I be able to distinguish between how things stand whether our inference to the mind of another ani- with me (my subjective perception of the situation) mal is correct. Moreover, the notion of an “analogy” and how things stand with the world (what is objec- or similarity is of little help, as philosophers of sci- tively the case). But to distinguish in this way be- ence have long since discovered; there will always be tween what is objective and what is merely subjec- an innumerable number of both similarities and dis- tive requires concepts and eventually a language. similarities between any two pieces of behavior, and KANT put it this way: only beings capable of making only some prior agreement about what constitutes judgments (about what is the case) are capable of the relevant similarities or dissimilarities will provide

Evolution and Cognition ❘ 111 ❘ 1999, Vol. 5, No. 2 Gordon G. Brittan, Jr. the foundations for an inference. But to assume such The first is that only commitment to an otherwise prior agreement is already to have “interpreted” the implausible philosophical theory prevents us from behavior in terms of what it expresses. Most impor- seeing that animals make judgments. KANT charac- tant, the argument from analogy suggests, once terizes the making of judgments in terms of the com- again, that the mind is hidden from view, somehow paring of “representations”. But animals compare “trapped inside” the body, and only an inference representations all the time. How else are we to un- from the “outside” can reach it, an inference which, derstand, for example, their indecision when faced, we have just noted, can never be justified. as was BURIDAN’s ass, with alternative, and equally But we can justify the attribution of mind, even attractive, courses of action. with respect to species which seem dissimilar to us The second point concerns the claim that one in many ways. It is a question of “reading” their be- cannot make judgments without using language. havior, in the same sort of way that we read a text. This claim also is false. We humans can, and for most For just as a text does not stand to its author as effect of the time do, make judgements without engaging to cause or conclusion to premise, so too the behav- any linguistic capacity, although many (although I ior of a grizzly bear is an expression of its mind. It is don’t think all) of these judgments can be given a all a question of interpretation. The text is not a “re- linguistic expression. This is, I think, the case when port” on the author’s mind whose accuracy, in the we make “unconscious” judgments, for example nature of the case, can never be verified. In the same when we reach for the hard rolls instead of the crois- way, when we properly interpret some animal’s be- sants at the breakfast table. But even “conscious” havior in its spatial and temporal contexts, there is judgments rarely have the determinate character of little room for asking, yes, but does this really signify a sentence and, when pressed to make them deter- a mind? The behavior itself, contextually under- minate, we sometimes have difficulty doing so (not stood, answers that question. quite being able to find “the right words”). One might contend at this point that we still do The third point is that self-representation can be not “really know” what it is like to be a grizzly bear, ascribed to at least some animals, and then on cannot get “inside” its head. There are two points to grounds which have nothing to do with linguistic be made in this connection. One is that there is no competence. The pioneering work of GALLUP, for ex- difference in this respect between human beings and ample, has shown that chimpanzees and orangutans other animals, and this is all that is necessary for the can learn to recognize mirror-images as representa- purposes of my argument. In the sense indicated, we tions of their own and not of another animal’s body.6 cannot “get inside” the head of other human beings If having a concept or sense of self (or being able to either, even when they tell us what is going on “in- “represent” oneself or refer to oneself) involves beside”. Language is no more conclusive than other ing able to distinguish between self and other, then forms of behavior. The other point is that we can, these animals clearly have a concept or sense of self, and often do, “get inside” the heads of both human and in this respect are “self-conscious”. Are they beings and other animals. We “read” the behavior of aware that they are primates, among other things? those around us all the time, sometimes rightly, Presumably not. But “self-consciousness” is neces- sometimes wrongly, through the behavioral texts sarily a matter of degree. Certainly none of us hu- they supply us with. Reading texts, whether “verbal” mans is completely self-conscious, as individuals or or simply behavioral, and entering into the minds as members of a species. are two aspects of one and the same activity. Experi- Not all primates recognize their mirror images. enced trainers are especially good at “reading” ani- since there is no reason to think that they are any mals, but anyone who has lived in close proximity less self-conscious than chimpanzees and orangu- to a pet for a number of years becomes good at it out tans, we have to say that mirror self-recognition is a of personal necessity. At the same time, both experi- sufficient, but not also a necessary condition of at- enced trainers and perceptive pet owners come to tributing a minimal sense of self. Members of other realize that animals come to “read” their behavior in species indicate that they have some sense of self in a variety of acute ways, occasionally anticipating its other ways. One which intrigues me is the grizzly sequence in ways that were not originally clear to the habit of seeking out positions from which it can see trainers and owners themselves. without being seen. Grizzlies make a special effort to Now to the DESCARTES–KANT–DAVIDSON line of ar- get their entire body behind the available rocks or gument connecting self-consciousness to language. vegetation, as if they had some (perhaps slight) con- There are four points I wish to make about it. ception of themselves as wholes. Whether or not the

Evolution and Cognition ❘ 112 ❘ 1999, Vol. 5, No. 2 The Mental Life of Other Animals bear has a language in which to represent and then jective” picture of things to the extent that it differs, reflect on its behavior, it would seem to me that con- on interpretation, from that produced by the major- cealment begins to reveal a sense of self, just as its ity of its kind. At the very least, it would take an capacity for being surprised reveals that it has some argument to show that this is not also what we do in notion of the difference between true and false be- the case of human beings. liefs. The fourth point is that we can begin to grasp an 8. It is thus in philosophy. We make progress by way animal’s subjective perception of the situation and of showing that certain claims are wrong, in this case thus begin to grasp the distinction that it makes be- the claims advanced by DESCARTES, KANT, and DAVID- tween how things stand with it and how things SON that animals do not really have a mental life. The stand with the world. What is “subjective” is mea- arguments for these claims are not persuasive. Now sured by the distance between the world described the way is open for us to look more closely at animal by the behavioral text that it produces and what behavior in all its richness and to discover once again might be described as the that it is in many respects “real” world. What counts as much like ours, indicating in- Author’s address the “real” world is going to telligence and expressing per- vary; to claim our ordinary im- sonality. Of course, given the Gordon G. Brittan, Jr., Department of Histo- age of things as reality would sorry record of our own species ry and Philosophy, Montana State Universi- be unduly anthropocentric. ty, Bozeman, Montana 59717, USA. on this planet, to say this But we do not need to. We will Email: [email protected] about other animals is not par- know about an animal’s “sub- ticularly to flatter them.7

Notes 6 Someone might object: but don’t GALLUP’s experiments merely show that an animal can recognize its body (and 1 “Spontaneity” is invoked precisely to indicate that in some not also that it can distinguish between itself and others)? cases no explanation can, or need, be given. I understand There is undoubtedly more to self-consciousness than that the purring of cats is not necessarily spontaneous. learning to recognize one’s own body, but it’s a start, just 2 If one were to say that the introduction of “want” at this as it is a start that one begins to make a distinction be- point is question-begging, and that at best the tidbit serves tween those things that are under one’s control and those to give the animal pleasure, then I think we would still need that are not, that one begins to take into consideration the to add that the animal wants or desires things that give it nature and limits of one’s abilities to effect changes, and pleasure. The fact that something gives an animal pleasure so on. At least some animals besides ourselves are going to is not by itself enough to motivate the animal to take steps have to make just these starts on self-consciousness if they to acquire it. are to reproduce themselves successfully. At the same 3 I have changed “presentation” to “representation,” the time, the distinctions involved (e.g., between what is sub- usual translation of KANT’s “Vorstellung”. Otherwise, the ject to my will and what is not) invoke notions of sub- translation is (PLUHAR 1987, p356). stance, causality, and reciprocity (as KANT held) as well as 4 For a summary of the evidence that language is unique to truth (as DAVIDSON contends) even though languageless humans, see (PINKER 1995), chapter 11. animals are of course not able to state, much less to anal- 5 “Expectation” is, of course, an intentional notion and thus yse, these notions or lay out the truth-conditions for their might seem to imply without argument that the bear has a various beliefs. mental life. But again, there are objective criteria which, 7 An earlier version of this paper was read at the College de when satisfied, allow us to apply this notion, and it is diffi- France in 1997, under the title “La vie mentale des ani- cult to see how one might avoid its use. The “hope of eat- maux”. I’m grateful for their comments on that version to ing” which DESCARTES attributes to the magpie is just James ALLARD, Christopher PINET, and Jules VUILLEMIN. another way of talking about the bird’s expectations.

References North-Holland: Amsterdam. Kenny, A. (1970) Descartes: Philosophical Letters. Clarendon Davidson, D. (1982) Rational animals. Dialectica (36) 4. Press: Oxford. Dennett, D. (1978) Intentional systems. Brainstorms. Brad- Pinker, S. (1995) The Language Instinct. Harper: New York. ford Books: Cambridge MA. Pluhar, W. (1987) Kant’s Critique of Judgment. Hackett Pub- Gallup, G. G. Jr. (1983) Towards a comparative psychology of lishing Company: Indianapolis. mind. Melgren, ed., Animal Cognition and Behavior.

Evolution and Cognition ❘ 113 ❘ 1999, Vol. 5, No. 2 Gordon M. Burghardt

Conceptions of Play and the Evolution of Animal Minds

HE FOLLOWING THREE Abstract comprehended process of T quotations from a operation (natural selec- century ago set the stage Issues of the origins of play must be separated from tion), and made a strong for the topic of this essay. issues of the functions of play in animal lives. A com- case for including behav- “A satisfactory theory parative analysis of the role of play in the evolution of ioral and psychological of play is still wanting, animal minds, however, is only as successful as the processes as key elements and yet a man does not conception of play that is used. Here a set of five crite- of a unified evolutionary learn through any kind of ria for recognizing playful acts wherever they may oc- biology. Although later instruction or study in cur phylogenetically is presented. In addition, downplayed by many bi- later life anything like so primary processes deriving from life history, physio- ologists themselves, much as the child learns logical, and psychological factors are distinguished philosophical and psy- in the first four years of from derived secondary processes. The latter may be chological issues were the his careless existence, the source of the behavioral, cognitive, and emotional critical goads to DARWIN through the perceptions changes resulting from play. Surplus Resource Theory in his compulsion to ex- and ideas acquired in his is described as a means of accommodating both pri- plain the origins of spe- play” (PREYER 1893, p42). mary and secondary processes in play and also aids in cies. This is evident in the “Science has shown predicting the conditions under which playlike behav- notebooks he kept in the that the embryonic pe- ior is observed. Brain size and intelligence are not nec- 1830s and 1840s. riod of physical develop- essarily linked with play, nevertheless, playful Although DARWIN ment is a masquerade of behavior may be linked to creativity and novel behav- (1859) developed detailed long-vanished forms of ior throughout both evolutionary history and individ- explanations for the ori- life. In like manner the ual lives. gin of many instinctive children of each new gen- behaviors in wild and do- eration seek instinctively Key words mesticated animals (e.g., to revive the life that is be- slavemaking in ants, hive- Play, evolution, cognition, creativity, development. hind them and in their fa- making in bees, pointing vorite occupations and in hunting dogs, and amusements re-enact the prehistoric experiences of pouting in pigeons) he was less successful in explain- mankind… Can we doubt that the order of history ing the origin of the complex mental life of the should be the order of education, and that before we “higher” animals and people. His two books dealing teach the child to read and write we should aid his with psychological topics (DARWIN 1871; 1872) fo- efforts to repeat in outline the earlier stages of hu- cused much more on the similarities between human development” (BLOW 1894, pp125–126). man behavior and that of other mammals than on “Play being the all important business of childish the origin of the comparable abilities. Indeed, al- life, and all play consisting more or less in acts, it is though the evolution of intelligence and mental play then, above all, that we must seek for the begin- abilities were a prime interest of the early POSTDAR- nings of voluntary activity” (COMPAYRÉ 1902, p151). WINIAN writers, play did not figure prominently in The search for understanding animal mental life this picture (e.g., ROMANES 1883; MORGAN 1894). By and cognitive abilities became a scientifically criti- the time of Mark BALDWIN’s great Dictionary of Philos- cal issue in the aftermath of the DARWINIAN revolu- ophy and Psychology, however (BALDWIN 1902), the tion. DARWIN’s writings legitimized evolutionary importance of play in the origins of diverse creative approaches to animal diversity, provided an easily and artistic expressions was seriously discussed. This

Evolution and Cognition ❘ 114 ❘ 1999, Vol. 5, No. 2 Conceptions of Play and the Evolution of Animal Minds aspect of play was one of the few in which the three the origins and evolution of play I think it possible major early play theorists, Herbert SPENCER (1872), to outline a testable model of both how playful ac- Karl GROOS (1901), and G. Stanley HALL (1904), could tivities arise and how they can shape the future be- all agree. In human developmental psychology the havior patterns of animals, including facilitating role of play in the early learning and mental life of complex mental processing. I will present a way of infants and young children was also widely accepted looking at play as a potential source of the rapid during this period (e.g., PREYER 1893; BLOW 1894; evolution of behavioral complexity in endothermic COMPAYRÉ 1902) as suggested by the opening quota- vertebrates and as a resource for ontogenetic tions. change in behavior that can be both conservative In the current century many formulations of the and novel. role of play in the development of virtually every In order to evaluate the role of play in creating or aspect of human behavior have been advanced influencing the animal mind we must first identify (HARTLEY/FRANK/GOLDENSON 1952). This interest in- the phenomena to be considered playful. This is not tensified after mid-century, especially following the a trivial exercise, especially if one wants to discuss influential lead of PIAGET (e.g., 1962), the list of re- the origins of behavioral phenotypes (BURGHARDT cent books is long and diverse (e.g., PIERS 1972; GÖR- 1998a, b). Sometimes even the most accepted and LITZ/WOHLWILL 1987). Theoretical attempts to derive clear examples of play can be problematic. The tra- art from play have likewise intensified (HUIZINGA ditional categories of play are locomotor, object (in- 1949; ALLAND 1977; DISSANAYAKE 1992). For over a cluding predatory), and social; yet these are fre- century, then, claims have been made concerning quently not distinct. In order to compare play across the role of play in the mental life and behavior of species we must have some way of isolating play be- human and, less often, nonhuman animals. These havior using a definition that is not specific to any claims, derived from theory, have remained contro- one kind of play (such as social play). Another im- versial and largely unsupported with empirical data portant problem is that the causes and functions of (MARTIN/CARO 1985; SMITH 1988, 1996). Empirical play may differ for different kinds of play, differ support that is advanced is often correlational or ex- throughout ontogeny, and differ across species. perimentally flawed. Nevertheless, since the time of With humans the problems of characterizing play FROEBEL and the Kindergarten movement (BLOW often seem insurmountable. Some have argued that 1894; FROEBEL 1895), many countries have included play is a romantic notion of middle-class Western play in daycare and educational systems as a means industrialized societies deriving from an artificial of facilitating learning. Claims have been made that work-play dichotomy (SUTTON-SMITH/KELLY-BRYNE play facilitates learning (including reading and 1984), others have held that play is the real ‘work’ of mathematics), imagination, socialization, behav- other species (GROOS 1898)! Playfulness, conceptual- ioral flexibility, mental agility, and creativity. Evolu- ized as participating in games, has been viewed as tionary arguments based on selected comparative intrinsic to war, competition, and rituals of all types data often accompany these assertions (c.f., FAGEN (cf. HUIZINGA 1949). Nursery teachers “adhere to the 1984). Certainly, such claims, in spite of the lack of term ‘play’ to describe any or every activity of young experimental support, have been viewed as plausible children (excepting biological necessities such as by many who have studied play. Even those most eating, sleeping, and elimination)” (HUTT/TYLER/ responsible for disconfirming experiments find it HUTT/CHRISTOPHERSON 1989, p221). Such rhetorical difficult to completely dismiss these putative bene- ploys, along with legitimate difficulties in defining, fits (e.g., SMITH 1996). describing, and analyzing playful behavior, has con- To make progress in understanding the role of tributed to the lack of serious and careful scientific playfulness in cognitive life we need to ground the attention to the role of play in development and evo- phenomena of play in a phylogenetic context. lution. Thus we first need to agree on how to recog- What are the processes leading to playfulness nize play. throughout vertebrate evolution? This question is prior to the search for putative consequences of Deﬁning play play. Elsewhere I am preparing an extended treatment of animal play, and the historical, defini- After reading many treatments of play one is proba- tional, theoretical, and comparative aspects will be bly best advised to throw up one’s hands and say, explored in greater depth and with supporting data play just cannot be usefully defined. E. O. WILSON and citations (BURGHARDT, in press). By focusing on (1975, p164) has written: “No behavioral concept

Evolution and Cognition ❘ 115 ❘ 1999, Vol. 5, No. 2 Gordon M. Burghardt has proved more ill-defined, elusive, controversial, pressed; that is, it includes elements, or is directed and even unfashionable.” More recently, MITCHELL toward stimuli, that do not seem to contribute to (1990, p197) has averred that “Play is the hobgoblin current survival. of animal behavior, mischievously tempting us to The second criterion for recognizing play is that the succeed in what, judging from the number of failed behavior appears to be spontaneous, voluntary, in- attempts, seems a futile task: defining play.” Never- tentional, pleasurable, rewarding or autotelic (“done theless, MITCHELL and many leading play research- for its own sake”). ers have advanced their own definitions. FAGEN The third criterion for recognizing play is that it (1981) has gathered together many of these. None differs from the “serious” performance of ethotypic of these list the minimal criteria necessary to set behavior in at least one respect: incomplete (gener- play apart from other phenomena in a rather precise ally through inhibited or dropped final elements), way. Here I cannot review even the most recent and exaggerated, awkward, precocious, or involves be- influential definitions of animal play in any depth. havior patterns with modified form, sequencing or For illustrative purposes, however, I will quote what targeting. is arguably the currently most widely cited defini- The fourth criterion for recognizing play is that it tion, one that I have used repeatedly myself be repeatedly observed in a similar, but not rigidly (BEKOFF/BYERS 1981, p300): stereotyped, manner, during at least a portion of the “Play is all motor activity performed postnatally animal’s ontogeny. that appears to be purposeless in which motor pat- The fifth criterion for recognizing play is that it is terns from other contexts may often be used in mod- observed when the animals are adequately fed, ified form and altered temporal sequencing.” healthy, and free from stress (e.g., predator threat, Note that the use of “may” hedges our ability to harsh microclimate, crowding, social instability), or apply the definition to phenomena that we do not intense competing motivations (e.g., feeding, court- already consider play; in actuality we are left with ship, resource competition, nestbuilding): In other the apparent lack of purpose in the target behavior words, the animals are in a “relaxed field.” Note, as the central and essential component. Stereotyped however, that this does not mean that the play itself dysfunctional abnormal behavior such as feather- is free from stress or risk. picking in birds cannot be excluded in this defini- Now while these criteria do not provide a nice tion as BEKOFF/ALLEN (1998) acknowledge. The mod- crisp one line definition, they seem to cover every ification of this definition by MARTIN/CARO (1985) accepted example of play and to exclude, with con- only compounded this problem. sistent reasoning, much of the problematic and con- In fact, the bottom line of most definitions is that troversial. It is not limited to social play, or juvenile the behavior seems to lack an immediate function or play, or play that involves special signals. It excludes purpose. Some writers follow theorists such as PIAGET stereotyped abnormal behavior as well as explora- (1962) and list individual criteria and pronounce tion. Although much behavior called play is vigor- each one inadequate in defining play as the category ous and energetic from a human perspective, this we think it to be. But what if we package the many also is not required. The apparent fuzziness of the different criteria into sets of similar type and see if boundaries in applying individual criteria in some we can now specify those that are essential? Since for cases reflects the fact that playfulness is on a contin- my purposes I need to identify possible candidates uum with a variety of nonplay behaviors. In fact, the for play in a wide diversity of species, I extracted relationship of play to both exploratory behavior from the literature on animal and human play, and and curiosity (BURGHARDT 1984; HUTT et al. 1989) my own 35 years of behavioral study, a set of diverse and stereotyped behavior need to be more fully re- criteria and arranged them in a formal and definitive solved if we hope to make progress in understanding manner. Here I present for, the first time in print, the both the phylogeny and ontogeny of play. But if all following five criteria, all of which must be satisfied in the above criteria are met the behavior deserves to at least one respect before a given behavior is accept- be called play, a distinct category. able as an instance of play. It is important to note Note the following about the above criteria. They that none of them alone are sufficient for applying make no assumptions about the function of play. the play designation. They avoid the controversy of whether or not there The first criterion for recognizing play is that the is a separate motivation underlying play. By includ- performance of the behavior does not appear coming terms that some find essential (pleasure, inten- pletely adaptive in the context in which it is ex- tionality, voluntary) but which other scientists find

Evolution and Cognition ❘ 116 ❘ 1999, Vol. 5, No. 2 Conceptions of Play and the Evolution of Animal Minds unclear, unscientific, or difficult to apply (as in a Distinguishing between primary and fish), the different conceptual and labeling systems secondary processes issues in the study scientists use can be accommodated. I am confident of play that if new behavioral labels arise they can be readily added to one of the five criteria sets. Of course, the Play may be both a product and cause of evolution- reliability of the observations is also important, es- ary processes; that is, playful activities may be both pecially in the more exotic and less-studied animals sources of enhanced behavioral and mental or those that have been generally dismissed as not improvements as well as a byproduct or remnant of playing (e.g., ants, reptiles). Thus independent ob- prior evolutionary and ontogenetic events. It is the servations by several qualified investigators and conflict between these two views of play that I view their documentation are important in firmly accept- as being at the root of the paradoxical nature of ing claims for some species. play. The early conflicts between the functional The end result of this endeavor has been that, for view of GROOS (1898) and the recapitulation view of me, the five criteria have been particularly valuable HALL (1904) were over misunderstandings of these in evaluating claims and evidence for play in prob- very issues, and the legacy of this controversy still lematic groups such as ants, fishes, monotremes, and haunts the study of play. To break the impasse and reptiles (e.g., BURGHARDT 1998b). They can also help make genuine progress, we need to keep conceptu- guard against the anthropomorphism that has been ally separate the primary processes by which “play” endemic to the study of animal play since its incep- originated and evolved in ancient vertebrates and tion. However, critical anthropomorphism (BURG- their modern descendants and the ways in which HARDT 1985; 1997) can be used to retain the motiva- these may have provided an essential scaffolding tional and hedonic aspects of play without losing an from which the highly diverse and complex struc- objective and public assessment of play. These latter tures of mammalian play could be built through a aspects of play will become clarified as physiological series of secondary or derived processes. It is these and neural measures of internal states become more derived processes that may provide cognitive refined (BURGHARDT 1995). advantages to those animals engaging in play. But it It is also important to note that there is no impli- is a mistake to think that play originally evolved in cation that all behaviors satisfying the criteria for order to provide such advantages. being classified as play are necessarily the same in Without going into all the supporting details, I their causal (neurobiological) mechanisms or that will approach play in animals from an evolutionary/ they are homologous. Different species will play in developmental perspective called Surplus Resource different ways depending on their normal behavior, Theory or SRT. This view has several components ecology, ontogenetic development, metabolism, including a reliance on studying animals, including neural organization, and phylogeny. In fact, it is be- ectothermic vertebrates, with less rich play in con- cause play is a heterogeneous category that arises trast to the exclusive focus on highly playful taxa repeatedly in vertebrate evolution that we need such (primates, canids, felids, ungulates) usually pursued. a litmus test. The criteria help us to sort out the pro- SRT incorporates physiology (e.g., activity metabo- cesses and variables influencing playfulness and lism, thermoregulation), life history (e.g., parental point the way to experiments and critical phyletic care, altriciality, food niche), and psychology (e.g., comparisons. level of stimulation, learning). As in DARWIN’s dis- Now that we have (hopefully) criteria that allow covery of natural selection, domestication has pro- us to confidently identify play and target missing or duced some intriguing parallels to what may have less known aspects of the putative play, what can we taken place during vertebrate phylogeny. do to explore the role of play in the evolution of The next move is to distinguish between primary animal mentality? One approach would be to clas- and secondary processes in play. It is the latter that sify the kinds of play in a species and study them are most often viewed as leading to the claims made closely to see what they contribute to cognitive and in the first paragraph. An example of a secondary pro- affective life. An alternative, followed here, is to see cess would be that vigorous rough and tumble play of if there are commonalities in the life history, physi- young rats and dogs enhances adult performance, ology, behavior, and psychology of the animal promotes socialization, or increases behavioral flexi- groups in which play has evolved and whether these bility. However, it is precisely because the more pri- commonalities can help us understand the evolu- mary processes of play have been ignored that prediction of animal mentality. tions made from secondary processes have fared

Evolution and Cognition ❘ 117 ❘ 1999, Vol. 5, No. 2 Gordon M. Burghardt poorly. An example of an important primary process The phylogenetic origins of play derived from SRT would be the role of metabolic rate or parental care in production of ‘surplus’ behavioral Unlike in the nineteenth century, in this century “mutants” that could in turn be selected ontogeneti- play has been viewed as primarily a property of cally and phylogenetically. A secondary process de- mammals. GRIFFITHS (1978) even listed play as one rived from a detailed consideration of primary proof the defining attributes of mammals and thus a cesses would be the recent correlation, by BYERS/ reason why monotremes should be considered true WALKER (1995), of the onset of vigorous motor play mammals and not “quasi-mammals”. Birds have with the period in ontogeny with the timing of per- been acknowledged as engaging in play (reviews in manent long-term changes in the muscular and cere- FICKEN 1977; FAGEN 1981; ORTEGA/BEKOFF 1987). bellar systems of several species of domesticated ani- However, a clear discontinuity was recognized in mals. With children, the most famous secondary that play was considered common in the endother- process would be PIAGET’s conception of assimilation mic vertebrates (especially mammals) and absent in and accommodation as the processes by which chil- the traditional “cold-blooded” vertebrates, includ- dren cognitively adapt to their environment (PIAGET ing fish, amphibians, and nonavian reptiles (rep- 1962). Play, especially during the early sensorimotor tiles from here on). As someone who has spent stage, is essential for this process. In this early stage decades studying reptile behavioral ontogeny, and reflexes and circular reactions are the basic building being impressed with the degree of complexity and blocks. Piagetian processes have been explored in an- learning in their behavior (BURGHARDT 1977), I had imals, especially nonhuman primates. However, to agree that I had seen nothing that appeared as while study of these stages can help uncover the on- playful. Why? I developed a list of life history, togenetic origins of mental life, they tell us little about physiological, and psychological factors that how these processes have been shaped by evolution appeared to be different between the typical reptile to have these essential roles. Interestingly, PIAGET and the typical mammal. Although the full array of himself (PIAGET 1972, p27) does not see his work has evidence underlying these differences is complex having any significant role in education, stating “I am (BURGHARDT 1984; 1988) the major premises can be not an educator. I have no advice to give.” He does quickly summarized. state that active methods involving self-discovery by First, except for crocodilians, most neonatal rep- the child seem most useful for long-term retention. tiles are not cared for by their parents and conse- In short, my approach is more basal than PIAGET’s quently must devote their activities, immediately with children or the functional studies on highly upon birth or hatching, toward obtaining food, derived play in mammals. The study of the role of finding shelter, avoiding predation, and growing play in the evolution of animal minds should ini- rapidly on their own. As a result there is limited tially focus on the primary processes leading to be- “safe” time or opportunity for practicing or perfect- havior satisfying the five criteria above. Some writers ing behaviors to be used in an uncertain future. have argued that play is a “random process genera- Most young reptiles are very small and selection will tor” and similar views abound. A more sophisticated have shaped abilities that enhanced juvenile sur- version is FAGEN’s (1974) comparison of the variabil- vival, such as remaining quiet and inconspicuous. ity seen in play to that found in genetic systems. Such behaviors are, of course, incompatible with That is, in play we see similarities to chromosomal vigorous play. There is also no source of high fat, inheritance in the sense that in play sequences we high protein nourishment at little cost, as in juve- can we find recombination, fragmentation, transloca- nile birds and mammals. Nonavian reptiles are tion, and duplication. Such variation can be the raw highly precocial and superficially resemble, and be- materials for natural selection to operate upon inso- have like, miniature adults. This is not to say that far as the behavioral variation also has some inher- young reptiles do not learn many things as they go ited component. However, play is far from random about their serious activities. Furthermore, reptiles and it seems to operate within fairly tight bound- grow as rapidly as food resources allow; unlike mam- aries. PELLIS (1993) has documented how play fight- mals and birds there is no necessary age defined pe- ing can differ among closely related species with lit- riod of youth. Young reptiles, unlike young mam- tle or no overlap. Thus the challenge is not just to mals, do not get fat; they just grow fast. This state that play creates novel behavioral phenotypes characteristic of reptiles is termed nondeterminate but to uncover the actual behavioral processes un- growth. The last life history characteristic I will derlying such behavioral variation. mention is that reptiles typically have large litter

Evolution and Cognition ❘ 118 ❘ 1999, Vol. 5, No. 2 Conceptions of Play and the Evolution of Animal Minds and clutch sizes with high juvenile mortality. Thus rich repertoire of possible movements of limbs and delayed benefits of play would be less worth a cur- face seen in many mammals. The lack of parental rent risk. care may have prevented the evolution of social Second, reptiles typically are metabolically con- bonding and affiliation to the extent seen in birds strained from performing vigorous, energetically and mammals. Certain kinds of complex ecological expensive behaviors, especially those of any dura- and social interactions and operations upon the en- tion. This is due to their low resting and maximal vironment seem to characterize the more playful metabolic rates, limited aerobic capacity, and long species. Interestingly it is those reptiles with more recovery times after anaerobic expenditures. Rep- variable diets and foraging techniques in which tiles have about 10% the metabolic rate of a compa- more exploration, curiosity, and, rarely, play, are rably sized mammal. But, in fact, most reptiles are found. much smaller than mammals. Both metabolic rates The chief difference of this approach from previ- and typical body sizes are further reduced in am- ous ones is that the initial advantages of incipient phibians. Small body size leads to greater heat loss playlike behavior did not involve any particular and even more efforts to behaviorally thermoregu- functions (such as perfecting later behavior, increas- late. Taken together, these factors constrain the per- ing endurance, and incorporating flexibility). Play formance, and selection for, vigorous playlike be- can evolve independently whenever physiological havior with no current function (BURGHARDT 1984). (including neural), life history, metabolic, and psy- Third, animals in boring, unstimulating environ- chological conditions, in conjunction with a spe- ments would be most likely to engage in behavior cies’ behavioral repertoire, reach a threshold level. to relieve sensory and response deprivation and to In mammals, however, play was also a disparate increase arousal. Such boredom might be expected constellation of ad hoc mechanisms that could be in the well provisioned and protected environments used to maintain the continuity of endothermic provided by endotherm parents. This factor may be and behavioral systems bridging the periods of ju- the critical one in the consistent finding that well- venile dependence and adult responsibilities. The cared for captive animals play much more than their deterioration of certain neonate response systems wild counterparts. If the juvenile animals being through parental care, the lowering of thresholds buffered from the demands of survival have a com- and the broadening of the effective stimuli, and the plex repertoire of evolved and active behavior pat- increased aerobic metabolic capacities resulting terns, then they have a behavioral resource to draw from endothermy led to a reorganization of devel- on when deprived of stimulation. opmental processes so that play, as well as other As a guide to identifying behavior in nonavian experiential avenues, was not only available to ho- reptiles that could be candidates for traditional ap- meotherms but may well have had to be exploited pearing play, I subsequently (BURGHARDT 1988) used by them for continued survival, by replacing lost, the above and related physiological, psychological, suppressed, or maturationally delayed “instincts.” and life history processes, to predict that mammalian or avian-like play in reptiles, whether locomo- The role of play in the origins of tor, object, or social, should be rare and occur only in specific contexts in which those factors facilitat- “higher” vertebrate minds ing play in mammals and birds are also present. A If I am correct, the rapid evolution of cognitive series of predictions was supported that suggest that capacities in mammals is due to processes initiated the three factors listed above are involved in facili- by the advent of parental care and the opportunity tating play. What they allow us to do is to predict in to use the surplus resources of time, energy, and what mammalian groups we would expect to find innate propensities already present from a preco- the most complex play and the most time spent in cial existence. This process probably took initially play. For example, on energetic grounds one would many millions of years and was rather slow in com- expect that aquatic mammals would be particularly ing. Many species and radiations of mammals and playful—and they are. The play we have found in birds developed along these lines and today we see turtles has been in aquatic turtles (BURGHARDT/ snapshots of where they are along a continuum of WARD/ROSSCOE 1996; BURGHARDT 1998b). increasing amounts of such playfulness. Play Object play is found most often in predators and behavior, as defined by the 5 criteria, became more generalist feeders that rely on manipulation of prominent. First it was an epiphenomenal byprod- limbs and mouth. Again, reptiles do not have the uct that may have been produced by the surplus

Evolution and Cognition ❘ 119 ❘ 1999, Vol. 5, No. 2 Gordon M. Burghardt

energy process postulated by SPENCER (1872), but reared dogs, especially to novel situations; they are almost immediately could be utilized as a source of less neophobic. Actually, it has been shown that behavioral neophenotypes. such dogs are rather stupid; even the most trivial Play can be viewed as providing opportunities items amuse them persistently (HEBB 1972). There is that can be exploited in adapting to changing phys- also considerable evidence that habituation, rapid iological, environmental, and developmental con- boredom with stimuli, is a good measure of intelli- ditions. Perhaps in species in which play is promi- gence and can be used with preverbal infants with nent it is a directed, even intentional, source of considerable success (COLOMBO 1993). If becoming adaptive variability. I actually think this is possible bored is a sign that one has mastered what there is only rarely. First, play, especially social play, can to know or do with an object, then choosing stimuli itself become highly ritualized, have important sig- that are more complex might be the way to deal with nal functions (BEKOFF 1995), and show consistent this situation. This is exactly what Piagetian mastery species and age-related patterns (PELLIS 1993; play is directed towards. THOMPSON (1998) has ad- BURGHARDT 1998a, b). This may give individuals op- vocated a similar process in animals she terms self- portunities to learn about conspecifics and how to assessment. But curiosity killed the cat. Rodents, respond to them. It may provide opportunities to fish, and tadpoles that stray too far from secure re- exploit others. But it is also possible that social play treats have all been shown to be more at risk of pre- is in large measure a response to social needs and dation. Thus, prudent behavior has some advan- motivational system legacies from when such be- tages over the exuberant “joy of life” excitement haviors needed to be performed for serious and im- many view as the essence of play. mediate functional reasons. In this way HALL’s But perhaps the most telling argument against a (1904) recapitulation view has some merit once the simple brain-size and intelligence relationship with rigid sequencing is eliminated as necessary. Only in play can be found by using domestication as a play’s most derived state does it do the kind of ser- model system for testing surplus resource theory. vice that the prevailing practice theory, that goes Many of the processes that I postulate as having back to GROOS (1898), requires. occurred in juvenile mammals with the onset of parental care and the consequent buffering from Play and intelligence are not the demands of life, are also found in domesticated species. PRICE (1984) has documented these pro- necessarily linked cesses completely independently of the theory out- Support for the role of play in cognition is often lined here. The match is remarkable. The retention advanced in the following way. Play is found most of juvenile morphology and behavior into adult- commonly in birds and mammals that have large hood is seen particularly clearly in dogs. Litter sizes brains and are considered “intelligent” (FAGEN increase and many behavioral skills shown in wild 1981). Play and curiosity are thus marks of intelli- populations show deterioration. But playfulness in- gence and necessary for the development of a creases! Brain size decreases! KRUSKA (1987) has sophisticated mind. The rough correlations are cer- shown that in many mammals domestic popula- tainly impressive. Primates and aquatic mammals tions have brain sizes 5% less than wild populations have large brains and are considered highly playful, after only a few generations. This is shown in both as are the big-brained birds in the parrot and corvid a rodent (bank vole) and carnivore (polecat). Ferrets families (FAGEN 1981). Sloths, insectivores, and were domesticated from wild polecats 2500 years many less playful mammals and birds have smaller ago and now have brain sizes 30% smaller than wild brains. The best evidence for fish play is in mormy- animals. It would seem that a careful study of play rid fish (e.g., MEDER 1958) that have huge brains behavior in wild animals undergoing domestica- larger, relatively, than most mammals (BUTLER/ tion might be a most useful method to see how HODOS 1996) and may even exceed the brain/body buffering animals from the harsher aspects of exist- weight ratio of adult human beings (BULLOCK ence may change the amount, type, and frequency 1977)! Furthermore, animals with more altricial of play as well as cognitive capacities. It is impor- young play more or more complexly than even tant that critically anthropomorphic methods be close relatives that are more precocial. used. We may find it all too easy to conclude that Yet there are some problems with this putative the more affectionate or compliant dog is smarter relationship. Dogs raised in social isolation appear than the more elusive, unpredictable high-strung to be far more social and curious than normally wolf or devious coyote.

Evolution and Cognition ❘ 120 ❘ 1999, Vol. 5, No. 2 Conceptions of Play and the Evolution of Animal Minds

Back to creativity ent reasons than they advanced. I have concluded that after a slow period of evolutionary reorganiza- The view of play and its origins outlined here sug- tion in behavioral ontogeny (BURGHARDT 1988), gests a rather slow development through out verte- play facilitated rapid behavioral and mental devel- brate evolution of playlike behavior under suitable opment. ecological and physiological conditions. Using the There are some intriguing parallels between above criteria for play, I think there is convincing recent work exploring creativity by individual evidence that play evolved independently in sev- human beings (CSIKSZENTMIHALYI 1996) and the role eral lines of fishes, turtles, squamate reptiles, croc- of play facilitating rapid evolutionary creativity. odilians, birds, and marsupials, as well as virtually Creativity is often evidenced by periods of informa- all families of placental mammals (BURGHARDT, in tion gathering, rehearsal, and consolidation before press). Endothermy and parental care interacted a rapid productive phase, and this can be simulated with this early incipient play and gave it new roles in a neural model (MACLENNAN, in prep.). Once and functions, a process still ongoing. The rapid begun, creative achievements may flow rapidly. rate of evolutionary changes in endothermic ani- Thus like the model of play, after a slow process “up mals, especially mammals, in genome size, brain a mountain” a turning point is reached at the top size, and behavioral complexity are remarkable and the trip down is much faster than that going up. and still little understood. Play may have been a We can also extend this to cultural change where, major engine in this rapid cascade of evolutionary after an innovation slowly spreads across a popula- change that led to increased cognitive complexity. tion, a point is reached in which the society is rapidly This may occur by moving initially playful and irreversibly transformed. This seems to have responses to more serious endeavors and functions been the case with the automobile, air travel, the so that the once playful behavior has been trans- telephone, and, most recently, the computerization formed and is now outside of the realm of play of many of our activities. But these changes, due to using the criteria developed above. Thus a paradox: the efforts of countless creative individuals, also play may be essential in the evolution of mind but have their origins in the playful tinkering and sym- what remains as play may bolic rehearsals that derive not be as valuable as often from processes underlying the believed. Much play leads Author’s address criteria for recognizing play in nowhere. We have to avoid even the “lowliest” animals! Gordon M. Burghardt, University of Tennes- the idealization of play, as Tracing out these connections see, Knoxville, Tennessee 37996-0900, USA SUTTON-SMITH/KELLY-BYRNE is a most exciting and signifi- Email: [email protected] (1984) argued, but for differ- cant challenge.

Note

Based on a talk given at the “Animal Mind” workshop, Insti- tute for Brain Research at the University of Bremen, Germany, September 19, 1997.

References ny and phylogeny of mammalian social and locomotor play: an ethological hornet’s nest. In: Immelmann, K./Bar- Alland, A. Jr. (1977) The artistic animal: an inquiry into the bi- low, G. W./Petrinovich, L./M. Main (eds) Behavioral devel- ological roots of art. Anchor Press: Garden City NY. opment: the Bielefeld interdisciplinary project (pp. 269– Baldwin, J. M. (ed) (1902) Dictionary of Philosophy and Psy- 337). Cambridge University Press: Cambridge. chology. Macmillan: New York. Blow, S. E. (1894) Symbolic education. A commentary on Bekoff, M. (1995) Play signals as punctuation: The structure of Froebel’s “Mother play”. D. Appleton: New York. social play in canids. Behaviour 132: 419–429. Bullock, T. H. (1977) Introduction to nervous systems. San Bekoff, M./Allen, C. (1998) Intentional communication and Francisco: Freeman, W. H. social play: how and why animals negotiate and agree to Burghardt, G. M. (1977) Learning processes in reptiles. In: play. In: Bekoff, M./Byers, J. A. (eds) Animal play: Evolu- Gans, C./Tinkle, D. (eds) The biology of the Reptilia. Vol. 7, tionary, comparative, and ecological perspectives (pp. 97– ecology and behavior. Academic Press: New York, pp. 555– 114). Cambridge University Press: Cambridge. 681. Bekoff, M./Byers, J. (1981) A critical reanalysis of the ontoge- Burghardt, G. M. (1984) On the origins of play. In: Smith, P.

Evolution and Cognition ❘ 121 ❘ 1999, Vol. 5, No. 2 Gordon M. Burghardt

K. (ed) Play in animals and humans. Basil Blackwell: Lon- Groos, K. (1901) The play of man. (Translated by E. L. Bald- don, pp. 5–41. win.) D. Appleton: New York. Burghardt, G. M. (1985) Animal awareness: Current percep- Hall, G. S. (1904) Adolescence: Its psychology and its relations tions and historical perspective. American Psychologist 40: to physiology, anthropology, sociology, sex, crime, religion 905–919. and education. D. Appleton: New York. Burghardt, G. M. (1988) Precocity, play, and the ectotherm- Hartley, R. E./Frank, L. K./Goldenson, R. M. (1952) Under- endotherm transition: Superficial adaptation or profound standing children’s play. Columbia University Press: New reorganization? In: Blass, E. M. (ed) Handbook of behavior- York. al neurobiology. Vol. 9. Plenum: New York, pp. 107–148. Hebb, D. O. (1972) Textbook of psychology. 3rd edition. W. Burghardt, G. M. (1995) Brain imaging, ethology, and the B. Saunders: Philadelphia. nonhuman mind. Behavioral and Brain Sciences 18: 339– Huizinga, J. (1949) Homo ludens. (Translated by R. F. C. Hull.) 340. Routledge: London. Burghardt, G. M. (1997) Amending Tinbergen: A fifth aim for Hutt, S. J./Tyler, S./Hutt, C./Christopherson, H. (1989) Play, ethology. In: Mitchell, R. W./Thompson, N. S./H. Miles, L. exploration and learning: A natural history of the pre- (eds) Anthropomorphism, anecdotes, and animals. SUNY school. Routledge: London. Press: New York, pp. 254–276. Kruska, D. (1987) How fast can total brain size change in Burghardt, G. M. (1998a) The evolutionary origins of play re- mammals? Journal für Hirnforschung 28: 59–70. visited: Lessons from turtles. In: Bekoff, M./Byers, J. A. (eds) Martin, P./Caro, T. M. (1985) On the function of play and its Animal play: Evolutionary, comparative, and ecological role in behavioral development. Advances in the Study of perspectives. Cambridge University Press: Cambridge, pp. Behavior 15: 59–103. 1–26. Meder, E. (1958) Gnathonemus petersii (Günter). Zeitschrift Burghardt, G. M. (1998b) Play. In: Greenberg, G./Haraway, für Vivaristik 4: 161–171. M. (eds) Comparative psychology: A handbook. Garland: Mitchell, R. W. (1990) A theory of play. In: Bekoff, M./Jamie- New York, pp. 757–767. son, D. (eds) Interpretation and explanation in the study of Burghardt, G. M. (in press) The genesis of animal play. Cam- animal behavior. Vol. 1: Interpretation, intentionality, and bridge, MA: MIT Press. communication. Westview Press: Boulder, pp. 197–227. Burghardt, G. M./Ward, B./Rosccoe, R. (1996) Problem of Morgan, C. L. (1894) An introduction to comparative psy- reptile play: Environmental enrichment and play behavior chology. Walter Scott: London. in a captive Nile soft-shelled turtle, (Trionyx triunguis). Zoo Ortega, J. C./Bekoff, M. (1987) Avian play: Comparative evo- Biology 15: 223–238. lutionary and developmental trends. Auk 104: 338–341. Butler, A./B./Hodos, W. (1996) Comparative vertebrate neu- Pellis, S. M. (1993) Sex and the evolution of play fighting: A roanatomy: Evolution and adaptation. Wiley: New York- review and model based on the behavior of muroid ro- Liss. dents. Play Theory and Research 1: 55–75. Byers, J. A./Walker, C. (1995) Refining the motor training hy- Piaget, J. (1962) Play, dreams and imitation in childhood. pothesis for the evolution of play. American Naturalist 146: (Translated by C. Gattegno and F. M. Hodgson) W. W. 25–40. Norton: New York. Colombo, J. (1993) Infant cognition: Predicting later intellec- Piaget, J. (1972) Some aspects of operations. In: Piers, M. W. tual functioning. Sage: London. (ed) Play and development. W. W. Norton: New York, pp. Compayré, G. (1902) Development of the child in later infan- 15–27 cy. (Translated by Mary E. Wilson.) D. Appleton: New York. Piers, M. W. (Ed.). (1972) Play and development. W. W. Csikszentmihalyi, M. (1996) Creativity: Flow and the psychol- Norton: New York. ogy of discovery and invention. HarperCollins: New York. Preyer, W. (1893) Mental development in the child. (Translat- Darwin, C. (1859) On the origin of species. Murray: London. ed by H. W. Brown.) D. Appleton: New York. Darwin, C. (1871) The descent of man and selection in rela- Price, E. O. (1984) Behavioral aspects of domestication. Quar- tion to sex. Murray: London. terly Review of Biology 59: 1–32. Darwin, C. (1872) The expression of the emotions in man and Romanes, G. J. (1883) Mental life of animals. Kegan, Paul, animals. Murray: London. Trench, Trübner & Co: London. Dissanayake, E. (1992) Homo aestheticus: Where art comes Smith, P. K. (1988) Children’s play and its role in early devel- from and why. Free Press: New York. opment: A reevaluation of the ‘play ethos’. In: Pellegrini, A. Fagen, R. M. (1974) Selective and evolutionary aspects of ani- D. (ed) Psychological bases for early education (pp. 207– mal play. American Naturalist 108: 850–858. 226). Chichester: John Wiley and Sons. Fagen, R. M. (1981) Animal play behavior. Oxford University Smith, P. K. (1996) Play, ethology, and education: A personal Press: New York. account. In: Pellegrini, A. D. (ed) The future of play theory Fagen, R. M. (1984) Play and behavioral flexibility. In: Smith, (pp. 3–21). Albany: SUNY Press. P. K. (ed) Play in animals and humans.Basil Blackwell: Lon- Spencer, H. (1872) Principles of psychology. (second ed.). don, pp. 159–173. (Vol. 2). D. Appleton: New York. Ficken, M. S. (1977) Avian play. Auk 94: 573–582. Sutton-Smith, B./Kelly-Byrne, D. (1984) The idealization of Froebel, F. (1895) The mottos and commentaries of Friedrich play. In: Smith, P. K. (ed) Play in animals and humans (pp. Froebel’s Mother Play. (Translated by S. E. Blow and H. R. 305–321). Oxford: Basil Blackwell. Eliot.) D. Appleton: New York. Thompson, K. V. (1998) Self-assessment in juvenile play. In: Görlitz, D./Wohlwill, J. F. (eds) (1987) Curiosity, imagina- Bekoff, M./Byers, J. A. (eds) Animal play: Evolutionary, tion, and play. Lawrence Erlbaum Associates: Hillsdale NJ. comparative, and ecological perspectives (pp. 183–204). Griffiths, M. (1978) The biology of the monotremes. Academ- Cambridge University Press: Cambridge. ic Press: New York. Wilson, E. O. (1975) Sociobiology: The new synthesis. Cam- Groos, K. (1898) The play of animals. (Translated by E. L. Bald- bridge: Belknap Press. win.) D. Appleton: New York.

Evolution and Cognition ❘ 122 ❘ 1999, Vol. 5, No. 2 Colin Allen

Animal Concepts Revisited1

The Use of Self Monitoring as an Empirical Approach

PIGEON, TRAINED TO Abstract linguistic animals have A peck for a food concepts” and that “we reward in response to pic- Many psychologists and philosophers believe that the simply do not know how tures of trees, generalizes close correlation between human language and hu- to turn the claim that to new instances, man concepts makes the attribution of concepts to nonlinguistic animals responding appropriately nonhuman animals highly questionable. I argue for a have concepts into an to pictures that it has three-part approach to attributing concepts to ani- empirically substantive never seen before (HERRN- mals. The approach goes beyond the usual discrimi- question”. STEIN et al. 1976). Does it nation tests by seeking evidence for self-monitoring of In the face of such crit- have a concept of TREE? discrimination errors. Such evidence can be collected icism, should compara- A pig, trained to respond without relying on language and, I argue, the capacity tive psychologists and to a pair of objects based for error-detection can only be explained by attribut- cognitive ethologists on whether they are simi- ing a kind of internal representation that is reason- abandon the notion of lar in shape and color, ably identified as a concept. Thus I hope to have animal concepts? There also generalizes to object shown that worries about the empirical intractability would be costs to doing pairs that are new to it of concepts in languageless animals are misplaced. so. First, insofar as cogni- (KEDDY-HECTOR et al. tive psychologists have 1999). Does it have a con- Key words considered concepts cen- cept of SAMENESS? Many tral to their theories of comparative psycholo- Animal concepts, animal learning, error discrimina- human psychology, the tion, language. gists consider it reason- abandonment of animal able to attribute concepts concepts would make it to nonhuman animals on the basis of such results. harder to relate studies of animal cognition to at- But questions remain about the adequacy of such tempts to understand the evolution of human con- approaches to support conclusions about animal cepts. Second, there is a very tight link between no- concepts. tions of concept and intentional content. Concepts Many of the skeptical arguments that have been are constituents of the so-called intentional states put forward focus on the relationship of language to (beliefs, desires, etc.): for instance, the thought that human concepts. For example, CHATER/HEYES there is a squirrel in the tree is constituted by a struc- (1994) argue that because nonhuman animals lack tured arrangement of the concepts of squirrel, tree, language it is not possible to assign content to ani- and in-ness. Giving up on the notion of animal con- mal concepts in a rigorous way. Reviewing theories cepts would make it difficult (although perhaps not of concept drawn from philosophy, cognitive psy- impossible) to see the form of a theory of content chology, comparative psychology, and cognitive for the intentional states of animals. Conversely, a ethology, CHATER and HEYES argue that none of the workable theory of animal concepts would be a sig- proposed theories can simultaneously satisfy the nificant contribution to the theory of animal inten- three desiderata of (1) accounting for the close rela- tionality generally. tionship between language and concepts in hu- Previously, Marc HAUSER and I also disputed the mans, (2) being applicable to non-linguistic ani- adequacy of standard comparative psychological re- mals, and (3) being empirically tractable by sults to support claims about concepts (ALLEN/ behavioral methods. They conclude (p237) that “no HAUSER 1991), but did not embrace the skeptical clear sense has been provided for the claim that non- conclusion suggested by CHATER and HEYES. Rather,

Evolution and Cognition ❘ 123 ❘ 1999, Vol. 5, No. 2 Colin Allen we undertook to provide suggestions for the empir- Both social and individual conceptions of con- ical investigation of concepts in nonhuman ani- cept are important to scientific psychology. Social mals. Specifically, we maintained that the attribu- concepts play a role in explaining communication tion of concepts is justified “if there is evidence and cooperation among individuals, while individ- supporting the presence of a mental representation ual concepts are implicated in the structure of indi- that is independent of solely perceptual informa- vidual behavior and differences between the behav- tion” and by means of a pair of thought experiments ior of individuals. Both conceptions also have we indicated how we thought such evidence could intuitive counterparts in animal psychology. For in- be collected. Both of these thought experiments instance, assuming that it will be possible to make volved interpreting animals’ reactions to perceptual sense of animal concepts, one might consider what evidence that would seem to indicate conflicting concepts are implicated in the alarm calling systems conclusions. Both of our thought experiments re- of a particular species, such as the vervet. As in hu- volved around providing information to an individ- man social groups, the group concept involves a ual organism that a familiar conspecific was not centralizing tendency: vervets communicate and dead despite the appearance of death, and looking cooperate with each other in ways that tend to for changes in future reactions to subsequent pre- maintain a stable reference for specific alarm calls. sentation of the markers for death (or the absence Yet there is also room for individual difference: of markers for life). some vervets may be more likely than others to dis- The theoretical basis underlying the inference tinguish between different species of raptor for the from these experiments to claims about concepts production of their avian predator alarm calls, and was not carefully articulated in that 1991 paper. this may not be simply a function of differences in Instead we implicitly relied upon a comparative perception. approach, which justified the attribution of con- In scientific contexts (particularly cognitive psy- cepts to nonhuman animals by using human be- chology) where the focus is ostensibly on individual havior in similar circumstances as the benchmark. concepts it is quite common to find the use of idi- It was not, however, our intention to provide an oms that are more suited to the social conception. account of animal concepts that was essentially an- Philosophical arguments about animal cognition thropocentric for it was our belief that the basic are also plagued by failure to heed the distinction. strategy of using responses to conflicting informa- For example, several philosophers have been tion as a basis for concept attribution could stand tempted by the argument that animals do not have alone. The present paper attempts to give a more beliefs because they lack the constituent concepts: explicit statement of the framework underlying for example, a dog does not believe there is a squirrel this suggestion about empirical research into ani- in the tree because it lacks “the” (presumably the mal concepts. authors mean human) concept of squirrel. But there Before turning to the substance of the proposal, is no reason to think that having the belief requires it will be useful to make a distinction that is funda- that animals have that specific concept, nor that mental to any discussion of concepts, but that is lacking the canonical concept of squirrel means that often neglected nonetheless. The distinction is they lack any concept whatsoever (see ALLEN 1992 marked in the questions “What is the (or our) con- and ALLEN/BEKOFF 1997 for discussion). cept of tree?” versus “What is Fred’s concept of tree?” The notion under discussion in this paper is that which deal respectively with social and individual of the individual conception of concept. Of partic- notions of concept. Whenever the idiom “the con- ular interest is the relationship between concepts cept” is used, the definite article suggests that there and perception. Most, and perhaps all organisms are is one concept which many individuals share (like- capable of categorizing stimuli into perceptual wise for “our concept”). But it is also common to equivalence classes—a basic requirement for pro- hear it said that everyone has a different concept of ducing consistent responses to stimuli. But some or- X. From the perspective of this claim about individ- ganisms construct categorization schemes that, in a ual concepts, it can seem unlikely that there should sense to be explained below, transcend particular be such a thing as “the” concept of anything rather perceptual stimuli. Concepts are the mental repre- than just a collection of more or less related individ- sentations constituting the nodes in such categori- ual concepts of X. In commonsense psychology this zation schemes. tension between social and individual conceptions For example, consider the representation of of concept is mostly unanalyzed. death (ALLEN/HAUSER 1991). Many organisms have

Evolution and Cognition ❘ 124 ❘ 1999, Vol. 5, No. 2 Animal Concepts Revisited perceptual mechanisms whose function it is to al- attribute a concept to an animal is a distinct ques- low them to respond differentially to dead or alive tion from that of what it means for an animal to conspecifics. Ants, for example, respond to the pres- possess a concept, just as the question of when it is ence of acidic byproducts of decomposition, allow- reasonable to believe that someone is a murderer ing them to detect and remove dead conspecifics differs from the question of what it means to be a from the nest. Yet ants almost certainly have no murderer. Meeting conditions (i)–(iii) above may concept of death; that is they provide no evidence provide good grounds for attributing concepts to of a capacity to represent death independently of animals, even though these conditions need be nei- their perception of the chemical indicator. Thus, in ther necessary nor sufficient for concept possession, this respect, they do not transcend perceptual stim- just as finding a victim’s blood on a pair of socks uli. This is indicated by the fact that ants will remove may provide good grounds for believing the sock from the nest anything that has oleic acid painted owner to be a murderer even though blood on sock on it, including other live ants. Although the ants is neither a necessary nor sufficient condition for get other information that would tend to count being a murderer (ALLEN/BEKOFF 1997). To explain against the assessment that the acid-treated conspe- why satisfaction of the three conditions should be cific is dead (they are for instance capable of detect- considered good evidence for concept possession in ing its motion), they are incapable of using this in- animals it is necessary to say something about what formation to modify the removal response. While it it means to possess a concept, but much less than a is the biological function of their chemosensors to full analysis needs to be provided. detect dead conspecifics, this detection ability is Investigation of the capacities specified in clauses merely perceptual not conceptual. (i)–(iii) is empirically tractable in languageless crea- In contrast, humans can represent the status of tures. Satisfaction of clause (i) has been extensively an organism independently of particular perceptual studied in animals by the kinds of experiments re- stimuli. A body that appears not to be breathing ferred to in the first paragraph of this paper. For may nonetheless be judged to be alive, and one that instance, pigeons shown photographs of human appears to be breathing may be judged dead. (There faces interspersed with photographs of other items is an emphasis here on the word “appears”.) Con- can be taught to peck a response key only when a ceptual representation of death allows the organism face is presented. The animals generalize their cate- to transcend any particular stimuli. This in turn fa- gorization responses to new photographs contain- cilitates more flexible behavior. The capacity to rep- ing faces that were not seen during training. This resent death independently of specific perceptual shows that the discriminations are systematic, not indicators allows the organism to learn new rela- based on having learned piecemeal the particular tionships between perceptual indicators and the un- response for each individual member of the set of derlying condition of being dead. training stimuli. Likewise, KEDDY-HECTOR et al. We are now in a position to see how one might (1999) show that pigs learned to respond differen- overcome the objection that no empirical content tially to stimuli consisting of a pair objects, depend- can be given to the claim that a languageless crea- ing on whether the two objects differed in one of ture possesses concepts: one seeks evidence for the the properties of shape, size, or color, or whether non-perceptual representations just mentioned. they were identical in those respects, and that the There may be several ways to do this, but here is one pigs generalized the discrimination of sameness and suggested schema for the investigation of concepts: difference to novel pairs of familiar objects and to An organism O may reasonably be attributed a pairs of novel objects. Although the same/different concept of X (e.g., TREE) whenever: categorization is of an abstract relationship between (i) O systematically discriminates some Xs from perceptual properties, the discrimination still re- some non-Xs; and mains tied to perception in the sense that we have (ii) O is capable of detecting some of its own dis- no evidence that these animals can judge that two crimination errors between Xs and non-Xs; and things are really the same despite appearing differ- (iii) O is capable of learning to better discriminate ent, or vice versa. Thus these discoveries satisfy Xs from non-Xs as a consequence of its capacity (ii). clause (i) of the schema but not either of the other It is important to be clear that the purpose of clauses. these three clauses is not to provide a philosophical The capacities specified in clauses (ii) and (iii) analysis of what it is for an organism to possess a have not been extensively investigated, but in com- concept. The question of when it is reasonable to bination they provide a basis for empirical investi-

Evolution and Cognition ❘ 125 ❘ 1999, Vol. 5, No. 2 Colin Allen gation that allows us to resist the pessimistic conclu- for better performance on (i) then we have evidence sion drawn by CHATER and HEYES. By helping to that there is an integrated processing mechanism determine the boundaries of the animals’ represen- linking perceptual categorization and the recogni- tations, the joint investigation of capacities (ii) and tion of perceptual error, and hence a representation (iii) can help to settle questions about the content system that compares the perceptual content with of the representations. Clause (ii) concerns detec- an independent representation of what the perception of error. There are various ways that an animal tion is supposed to represent, i.e., a concept. might be informed by an external cue that it has The internal states implicated in the explanation miscategorized something, but in such cases there of these capacities are worthy of being designated as is a challenge to any empirical inference because of concepts. For these capacities to be implemented it the possibility that the cue indicating the error is appears that there must be an internal standard of sufficient alone to explain any subsequent behavior comparison that represents the organism’s world in- by an animal. (It is rather like the chemical cue being dependently of its perceptual representation at any sufficient alone to explain why ants remove dead given moment. Thus, such evidence supports the conspecifics—no cognitive representation of death claim that organisms with these capacities possess is required.) representations of the world that are detached from It is, however, possible for animals to show evi- immediate perceptual information. dence of error detection without external cuing, i.e., The difference between perceptual representa- they can show self-monitoring of their perfor- tions and concepts correspond to longstanding mance. Pigs again illustrate the point, although in philosophical ideas about concepts (WATSON 1995). this example it is knowledge of response error rather Descartes, for example, discusses the power of the than perceptual error that is demonstrated. Pigs mind to represent a chiliagon even though the vi- trained on the same/different task just described sual presentation of such an object would be indis- would still occasionally make errors (wrong choices) tinguishable from that of a 999-sided figure. Among while performing at an overall rate at or near 90% cognitive psychologists too, the major current the- correct. After committing to a response but before ories of concepts are all concerned with the way in any feedback was provided, some pigs would at- which concepts serve to unite varied perceptual pre- tempt to back away from the choice they had made. sentations of instances. Analysis of 22 cases of the backout behavior showed The close connection of language to concepts in that in only one of these cases occurred after the pig humans has seduced many into thinking that the made a correct choice (KEDDY-HECTOR et al 1999). It two notions of language and concept cannot be dis- is intriguing to try to account for this behavior, but entangled. This close connection may be explained regardless of what it shows in this particular exper- on the current schema in virtue of the fact that lan- imental set up, it indicates that under certain con- guages provide a structure that has a vast number of ditions some animals can provide non-linguistic ev- degrees of freedom with respect to immediate per- idence that supports the attribution of endogenous ception. Linguistic representation is, then, the basis error detection capacity. for the most fine-grained system of conceptual rep- Clause (iii) is the most difficult to articulate and resentation that we know. But it would be a mistake defend. Indeed, some participants at workshops to think that it is the only basis for conceptual rep- based on the ideas in this paper were satisfied that resentation available. It is well within the bounds of concepts are reasonably attributed on the basis of possibility that other species have ways of structur- the other clauses. Some were satisfied with (i) alone ing their experiences that transcend merely group- (and some prefer other criteria—those are fine for ing those experiences into equivalence classes for the point here is not to provide either necessary or the purposes of producing immediate behavioral re- jointly sufficient conditions for concept posses- sponses. Such a capacity contains the rudiments of sion). It is my view, however, a conceptual scheme. Thus, I that clause (iii), when satis- Author’s address argue, there is a clear sense for fied, provides a stronger case the claim in which language- for the attribution of con- less animals might possess Colin Allen, Department of Philosophy, cepts because it provides a Texas A&M University, College Station, TX concepts, and we do know link between the first two ca- 77843-4237, USA. how to turn this claim into an pacities. If the capacity to sat- Email: [email protected] empirically substantive ques- isfy (ii) is causally responsible tion.

Evolution and Cognition ❘ 126 ❘ 1999, Vol. 5, No. 2 Animal Concepts Revisited

Note

1 Reprinted with kind permission of Kluwer Academic Pub- lishers. The paper originally appeared 1999 in Erkenntnis 51: 33–40.

References Chater, N./Heyes, C. M. (1994) Animal concepts: content and discontent. Mind and Language 9: 209–246. Allen, C. (1992) Mental content. British Journal for the Phi- Herrnstein, R. J./Loveland, D. H./Cable, C. (1976) Natural losophy of Science 43: 537–553. concepts in pigeons’, Journal of Experimental Psychology. Allen, C./Bekoff, M. (1997) Species of Mind: the philosophy Animal Behavior Processes 2: 385–302. and biology of cognitive ethology. MIT Press: Cambridge Keddy-Hector, A./Allen, C./Friend, T. H. (1999) Cognition in MA. Domestic Pigs: Relational Concepts and Error Recogni- Allen, C./Hauser, M. D. (1991) Concept attribution in non- tion. Submitted. human animals: theoretical and methodological prob- Watson, R. A. (1995) Representational Ideas: from Plato to lems in ascribing complex mental processes. Philosophy Patricia Churchland. Kluwer: Dordrecht. of Science 58: 221–240.

Evolution and Cognition ❘ 127 ❘ 1999, Vol. 5, No. 2 Marc Bekoff

Social Cognition: Exchanging and Sharing Information on the Run1

Introduction Abstract Social play behavior: How individuals Individuals in many ani- In this essay I consider various aspects of the rapidly negotiate mal taxa need to be able growing field of cognitive ethology, concentrating cooperative to exchange social infor- mainly on evolutionary and comparative discussion agreements on the mation. While many of the notion of intentionality. I am not concerned run details of transmitting, with consciousness, per se, for a concentration on con- receiving, and processing sciousness deflects attention from other, and in many In recent years, the scien- information can be cases more interesting, problems in the study of ani- tific study of play behav- explained by appealing mal cognition. I consider how, when, where, and (at- ior has undergone many to hard-wired evolved tempt to discuss) why individuals from different taxa significant changes. mechanisms or to rules– exchange social information concerning their beliefs, Researchers in various dis- of–thumb acquired dur- desires, and goals. My main examples come from ciplines are conducting ing past exposures to var- studies of social play in mammals and antipredator detailed observational, ious stimulus–response behavior in birds. Basically, I argue that although not experimental, and theo- contingencies, in other all individuals always display behavior patterns that retical analyses of play pri- instances explanations are best explained by appeals to intentionality, it is marily in mammals that appeal to the philo- misleading to argue that such explanations have no (including humans) and sophical notion of inten- place in the study of animal cognition. birds. They are concerned tionality (beliefs and with such topics as evolu- desires) more easily and Key words tion, ecology, develop- parsimoniously explain ment, social communica- Social cognition, cognitive ethology, play, antipreda- observed patterns of tion, individual well- tor behavior, vigilance, canids, birds. behavior. While there being, neurobiology, has been a good deal of learning, and cognition concentration on the cognitive abilities of non- (for references see BEKOFF/BYERS 1981, 1998; FAGEN human primates (BYRNE/WHITEN 1988; BYRNE 1995; 1981; BURGHARDT 1998; BEKOFF/ALLEN 1992, 1998; RUSSON et al. 1996), there also is a wealth of data on ALLEN/BEKOFF 1994, 1997; PELLIGRINI 1995; SUTTON- nonprimates that is important for studies of non- SMITH 1998; BURGHARDT 1999). New data are forcing human animal (hereafter animal) cognition people to give up old ideas and set ways of thinking (BEKOFF/JAMIESON 1990, 1996; RISTAU 1991a; GRIFFIN about this phenomenon. Most researchers recognize 1992; ALLEN/BEKOFF 1997; BEKOFF/BYERS 1998). the importance of rigorous interdisciplinary collabo- In this essay I discuss two different types of be- ration in play research (PELLIGRINI, 1995; BEKOFF/ havior, mammalian social play and antipredator be- BYERS, 1998; BURGHARDT, 1998), and the interactions havior in birds, to make the case that the time has among those studying humans and non-humans are come to expand our taxonomic interests in cogni- producing exciting new results. tive ethological inquiries. These examples show Social play behavior is performed by individuals in how individuals exchange information during on- many mammalian, avian, and perhaps other verte- going social encounters—how analyses of social brate taxa (FAGEN 1981; BEKOFF/BYERS 1998; cognition on the run can inform comparative in- BURGHARDT 1998). It is a behavioral phenotype that quiries into what might be going on in animals’ lends itself to detailed empirical study (ALLEN/BEKOFF minds. 1997; BEKOFF/BYERS 1998; BURGHARDT 1999). Follow-

Evolution and Cognition ❘ 128 ❘ 1999, Vol. 5, No. 2 Social Cognition: Exchanging and Sharing Information on the Run

ing BEKOFF/BYERS (1981), I will define play as follows: is difficult to know whether arboreal clambering or “Play is all motor activity performed postnatally that the performance of various acrobatic movements appears (my emphasis) to be purposeless, in which during play may more be related to the evolution of motor patterns from other contexts may often be (mere) body awareness (e.g., knowing one’s place in used in modified forms and altered temporal sequenc- space) and not a concept of self. ing. If the activity is directed toward another living The study of social play also lends itself to Colin being it is called social play.” This definition centers ALLEN’s work on concepts (this volume). ALLEN ex- on the structure of play sequences—what animals do plores the idea that an organism O may reasonably when they play—and not on possible functions of be attributed a concept of X (e.g., play) whenever: (1) play. It concerns what play is, rather than what it is O systematically discriminates some instances of X not. The flexibility and versatility of social play make from some non-Xs; (2) O is capable of detecting it a good candidate for comparative and evolutionary some of its own discrimination errors between Xs cognitive studies including those that are concerned and non-Xs; and (3) O is capable of learning to better with ways in which individuals may negotiate agree- discriminate Xs from non-Xs as a consequence of its ments to engage in cooperative (not merely coordi- capacity (2). nated) social interactions. FAGEN (1993, p192) has With respect to ALLEN’s point (1), individual noted that “Levels of cooperation in play of juvenile canids, especially coyotes, seem to be able to discrim- primates may exceed those predicted by simple evo- inate some cases of play from non-play. They clearly lutionary arguments…” Play cannot occur if the in- respond differently to long and vigorous bites and teracting animals do not agree to cooperate. long stand-overs (an action that is performed signif- Cognitive ethological approaches are useful for icantly more by dominant individuals during aggres- gaining an understanding of social play for various sive interactions). Bites directed toward the body and of reasons including: (1) it exemplifies many of the stand-overs during play are of significantly shorter theoretical issues faced by cognitive ethologists; (2) duration during play than during aggressive encoun- empirical research on social play has and will benefit ters (HILL/BEKOFF 1977). Concerning ALLEN’s point from a cognitive approach, because play involves (2), individuals seem to have the capacity to recog- communication, intention, role-playing, and coop- nize that they have misclassified some encounters as eration; (3) detailed analyses of social play may pro- play or non-play. For example, when it seems likely vide more promising evidence of animal minds than that a playful encounter will escalate into aggression, research in many other areas, for it may yield clues individuals tend to perform a play bow (see below). about the ability of animals to understand one an- ALLEN (personal communication) pointed out that other’s intentions; and (4) play occurs in a wide range this might not be the case if play escalating into ag- of mammalian species and in a number of avian spe- gression is no longer perceived to be play by the in- cies, and thus it affords the opportunity for a com- teracting animals; there might be no misclassifcation parative investigation of cognitive abilities extend- in this case. With respect to point (3), individuals also ing beyond the narrow focus on primates that often appear to learn to discriminate play from non-play as dominates discussions of non-human cognition. For a consequence of recognizing their own previous example, during social play, many nonprimates (and misclassifications. This is especially so in cases where primates) engage in self-handicapping and role-re- they use play bows or other actions to indicate “this versals, two behavior patterns that are often used to is play” with individuals with whom they had previ- make inferences about intentionality (and con- ously made errors. That is, individuals seem more sciousness and self-consciousness). likely to assess more carefully the situation if they Along these lines, POVINELLI/CANT (1995, p400) have made previous category (play/non-play) errors suggest that the performance by arboreal ancestors of (BEKOFF 1975). the great ape/human clade of “unusual locomotor solutions … drove the evolution of self-conception”. The communication of the intention to Many nonprimate mammals also perform complex, flexible, and unusual acrobatic motor patterns (loco- play: The meaning of bows motor-rotational movements; WILSON/KLEIMAN When individuals play they typically use action 1974) during social play, and it would be premature patterns that are also used in other contexts, such to rule out the possibility that the performance of as predatory behavior, antipredatory behavior, and these behaviors is also important to the evolution of mating. These action patterns may not be intrinsi- self-conception in nonprimates. In some instances it cally different across different contexts, or they

Evolution and Cognition ❘ 129 ❘ 1999, Vol. 5, No. 2 Marc Bekoff may be hard to discriminate even for the partici- animals are performing them from a wide variety of pants. preceding postures, and (3) there is less of a need to How do animals know that they are playing? How communicate that “this is still play” than there is do they communicate their desires or intentions to when trying to initiate a new interaction. These ex- play? Individuals might know that they are playing planations are not exclusive alternatives. because the actions that are performed differ when Play bows occur almost exclusively in the context they are performed during play when compared to of play, and it is common to attribute to play-solic- other contexts (HILL/BEKOFF 1977), or the order in iting signals the message “what follows is play” or which motor patterns are performed differs from, “this is still play”. What is the significance of these and might be more variable than, the order in which messages for the players themselves? Are they in any they are performed during the performance of, for way aware of the meaning of the play bows, or are example, serious aggression or serious predation they simply conditioned to respond differently, for (BEKOFF/BYERS 1981). While there still are very few example, less aggressively or less sexually, when a empirical data that speak to either of these alterna- specific action such as a bite or a mount is preceded tives (BEKOFF/BYERS 1981, 1998), numerous people by a play bow? who study play come away with the impression that One way to approach these questions is to ask these possibilities would have strong and wide sup- whether play signals such as bows are used to main- port if they were studied rigorously. Suffice it to say, tain social play in situations where the performance playing animals seem to fine-tune on-going play se- of a specific behavior during a play bout could be mis- quences to maintain play and to prevent play from interpreted or when an individual makes an error. A escalating in real aggression. recent study of the structure of play sequences (BEKOFF Because there is a chance that various behavior 1995a) showed that bows in some canids (infant and patterns that are performed during on-going social adult domestic dogs, infant coyotes, and infant play can be misinterpreted, individuals need to tell wolves) often are used immediately before and imme- others “I want to play”, “this is still play no matter diately after an action that can be misinterpreted and what I am going to do to you”, or “this is still play disrupt ongoing social play. For example, in infant regardless of what I just did to you”. An agreement coyotes, wolves, and domestic dogs, the “bow” was to play, rather than to fight, mate, or engage in pred- not performed randomly, but rather immediately pre- atory activities, can be negotiated in various ways. ceded or immediately followed behavior patterns that Individuals may use various behavior patterns—play can be misinterpreted (for example, biting accompa- markers—to initiate play or to maintain the play nied by vigorous side–to–side shaking of the head). mood (BEKOFF 1977, 1995a; CHALMERS 1980; NEW- Recall that the social play of canids (and of other BERRY/WOOD-GUSH/HALL 1988; ALLEN/BEKOFF 1997) mammals) contains actions, primarily bites, that are by punctuating play sequences with these actions used in other contexts that do not contain bows (e.g., when it is likely that a particular behavior may have agonistic, predatory, or antipredatory). Actions such been, or will be, misinterpreted. There is little evi- as biting accompanied by rapid side–to–side shaking dence that play signals are used to deceive others. of the head are used in aggressive interactions and The “bow”, a highly ritualized and stereotyped also during predation and could be misinterpreted movement that seems to function to stimulate recip- when used in play. While there are few comparative ients to engage (or to continue to engage) in social data, in most species in which play has been de- play, provides an excellent example of a signal and scribed, play-soliciting signals appear to foster some it has been extensively studied in various canids in degree of cooperation between players so that each this context. Bows (the animal crouches on her fore- responds to the other in a way consistent with play limbs and elevates her hindlimbs) occur throughout and different from the responses the same actions play sequences, but most commonly at the begin- would elicit in other contexts. ning or towards the middle of playful encounters. In a detailed analysis of the form and duration of play Sharing intentions bows (BEKOFF 1977) it was shown that duration was more variable than form, and that play bows were How might a play bow serve to provide information always less variable when performed at the begin- to its recipient about the sender’s intentional state— ning, rather than in the middle of, ongoing play se- his intentions, desires, and beliefs? Perhaps one’s quences. Three possible explanations for this change own experiences with play can promote learning in variability include: (1) fatigue, (2) the fact that about the intentions of others. Perhaps the recipient

Evolution and Cognition ❘ 130 ❘ 1999, Vol. 5, No. 2 Social Cognition: Exchanging and Sharing Information on the Run shares the intentions (beliefs, desires) of the sender Self-handicapping occurs when an individual per- based on the recipient’s own prior experiences of sit- forms a behavior patterns that might compromise uations in which she performed bows. It may be rea- herself. For example, a coyote might not bite her sonable to attribute a very specific second-order play partner as hard as she can, or she might not play inference of the form “when I bow I want to play so as vigorously as she can. WATSON/CROFT (1996) when you bow you also want to play” without being found that red-neck wallabies adjusted their play to committed to a general capacity for the possession of the age of their partner. When a partner was second-order mental states in these animals. younger, the older animal adopted a defensive, flat- GOPNIK (1993) has suggested that others’ body footed posture, and pawing rather than sparring oc- movements are mapped onto one’s own kinesthetic curred. In addition, the older player was more toler- sensations based on prior experience of the observer. ant of its partners tactics and took the initiative in She has claimed (p275): “In particular, we innately prolonging interactions. (While more data are map the body movements of others onto our own needed, this study also suggests that the benefits of kinesthetic sensations. This initial bridge between the play may vary according to the age of the player.) inside and the outside, the self and other, underlies Role-reversing occurs when a dominant animal our later conviction that all mental states are things performs a action during play that would not nor- both we and others share.” TOMASELLO/GUST/FROST mally occur during real aggression. For example, a (1989) also have noted that some gestures in chim- dominant animal might voluntarily not roll-over on panzees may be learned by “second-person imita- his back during fighting, but would do so while play- tion”—”an individual copying a behavior directed to ing. it by another individual” (p35). They concluded (p45) In some instances self-handicapping and role-re- that chimpanzees “… rely on the sophisticated pow- versing might occur together. For example, a domi- ers of social cognition they employ in determining nant individual might roll over while playing with a what is perceived by a conspecific and how that con- subordinate animal or inhibit the intensity of a bite. specific is likely to react to various types of informa- In other situations, self-handicapping and role-re- tion…” FLANAGAN (1992, pp102ff) also has been inter- versing might not be performed together—a subor- ested in ways in which mental states can be shared, dinate animal restrains herself during play so as not and introduced the notion of a “mental detector” that to allow the interaction to escalate into a real fight. is used to detect others’ invisible mental states. There are very few data that are concerned with self- There is research that suggests a neurobiological handicapping and role-reversing in play. In canids, basis for the understanding of others’ mental states. infant coyotes, when compared to infant wolves and “Mirror neurons”, found in macaques, fire when a beagles, are more aggressive and engage in serious monkey executes an action and also when the mon- fights that determine social rank (BEKOFF 1974). key observes the same action performed by another Dominant coyote pups use rolling-over as self-hand- monkey (GALLESE 1998; GALLESE/GOLDMAN 1998; see icapping and role-reversing to get subordinate ani- also JEANNEROD 1994 for discussion of possible neural mals to play. In once case, 38/38 instances of rolling- bases for kinesthetic–visual matching). These find- over by a dominant coyote were used in the context ings are intriguing, and more comparative data are of social play. Furthermore, 34/42 (81%) approach/ needed to determine if mirror neurons are found in withdrawals were used in social play and resulted in other taxa and if they might actually play a role in the chasing by the subordinate coyote. In real aggressive sharing of intentions between individuals engaged in encounters, dominant animals do not allow others an on-going social interaction. to chase them. Both self-handicapping and role-reversing are Self-handicapping and also responsible for altering the temporal sequencing of play and might serve to call further attention role-reversing in play to the intentions of the animal who engages in these Two behavior patterns that have been reported to activities. That is, not only are the behavioral se- occur during on-going social play in various mam- quences performed in play different from those per- mals are self-handicapping and role-reversing (for formed during serious aggression, but play se- reviews see BEKOFF/ALLEN 1998; BIBEN 1998; PELLIS/ quences are punctuated with behavior patterns that PELLIS 1998; THOMPSON 1998; WATSON 1998). Both indicate that an individual is willing either to hand- might also be important for maintaining a play icap himself or to engage in role-reversing. From a mood. functional perspective, self-handicapping and role-

Evolution and Cognition ❘ 131 ❘ 1999, Vol. 5, No. 2 Marc Bekoff reversing, similar to using specific play invitation from her nest or young”) rather than as a hard-wired signals or altering behavioral sequences, might serve reflexive response to the presence of a particular to signal an individual’s intention to continue to stimulus, a potentially intruding predator. She stud- play. ied the direction in which birds moved during the The performance of self-handicapping and role- broken-wing display, how they monitored the loca- reversing suggests that an individual is able to make tion of the predator, and the flexibility of the an assessment of himself and others and manage or response. RISTAU found that birds usually performed regulate how she plays (PELLIS/PELLIS 1998, THOMP- the display in the direction that would lead an SON 1998), and detailed analyses of these activities intruder who was following them further away from will undoubtedly inform arguments about the cog- the threatened nest or young, and also that birds nitive capacities of animal who engage in them monitored the intruder’s approach and modified (ALLEN/BEKOFF 1997). Certainly, it appears that indi- their behavior in responses to variations in the viduals trust others to maintain the rules of the intruder’s movements. These and other data led RIS- game. BIBEN (1998) stresses the importance of devel- TAU to conclude that the plover’s broken-wing dis- opmental flexibility in the social behavior of rhesus play lent itself to an intentional explanation—that monkeys, and suggests that individuals use play as a plovers purposely lead intruders away from their way to learn strategies of social interaction that are nests or young and modified their behavior in order related to self-competence and self-preservation. to do so. To sum up briefly, I found that at least some In another field study of antipredator behavior in canids (and, I expect other mammals also would) birds, I (BEKOFF 1995b, 1996) found that western cooperate when they engage in social play, and that evening grosbeaks modified their vigilance or scan- individuals may negotiate these cooperative ven- ning behavior depending on the way in which indi- tures by sharing their intentions. In general, animals viduals were positioned with respect to one another. engaged in social play use specific signals to modu- Grosbeaks and other birds often trade-off scanning late the effects of behavior patterns that are typically for potential predators and feeding. Basically (but performed in other contexts, but whose meaning is oversimplified), some birds scan while others feed changed in the context of play. These signals are of- and some birds feed when others scan. Thus, it can ten flexibly related to the occurrence of events in a be hypothesized that individuals want to know what play sequence that might violate expectations others are doing and learn about others’ behavior by within that sequence. Furthermore, the relationship trying to watch them. of play to a cognitive appreciation of the distinction My study of grosbeaks showed that when a flock between reality and pretense provides an important contained four or more birds, there were large link to other cognitive abilities, such as the ability to changes in scanning and other patterns of behavior detect deception (which is extremely rare in social that seemed to be related to ways in which grosbeaks play) or to detect errors. Given these connections, attempted to gather information about other flock further detailed analyses of various aspects of social members. When birds were arranged in a circular play might help promote the development of more array so that they could see one another easily com- sophisticated theories of intentionality, representa- pared to when they were arranged in a line that made tion, communication, and (perhaps even) con- visual monitoring of flock members more difficult, sciousness. birds who had difficulty seeing one another were (1) more vigilant, (2) changed their head and body po- Antipredatory behavior in birds: sitions more often, (3) reacted to changes in group size more slowly, (4) showed less coordination in Cognition on the ﬂy head movements, and (5) showed more variability Antipredator behavior in birds also provides a good in all measures. example of the usefulness of comparative cognitive ethological approaches. For example, RISTAU Behavioral ﬂexibility and intentional (1991b) conducted field studies of injury-feigning in piping plovers (the broken-wing display) and explanations wanted to know if she could learn more about The differences in behavior between birds organized deceptive injury-feigning if she viewed the broken- in circular arrays when compared to birds organized wing display as an intentional or purposeful behav- in linear arrays were best explained by accounting ior (“the plover wants to lead the intruder away for individuals’ attempts to learn, via visual moni-

Evolution and Cognition ❘ 132 ❘ 1999, Vol. 5, No. 2 Social Cognition: Exchanging and Sharing Information on the Run toring, about what other flock members were doing. have included explicit conditioning about the fac- This may say something about if and how birds tors that influence when and how it should scan or attempt to represent their flock, or at least certain move its head and body. other individuals, to themselves. It may be that individuals form beliefs about what others are most Towards a broad and comparative likely doing and predicate their own behavior on these beliefs. I have argued that cognitive explana- cognitive ethology: The relationship tions were simpler and less cumbersome than non- between what’s happening “out there” cognitive rule–of–thumb explanations (e.g., “scan and “in here” this way if there are this number of birds in this geo- In his early work, Niko TINBERGEN (1951/1989; 1963) metric array” or “scan that way if there are that num- identified four overlapping areas with which etho- ber of birds in that geometric array.” Non-cognitive logical investigations should be concerned, namely, rule–of–thumb explanations did not seem to evolution (phylogeny), adaptation (function), cau- account for the flexibility in animals’ behavior as sation, and development (ontogeny), and his well or as simply as explanations that appealed to framework also is useful for those interested in ani- the intentions of the animals under study. mal cognition (JAMIESON/BEKOFF 1993; ALLEN/BEKOFF In many other instances it also may be more eco- 1997). BURGHARDT (1997) suggested adding a fifth nomical or parsimonious to assume that not every- area, private experience. He (p276) noted that “The thing that an individual needs to be able to do in all fifth aim is nothing less than a deliberate attempt to situations in which it finds itself is preprogrammed; understand the private experience, including the intentional explanations can be simpler than cum- perceptual world and mental states, of other organ- bersome stimulus–response explanations (DE WAAL isms. The term private experience is advanced as a 1991; ALLEN/BEKOFF 1997; BEKOFF/ALLEN 1997; BEKOFF preferred label that is most inclusive of the full 1998a). While general rules of thumb may be laid range of phenomena that have been identified down genetically during evolution, specific rules of without prejudging any particular theoretical or conduct that account for all possible contingencies methodological approach.” may be too numerous to be hard-wired. Further- BURGHARDT also noted that calling the fifth area more, while behavioristic learning schemes appeal- “cognitive ethology” is not radical enough because ing to notions such as conditioning, generalizing, there may be some historical baggage associated and substituting can account to a limited extent for with the use of this term. For one, it is frequently behavioral flexibility, behavioral integration, and associated with only the study of animal conscious- the use of internal states and images of absent ob- ness. However, cognitive ethology is much more jects in some organisms (e.g., HOLLAND 1990), learn- than the study of animal consciousness (ALLEN/ ing at high degrees of abstraction from sensory stim- BEKOFF 1997; BEKOFF/ALLEN 1997; BEKOFF 1998a) al- ulation seems less amenable to behavioristic analysis though many do not mark this distinction. (BEKOFF/ALLEN 1992). My studies of animal cognition have taught me a Cognitive models of learning provide explanatory number of important lessons. The first is that de- schemes for such cases. It might actually be more tailed observational and descriptive research are parsimonious to appeal to intentional explanations mandatory, especially under field conditions. We in terms of accounting for complex patterns of be- must know what animals do “out there” in order to havior with fewer explanations. For example, the re- learn more about what is going on “in here”—in sults of my study on grosbeaks, namely that flock their minds. Of course, experimental work is essen- geometry influences various patterns of behavior, tial as well, but before we understand the lives of the are more simply explained in terms of the represen- animals we study much experimental work is prema- tational needs or information-gathering goals of ture and can even be misleading. A second lesson is grosbeaks than by an attempt to account for the ef- that the sort of research that is needed in order to fects of group geometry in terms of numerous and study different aspects of animal cognition also is diverse stimulus–response contingencies or general- extremely tedious and time-consuming, but there izations from earlier experiences (e.g., the birds are are no substitutes for this work. For example, in a somehow conditioned [or innately predisposed] to current project on vigilance in western evening gros- produce certain behavioral patterns in response to beaks, I am finding that individuals exchange infor- group geometry). In this case it is difficult to con- mation very rapidly. Information on head (and eye) ceive that an individual grosbeak’s experience could and body position is lost if frames are not analyzed

Evolution and Cognition ❘ 133 ❘ 1999, Vol. 5, No. 2 Marc Bekoff

every 1/30th of a second (BEKOFF, unpublished data). claims about smartness or intelligence, we need to I hope that my colleagues who study animal cogni- be very careful about making such statements as tion as well as (especially?) those who are extremely “apes are smarter than monkeys or dogs” for each critical of the study of animal cognition and of the can do things the other cannot. Thus, MARLER (1996, notion of animal minds will engage in, and in some p22) concluded his review of social cognition in instances return to, this type of research as we head non-human primates and birds as follows: “I am into the 21st century. The third lesson concerns the driven to conclude, at least provisionally, that there use of old data. There are many existing data that can are more similarities than differences between birds inform cognitive ethological inquiries and research- and primates. Each taxon has significant advantages ers might consider revisiting them while planning that the other lacks.” Along these lines, TOMASELLO/ future studies. Sometimes a fresh look at old data can CALL (1997, pp399–400) summarized their compre- be extremely revealing. hensive review of primate cognition by noting that All in all, we are a long way from having an ade- “The experimental foundation for claims that apes quate data base from which hard-and-fast claims are ‘more intelligent’ than monkeys is not a solid about the taxonomic distribution of various cognitive one, and there are few if any naturalistic observa- skills, or about the having of a theory of mind, can be tions that would substantiate such broad-based, spe- put forth. BYRNE (1995), who otherwise takes a cies-general claims”. strongly primatocentric view of animal cognition, There are many examples other than social play agrees with the importance of conducting broadly and antipredator behavior that could have been cho- comparative research. Consider studies that show sen for illustration, but these two behavior patterns that some monkeys cannot perform imitation tasks make the case that chauvinism on either side of the that some mice can (WHITEN/HAM 1992). If the point debate concerning how to explain animal behavior is to answer the question are monkeys smarter than is unwarranted. Sometimes some non-humans (and mice or not it is misleading, for there is no reason to some humans) behave as stimulus–response ma- expect a single linear scale of intelligence (BEKOFF chines and at other times some non-humans (and 1998b,c). In the world of mice it may be more impor- some humans) behave in ways that are best extant to be able to do some things than it is in the world plained using a rich cognitive vocabulary. A plural- of monkeys, but in other respects a monkey may have istic approach will result in the best understanding cognitive capacities that a mouse would lack. There of the minds of other animals. also is much individual variation within species, and this also must be documented more fully. Acknowledgments To sum up, speciesist cognitivism must be re- sisted. We should strive to study individual animals I thank the organizers of this meeting for inviting to learn more about their lives in their own worlds me to Bremen in September 1997, and Colin ALLEN (and their abilities to feel pain and to suffer psycho- for all the discussions we have had and continue to logically and physically). Methodological pluralism have on animal minds and cognitive ethology. is required: species-fair meth- Marek SPINKA and I also have ods need to be tailored to the discussed numerous issues questions and the animals un- Author’s address dealing with play and I thank der consideration, and com- him for these fruitful peting hypotheses and expla- Marc Bekoff, Department of Environmental, exchanges. I also thank Anne nations need always to be Population, and Organismic Biology, Uni- BEKOFF for allowing me to use considered. In the absence of versity of Colorado, Boulder, Colorado her Peak Motus Motion Ana- suitable criteria and empirical 80309-0334 USA lyzer System in some of my data for making comparative Email: [email protected] work on avian vigilance.

Note

1 Reprinted with kind permission of Kluwer Academic Pub- lishers. The paper originally appeared 1999 in Erkenntnis 51: 113–128.

Evolution and Cognition ❘ 134 ❘ 1999, Vol. 5, No. 2 Social Cognition: Exchanging and Sharing Information on the Run

References Anthropomorphism, Anecdote, and Animals. SUNY Press: Albany, New York, pp. 254–276. Allen, C./Bekoff, M. (1994) Intentionality, social play, and Burghardt, G. M. (1998) Play. In: Greenberg, G./Haraway, M. definition. Biology and Philosophy 9: 63–74. (eds) Encyclopedia of Comparative Psychology. New York, Allen, C./Bekoff, M. (1997) Species of Mind: The Philosophy Garland, pp. 725–735. and Biology of Cognitive Ethology. MIT Press: Cambridge Burghardt, G. M. (1999) The Genesis of Animal Play. Rout- MA. ledge: New York. Bekoff, M. (1974) Social play and play-soliciting by infant Byrne, R. (1995) The Thinking Ape: Evolutionary Origins of canids. American Zoologist 14: 323–340. Intelligence. Oxford University Press: New York. Bekoff, M. (1975) The communication of play intention: Are Byrne, R./Whiten, A. (eds) (1988) Machiavellian intelli- play signals functional? Semiotica 15: 231–239. gence: social expertise and the evolution of intellect in Bekoff, M. (1977) Social communication in canids: Evidence monkeys, apes, and humans. Oxford University Press: New for the evolution of a stereotyped mammalian display. Sci- York. ence 197: 1097–1099. Chalmers, N. R. (1980) The ontogeny of play in feral olive ba- Bekoff, M. (1995a) Play signals as punctuation: The structure boons. Animal Behaviour 28: 570–585. of social play in canids. Behaviour 132: 419–429. Fagen, R. M. (1981) Animal Play Behavior. Oxford University Bekoff, M. (1995b) Vigilance, flock size, and flock geometry: Press: New York. Information gathering by western evening grosbeaks Fagen, R. (1993) Primate juveniles and primate play. In: (Aves, fringillidae). Ethology 99: 150–161. Pereira, M. E./Fairbanks, L. A. (eds) Juvenile Primates: Life Bekoff, M. (1996) Cognitive ethology, vigilance, information History, Development, and Behavior. Oxford University gathering, and representation: Who might know what and Press: New York, pp. 183–196. why? Behavioural Processes 35: 225–237. Flanagan, O. (1992) Consciousness Reconsidered. MIT Press: Bekoff, M. (1998a) Cognitive ethology: In: Bechtel, W./Gra- Cambridge MA. ham, G. (eds) Blackwell Companion to Cognitive Science. Gallese, V. (1998) Mirror neurons, from grasping to language. Blackwell Publishers: Oxford, pp. 371–379. Consciousness Bulletin Fall: 3–4. Bekoff, M. (1998b) Resisting speciesism and expanding the Gallese, V./Goldman, A. (1998) Mirror Neurons and the Sim- community of equals. BioScience 48: 638–641. ulation Theory of Mind-Reading. Trends in Cognitive Sci- Bekoff, M. (1998c) Deep ethology, animal rights, and the ence 2: 493–501. Great Ape/Animal Project: Resisting speciesism and ex- Gopnik, A. (1993) Psychopsychology. Consciousness and panding the community of equals. Journal of Agricultural Cognition 2: 264–280. and Environmental Ethics 10: 269–296. Griffin, D. R. (1992) Animal Minds. University of Chicago Bekoff, M./Allen, C. (1992) Intentional icons: towards an Press: Chicago. evolutionary cognitive ethology. Ethology 91: 1–16. Hill, H. L./Bekoff, M. (1977) The variability of some motor Bekoff, M./Allen C. (1997) Cognitive ethology: Slayers, skep- components of social play and agonistic behaviour in East- tics, and proponents. In: Mitchell, R. W./Thompson, N./ ern coyotes, Canis latrans var. Animal Behaviour 25: 907– Miles, L. (eds) Anthropomorphism, Anecdote, and Ani- 909. mals: The Emperor’s New Clothes? SUNY Press: Albany: Holland, P. C. (1990) Event representation in Pavlovian con- New York. Pages. 313–334. ditioning: Image and action. Cognition 37: 105–131. Bekoff, M./Allen, C. (1998) Intentional communication and Jamieson, D./Bekoff, M. (1993) On aims and methods of cog- social play: How and why animals negotiate and agree to nitive ethology. Philosophy of Science Association 2: 110– play. In: Bekoff, M./Byers, J. A. (eds) Animal Play: Evolu- 124. tionary, Comparative, and Ecological Perspectives. Cam- Jeannerod, M. (1994) The representing brain, neural corre- bridge University Press: Cambridge, New York, pp. 97–114. lates of motor intention andimagery. Behavioral and Brain Bekoff, M./Byers, J. A. (1981) A critical reanalysis of the on- Sciences 17: 187–245. togeny of mammaliansocial and locomotor play: An etho- Marler, P. (1996) Social cognition: Are primates smarter than logical hornet’s nest. In: Immelmann, K./Barlow, G. W./ birds? In: Nolan, V. Jr./Ketterson, E. D. (eds) Current Orni- Petrinovich, L./Main, M. (eds) Behavioral Development: thology Volume 13. Plenum Press: New York, pp. 1–32. The Bielefeld Interdisciplinary Project. Cambridge Univer- Newberry, R. C./Wood-Gush, D. G. M./Hall, J. W. (1988) Play- sity. Press: New York, pp. 296–337. ful behaviour ofpiglets. Behavioural Processes 17: 205– Bekoff, M./Byers, J. A. (eds) (1998) Animal Play: Evolution- 216. ary, Comparative, and Ecological Approaches. Cambridge Pelligrini, A. D. (ed) (1995) The Future of Play Theory: A Mul- University Press: New York, Cambridge. tidisciplinary Inquiry into the Contributions of Brian Sut- Bekoff, M./Jamieson, D. (eds) (1990) Interpretation and Ex- ton-Smith. SUNY Press: Albany: New York. planation in the Study of Animal Behavior. Volume I: In- Pellis, S. M./Pellis, V. C. (1998) The structure–function inter- terpretation, Intentionality, and Communication. 505 face in the analysisof play fighting. In: Bekoff, M./Byers, J. Pages; Volume II: Explanation, Evolution, and Adaptation. A. (eds) Animal Play: Evolutionary, Comparative, and Eco- 465 Pages. Westview Press: Boulder CO. logical Perspectives. Cambridge University Press: Cam- Bekoff, M./Jamieson, D. (eds) (1996b) Readings in animal bridge, New York, pp. 115–140. cognition. MIT Press: Cambridge MA. Povinelli, D. J./Cant, J. G. H. (1995) Arboreal clambering and Biben, M. (1998) Squirrel monkey playfighting: Making the the evolution of self-conception. Quarterly Review of Biol- case for a cognitive trainingfunction for play. In: Bekoff, ogy 70: 393–421. M./Byers, J. A. (eds) Animal play: Evolutionary, Compara- Ristau, C. (ed) (1991a) Cognitive Ethology: The Minds of tive, and Ecological Perspectives. Cambridge University Other Animals. Essays in Honor of Donald Griffin, R.. Press: Cambridge, New York, pp. 161–182. Lawrence Erlbaum: Hillsdale NJ. Burghardt, G. M. (1997) Amending Tinbergen: A fifth aim for Ristau, C. (1991b) Aspects of the cognitive ethology of an in- ethology. In: Mitchell, R. W./Thompson, N./Miles, L. (eds) jury-feigning bird, the piping plovers. In: Ristau, C. A. (ed)

Evolution and Cognition ❘ 135 ❘ 1999, Vol. 5, No. 2 Marc Bekoff

Cognitive Ethology: The Minds of Other Animals. Essays University Press: New York. in Honor of Donald Griffin, R. Lawrence Erlbaum: Hills- de Waal, F. B. (1991) Complementary methods and conver- dale NJ, Pages 91–126. gent evidence in the study of primate social cognition. Be- Russon, A. R./Bard, K. A./Parker, S. T. (1996) Reaching into haviour 118: 297–320. Thought: The Minds of the Great Apes. Cambridge Univer- Watson, D. M. (1998) Kangaroos at play: Play behaviour in sity Press: New York. the Macropodoidea. In: Bekoff, M./Byers, J. A. (eds) Animal Sutton-Smith, B. (1998) The Ambiguity of Play. Harvard Uni- Play: Evolutionary, Comparative, and Ecological Perspec- versity Press: Cambridge MA. tives. Cambridge University Press: Cambridge, New York, Tinbergen, N. (1951/1989) The Study of Instinct. Oxford pp. 61–95. University Press: New York. Watson, D. M./Croft, D. B. (1996) Age-related differences in Tinbergen, N. (1963) On aims and methods of ethology. playfighting strategies of captive male red-necked walla- Zeitschrift für Tierpsychologie 20: 410–433. bies (Macropus rufogriseus banksianus). Ethology 102: 33– Thompson, K. V. (1998) Self assessment in juvenile play. In: 346. Bekoff, M./Byers, J. A. (eds) Animal Play: Evolutionary, Whiten, A./Ham, R. (1992) On the nature and evolution of Comparative, and Ecological Perspectives. Cambridge Uni- imitation in the animal kingdom: reappraisal of a century versity Press: Cambridge, New York, pp. 183–204. of research. Advances in the Study of Behavior 21: 239– Tomasello, M./Gust, D./Frost, G. T. (1989) A longitudinal in- 283. vestigation of gesturalcommunication in young chimpan- Wilson, S. C./Kleiman, D. G. (1974) Eliciting play: A compar- zees. Primates 30: 35–50. ative study. American Zoologist 14: 341–370. Tomassello, M./Call, J. (1997) Primate Cognition. Oxford

Evolution and Cognition ❘ 136 ❘ 1999, Vol. 5, No. 2 Gerhard Roth

The Neurobiological Basis of Consciousness in Man and Animals

HE QUESTION OF Abstract consciousness. Finally, I T whether animals will investigate whether possess consciousness is In humans, consciousness includes diverse states, the different groups of as old and as difficult as from alertness to self-reflection. These states are pro- vertebrates possess brain the mind–body problem. duced by the interaction of many cortical and subcor- centers that are homolo- Many philosophers, psy- tical (predominantly limbic) brain centers. However, gous to those that in chologists and even some only processes taking place in the associative parietal, humans are correlated neurobiologists strictly temporal and frontal isocortex are accompanied by with states of conscious- deny the existence of con- consciousness. Consciousness and attention arise ness and ask how this sciousness in animals and whenever the human brain is confronted with tasks information can be used believe that this phenom- for which no pre-established “solutions” and respec- for answering the central enon is unique to tive neural networks exist. Conscious states are ac- question about the exist- humans (for very differ- companied by the formation of new neural networks ence of consciousness in ent reasons, though). based on changes in synaptic connectivity. Most cen- animals. Others say: Maybe ani- ters of the human brain involved in consciousness are present in the brains of all tetrapods (amphibians, mals have consciousness, 1. States of but we will never know reptiles, birds, mammals). Accordingly, it is likely for sure, because even in that most tetrapods possess at least simple states of consciousness humans consciousness is consciousness such as awareness of sensory events, Consciousness or aware- only accessible to the attention, knowledge representation and analogical ness varies widely with individual that has it. thinking. Higher states of consciousness such as tak- respect to intensity and Again others state: For a ing the perspective of the other and anticipation of fu- content ranging from number of reasons it is ture events are only found in primates and, therefore, deep coma to the highest very likely that at least may require a large isocortex. Finally, conscious degree of concentration, some groups of animals states such as comprehension of underlying mecha- from alertness to self- have some of those states nisms, knowledge attribution, self-awareness and the reflection. The most gen- use of simple syntactical language seem to be restrict- of consciousness found in eral form of conscious- ed to the great apes and may require a large prefrontal humans. Now, the ques- ness is wakefulness or vig- cortex. The use of a complex syntactical language tion is: What kind of hard ilance. It is characterized seems to be restricted to humans. However, there is ev- evidence does exist for by a general responsive- idence that non-human primates have isocortical the presence of con- ness to sensory stimuli. structures that are homologous to the human WER- sciousness in at least There are different states NICKE and part of the BROCA speech centers. The rela- some groups of animals? tionship between higher states of consciousness and of reduced consciousness In discussing the pieces of syntactical language remains unclear, but both may such as somnolence and evidence within the be due to a much longer period of synaptic plasticity stupor and different types domain of neurobiology, in humans compared to non-human primates. of coma. Vigilance is usu- I will proceed as follows: ally combined with sub- First, I will give a brief Key words jective awareness or con- phenomenology of states scious experience of of consciousness. Then I Animal consciousness, consciousness brain centers, something. This “some- will discuss what is function of consciousness. thing” includes external known about the neuro- as well as internal bodily biological basis of consciousness in man and stimuli, my own emotions and my mental activity. present two hypotheses about the functional role of From this awareness results the experience of my

Evolution and Cognition ❘ 137 ❘ 1999, Vol. 5, No. 2 Gerhard Roth own presence in the world. Attention is the most below). The function of the medial column is the characteristic state of increased awareness. A more control of wake–sleep cycle and of general cortical special type of consciousness is body-identity aware- activity. The median column is formed by the so- ness, i.e. the belief that I belong to the body that called raphe nuclei. These nuclei, predominantly apparently surrounds me. There is autobiographic the dorsal raphe nucleus, send serotonergic fibers to consciousness, i.e. the conviction that I am the one all parts of the limbic system that are involved in who existed yesterday. There is awareness of what cognitive functions, e.g., hippocampus, amygdala, was going on in the past and is happening in the basal forebrain, limbic thalamic nuclei, cingulate world surrounding me. There is awareness of volun- and entorhinal cortex, frontal, parietal and occipital tary control of movements and actions, of being the isocortex. The raphe nuclei are supposed to play a author of my thoughts and deeds. Finally, there is self- modulatory role in the context of behaviorally rel- awareness, i.e. the ability of self-recognition and self- evant events, apparently by counteracting and reflection. dampening the arousing effect of the other systems These different aspects of consciousness can dis- and, thus, increasing the signal–to–noise ratio in sociate, i.e., they can be lost independent of each the cortex (ROBBINS 1997). The lateral column con- other after damage of different parts of the brain tains the noradrenergic locus coeruleus complex, (KINSBOURNE 1995; KNIGHT/GRABOWECKY 1995; MOS- which again projects to all parts of the limbic–cog- COVITCH 1995). Accordingly, there are patients who nitive system. The locus coeruleus is supposed to have all normal states of cognition, consciousness exert a role in controlling attention and continu- and intelligence except that they deny belonging to ously “monitoring” the external and internal envi- their own body, and others who do not know who ronment for important events. Its projection to the or where they are. prefrontal cortex in particular may mediate information about the relevance of complex sensory 2. The neurobiological basis of the events and situations (ROBBINS 1997). The next most important brain centers for the different states and appearances of control of consciousness are the limbic intralaminar consciousness and midline thalamic nuclei, because they are the It appears that we are only aware of those things most important relay station for the ascending pro- that are bound to the activity of the so-called asso- jections of the reticular formation. They receive in- ciative (i. e., parietal, temporal and frontal) isocor- put from the entire isocortex and project, with some tex. However, this is not to say that the associative topography, back to it, predominantly to the pre- cortex is the only “producer” of conscious experi- frontal cortex, and additionally to the striatum; ence; its activity is a necessary, but not sufficient they are also connected to the entire limbic system. prerequisite for consciousness. Many subcortical The nucleus reticularis thalami surrounds the en- parts of the brain, the activity of which is never tire lateral part of the thalamus in a bowl-like fash- accompanied by consciousness, contribute in very ion. It receives collaterals from thalamo–cortical as different ways to the origin of consciousness in the well as cortico–thalamic tracts and has reciprocal associative cortex. This is true, above all, for the connections with the sensory and limbic thalamic reticular formation, because their destruction leads nuclei. It does not project to the cortex, rather, it to a general loss of consciousness, i.e. to coma (HAS- exerts inhibitory control over most thalamic nuclei SLER 1978). via GABA-ergic fibers. It is assumed to function as a The reticular formation consists of three columns “filter” for various kinds of information coming of nuclei, a median, medial and lateral column that from the sensory periphery and brainstem, acting extend from the anterior mesencephalon through under the control of the isocortex and the limbic the pons to the medulla oblongata and rostral spinal system. It may, therefore, be involved in guidance cord (NIEUWENHUYS et al. 1991). The medial column of attention (GUILLERY et al. 1998). receives input from all sensory modalities and the The brain centers mentioned so far belong to the cerebellum as well as descending cortical input via limbic system in a wider sense, which in the verte- the pyramidal tract. This system called the “extrale- brate brain is the system that subconsciously con- mniscal” or ascending reticular activating system, trols all aspects of cognitive and emotional states projects directly or indirectly to the intralaminar and, accordingly, “voluntary” behavior. The limbic thalamic nuclei, which, in turn, project (with some system evaluates what the organism does and stores topography) to the striatum and the isocortex (see the result of this evaluation in the different kinds of

Evolution and Cognition ❘ 138 ❘ 1999, Vol. 5, No. 2 The Neurobiological Basis of Consciousness in Man and Animals memory. Other parts of the limbic centers discussed cordingly, damage to these subcortical centers usu- in the following contribute in more specific ways to ally produces either complete loss of consciousness the different states of consciousness. or profound impairment of conscious cognitive and The basal forebrain–septal nuclei complex is recip- emotional functions. This may include the inability rocally connected with the hippocampus and the to recognize positive or negative consequences of amygdala as well as with the centers of the reticular action (for example as a consequence of lesions in formation already mentioned. Its cholinergic fibers the amygdala or nucleus accumbens), impairment project to all parts of the isocortex. The basal fore- of attention and loss of declarative memory. Impor- brain is involved in the control of attention and of tantly, patients are usually unaware of these deficits. activity of neocortical neuronal network, primarily Although activity in the isocortex is necessary for in the context of earlier experience (VOYTKO 1996). consciousness, we are likewise unaware of processes The amygdala is a complex of many different nu- in the primary and secondary sensory and motor clei and is reciprocally connected with the associa- areas of the isocortex, although these processes are tive isocortex, particularly with the orbitofrontal necessary for the specific contents of awareness of prefrontal cortex (either directly or via the me- events inside or outside our body. We are only aware diodorsal thalamic nucleus) and the hippocampal of processes bound to the activity of the cingulate formation (AGGLETON 1992). It strongly influences and the associative cortex, and only of some of those the sensory (visual, auditory, gustatory) isocortex. processes. Subcortically, it receives input from the olfactory The cingulate cortex is that part of the cortex that system, the medial, intralaminar and midline tha- surrounds the subcortical parts of the telencephalon lamic nuclei and from the rest of the limbic system. and the thalamus. It is tightly connected with the Regarding cognitive and emotional functions, it is prefrontal and parahippocampal cortex, the basal believed to be an important center (together with forebrain–septal region, the amygdala, the in- the anterior cingulate cortex, see below) for evalua- tralaminar and midline thalamic nuclei and the re- tion and perhaps storage of negative experience, e.g., ticular formation. The anterior part (BRODMANN area in the context of fear conditioning and anxiety 24) is involved in the sensation of pain (in combi- (AGGLETON 1992, 1993). The hippocampal formation nation with the somatosensory cortex, the medial (AMMON’s horn, subiculum, dentate gyrus) and the thalamic nuclei and the so-called central tegmental surrounding parahippocampal and perirhinal (in- gray), and in memory of painful events. PET and cluding entorhinal) cortex are important centers for functional MRI studies demonstrate that it is always the formation and the consolidation of traces of de- active in tasks requiring attention (POSNER 1994). clarative memory inside the isocortex, i.e., of those The posterior parietal cortex (PP) shows strong kinds of memory that in principle can be con- hemispheric asymmetry. The left PP is involved in sciously retrieved and reported (SQUIRE 1987; symbolic–analytic information processing such as MARKOWITSCH 1992, 1999; SQUIRE/KNOWLTON 1995). mathematics, language, meaning of drawings and The dorsal parts of the basal ganglia, i.e. putamen, symbols. Lesions impair reading and writing and nucleus caudatus, globus pallidus, are closely asso- respective memory functions. The right PP involves ciated with the prefrontal, premotor and parietal real and mental spatial orientation, the control of isocortex as well as with the entire limbic system. hand and eye movement, change of perspective and The basal ganglia have to do with subconscious control of attention. Lesions of inferior right PP pro- planning and final decision in voluntary action un- duce neglect, e.g. ignoring the contralateral half of der the influence of the limbic system. The so-called the body or events in the contralateral visual hemi- mesolimbic system (i.e., nucleus accumbens, lateral field, or anosognosia, i.e. lack of insight or denial of hypothalamus, ventral tegmental area) is character- disturbances. ized by the neuromodulator dopamine. This system The superior and middle temporal cortex includes has strong connections with the orbitofrontal cor- perception of complex auditory stimuli including tex and is involved in the formation of positive (generally on the left side) WERNICKE’s semantic memories and pleasure and perhaps in the control speech center, which is crucial for the understand- of attention in the context of new events (ROBBINS/ ing and the production of meaningful written and EVERITT 1995, 1996). spoken language. Perception of music usually in- All these parts of the brain outside the isocortex volves the right medial temporal cortex. The inferior substantially contribute to consciousness, while temporal cortex (IT) is decisive for complex visual their activities remain completely unconscious. Ac- information regarding non-spatial properties of vi-

Evolution and Cognition ❘ 139 ❘ 1999, Vol. 5, No. 2 Gerhard Roth sual objects and scenes including their meaning and nantly oriented toward “internal” emotional and correct interpretation. Lesions in IT produce object social aspects of life. agnosia (left IT), color agnosia (right IT), prosopag- In conclusion, there is neither consciousness per se nosia, i.e. inability to recognize faces (right or bilat- nor a highest brain center producing consciousness. eral), deficits in categorization, changes in person- Different parts of the associative isocortex contrib- ality and emotionality and deficits in the use of ute in different ways to the high diversity and con- contextual information. tent of consciousness including awareness of exter- The prefrontal cortex (PFC) includes the largest nal and internal sensory events, consequences of portion of the isocortex (about 30% in man). Usu- one’s own behavior, autobiographic, body and ally, two major parts are distinguished in the pri- “ego” identity, action planning and authorship of mate, including human, brain: a dorsolateral and a own actions. It does this under the strong influence ventral–orbitofrontal part. The dorsolateral PFC of the primary and secondary sensory and motor (Brodmann areas 9, 10, 46, and 8) receives its major cortices as well as of the subcortical centers men- isocortical input from posterior parietal cortex, so- tioned. matosensory cortex related to trunk and limbs (superior parietal cortex), associative visual areas (V1, 3. The functional role of consciousness 3 and 4) and the anterior cingulate cortex. Subcor- tical input comes from the middle (small-celled) In the following, I will briefly present two hypothe- part of the mediodorsal thalamic nucleus, the ven- ses—a lower and a higher order one—about the tral anterior and parafascicular thalamic nuclei, functional role of awareness of sensory events and which mediate information from the basal ganglia particularly attention as the most prominent states (ALEXANDER et al. 1990; ROBERTS et al. 1998). The of consciousness,. The lower order hypothesis is the dorsolateral PFC appears to be involved in: (1) atten- following: States of conscious awareness are bound tion and selective control of sensory experience, (2) to a high degree of neural activity in associative action planning and decision making, (3) temporal areas which in turn is bound to three factors: (1) coding of events, (4) judgement and insight, partic- sufficient sensory stimulation through specific tha- ularly with respect to reality, (5) spontaneity of be- lamic nuclei and primary and secondary cortical havior, (6) strategic thinking, (7) associative think- areas; (2) adequate stimulation by the reticular for- ing, and (8) short-term or working memory. Thus, mation, reticular and limbic thalamic nuclei, the the dorsolateral PFC is predominantly, though not limbic system, basal ganglia etc. (3) high local met- exclusively, oriented toward the external world and abolic activity based on sufficient supply of oxygen its demands including short-term memory. and glucose as a consequence of high local cerebral The orbitofrontal PFC (Brodmann areas 12–14, 10, blood flow. These facts underlie the known imag- 11, 46, and 8) receives its main intracortical input ing techniques such as functional MRI and PET from the anterior cingulate cortex, the inferotempo- (POSNER 1994; MAGISTRETTI et al. 1999). ral cortex, the inferior somatosensory cortical areas Why is increased local metabolic and neural ac- (head, neck, face), the anterior pole of the temporal tivity of associative isocortical networks a necessary lobe and the parahippocampal and perirhinal/en- prerequisite for specific states of consciousness? torhinal cortex. Subcortical input is mediated These states occur only when our brain is con- mainly by the medial (large-celled) part and por- fronted with something that is sufficiently impor- tions of the lateral (small-celled) part of the me- tant for the brain to deal with and for which the diodorsal thalamic nucleus (ALEXANDER et al. 1990), brain does not have already existing neuronal net- which mediate information coming from the entire works and accordingly pre-established “solutions” limbic system regarding emotion, motivation, se- (e. g., in the realm of understanding of sentences, lective attention. The orbitofrontal PFC is involved motor skill or the recognition of faces). When such in social behavior, ethical considerations, divergent pre-set networks exist, we master these demands thinking, risk assessment, awareness of conse- without paying attention or even without any quences of behavior, emotional life and emotional awareness at all. Rather, attention and directed con- control of behavior. Accordingly, damage to the or- sciousness often will disturb the performance of au- bitofrontal PFC results in loss of interest in impor- tomatized motor skills. Accordingly, localized met- tant life events, loss of “ego”, “immoral” behavior abolic and neural cortical activity is low during and disregard of negative consequences of own be- these routine behaviors, as can be visualized havior. Thus, the orbitofrontal PFC is predomi- through imaging techniques (POSNER 1994).

Evolution and Cognition ❘ 140 ❘ 1999, Vol. 5, No. 2 The Neurobiological Basis of Consciousness in Man and Animals

However, when our brain is confronted with cles work, and it would be absolutely impossible for something new and important, e.g., an unexpected us to consciously follow sensory processing or to or unfamiliar perception, the meaning of a newly consciously activate or inhibit all muscles involved heard sentence, handling an unfamiliar tool or prac- even in a simple movement. The phenomenal world ticing a new piece of piano music, we can cope with works on the basis of a user-friendly code as opposed these tasks only with attention. As a consequence, to the machine code of the brain. The creation of a certain brain regions are activated above normal lev- phenomenal world, of “actuality”, seems to be els, and this activation is accompanied by an in- highly favorable or even necessary for action plan- crease in blood flow and cellular metabolism at the ning, tactic and strategic thinking and speaking via expense of other brain regions. In this state, the a complex syntactical language, which would be im- brain apparently modifies the pattern of synaptic possible in terms of the actual neuronal code. coupling among neurons in those centers that are This higher-order hypothesis includes the as- committed to these functions. For example, when sumption that during the evolution of great apes an we look at a new face that is important for us, then increase in demands of action planning, imagina- synaptic coupling in specialized networks in the tion, strategic thinking and complex syntactical caudal portion of the lower temporal lobe are mod- language took place, and this necessitated the con- ified. This also happens in the BROCA and WERNICKE struction of a phenomenal world, in which a virtual speech areas, when we try to understand the mean- actor, the ego, plans, acts and communicates with- ing of a sentence. out caring about how this is actually done by the In contrast to “normal” neuronal activity, brain in terms of neuronal activity. Apparently, the changes in the efficiency of synaptic coupling re- invention of a conscious, phenomenal world was a quire increased metabolic energy (sugar and oxy- necessary prerequisite for survival in a complex bi- gen), probably due to an increase in intracellular ological and social world. This hypothesis is largely signal processing, an increase in receptor density consistent with the view of consciousness as some- etc. Such synaptic changes are required to form new thing like a “self model” or “monitoring system”, as network properties that deal with the task at hand. has been developed by various authors (cf. METZ- If these processes are inhibited by the administra- INGER 1993; BAARS 1997). The model presented here tion of certain drugs, then awareness, attention and in particular stresses the need for reduction in com- the ability to master new cognitive and motor tasks plexity at planning and execution of action. are likewise hindered. This hypothesis addresses the necessary neuronal 4. Cognition and consciousness in preconditions and correlates of consciousness; it does, of course, not answer the question, why we are animals conscious at all when we are confronted with new Checking recent reviews on “animal mind” (STAMP and important cognitive and motor tasks. A possible DAWKINS 1993; PARKER et al. 1994; BYRNE 1995; answer to this question comes up, when we care- PEARCE 1997; for criticism see MACPHAIL 1998), it fully study the properties of our phenomenal, sub- seems that all vertebrates and probably inverte- jectively experienced world. In this phenomenal brates with large brains (e.g., cephalopods) display world, we are completely unaware of the over- sensory or focused attention, extended memory, whelmingly complex processing of information categorization and the formation of cognitive from sense organs to the associative cortex. Rather, maps. Whether this is accompanied by some kind the sensory world seems to directly enter our mind. of consciousness is difficult to determine in fishes, We may recall the finding of Benjamin LIBET from amphibians or reptiles. Many of so-called higher the late 70s that our consciousness simply ignores cognitive functions such as concept learning, the 300–500 ms that are needed for sensory infor- knowledge representation, analogical thinking, the mation from sense organs to cortical consciousness formation of abstract representations, and imita- and acts as if we would directly sense the sensory tion in the sense of copying a behavior are found at world (LIBET 1978). Similarly, our will seems to di- least among birds and mammals. Even in humans rectly move our muscles, and we are completely un- these higher cognitive functions are not necessarily aware of the hyper-complex actions of the cortical accompanied by consciousness and subcortical motor systems. Things in this phe- There are, however, cognitive functions that in nomenal world indeed happen radically differently humans require consciousness and for which it is from how the brain, the sense organs and the mus- hard to believe that animals but not humans could

Evolution and Cognition ❘ 141 ❘ 1999, Vol. 5, No. 2 Gerhard Roth carry them out unconsciously. These functions in- more important are the experiments concerning clude: (1) Imitation in the sense of task structure or mirror self-recognition (PARKER et al. 1994). Mon- task principle learning and tool learning. This is keys are capable of making use of mirrors, e.g. in found in macaque and capuchin monkeys, apes and order to look behind objects that otherwise are in- maybe in some other mammals (otter etc.). (2) Tak- accessible (BYRNE 1995). Kummer reports that ba- ing the perspective of other individuals in deception boons recognize a group member on a slide and and counterdeception. This is found in monkeys identify themselves as mother or child on slides (e.g., baboons) and great apes. (3) Anticipation of without difficulty. Yet, primatologists agree that future events (preparation of tools in advance etc.). monkeys show no sign of mirror self-recognition. This has been found in the great apes and may be Among apes, chimpanzees and orangutans show present in some monkeys. (4) Comprehension of mirror self-recognition, while among gorillas only underlying mechanisms, for example in the use of Koko does (it is assumed that gorillas often are too tools: this has been reported only in great apes. (5) shy). Chimpanzees show great interest in their mir- Knowledge attribution/theory of mind: This is ror image and pass the marking test (i. e., removing found in great apes, particularly chimpanzees. (6) marks of paint from face or body using the mirror Self-recognition in mirror: This is found in great image) well. Interestingly, at least some dolphins are apes and dolphins. (7) Distinction between appear- reported to show mirror self-recognition, too, but ance and reality: This is present only in chimps and their behavior appears to be very different in front dolphins. (8) Playing with imaginated objects: This of a mirror. They show no natural interest in their is found only in chimpanzees. (9) Teaching: This mirror image and dislike the marking test (MARTEN/ again is found only in chimpanzees. (10) Under- PSAKAROS 1994). The reasons for these differences are standing and using simple syntactical language (up unclear, besides many methodological and concep- to three-word sentences): This is found in great apes tual problems. and dolphins. (11) Using complex syntactical lan- The presence of knowledge attribution or theory guage: This is found only in humans. of mind to others (BARON-COHEN et al. 1985) is dif- Thus, a few of the mentioned “higher cognitive ficult to distinguish from the capability of taking functions” may be found in birds and all mammals, the perspective of the other. For most animals with but most of them are found either only in primates complex behavior, it is important to guess what the and dolphins, only in the great apes and dolphins, other is going to do, e.g. in agonistic behavior only in great apes or even only in chimpanzees and (STAMP DAWKINS 1993). However, primatologists humans, and at least one function (viz., the use of (e.g., KUMMER) definitely state that monkeys (e.g., complex syntactical language) only in man. Being baboons) do not take into account what the other is aware of the risk of an anthropomorphic perspec- “thinking”. Important in this context is the perfor- tive, we might speculate about “jumps” in cognition mance of very young children. According to PARKER and states of consciousness among tetrapods from et al. (1994), in children self-detection starts with 3 amphibians and reptiles to birds and mammals, months, and mirror self-recognition is shown at an among mammals to cetaceans and primates and age of 18 months on average. However, little chil- among primates from monkeys to the great apes and dren show embarrassment and coy reactions earlier from non-human apes to Homo sapiens. than that. Self-recognition in photos starts at 24 Functions (4), (5), (6) and (10) have been exten- months, followed by signs of self-evaluative emo- sively studied in the recent past and are of special tions, e.g., shame and pride. A true “theory of mind” interest, because they might draw a borderline be- is said to emerge at 4 years (MELTZOFF/GOPNIK 1993). tween monkeys and great apes among the primates. The authors stress that these events occur in the Such a borderline is less emotionally laden than that same sequence in non-human and human apes, between humans and non-human apes. Primatolo- though at a much slower rate in the former. gists almost unanimously agree that monkeys (e.g., Much has been written about the presence or ab- capuchins) exhibit the use of tools (as many other sence of a syntactically complex, language in non- animals do), but without an understanding of the human primates (and other animals). Monkeys underlying mechanism. They do not seem to have complex systems for intraspecific vocal com- “know” why one tool is effective and another is not munication that are able to express relatively com- (VISALBERGHI/LIMONGELLI 1994). A similar lack of plicated types of meaning including symbolic infor- “insight” has been reported by KUMMER on baboons mation (i.e., about objects and events that are not (using a stick for getting into reach of apples). Even present) or about relationships between events (e.g.,

Evolution and Cognition ❘ 142 ❘ 1999, Vol. 5, No. 2 The Neurobiological Basis of Consciousness in Man and Animals parental relationships). Many of these calls have to Are the above differences due to differences in (1) be learned by the infant monkey and—as in song- overall organization of brains, (2) absolute or rela- birds—dialects exist (cf. ZIMMERMANN et al. 1995). tive brain size, (3) absolute or relative size of parts of Most authors agree that sentences consisting of up the brain, e.g., cortex or subcortical centers, (4) size to three words are understood and used by chim- of the association cortex and prefrontal association panzees, gorillas and dolphins. Whether this is a cortex in particular, or (5) anatomical or physiolog- sign for a “simple” syntax is a matter of debate. ical features of the cortex or subcortical centers? Chimpanzees do combine words to form new words (1) All tetrapods have brains that—despite enor- (GARDNER et al. 1989), but they do not go beyond mous differences in outer appearance, overall size the linguistic capabilities of a 3-year-old child, even and relative size of major parts of the brain—are very after intense training (SAVAGE-RUMBAUGH 1984). similar in their overall organization and even in Of great interest in this context is the develop- many details (ROTH/WULLIMANN 1996). All tetrapod ment of language in human children (LOCKE 1995; brains possess a median, medial and lateral reticular STROMSWOLD 1995). In humans, language learning formation inside the medulla oblongata and the starts long before speaking, viz., in utero. Facial ventral mesencephalon. In the dorsal diencepha- learning starts immediately after birth; babies seem lon, an epithalamus, a dorsal and ventral thalamus to be pre-adapted to it. At 6 months after birth, chil- and a subthalamus are found, and it seems that they dren show a preference for familiar over foreign lan- also share the distinction between “specific” (i.e., guage. First voweling sounds are produced around those involved in sensory and motor functions) and 4–6 months, when “babbling” starts, first conso- “non-specific”, i.e., limbic nuclei. There is a corpus nant-like sounds occur at 9–12 months. First words striatum, globus pallidus, nucleus accumbens, basal are produced between 8–20 months, (average 12 forebrain/septum and amygdala within the ventral months). Around 18–23 months, children begin to telencephalon, a lateral pallium, homologous to the combine words and form 2-word utterances. Then, olfactory cortex of mammals, and a medial pallium, they gradually begin to use sentences longer than homologous to the hippocampal formation (at least two words, but typically in a “telegraphic” style sim- AMMON’s horn and subiculum). Thus, all structures ilar to that found in patients with lesions of BROCA’s required for attention, declarative memory (or its speech area. In the third year, children start distin- equivalents in animals), motivation, guidance of guishing singular and plural and asking the famous voluntary actions and evaluation of actions are question “what’s that?”. From there on, their lan- present in the tetrapod brain. These structures have guage syntactically becomes rapidly more elaborate. essentially the same connectivity and distribution Between 3 and 4 years, utterances are completely of transmitters, neuromodulators and neuropep- grammatical. tides in the different groups of tetrapods. Thus, it seems that apes have simple linguistic A more difficult problem is the presence of a ho- capabilities comparable to a human child of 2–3 mologon to the mammalian isocortex and associa- years, but from then on the child rapidly develops tive cortex in the telencephalon of other tetrapods. a fully syntactical language that is far superior to any Amphibians possess a dorsal pallium, reptiles have communicative system known from animals. a dorsal cortex plus a dorsal ventricular ridge (DVR), birds have a wulst and a DVR, and these structures 5. Animal brains and human brain are believed by many comparative neurobiologists to be homologous to the isocortex of mammals Are we able to correlate the above discussed differ- (KARTEN 1991; NORTHCUTT/KAAS 1995). However, ences in higher cognitive functions including con- major differences exist with regard to cytoarchitec- sciousness among groups of vertebrates with ture and size of the dorsal pallium/isocortex. In am- properties of their brains? When answering this phibians, the dorsal pallium is small and unlami- question, I will restrict myself to the discussion of nated; in lizards it is relatively larger and in some the tetrapod brain, because the components of the groups it shows a three-layered structure. In birds, forebrain of cartilaginous and bony fishes are diffi- those parts assumed to be homologous to the mam- cult to homologize with those of tetrapods. In the malian isocortex (i. e., DVR and wulst) are large, but following, I will briefly discuss the following unlaminated. In mammals excluding insectivores hypotheses which in the past have been forwarded and cetaceans, the dorsal pallium or isocortex shows in order to explain the alleged evolutionary the characteristic six-layered structure. But what increase in cognitive capabilities from frog to man. does a six-layered cortex mean? It has not been dem-

Evolution and Cognition ❘ 143 ❘ 1999, Vol. 5, No. 2 Gerhard Roth onstrated that mammals are “smarter” than birds brains, reptiles have small brains, birds relatively with the same brain size or size of cortex/pallium. large brains (particularly passerine birds and par- To date, there is no convincing answer to this im- rots; the “intelligent” pigeon, however, has a re- portant question. Furthermore, the pigeon is con- markably small brain), as do mammals in general. sidered by many behaviorists to be very smart, al- Among mammals, very large brains both in absolute though it has a brain and accordingly an and relative terms are found in primates (1.4 kg in unlaminated dorsal pallium/cortex that is much Homo sapiens), in elephants (up to 5.7 kg).and in smaller than that of a rat both in absolute and rela- whales including dolphins (up to 10 kg). Primates tive terms. Apparently, the same or very similar cog- generally have larger brains in relative terms than nitive functions are performed by anatomically very any other group of mammals except dolphins. Con- different kinds of pallium/cortex. trary to a common belief, new-world monkeys have (2) Brain size—either in absolute terms or relative larger brains than old-world ones (which include to body size—has often been correlated with “intel- Homo sapiens). Apes do not exceed monkeys in rela- ligence” or “higher cognitive abilities”, but the rea- tive brain size except man who has a brain that is 2– sons for increase in brain size are unclear. Body size 3 times larger than an ape or monkey of the same appears to be the single most important factor influ- size. During 3.5 million years of evolution of Homo encing brain size, i.e., large animals generally have sapiens, brain size has increased by a factor of 3–4, large brains. Increases and decreases in body size from 350 to 1400 ccm, but despite many scenarios, have occurred many times in virtually all animal the reasons for this increase remain obscure. groups and seems to happen fast (K-selection seems (3) The isocortex of “higher” mammals and par- to favor large, r-selection small bodies), and the ticularly man has been the target of many anatom- brain seems to follow these changes with some delay ical and physiological studies, because it is believed (VAN DONGEN 1998). However, increase in brain size to be the seat of higher cognitive functions (see does not linearly parallel increase in body size, but above discussion). However, while there are enor- only to the power of about 0.67, which means that mous differences both in absolute and relative brain small animals have relatively large brains and large size among tetrapods and mammals in particular, animals have relatively small brains (JERISON 1973). changes in relative size of the isocortex are relatively Among mammals, body size ranges from 3 g (the inconspicuous, because at least in mammals they insectivore Suncus etruscus) to 150 tons (Bal- more or less strictly follow changes in brain size. In aenopterus musculus, blue whale), which is a ratio of mammals, the correlation coefficient between brain 1: 50 millions; brain size ranges between 74 mg (the volume and cortical volume is almost exactly 1 bat Tylonycteris pachypus) to 10 kg (sperm whale), (JERISON 1997). This means that cortex size increases which is a ratio of 1: 130.000, i.e., still an enormous more or less isometrically with brain size. In man, but much smaller range than body size. Among the amount of neocortex is only slightly increased mammals, this is reflected by the fact that in very (76% compared to 72% in chimpanzee). Humans small rodents brains occupy up to 10% of body have a large isocortex because they have a large mass, in pigs 0.1% and in the blue whale, the largest brain, and accordingly elephants and most ceta- living mammal and animal 0.01%. ceans have much larger cortices than man. Isomet- Nobody really knows what relative—rather than ric increase relative to total brain volume is found absolute—brain size means. Since cell size does not in many other brain structures including the thala- generally increase with brain size, absolutely larger mus (which is tightly connected to the isocortex), brains mean more neurons, and there is good reason the basal ganglia (the largest subcortical structure in to believe that more neurons are better than fewer, the telencephalon) and the hippocampus (JERISON be that for faster and more sophisticated informa- 1997). tion processing, storage of memory traces, higher In the cortex of large-brained mammals, sensory, neuronal plasticity, better compensation of lesions. motor and association areas show differences in cy- Among vertebrates, we find many independent toarchitecture. Typically, with respect to the pres- cases of increases (as well as decreases) in absolute ence and thickness of layers with small or larger cell and relative brain size. In relative terms, chondrich- bodies, a heterotypic, granular motor cortex, a ho- thyans (sharks and rays) on average have large motypic, agranular frontal, parietal and temporal brains (with large telencephala), while osteichthyes association cortex and a heterotypic, granular sen- (bony fishes) have rather small brains (with notable sory cortex (also called “koniocortex”) are distin- exceptions). Amphibians have small to very small guished in the cortex of large-brained mammals

Evolution and Cognition ❘ 144 ❘ 1999, Vol. 5, No. 2 The Neurobiological Basis of Consciousness in Man and Animals

(CREUTZFELDT 1983). Besides these differences in cy- in the cortices of “higher” mammals that could be toarchitecture, there is great uniformity of cortex correlated with the observed “higher” cognitive with regard to the presence of cortical cells except abilities. So far, only two features have been discov- for the motor cortex, where small pyramidal cells ered that could distinguish the human cortex/brain are nearly absent. Rather, the different parts of the from that of other primates, viz., (1) differences in cortex are mainly characterized by their major tha- growth rate and (2) the presence of speech centers. lamic input: The prefrontal cortex receives its main As to growth, in humans, the isocortex is imma- input from the medial dorsal thalamic nucleus, the ture at birth, with simple afferent connections and precentral–motor cortex from the ventral lateral nu- “primitive” intracortical connections (CREUTZFELDT cleus, the anterior parietal cortex from the ventral 1983). Postnatal maturation consists in explosion- posterior nucleus, the parieto–temporo–occipital like increase in number of synapses in all parts of the cortex from the pulvinar, the supratemporal–audi- brain, increase in size of pyramidal neurons and di- tory cortex from the medial geniculate, the occipi- ameter of axons, increase in size in smaller cortical tal–visual cortex from the lateral geniculate and the neurons, increase in number of glial cells, growth of limbic–cingular cortices including the temporal intracortical capillary network and therefore strong pole from the anterior thalamic nucleus. There is no decrease in neuron density. This maturation is sign of something unique to humans or other pri- nearly completed 3 months after birth in monkeys mates in regard to these cortical features including and at 3 years in humans, but the human brain still those of the association cortex. continues to mature until the age of 20, which is (4) There is a common belief that during hominid much longer than in any other primate. A critical brain evolution the association cortex has increased phase in the development of the human brain seems dramatically. However, there are no precise criteria to occur around 2.5 years. At this time, major ana- for distinguishing primary and secondary sensory tomical rearrangements in the association cortex cortical areas from true association areas. Recently, have come to a stop, and the period of “fine-wiring” KAAS (1995) has argued that the number of isocorti- seems to start, particularly in layer 3 of the prefron- cal areas has strongly increased from the hypothet- tal cortex (MRZLJAK et al. 1990). As mentioned ical insectivore-like ancestor with about 20 such ar- above, at this time, human children cognitively eas to more than 60 in primates. However, what— “take off” compared to non-human primates. according to KAAS—has increased is the number of The other exception concerns the presence of so- functionally “intermediate” areas, but not the pri- called speech centers in the human brain, particu- mary or “highly associative” ones. Furthermore—as larly the BROCA speech center in the frontal lobe KAAS rightly emphasizes –, there is the danger to responsible for temporal aspects of language includ- strongly underestimate the number of functionally ing syntax, and the WERNICKE speech center in the different cortical areas in small-brained mammals. temporal lobe responsible for the meaning of words Thus, whether associative cortex has in fact substan- and sentences (although meaning is likewise depen- tially increased in relative terms with increase in dent on syntax and grammar). However, it is to date brain and cortical size, remains undecided so far. unclear whether these speech centers really are evo- The difficulties with defining the prefrontal cor- lutionary novelties. Several explanations have been tex anatomically in the various groups of mammals offered for the absence or strong limitations of hu- are even greater. UYLINGS/VAN EDEN (1990) define— man-like speech in non-human primates. It has according to the neuroanatomical tradition—the been argued that the vocal tract of non-human pri- PFC as the cortical area with major (though not ex- mates is incapable of producing sounds typical of clusive) input from mediodorsal thalamic nucleus. and necessary for the human language. However, Using this definition, the volume of PFC relative to the capabilities of the vocal tract of non-human pri- the rest of the isocortex does not differ dramatically mates appear to exceed previous assumptions and between the rat (24%) and Homo sapiens (30%). Fur- the articulatory gestures associated with the non- thermore, the PFC of the orangutan is equal to hu- human primate vocal output are more complicated mans in relative terms, while in monkeys it varies than originally thought. Yet, some inherent sound- between 11 and 18%. In general, PFC increases iso- producing limitations of this apparatus cannot be metrically with an increase in cortical and overall denied, particularly with respect to vowels and for- brain volume (JERISON 1997). mant transitions. If these were the only limitations, (5) Remains the question of any cortical or sub- we would expect the great apes to naturally produce cortical anatomical or physiological specializations or learn from humans a much richer vocal language

Evolution and Cognition ❘ 145 ❘ 1999, Vol. 5, No. 2 Gerhard Roth mostly based on consonants. Thus, there seem to be According to this view, the evolutionary basis for differences in the neural control of vocal language human language was an emotionally driven lan- in addition to the restrictions of the vocal apparatus. guage typical of non-human primates. During hom- What could these be? inid evolution, the isocortex gained control over All mammals studied so far have a center for in- this system such that beyond the initiation of intraspecific communication within the temporal nate vocalization a “free” production of sounds and lobe (mostly left) which may be homologous to the their sequences became possible. Such an interpre- WERNICKE center for semantics. It has been reported tation, however, is in contrast with recent evidence that destruction of these areas leads to deficits in of a high degree of sound learning in monkeys (ZIM- intraspecific vocal communication (HEFFNER/HEF- MERMANN 1995) and the mentioned consequences FNER 1995). In addition, it has long been argued that of destruction of left-hemispheric, WERNICKE-like the posterior part (A 44) of the BROCA speech center temporal areas. in humans and the ventral premotor area (PMV) of Leaving this problem unsolved, non-human pri- non-human primates are similar in location and cy- mates including the great apes are strongly limited toarchitecture and probably are homologous even in non-vocal speech based on the use of sign (PREUSS 1995). PMV controls movement of fore- language or symbols, and these limitations seem to limbs, face and mouth, which is the case with the concern mostly syntax. Accordingly, if something posterior portion of the BROCA area. However, JÜR- relatively new concerning language developed or GENS (1995) emphasizes that in non-human pri- became substantially modified in the human brain, mates (e.g., squirrel monkeys), intraspecific vocal it was probably the BROCA center and not so much communication does not involve isocortical func- the WERNICKE center, which already may have ex- tions. According to JÜRGENS, in primates including isted in non-human primates. Such an assumption man, critical structures for “innate” sound produc- is consistent with the fact that the most clear-cut tion are the mesencephalic periaqueductal gray differences between humans and non-human pri- (PAG) and laterally bordering tegmentum. Destruc- mates concern syntactical complexity of language tion of these two centers alone produces mutism in (see above). Thus, a reorganization of the frontal– monkeys and man. The PAG is controlled by sen- prefrontal cortex may have happened such that the sory input via mesencephalon and medulla oblon- facial and oral motor cortex and the related subcor- gata (e.g., colliculus superior and inferior) and in tical speech centers came under the control of a kind turn controls the motor nuclei innervating the la- of isocortex that is specialized in any aspects of tem- ryngeal, oral and respiratory muscles involved in poral sequence of events including the sequence of vocalization. action (DEACON 1990). This would be consistent While in humans, isocortical centers (e.g., BROCA with the fact that the posterior part of BROCA’s area and WERNICKE) are obligatory for voluntary control appears to be homologous with the ventral premo- of vocalization, in non-human primates, only the tor area of non-human primates. Thus, only the an- anterior cingulate cortex is believed to be involved terior part of BROCA’s area would be a “new” struc- in this function. Destruction of this cortical area ture. To date, there is no good idea when and how abolishes self-initiated vocalization in the monkeys, such an evolutionary step might have occurred. while in man, it abolishes all kinds of emotional For many reasons, it is unlikely that conscious- control of vocalization, and utterances become very ness is strictly bound to the existence of syntactical monotonous, but patients can still speak. JÜRGENS language and originated with it. It is clear, however, (1995) concludes that the anterior cingulate cortex that the evolution of a syntactical language strongly is the main center for voluntary control of emotional favored higher states of consciousness including utterances. This structure is directly connected with self-reflection, thinking and action-planning. the PAG and functions in parallel to the isocortical While thinking is not necessarily bound to lan- speech centers. According to JÜRGENS, non-human guage, most people think and plan verbally. Fur- primates lack a direct connection between the mo- thermore, many concepts typical of human mind tor cortex and the laryngeal motor neurons. In man, “exist” only linguistically, i.e., because we can talk bilateral destruction of the facial motor cortex abol- about them, e.g., future events or abstract entities ishes the capacity to produce learned vocalization such as society and freedom. including speech or humming a melody, while a Certainly, the presence states of consciousness were similar destruction in monkeys has no such conse- and are of great advantage for an individual, particu- quence. larly regarding social life. In this direction goes the

Evolution and Cognition ❘ 146 ❘ 1999, Vol. 5, No. 2 The Neurobiological Basis of Consciousness in Man and Animals

“MACHIAVELLIAN intelligence” hypothesis of BYRNE with a complex natural and social environment, to (1995). According to this hypothesis, social environ- make complicated choices and to carefully plan one’s ment exerted a strong selection pressure resulting in own actions based on previous experience, to modify self-recognition, empathy, imi- behavioral strategies quickly at tation, pretend play, decep- new demands. tion, theory of mind, and Author’s address teaching. All this was and is Gerhard Roth, University of Bremen, Brain used for social manipulation. Research Institute, D-28334 Bremen, Ger- However, equally important many. Email: [email protected] was and is the ability to deal

References Neurobiology, and Behavior. Brookes Publ. Company: Bal- timore, London, Toronto, Sydney, pp. 9–26. Aggleton, J. P. (1992) The Amygdala: Neurobiological As- Jürgens, U. (1995) Neuronal control of vocal production in pects of Emotion, Memory, and Mental Dysfunction. non-human and human primates. In: Zimmermann, E./ Wiley-Liss: New York, Chichester. Newman, J. D./Jürgens, U. (eds) Current Topics in Primate Aggleton, J. P. (1993) The contribution of the amygdala to Vocal Communication. Plenum Press: New York, London, normal and abnormal emotional states. Trends in Neuro- pp. 199–206. sciences 16: 328–333. Kaas, J. H. (1995) The evolution of isocortex. Brain Behav. Alexander, G. E./Crutcher, M. D./DeLong, M. R. (1990) Bas- Evol. 46: 187–196. al ganglia–thalamocortical circuits: Parallel substrates for Karten, H. J. (1991) Homology and evolutionary origins of motor, oculomotor, “prefrontal” and “limbic” functions. the “neocortex”. Brain Behav. Evol. 38: 264–272. In: Uylings, H. B. M./van Eden, C. G./de Bruin J. P. C./ Kinsbourne, M. (1995) Models of consciousness: Serial or par- Corner, M. A./Feenstra, M. G. P. (eds) The Prefrontal Cor- allel in the brain? In: Gazzaniga, M. S. et al. (eds) The Cog- tex. Its Structure, Function and Pathology. Elsevier: Am- nitive Neurosciences. MIT Press: Cambridge MA, pp. 1321– sterdam, New York, Oxford, pp. 119–146. 1329. Baars, B. J. (1997) In the Theater of Consciousness: The Knight, R. T./Grabowecky M. (1995) Escape from linear time: Workspace of the Mind. Oxford University Press: Oxford. Prefrontal cortex and conscious experience. In: Gazzaniga, Baron-Cohen, S./Leslie, A. M./Frith, U. (1985) Does the autis- M. S. et al. (eds) The Cognitive Neurosciences. MIT Press: tic child have a “theory of mind”? Cognition 21: 37–46. Cambridge MA, pp. 1357–1371. Byrne, R. (1995) The Thinking Ape. Evolutionary Origins of Libet, B. (1978) Neuronal vs. subjective timing for a con- Intelligence. Oxford University Press: Oxford, New York, scious sensory experience. In: Buser P. A./Rougeul-Buser, Tokyo. A. (eds) Cerebral Correlates of Conscious Experience. Creutzfeldt, O. D. (1983) Cortex Cerebri. Leistung, struk- Elsevier/North-Holland: Amsterdam, pp. 69–82. turelle und funktionelle Organisation der Hirnrinde. Locke, J. L. (1995) Linguistic capacity: An ontogenetic theory Springer: Berlin, Heidelberg, New York. with evolutionary implications. In: Zimmermann, E./New- Deacon, T. W. (1990) Rethinking mammalian brain evolu- man/J. D./Jürgens, U. (eds) Current Topics in Primate Vo- tion. American Zoologist 30: 629–705. cal Communication. Plenum Press: New York, London, Gardner, R. A./Gardner T. B./van Cantfort T. E. (1989) Teach- pp. 253–272. ing Sign Language to Chimpanzees. State Univ. New York MacPhail, E. (1998) The Evolution of Consciousness. Oxford Press: New York. University Press, Oxford: New York, Tokyo. Guillery, R. W./Feig, S. L./Lozsádi D. A. (1998) Paying atten- Magistretti, P. J./Pellerin, L./Rothman, D. L./Shulman, R. G. tion to the thalamic reticular nuclei. Trends Neurosci. 21: (1999) Energy on demand. Science 283: 496–497. 28–32. Markowitsch, H. J. (1992) Neuropsychologie des Gedächtnis- Hassler, R. (1978) Interaction of reticular activating system ses. Hogrefe: Göttingen. for vigilance and the truncothalamic and pallidal systems Markowitsch, H. J. (1999) Gedächtnisstörungen. Kohlham- for directing awareness and attention under striatal con- mer: Stuttgart. trol. In: Buser, P. A./Rougeul-Buser, A. (eds) Cerebral Corre- Marten, K./Psakaros, S. (1994) Evidence of self-awarenes in lates of Conscious Experience. Elsevier/North-Holland: the bottlenose dolphin (Tursiops truncatus). In: Parker, T./ Amsterdam, pp. 111–129. Mitchell, R. W./Boccia, M. L. (eds) Self-Awareness in Ani- Heffner, H. E./Heffner R. S. (1995) Role of auditory cortex in mals and Humans: Developmental Perspectives. Cam- the perception of vocalization by Japanese Macaques. In: bridge Univ. Press: Cambridge, pp. 361–379. Zimmermann, E./Newman, J. D./U. Jürgens (eds) Current Meltzoff, A./Gopnik, A. (1993) The role of imitation in un- Topics in Primate Vocal Communication. Plenum Press: derstanding persons and developing a theory of mind. In: New York, London, pp. 207–219. Baron-Cohen S./Tager-Flusberg, H./Cohen, D. J. (eds) Un- Jerison, H. J. (1973) Evolution of the Brain and Intelligence. derstanding Other Minds: Perspectives from Autism. Ox- Academic Press: New York. ford University Press: Oxford, pp. 335–366. Jerison, H. J. (1997) Evolution of prefrontal cortex. In: Kras- Metzinger, T (1993) Subjekt und Selbstmodell. Die Perspek- negor, N. A./Lyon, G. R/Goldman-Rakic, P. S. (eds) Devel- tivität phänomenalen Bewußtseins vor dem Hintergrund opment of the Prefrontal Cortex: Evolution, einer naturalistischen Theorie mentaler Repräsentation.

Evolution and Cognition ❘ 147 ❘ 1999, Vol. 5, No. 2 Gerhard Roth

Schöningh: Paderborn. VCH: Weinheim, pp. 1–31. Moscovitch, M. (1995) Models of consciousness and memo- Savage-Rumbaugh, S. (1984) Acquisition of functional sym- ry. In: Gazzaniga, M. S. et al. (eds) The Cognitive Neuro- bol usage in apes and children. In: Roitblat, H. L./Bever, T. sciences. MIT Press: Cambridge MA, pp. 1341–1356. G./Terrace, H. S (eds. ): Animal Cognition. Earlbaum: Hills- Mrzljak, L./Uylings, H. B. M./van Eden, C. G./Judás, M. dale, New Jersey, pp. 291–310. (1990) Neuronal development in human prefrontal cortex Squire, L. R. (1987) Memory and Brain. Oxford University in prenatal and postnatal stages. In: Uylings, H. B. M./van Press: New York. Eden, C. G./de Bruin J. P. C./Corner, M. A./Feenstra, M. G. Squire, L. R./Knowlton, B. (1995) Memory, hippocampus, P. (eds) The Prefrontal Cortex. Its Structure, Function and and brain systems. In: Gazzaniga, M. S. et al. (eds) The Cog- Pathology. Elsevier: Amsterdam, New York, Oxford, pp. nitive Neurosciences. MIT Press, Cambridge MA, 825–836. 185–222. Stamp Dawkins, M. (1993) Through Our Eyes Only? The Nieuwenhuys, R./Voogd J./van Huijzen, C. (1988) The Hu- Search for Animal Consciousness. W. H. Freeman/Spek- man Central Nervous System. Springer: Berlin, Heidelberg, trum: Oxford, New York, Heidelberg. New York. German edition: Nieuwenhuys, R., Voogd, J./ Stromswold, K. (1995) The cognitive and neural bases of lan- Huijzen, C. van (1991) Das Zentralnervensystem des Men- guage acquisition. In: Gazzaniga, M. S. et al. (eds) The Cog- schen. Springer: Berlin, Heidelberg, New York. nitive Neurosciences. MIT Press: Cambridge MA, pp. 855– Northcutt, R. G./Kaas, J. H. (1995) The emergence and evolu- 870. tion of mammalian isocortex. Trends in Neurosciences 18: Uylings, H. B. M./van Eden, C. G. (1990) Qualitative and 373–379. quantitative comparison of the prefrontal cortex in rat and Parker, S. T./Mitchell, R. W./Boccia, M. L. (1994) Self-aware- in primates, including humans. In: Uylings, H. B. M./van ness in Animals and Humans: Developmental Perspec- Eden, C. G./de Bruin J. P. C./Corner, M. A./Feenstra, M. G. tives. Cambridge Univ. Press: Cambridge MA. P. (eds) The Prefrontal Cortex. Its Structure, Function and Pearce, J. M. (1997) Animal Learning and Cognition. Psy- Pathology. Elsevier: Amsterdam, New York, Oxford, pp. chology Press: Exeter. 31–62. Posner, M. I. (1994) Seeing the mind. Science 262, 673–674. Van Dongen, P. A. M. (1998) Brain size in Vertebrates. In: Preuss, T. M. (1995) Do rats have a prefrontal cortex? The Nieuwenhuys, R./ten Donkelaar, H. J./Nicholson, C. Rose–Woolsey–Akert program reconsidered. J. Cognitive (1998) The Central Nervous System of Vertebrates, Vol. 3. Neurosci. 7: 1–24. Springer: Berlin, pp. 2099–2134. Robbins, T. W. (1997) Arousal systems and attentional pro- Visalberghi, E./Limongelli, L. (1994) Lack of comprehension cesses. Biol. Psychology 45: 57–71. of cause–effect relationships in tool-using capuchin mon- Robbins, T. W./Everitt, B. J. (1995) Arousal systems and at- keys (Cebus apella). J. Comp. Psychol. 108: 15–22. tention. In: Gazzaniga, M. S. et al. (eds) The Cognitive Voytko, M. L. (1996) Cognitive functions of the basal fore- Neurosciences. MIT Press: Cambridge MA, pp. 703–720. brain cholinergic system in monkeys: Memory or atten- Robbins, T. W./Everitt, B. J. (1996) Neurobehavioural mecha- tion? Behav. Brain Res. 75: 13–25. nisms of reward and motivation. Current Opinion in Neu- Zimmermann, E./Newman, J. D./Jürgens, U. (1995) Current robiology 6: 228–236. Topics in Primate Vocal Communication. Plenum Press: Roberts, A. C./Robbins, T. W./Weiskrantz, L. (1998) The Pre- New York, London. frontal Cortex. Executive and Cognitive Functions. Oxford Zimmermann, E. (1995) Loud calls in nocturnal prosimians: University Press: Oxford, New York, Tokyo. Structure, evolution and ontogeny. In: Zimmermann, E./ Roth, G./Wullimann, M. F. (1996) Die Evolution des Nerven- Newman, J. D./Jürgens, U. (eds) Current Topics in Primate systems und der Sinnesorgane. In: Dudel J./Menzel, R./ Vocal Communication. Plenum Press: New York, London, Schmidt, R. F. (eds. ), Lehrbuch der Neurowissenschaft. pp. 47–72.

Evolution and Cognition ❘ 148 ❘ 1999, Vol. 5, No. 2 Franz Seitelberger

Neurobiological Comments on Empirical Consciousness

HE PHENOMENON OF Abstract one understands a united T consciousness is con- individual experience sidered as the main cross- The unexplainable categorial difference in the struc- consisting of two compo- ing point between neuro- tural fundament of the human mental life is charac- nents: sciences and cognitive sci- terised and discussed regarding I. KANT´s transcenden- (a) someone who has, ences. The wealth of tal concept and some topical thories of the i.e., disposes conscious- pertinent literature hardly psychophysical problem. Respective facts of neurosci- ness, a quality of the liv- can be surveyed and the entific brain research and recent results of studies con- ing subject, and problem solving strategies cerning the unique human capacity of language and of (b) somewhat which to the respective para- interindividual communication are mentioned. occurs in such a con- mount problems are fre- sciousness, i.e., the con- quently repeated without Key words tents of consciousness, an outlook of progress. resp. objects for a subject. Before starting a re- Conciousness, self-sentience, object–subject–ego rela- In the psychological newed discussion on cru- tion; brain–mind duality, information processing. consciousness thus a cial points of the subject “self-conscious” part as it seems advisable to look at the primary notion of concrete subjective experience is given, namely not the word consciousness as central object of the fol- as an object of perception but as a concomitant or, lowing remarks. Linguistically, ‘consciousness’ is a obligatorily enclosed self-conscious part as concrete so-called abstract–noun of predicative–adjectival experience. This “self conscious” part is not a object significance. Consciousness means to be conscious. of sensation, but a concomitant or a obligatorily en- Accordingly there are facts given in the conscious- closed self-sentience of the conscious subject con- ness but there doesn’t exist the consciousness given nected with an indistinct bodily and intentional as an object of science. This has clearly been ex- feeling vis–a–vis or directed to the appearing objects: pressed by FICHTE: “In each consciousness some- in any conscious state this ego-feeling appears in dif- thing is given as conscious in somebody, which, ferent degrees of clearness. It is discussed as faint however, is not the consciousness itself.” (FICHTE allusion of the individual´s private life process. The 1845/46) occurring consciousness continuously unifies both In WUNDT’s words: “Consciousness always is con- unequal parts into the process of individual consciousness of something, namely of our inner expe- sciousness. In the sensation–perception of pain riences of the states and processes in us!” (WUNDT which is not a sense like vision or audition the self- 1873/74). feeling becomes prominent, initiates self-protection and is experienced as one source of the self- Empirical consciousness consciousness, whereas the other one consists in the neuronal self-representation of the conscious What, however, in phenomenological analysis is the ego. substrate of this abstraction? Consciousness refers to From the basic duality of consciousness the per- a certain quality of a living being as a manifestation manently moved and intangible character of the of it`s individual behavior. This is the primary expe- ego results. The ego, because it itself is the factual rience underlying the empirical or biological notion subject and lacks a sensory device, due to his imme- of consciousness. With psychological consciousness diacy never can become an object to the conscious-

Evolution and Cognition ❘ 149 ❘ 1999, Vol. 5, No. 2 ness. The body too is not the ego but is the percep- Thus the transcendental ego does not mean an tually objectivated ego-experience. empirical ego but concerns an impersonal formal The other part of consciousness consists of the subject. The transcendental consciousness, how- conscious contents, i.e., events, imaginations, ever, is not an empty idealistic construct but the thoughts, emotions, drives etc.. These are the ob- result of high-level abstraction processes, in KANT’s jects themselves, experienced in external percep- often misunderstood words “an idea of reason”. tion, embedded into subjective conscious experi- This concept represents the fundamental fact of ence. Only within the contents of consciousness “esse est percipi” and the inherent subject–object physical entities and reality is immediately given to correlation in our perception of reality it appears as us, as one fundamental element of our dual experi- constitutive for the possibility of experience and science consisting of the reciprocal opposition be- entific exploration of the world, because, following tween subject and object. KANT, the empirical consciousness, i.e., the subject– As already indicated, the quantitative portion and object relation, possesses “a necessary relation to a distinctness of the basic stationary self-awareness transcendental consciousness, namely to the con- given in consciousness may vary considerably over sciousness of my self as the primary apperception” time, e.g., depending on the vital impact of autono- (KANT 1922). And: “The synthetic sentence, that mous cognitive activities. This also reflects the asym- each different empirical consciousness had to be metric relation between the conscious subject and connected to an only self-consciousness, is plainly his surrounding world. Even more important, it the first and synthetic principle of our thinking at points to the apparent paradox that this inner dual- all.” (KANT 1922) ity of consciousness regarding self-sentience versus This triangular or ‘elliptic’ relation between sub- events and self-activity forms the unifying clamp of ject, object and reflection is the fundament of hu- the different aspects of individual existence, so to say man conscious functions. the invariant function of the human living system in spatial–temporal dimension. Neuropsychological characters and the duality of consciousness Excursion on transcendental The critical feature of the phenomenon of con- consciousness sciousness is “self-sentience”, sometimes also called Although in this treatise a systematical philosophi- “the primarily psychic”. Self-sentience is lacking cal discourse on consciousness is not intended any sensory quality; it is not experienced as repre- KANT’s transcendental notion of consciousness sentation of a concrete object and therefore does should be mentioned because it possesses an indis- not possess the type of information as present in pensable function for orientation within the prob- consciously perceived contents of outer or inner lem of consciousness in general. The central reality: there doesn’t exist any expertise of that formulation by I. KANT in the original text reads as kind about self-sentience and equally not about follows: respective cognitive products. ”Das ich denke muß alle meine Vorstellungen Regarding the compact abundance of its con- begleiten können denn sonst würde etwas in mir tents, the dynamics of consciousness become re- vorgestellt werden, was gar nicht gedacht werden duced in states of rest and high concentration. könnte welches ebensoviel heißt, als: Die Vorstel- However, there doesn’t exist consciousness with- lung würde entweder unmöglich oder wenigstens out any contents, if not as a symptom of psychic für mich nichts sein.” (KANT 1922) disease. The criterion ‘conscious’ doesn’t mean a In other words: Everything existing for me has to quality of defined contents, caused by a proper neu- be a phenomenon in my consciousness. ‘Conscious’ ral pattern of activity, as it is sometimes assumed by in this context means to be an object for a subject or reductionists. Rather it offers the transcendental for an ego, respectively. The latter is possibly not situation of the perceiving subject as experiencing something real. Rather, it is the virtual frame of var- himself as the ego. Thus it appears to be more cor- ious conscious experiences. If the subject by reflec- rect to speak of “representations in consciousness” tion towards his self changes to an object, then it will instead of “conscious representation”. The psycho- loose the ego-character and will exist in this context genetic hypothesis of a “primary psychic appercep- only as representation of oneself but not as the ade- tion” consisting in vague sensation of the outer quate conscious self. world and indistinct self-sentience therefore means

Evolution and Cognition ❘ 150 ❘ 1999, Vol. 5, No. 2 Neurobiological Comments on Empirical Consciousness a retrograde projection of high probability; it can Finally the experience of time and temporal be recognized already in most primitive organisms events has to be emphasized. In consciousness we in self protecting behavior as “instinctive care of perceive ourselves as immediately participating or ones self”. During evolution towards humans this engaged in an uninterrupted stream of events lead- vague self-sentience in co-evolution with the brain ing from the past into the future, containing remem- by means of concrete conscious experience has bered as well as expected contents. been implemented to the conscious self-concept What is experienced as presence is not an imagi- and and since about 200 000 years a neurobiologi- nary borderline between the gone and the coming, cally important element for the strategies of human but a distinct entity of the conscious process be- self-representation. tween the elapsed past and the not yet arrived future In the latter phase, contrary to the previous trend of a duration of about 4 sec. This time period appears of evolution, a transition from hierarchical organi- necessary for generating the neuronal correlates of zation of neural networks to combinatorical net- the actual contents of experience. E.g., of the sensory works with self-referential metafunctions occurred. input, events, thought processes, impulses of action This can be considered as one of the cornerstones of etc. and likewise for fixation and evaluation of the cultural evolution. conscious contents regarding self-sentience. In con- The “psycho–physical problem”, i.e., the ques- sciousness we immediately experience our own relation regarding nature and connections of both parts tion to the world as the frame of relevant behavior of consciousness, distinguished as sentience and organized by higher brain functions. The basic con- imagination, occupies occidental thinking since scious brain processes, however, cannot be clarified more than 2500 years. Due to the progress in brain by our observation: there is not any insight into the research now also natural sciences are increasingly conscious process. Yet in the dynamics of relevant interested in the “body–mind problem”. Interdisci- conscious contents a comprehensive selection of plinary research in cognition unequivocally proved disparate results of neuronal processing mechanisms that consciousness—understood as inner and inter- is offered. This also has been discussed as the ‘phe- subjective experience and control of behavior—is nomenon of binding’. Yet conscious experience not connected with cerebral processes. Because these ex- only contains actual data but also higher psychic clusively are defined as biological physico–chemical processes, e.g., the above mentioned time-related be- processes consciousness cannot arise from or occur havior, the phenomenon of representation of the assisted by processes of different essence. This con- world and the ego on several levels, connected with clusion is basic to the position of “psycho–physical the reflection of these relations in the self-concept. parallelism” which interprets the psychic states and Consciousness in general can be defined as the pre- actions of behavior as special systemic processes oc- sentation of objectivity in the medium of subjectiv- curring parallel to neuronal processes. ity: “The being–for–me of something as something.” A modified version of this position appears in the monistic identity theory, considering both men- Neurobiological correlates of cognition tioned parts of consciousness as two aspects of one and the same brain. Conscious phenomena, as exposed before, are In contrast the hypothesis of substantial “dual- objects of neuropsychology. Due to their inherent ism” affirms that the immaterial soul acts upon the basic duality principally new aspects or attempts are brain and by that on the behavior. This, however, scarcely available for the next time. Cerebral corre- would demand the unacceptable presupposition lates of cognitive processes, however, are objects of that in this case the laws of physics could not be neurophysiology, a big domain of striking complex- valid and other influences would become effective. ity but in rapid progress. Before presenting a few Also the attempt to explain the duality of conscious- actual problematic fields the instrumental complex- ness by a dualism of neuronal information process- ity of the neuronal system has to be shown. ing, i.e., by the supposition that self-sentience is The central position of the brain regarding con- founded in not yet known, specific neural patterns trol of vital functions as well as of mental perfor- of activity would not clarify the impenetrable na- mances is generally accepted. Nevertheless it has to ture of conscious experience: The problem only be mentioned, that the human brain compared with would be shifted to the level of a physicalistic reduc- the brains of our phylogenetic ancestors exhibits es- tion where no sound explanation and definition sential innovations and structural completions, e.g., can be found. total corticalization of the input from environment

Evolution and Cognition ❘ 151 ❘ 1999, Vol. 5, No. 2 and from the body as basis of complete central con- ferred to coded patterns of electrochemical pulses. scious as well as unconscious integration of informa- The same is true for the products of autonomous, tion regarding self-representation and individual be- e.g., cognitive brain activity. The specific function havior—also including vegetative functions and of the brain completely occurs in this translated drives. In the following only main facts and essential form, whose processes of transformation are not features of the specialized organic brain structures completely explored. Neuronal information is and of the immense realm of their functions will be totally different to the physical quality of its natural discussed. sources and lacks a definable physical nature, yet Two general aspects of the complex neurofunc- nevertheless is a constitutive physiological phe- tional central system have to be emphasized: nomenon. Neuronal networks are realized in the (a) the neurobiological substrate, the brain, is a brain in countless samples differing in relation to hierarchically organized mega-system consisting of complexity, functional architecture and diverse pat- about 12 billions of nerve cells, i.e., neurons of dif- terns of connectivity with other networks of corti- ferent types arranged in numerous variant accumu- ces, grisea and nuclei. Brain morphology and lations of different size: cortices, nuclei, tractus, bun- physiology demonstrates a hierarchy of structural dles etc. Each single neuron may have more than levels, with the cerebral cortex in the top position. 10,000 contacts, i.e., synapses, to others; the total The prominent grey formation contains half the number of these amounting to several 1000 billions. number of brain neurons and in humans is con- Due to that the number of possible different states nected with all other brain regions. However it does of brain function runs to 103000 what also means that not exert an absolutistic functional dominance but an exact decoding of micromorphological brain in contrary a widely distributed democratic regime states is beyond the borderline of intracosmic possi- consisting of many unexplained rules. There is no bilities. The instrumental complexity of brain struc- central processing unit and no localized site of ture is reflected in the organ-specific biochemistry of authority. various physiological trophic and regenerative func- In spite of remarkable progress of neurosciences tions which can be studied exactly. and of bioinformatics the chances to follow and to (b) However, with regard to systemic functions, in localize the complicated way of single signals and all contrast to other organs and equally to the com- the more to identify widely distributed programs of puter, the brain system beyond its organic embed- higher level-activities are mean, although today ment into the life process does not exhibit an own more exact instruments for registration of activation concrete physical product of its systemic activity but pattern in the cerebral cortex are used. Today the only metaorganic results of its proper function man- exact knowledge of brain activities is very limited ifested in the discussed phenomena of the con- and their complete description and exact analysis so sciousness and in form of human cognition. far impossible. Thus the metaphor of C.H. SHER- RINGTON of brain activity as an “enchanted loom” Information and information processing still appears as pertinent. In general brain activities do not produce images The basic level of the brain considered as a system but compatible abstract reconstructions of reality. consists of the mentioned enormous connectivity Concerning the subject–object relation it must be of neurons and interneuronal exchange of ‘excita- emphasized that this pragmatic reconstruction is tion’, demonstrated by neurophysiological meth- not an arbitrary one but is based on procedures as ods. The total of respective events in this level of outcomes from millions of years of adaptation con- neural interconnected actions is named Information taining creative and optimizing procedures. Processing. The underlying structure of this function is the neural network formed of cooperating neu- Cognitive brain performances rons in variant designs. Network activity, i.e., information, generated by figuration, dynamics and The first step of information processing is sensory transport of excitation patterns, lacks the usual cri- perception whose adaptability does not depend on teria of objects of natural science: organic material the precision of single partial senses e.g., visual or qualities. Information contains the wide spectrum olfactory acuity but on the quality of the unity of of different physical stimuli from the outer world as perception (in German: Gestalt-Qualität). well as from the body, both reaching the brain. In humans the object of perception, i.e., the prod- These network processes become selectively trans- uct of neural reconstruction of reality means an

Evolution and Cognition ❘ 152 ❘ 1999, Vol. 5, No. 2 Neurobiological Comments on Empirical Consciousness enormously complex synthesis of multisensory in- Symbol competence, language capacity put enriched and instructed by an effective concert and human cognition of inborn and learned strategies, memories and respective emotional and rational evaluations. It By means of decisive improvement of the ability of should not be neglected that the resulting human symbol formation and related language capacities a “object of perception” contains more information new level and capacity of cognition emerged. than that of higher mammals, monkeys included. Regarding size, dimension and universal availabil- Moreover the complex neural patterns representing ity this symbol faculty led to enormous progress external events in humans lead to an own functional surpassing all other types of neuronal processes. and ontological status, with increasing indepen- The evolution of phonetic language is based on dence from the underlying levels, e.g., reflex-like the non-neuronal transcodification of products of mechanisms. The new combinatoric capacities form neural information processing, especially of sym- the basis of higher cognitive processes, like compar- bolic cognitive products, into acoustic signals in lin- ison and classification processes as well as related guistic, syntactical order. E.g., the word “house”, as abstractive abilities. a simple acoustic pattern represents an immensely On this step of cognition the diversity and com- rich and branched network pattern which, how- plexity of the reality is simplified by reductive ab- ever, in its non-neuronal form easily can overcome straction yet without loosing significance and essen- the network-cascades and produce a wealth of new tial contents but on the contrary offering new relations and contents. Language is the richest aspects of objects. We speak of models equivalent to source of relevant empirical conscious contents reality, in short of real-models of objects of percep- whose prominent selective importance for the hu- tion and events. In the frame of our spatial–temporal man cognitive evolution appears as evident. The orientation (not discussed here) these models are the funds of dispositions of human behavior by the lan- source of the diversity of objects and their structural guage-bound faculty of symbols as well as by in- arrangement in our perceived world. crease of the neuronal capacity of learning and Moreover single elements of the these processes memory has been extraordinarily enlarged; what become part of mentally constructed worlds. These also promoted the representation of temporal products are named model-objects: together with im- events and opened the chance to anticipate future proved memory and the extrasensory faculty of needs. The central part of conscious utilization of time experience, which opens the dimension of fu- symbolic products underlies rational thinking. Last ture, the mentioned model-types enable humans of but not least by means of language also the consti- integrative thinking processes, i.e., pragmatic and tutive isolation of the private inner individual world innovative use of products of imagination, includ- has been partially opened: language makes the con- ing instant planning of behavior and future activi- sciousness of the fellow transparent and communi- ties. cable, forming a key event of human empathy and With regard to the human level of cognition the of the socio–cultural evolution: “At the beginning fundamental role of the body has to be sketched. The existed the word.” brain–body relation in humans is the most developed. In relation to the development of language it has Compared to other animals the sensorimotor repre- to be noticed that one of its most important precur- sentation of the human body is the most perfect one. sor faculties probably has been gesticulation—what Beside this, in humans the vegetative functions, in- can already be demonstrated in primates. In the pri- cluding the vital drives are represented in the cere- mate cortex also so-called mirror neurons were bral cortex, in the already mentioned archaic form found which are activated not only by direct func- of a conscious emotional feeling (German: Befind- tional excitation but also in case of sensory trans- lichkeit) and accidentally in the subcognitive sensa- mission of respective events in the environment, tion of pain. e.g., gestures of social partners. This type of trans- The so-called higher brain processes therefore can personalization is indispensable for signal commu- be interpreted as results of the archaic body–brain– nication. complex and of its self-referential unity. Finally the The structural fundaments of the unique evolu- rational constitution of an individual self-concept tionary event of language capacity are manifest in and the reflexive representation of the world mani- the brain: the human cortex cerebri not only be- fested the inherent duality of the human conscious came distinctly enlarged regarding extension, area experience. differentiation and quantity of neurons, but before

Evolution and Cognition ❘ 153 ❘ 1999, Vol. 5, No. 2 all was qualitatively furnished by new and increased Final remarks on consciousness connection systems as well with equally enlarged other cortical association areas in the left hemi- For a further rational approach to consciousness we sphere and with brain regions which that far lacked have to start from the assumption that scientific the- cortical connections. The result was complete corti- ories are rendered possible by language and mathe- calization of the outer and inner input. Altogether matics, i.e., by simplifying coding processes of these innovations form the mighty structural axis reality, mediated by conscious experience. The self- of the integration of linguistic functions into the sentient component of psychic experience, however, already lateralized functional specialization of the is not a sort of reality appropriate for these proce- human brain. dures: it is associated with objects or features of con- Language competence includes speaking: Yet the scious contents but is not an element of this part of structures of articulation of the anthropoid apes consciousness. Thus consciousness is something like lack the ability of producing the sounds of human a medium in which the products of information pro- articulation. In the course of brain–language co- cessing procedures are becoming evident, or some- evolution a new direct fibre tract from the BROCA thing like a forum of initiation into full membership area of the motor cortex towards the augmented to the conscious contents. Furthermore conscious- muscle bundles of the vocal cords of the larynx, the ness cannot be something belonging to single organ of articulation, as the instrumental precondi- defined imaginations themselves, i.e., an attribute of tion for production of human language and singing contents and by that a neural pattern of activity cor- was established. This innovation among others was responding to conscious experience. supported by translocation and transformation of The neurophysiological findings therefore once the human larynx including better separation of it more take into account just the tight functional rela- from the route of breathing. tion between certain neuronal processes and con- Probably synchronously with the already men- scious events with any regard to the categorical dual- tioned symbol capacity a new type of neuronal in- ity evident in conscious experience and the formation processing mechanisms developed, i.e., demonstrated way and form of its synthesis. In a sim- linear serial processing of excitation patterns. ilar way the highly speculative assumption of the ego- This enabled ordered sequences of single logical feeling part of consciousness as another particular steps of decision–making processes forming the type of neural information processing, e.g., based on proper tool and standard form of human rational- a hypothetical super-imposition by quantum effects ity, logical thinking and communicative interpreta- occurring in neuroaxonal tubuli, only shifts the prob- tion of reality. lem to the level of physicalistic reduction which does In this context the metaorganic products of the not provide a convincing explanation, all the more as cognitive processes will not be described. probably quantum processes are also inherent in in- The phenomena of this realm although they miss formation processing procedures. (DENNET 1991) as shown some items of physical nature are based Concerning the distinctive physiological basis of on the organic structure and intriguing functions of self—sentience (not self consciousness), whose or- the brain. Therefore, due to physical impact in case ganic basis seems to be scientifically enigmatic it of an organic injury, disorder, aging and disease the seems to manifest itself in a faint emotional form and instrumental generation of them becomes dis- only occasionally—e.g., in pain—as a common sen- turbed, distorted or even destroyed: Agnosias, sation. anosognosia, split-brain syndrome etc. are causal With that supposition questions concerning the consequences of certain structural organic defects. nature of the problematic phenomenon of self-sen- The respective alterations pertain to distorted or de- tience in living systems is not yet answered. In “ego- fective configurations of the contents of the con- feeling” too the case of the elliptic situation of a dou- sciousness which—although pathological—never- ble reflection is evident: (a) the transcendental posi- theless are products of perception of reality and do tion as subject versus objects in connection with the not reach the primary ego of consciousness. On the reflection of (b) the immediate inner certitude of ex- other hand organic disorders of the so-called uncon- periencing the subject itself. Concerning empirical scious brain regions, most of them localized in the consciousness this type of a apparently indetermin- deeper brain parts, diencephalon and brainstem, able relation points to GÖDEL’s principle of incom- can alternate the emotional life regarding e.g., self- pleteness, especially to the impossibility to explain feeling and/or manifestations of drives. consciousness on the level of consciousness.

Evolution and Cognition ❘ 154 ❘ 1999, Vol. 5, No. 2 Neurobiological Comments on Empirical Consciousness

Probably the descriptive hypothesis of a quality self-consciousness of human reflection. If this wid- inherent in systems of extreme complexity and au- ened concept of consciousness would prove to be tonomous self-control might prove itself as an ac- correct, consciousness could be seen as an a-priori ceptable metaphor. Thus the quality of life processes in earlier mentioned conjecture general and increasing of an “archepsychic experi- Author’s address knowledge of the nature of ence” consisting of indistinct life would lead to a better ex- Franz Seitelberger, Klinisches Institut für awareness and self-reference planation of the particular Neurologie, Universität Wien, Schwarz- appears as a plausible retro- dual quality of conscious- spanierstr. 97, A-1090 Wien, Austria. spective projection from the ness.

References schaft: Darmstadt. Seitelberger, F. (1987) Wie geschieht Bewußtsein? Psychol. Dennet, D. C. (1991) Consciousness explained. Little Brown: in Österreich 7: 6–19. Boston. Seitelberger, F. (1993) Gehirn, Verhalten und Bewußtsein. Fichte, G. (1845/46) Thatsachen des Bewusstseyns. In: Sämt- In: Philosophia Perennis. Teil I. Peter Lang Verlag: Frank- liche Werke (Edited by J. H. Fichte). De Gruyter: Berlin. furt/M., pp. 368–381. Kant, I. (1922) Kritik der reinen Vernunft. Grossherzog Wil- Seitelberger, F. (1994) Das Bewußtsein. Freie Akademie, helm Ernst Ausgabe. Bd. 3. Insel: Leipzig. Schriftenreihe Bd.16: 29–47. Oeser, E./Seitelberger, F. (1995) Gehirn, Bewußtsein und Wundt, W. (1873/74) Grundzüge der physiologischen Psy- Erkenntnis. 2nd edition. Wissenschaftliche Buchgesell- chologie. Engelmann: Leipzig.

Evolution and Cognition ❘ 155 ❘ 1999, Vol. 5, No. 2 Gunther S. Stent

Are Animals Persons?

HE QUESTION WHETHER ject to natural laws. Abstract T animals are persons “Supernaturalistic” refers or things epitomizes the to phenomena that exist The starting point is the question of personhood in problem of their ethical only in the metaphysical human and non human beings. Referring to KANT, standing, since the moral naturalistic and supernaturalistic phenomena attrib- world of thought and law applies only to per- uted to persons are distinguished, and the role of su- transcend natural laws. sons but not to things pernaturalistic phenomena, especially those in According to this bipartite (GOODMAN 1988; LOMAR- relation to the intersection between biology and eth- classification, all of the SKY 1987; 1992). If the cor- ics, is discussed. properties we attribute to rect answer to the the person’s body are nat- question were: “Animals Key words uralistic phenomena and are mere things, because amenable to study by sci- only human beings are Dualism, KANT, personhood, naturalistic phenome- entific methods, as are (at persons”, it would be non- na, supernaturalistic phenomena, ethics, animal least in principle) some of sensical to speak of “Ani- rights. the properties we attribute mal Rights”. But this is not to the person’s mind, such the correct answer. An animal might be a person and as consciousness, rational thought, intentions and have rights, since in moral praxis (as opposed to feelings. However, we also attribute some other prop- moral theory) the category of person neither includes erties to the mind that are supernaturalistic phenom- automatically all human beings nor excludes auto- ena, which are not amenable to study by scientific matically all non-human beings. For personhood is methods. They include the categorical properties not an objectively inherent trait that some creatures identified by Immanuel KANT (1949) as applicable have and others lack. Instead, it is a subjectively only to persons but not to things, such as uniqueness, bestowed trait with which we endow some creatures sacredness (or being an end in itself), and free will. and withhold from others. Moreover, personhood is According to KANT, persons owe their ethical not an all–or–none trait, which we either wholly standing—having moral duties as well as moral bestow on or wholly withhold from a creature. rights—to their being invested with these supernat- Rather, we bestow it to a greater or lesser degree on an uralistic phenomena of the mind. The ensemble of animal or a human being, with the ethical standing KANTIAN supernaturalistic phenomena used to be of the creature rising with the degree of personhood called the person’s “soul”. Despite the haughty deri- with which we have endowed it. sion of the soul by most modern, down–to–earth, no- nonsense biologists and psychologists as an out- The Soul dated, patently absurd, irrational—at best, poetic— locution, I believe that it remains a helpful, if not The modern view of personhood is based on DES- downright unavoidable concept for any discussion of CARTES’ notion that the person comprises a mind, the intersection of biology and ethics. which is responsible for consciousness, feelings and As KANT showed, belief in the soul, i.e., in a mind rational thought, as well as a body, which provides with supernaturalistic attributes, is not only not irra- for the person’s physical presence and is a kind of tional but its acceptance is actually a prerequisite for automaton. The phenomena that we attribute to the a rationally coherent account of morality. He person’s body and mind can be classified according pointed out that we spend our lives in two metaphys- to whether they are naturalistic or supernaturalistic ically distinct realms of what he called the intellgible (STENT 1998). “Naturalistic” refers to phenomena world, both of which are constructed by our mind. that exist in the physical world of things and are sub- One of them—the amoral realm of things—is con-

Evolution and Cognition ❘ 156 ❘ 1999, Vol. 5, No. 2 Are Animals Persons? structed by our mind’s theoretical reason. In that pensants all over, thought ought to be illegal. So their realm, the relations between things (or objects) are avowed aim was to provide the Assemblée Nationale governed by the causal determination of natural with an objective, scientific criterion for the critical laws. The other realm—the moral realm of persons— stage in prenatal or perinatal human development at is constructed by the mind’s practical reason. In that which personhood begins, i.e., when the biologi- realm, the relations between persons (or subjects) are cally. immature human being acquires its soul. Prior governed by the laws of freedom self-imposed by au- to that stage—so their reasoning went—there could tonomous minds. be no moral objection to abortion. Thus our interpretation of a being’s actions can be One participant, the eminent cytogeneticist, fundamentally different, depending on whether we Michel LEJEUNE—discoverer of the chromosomal ba- encounter it in an amoral context or in a moral con- sis of Down’s Syndrome—held that fertilization of text. In an amoral context—especially in a scientific the human egg is that critical stage, a criterion that setting—we can regard all beings as things, whose would rule out abortion altogether. LEJEUNE, a devout actions are attributable to inflexible natural laws. But Catholic and a leader of the French anti-abortion in a moral context—especially in a social setting—we movement at that time, argued that personhood behave to give special consideration to beings who gins with the diploid zygote and formation of the qualify as persons and whose actions are attributable unique genetic identity of the future individual. to minds with supernaturalistic, autonomous free Other participants proposed various later develop- will. Although our attributions of free will to the soul mental stages as critical, ranging from gastrulation of and of natural causal necessity to the body are obvi- the embryo, through the onset of fetal heart beat or ously incoherent, as rational beings we can no more electrical brain activity, to parturition. Their propos- abandon the idea of moral freedom than we can als would allow abortion at ever-later stages of hu- abandon the idea of amoral causal necessity. Indeed, man development. those rare individuals who have abandoned either of It was obviously futile for the conference to seek a these ideas are usually diagnosed as psychopaths, biological criterion for definition of the critical stage and most of them live in prisons or mental institu- in development at which the human embryo or fetus tions. turns into a person, since the soul is a supernaturalistic concept of KANTIAN practical reason. At the ear- Personhood of Humans liest developmental stages—zygote, blastula, or gas- trula—the embryo is certainly alive and contains a Abortion. The controversy regarding the ethics of human genome which (as LEJEUNE emphasized) will abortion, in particular the controversial thesis of the eventually give rise to a biologically unique person. “right to life” of a human fetus, provides one exam- But the embryo is not yet a person. It is a mere blob ple of our subjective endowment of some beings with of protoplasm, which, despite its covert genetic personhood while withholding it from other beings. uniqueness, is devoid of any overt signs of behavioral I first became aware of the critical role of the subjec- or morphological uniqueness. Sometime later in de- tive assessment of personhood in the context of the velopment, the fetus does assume human. features abortion debate when I attended a conference on that may make it look like a person. But at which “Biology and the Future of Man,” held in the Grand developmental stage the fetus does begin to look that Amphithéatre of the Sorbonne. (GALPÉRINE 1976). As I way is a highly subjective call. For most, though— soon discovered, the hidden agenda of the confer- transhistorically and transculturally—not necessar- ence’s organizers was to create a favorable intellec- ily all people, that stage will be reached no later than tual climate for the legalization of abortion in France, the time of birth. then under consideration by the Assemblée Nationale. The subjectivity of the criterion by which person- The international roster of participants included hood of an immature human creature is actually as- only biologists and biomedical scientists, because the sessed is evident from the records of trials in which organizers had evidently thought that your woolly- abortionists were charged with the murder of a fetus. headed philosophers and theologians had nothing Very often, the prosecution showed photographs of useful to contribute to the difficult moral problem the abortus to the jury, and—not surprisingly—the under discussion. chance of a verdict of “guilty” was the greater the The participants were generally agreed that a per- more advanced was its developmental stage. When son has a right to life and that the intentional termi- asked for the grounds on which they reached their nation of that life is murder, which they, like bien- verdict, the jurors rarely mentioned any of the (ob-

Evolution and Cognition ❘ 157 ❘ 1999, Vol. 5, No. 2 jective) biological criteria put forward at the Sor- ating cowboys, card sharks, bandits and whores. The bonne conference. Rather, most often they said that guests are informed that the town’s inhabitants are they found for “guilty” when the picture evoked the actually mechanical robots, manipulated electroni- (subjective) impression that the abortus had a recog- cally by human handlers in an underground control nizably human face. center. The robots look and behave so realistically It seemed ironic that the participants of the Paris that only the handlers, but not the guests, can tell conference (not excepting even the devout LEJEUNE) them apart from real humans. Since the guests be- addressed the ethics of abortion by resorting to lieve (on hearsay, without objective proof), that the purely bio-scientistic arguments, while facing the people they run into at Westworld are things rather Grand Amphithéatre’s bigger than–life–size statue of than persons, they feel free act out their repressed DESCARTES. I was half-expecting the statue, like that immoral desires. So they have an enjoyable holiday of Don GIOVANNI’s Commendatore, to quicken in an- away from the office, raping, and killing, guilt-free ger and drag the discussants to Hell for not paying and ad libitum. heed to DESCARTES’ admonition that the moral law Racism and Slavery. There is (or at least used to be) applies only to persons and not to machines—not an American saying that “all Chinese waiters looks because only persons have this or that body part that alike”, which, though it may seem like a harmless machines lack, but because they (and only they) have opinion, is, in fact, a deeply racist revelation. Not to souls. recognize the unique individuality of members of the War. Another example of the subjective nature of foreign race makes it possible to exclude them from our endowment of some human beings with person- the category of persons protected by the moral law. hood while withholding it from others is provided by A similar, though much more egregious instance the suspension of the moral law of the sacredness of of the failure to grant full personhood, and hence the person by soldiers in brief war time encounters protection under the moral law, to what are obvi- with an unknown, or even invisible foe. They feel ously fellow human beings is provided by slavery. duty-bound to kill the faceless enemies, who are per- Here too the denial of personhood is usually linked ceived as dangerous things, or vermin, outside the with a failure or refusal to perceive the uniqueness of bounds of morality. But if an opportunity has been individual slaves (who are, in any case, often mem- afforded for the soldier to recognize his enemy as a bers of a foreign race). In this way, the slave masters unique individual, the foe-thing may acquire person- make themselves feel morally righteous in the belief hood and come under the protection of the moral that the beings they cruelly exploit are mere, com- law. monplace, replaceable things, rather than sacred per- This process of repersonalization of a depersonal- sons with unique souls who are ends in themselves. ized enemy was rendered in a poignant scene of the The bestowal or withdrawal of personhood is not 1930 Universal Pictures movie, All Quiet on the West- necessarily an all–or–none matter, however. For in- ern Front. In the dark of night, a German soldier tum- stance, while prior to their emancipation, the African bles into a front-line fox hole already occupied by a slaves in the Southern United States had no rights French soldier. The Frenchman pulls his bayonet and and were not recognized as legal persons by the state, is about to plunge it into the German’s chest when some slaves were invested with partial personhood an exploding grenade lights up the scene. The poilu by their masters. Most of these semi-persons were sees the terrified face of the boche. He recognizes that house slaves, whose uniqueness was appreciated by the other man, hitherto identified generically as an virtue of their close daily contacts with their masters’ enemy by his military uniform and style of helmet, families and with whom they established even some is a unique person and cannot bring himself to kill emotional ties. him. Instead, the poilu offers the boche a cigarette. Human clones. Ever since the rise of molecular and Robots. The 1973 MGM movie, Westworld, pre- cell biology in the 1950s and 1960s, the prospect of sented a further cinematic instantiation of the sub- cloning mammals was held out as an exciting bio- jectivity of our judgment of personhood. The scene technological prospect. Cloning would replace old- is set in a holiday resort in Arizona for (male) million- time, conventional sexual reproduction and its ge- aire guests, where, in Disneyland style, an old-time, netically uncertain outcome with a radically novel Wild West frontier town, complete with bars and approach to procreation that guarantees the genetic bordellos, has been re-created. On arrival, the guests identity of the progeny. This project seemed well exchange their business suits for ranch-hand outfits within the range of possible experimental proce- and mingle with the resort’s staff, who are imperson- dures, by extracting cell nuclei from the somatic tis-

Evolution and Cognition ❘ 158 ❘ 1999, Vol. 5, No. 2 Are Animals Persons? sues of an adult male or female donor animal and still fail to qualify for personhood and lack the ethical transfer them into enucleated eggs provided by a fe- standing claimed on their behalf by the advocates of male host animal. Upon embryological development “Animal Rights” (RADNER/RADNER 1989). But just as of these transgenically manipulated eggs would arise we subjectively depersonalize some humans and a clone of offspring individuals, all of whom would deprive them of their moral status on the basis of be genetically identical to the donor. their perceived lack of a unique soul, so we subjec- The recently claimed success in cloning mammals tively personalize some animals, endow them with a by this procedure after, several decades of failed at- unique soul, and confer moral status on them. tempts (WILMUT et al. 1997) has given rise to great Household Pets. Relatively few kinds of animals are hopes for genetically engineered animal husbandry. likely candidates for elevation to honorary person- But it has also aroused to great fears that this proce- hood. Most kinds of animals belong to invertebrate dure will be applied to humans as well and that we phyla, such as the annelids, mollusks and insects, may be at the threshold of a monstrous Brave New about which it is hard to believe that they have a World populated by hordes of genetically identical mind that generates consciousness and rational people. thought. And even among the vertebrates there is Why is it that whereas I would welcome having only one class—the mammals—that includes some EINSTEIN, BEETHOVEN, Marilyn MONROE and Clark GA- species whose behavior is sufficiently complex that BLE reside in my neighborhood and enjoy running attributing to them a conscious and rational mind into them at the Supermarket or at the local bar, the might cross the threshold of credibility. Yet even idea of living in a town with a hundred of their among those candidate species few specimens attain cloned replicas seems obscene? What is the source of personhood status, because of their lack of opportu- this revulsion? It is that these hundreds of replica nity to be recognized as, a unique individuals by a EINSTEINS, BEETHOVENS, MONROES and GABLES would human being and be granted moral standing by vir- not really be Albert, Ludwig, Marilyn and Clark. They tue of the authority vested in us to confer that honor would merely be CARTESIAN automata in human on them. form, because they could not, each of them, be, en- The most common examples of personalized ani- dowed with a unique soul. I would doubt the human- mals are provided by household pets, such as cats and ity of these Doppelgänger, because I could not tell dogs. They are in a particularly favorable position for them apart and could not conceive of so many differ- elevation to personhood, not least because they were ent, unique minds all inhabiting look-alike bodies bred from their wild ancestor species for the very pur- with identical faces. pose of providing companionship for their human masters. Their masters generally acknowledge the an- Personhood of Animals imals’ individuality, perceive them as unique specimens of their kind, and endow them with a mind Prior to the end of the 18th century, philosophical with supernaturalistic properties, such as being ends discussions regarding personhood did not concern in themselves and having free will. The protection of animals, except as mindless counterexamples to the moral law is extended to the pet, which is given thinking human beings. DESCARTES, like his philo- a personal name and treated with the same consider- sophical predecessors and contemporaries, still con- ation as are the human members of the household. sidered animals as unthinking automata whose Moreover, the concept of moral responsibility tends control of behavior is wholly accounted for by bodily to be extended to the pet as well. The master may functions. The possibility that animals have minds scold, or even punish the pet for soiling the carpet or was not taken seriously until the latter part of the killing a song-bird in the garden, while also praising, 19th century, upon the acceptance of the idea of the or even rewarding it for killing household vermin. evolutionary continuity between animals and humans and the subsequent rise of the discipline that Animal Experimentation came to be known as “ethology”. And yet, even if animals do have minds capable generating conscious- The personhood of animals is at stake too in the eth- ness, rational thoughts, intentions and feelings, it ical struggle about the permissibility of animal exper- does not necessarily follow that their minds include iments. The parties to this dispute generally try to also the supernaturalistic properties such as unique- disqualify their opponents by claiming that they are ness, sacredness, and free will that we attribute to the morally unworthy discussion partners. On the one soul. That is, even if animals have minds, they might hand, the Animal-Rights opponents of animal exper-

Evolution and Cognition ❘ 159 ❘ 1999, Vol. 5, No. 2 imentation usually attribute selfish motives to their mera, the adherents the Animal Rights movement adversaries, even claiming that biomedical scientists point out that it was not all that long ago in the do their experiments mainly to satisfy their sadistic United States that putting black Africans and women yen for torturing animals. Accordingly, they allege on a par with white males was also considered a chi- that one cannot expect that just those researchers mera. Admittedly, this historical comparison does who are most qualified to evaluate the scientific or not provide a logically compelling argument, yet the medical need for animal experimentation would claim that the recognition of Animal Rights is a goal actually meet their ethical responsibility vis a vis the representing moral progress over Biblical tradition creatures they victimize. On the other hand, the bio- cannot be dismissed all that easily. After all, there do medical scientists under attack usually claim that exist Oriental cultures whose moral traditions are they have a highly ethical attitude towards the pro- based on the atheistic ethics of Buddhism, according tection of life in general, show great concern for the to which animals and humans are considered on a welfare of animals in particular, including those that par to a much greater extent than they are in the they use in their experiments, and that the principal West. Since Buddhists don’t believe in God in the professional goal of their whole research effort is pre- first place, they can hardly hold that He created man- cisely to avert pain and suffering of humans as well kind, but not the animals, in His own image. All the as animals. According to them, the Animal Rights same, putting mankind and animals on a moral par movement consists mainly of sentimental fools and is a belief whose general acceptance would bring hypocrites, who, though they have managed to gain about a radical transformation of the Occidental life a disproportionately powerful political influence, style that even the adherents of the. Animal Rights cannot be engaged in any rational discourse. doctrine would be unlikely to welcome. As a practicing neurobiologist, I recognize the un- A comparison of Sweden and the United States is avoidable necessity of the conduct of animal experi- instructive in this connection. On the one hand, the ments, without which it is hard to imagine progress regulations governing animal experimentation are in biomedical research. Yet, I was dismayed to note much more stringent in Sweden than in the United that the arguments put forward in defense of animal States. As far as I know, neurophysiological experi- experiments by my colleagues were easily defeated by ments with higher mammals, not to mention pri- representatives of the Animal Rights movement mates, are almost impossible to carry out in Sweden. skilled in ethical disputation. For instance, many bio- On the other hand, the regulations governing hu- medical scientists concede that, in view of their pro- man experimentation are much more liberal in Swe- fessing a moral duty to protect animals (in fact, life den than in the United States. Thus the Swedish leg- in general), they see a moral problem in our killing islature is apparently much closer to the chimera of of animals to provide us with food and clothing, de- the moral parity of humans and animals than is the stroying them as pests, and vermin, and condemning American. them to extinction in their natural state by ever expanding our living space. Nonetheless, they claim Coda that placing humans and animals on a moral par is a chimera that puts into question the human primacy I give it as my view that a creature—human or ani- embedded in Occidental moral concepts, and thus of mal—has full ethical standing, i.e., has moral duties human existence on the whole. For, as we are told in as well as moral rights, if and only if we attribute to it Genesis 1:26. “And God said, let us make mankind in a mind with the supernaturalistic properties identi- our image, after our likeness. And let them have do- fied by KANT as applicable to persons but not to minion over the fish in the sea, and over the fowl of things. These properties are uniqueness, being an the air, and over the cattle, and over all the earth, and end in itself, and free will, which belong to what used over every creeping thing that creeps upon the to be called “the soul”. This statement is not meant earth.” Thus, within the context of Biblical tradition, to be normative, but intended merely to call atten- animals are not ends in them- tion to the subjective opera- selves, and their use for human tional criteria by which the Author’s address utility presents no moral prob- judgment of personhood is lem. actually made, even by those Gunther S. Stent, Department of Molecular In response to the scientists’ and Cell Biology, University of California, people who may profess other claim that placing humans Berkeley, CA 94708-3200, USA. criteria on religious, biological and animals on a par is a chi- or humanistic grounds.

Evolution and Cognition ❘ 160 ❘ 1999, Vol. 5, No. 2 Are Animals Persons?

References Lomarsky, L. E. (1992) Concept of Person. In: Becker, L.C./ Becker, C. B. (eds) Encyclopedia of Ethics. Vol. II. Garland Galpérine, C. (1976) Biology and the Future of Man. Universi- Publishing: New York, pp. 950–56. ties of Paris: Paris. Radner, D./Radner, M. (1989) Animal Consciousness. Prome- Goodman, M.F. (ed) (1988) What is a Person? Humana Press: theus Books: New York. Clifton, NJ. Stent, G. S. (1998) Epistemic Dualism of Mind and Body. Proc Kant, I. (1949) Critique of Practical Reason. (Translated by L. Am. Phil. Soc. 142 (4): 578–588. W. Beck.) University of Chicago Press: Chicago. Wilmut, I./Schnieke, A. E./McWhir, I./Kind, A. J./Campbell, K. Lomarsky, L. E. (1987) Persons, Rights, and the Moral Commu- H. S. (1997) Viable offspring derived from fetal and adult nity. Oxford University Press: Oxford. mammalian cells. Nature 385: 810–813.

Evolution and Cognition ❘ 161 ❘ 1999, Vol. 5, No. 2 Holk Cruse

Feeling Our Body—The Basis of Cognition?

Introduction Abstract but merely to obtain an understanding of the first To be able to feel some- The paper introduces the distinction between HIP sys- step of the transition thing implies that the tems, i.e., systems with first-person perspective, and from a pure, though pos- system which feels has an NIP systems, i.e., systems not having an internal per- sibly complex, reflex internal aspect or a first- spective. This leads to the question as to what may be machine-like system to a person perspective. Biolo- the conditions which are responsible to make up a system which can have a gists, of course, do not HIP system. subjective experience, deny the existence of an First it is argued that the traditional distinction be- even if this experience internal aspect, but are tween the two epistemological domains, the objective occurs in only a prototyp- trained not to consider and the subjective world, should be replaced by the ical form. the phenomena of the view that the so-called objective world is not a distinct What is meant by cog- first-person perspective domain, but a subset of the subjective world. It is fur- nition? Cognition is not in their scientific work, ther argued that the critical condition to become a HIP an easy term to be used, because they do not system is to have an internal world model the basis of because many different belong to the objective which is a model of the own body. A neural network is definitions can be found. world and are therefore explained which can be used as a manipulable body Nevertheless, there ap- assumed not to be tracta- model. The output of this model is compared with pears to be some agree- ble by scientific methods. psychological findings discussed to be important de- ment such that cognition Therefore neurobiolo- scriptions of the internal perspective as are perspectiv- has to do with feeling, gists when looking at alness and transparency, and with recent thinking, problem solv- brain functions keep observations of patients suffering from brain lesions. ing, usually in connection away from this suspicious This supports the hypothesis that the content of the with conscious percep- field. Philosophers, on world model which includes the model of ourselves is tion and awareness. There what we experience as a first-person perspective. In the other hand, take this are, however, also strict other words, we do not have the experience, we are the phenomenon as a given operational definitions experience. This hypothesis agrees with the idea of basis and investigate the which do not rely on METZINGER (2000) who states that humans “are sys- rich complexity of phe- awareness. MCFARLAND/ tems, which were configured by evolution in such a nomena they find within BÖSSER (1993), for exam- way that they constantly confuse themselves with the this world, but are usually ple, define cognition as content of their phenomenal self-model”. not interested in any pos- manipulation of declara- sible physiological Key words tive knowledge. In order grounding of these phe- to avoid the error to re- nomena. As in this paper First-person perspective, third-person perspective, strict our view on a possi- my goal is to make an world model, body model, perspectivalness. bly too narrow range, I attempt to contribute to would like to leave this close the gap between problem in the following these two worlds, the paper bears the danger of as open as possible. Thus, the term cognition is only being of interest neither for the neurobiologist nor used here in an intuitive and loosely defined way. for the philosopher. This even more, as, when start- One can distinguish between different systems ing to work on this problem, it is indispensable to which are able to control behavior. For example, dissect it into smaller parts. So my goal is not try to cognitive systems are able to control, in particular, understand the first-person perspective in its full intelligent behavior. There are, however, also two richness, i.e., including all aspects of consciousness other types of systems under discussion. These are and self-consciousness (see Introductory Chapter), the reactive, or data-driven systems on the one hand

Evolution and Cognition ❘ 162 ❘ 1999, Vol. 5, No. 2 Feeling Our Body—The Basis of Cognition? and symbol using systems as developed in tradi- Two points are usually mentioned when looking tional artificial intelligence (AI) on the other hand. for reasons why these AI systems lack nominal intel- In the following I shall briefly describe what is meant ligence, too. One is that, in contrast to the reactive by reactive and by symbol using systems. It should, systems mentioned, these systems are not however, be noted that the application of these two “grounded” or “situated” (e.g., PFEIFER 1996), they terms does not necessarily imply that there is always have no own body and therefore cannot react to the a clear distinction to be made between both types of world. This has led to postulate that, in order to gain systems. intelligent systems, they have to be built as real ro- Nowadays robots, and possibly also simple ani- bots which have not only sensory but also motor mals are typical reactive systems, sometimes also contact with the physical world (BROOKS 1991). dubbed reflex machines. The BRAITENBERG vehicles (BRAITENBERG 1984) are often mentioned examples First-person perspective and third- which, in spite of the extremely simple internal wiring connecting their sensor and motor units, can person perspective show quite complex behavior. However, in both, the Another argument is that these AI systems, as the field of autonomous robots, and the field of neuroe- reactive systems, are no cognitive systems because thology, more and more complicated “animats” or they do not have awareness. They lack the capabil- animals of this type are investigated. These reactive ity of having subjective experience. As the property systems show a rich behavioral spectrum ranging of a system to have feelings seems to be one prereq- from simple reflexes to complex navigation and op- uisite for the system to be a cognitive one, this portunistic behavior when following different goals. should be treated further. What does it mean to Such systems can even show displacement behavior have subjective experiences? and can solve complex motor control tasks in under- Apparently, the world can be divided into two constrained situations, i.e., in situations where deci- types of systems. There are systems, as definitely hu- sions are required (e.g., MAES 1991). Could such sys- mans, and most probably many other animals, tems be called intelligent? It has often been said that which can have awareness, or subjective experience these systems are attributed with intelligence in the or internal perspective. For short they might be adverbial sense but not in the nominal sense (LANZ called HIP systems (Having an Internal Perspective). 2000). They may behave intelligently, or behave as There are other systems, as most probably a pencil, if they were intelligent, but they have no “real“, a stone, or a clockwork, which do not have this abil- nominal intelligence in the sense of creating new ity. I will name them NIP systems (Not having an ideas, i.e., showing new solutions to given problems. Internal Perspective). They do not feel. Of course, we The second type mentioned is the symbol using cannot be definitely sure about the latter in any case, system such as expert systems or general problem but usually people agree on that and, to simplify the solvers developed in artificial intelligence. In con- following discussion, I want to take this as given. In trast to the “analog” reactive systems, traditional AI many cases, for example for an insect, we cannot systems cannot easily deal with blurred information. decide whether it belongs to the NIP systems acting In these systems input signals are first transformed like a reflex machine, as a clockwork, or whether it into symbols and these symbols are then treated fur- belongs to the second type, the HIP systems, and has ther. Given the symbolic information, known ratio- feelings. Such a decision is the more difficult, the nal methods can be applied such as the rules of logic more the brain under view differs from a human or specific mathematical tools. The general goal of brain. AI was to develop systems of nominal intelligence. However, one part of the problem might be However, although being quite successful in many solved in the following way. Philosophers usually cases, AI systems were found not to be able to solve distinguish between two epistemological domains, many “simple”, daily-life problems which can al- which they describe as the external perspective and ready be solved by young children or by many ani- the internal perspective, or the third-person per- mals. These AI systems are, the reader may permit spective and the first-person perspective, respec- this biological view, like parasites. They “feed” on tively. The third-person perspective concerns our symbols provided by humans, but they cannot “sur- daily intellectual activities dealing with “objective” vive” without symbols, i.e., in a natural world. They facts. They are called objective because, in principle, are specialists, apparently not equipped with a gen- every human has the possibility of mentally treating eral intelligence. these objects which means that information about

Evolution and Cognition ❘ 163 ❘ 1999, Vol. 5, No. 2 Holk Cruse these “facts” can be transmitted between persons. ability to experience. So, somewhere during ontog- Examples are the number of apples lying in a dish, eny this change has to occur. Furthermore, our ev- or the name of the color of a green pencil, or entities eryday change from (dreamless) sleep to the awake used when performing a mathematical calculation state shows such a transition. This means that even such as adding two numbers. These activities are our brain can, to some extent deliberately, switch objective or public because they can be looked at by between these states. The assumption that there are any outside observer. This “external perspective” two types of systems, some with and others without not only concerns the behavior of the complete sys- a first-person perspective leads to the consideration tem, but, in principle, also includes the possibility that having the first-person perspective is a system of looking inside the brain of a subject and measur- property—sometimes called an emergent prop- ing all the interesting neuronal activities, for exam- erty—that suddenly occurs if some specific condi- ple, when the subject judges the color of the pencil. tions are fulfilled. It would be of interest to know In this way, all details of the subject seeing “green” what are these conditions? could be determined. (Thus, looking inside in this Before I treat this question, I will expand on the way still belongs to the external perspective). In gen- distinction between the two domains determined eral, this view could also be applied to a correspond- by the external and the internal view. I would like ingly programmed robot. All these data form the to argue that this separation is appropriate to de- basis of the external perspective, the view of the scribe the historical development and that it is also third person or outside observer. On the other hand, appropriate as a didactic method to introduce into the subject seeing green does not see these neuronal the problems, but that it is actually misleading when activities, but experiences seeing green. This is called one tries to understand the problems raised here. the subjective quality of experience. In contrast to This separation was probably introduced by PLATO the items of the external perspective, the content of and worked out most stringently by DESCARTES. It this subjective or first-person perspective is only ac- can still be found today in most texts referring to cessible to the person himself or herself. Therefore, these problems. Even in texts tending to oppose DES- this is also called the private view in contrast to the CARTES’ view, this separation still exists in a more or public view. Nobody other than myself can judge less hidden form (e.g., DAMASIO 1994). Unhappy how I see green. The difference between the external with DESCARTES’ dualistic approach, philosophers perspective and the internal perspective may behave introduced the notion of ontological identity come even clearer in the example of pain. Again we explained with the metaphor of “these are two sides can consider all neuronal activities that occur when of a coin”. This metaphor is definitely helpful in the subject’s skin is stimulated, for example by a physics in trying to combine two items belonging to needle. One might, in principle, even look at one’s the “objective world“, for example the wave–cor- own action potentials, if oneself is the subject of this puscular dichotomy. However, the distinction be- experiment. But the experience when considering tween the first-person perspective and the third-per- all these neuronal activities is completely different son perspective is not cogently described by this from the pain one experiences at this moment. metaphor. Looking at the coin requires the capabil- Thus, self observation tells us that there exist sys- ity to look from the outside. This is just what we tems, namely humans, that can experience an inter- cannot do from the internal viewpoint. We do not nal perspective. As has already been mentioned have the experience, we are the experience. I thus above, intuition tells us that there are other systems, propose the following, alternative view which like a stone or a simple machine (including some might be called epistemological solipsism. In this clever nowadays robots) that may not have such an view, it is impossible to draw a clear line separating internal perspective. the two worlds in such a way that an item falls into Interestingly, there are at least three points where either one or the other domain. Rather, everything one can find a transition from a system without first- of our so-called objective or third-person perspec- person perspective, a NIP system, to a HIP system, tive is embedded in our subjective experiences. Ba- i.e., a system having first-person perspective. At least sically, we have subjective experiences, and those once during the evolution, anywhere between sim- we call objective are subjective experiences that can ple algae and human beings, this change must have be more easily separated from each other, therefore happened. Such a transition also occurs during the being measurable in many cases and more easily ontogeny of each individual human. A, say, four-cell transmittable between different persons. Most of- human embryo most probably does not yet have the ten, but not necessarily, these entities are denoted

Evolution and Cognition ❘ 164 ❘ 1999, Vol. 5, No. 2 Feeling Our Body—The Basis of Cognition? by symbols like words. Nevertheless, non of these experiences are non-subjective. We do not have ac- 7 B person 1 cess to the “real objective world” (provided such a world exists). Even an “objective” notion like “color is an electromagnetic wave” belongs to the subjective domain, the first-person perspective, because 7 we do not know what an electromagnetic wave re- person 2 B ally is. To say this by another, similar example: We are usually trained to formulate a sentence like: “light of the wavelength of 450 nm appears to be blue”. This formulation implies the first part of the B person 3 sentence, namely “light of the wavelength of 450 7 mn” to represent the hard fact, the second part to represent a much weaker, subjective aspect. How- Figure 1: The large ellipses represent the subjective world of ever, the statement could also be formulated in the three persons. Their objective worlds form subsets of these opposite way: “My experience of ‘blue’ can, in the worlds (broken lines). These objective worlds contain such en- framework of physics, be described by using the tities (e.g., a square, a number, a letter) which can be easily metaphor of a wave with a given length”. Clearly the communicated. first part, the subjective experience, can be experienced directly. But, as mentioned, also the content of the second part is represented in the domain of Simple neural network models subjective experience. At the same time, however, it belongs to a mental system consisting of declarative Let us now come back to the basic questions of how entities or symbols which we have developed during it is possible that some systems have the first-person our cultural evolution. Therefore, instead of speak- perspective which CHALMERS (1995) has termed the ing of two ontologically separate domains, a better really hard problem. What are the conditions allow- metaphor would be to consider the set of “objective” ing humans and, as I believe, “higher” animals, to phenomena as a part, a subset of the subjective experience an internal perspective (“have a soul”) world we experience. This epistemological solipsism and thus being able to become cognitive systems means that so-called “objective” phenomena are whereas other systems, like simple animals, dead subjective phenomena with specific properties that objects (e.g., a ball rolling in a bowl towards the low- do not belong to all subjective phenomena: objec- est point, i.e., behaving as if it had a “goal”), or tive phenomena are easily transmittable between robots, do not experience the first-person perspec- persons, because they can be connected to more or tive. I assume that this property depends exclusively less clearly defined symbols. Phenomena lacking on the architecture of the neuronal system, in par- these properties are those commonly called subjec- ticular on a special kind of world model constructed tive. In other words, what we call “objective” might within the neuronal system. To explain this, I be called “intersubjective” instead. This is illustrated would like to introduce three types of systems (Fig. in Fig. 1. The large ellipses represent the subjective 2), which can be distinguished according to their world of three persons. The broken lines mark sub- internal architecture (the concepts used in the field sets of these worlds containing such entities which of Artificial Neural Networks will be used to describe can be easily separated from each other and can this architecture). The first type contains a feed-for- therefore easily be described and communicated be- ward system, the two other types contain recurrent tween persons (this is illustrated by the double ar- networks and will be explained below. rows). In this way, these entities appear to be ele- A typical feed-forward network consists of an in- ments of an objective world. The difference between put (sensory) layer, one or more hidden layers and an objective and subjective items might correspond to output (motor) layer. Such a reactive system is sym- the observation that verbal expressions and memo- bolized in the lower section of Fig. 2. Placed in an rized items allow a coarser distinction than our at- environment and having the appropriate internal tentional contents do. To give an example: we are connections, such a system can exhibit quite sophis- able to perceive more subtle differences in color ticated behaviors (see the BRAITENBERG vehicles and shadings than we are able to memorize or to describe their more recent and advanced relatives). The inter- verbally. nal connections of this network may be interpreted

Evolution and Cognition ❘ 165 ❘ 1999, Vol. 5, No. 2 Holk Cruse

e.g., determined by the hormonal state) stimuli, and, in this respect, are also, however, more indirectly, data-driven. These two types of systems will be described as containing non-manipulable models, in contrast to a third type of world model to be ex- manipulable world model plained below. Both types of systems described up to now are intuitively assumed to have no first-person perspective. As a further step, one could now imagine expanding the system such that another type of internal simple world model world model exists that can be to a large degree uncoupled from the actual sensory input and the motor output (Fig. 2, upper part). For example, this is the case when a movement is planned in order to decide whether this movement, when executed, would lead reactive system to the desired result. This movement may, for instance, concern the grasping of an apple lying on the table possibly within reach, or, to take a more abstract example, the movement of a figure in a chess game. Another example is given by the task to men- world tally rotate a visually given object (SHEPARD/METZLER 1971). Such an internal world model may be called Figure 2: A three layered system acting onto and being influ- manipulable, because after the task is given as input, enced by the world. The lower part consists of a pure reactive it can be further uncoupled from actual sensory in- system. The section in the middle shows a system with a sim- put and can “play around in its virtual internal ple, fixed world model as, e.g., a circadian clock. The upper part denotes a system which contains a “manipulable” world world”. The results this model proposes after finish- model, i.e., a model which can be uncoupled from the motor ing its search can be judged by given evaluation cri- output. teria. Depending on these values, a decision might be made whether or not to actually perform the pro- as comprising an implicit world model (of that part posed activity. of the world that is of importance for the system), As long as this world model represents only infor- because in some more or less indirect way it repre- mation concerning the outer world and not proper- sents the properties of the world and the appropriate ties of the system itself, one still might not be pre- reactions. This corresponds to the internal represen- pared to attribute a first-person perspective to it. tations mentioned by J. PROUST (1999). However, this (Traditional, non-situated AI systems may be consid- “world model” is static. The reactive or data-driven ered examples of such cases). This situation might mode of this system does not allow for explicit pre- change if the system under view, first, has not only dictions in time. To make a prediction possible, ad- sensory but also motor contact with the physical ditional networks with internal recurrent connec- world and, second, some of the physical properties tions are necessary. Such recurrent networks allow of the system itself are also embedded into the inter- for time-dependent states that are no longer directly nal world model. To explain this, I would like to dependent on the actual input. In this way, they can restrict the following discussion to the experiences comprise world models that can be used by the sys- of the subject’s own body including planning and tem to predict sensory input before the correspond- performing movements, for example reaching ing change in the environment actually occurs and movements with the arm and the hand. I do this to enable the system to prepare the necessary actions because I assume that the body model that animals accordingly. A simple example is a central pattern including humans use when planning or performing generator used to control time-dependent behavior a movement may be the very basis of the phyloge- (e.g., our circadian clock). This not only concerns the netic evolution and possibly also the ontogenetic control of the motor output, but also the time-depen- development of a manipulable internal model. This dent control of sensory input. Usually these “world view has been proposed in a similar way by CONDIL- models” are confined to a given behavioral context LAC (1798). If we had only the possibility to receive elicited by a definite complex of sensory (or internal, acoustic or chemical stimuli, according to CONDIL-

Evolution and Cognition ❘ 166 ❘ 1999, Vol. 5, No. 2 Feeling Our Body—The Basis of Cognition?

LAC, we had not the possibility to distinguish be- forward model tween the self and the external world. For tactile y g xt stimulation the situation is quite different. During yt movements of the own body self contact may occur at b bt which leads to simultaneous tactile stimulation of gt a x different body sites. As CONDILLAC states this makes

task specification gbayx a fundamental difference between the tactile sense and the other senses. Such correlated simultaneous stimulations within one sensory modality are only possible in the tactile sense. The information from reset the mechano–sensory input, including the position senses, can be used to form an internal representa- xs tion of the own body allowing the distinction be- ys as tween the own body and the external world (active bs body moments may also play a role). This body gs model may then be expanded to include objects of sensory input the world outside the body. For example, if we close the eyes and hold a stick in the hand to grope the premotor layer environment, we seem to feel the tip of the stick. motor output Therefore, this stick seems to be a part of our body. Later on, other objects like a cup standing on the Figure 3: A schematic drawing showing the body model of a three joint arm (forward model plus bold recurrent connec- table but not (yet) connected to the body might be tions). The arm is depicted in the inset, the shoulder forming included into the world model. the origin of the coordinate system x/y. The output of this recurrent network is assumed to correspond to the phenomenon of subjective experience. The sensory input (joint angles The body model at shoulder, elbow and wrist, as, bs, gs, and the end point position xs and ys) is compared to the output of the body model How should a system having a manipulable model (subtraction symbols marked by the ellipse). of its own body be constructed? To explain the hypothesis more concretely, in the following an extremely basic model will be described. three-dimensional object. Here we deal with a corre- STEINKÜHLER et al. (1995) proposed a model which sponding problem in the motor domain. The model can deal with the problem of inverse kinematics, i.e. consists of a recurrent network that relaxes to adopt that can determine a geometrical arrangement of a a stable state corresponding to a geometrically cor- multilimbed arm to grasp a given object. This net- rect solution, even when the input does not fully work, called MMC net, provides a holistic internal constrain the solution1. This network, called body model of the kinematics of the body. Although the model in the following, can be used as a model for model can be extended to describe the geometry of the inverse kinematics by which for any visually a complete body (CRUSE et al. 1996), I will concen- given target, for example, an appropriate arm con- trate here on the example of moving a three-joint figuration can be found. planar arm (Fig. 3). The control of such an arm The MMC model will be shown in a simplified already involves the general problem of controlling form (for a detailed description of the quantitative a redundant system, because the end point of the model see STEINKÜHLER/CRUSE 1998). This presenta- planar arm is determined by two coordinates, x and tion permits to an explanation of the body model as y, whereas more than two independent values, a combination of a recurrent system (Fig. 3, bold namely the three joint angles, a, b, and g, have to be lines) and a simple nonlinear multi-layered feed-for- controlled (see inset Fig. 3). This “redundant” sys- ward system (Fig. 3, forward model). The forward tem thus has one extra degree of freedom which model uses as internal variables the three angle val- means that the model has to decide between differ- ues a, b, and g of shoulder, elbow, and wrist as well ent possible solutions. The necessity to cope with as the x/y position of the arm end point (see inset in underconstrained situations is an important prob- Fig. 3). These five values are then fed back to the lem. In the visual domain, for example, this prob- input to form a recurrent system (Fig. 3, bold lines). lem has to be solved when we interpret the two- Three sensory variables are called as, bs, and gs, de- dimensional line drawing of a Necker cube as a scribing the actual joint angles. This input layer may

Evolution and Cognition ❘ 167 ❘ 1999, Vol. 5, No. 2 Holk Cruse be regarded as to form a mechano-sensory layer. Fur- the case of mere observation of this action, a strong thermore, the position of the end point of the arm inhibitory influence has been found which selec- is given by CARTESIAN coordinates xs,ys provided by tively inhibits those motoneurons in the spinal the visual system. These sensory values can be used chord which were involved in the observed action to reset the recurrent connections of the body model execution. to adopt the values of the real world. This is indicated An open question is how the body model can by the open arrowheads which, when activated, sup- adapt to changes of body geometry, e.g., during press the recurrent signal (Fig. 3, reset). The output growth, or to sensory drift. At first sight, this may of the body model is compared with the sensory in- appear to be a mere technical problem, but the solu- put (marked by the ellipse in Fig. 3) to provide the tion is of interest here. This could be done in the motor output. following way. When, in a relaxed, static state, the How does the system work? When all internal difference between the output of the body model, variables, i.e., the angle values and the end point e.g., joint angles, and the actual sensory input is position coordinates, match, i.e., form a geometri- zero, the body model is appropriate (the symbols cally possible configuration, the recurrent network representing the subtractions are marked by the el- is in a relaxed state. Its “harmony value” (RUMEL- lipse in Fig. 3). In this case no changes of the internal HART/MCCLELLAND 1986) is high. Now assume that a connections of the forward model, its synaptic task is specified to move the arm to a new position weights, are necessary. The greater the differences, xt /yt. This corresponds to an input from the box the greater the necessary changes of the weights. Us- “task specification” in Fig. 3 which leads to a sup- ing these differences as error signals, the body model pression (open arrowheads) of the recurrent x/y val- could easily be updated using the classical backprop- ues. Now the harmony of the system is low because agation algorithm in accordance to a proposal by the actual angle values and the new x/y values do not KAWATO/GOMI (1992). In Fig. 3 these error signals are match. However, the network relaxes to a new state indicated by dashed lines. Apart from enabling us to with angle values fulfilling the external constraints. learn the properties of the body model, this compar- After having reached high harmony again, the out- ison between the output of the body model and the put could be used such that the premotor values are actual sensory input might also have another effect. given to the motor output to control the actual When the manipulable model proposes a new arm movement. By disconnecting the premotor layer position, the latter is usually different from the ac- from the motor output, the body model can, how- tual one represented by sensory input. Functionally, ever, also be used to “plan” the movement. It could, this difference may be used to control a correspond- in other words, be considered as the neural basis for ing movement. Furthermore, as explained below, it the production of an imagined movement (which may be used to direct attention. has been called Bewegungsphantasie by GEHLEN 1971). The appearance of the ﬁrst-person This system can be considered to provide a solution for the problem of sensor fusion, occurring perspective when different sensory input represent the same When applying the third-person perspective, this value. As an example, in the case of the three-joint may be a sufficient description of the system. At this arm, the position of the end-effector could be given point, however, let us also consider the first-person by the visual system (the coordinates xs and ys), but perspective. When planning a movement we can also by the mechanoreceptive input (the angles a, b, experience a body image, we “see”—or can imag- and g). These different values are integrated within ine—the position of our virtual arm moving around. this holistic system. The network concatenates both Why does this experience occur? Seen by an outside the visual and mechanoreceptively given spaces to observer, the situation can be described such that form a common, dynamic representation. Within two neuronal layers exist that both represent the the recurrent model, a holistic system, no distinc- subject’s own body. One is represented by the cur- tion between the sensory values and the motor val- rent activities of the actual sensory layers, the other ues can be made (CRUSE et al. 1998). Interestingly, it by the output of the forward model. The first may be has been reported by RIZZOLATTI/ARBIB (1998) that used to specify a particular behavior by direct con- premotor areas are activated whether an individual nections to the motor output not shown in Fig. 3. is about to perform an action or observes another Furthermore, they may be used to reset the body individual performing an action. Furthermore, in model and to update its weights. The second layer

Evolution and Cognition ❘ 168 ❘ 1999, Vol. 5, No. 2 Feeling Our Body—The Basis of Cognition? may be used to plan and/or to control motor behav- Even in awake state, we are not aware of all parts of ior. Following METZINGER (2000), I speculate that it is the body model in a given moment. There are “at- the output of this body model which yields the con- tentional islands” which can move to different re- dition for the system to experience a first-person gions. The attentional level may be increased in do- perspective, i.e., to make up a HIP system (see also mains where the difference between the sensory CRUSE 1979; for an excellent and very elaborated input and the model output is large. This means that description of this view, see METZINGER 1993). METZ- the sensory input is different from the expectation2. INGER (2000) claims “that—ontologically speaking— When undergoing mental relaxation, these islands no such things as selves exist in the world. What may become flattened and this exercise is quite often actually exists is a special kind of self-models and followed by the transition to sleep. A spectacular their contents, and this content makes us believe case of resetting the body model has been reported that we actually do have or are identical to a self.” In by RAMACHANDRAN et al. (1995). In a patient suffering other words, the hypothesis is that it is the occur- from phantom sensation in the amputated left arm, rence of this body model that makes the system feel. a lively subjective experience of moving this arm Is there a possibility to test this hypothesis? The could be elicited when the intact right arm was both answer is that there is probably no direct possibility actively moved and seen in a mirror such that the to test it. The only immediate way to prove the feasi- mirror image of the right arm appeared at the posi- bility of the hypothesis is to collect phenomenolog- tion where the left arm would have appeared if it had ical observations on the first-person perspective and been intact. The visual input appeared to be suffi- see whether corresponding properties of the model cient to activate the model of the arm. can be found. Several such observations will be Our hypothesis may be further supported by the explained. As mentioned, according to our hypoth- following observation. In cases of asomatognosia an esis only the output of the body model can be recog- accidental destruction of brain structures may lead nized, but not the sensory representations as such. to the result that a part of the body, e.g., the left leg, The latter are not cognitively available. This agrees can be visually recognized, but is not regarded as with the introspective observation that the con- belonging to the own body. The patient considers it scious I does not have the impression of viewing a as a foreign body. At the same time, the leg cannot movie when looking at the sensory input (like a voluntarily be moved, although the muscles and pe- homunculus would), but to be in the world. We ripheral nerves are intact. Following the above hy- “look through the screen”. This is called semantic pothesis, this result may be interpreted in such a way transparency (METZINGER 1993). The second, related that the neuronal basis of the body model of the leg property is the following. The I is always in the cen- had been destroyed by the accident. As a conse- ter of the experience. Philosophers call this perspec- quence, neither a voluntary movement nor a feeling tivalness, because we always see the world from our was possible. SACKS (1984) reported a self-observa- own perspective (METZINGER 1993). According to tion in which the neurological cause was unclear. In METZINGER (2000), perspectivalness means “the this case, motor control was regained and this lucky existence of a single, coherent and temporally stable experience was paralleled by the reoccurrence of the model of reality, which is representationally cen- subjective experience of the leg. As both properties tered around a single, coherent and temporally reoccurred together, it seems to be probable that extended phenomenal subject.” This also fits with both phenomena have a common basis. Most people the model, because the basis of the subjective expe- may have experienced the following less critical sit- rience is the data of the system’s own body model. uation: their arm has “fallen asleep”,—yet the arm This, by the way, may provide our impression of the, position that is seen and the one that is felt may not subjectively, constant I, although, seen from the coincide. There is also another observation made by third-person perspective, we actually do change a SACKS which fits the properties of the model. At the lot during our lifetime. Similarly, this could explain beginning of the reoccurrence of the subjective ex- the fact that after waking up in the morning we feel perience of his leg, it seemed to be either very long to be the same person as the day before. Probably we (many meters) or very short (some centimeters). As would have this impression even if we were dramat- time passed these “oscillations” became less distinct ically changed overnight. and finally came closer to the real leg length. Quali- Falling asleep corresponds to loosing the first-per- tatively, this corresponds to the dynamic behavior son perspective. According to our hypothesis this of the MMC net when recovering from a disturbance corresponds to the model output becoming zero. (CRUSE/STEINKÜHLER 1993).

Evolution and Cognition ❘ 169 ❘ 1999, Vol. 5, No. 2 Holk Cruse

Taken together, these observations suggest the The disadvantage of both examples is—as it is the following hypothesis: a system comprises a HIP sys- case for the above mentioned metaphor of the two tem, i.e., has a first-person perspective, when (a) it sides of a coin—that they describe the transition be- contains a manipulable internal world model that tween entities which both belong to the domain of includes properties of the system’s own body which the third-person perspective. Therefore, these exam- (b) can be used to compare the (“virtual”) data of this ples may not convince anybody taking a dualistic model with those provided by the “real” data from view. The following example might help to give a the sensory input (and the memory, see below). As a better intuition of a transition from the third-person weaker form of the hypothesis, one can assume that perspective to the first-person perspective. Imagine these conditions are necessary for the first-person you join for the first time a group of persons who perspective to appear as an emergent property. The already have well established social connections. At stronger version assumes that these conditions are the beginning, you look at this group and the per- also sufficient. In any case, our hypothesis suggests sonal relations between the group members in a kind that only such items have the ability to become ex- of third-person perspective. After becoming ac- perienced that are represented in the manipulable quainted with the different people, you develop an world model. In this way, such a system includes all internal model of the group and its internal rela- the elements proposed earlier by the different defi- tions, and, as you develop own relations to group nitions of cognition. This, in particular, means that members, you gradually experience the transition of the operational definition of MCFARLAND/BÖSSER is an individual becoming a member of a group. When justified insofar, as the latter concentrates on the doing so, you develop a kind of first-person perspec- functional aspect. Following our hypothesis, the tive for the group. neuronal system underlying the body model has a functional meaning, whereas the property of having Expansions of the model a first-person perspective comes for free as a kind of epiphenomenon. This means that, the fact that we This model does, of course, not describe all neces- experience an internal perspective has, as such, no sary elements of a system which is able to control functional meaning, and therefore has not necessar- behavior. It, if at all, describes an atom, a basic ily to be included in the definition. Another conse- element of a cognitive system. Therefore, several quence, following from this hypothesis is that, if it expansions could be envisaged. First, there are was correct, also artificial systems could be con- reasons to believe that not all movements are structed such as to have first-person perspective, i.e., controlled by such a body model. On the to become HIP systems. contrary, for urgent actions there are presumably I do not see a way to make it intuitively clear that, also many parallel systems for direct specifica- under the conditions described, a first-person per- tions of actions. These already mentioned reactive spective necessarily develops. However, this is similar systems, however, are normally not cognitively in other cases where unexpected system properties penetrable and are therefore called non-manipu- emerge as, for example, the property of resonance lable (see Fig. 2). They form direct connections frequency in an electronic circuit. After having ob- between sensors and motor layers, possibly served and described the resonance phenomena, in- including hidden layers. These connections are cluding constructing a pendulum-like mental model, not shown in Fig. 3. The alien limb phenomenon we have the impression of having understood the might be explained by such influences. In these property of resonance. Another, possibly even better cases a limb of the patient, for example the left example concerns the occurrence of the phenome- hand, moves, but the patient has the impression non of life. It was a serious and intensively debated that this movement is not influenced by himself, question between Vitalists and Mechanists at the be- but by some external agent. This means that the ginning of our century, how nonliving matter could body model appears to be intact because the hand be transformed into living matter. The Vitalists ar- is perceived as the own limb. Voluntary control of gued that a special force, termed vis vitalis, is re- movement, however, is not possible. As a hypo- quired. After intensive investigations, we now have thetical explanation one could assume that in this developed a mental model showing that a special col- case the connection between body model and lection and arrangement of molecules has the prop- motor output is disturbed and hand movement is erty of living. We do not need a vis vitalis. Liveliness only controlled by such direct, non-cognitive is understood as an emergent property. influences.

Evolution and Cognition ❘ 170 ❘ 1999, Vol. 5, No. 2 Feeling Our Body—The Basis of Cognition?

Second, the body model itself has to be expanded, eas and the parietal lobe. Imagination of a visual too. The exact border dividing objects belonging to stimulus, the picture of a tree, for example, activates the own body and belonging to the environment is the visual areas. During calculation parts of the pre- not strictly fixed. A hammer in the hand can be ex- frontal area and of the temporal lobe are activated, perienced in the body image as an extension of the and so on. Therefore, the model that is explained body. Also, a car driver may feel his body extended here for a special case only, namely the representa- to the outer margin of the car’s body. One may even tion and the control of arm position, may be applied feel kind of sick when the car does not work well. in a similar way to other items and other cortical Depending on cultural influences, these borderlines areas. Thus, these world models might form the basis may even include other people; an extreme exam- of memory. Various of such models might be con- ple, for members of the western culture, may be the nected such that the output of one model (“a con- experience of labor-pains by males in the case of cept”) forms the input to another model by replac- couvade. ing the sensory input of the latter. For example, Furthermore, as mentioned earlier, the body instead of the sensory input describing the actual model is considered as being part of the world arm position the angle values of an arm position model. From this point of view, the model described memorized in another context might be used. up to now which only has knowledge concerning its own body including the possible expansions men- Discussion tioned, comprises the most primitive version of a cognitive system. However, to be applicable for con- Whereas the considerations of ROTH (this volume) trolling the movement to reach an object, the model and FLOHR (1991; 1996) concern the question in has to be enlarged by the capability of imbedding which part of our brain the conscious events take external objects which occur in the workspace. In place and whereas FLOHR in particular asks for corre- the course of further development, more distant ob- lated changes on the molecular level, and, further- jects could be integrated to permit, for example, ori- more, convincingly argues that under some entation. Already at this stage of development, this conditions a first-person perspective necessarily will model contains internal representations or “con- occur, I address the question of how the local mech- cepts” (cf. C. ALLEN, this volume) of these objects, anisms might work and how the transition between although application of language is not yet consid- the third-person perspective and the first-person ered. The invention of spoken words by humans in- perspective could be understood. The proposal is creased the manipulability of these virtual objects, made that it is the output of the world model which allows for finer distinction between items, and im- corresponds to the first-person perspective. proves the possibility of communicating between Are we really prepared to attribute a first-person subjects. In this way, the internal world model start- perspective to an artificial system? There are two ing with representations of only its own body, is ex- “opposite” ways to come to a wrong conclusion. We panded to contain near and distant extracorporeal may consider even a simple robot in an animistic objects as well as abstract entities. At least on this way, i.e., as having a plan or a will, at least, as long level, items stored in the manipulable model can be as we observe only the behavior of the system (think called “objective”.3 of the arguments you have with your computer). Be- Apart from this functional view, one might also ing social animals we apparently are constructed ask in which part of the brain this model could be that way. So when applying common sense, we may realized. As mentioned by ROTH (this volume), con- be inclined to falsely assume the existence of a scious experiences appear to be correlated with neu- “soul”, an internal view even in the case of an ex- ronal activities in the whole cortex (see also FLOHR tremely simple system. On the other hand, the 1996), but this may not be true for the primary areas. reader may not be intuitively prepared to attribute a Furthermore, the content of first-person perspective to the consciousness is correlated simple system shown in Fig. 3. with the locus of that activa- Author’s address Indeed, most probably each tion. As shown by brain imag- concrete circuit, even if it re- Holk Cruse, Fakultät für Biologie, Univer- ing techniques, imagination ally described a HIP system, sität Bielefeld, Postfach 100 131, D-33501 of arm movements, for exam- would probably lead to the in- Bielefeld, Germany ple, are correlated with activ- Email: [email protected] tuition of a system not having ity in parts of the premotor ar- a first-person perspective. We

Evolution and Cognition ❘ 171 ❘ 1999, Vol. 5, No. 2 Holk Cruse tend to consider any hardwired system as possibly which we also experience during our daily life, in being complicated, but still of the clockwork type. particular, when we change from sleep to the awake Therefore, both observations show that intuition is state. a problematic advisor. The MMC net for the arm can be regarded as a One reason why our intuition might have these “general purpose” representation of the Euclidean problems could be that we compare the possible “in- space in which the arm can move. The angle values ner life” of this primitive model to our rich subjec- (together with the fixed segment lengths) represent tive experiences. However, this direct comparison the workspace of the arm. The x–y coordinates de- may not be justified. Such a simple system may in- scribe a larger, visually given space. The recurrent deed have crossed the principal threshold, but the network concatenates both spaces to form one liveliness of its first-person perspective might actu- unique space. In this way the MMC net corresponds ally be very weak. The liveliness might become more to a mental representation as proposed by J. PROUST intense and the feeling of the system may become (1999). more and more real, the more different sensory modalities contribute to the description of a given situ- Acknowledgments ation. This is supported by the observation of patients with sensory deficiencies who report that their I would like to thank the Wissenschaftskolleg zu feeling seems to be less real. In this way, a gradual Berlin for providing the possibility to work on this change is conceivable, ranging from a very weak subject and Dr. Thomas METZINGER for very helpful first-person perspective to full awareness, a change discussions.

Notes 2 According to GOMBRICH (1960), only those sensory inputs trigger our attention which are not expected. 1 The underlying idea is that several geometrical relation- 3 Another possible expansion, not considered here, might be ships are used to calculate the same value (e.g., the angle of that the signals to be compared with the output of the the elbow joint) several times in parallel. The final output world model may not only come directly from the sensors, is then obtained by calculation of the mean value of these but also from other memories which then may replace the multiple computations (MMC). This is then fed back for the sensory input. next iteration. Any special constraint of the variables (e.g., joint limits) can be introduced.

References From animals to animats 4. MIT Press: Cambridge MA, pp. 84–93. Allen, C. (1999) this volume. Cruse, H./Steinkühler, U. (1993) Solution of the direct and Braitenberg, V. (1984) Vehicles: experiments in synthetic inverse kinematic problem by a unique algorithm using psychology. MIT Press: Cambridge MA. the mean of multiple computation method. Biol. Cybern. Brooks R. A. (1991) Intelligence without reason. Proc. Int. 69: 345–351. Joint Conf. on Artificial Intelligence. pp. 569–595. Damasio (1994) Descartes’ error. Emotion, Reason and the Chalmers, D. J. (1995) The puzzle of conscious experience. human brain. G. P. Putnam’s Son: New York. Sci. Am. : Dec.: 62–68. Flohr, H. (1991) Brain Processes and Phenomenal Conscious- Condillac, E. B. de (1798/1983) Abhandlungen über die ness. Theory and Psychology 1: 245–262. Empfindungen. Felix Meiner Verlag: Hamburg. Flohr, H. (1996) Ignorabimus? In: Roth, G./W. Prinz, W. (eds) Cruse, H. (1979) Modellvorstellungen zu Bewußtseins- Kopf-Arbeit—Gehirnfunktionen und kognitive Leistun- vorgängen. Naturw. Rundschau 32: 45–54. gen. Spektrum Akademischer Verlag: Heidelberg, Berlin, Cruse, H./Steinkühler, U./Burkamp, C. (1998) MMC—a re- Oxford. current neural network which can be used as manipulable Gehlen, A. (1971) Der Mensch. Athenäum: Frankfurt. body model. In: Pfeifer,R./Blumberg, B./Meyer, J.-A./Wil- Gombrich, E. H. (1960) Art and Illusion: a study in the psy- son, S. W. (eds) From animals to animats 5. MIT Press: chology of pictorial representation. Princeton University Cambridge MA, pp. 381–389 Press: Princeton. Cruse, H./Bartling, C./Dean, J./Kindermann, T./Schmitz, J./ Kawato, M./Gomi, H. (1992) The cerebellum and VOR/OKR Schumm, M./Wagner, H. (1996) Coordination in a six- learning model. Trends in Neurosciences 15: 455–453. legged walking system. Simple solutions to complex prob- Lanz, P. (2000) The concept of intelligence in psychology lems by exploitation of physical properties. In: Maes, P./ and philosophy. In: Cruse, H./Dean, J./Ritter, H. (eds) Pre- Mataric, M. J./Meyer, J.-A./Pollack, J./Wilson, S. W. (eds) rational Intelligence: Interdisciplinary perspectives on the

Evolution and Cognition ❘ 172 ❘ 1999, Vol. 5, No. 2 Feeling Our Body—The Basis of Cognition?

behavior of natural and artificial systems. Kluwer Press (in Ramachandran, V. S./Rogers-Ramachandran, D./Cobb, S. press). (1995) Touching the phantom limb. Nature 377: 489–90. Maes, P. (1991) A bottom-up mechanisms for behavior selec- Rizzolatti, G, Arbib, M. A. (1998) Language within our grasp. tion in an artificial creature. In: Meyer, J.-A., Wilson, S. Trends in Neurosciences 21: 188–194 (eds) From animals to animats. MIT Press: Cambridge MA. Roth, G. (1999) this volume. McFarland, D./Bösser, Th (1993) Intelligent behavior in ani- Rumelhart, D. E./McClelland, J. L. (1986) Parallel distributed mals and robots. MIT Press: Cambridge, MA. Processing. Explorations in the microstructure of cogni- Metzinger, T. (1993) Subjekt und Selbstmodell. F. Schön- tion, Vol. 1. MIT Press: Cambridge MA. ingh: Paderborn, München, Wien, Zürich. Sacks, O. (1984) A leg to stand on. Gerald Duckworth: Lon- Metzinger, T. (2000) The subjectivity of subjective experi- don. ence: A representationalist analysis of the first-person per- Shepard, R. N./Metzler, J. (1971) Mental rotation of three-di- spective. In: Metzinger, T. (ed) Neural Correlates of mensional objects. Science 171: 701–703 Consciousness: Empirical and Conceptual Questions. Steinkühler, U./Beyn, W.-J./Cruse, H. (1995) A simplified Cambridge MA: MIT Press. MMC model for the control of an arm with redundant de- Pfeifer, R. (1996) Building “Fungus Eaters”: design principles grees of freedom. Neural Processing Letters 2: 11–15. of autonomous agents. In: Maes, P./Mataric, M. J./Meyer, Steinkühler, U./Cruse, H. (1998) A holistic model for an inter- J.-A./Pollack, J./Wilson, S. W. (eds) From animals to ani- nal representation to control the movement of a manipu- mats 4. MIT Press: Cambridge MA. lator with redundant degrees of freedom. Biol. Cybernetics Proust, J. (1999) Mind, space and objectivity in non-human 79: 457–466. animals. Erkenntnis 51: 41–58.

Evolution and Cognition ❘ 173 ❘ 1999, Vol. 5, No. 2 Hans-Johann Glock

Animal Minds: Conceptual Problems

1. Conceptual Abstract Among the contested problems and the concepts that are impor- problem of This article discusses the problems which concepts tant to the issue of animal concepts pose for the attribution of thoughts to animals. It first minds, the concept of a locates these problems within a range of other issues concept occupies an im- Concepts have a two-fold concerning animal minds (sct. 1). It then distinguish- portant place. More spe- significance for animal es three positions: DAVIDSONIAN lingualism denies cifically, concepts seem to minds. Like most philo- that non-linguistic animals have any thoughts or con- pose a distinct obstacle to sophically interesting cepts; mentalism maintains that their thoughts differ those theories that ascribe topics, animal mentality from ours only in degree; an intermediate position to animals a wealth of poses problems of a dis- maintains that animals can have thoughts of a sim- mental properties and ca- tinctively conceptual ple kind (sct. 2). I opt for an intermediate position on pacities. On the one hand, kind. What mentality, if the basis of considering those arguments for lingual- many of these phenom- any, one can attribute to ism which trade on the connection between thoughts ena seem to involve or non-human animals and concepts: the argument from the intensional na- presuppose the posses- (henceforth “animals”) ture of thought (sct. 3); the idea that thoughts involve sion of concepts, but on depends not just on concepts (scts. 4–6); the argument from the logical the other hand, concepts empirical findings connections created by concepts (sct. 7). I conclude and conceptual capacities (whether observations in that there is a kind of perceptual thought that does not seem to be highly sophis- the field or experiments require concepts and that concept-possession is not ticated features closely in the laboratory) and sci- tied to language, but to the capacity of making dis- connected to language, entific theories, but also criminations that are subject to normative assess- and hence beyond the on what one makes of ment. At the same time, even the ascription of simple ken of animals. In this es- heavily contested con- thoughts to animals employs a rich idiom with con- say I want to discuss both cepts like that of a mind, ceptual connections that go beyond the phenomena to conceptual problems in the which it is applied (sct. 8). of thought, of conscious- field of animal minds, and ness, of behavior, etc. The the problem of concepts for Key words philosophical task in this animal minds. area consists not in col- Animal Minds, concepts, normativity, DAVIDSON. Do at least some ani- lecting new empirical mals have minds which data about animal behav- are comparable to those of ior, its neurological causes or its evolutionary ori- humans? The question is both complex and vexed. gins, but in clarifying what it is to possess various It involves many different problems that have been mental properties, and hence in clarifying under treated from a variety of methodological perspec- what conditions such properties can be ascribed to tives, in subjects ranging from evolutionary biology organisms. This is not to say that one needs a cast- and neurophysiology through ethology, psychology iron precise definition of these properties in advance and linguistics to the philosophy of mind and lan- of empirical theory-building. But it is to say that such guage. Particularly baneful has been the intrusion of theory-building must be accompanied by concep- moral concerns. It is perfectly legitimate to discuss tual reflection on the provisional understanding of what implications the possession or lack of mental- mental concepts that informs specific lines of ity should have for our treatment of animals. What research, methods and conclusions (see ALLEN/ is illegitimate is to tailor one’s views of animal minds BEKOFF 1996, pp236–7). to suit one’s ethical (or unethical) prejudices. Alas,

Evolution and Cognition ❘ 174 ❘ 1999, Vol. 5, No. 2 Animal Minds: Conceptual Problems

in many cases it is obvious that verdicts on animal to insist that loving MALCOLM X or intending to minds are motivated neither by empirical observa- climb a tree are in the final analysis attitudes towards tions nor by conceptual clarifications, but by a desire propositions. In any event, the capacity for thought to affirm or deny that animals are worthy of various is a precondition for the capacity of reasoning. In degrees of moral and legal consideration. both theoretical and practical reasoning, one moves As one leaves behind the murky waters of applied from one or more thoughts, the premises, to another ethics and animal welfare, other complications thought, the conclusion. Reasoning for its part must loom. The concept of a mind may be vague, but it is be distinguished from intelligence. Roughly speak- certainly wide-ranging. In dealing with the question ing, intelligence is the ability to solve hitherto unen- of animal minds, we must therefore specify what countered problems in a flexible way, notably one mental phenomena and capacities are at issue. One which is not predetermined by the genetic outfit of possibility here is to simply provide a list, e.g., sen- the species (similarly DUPRÉ 1996, pp329–30). Intel- sation, perception, memory, mental imagery, belief, ligence in this general sense covers diverse phenom- emotion, desire, intention, consciousness, self-con- ena ranging from problem-solving through trial and sciousness, awareness of other minds, reason, etc. error, as with the Cappuchin monkeys in the trap But it may also be possible to proceed in a more sys- tube task, to the kind of insight and foresight dis- tematic fashion. played by chimpanzees in the fashioning of tools For one thing, we must distinguish cognitive ca- (see TOMASSELLO/CALL 1997, pp10–11, ch. 3). But pacities, namely capacities to gain information or even in the latter case, it is an open question whether knowledge, from conative capacities, that is, capac- reasoning from premises to conclusions is involved ities to have evaluative attitudes of pursuit or avoid- (see sct. 6). ance. The popular term “animal cognition” is mis- Because of my interest in concepts, I shall focus leading in that it seems to refer exclusively to the on animal thought, addressing the issues of sensa- former, when in fact most of the debate under that tion, perception, intelligence and reason only in so title is as much concerned with animal desires and far as they bear on the theme of concepts and intentions as it is with animal belief. thought. That vertebrates have sensations and are Both cognitive and conative states are what phi- capable of perceiving their environment should be losophers call intentional; that is to say, they have a obvious, and so is that many of them, especially particular object or content. One believes or knows higher mammals, are capable of intelligent behavior that something is the case, one desires that some- of various degrees. But only in the case of thought is thing be the case, one intends to do something. In there any prima facie reason for believing that con- cases in which this content is expressed by a that- cepts must be in play. Conversely, the capacity to clause, it is common to use the (in my view mislead- reason clearly presupposes conceptual capacities, ing) term “propositional attitudes”. However, there but only because reasoning involves thoughts. are other mental phenomena which have no such Even within the area of animal thought as here object or content, whether propositional or other. understood, there are several issues. From DESCARTES This holds not just for sensations like pains or tickles, onwards, the main philosophical argument against but also for relatively complex phenomena such as crediting animals with thoughts and concepts has moods. One of the problems with concepts like ex- been that both require language. However, recent perience and consciousness is that they seem to research in cognitive ethology suggests that chim- cover a variety of phenomena ranging from simple panzees and bonobos are capable of acquiring ele- sensations through perception to full-blown propo- mentary linguistic skills (SAVAGE-RUMBAUGH/LEWIN sitional attitudes. It is one thing to be conscious in 1994; SAVAGE-RUMBAUGH/SHANKER/TAYLOR 1998). the sense of being capable of experiencing pain, These findings have been disputed by CHOMSKIAN which may simply be a mark of all sentient life. It is linguists (PINKER 1994, ch.11), who complain in par- quite another thing to be capable of perceiving cer- ticular that these apes lack the syntactical capacities tain features of one’s environment, and possibly yet characteristic of language. So one question is another thing to be capable of believing that such– whether some animals possess language. If that and–such is the case. question is answered in the affirmative, then several I shall refer to these intentional states as thoughts arguments against animal thought simply do not ap- rather than propositional attitudes. It is unclear ply to these cases (see STEPHAN 1999, pp88–90). whether they are attitudes towards entities (see sct. The other question, which is my exclusive con- 4); and even if they are, it would be sheer dogmatism cern here, is whether non-linguistic creatures are ca-

Evolution and Cognition ❘ 175 ❘ 1999, Vol. 5, No. 2 Hans-Johann Glock pable of thought. This keeps out signing apes, but approximate to our mental life. For one thing, the brings in non-linguistic humans. However, with the extent to which biological evolution can involve exception of occasional side-glances, I shall focus on genuine leaps is controversial, as the debate over the non-linguistic animals, partly because non-linguistic possibility of punctuated equilibria shows (see humans raise problems of an entirely different kind. GOULD/ELDRIDGE 1977) For another, even if natura For example, it is unclear to what extent our attribu- non facit saltus, continuity along lineages of evolution of thoughts to pre-linguistic children is based tionary development has no implications for the on an anticipation of their later linguistic skills. At mental capacities of the animals around us. Al- the same time, I shall invoke not just actual forms of though it is probable that our closest evolutionary animal behavior, but also forms of behavior that a ancestors without language shared many of our non-linguistic creature could engage in. For it is this other mental capacities, these ancestors are extinct; possibility which is crucial to the conceptual ques- and there is no guarantee that the biologically clos- tion of whether our mental concepts preclude ab ini- est extant species is mentally close to us. If all verte- tio the possibility of non-linguistic animal thought. brates except homo sapiens had been exterminated by a wayward meteorite, it would be absurd to con- 2. Options and perspectives clude that starfish and sea cucumbers must be mentally close to us (see PINKER 1994, p346). At long last, we have a question which may shed It so happens that our closest extant relatives, the light on the role of concepts for animal minds with- chimpanzees, share 98% of our DNA. However, it out being absurdly general. Can animals without does not follow that they share 98% of our mental language have thoughts? For most of this century, life, simply because small biochemical differences in there has been a tendency to assume that any posi- genotype may lead to significant differences in phe- tive answer must be based on anthropomorphism, notype. In fact, this already holds at the level of neu- an illegitimate projection of human features onto rophysiology. Although the brains of chimpanzees animals. However, with the rise of cognitive psy- are made of the same matter, they are significantly chology and ethology the tide has turned. It has smaller than those of humans (on average, 400ccm been shown, for example, that the blanket charge of to 1400ccm), even if body-size is taken into consid- anthropomorphism is both diffuse and ill-founded eration. In any event, however, our mental vocabu- (FISHER 1996). lary captures neither genetic nor neurophysiological Alas, at present there is an opposite danger. Cur- differences, but differences in the kinds of behav- rent debates about the mind are increasingly domi- ioral and perceptual capacities we humans are inter- nated by evolutionary theory and by naturalism, the ested in. To that extent, our mental concepts them- idea that philosophy has no distinctive contribution selves may be anthropocentric; yet it does not follow to make over and above that of the special sciences. that it is anthropocentric to insist that these con- In this climate it appears that opponents of animal cepts preclude application to non-linguistic crea- thought are misguided “human exceptionalists” tures (see DAVIDSON 1985, p473). Of course, some who ignore the “continuity across species” that has naturalistic philosophers have suggested that we been discovered by evolutionary biology and neuro- should replace our mental vocabulary with a more physiology (JAMIESON 1998). In this vein, various bi- scientific jargon based on neurophysiology. But ological principles of continuity have been invoked whatever the merits of that proposal, it will not solve to show that the mental differences between hu- the problem that fuels the debate about animal mans and animals can only be a matter of degree minds, namely whether it is legitimate to apply our (CRISP 1996). From this perspective, any attempt to current mental concepts to animals. draw qualitative distinctions between humans and As regards that question, one can distinguish not animals appears deplorably anthropocentric and out just proponents and opponents, but a whole spec- of touch with natural science. trum of views. At one end we find lingualists like This blanket charge of anthropocentrism is no DAVIDSON, who deny that non-linguistic animals more warranted than the opposing charge of anthro- have any thoughts. The other, mentalist end is oc- pomorphism. There is no gainsaying the empirical cupied by empiricists like HUME, who think that the fact that there is both biological (in particular, ge- thoughts of animals differ from those of humans netic and neurophysiological) similarity and evolu- only in degree, due to their different perceptual in- tionary continuity between us and certain non-lin- puts. Oysters don’t have thoughts about bicycles, guistic animals. But it does not follow that they must simply because they cannot perceive bicycles. In one

Evolution and Cognition ❘ 176 ❘ 1999, Vol. 5, No. 2 Animal Minds: Conceptual Problems sense it is also occupied by those cognitive represen- These days, the harsh charge of nonsense is much tationalists who explain even simple animal behav- less popular among analytic philosophers than it ior by reference to a rich variety of complex thoughts used to be. But even those who are prepared to ac- and calculations, except that these thoughts are held cept that some thought ascriptions to animals are to be in a language of thought, not in a public lan- neither conceptually incoherent nor empirically guage. In addition to these two extremes, however, false often insist that they cannot be taken literally, there is an intermediate position, adopted by a coa- but must be regarded as figurative, metaphorical or lition (rare, some might say) of common sense and secondary. In a similar vein, Bede RUNDLE has sug- WITTGENSTEIN. It holds that animals are capable of gested that ascribing thoughts to animals is a mere having thoughts of a simple kind, namely those that description or redescription of behavior, whereas can be expressed in non-linguistic behavior. ascribing thoughts to humans can be a genuine ex- A second complication arises because of the vari- planatory hypothesis, to be confirmed by what the ous dimensions of the concept of thought. Even if subject is prepared to divulge (1997, ch. 4). On ei- we leave aside thought as the capacity for reasoning ther view, to say that an animal believes that p may and confine ourselves to intentional states, we have be no more than a façon de parler. to distinguish between two parameters, namely the This stance is illustrated by DAVIDSON’s reaction kind of state on the one hand, the kind of “content” to the following tale from Norman MALCOLM. on the other. One question is what intentional verbs “Suppose our dog is chasing the neighbor’s cat. can be applied to animals; another question is what The latter runs full tilt toward the oak tree, but sud- that-clauses, singular terms or infinitives can follow denly swerves at the last moment and disappears up these intentional verbs. Concerning the first param- a nearby maple. The dog doesn’t see this maneuver eter, it is possible to grant that a dog can know, be- and on arriving at the oak tree he rears up on his lieve or see that p, but to deny that it can think or hind feet, paws at the trunk as if trying to scale it, hope that p. Concerning the second, WITTGENSTEIN and barks excitedly into the branches above. We famously suggested that a dog can believe that its who observe this whole episode from a window say, master is at the door, but not that its master will “He thinks that the cat went up that oak tree” return the day after tomorrow (1967, p174; see also (1972–73, p13). RUNDLE 1997, ch. 3; SAVIGNY 1995). MALCOLM claims that we would be right to say Finally, there are diverse views on the status we this, and DAVIDSON acknowledges that it is prima should accord to ascriptions of thoughts to animals. facie plausible. Nevertheless, he insists that strictly It is possible to regard the ascriptions of some types speaking MALCOLM’s dog cannot believe anything, of thoughts to some species of animals as empirically because he does not possess a language (see 1985, false. This is the case, for example, when ethologists pp474–6, 478; 1984, p155). like SEYFARTH/CHENEY (1996, pp340–3) use observa- An immediate objection to DAVIDSON is this: an- tions to deny that vervet monkeys have a “theory of imals must be capable of having thoughts because mind”, that is, beliefs about the beliefs of conspecif- we have no better way of explaining and predicting ics. It is not easy to see how one could regard the their behavior than by attributing thoughts to ascription of all types of thoughts to all species of them (e.g., BENNETT 1976, §§7–8; FODOR 1975, ch. animals as empirically false. In any event, however, 1). According to DAVIDSON, this provides a prag- most opponents of animal thought do not take this matic justification for our attribution of thoughts line. They often regard such attributions not as false, to animals, but it does not show that animals can but as suffering from a more basic defect, namely have thoughts. In attributing thoughts to animals, that of being meaningless, nonsensical or based on we merely treat them as if they were capable of act- a category-mistake (see FISHER 1996, pp4–8). This is ing for reasons (beliefs and desires), just as one part and parcel of the aforementioned fact that dis- might explain the movements of a heat-seeking putes about animal minds are often rooted not in missile by ascribing to it the desire to destroy an disagreement about the empirical evidence, but in a airplane. In this way, “we can continue to explain diverse understanding of the concepts involved. If the behavior of speechless creatures by attributing the concept of thought is such that it precludes ap- propositional attitudes to them while at the same plication to non-linguistic creatures, nothing could time recognizing that such creatures do not actually count as evidence that some animals have thoughts. have propositional attitudes. We will be bound to Ascribing thoughts to animals would then make no acknowledge that we are applying a pattern of ex- more sense than ascribing a color to a number. planation that is far stronger than the observed be-

Evolution and Cognition ❘ 177 ❘ 1999, Vol. 5, No. 2 Hans-Johann Glock havior requires, and to which the observed behav- ple. For one thing, one cannot make sense of the ior is not subtle enough to give point” (1985, notion of a belief as a private attitude completely pp477–8). detached from behavior and its explanation (DAVID- This account treats thought-attributions to ani- SON 1984, p170; 1985, p476). For another, although mals as useful fictions. However, DAVIDSON himself thoughts can be more or less definite, we can ascribe concedes that animal behavior is much closer to a specified thought that p to a creature a only if some- human behavior than the movements of heat-seek- thing counts as a thinking that p rather than that q ing missiles, and that we know of no better way of (see GLOCK 1997, pp166–8). Finally, even if attribut- explaining the behavior of animals than by attrib- ing beliefs or concepts to creatures which are totally uting thoughts to them. But these concessions in- incapable of manifesting them is not strictly mean- vite an objection DAVIDSON ignores. ingless, it is vacuous and cannot be part of a proper We regard attributing thoughts to animals not behavioral science. just as convenient, as he would have it, but as entirely This third-person perspective contrasts with a justified. For, unlike attributing desires and beliefs mentalist approach that posits mental faculties and to complex missiles, such attributions are not based processes which are unreflected in behavior and lie on technological ignorance but on a biological insight, forever beyond human consciousness. I shall not at- namely that the life and behavior of animals shows tempt to defend the third-person approach against them to have both wants and perceptual capacities. this alternative here.1 But it is important to note that DAVIDSON might reply that the alleged insight is such a third-person perspective is not tantamount to merely an illusion of ARISTOTELIAN folk-biology, behaviorism: thoughts can be ascribed on the basis since animal behavior could be fully explained by of behavior, without therefore being reducible to be- reference to physiological processes, if our knowl- havioral dispositions (DAVIDSON 1985, p476). Fur- edge of physiology were sufficiently advanced. thermore, the third person perspective is adopted not However, this invites the question of why human just by common folks, but also by cognitive ethology behavior should resist such explanation. And a lin- in its scientific investigations of animal mentality. gualist who responds that human behavior resists To be sure, many cognitive ethologists make free mechanical explanations because it rests on use of the notion of a mental representation, espe- thoughts would be begging the question at issue. cially with respect to concepts. One reason is that Nevertheless, there may be a kernel of truth in they regard this as an anti-dote to the behaviorist DAVIDSON’s suggestion that the pattern of explana- perspective which repudiates mental predicates and tion we employ with respect to animal behavior in describes animal behavior exclusively in terms of some sense outstrips the explanandum, because it stimulus and response. But in contrast to mentalist is originally tailored to the explanation of the more philosophers (e.g., CARRUTHERS 1989), their postula- complex behavior of linguistic creatures like our- tion of mental representations remains tied to the selves. In this article I want to bring out that kernel explanation of animal behavior. Furthermore, cir- of truth. At the same time, I shall resist DAVIDSON’s cumspect practitioners stress that this notion of a lingualism. Instead, I opt for a version of the inter- mental representation is a modest one, which does mediate position, both with respect to the kinds of not require the imputation of “pictures in the head” thoughts one can ascribe to animals, and with re- or of symbols in a “language of thought” (TOMA- spect to the status of such descriptions. We can as- SELLO/CALL 1997, pp7–12). Instead, the main pur- cribe thoughts and concepts to animals, but these pose of the terminology seems to indicate that ani- are restricted to a simple kind. For only simple mal cognition goes beyond immediate sense– thoughts can be identifiably manifested in non-lin- perception.2 For this reason, in most contexts talk of guistic behavior. Furthermore, although the ascrip- mental representations can be replaced by talk about tion of thoughts and concepts to animals can be higher cognitive and mental capacities, such as the more than a useful fiction, it is incongruous in that ability to remember, to reason with foresight, etc. As the rich mental idiom we employ has conceptual we shall see (sct. 6), such abilities are importantly connections that go beyond the phenomena to different from behavioral dispositions. Whether the which it is applied. causal explanation of those abilities refers to physi- I shall argue for this position from a third-person cal tokens of computational types in the brain, as perspective. That is to say, I shall not appeal to phe- cognitive representationalists have it, or whether it nomena—whether mental or neurophysiological— involves other factors, as connectionists and many that cannot be manifested in behavior even in princi- neuroscientists maintain, is a separate question.

Evolution and Cognition ❘ 178 ❘ 1999, Vol. 5, No. 2 Animal Minds: Conceptual Problems

Moreover, it is a question that can and should be left believes that Tully was Roman”). In the case of ani- be open when the issue is whether animals have mals, by contrast, substitution of co-referential thoughts or concepts. expressions often leads from attributions which we In any event, to talk about abilities rather than commonly regard as true to attributions which are representations avoids the suggestion that humans absurd or unintelligible. The oak tree that the cat and animals alike relate to the world not directly, went up also happens to be the oldest tree in sight through exercising their cognitive and conative ca- and the same tree the cat went up last time the dog pacities, but through “inner” intermediaries. In my chased it. But does MALCOLM’s dog believe that the view, this shift of perspective is especially apposite cat went up the oldest tree in sight, or the one it in the case of concepts. There is a mentalist tradition, went up last time? Equally, a dog can know that its which regards concepts as representations in the master is at the door. But does it also know that the minds or brains of individuals, and a PLATONIST tra- president of the bank is at the door? “We have no dition, according to which they are building-blocks real idea how to settle, or make sense of, these ques- of abstract entities. The former view has problems in tions” (DAVIDSON 1984, p163; see also CHATER/ doing justice to the objective nature of concepts, the HEYES 1994). The reason is that the dog can think fact that they can be shared between subjects which neither that its master is the president, nor that he could not share private mental ideas and need not is not. share patterns of electro–chemical activity in the One response to this failure of intensionality is to brain.3 The latter view has difficulties in explaining hold that in the sentence the role concepts play in thought and action. Both (1) The dog thinks that the cat went up that oak tree types of problems are avoided by a third position, the expression “that oak tree” occurs transpar- going back to ARISTOTLE and KANT, which regards ently (in QUINE’s terminology). Accordingly, (1) is concepts as principles or rules of classification and paraphrased so as to avoid problems of intensional- explains the possession of concepts as the possession ity, e.g., as of classificatory abilities. (1’) The dog thinks, with respect to that oak tree, In this essay, I cannot justify either my animad- that the cat went up it. versions against inner representations or my prefer- But as DAVIDSON (1985, pp474–5) points out, this ence for the notion of abilities that can be mani- response presupposes that “the de re description fested in behavior. My aim is more modest. I want to picks out an object the believer could somehow pick show that even from such a third person perspective, out”. If MALCOLM’s dog could not distinguish the the various connections between thoughts and con- oak tree from among other objects (e.g., the pine cepts point in the direction not of lingualism, but of tree or the garden fence), we might still causally ex- an intermediate position. The three connections plain his behavior by reference to the oak tree, just that threaten the possibility of animal thought have as we explain the convulsions of an oyster by refer- featured in several authors, but are especially prom- ence to its being pricked with a needle. But (1’) inent in the work of DAVIDSON, who is the most im- would no longer be appropriate. For de re construc- portant current lingualist.4 I shall discuss in turn: tions like “with respect to” or “of” require an ana- B the intensional nature of thought (sct. 3); phoric referent in the subsequent content-clause, an B the claim that thoughts must involve concepts “it” which the now disabled dog could not distin- (scts. 4–6); guish from other things. B the holistic connections between thoughts (sct. 7). However, it remains an open question whether this requirement might not be met by non-linguis- 3. The intensional nature of thoughts tic creatures on account of their possessing certain discriminatory capacities. The dog believes some- One problem with attributing thoughts to animals thing of the tree “under some description”, namely is that without verbal responses we cannot make one that expresses those features by which the dog the fine distinctions between different thoughts recognizes the tree and distinguishes it from other (beliefs, desires) expressed in the same non-verbal objects. This is the point at which concepts come in. behavior. Thought-attributions to humans create If the dog is to be capable of recognizing the tree by intensional contexts: if we substitute co-referential certain features, it seems that it must have the con- terms within the content-clause, this may lead cept of a tree. It needs not only to distinguish trees from a true attribution (e.g., “Sarah believes that from other objects, but to recognize them by certain Cicero was Roman”) to a false one (e.g., “Sarah features. If this is granted, it might even provide for

Evolution and Cognition ❘ 179 ❘ 1999, Vol. 5, No. 2 Hans-Johann Glock a non-linguistic analogue of intensionality, at least amount to one and the same thing, and both are in simple cases. Animals cannot know objects by confined to language-users (DUMMETT 1993, chs. different descriptions, but they might recognize 12–3; DAVIDSON 1997, pp24–5; 1999, pp7–8). them by different features, and they might fail to This raises two problems. Firstly, can animals pos- recognize that these are features of one and the same sess concepts at all, and, if so, what kind of concepts? object. Thus it is conceivable that MALCOLM’s dog Secondly, if they cannot have concepts of any kind, reacts in one way to a man with heavy foot-steps does that really preclude them from having thoughts being at the door, in another to his master being at or beliefs? the door, because he has not yet realized that the One will have to answer the second question in heavy stepper is his master. In that case it may be the affirmative if one accepts the mentalist idea that true that he believes that the heavy stepper is at the thoughts are occurrences in the mind or brain, or the door, but not that his master is at the door, or vice PLATONIST idea that they are abstract entities which versa. have concepts as their components. On either ac- But according to lingualists, crediting the dog count, one cannot have or grasp the thought with- with the concept of a tree makes no more sense than out having or grasping its constituent concepts. crediting him with beliefs about trees. The dog can- However, the PLATONIST picture has no explanatory not believe of an object that it is a tree, “unless we value. It transposes the relation of spatial part and suppose the dog has many general beliefs about whole from an area where it is perfectly clear—ma- trees: that they are growing things, that they need terial objects—onto an area where spatial concepts soil and water, that they have leaves or needles, that ex hypothesis make no sense. At least in its modern they burn. There is no fixed list of things someone version, the mentalist picture looks very concrete: with the concept of a tree must believe, but without concepts are temporal stages of physical processes in many general beliefs there would be no reason to the brain. But even if identifiable neurological pro- identify a belief as a belief about a tree, much less an cesses correspond to having parts of a thought as oak tree” (DAVIDSON 1985, p475). opposed to having a whole thought (something Two lingualist objections emerge here: a general which has been doubted on both empirical and con- one concerning the holistic connections between ceptual grounds), these are only stages of the think- thoughts created by concepts, which will be dis- ing of a thought, not of its content. cussed in section 6, and a more specific one concern- Furthermore, DAVIDSON himself has maintained ing concepts to which I turn now. Consequently, that “propositional attitudes” are not relations be- the force of the argument from intensionality de- tween a subject and an abstract or mental phenome- pends on the arguments from concepts and from non, but rather modifications of a person (1994, holism. The original argument questions whether p232). In my view, he is right. Although that-clauses there is a that-clause which would capture the con- are grammatically speaking noun-phrases and can tent of an animal’s thought. This question now function as accusatives, they no more refer to a gen- gives way to the question of whether such clauses uine object than noun-phrases like “everything” or would not be implying too much conceptual so- “the past” (see GLOCK 1997a). If this is correct, for a phistication, and to the question whether they to believe that p, a need not stand in a relation to an would not drag in too many other thoughts. object (a proposition) which would involve standing in a relation to components of that object (concepts). 4. Thoughts and concepts The building-block model according to which small entities—concepts—combine to form large en- One prominent obstacle obstacle which concepts tities—thoughts—is not a truism but a problematic create for animal thought is this. The thoughts we metaphor. What seems to give content to that meta- ascribe to animals in common parlance involve phor is exclusively the fact that the linguistic expres- concepts with which the animal cannot be credited. sions of thoughts—namely sentences—have compo- Attributing thoughts to animals on the basis of nents—namely words. But as lingualists like non-linguistic behavior is misguided, since these DUMMETT and DAVIDSON themselves have stressed, at thoughts involve concepts which themselves can- the linguistic level the building-block model is prob- not be attributed on such a basis, because they lematic. We do not construct the senses of sentences require general beliefs which could only be mani- out of the meanings of words; rather, we assign fested through language. Accordingly, concept- meanings to words by noting their role within sen- possession and the ability to have thoughts tences. For this reason, DAVIDSON’s approach to hu-

Evolution and Cognition ❘ 180 ❘ 1999, Vol. 5, No. 2 Animal Minds: Conceptual Problems man belief is “holophrastic”: we ascribe beliefs to lin- ceptual experience—or to a conceptual, linguistic guistic creatures on the basis of their assenting to one like full-blown thought. sentences as a whole (1984, pp4, 22, 220–5; 1997, In this vein, MALCOLM suggests that while the dog p25). can “believe” that the cat went up the oak tree, only The possibility he ignores is that rejecting the humans can “have the thought” that it went up the building-block picture in the linguistic case invites oak tree. Similarly, DUMMETT maintains that while an analogous move in the case of animals, a holodox- humans can have thoughts consisting of concepts, astic approach that starts out from the whole belief. animals have mere “protothoughts” consisting of Concepts are not the building-blocks of thoughts, spatial representations. But this type of distinction they are abstractions from thoughts, because, as seems to count against ascribing one and the same KANT noted, their sole function is to be used in judge- belief to humans and animals. It suggests that “Both ments. According to such an approach, what matters Sarah and the dog believe that p” is not so much a is precisely a kind of “modification”: if a creature can falsehood as a zeugma, like “Both the exam and the be correct or mistaken as to how things are, it can chair were hard”. For “Sarah believes that p” comes have beliefs. Although the sentences we use in as- out as “Sarah has the thought that p” while “The dog cribing thoughts have components, our ascriptions believes that p” comes out as “The dog has the pro- are not based on a prior ascription of these compo- tothought that p”. nents. Instead, they are based on the subject mani- In my view, however, we can grant that there are festing certain perceptual capacities, attitudes and important differences between the beliefs of concep- emotions. tual and those of non-conceptual creatures, yet resist In the non-linguistic case, these manifestations the pressure towards postulating distinct objects and will obviously not include assent to sentences. But hence distinct attitudes. A certain disparity between they will include forms of behavior, postures and fa- the terms used in a belief report and those that could cial expressions which higher animals share with hu- be used by the subject is present even in the linguis- man beings. When we say that MALCOLM’s dog be- tic case, without constituting a fundamental incon- lieves that the cat went up that oak tree, we do not do gruity. The terms which occur in the content-clause so on the grounds that it picks out objects and classi- are in general dictated not so much by the creature fies them in a way that corresponds to the singular whose belief we report, but by the concerns of and general terms we use in the attribution—that is speaker and audience. Thus, “Sarah thinks that the why, for philosophical purposes, it may be more ac- charlatan you introduced me to is about to give her curate to rephrase (1) in the transparent manner of a biscuit” can be in order, whether Sarah is an adult, (1’). Rather, we simply note the dog’s reaction to its a child that lacks the concept of a charlatan, or a dog environment. We regard these reactions as directed (RUNDLE 1997, p83). towards particular objects, creatures and events, be- Consequently, it is far from obvious that attribu- cause we assume that dogs have certain perceptual tion of beliefs requires attribution of concepts. Still, capacities and wants, assumptions which require ru- the question whether animals can possess concepts dimentary knowledge of the way dogs live (what they remains relevant, for two reasons. When it comes to can recognize, what they tend to dislike, etc.). attributing beliefs to animals, some terms are more Because of its reliance on behavioral reactions, the absurd than others. This suggests that animals can holodoxastic move is confined to simple beliefs, no- be credited with some concepts but not others. Fur- tably about perceptible features of the subject’s en- thermore, if some animals have conceptual capaci- vironment. But it suffices to blunt the force of the ties, the lingualist argument from concepts fails even line “No thoughts without concepts!”. It may seem if it is right to tie beliefs to concepts. that a problem remains nonetheless. Granting thoughts to animals while denying them concepts 5. Animal concepts suggests that there is an incongruity between ascribing thoughts to animals and ascribing thoughts to With respect to animal concepts, one finds the same linguistic creatures. In the second case, our ascrip- spectrum of opinion we already encountered con- tions impute to the believer a grasp of the concepts cerning animal thoughts. KANT, FREGE, DAVIDSON involved, whereas in the first they do not. This cre- and DUMMETT occupy the lingualist corner. Accord- ates a pressure for holding that intentional verbs like ing to them, animals can perceive, but lack concepts “believes” are ambiguous, referring either to a holo- of any kind. In the other corner are cognitivists, who doxastic, behavioral phenomenon—notably per- have no qualms about ascribing complex concepts

Evolution and Cognition ❘ 181 ❘ 1999, Vol. 5, No. 2 Hans-Johann Glock to animals. An intermediate position is occupied, this absurdity follows from treating concepts as pow- for example, by Anthony KENNY, who maintains ers of discrimination. Olive trees do not discriminate that animals can possess some concepts, namely between moist and dry soil, since discrimination is those that can be manifested in non-linguistic a prerogative of sentient creatures, that is, animals. behavior (1989, pp36–7; see also DEGRAZIA 1996, We must distinguish between mere differential reac- pp154–6). tion to causal inputs, which is a universal feature of Proponents of this position have to concede that physical phenomena, and discrimination, which is the concepts animals have are often not the ones we tied to creatures with perceptual capacities. use in ascribing thoughts to them. The discrimina- Nevertheless, even proponents of animal con- tions which underlie animal behavior may not coin- cepts have come to accept DAVIDSON’s conclusion, cide either extensionally or intensionally with our namely that conceptualization requires more than verbal classifications. A dog might group cats to- discrimination (e.g., ALLEN/HAUSER 1996; ALLEN gether with hamsters or distinguish black cats from 1999; STEPHAN 1999). But there is much less agree- all others; and even if it groups all and only cats ment on what more is needed. together, it might recognize them by smell rather One noteworthy proposal is that of ALLEN and than visually. But this by itself is no obstacle to as- HAUSER. They distinguish between “recognizing an cribing to them concepts that differ from ours. For X” and “recognizing something as an X or recogniz- example, when Sue SAVAGE-RUMBAUGH’s chimpan- ing it to be an X”. The former is merely the “exten- zees distinguish foodstuffs and tools, the operative sional characterization of a discriminatory ability”, difference seems to be simply that between the edi- while the latter requires the ability to recognize an X ble and the inedible (1986, p257). Accordingly, what on the basis of several different properties, notably kind of concepts we should ascribe to avoid anthro- on the basis of properties which transcend percep- pocentrism is a matter for ethological discoveries. tion. Perhaps these properties should even be essen- We discover animal concepts by discovering the pa- tial rather than accidental to Xs (1996, p51). rameters governing their discriminatory behavior. I find this proposal unconvincing. For one thing, Such considerations are likely to indicate that our it is far from clear that “recognizing an X” is exten- ordinary ascriptions require qualification, but not sional: it would seem, for example, that one can rec- that they involve the kind of convenient pretence ognize a flash of lightning without recognizing a cer- lingualists have diagnosed. tain type of electric discharge. For another, the Whether this criticism holds water naturally de- proposal rules out the possibility of distinguishing pends on what one makes of concepts and concept- either between perceptual and more abstract con- possession. According to one construal, concepts are cepts or between having more or less rich concepts principles of discrimination, and to possess a con- of an X. Thus it implies that before the advent of cept is to have the ability to recognize or discrimi- instruments to measure the length of electromag- nate different types of things (PRICE 1953, p355; DU- netic waves, humans had no color concepts, because PRÉ 1996, p331). This construal is prima facie they could recognize colors in only one way—by plausible, and has served as a starting point even for looking—and one which is purely perceptual. I find many who regard concepts as a type of mental rep- it even less plausible to suppose that one has the resentations. It certainly implies that animals can concept of an X only if one distinguishes Xs by those possess concepts. Both in the wild and in the labora- features which we regard as essential to Xs. tory, animals distinguish between a host of different What is correct, however, is that for creatures that colors, tastes, sounds, shapes, stuffs, quantities, are capable of distinguishing between essential and types of creatures, etc. Moreover, many of these ca- accidental properties, what concept of an X they pacities are learned rather than innate.5 possess is determined by what properties they regard DAVIDSON thinks that this account of concepts as essential to being an X. Now, the distinction be- falls prey to a reductio ad absurdum. “Unless we want tween essential and accidental properties is beyond to attribute concepts to butterflies and olive trees, we the grasp of animals. But this should not count should not count mere ability to discriminate be- against their having concepts, since even many phi- tween red and green or moist and dry as having a losophers, especially radical empiricists, have at best concept, not even if such selective behavior is a tenuous grasp of the difference. Distinguishing es- learned” (1997, p25). Leaving aside butterflies for sential from accidental properties is, in my opinion, the moment, I agree that it would be absurd to credit crucial to a proper understanding of concepts, but it olive trees with concepts. But I do not accept that cannot be crucial to their possession.

Evolution and Cognition ❘ 182 ❘ 1999, Vol. 5, No. 2 Animal Minds: Conceptual Problems

6. Concepts and normativity tention to perform in this way, such failures are not misapplications of a rule. For these reasons, I favor a different way of separat- However, the crucial question is whether such ing conceptualization from discrimination, normative behavior is the prerogative of linguistic namely one which stresses the normative dimen- creatures. DAVIDSON thinks it is, because the behav- sion of conceptual classification. This idea goes ior of animals and children is purely mechanical, no back to WITTGENSTEIN, and it has been succinctly different in principle from that of a heat seeking mis- expressed by DAVIDSON. “To have a concept is to sile (1997, p25). This argument rightly assumes that classify objects or properties or events or situa- only intentional discriminations can be corrected in tions”, or, more accurately, to be able to do so. the relevant sense, because only intentional behavior Powers of discrimination, by contrast, are mere can be accused of misapplying a principle of classifi- “dispositions”, and therefore, “as WITTGENSTEIN cation. It also intimates, again rightly, that such be- emphasized, have no normative force”. Such dis- havior must be voluntary in the sense that the agent positions do not involve the ability to recognize a could have done otherwise. Mechanical behavior— mistake, and hence no knowledge of the difference no matter whether unconditioned or conditioned— between correct and incorrect behavior (1997, cannot be accused of failing to live up to a principle, pp24–5; see 1985, p480). roughly because ought implies can. This is why clas- It is at least partly the absence of such classifica- sification is not the exercise of a mere disposition to tion which makes us reluctant to credit butterflies react differently to distinct external influences, even with concepts. Furthermore, the normativity re- if, as with animals, these influences are stimuli per- quired for such classification presupposes that the ceived by a sentient creature. Rather, classification is classifier can make a mistake which she is capable of the exercise of an ability. Unlike dispositions, abili- recognizing as such. A WITTGENSTEINIAN distinction ties are not automatically exercised under specified can help to clarify this point.6 To be capable of clas- conditions; the agent can intentionally exercise sifying or misclassifying things, a creature a must them or refrain from exercising them.7 not just have a disposition to behave in accordance At the same time, DAVIDSON is wrong to hold that with a rule—as butterflies do when they land only non-linguistic creatures have only dispositions and on red petals—but of following a rule. That is to say, lack abilities. The behavior of non-linguistic crea- the principle which distinguishes Fs from non-Fs tures is not always explicable solely by reference to must be part of a’s reason for differentiating between immediate biological imperatives. Both prelinguis- Fs and non-Fs, not just a law to which its discrimi- tic children and the great apes are capable of volun- nating behavior conforms. tary action, because they can refrain from a particu- If a can classify things into those which are (an) F lar action, either by pursuing their goals in a and those which are not, it must be possible that a different way or by forsaking them, at least tempo- should be mistaken, namely in taking something as rarily.8 By the same token, in a particular situation not being F which is in fact F, or in taking something they are capable of either heeding or disregarding a as F which is in fact not. But a can be accused of difference. That is to say, they are capable of distin- making a mistake in applying the rule which distinguishing objects of types F and G in one situation, guishes Fs from non-Fs only if a is also capable in and of ignoring the difference in another. It would principle of recognizing that she has violated that be wrong to ascribe such classification to butterflies. rule. Only given that possibility can a be said to di- But it seems equally wrong to deny classification to verge from a rule which she was trying to follow, i.e., chimpanzees capable of selecting or making tools in to have acted contrary to her own intentions. Oth- advance of attempting a task. For these creatures de- erwise, a is merely diverging from our expectations liberately distinguish between different kinds of ob- or from a statistical norm. As DAVIDSON points out, a jects (e.g., different types of stones) in some situa- slippery road may be a danger or a nuisance, but it tions, but may disregard the difference in others, or does not commit a mistake. Mutatis mutandis, a but- if they are not in the mood (see BYRNE 1995, pp150, terfly that fails to discriminate between red and 187–9, 225, ch. 7; TOMASSELLO/CALL 1997, pp36, 78). green may reduce its biological fitness, but it does And here the question of whether a discrimination not violate a principle to which it has committed has been learned is important. An unconditioned itself. There are of course types of mistakes that do reflex cannot be the exercise of an ability, but a not require this possibility, e.g., failures to perform learned pattern of response can, because it is not nec- in line with evolutionary design. But without an in- essarily automatic.

Evolution and Cognition ❘ 183 ❘ 1999, Vol. 5, No. 2 Hans-Johann Glock

This response may confine non-linguistic con- behavior that go together with the correction of cept possession in the normative sense to infants error among humans, such as hesitation, displea- and the great apes. However, it does not make con- sure, discarding one type of tool in favor of another, cept possession dependant on language possession, etc. More generally, it is far from obvious that ani- but on discriminatory behavior that is sufficiently mals are incapable of reasoning. Chimpanzees complex and flexible to be subject to normative as- seem to do just that in their construction and em- sessment. Once more, the lingualist could not reply ployment of tools in advance of feedback from the that the appearance of flexible behavior among an- task itself. imals is deceptive, without inviting the same chal- Finally, consider the story of CHRYSIPPUS’ hunt- lenge concerning humans. ing-dog (see SORABJI 1994). In chasing a prey of The idea that cognition in animals is a measure which it has lost the scent, this dog reaches a cross- of the complexity and flexibility of their behavior roads; it sniffs down the first path, then sniffs down is widespread (e.g., TOMASELLO/CALL 1997, pp7–12; the second path, then it immediately follows the DUPRÉ 1996). DAVIDSON’s point that the ability to third without sniffing. In the case of dogs, perhaps recognize and correct mistakes distinguishes classi- such behavior could only be a rigid conditioned fication from discrimination has also been taken on reflex. But I can see no reason for denying that this board in recent contributions (ALLEN 1999; STEPHAN is an intelligible form of behavior for a non-linguis- 1999, p87). I want to suggest that these two points tic creature capable of voluntary action. And if it is, are connected through two requirements: classifi- what is wrong with the explanation that the behav- cation must be rule-guided, and rule-guided behavior evinces a disjunctive inference (“p or q or r; nei- ior must be intentional. ther p nor q; ergo r”)? We might grant that there is However, these requirements may yet prove to a difficulty in describing such a creature as silently be the undoing of animal concepts.9 For it is possi- consulting a principle. But as Ryle has convincingly ble to argue as follows. Even if animals are capable argued, even the intelligent performances of hu- of acting voluntarily, in the sense of doing other- mans are rarely accompanied by conscious consul- wise, and of acting intentionally, in the sense of act- tations of this kind. ing for a purpose, they are incapable of acting in- However, a related problem has been pointed tentionally in the stronger sense of acting for a out to me by Anthony KENNY. Although humans reason. We explain the behavior of animals by ref- need not actually verbalize their reasoning, they erence to reasons, (e.g., “the dog runs to the oak tree are capable of doing so. In the absence of this ca- because he wants to catch the cat”). But in doing so pacity, the question arises of what in an animal’s we indicate only what their purposes or goals are, behavior could correspond to the ergo of linguistic not how they have reasoned, i.e., what their justifica- reasoning. This point is unanswerable with respect tion is for acting as they do. For, as RUNDLE (1997, to creatures like dogs. But in the case of chimpan- ch. 4) has argued, that would presuppose that they zees there can be an analogue to our ergo, however are in principle capable of stating such reasons. If thin. In the context of encountering and ponder- this is correct, we can rule out animal concepts by ing a problem, certain gestures and grimaces, fol- ruling out animal reasoning. Animals might dis- lowed by renewed activity, can be interpreted as criminate for a purpose (e.g., to reap certain remarking the point when the shilling dropped. wards), but they cannot reason. Even though their Even if this is an anthropomorphic interpretation discriminations may be voluntary, they do not fol- in the case of chimpanzees, we can easily imagine low rules: they do not distinguish Fs from non-Fs a non-linguistic hominoid whose facial expres- for the reason that Fs possess certain distinctive fea- sions and gestures are so close to ours as to make tures. such a description inevitable. Furthermore, even But is this line of argument correct? Can a crea- without the ergo, as regards context (problem solv- ture only act for a reason if it is capable of commu- ing), demeanor (e.g., head scratching) and result nicating this reason? Take a chimpanzee that has (problem solution), the deliberations of chimpan- learned to use different tools in the pursuit of dory- zees are close enough to those of humans to qualify lus ants and macrotermes termites. It is plausible to as instrumental reasoning. maintain that its reason or justification for match- In my view, therefore, there is no compelling ing tool and prey is that they possess certain fea- case for claiming that animals cannot possess con- tures. This impression is strengthened by the fact cepts. Moreover, even if there is, it does not suffice that chimpanzees display non-linguistic forms of to deny them thoughts, because of the possibility

Evolution and Cognition ❘ 184 ❘ 1999, Vol. 5, No. 2 Animal Minds: Conceptual Problems of holodoxastic belief of perceptual kind. Lingual- up the oak tree without knowing that trees burn, or ists are right to insist that “concept-formation is that one can believe that a cloud is passing before not a way station between mere dispositions … the sun without knowing that clouds are made of and judgements” (DAVIDSON 1997, p25). Concepts water vapor. Such a radical conceptual holism also and judgements remain on a par. A chimpanzee creates a general problem, namely that any empir- capable of classifying things into sticks and knifes, ical discovery amounts to a conceptual change, and hence of concepts like “stick”, can also believe with the possible consequence that scientific theo- that the object it confronts is a knife, or wish that ries talk about different things as soon as some of it were a stick. My point is rather that it has yet to their empirical claims are at odds (FODOR/LEPORE be shown either that concept-formation and 1992, ch. 1). judgement require linguistic judgement or that ho- DAVIDSON is on firmer ground when he invokes lodoxastic belief cannot be a way station between general beliefs which are clearly conceptual, e.g., mere dispositions and judgement involving con- that cats are animals or continuing physical objects cepts. that move in certain ways (1999, pp8–9). Animals cannot recognize the conceptual status of such be- 7. The holistic nature of thought liefs, nor do they need to in order to have thoughts. But as DAVIDSON himself admits, they can “general- There is one last argument against animal concepts ize” in the sense of reacting similarly to similar and, by implication, against the possibility of con- stimuli (e.g., 1985, p480). By this token, they can ceptual thought among animals. It is that the con- also in principle distinguish not just between, e.g., cepts which feature in the beliefs we commonly mice and cats, but between animals, plants and in- ascribe to animals require certain general beliefs animate objects. The notion of a continuing phys- with which we cannot credit them. This is part of a ical object is more problematic, since it is in many wider qualm, namely that attributing thoughts to respects a result of philosophical reflection that ex- animals is incompatible with “the intrinsically ceeds the requirements of ordinary human thought holistic character of the propositional attitudes”, (see STRAWSON 1992, ch. 2). But if it is spelled out in the alleged fact that “to have one is to have a full a pedestrian way, it is clear that chimpanzees can complement” (DAVIDSON 1985, p473; see also STICH learn to distinguish, for example, between physical 1979). Since at least some members of that comple- objects on the one hand, mirror reflections or TV ment are definitely beyond their pale, animals can- images on the other. Finally, we must remember not even have the simple beliefs commonly that a creature could have concepts without having ascribed to them. our concepts. Elsewhere I have argued that this general line of This means that there is no holistic argument argument rests on holistic principles which are ei- against animal concepts over and above the argu- ther too strong for the lingualist case, because they ments discussed in section 5. More generally, ho- preclude plausible cases of human thought, or too lism does not provide a compelling objection weak, because they allow for some forms of animal against the possibility of animal thought. This is no thought (GLOCK 2000, sct. 4).This leaves more spe- reason to abandon holism altogether. There are cific qualms about the holistic costs involved in plausible holistic principles which exclude the pos- crediting animals with beliefs involving specific sibility of a creature having just a single belief. In concepts. That argument does not threaten the pos- any event, the complexity required for conceptual sibility of holodoxastic belief (sct. 4). To assess its belief is incompatible with a behavioral repertoire consequences for conceptual belief, one should dis- capable of exhibiting just a single belief. But these tinguish between two types of general beliefs, reflections do not establish that the web of which namely conceptual and empirical (without deny- any belief must be part need extend as far as the web ing that some of the examples are borderline cases, of sophisticated human thought. There may be e.g., that trees are growing things that need soil and larger and smaller webs. What kind of web is re- water). quired may depend on the belief and the creature DAVIDSON insists that beliefs presuppose certain concerned. From the fact that an animal lacks our general empirical beliefs on account of their con- web of beliefs and our concepts, it does not follow stituent concepts. This is implausible for many of that it has no beliefs and no concepts (see BEKOFF/ his examples (see 1985, p475; 1984, p200). It JAMIESON 1991, pp19–20; DUPRÉ 1996, p332; DEGRA- would seem that one can believe that the cat went ZIA 1996, pp154–8; ALLEN 1999, p39).

Evolution and Cognition ❘ 185 ❘ 1999, Vol. 5, No. 2 Hans-Johann Glock

8. Conclusion as deliberately as it does between red and black ants. But we are more inclined to ascribe to it the concept Where does all this leave the lingualist thesis that of redness than the concept of a keeper, because the capacity for thought requires the capacity for there is so much more to explain with respect to the language, especially on account of concepts? As latter. Finally, even those thoughts animals might regards intensionality (sct. 3), the problem is not be credited with stand in a smaller logical space. that a non-linguistic analogue (recognizing the They lack the kind of context which characterizes same objects through different features) is ruled sophisticated linguistic thought. out in principle, but that the range of distinctions Accordingly, attributions of simple thoughts to that can be displayed in non-linguistic behavior is animals are neither intensional, nor conceptual, restricted. The same message emerges from my nor holistic in the way thought-attributions to hu- moderate holism, according to which thoughts mans are. However, the best analogy is not the an- come in larger or smaller packages (sct. 7). Pack- thropomorphic explanation of missiles, but one ages that include beliefs manifestable only in lin- DAVIDSON has suggested in discussion. Attributing guistic behavior are the preserve of language-users. thoughts to animals is like using numerals for the Again, rejecting the building-block conception of purpose of labeling members of a football team. Al- thoughts and concepts makes room for employing though natural numbers stand in complex rela- intentional verbs like “believes”, “wants” or tions of order and numerical difference, these rela- “intends” to animals, without necessarily imput- tions are ignored in this context. What matters ing concepts (sct. 4). Finally, whether or not ani- here is not the numerical difference between two mals can have concepts of a simple kind depends numbers, nor even which one is greater, but only not on their having a language, but to the extent to that no two numbers should be used for the same which their discriminations are rule-governed and player. hence intentional (scts. 5–6). The analogy is illuminating. Thought attributions To this extent, our discussion reinforces the in- to animals employ a rich conceptual apparatus to an termediate position. At the same time, they also area in which many of the logical connections which support DAVIDSON’s suggestion that animals do not constitute that apparatus do not apply. But it breaks just have thoughts of a simpler kind, their having down in one important respect. Attributing these thoughts amounts to something simpler. In thoughts to animals is not simply an impoverished the case of animals, there is at most an analogue of application of a rich technique. For that richer tech- intensionality. In so far as thought-ascriptions to nique evolves around a central core of cases in which animals are holodoxastic, they are not only re- creatures believe, know or desire things on account stricted to thoughts about perceptible features of of their wants and perceptual capacities. These bio- the environment, they also lack conceptual con- logical basics of belief are shared by humans and an- nections which apply in the human case: we cannot imals. At the same time, when we move from this infer from the fact that the dog thinks that x is F core area in the direction of conceptual thought, we that the dog grasps the concept F. Furthermore, also move in the direction of linguistic thought. The even if animals can have concepts, these are not just features which non-linguistic creatures must possess confined to concepts of a (roughly speaking) per- to be capable of conceptual thought—intentional- ceptual kind. Animals are also incapable of satisfy- ity, complexity, flexibility—correspond to those fea- ing one of the two criteria which we standardly use tures by which theorists from DESCARTES to CHOMSKY in attributing concepts to hu- have distinguished language mans. They may be able to ap- from more basic systems of ply principles of classifica- Author’s address communication. In this re- tion, but not to explain them. spect, at least, our reflections Hans-Johann Glock, Department of Philoso- In fact, the two restrictions tend to confirm rather than to phy, The University of Reading, Reading are linked. A chimpanzee may negate the connection be- RG6 6AA, United Kingdom. discriminate between its tween thought and lan- Email: [email protected] keeper and other humans just guage.10

Evolution and Cognition ❘ 186 ❘ 1999, Vol. 5, No. 2 Animal Minds: Conceptual Problems

Notes 5 For primates, see TOMASSELLO/CALL (1997), chs. 4–5), and for the even more surprising achievements of pigeons, see 1 For a defense of a third person perspective on animal HERNSTEIN/LOVELAND/CABLE (1976). thought, see DUPRÉ (1996). For a critique of the explanation 6 The distinction is explained in detail by BAKER/HACKER of animal behavior by reference to complex calculations, (1985, pp154–8). see SEARLE (1997), and for a judicious assessment of cogni- 7 For this distinction, see KENNY (1975). In ordinary parlance, tive representationalism BECKERMANN (1999), ch. 10. dispositions include character traits which are half-way be- 2 As far as I can tell, this holds true even for authors wedded tween dispositions in this technical sense and abilities: they to a stronger notion of mental representation, e.g., in are neither realized automatically nor simply subject to the ALLEN/HAUSER (1996, pp54–5). will. 3 These qualms are not due to a failure to heed ALLEN’s dis- 8 This will be obvious to the parents of any pre-linguistic two- tinction between social and individual concepts (1999, year old. As regards the great apes, see GOODALL (1986), p35). Even a concept that is not actually shared between MENZEL (1974), BYRNE (1995), chs. 8–9. several individuals must be capable of being shared, and the 9 Even if this requirement is met in the case of certain ani- extent to which individuals share a concept can be deter- mals, it still remains to be shown that they really correct mined independently of establishing patterns of neural ac- mistakes, rather than simply alter their behavior. And if tivity in the brain. they do, does this require that these animals have second- 4 DAVIDSON’s work also features other arguments in favor of order beliefs? For a discussion of these questions, see GLOCK lingualism, including the argument that having a belief (2000), sct. 5. The role of rules for animal behavior also requires having the concept of a belief and that the latter plays a role in HENDRICHS (1999). requires language because of the need for “triangulation”. 10 For advice and comments on related material, I should like I discuss these arguments in GLOCK (2000), but leave them to thank Ansgar BECKERMANN, Peter HACKER, John HYMAN, aside here, since in my view they are less revealing about Sir Anthony KENNY, John PRESTON, Eike von SAVIGNY, Stuart the connection between thoughts, concepts and lan- SHANKER and Michael TOMASSELLO. This paper was written guage. during tenure of a research fellowship from the Alexander VON HUMBOLDT Foundation, for which I am grateful.

References McLaughlin, B. (eds) Actions and Events. Blackwell: Ox- ford, pp. 473–480. Allen, C. (1999) Animal Concepts Revisited: The Use of Self- Davidson, D. (1994) Davidson, Donald. In: Guttenplan, S. Monitoring as an Empirical Approach. Erkenntnis 51: 33– (ed) A Companion to the Philosophy of Mind. Blackwell: 40. Oxford, pp. 231–236. Allen, C./Bekoff, M. (1996) Intentionality, Social Play and Davidson, D. (1997) Seeing Through Language. In: Preston, J. Definition. In: Bekoff, M./Jamieson, D. (eds) Readings in (ed) Thought and Language. Cambridge University Press: Animal Psychology. MIT Press: Cambridge MA, pp. 229– Cambridge, pp. 15–27. 239. Davidson, D. (1999) The Emergence of Thought. Erkenntnis Allen, C./Hauser, M. (1996) Concept Attribution in Nonhu- 51: 7–17. man Animals: Theoretical and Methodological Problems DeGrazia, D. (1996) Taking Animals Seriously. Cambridge in Ascribing Complex Mental Processes. In: Bekoff, M./ University Press: Cambridge. Jamieson, D. (eds) Readings in Animal Psychology. MIT Dummett, M. (1993) Origins of Analytical Philosophy. Duck- Press: Cambridge MA, pp. 47–62. worth: London. Baker, G. P./Hacker, P. M. S. (1985) Rules, Grammar and Ne- Dupré, J. (1996) The Mental Lives of Nonhuman Animals. In: cessity. Blackwell: Oxford. Bekoff, M./Jamieson, D. (eds) Readings in Animal Psychol- Beckermann, A. (1999) Analytische Einführung in die Philos- ogy. MIT Press: Cambridge MA, pp. 323–336. ophie des Geistes. de Gruyter: Berlin, New York. Fisher, J. A. (1996) The Myth of Anthropomorphism. In: Bekoff, M./Jamieson, D. (1991) Reflective Ethology, Applied Bekoff, M./Jamieson, D. (eds) Readings in Animal Psychol- Philosophy, and the Moral Status of Animals. Perspectives ogy. MIT Press: Cambridge MA, pp. 3–16. in Ethology 9: 1–47. Fodor, J. (1975) The Language of Thought. Crowell: New Bennett, J. (1976) Linguistic Behavior. Cambridge University York. Press: Cambridge. Fodor, J./LePore, E. (1992) Holism: a Shopper’s Guide. Black- Byrne, R. (1995) The Thinking Ape Oxford University Press: well: Oxford. Oxford. Glock, H. J. (1997) Philosophy, Thought and Language. In: Carruthers, P. (1989) Brute Experience. Journal of Philoso- Preston, J. (ed) Thought and Language. Cambridge Univer- phy 86: 258–269. sity Press: Cambridge, pp. 151–169. Chater, N./Heyes, C. M. (1994) Animal Concepts: Content Glock, H. J. (1997a) Truth without People? Philosophy 72: and Discontent. Mind and Language 9: 209–246. 85–104. Crisp, R. (1996) Evolution and Psychological Unity. In: Glock, H. J. (2000) Animals, Thoughts and Concepts. Syn- Bekoff, M./Jamieson, D. (eds) Readings in Animal Psychol- these 119. ogy. MIT Press: Cambridge MA, pp. 309–321. Gould, S. J./Eldredge, N. (1977) Punctuated Equilibria: The Davidson, D. (1984) Inquiries into Truth and Interpretation. Tempo and Mode of Evolution Reconsidered. Paleobiology Oxford University Press: Oxford. 3: 115–51. Davidson, D. (1985) Rational Animals. In: LePore, E./ Goodall, J. (1986) The Chimpanzees of Gombe: Patterns of

Evolution and Cognition ❘ 187 ❘ 1999, Vol. 5, No. 2 Hans-Johann Glock

Behavior. Harvard University Press: Cambridge MA. Savage-Rumbaugh, S./Shanker, S./Taylor, T. (1998) Apes, Hendrichs, H. (1999) Different Roots of Human Intentionali- Language and the Human Mind. Oxford University Press: ty in Mammalian Mentality. Erkenntnis 51: 145–164. Oxford. Jamieson, D. (1998) Animal language and thought. In: Craig, Savigny, E. von (1995) Doubts Concerning Hopeless Dogs. E. (ed) The Routledge Encyclopedia of Philosophy. Rout- In: Hintikka, J./Puhl, K. (eds) The British Tradition in 20th ledge: London. Century Philosophy. Hölder-Pichler-Tempsky: Vienna, pp. Kenny, A. J. P. (1975) Will, Freedom and Power. Blackwell: 207–12. Oxford. Searle, J. (1997) The Explanation of Cognition. In: Preston, J. Kenny, A. J. P. (1989) The Metaphysics of Mind. Oxford Uni- (ed) Thought and Language. Cambridge University Press: versity Press: Oxford. Cambridge, pp. 103–126. Malcolm, N. (1972–3) Thoughtless Brutes. Proceedings and Seyfarth, R./Cheney, D. (1996) Inside the Mind of a Monkey. Addresses of the American Philosophical Society 46. In: Bekoff, M./Jamieson, D. (eds) Readings in Animal Psy- Menzel, E. W. (1974) A Group of Chimpanzees in a 1-acre chology. MIT Press: Cambridge MA, pp. 337–343 Field. In: Schrier, A. M./Stollnizt, F. (eds) Behavior of Non- Sorabji, R. (1994) Animal Minds and Human Morals. Rout- human Primates. Academic Press: New York, pp. 83–153. ledge: London. Pinker, S. (1994) The Language Instinct. Penguin: Middlesex. Stephan, A. (1999) Are Animals Capable of Concepts? Price, H. H. (1953) Thinking and Experience. Hutchinson: Erkenntnis 51: 79–92. London. Stich, S. (1979) Do Animals Have Beliefs? Australasian Jour- Rundle, B. (1997) Mind in Action. Oxford University Press: nal of Philosophy 57: 15–28. Oxford. Strawson, P. F. (1992) Analysis and Metaphysics. Oxford Savage-Rumbaugh, S. (1986) Ape Language: From Condi- University Press: Oxford. tioned Response to Symbol. Oxford University Press: Ox- Tomassello, M./Call, J. (1997) Primate Cognition. Oxford ford. University Press: Oxford. Savage-Rumbaugh, S./Lewin, R. (1994) Kanzi. Doubleday: Wittgenstein, L. (1967) Philosophical Investigations. Black- London. well: Oxford.

Evolution and Cognition ❘ 188 ❘ 1999, Vol. 5, No. 2 Rolf Lachmann

Animal Perception as Value Perception

HE ASSUMPTION THAT Abstract of approach, hypotheses Tanimals are con- can be formulated which, scious beings is widely Understanding animal consciousness demands a so- in turn, have to be veri- accepted. However, many phisticated methodology. One approach takes evi- fied with available biolog- also hold that their mode dence steming from human consciousness as its ical knowledge such as on of consciousness differs starting point. Here again various approaches are pos- the behavioral capability from our human con- sible. This paper claims the fruitfulness of a phenom- or deficiency in animals. sciousness. The question enological investigation and illustrates it concerning The final outcome is a “re- arises from these two perceptual consciousness. The phenomenological in- flective equilibrium”: a assumptions of how to vestigations of Alfred North WHITEHEAD, Arnold GE- plausible and highly cor- characterize animal con- HLEN and Maurice MERLEAU-PONTY converge in their roborated interpretation. sciousness. Human con- demonstration of the primitive layer of consciousness In the long run, such in- sciousness has turned out as pragmatic value perception. The hypothetical ex- vestigations have to be re- to be a very tricky subject tension of human perception to animal perception is stricted to particular spe- and difficult to grasp. supported by Susanne K. LANGER’s interpretation of cies. Nevertheless, it Thus a large part of twen- animal consciousness which gives additional weight seems to be possible to de- tieth century psychology to this interpretation by pointing to evidence steming velop plausible hypothe- has tried to evade any ref- from experiments on human perception. ses on animal perception erence to consciousness and, until such time as altogether in order to neurophysiology has de- become an objective sci- Key words veloped more substantial ence. While understand- approaches, these are the Pre-objective knowledege, phenomenological method, ing our own conscious- best interpretations we primitive layer of perception. ness is difficult enough, can accomplish. animal consciousness Restricting this paper poses an even larger problem. We cannot even rely on animal perceptual consciousness has the advan- on direct “knowledge by acquaintance” as we can in tage that here in both heuristic approaches one the case of our own conscious life. does not have to start at the very beginning. There An adequate understanding of animal conscious- are numerous experimental results regarding hu- ness can only be reached by a combination of evi- man perception. There are also penetrating phe- dence and research results from various approaches. nomenological investigations of human perceptual In this paper I will depict one approach to animal consciousness. In this paper I will first give an ac- consciousness and illustrate it by developing a hy- count of various interpretations of human percep- pothesis on animal perception. This approach as- tual consciousness which are either explicitly phe- sumes that human consciousness still includes nomenological or can at least be understood to be traces of animal consciousness which are, however, such. In outlining the interpretations of Alfred not open to immediate inspection but have to be North WHITEHEAD, Arnold GEHLEN and Maurice discovered in the course of a systematic investiga- MERLEAU-PONTY I will draw attention to the primi- tion. Two possible forms of systematic investigation tive layer of perceptual consciousness as under- are the study of experimental results originating in stood in these interpretations. Despite slight shifts human psychological research and the phenomeno- in emphasis, their interpretations lead to a conver- logical method developed by philosophers in the gent view which serves as a plausible candidate for first half of this century. On the basis of first hints an interpretation of animal perception. I will then and suggestions gained through these two avenues turn to Susanne K. LANGER’s interpretation of ani-

Evolution and Cognition ❘ 189 ❘ 1999, Vol. 5, No. 2 Rolf Lachmann

mal perception. In her interpretation LANGER relies an “isolated given” but always replete with refer- a great deal on findings which stem from experiences and complexities. All three philosophers draw ments on human perception. Indeed, the leading attention to the essential relatedness of perception idea of this paper is derived from LANGER’s theory and regard this as the basis for symbolic references. and intends to give her interpretation additional In the following I will ignore this aspect of their in- support. The goal of this paper is both to clarify the terpretations and focus on their depiction of the methodological position and to form a hypothesis primitive layer of human perception. on animal perception. 2. Alfred North Whitehead: The 1. Phenomenological investigations of emotional basis of perception human perception WHITEHEAD’s position that perceptual conscious- Phenomenological investigation is essentially a ness includes much more than our sense percep- methodologically guided attempt to an unpreju- tions is central to his interpretation of human diced description. The major sources of prejudices perceptual consciousness. Over and above our clear against which phenomenology directs its objec- and distinct perceptions of form, color etc. our per- tions are the traditional interpretations of philo- ceptual consciousness also includes numerous sophical theories, the theoretical accounts of indefinite and vague impressions. In order to clar- science and the prejudices of common sense. The ify the different perceptual layers WHITEHEAD dis- central goal of phenomenology is a fresh, compre- tinguishes between two forms of perception: hensive and unbiased look at its particular subject. perceptions in the mode of “presentational imme- It tries to ignore the numerous interpretations diacy” and perceptions in the mode of “causal effi- already given and steps back, for example, from the cacy”. ordinary understanding of human perception. Max Perceptions in the mode of presentational imme- SCHELER, one of the founders of the “phenomeno- diacy are the range of distinct perceptions in the logical movement”1, criticized both the philosophi- focus of our waking consciousness when we see col- cal and psychological theories of perception of his ors and forms or have smell, taste, acoustic or touch time in that they never ask “What is in fact given in impressions (see WHITEHEAD 1985, p25). They are our perceptual consciousness?” but only “What can what is ordinarily understood as perception. These be given in our perception?” The answer to the latter perceptions appear in our waking consciousness as question was, of course, that only what is provided “directly given” or “immediately present”. There- by our sense organs can be given. Hence, man fore WHITEHEAD calls them perceptions in the mode receives various visual, acoustic or kinesthetical of “presentational immediacy”. They are relatively impressions and, by combining them, organizes his clear and distinct. Here we are confronted with a knowledge. This view, however, is in sharp contra- complex variety of distinct perceptions, which are, diction to perceptual consciousness as we experi- to a large extent, controllable at will by closing our ence it. Phenomenology protests against any eyes, ears or nose. Moreover, these perceptions are account of human perception which ignores the relatively insignificant for our existence. Loss of eye- factual complexities and appearances of our percep- sight does, of course, impair our ability to orient tual life (SCHELER 1973, p55). The shared opinion of ourselves in our surroundings. It does not, however, authors who try to take a new and more penetrating threaten our physical existence. WHITEHEAD charac- view of human perception is that the one-sidedness terizes perceptions in the mode of presentational and the difficulties of many modern philosophical immediacy as “vivid, precise, and barren” (WHITE- problems derive from a highly artificial view of HEAD 1985, p23). They present a complex richness human perception. Therefore, a new look at the real of distinct perceptions but are without intrinsic rel- ingredients and structures of human perception is evance. They are in themselves nothing but a spec- needed. tacular “adventitious show”. Despite the fact that WHITEHEAD and GEHLEN did But there is another perceptual sphere. WHITE- not consider themselves primarily as phenomenol- HEAD refers to its content as perceptions in the mode ogists, their contributions can be regarded as phe- of “causal efficacy”. In this case we feel forces acting nomenological investigations. In combination with upon us, sense a causal influence. Our feeling of MERLEAU-PONTY’s phenomenology of perception the causal influence is not in the focus of our conscious shared view emerges that human perception is never attention but it is always present. We feel that each

Evolution and Cognition ❘ 190 ❘ 1999, Vol. 5, No. 2 Animal Perception as Value Perception concrete moment is influenced and shaped by the efficacy is seen in our emotions of anger, hate, fear, contents of its immediate predecessor. The pain we love or attraction. “The basis of experience is emo- had a moment ago impresses itself on each succeed- tional. Stated more generally, the basic fact is the ing moment and remains effective until it fades rise of an affective tone originating from things away or is ruled out by opposing forces. In every whose relevance is given.” (WHITEHEAD 1967, p176). actual moment we feel that the pain has been forced At the basis of our consciousness lies a vectorial and upon us. Here we feel the causal insistence of the directing layer. The intrinsic emptiness and barren- past and its determining force. WHITEHEAD draws at- ness of our clear perceptions contrasts with the deep tention to the fact that every instant is a continua- emotional significance of our causal feelings. tion and reproduction of its predecessors in the im- In order to act successfully we have to take the mediate past: “Our primitive perception is that of surrounding forces into consideration. Therefore ‘conformation’ vaguely…” (WHITEHEAD 1985, p43). we use visual and acoustic perceptions in order to In contrast to the “vivid, precise, and barren” pre- anticipate causal forces.2 This is the elementary form sentations of forms and colors, perceptions in the of symbolic reference: the use of distinct percep- mode of causal efficacy are vague, massive and im- tions to indicate surrounding forces. portant. They are “…heavy with the contact of the things gone by, which lay their grip on our imme- 3. Arnold Gehlen: Perceptual diate selves.” (WHITEHEAD 1985, p44). In this perceptual mode we have the vague but insistent feeling of enrichment through practical behavior forces acting upon us. Hence, they are felt with great This last aspect of practical orientation, a combina- emotional weight. Moreover, we cannot control or tion of distinct presentations and the feeling of ignore these perceptions in the same way as we can forces is also a central theme of Arnold GEHLEN’s control perceptions in the mode of presentational interpretation of human perceptual consciousness. immediacy. Perceptions in the mode of causal effi- GEHLEN develops his interpretation as part of his cacy create a continuous background of feeling at question concerning the biological “special place” the base of our consciousness. These perceptions are of man. The central idea of his theory is the demon- closely connected with our bodily feelings (WHITE- stration that man is biologically maladjusted. In his HEAD 1968, pp152–153). organic constitution man is a “deficient being”. It may appear odd to call these feelings of causal This is particularly obvious in man’s almost total efficacy perceptions since they are not perceptions loss of instinctive reaction, a fact that also has far- in the sense of being derived from our sense organs. reaching consequences for his perceptual con- It might seem more adequate to speak of an “feeling sciousness. Whereas for animals perceptions func- of efficacy” or “experience of efficacy”. Indeed, tion essentially as signs, e.g., as triggers for WHITEHEAD describes this perceptual mode as “non- genetically determined instinctive reactions, they sensuous perception” (WHITEHEAD 1967, p180). have no biologically prepared function for man. Nevertheless, it must be regarded as a form of per- Because of man’s loss of instinctive reaction percep- ception because it is not a self-invented product but tions have no biologically prepared behavioral sig- a reception of factors which are external to each nificance. This leads to “man’s unusual receptivity particular moment in question. to perceptions” (see GEHLEN 1988, p27). GEHLEN WHITEHEAD adds another distinguishing feature interprets the human situation as “flooded with in the perceptual modes. Whereas perceptions in stimulation” in “an abundance of impressions” the mode of presentational immediacy are simply (GEHLEN 1988, p28), to which the infant is exposed “given” without any reference to a past or a future, after birth. “Since he cannot rely upon instincts for perceptions in the mode of causal efficacy are full of understanding his environment” (GEHLEN 1988, moving and directing force. They are vectorial in p28)3 the child has to be relieved of the burden- character. At the root of our consciousness there is some mass of perceptions. a vague but constant feeling of movement, activity, In order for man to be relieved of the burden of expansion towards, retreating from, and passage. sensory impressions, learning becomes imperative. Hence, perceptions in the mode of causal efficacy Contrary to animals man has to build up his life are overpowering, inescapable and motivating. In independently and shape it step by step. The essen- contrast to the “adventitious show”, the superficial tial task is to enrich his perceptions with meaning. and secondary product of complex subtleties, the Since perceptions have no biologically determined directing force of perceptions in the mode of causal “meaning”, the perceptual enrichment with mean-

Evolution and Cognition ❘ 191 ❘ 1999, Vol. 5, No. 2 Rolf Lachmann ing has to be reached by active manipulation of the ‘symbols’—of outcomes and reactions we can ex- surrounding situations.4 GEHLEN gives a detailed ac- pect, of resistance we might encounter. With the aid count of the endless manipulation and “working of these symbols, we can initiate and regulate move- through” of things by children. In their developments, even before success of failure has occurred.” ment a considerable amount of time is spent on (GEHLEN 1988, p158). In the course of this develop- touching, moving, seeing, manipulating and de- ment our perceptual “contact” with surrounding forming everything within their reach in an attempt things is reduced to a minimum. Our perceptual to experience all there is to know about their prop- consciousness is relieved of the burden of too many erties, their uses, and their effects. These activities functionally meaningless sense perceptions. are not motivated by any direct practical necessity. “In general, these efforts consist of the ‘movements’ 4. Maurice Merleau-Ponty: The that occupy man’s childhood, movements through which man gradually experiences the world around biological formation of perception him. In this process, objects are seen, touched, GEHLEN’s view that our daily perception is saturated moved, handled in communicative movements of with “pragmatic values” reinforces WHITEHEAD’s interaction … As a result, the world is ‘worked position that perceptions in the mode of presenta- through’ with an eye toward its potential availabil- tional immediacy function as directing our aware- ity and usefulness to man. In succession, objects are ness to the causally relevant aspects of our experienced by man and then set aside; the objects environment. MERLEAU-PONTY’s phenomenological are thereby unwittingly endowed with a high degree investigation of our perception clarifies another of symbolism such that, eventually, the eye alone aspect. Similar to WHITEHEAD, MERLEAU-PONTY’s (an effortless sense) can take them in and quickly main theme is his opposition to the empiricist psy- assess their potential usefulness and value.” (GE- chology of association. The assumption that our HLEN 1988, p32). The result of this extended manip- perceptual consciousness consists of clear, distinct ulative behavior is an enrichment of our distinct and isolated sense perceptions contrasts sharply perceptions with pragmatic “values”. He illustrates with the real nature of our perceptual conscious- his interpretation by the perception of a teacup. In ness. MERLEAU-PONTY claims that the assumption of our ordinary perception we see the teacup ex- an isolated perception of distinct features is a myth. tremely selectively. For example, we ignore the vari- We never have a completely isolated perception. ation of its color by the distribution of light and On the contrary, even the most simple perception, shade, its ornamentation and other esthetic fea- say of a red spot, includes structures and relations. tures. On the other hand, without closer examina- One elementary structure is that we see the spot tion we immediately grasp its pragmatically rele- against a background, or the spot has an outline, a vant features such as its weight, material structure, border, internal shading etc. We spontaneously stability and handling possibilities. The pragmati- judge all the points as either belonging to the spot cally relevant aspects of surrounding things are the or to the surrounding space. The elementary focus of our ordinary perceptual attention. The tea- embeddedness of the spot in a background and all cup exists for us primarily as a condensation of our other structurings have the quality of a “phenome- previous handling experiences, as a complex of var- nal field”. “The perceptual ‘something’ is always in ious “visual indications of movement” (GEHLEN the middle of something else, it always forms part 1988, p55 and p160). of a ‘field’.” (MERLEAU-PONTY 1962, p4). Thus, in the course of our manipulating experi- Despite the elementary relatedness of any percep- ences, our perceptions become “symbolic”, they are tion it is far from true that every perceptual structur- loaded with “meaning”. At first glance, the percep- ing is completely distinct and determined. On the tual constellation informs us about the pragmatic contrary, most elements of our perceptual con- aspects of a thing or situation. “We orient ourselves sciousness remain vague and indeterminate. Our in the sensory world through certain minimal opti- perceptual consciousness is not a range of clear and cal, acoustical, tactile, etc. symbols. From a biologi- distinct perceptions but is largely amorphous and cal perspective, this is highly expedient: it saves us ambivalent. “The perceived, by its nature, admits of from getting caught up in the profuse abundance of the ambiguous, the shifting and is shaped by its con- things; furthermore, the greatest possible degree of text.” (MERLEAU-PONTY 1962, p11) MERLEAU-PONTY receptivity to stimuli in an organism is not the pur- calls this vaguely structured perceptual conscious- pose of perception. Our visual perceptions give us ness a “pre-objective realm” (MERLEAU-PONTY 1962,

Evolution and Cognition ❘ 192 ❘ 1999, Vol. 5, No. 2 Animal Perception as Value Perception p12). He illustrates his interpretation by referring to any elements which do not fit into the scheme of the experience of vague localization. Normally, our primary orientation. when a particular part of our skin is touched, we can MERLEAU-PONTY characterizes the pre-objective locate it spatially even if we cannot see the touch- structuring as an endowment of meaning. This ing. We have a spatial impression. We can also feel again can be illustrated by the perception of the red the distance between various stimulated spots. Nev- spot. We are strongly inclined to assume the back- ertheless, these impressions often remain vague and ground continuing behind the spot despite the fact imprecise. This becomes evident when we are asked that we cannot see that. We do not usually see the to point our finger to the spot. Our finger often has background as being interrupted by the spot. The to move around that area of the skin in order to same is true of the fact that our perceptions have identify the exact spot. “A vaguely located spot is a expressive qualities. The red spot is felt to be contradictory phenomenon and reveals a pre-objec- “warm”, “cold”, “attracting” or “repulsing”. These tive space where there is extension, since several examples show that our perception is full of aspects points on the body touched together are not con- and assumptions which are not validated by the per- fused by the subject but as yet no univocal position, ceptions themselves. MERLEAU-PONTY calls these as- because no spatial framework, persists from one per- pects which are “read in” to our perceptions “mean- ception to another.” (MERLEAU-PONTY 1962, p50).5 ing” (MERLEAU-PONTY 1962, p4). In fact our Although our perceptual experience is more or less inclination to meaning is the very basis of our per- structured it remains obvious that it is not objec- ception. tively determined and clearly given. MERLEAU- It would be false to regard the pre-objective phase PONTY calls this layer of consciousness a “non-pos- of our perceptual consciousness as disturbing. Quite iting consciousness” (MERLEAU-PONTY 1962, p49). the contrary, it is at this layer that the direction MERLEAU-PONTY describes the process of reaching which determines what we are going to perceive a distinct and objective identification of qualities more closely in the first place and how the percep- or relations as a “passage from the indeterminate to tions are synthesized is fixed. The structuring pro- the determinate” (MERLEAU-PONTY 1962, p24) or as cess does not start with our distinct perceptions but a “passage from indistinctness to clarity” (MERLEAU- is the very basis of having particular perceptions at PONTY 1962, p27). This transition to an explicit ap- all. It is set up in the pre-objective phase of percep- prehension is exceptional in our everyday life. Here tion. Every explicit perception is “already endowed we mostly remain in the pragmatically sufficient with a meaning” (MERLEAU-PONTY 1962, p9), which sphere of a vaguely structured perceptual con- means that every explicit perception is motivated sciousness. Only in rare moments our perceptual by a disposition to a certain meaningful constella- consciousness has the character of an explicit ap- tion. “The elementary event is already invested with prehension of the distinct features of surrounding meaning, and the higher function will bring into things. being only a more integrated mode of existence or This position is corroborated by an earlier analy- a more valid adaptation, by using and sublimating sis of Max SCHELER. SCHELER talks of the “value per- the subordinate operations.” (MERLEAU-PONTY 1962, ception”, which is active before the higher develop- p10). ment of “form perception” takes place. He claims MERLEAU-PONTY goes a step further in characteriz- that the more primitive value perception deter- ing the primary orientation towards meaning. The mines the range of what we then perceive distinctly, pre-objective orientation, which determines what if we reach the phase of distinct perception at all. we perceive and in which way, is closely connected Anything that does not lie in the direction of our with our bodily constitution. The meaning that we value perception is neglected and ruled out for per- endow our perceptions with depends on the “bio- ception. He illustrates this “suppressing” force by logical meaning of the situation” (MERLEAU-PONTY referring to a common experience. “Thus, when oc- 1962, p11). Every perception is penetrated by bodily cupied in important affairs, for instance, we may feelings. This is the reason why we see a wheel lying feel a ‘pull’ (perhaps caused by the facial expression on the ground differently from a wheel carrying a of another person) toward a specific direction in our heavy burden. MERLEAU-PONTY calls the bodily relat- environment [Umwelt]; but we do not follow it, and edness of our perceptions their “vital value” (MER- so this picture-content of what is striven for does not LEAU-PONTY 1962, p52). At other times he refers to come about.” (SCHELER 1973, p34). The actual focus the effects of the elementary imposition of meaning of attention suppresses the explicit perception of as the “affective value or biological meaning” (MER-

Evolution and Cognition ❘ 193 ❘ 1999, Vol. 5, No. 2 Rolf Lachmann

LEAU-PONTY 1962, p123) of perceptions. He thereby 1. Our perceptual consciousness is much more en- intends to make clear that our perceptions are di- compassing than the sum of our distinct percep- rectly related to our emotions and our motor activ- tions derived from our sense organs. ity, and to our entire bodily constitution. This is due 2. The perceptual layer at the base of our distinct to the fact that our body is not an inert and passive perceptions is vague and highly indeterminate thing but a “momentum of existence”. “Existence”, but it is not totally unstructured. a central concept of MERLEAU-PONTY, means the or- 3. Primitive perception is of great massiveness and ganization of our perceptions by the predispositions emotional force. It motivates the direction of of our body. “When we say that an animal exists, what we then perceive distinctly. that it has a world, or that it belongs to a world, we 4. The elementary layer of perception has a prag- do not mean that it has a perception or objective matic and behavioral significance. consciousness of that world. The situation which 5. The perception of pragmatic meaning has the unleashes instinctive operations is not entirely ar- character of perceiving the pragmatic “values” of ticulate and determinate, its total meaning is not things and situations. possessed, as is adequately shown by the mistakes The outlined positions are interpretations of and the blindness of instinct. It presents only a prac- human perceptual consciousness.6 Putting them in tical significance; it asks for only bodily recognition; the context of the leading question of this essay it is experienced as an ‘open’ situation, and ‘re- one can ask (1) whether it can be assumed that the quires’ the animal’s movements, just as the first layer of value perception exists in animal percep- notes of a melody require a certain kind of resolu- tion as well. (2) Moreover, if one is inclined to tion, without its being known in itself…” (MERLEAU- answer in the affirmative one can go further and PONTY 1962, p78). ask how much weight this form of perception has In conclusion one can say that perception of dis- in animal perception. One extreme position is that tinct qualities (perceptions in the mode of presenta- animal perception remains entirely within the tional immediacy) is not elementary but a late de- bounds of value perception and does not extend to velopment and, in fact, exceptional. It is an any distinct perception. However, more modest elaboration of a pre-objective and vague perception interpretations are also possible. An adequate in which the higher phases of perception are already answer cannot be given without a comprehensive structured and directed. This structuring is basically investigation of the evidence from biological due to our bodily constitution. Hence perception is research. essentially a pragmatic value perception. One such investigation based both on a survey of ethological data and on a consideration of results 5. Results of the from experiments on human perception was carried out by Susanne K. LANGER in the context of her nat- phenomenological investigations uralistic philosophy of mind. LANGER undertook an Although originating in different philosophical interpretation of general principles of animal per- contexts, the interpretations of human perceptual ception and animal social behavior in order to fix consciousness outlined above are mutually corrob- the dividing line between animal and man. She orative. WHITEHEAD’s analysis draws attention to reaches the position that the elementary form of the fact that besides the perceptions in the mode of human perceptual consciousness extends back to presentational immediacy there are other percep- animal perception. Indeed, although LANGER does tions which have a directing and motivating not refer to the authors considered above or any power. This can be found in GEHLEN in the context other philosophical theory, they were of major im- of the ontogenetical development. Finally, MER- portance for her philosophy.7 LEAU-PONTY argues that distinct perceptions do not simply enter our consciousness but are the out- 6. Susanne K. Langer’s interpretation come of earlier processes of structuring which stem of animal perception from our bodily constitution. MERLEAU-PONTY’s theme is the intermediate sphere between the per- In her Mind: An Essay on Human Feeling, which aims ceptions in the mode of presentational immediacy at a “biologically” based philosophy of mind, and perceptions in the mode of causal efficacy. The LANGER discusses the dividing line between man given depiction gives sufficient evidence for the fol- and animal.8 Before the field of “cognitive ethol- lowing account: ogy” had emerged LANGER objected to the behavior-

Evolution and Cognition ❘ 194 ❘ 1999, Vol. 5, No. 2 Animal Perception as Value Perception

istic prohibition of investigating animal and tion.” (LANGER 1972, p55). Hence, animals do not human consciousness. Despite the complexities perceive distinct qualities in their individual appear- and difficulties of interpreting the subjective form ance but only in their value character for actual be- of feeling, a comprehensive view of animal intellect havior. LANGER, too, uses the value concept in order and human thinking has to include an analysis of to characterize the pragmatic structure of animal consciousness. perception.10 According to LANGER animal perception has to be LANGER refers to the observations of Hans VOLKELT understood as depending fully on physiological and for further evidence on the structure of animal per- behavioral conditions. The frame of what an animal ception. In his famous experiments with spiders perceives is always directly controlled by the in- VOLKELT showed that they obviously did not recog- traorganic motivation i.e., by hormonal cycles. nize the same thing when it appeared in different Both seasonally changing sensibilities and the di- behavioral situations. If a fly was put into the web in rect physiological situation (e.g., hunger) sensitize front of the spider, the spider performed as usual, animals for specific sensory stimuli. Hunger sensi- paralyzing and ensnaring its prey. However, if the fly tizes perception for everything that can be eaten was put into the rear of the lair the spider showed a and the seasonally changing motivations activate difference in behavior and retreated from the fly. So dispositions to specific behavior such as nest-build- what the spider perceived was evidently not one and ing, mating, breeding or migration. “The currently the same fly appearing in different situations. prevailing chemical balance is a general condition VOLKELT concluded that the spider “perceived” to- underlying the play of perceptions, and as these are tally different situations the elements of which have essentially emotive perceptions of momentary val- different values for action. Felix KRÜGER called this ues the animal’s total awareness is governed by the relational dependence of perceived elements the inward pattern of sensitization.” (LANGER 1972, “complex-quality”.11 This implies that for animals p128). This entails a constantly changing evalua- perceptions have no stabile and invariant meaning. tion of the environment. What perception means to them is always dependent LANGER’s central hypothesis here is that the per- on their actual behavior. The value of a perception is ceptual consciousness of animals is completely inte- its role in a larger complex of factors determined by grated in animal behavior. She speaks of the “inher- behavioral orientation. No aspect has an indepen- ence” of “sensory acts in larger behavioral acts” dent meaning. Any minor change in the constella- (LANGER 1972, p59). For animals, perceptions de- tion can lead to a total shift of meaning. LANGER pend directly on current behavior. This total integra- writes: “…animal perception might be normally a tion in action is illustrated by space perception. Most matter of locating situations for action, in which a animals are not able to assess the structure of a loca- center of highest value draws the agent’s interest; tion just by looking at it. Carrying a cat around a that center (for us, the ‘object’) presumably has sen- location to make it familiar with the area will not sory properties which the animal recognizes without teach it anything. For the cat, acquaintance is a mat- conceiving them descriptively, i.e., without distin- ter of moving and behaving in the location. The lo- guishing them as shapes, colors, surface feelings or cation exists only as a behavioral space, not as ab- even characteristic smells…” (LANGER 1972, p116). stract knowledge. Animal perceptual consciousness is restricted essen- Animals do not perceive their environment in its tially to the pragmatic layer of perception which is at individual appearance of forms and qualities. Per- the root of distinct perception. However, the percep- ceptions function only in their pragmatic relevance tual process originating in that layer is interrupted as “values” for action.9 A cat that is sensitized to chas- before it reaches the higher elaboration of distinct ing behavior chases anything that runs or flutters. qualities as takes place in human perception. The individual perceptual qualities are compara- LANGER gives additional weight to her interpreta- tively insignificant. What counts is its pragmatic tion of animal perception by referring to findings value for particular behavioral action. Thus, LANGER from experiments on human perception. Here, it is maintains that: “…the primary characteristics, a well-established fact that conscious perception which animals see are values, and all the qualities of follows a phase of unconscious evaluation. Under- form, color, shape, sound, warmth, and even smell, standing words, for instance, is influenced by a pre- by which we would naturally expect them to recog- monition of their positive, negative or inappropri- nize things, enter into their perceptual acts only as ate value. Motorists on the highway recognize more they enter into their overt behavior as values for ac- quickly the message “keep right” than contextually

Evolution and Cognition ❘ 195 ❘ 1999, Vol. 5, No. 2 Rolf Lachmann inappropriate words. Other experiments add to the spontaneous recognition of the pragmatic values of evidence that the value of a perception is recog- a behavioral situation.12 nized even before it is distinctly perceived. “The LANGER’s interpretation of animal perception in significance of these findings in human psychology terms of primitive phases of human perception has for our necessarily speculative judgments of animal far-reaching consequences. If animals do not have feeling, perception and motivation lies in the ob- distinct perceptions then one also has to deny that jective demonstration that value may be adum- they have images, conceptions or a memory in the brated before perception of forms is complete; in- same sense as humans do. Past and future situations deed, the expectant, covert anticipation of the full cannot be represented in terms of having images of percept appears to be an emotively tinged process, those situations, as in the case of humans. Then, missed in our ordinary introspective analysis be- memory is only possible for animals in the sense of a cause of its minuteness and transitory character.” behavioral familiarity with situations. Past experi- (LANGER 1972, p115). LANGER suggests that this ences are integrated in the behavioral structure. Mem- phase of preconscious perception is probably much ory is nothing but behavioral reproduction—without more dominant and perhaps even more elaborate any accompanying imaginary consciousness.13 in animal perception because contrary to man it is not a transitory phase for animals. Perhaps animals 7. Results never enter the phase of distinct presentational perception at all. They may remain in the more prim- The depicted interpretations served both to (1) clar- itive phase of value perception since the more ify a methodological position on how to reach highly elaborated phase is constantly interrupted hypotheses on animal consciousness and (2) a par- by the fact that animals only perceive in the frame- ticular hypothesis concerning animal perception. If work of a constantly changing behavioral situation. human consciousness includes traces of animal Perception is permanently fixed to a constantly consciousness (which is plausible from the evolu- shifting behavioral constellation. This hypotheti- tionary constellation (then one can reach a hypoth- cal interpretation does not suggest that animals do esis about animal consciousness by means of not have very differentiated perceptions. On the experimental and phenomenological investiga- contrary, in many respects animal perception is su- tions. The most plausible candidates for such perior to human perception. However, what is hypotheses are the more primitive aspects still meant here is that for animals perception does not found to be operating in human consciousness. arrive at full conscious apprehension of distinct This can serve as a starting point for the second step: qualities. Animals perceive highly sensitively but a corroboration or falsification of the hypothesis by do not enter a conscious perception of the qualities checking it with biological evidence and findings. of form and color in their individual characteristics The methodological position has been illustrated (see also GIBSON 1973, pp308–309). by dealing with the particular aspect of animal per- This interpretation of animal perception also ception. A phenomenological investigation as well suggests that animals are capable of generalization as experimental findings suggest that the recogni- and behavioral flexibility. Since situations never re- tion of value takes place before there is any distinct cur in exactly the same way, successful behavior perception of qualities. It has been suggested that depends on their ability to recognize different situ- this hypothesis is in accordance with the available ations as “equivalent” to each other. Flight behav- evidence on animal behavior and their intellectual ior can be brought to an end abilities. Even if this interpre- successfully in a variety of ac- tation is only the beginning tions of the same pragmatic Author’s address and still has to be specified value: climbing a tree, jump- with regard to a particular ani- Rolf Jürgen Lachmann, Lehrstuhl für Phil- ing into the water or running mal species, it does provide a osophische Anthropologie und Kultur- into a hole. Hence, successful general staring point for fur- philosophie, UL 6, Raum 2049, Humboldt- behavior may not be so much ther research with the hy- Universität Berlin, Unter den Linden 6, D- the outcome of complex and 10099 Berlin, Germany. pothesis that animal percep- explicit thinking, comparison Email: [email protected] tion is essentially value per- and judgment but of the ception.

Evolution and Cognition ❘ 196 ❘ 1999, Vol. 5, No. 2 Animal Perception as Value Perception

Notes background” for the development of her position on animal perception. 1 For a comprehensive historical account see SPIEGELBERG 8 LANGER’s theory is not restricted to animal perception. She (1984). also discusses the different forms of social behavior (sugges- 2 WHITEHEAD illustrates this with an example of traffic behav- tion versus imitation), the contagion of feeling (empathy ior. We do not wait in front of the traffic lights in a contem- versus sympathy) and of mutual exchange (communion plative attitude enjoying the various colors and tones. versus communication) in animal and man. Instead we look at them in order to get information con- 9 LANGER’s interpretation is influenced by GIBSON (1973) who cerning the forces that determine our behavioral environ- also refers to an elementary form of perception as the per- ment and are relevant to the success of our intentions. ception of pragmatic values. 3 GEHLEN also stresses a particular human “excessive nature 10 The value concept as it is used here does not include moral of impulses” which has to be processed. See GEHLEN (1988, values, which according to LANGER are restricted to the hu- p49ff). man sphere. 4 GEHLEN also gives an account of animal perception which 11 The concept “Komplexqualität” was introduced by F. is different in structure from human perception. He stresses KRÜGER and was used in VOLKELT (1914, p84). that animals have perceptions in their role as indications 12 LANGER’s position is corroborated by D. MCFARLAND’s posi- and triggers of instinctive reactions. See GEHLEN (1988, tion that animal behavior “…is directed, not by goal-repre- pp137–157). sentation, but in a holistic manner which takes into 5 This ambivalence is typical for the pre-objective percep- account all (relevant) aspects of the animal’s internal state, tion. See also MERLEAU-PONTY’s characterization of this pro- and of the (perceived) external situation.” (MCFARLAND cess as a “still ‘empty’ but already determined intention.” 1989, p49). MERLEAU-PONTY (1962, p28). 13 In this regard Langer is in sharp contradiction to the lead- 6 An exception is WHITEHEAD who does not restrict his anal- ing assumption of cognitive ethologists. See e.g., GRIFFIN ysis to human perception and includes animal perception (1991, p5). LANGER’s thesis that animals have no explicit as well. memory in terms of a conscious recollection of past events 7 This is particularly true of WHITEHEAD and GEHLEN. There is in accordance with Henri BERGSON’s distinction between are no indications, however, whether LANGER knew MER- two forms of memory. BERGSON regards the elementary LEAU-PONTY. But again, phenomenological philosophy form of memory as a motor habit or a system of move- played an important part in her thought. Therefore, the ments. This is different from an explicit memory in terms sketched positions can be regarded as a “philosophical of an imaginary remembrance. See BERGSON (1991, pp68– 80).

References Merleau-Ponty, M. (1962) Phenomenology of Perception. Routledge & Kegan Paul: London (Originally appeared in Bergson, H. (1991) Materie und Gedächtnis. Meiner: Ham- 1945). burg (Originally appeared in 1896). Spiegelberg, H. (1984) The Phenomenological Movement. A Gehlen, A. (1988) Man. His Nature and Place in the World. Historical Introduction. Martinus Nijhoff: The Hague Columbia University Press: New York (Originally appeared (Originally appeared in 1960). in 1940). Scheler, M. (1973) Formalism in Ethics and Non-Formal Eth- Gibson, J. J. (1973) Die Wahrnehmung der visuellen Welt. ics of Values. A New Attempt toward the Foundation of an Beltz Verlag: Weinheim (Originally appeared in 1950). Ethical Personalism. Northwestern University Press: Evan- Griffin, D. R. (1991) Progress Toward a Cognitive Ethology. ston (Originally appeared in 1913/1916). In: Ristau, C. A. (ed) Cognitive Ethology. Erlbaum: Hills- Whitehead, A. N. (1985) Symbolism. Its Meaning and Effect. dale. Fordham University Press: New York (Originally appeared Langer, S. K. (1967) Mind: An Essay on Human Feeling, vol. I. in 1927). Johns Hopkins Press: Baltimore. Whitehead, A. N. (1967) Adventures of Ideas. The Free Press: Langer, S. K. (1972) Mind: An Essay on Human Feeling, vol. New York (Originally appeared in 1933). II. Johns Hopkins Press: Baltimore. Whitehead, A. N. (1968) Modes of Thought. The Free Press: McFarland, D. (1989) Goals, No-Goals and Own Goals. In: New York (Originally appeared in 1938). Montefiore, A./Noble, D. (eds) Goals, No-Goals and Own Volkelt, H. (1914) Über die Vorstellungen der Tiere. Engel- Goals. Unwin Hyman: London, pp. 39–57. mann: Leipzig.

Evolution and Cognition ❘ 197 ❘ 1999, Vol. 5, No. 2 Hans Lenk

Humans as Metasymbolic and Superinterpreting Beings

He who understands the oped the thesis of “the apes, does more for meta- Abstract second nature”, i.e., cul- physics than LOCKE. ture, as highlighting the CASSIRER characterized “man” (i.e., the human being) (Charles DARWIN1) constitutive “eccentric- as the symbolic being (animal symbolicum). This is ity” and homelessness of analyzed and criticized according to new research WAS THE PHILOSOPHER with wild living and trained primates with regard to humans by the depen- It Ernst CASSIRER who using tools and establishing local cultures as well as dence on this kind of sec- characterized “man”, the utilizing symbolic gestures and sign languages. The ond nature and the origi- human being as the human being therefore cannot be characterized any nal “homelessness of “symbolic animal” (ani- more just as the symbolic animal. Instead, it can be man” making cultural de- mal symbolicum), i.e., the demarcated from primates by its capability of ascend- velopment necessary. being which construes a ing and transcending to higher meta-levels of interpre- “Man” would even be world of symbols, orients tation, cognition and language. It is rather the meta- “the apostate, defector of itself in it and has to interpreting being (animal meta-symbolicum) and su- nature, the troublemaker, weave a net of symbols perinterpreting being (trans- and super-interpreting). attention-seeker, achiev- or symbolic net (1944), Yet, this characterization has to be integrated within ing being” with all respec- even living in a new a rather pluralistic (non-monolithic) practical anthro- tive consequences and dimension of reality, in pology including results of many other anthropologi- phenomena of a “ten- “a symbolic universe” cal disciplines. dency of self-aggrandize- (1944, p221). (S)he can- ments of life in terms of not confront reality Key words the power instinct” directly, but (s)he has to (PLESSNER 1975, p320). interstice this artiﬁcial Animal symbolicum, CASSIRER, animal meta-sym- The consequence of the bolicum, symbol usage, the meta- or supra-interpret- medium between him or basic interpretation of the ing being, toolmaking, primate behavior (research), her and reality being human as the constitu- chimpanzees (Washoe), bonobos (Kanzi), gorillas basically—by nature— tively eccentric being (lo- (Koko). dependent on symbolic cating itself besides its means and capacities of own center) which funda- representation (i.e., on culture). mentally has to distance itself from itself only sec- Already in antiquity, POSEIDONIUS stressed the idea ondarily working towards the unity of life under that “man” is by nature necessarily and indispens- drafts and anticipations, which has to plan and to ably dependent on culture (REINHARDT 1953; POEHL- “lead” its life: “man as an eccentrically organized MANN 1970). POSEIDONIUS would as probably the first being he has only to make himself that what he al- author use the expression ‘second nature’ as a central ready is”, by not merely executing or performing his concept of anthropology and of his doctrine of the life according to the “law of natural artificiality”, in origin of culture. HERDER used this concept later on an “artificial form of existence”, but also by “lead- in his theses of humans as beings characterized by ing” his life (PLESSNER 1975, p319). GEHLEN not only deficiencies, insufficiencies and incompleteness (by took up the thesis of the “deficient being” the way a contention already implicitly seen by (“Mängelwesen”) from HERDER, but also this thesis PLATO) which had been used for the development of of the “second nature of man” from PLESSNER: “ex- language and culture. In modern philosophical an- actly at that locus, where for an animal the ‘environ- thropology, Helmuth PLESSNER (1975, p301) devel- ment’ (‘Umwelt’) figures, stands as regards man the

Evolution and Cognition ❘ 198 ❘ 1999, Vol. 5, No. 2 Humans as Metasymbolic and Superinterpreting Beings cultural world (‘Kulturwelt’), i.e., the sector of nature symbols” which renders rational and relational superwhelmed by him and creatively altered by him thinking, human grasping, reflection and culture to become supporting means of life” (GEHLEN 1962, possible, since “without a complex system of sym- p38). GEHLEN also develops this insight within the bols relational thought cannot arise at all, much less framework of his philosophical doctrine of institu- reach its full development” (1944, p38). tions interpreting the function and the central na- Moreover, CASSIRER refers to HERDER’s theory of ture of institutions in the respective relieving import “reflexive thinking” with regard to the origin of hu- and in the functions of mediating as well as consti- man language and culture. All these fundamental tuting and constructing artificial world structures capacities and explanatory factors depend on the and norms, values etc. Culture is understood as sec- ability and function of using symbols so masterly ond nature. The “background” or primary nature in and intriguingly analyzed by CASSIRER in his main GEHLEN takes up the role as a quasi KANTIAN a priori work The Philosophy of Symbolic Forms (1923). Any behind the only knowledgeable and available “sec- grasping and “objectification”, even representations ond nature” shaped by cultural versions of the world of nature states are only to be covered by “construc- (i.e., by world representations, or world versions). Cul- tive process(es)”, dependent on “symbolic con- tural means and sorts of representation (media of structs” (CASSIRER 1944, pp160, 62). Even “all classi- grasping and representing) are of course symbolic: fication systems are artificial. Nature as such does only by symbolic representation we can and may contain only individual, multifarious phenomena. grasp, constitute and structure our world versions; By abstracting and summarizing as well as conceiv- the second nature is necessarily a “symbolic uni- ing (‘zusammenfassen’) these phenomena under verse”, it is spanned by CASSIRER’s “symbol net” concepts of sets and general laws, we do not describe (“Symbolnetz”). The cultural world is the symbolic natural facts. “Every system is a work of art” or symbolically shaped world (version). (‘Kunstwerk’)—a result of conscious creative activ- CASSIRER cites the traditional sign-theory after ity” (1944, p209). “Theoretical” work always means PEIRCE, MORRIS (1938) and S. K. LANGER (1942) ac- also “constructive work”: “spontaneity and produc- cording to which philosophy all symbols are not tivity is the very center of all human activities. It is only vicarious substitutes for objects as a traditional man’s highest power and it designates at the same interpretation of PEIRCE’s triadic sign theory, but a time the natural boundary of our human world. In “vehicle for the grasping of objects” in general serv- language, in religion, in art, in science, man can do ing to grasp, symbolize and interpret a thing or situ- no more than to build up his own universe—a sym- ation, beyond the mere behavioristic reaction or re- bolic universe, that enables him to understand and sponse: it is the “conception”, not the things, what interpret, to articulate and organize, to synthesize symbols directly ‘mean’ (LANGER 1982, pp60f)2. In and universalize his human experience” (1944, this approach CASSIRER’s theses are anticipated—at pp220f, my italics). Therefore, it is indeed the hu- least in some sense. CASSIRER however uses this in- man and according to CASSIRER (s)he “alone” (1944, sight for the foundation of an enlarged symbolistic p33) who is the symbolic being, the “animal symboli- philosophical methodology by characterizing man cum” (1944, p26). This symbolic being needs a “sym- as the only being which by “a symbolic imagination bolic universe”, into which it has to perform its pro- and intelligence” (1944, p33): therefore “we instead of jections, in which it would “live” (in a secondary defining man as an animal rationale, we should de- meaning) and by which “the symbolic net” is ex- fine him as an animal symbolicum” (1944, p26). He tended (1944, p25). relates his symbolistic theory also to Charles MOR- CASSIRER (cf. e.g., 1944, pp28, 30ff. a. e.) wanted to RIS’s elaborating and differentiating PEIRCE’s semiot- substantiate and support his philosophic anthropol- ics or sign-theoretic approach (1944, pp31ff)3. In- ogy of the symbolic being by a comparison with the trinsically, characteristic for the principle of this most modern behavior research with animals, espe- “symbolic” is among others “universality”, the gen- cially chimpanzees (Yerkes) attributing to these pri- eral “validity and general applicability”, “the magic mates the ability to use signs to utilize “designators” world, the Open Sesame!” which allows an entrance but indeed not the capability of a symbolic interpre- to the specifically human world, “to the world of tation (with universality, functionality, variability human culture” and “by architectural structure” and general language functions). (1944, p36), the “general symbolic function” of sym- This certainly was plausible at the time (1944) bols and language as well as multisidedness and vari- when CASSIRER wrote that, but this does not hold any ability: it is also only a highly “complex system of more nowadays, for in the meantime research on

Evolution and Cognition ❘ 199 ❘ 1999, Vol. 5, No. 2 primates is much more advanced. For instance, one neighboring one. BOESCH (1991) and BOESCH/BOE- taught chimpanzees like Washoe and gorillas like SCH (1990; 1991) observed chimpanzee mothers car- Koko to communicate (the latter one with more rying instruments as, e.g., a stone which they used than 1000 “words”, “concepts” or “meaningful ges- to open nuts (19 types of tool uses and 6 ways of tures”) in American Sign Language (A. and E. GARD- making tools were recorded with chimps in Taï Na- NER and Francine PATTERSON). These primates are tional Park, Ivory Coast). The opening was taught to able by using a functional quasi symbolic combina- the youngsters which had for quite a while really to tion of meaningful signs and sentence parts to com- learn and practice a certain kind of skill to be able bine simple sentences up to six words or so or even to crack a hard nut. The mother occasionally hid add and subtract small numbers (e.g., Sheba, after S. such a tool and carried it along—defending and/or BOYSON). There occurs a sort of generalizations, gen- reclaiming it even as some kind of “private prop- eralizations and self-reflective utterances (designa- erty”—when migrating to another feeding place. tion of the sign-talking chimpanzee as a subject her- This seems to be local “culture”—because there was self or himself by a symbolically represented also a spreading of this utilization of tools carrying expression). In addition, besides the gestures lan- in some sense symbolic functional meaning (in- guage plastic chips or symbolic notations on the cluding a certain limited generalization). Whereas computer display in a computer game are used in in Senegal wild chimpanzees also used sticks and some kind of combination game like “Scrabble” with branches to dig for termites (Wrangham), respective chimpanzees and bonobos as RUMBAUGH and SAV- competitive groups of chimpanzees in Uganda did AGE-RUMBAUGH successfully experimented with the not develop this “cultural” utilization of a sophisti- chimpanzees Austin and Sherman as well as with the cated tool, although they had also termite heaps real primate genius, the bonobo Kanzi, and his sister available. SUGIYAMA (1981) in Cameroon even ob- Panbenisha (SAVAGE-RUMBAUGH et al. 1998). served chimpanzees using tools to make tools with Washoe, Koko and Kanzi not only occasionally several functions: like using a sharp cutting stone or played the fool, but also made a fool of the respective so to make a point for boring at one end of the twig interviewing partner or experimenter. For instance, and to hammer with a stone the other end to form Koko, the gorilla, communicated with American a certain kind of brush to retrieve the termites by Sign Language the wrong answer “red” of a white putting this end into the hole bored before (cf. handkerchief, laughing and necking the experi- BECKER 1993, p102). Therefore, primates not only menter PATTERSON. The latter one was annoyed and use tools, but they also understand a certain kind of wanted to give up already. Koko then picked all of a general function of tools and they even use tools to sudden a very tiny red thread from the handkerchief make tools, thereby falsifying FRANKLIN’s thesis and and signalled triumphally laughing “Red, red, red!”. definition of man as the (only) toolmaking animal. Chimpanzees and bonobos are not only able to Primates are therefore able to generalize and again use and even produce tools (e.g., to put together individualize functions of learned utilizations of bamboo sticks to get at bananas in the distance as tools and also to specify individualized meanings of W. KOEHLER had already earlier substantiated in his symbols and gestures by establishing little local-cul- famous primate experiments), but they are also able tural traditions etc. They even discover or invent creatively to apply tools to a respective situation type new artifacts or activities which in a limited sense are and single situations. They are also capable of using spread within the local colony or even neighboring in a limited sense at least means of representation groups. That is to say, the utilization of tools and to transfer and in a sense generalize them referred to symbols is not only characteristic for man, but also other situations. They are able consciously to make in a limited sense to be found in wild living primates. a fool of somebody (see above), to betray or deceive (To be sure, primates like chimpanzees or bonobos partners and rivals (DE WAAL 1983, 1990; SOMMER in zoos or experimental situations are much more 1992). They even developed an elementary “cul- extensively taught tools and symbols, but this is cer- ture” by establishing a certain tradition of learned tainly an artificially arranged situation, which is not and discovered uses of instruments and tools. Fa- so telling here in the first place, although it drasti- mous is the macaque “girl” Imo in Japan which had cally may refine and extend the use of symbols and discovered that potatoes would taste better after tools extraordinarily (see FOUTS-MILLS 1997; SAVAGE- they had been washed in salt water. Imo’s washing RUMBAUGH et al. 1994; 1998). By the way, some anal- activity soon spread out over the whole colony of yses of animal languages like the one of dolphins the respective group of macaques and even to a studied by LILLY, PRYOR and HERMAN or pairwise in-

Evolution and Cognition ❘ 200 ❘ 1999, Vol. 5, No. 2 Humans as Metasymbolic and Superinterpreting Beings dividualized speech or singing recognition patterns symbols in a higher-level representative form and with birds seem to substantiate the use of communi- analysis. Only man or woman is the being who is able cation gestures and symbols also with other species to ascend to higher symbolic levels in a rather unlim- (see e.g., GRIFFIN 1984). ited manner. To my mind, the human being is the That means that higher animals like primates and only being who may—without restriction—establish some others engender signs which they can in a lim- levels and metalevels of symbols and signs to symbol- ited sense and with respect to classified situations ize and refer to lower-level symbols and represented use in a symbolic way or even generalize in some objects (again in the form of a kind of object repre- (though rather restricted) sense. They use, to a cer- sentation).(S)he is able to perform a sort of “symbolic tain degree, symbols. Therefore, they are also within ascent” to abstract and go to higher-levels, to vary, some limits “symbolic beings” or in a at least capable again “objectify”, i.e., symbolically and abstractly to understand sign combinations as symbols or sym- designate lower-level activities and symbolizations bolic way and to use them more or less systematically in the form of new, higher-level symbolization pro- for the control of behavior and reaction. This is es- cesses and activities. Procedures of checking, evalu- pecially true for the highest primates like chimpan- ating, controlling and planning action in the sense zees, gorillas and the bonobos as has been exten- of an anticipatory and situation transgressing “antic- sively substantiated and confirmed by observation ipation” towards future behavioral and action pat- of wild living primate groups or experiments with terns are necessarily represented or incorporated by trained primates (see e.g., SOMMER 1992; DE WAAL symbols in processes of symbolization (symbolic rep- 1983, 1998;GRIFFIN 1984; FOUTS-MILLS 1998; SAVAGE- resentation) within cultural activities of structuring RUMBAUGH 1998; for tool use cf. BECKER 1993). The and interpreting norms and signs. To make it short: activities imply certain symbolic representations for the human being is not only the sign using symbolic grasping their functional “meaning” and manipu- being inventing, fashioning, varying and utilizing lating as well as controlling behavior—even includ- symbols and representing by symbols, but the being ing self-reference, reflexivity, and (sign) reciprocity which projects, establishes and changes symbols as (at least with chimpanzees, bonobos and orangu- the object of higher-level symbolization processes, tans)4, 5. That is to say, CASSIRER’s restriction of the use i.e., interprets these construed artificial “objects” on of symbols and the capacity to establish and gener- a metalevel. Man is not only the being who uses sym- alize as well as individualize symbolic articulations, bolic forms, but the one reflecting and projecting as representations and networks as being unique for well as varying symbols about and over symbolic humans and his attempt to characterize “man” by signs and symbols as well as symbolization processes. this symbolic ability or function turns out to be too The human being therefore is not only the animal rash, if not wrong. This thesis has to be modified. symbolicum but the animal meta-symbolicum, being Higher animals like primates indeed use signs and the being which not only interprets, after NIETZSCHE, objects in a symbolic or ritual manner and they may but who in turn interprets its interpretations and in- also in a limited sense perform transfers or generali- terpretation processes as well as the respective sym- zations and some kind of inferential discrimination bolizations and representations as higher-level and as well as simple logical inferences like exclusions or abstract “objects” and again analyzes these, attaches even contrapositions, but they certainly do not again meanings to these etc. This creature called human is analyze these symbols or symbolically “understood” the being capable of not only creating and using sym- functioning signs as objects of a symbolic analysis on bols and metasymbols as well as interpretations and a higher level (metalevel). This is particularly true for metainterpretations—(s)he is the animal super(inter- the gestures and signs as well as activities which are )pretans et symbola symbolisque creans. (S)he trans- not genetically fixed but “culturally” developed, so gresses the realm of mere sign utilization and sym- to speak. They do not make the symbols and the uti- bolic functions by trans- and superinterpreting, sym- lization of them including the rules of symbol use bolically transcending these forms and usages in a and symbolic functions again an object of a higher reflexive manner. (S)he is the being potentially able (meta)-level symbolic representation. They do not to transgress any levels of representation towards analyze and symbolize the application of signs per se ever-accumulating higher metalevels varying, pro- again. This seems to be a characteristic trait for the jecting, reflecting or rejecting “meaning”, schematiz- human being that (s)he is able to not only apply sym- ing and structuring as well as ordering—in short: in- bols on a higher metalevel but also to designate, terpreting—representations of the (mostly “mean” and interpret the utilization and function of proximate) next-lower level. The human being is the

Evolution and Cognition ❘ 201 ❘ 1999, Vol. 5, No. 2 being of meta-levels. (S)he does not only go beyond derstandings and difficulties of level interpretation certain sets and extensions of sets within one and the etc., it seems to be better to talk of “animal metasym- same level of designation and denotation but poten- bolicum or the transinterpreting and suprainterpret- tially and tendentiously is able to transcend beyond ing being” or “superinterpreting” (combining trans- different meta-levels of language by higher-level and supra-interpreting) being in general. symbolization and interpretation. If we refer to the Herewith, it is certainly not hypothesized or ex- transgressing of sets within one and the same level plicitly contended that this anthropological feature by the Latin syllable ‘trans’ and to the ascending to would be the only one: earlier (LENK 1983; 1986; higher-levels by ‘super’ or ‘supra’ respectively, we can 1989) I tried to develop a practice-oriented philo- call the human being the trans-interpreting and/or su- sophical anthropology of rather pluralistic prove- per- or supra-interpreting being, respectively, or for nience, based not on just one unique and only an- short: the meta-interpreting being (see LENK 1995; 1998, thropological factor, but a multifactorial and pp38ff). The human being does not only interpret multifunctional philosophical anthropology tak- within one level of interpretation and sign utiliza- ing into consideration many results of empirical tion, (s)he is not only the interpreting being re- anthropological disciplines. A philosophical an- stricted to just one symbolic plane, but the level- thropology cannot any more merely highlight one transgressing interpreting being and by that in turn unique trait in a monolithical manner or just rely really the reflecting being. Abstract reflections are on constitutive self-production and artificial self- only possible if you are able to transcend the actual identification of the human being in a traditional level at hand, if you transgress the levels by going idealistical–”subjective” or transcendent or a rei- meta-symbolic or superinterpreting. Therefore, it is fied transcendental understanding, but it has to most plausible to talk of the human being as the “an- take into account many and different functions of imal metasymbolicum” (a sort of extension of CAS- the so-called “second nature” or “symbol world” SIRER’s terminology) or as the superinterpreting being (including within the symbolic net higher- and (in extension of NIETZSCHE’s conception of the inter- lower-level symbols in a rather architectonic order preting being). (We could conceive of superinterpret- of meta-level character). The capability of tran- ing as including trans- and supra-interpreting, i.e., scending and ascending to ever higher-levels and not only to another set or denoting symbol within meta-levels of interpretation may be a unique sign the same level of denotation, but also to higher-level of delineation and demarcation of the human be- representations.) ing from primates well-suited to demarcate the Meta-levels and even methodological interpreta- higher(-level) human culture as a realm of meta- tions or analyses of reflective and interpretative lev- symbols and meta-interpretations, but this does els (so to speak meta-levels of interpretations) can not stand out as the only and unique function for well be coordinated with or even integrated within the constitution, development, and foundation of this approach. The interpretative levels of interpre- our “second nature” (i.e., “culture”). This feature of tation using higher-level symbols for representation meta-interpretation is to be sure central and indis- is certainly open with respect to the ever-transgress- pensable, essential and unique for man, but it is ing and ascending leveling of symbolizations up- certainly not the only characteristic feature of man. wards. One may again in a cumulative combination To summarize again: One may understand and de- so to speak incorporate these higher-level interpre- marcate, or distinguish, man from higher animals— tations in the talk of the “interpreting being” and including especially primates—by identifying and stay with NIETZSCHE’s terminology, but then one has characterizing the human being as the animal meta- explicitly to add that the capacity of interpretation symbolicum or as the meta-representative (meta-rep- in humans is not restricted to just one level of sym- resenting) or meta-interpreting or trans- and supra- bolization. NIETZSCHE and CAS- interpreting (“super-interpret- SIRER certainly would have un- ing”) being. It or (s)he is the derlined this capacity and Author’s address meta-level being of represen- would have included the tations and interpretations— Hans Lenk, Universität Karlsruhe Institut higher-level symbolic func- as well as the being of meta- für Philosophie, Geb.Nr. 20.12; Bau II am tions and utilizations in their Schloß, Postfach 6980, D-76128 Karlsruhe, cognitions and meta-ac- approaches. However, to Germany. Email: [email protected] tions—in short: the meta-level avoid terminological misun- being.

Evolution and Cognition ❘ 202 ❘ 1999, Vol. 5, No. 2 Humans as Metasymbolic and Superinterpreting Beings

Notes in front of a mirror. They even use a mirror in a goal-oriented manner, “by intention” so to speak, as an aid to clean, check 1 Notebook, my re-translation. and inspect hidden body parts (after GALLUP in STRAUSS/ 2 LANGER replaced the traditional semiotic triadic theory of GOTHAL 1991/92). However, gorillas do not recognize them- signs by a tetradic one in which beside the symbol, the selves individually in a mirror image. subject and the object indeed also “conception” figures as Lately, however, experiments with video movies moderated an important instance (1982, p64). She distinguished strict- this interpretation: Only in directly concomitant replays of ly between “sign meaning”, and “symbol meaning” the videos chimpanzees could recognize themselves—not if the latter one rather referring to “connotations”, namely to an replaying was postponed by even a relatively short time (cf. expression of graspings of meaning instead of denotations Frankfurter Allgemeine Zeitung, April 7, 1999). Like with or objects referred to. Indeed, according to her there is no human children (are only) under the age of three it seems “real denotation” whatsoever in the symbolic process of that chimps probably lack a sense of self-continuity or of grasping, but always only “connotation-mediated mean- (autobiographical) self-recognition within a movie repre- ing” (1982, pp65f). sentation over time: They apparently cannot individually 3 CASSIRER (1944, p32) conceives, after MORRIS, of “signals” as reidentify themselves on the movie-screen if the video is “operators, symbols are ‘designators’”. Signals, even when replayed a while after it had been taken whereas they understood and used as such, have nevertheless a sort of seemed to be able skillfully to do the reidentification on- physical or substantial being; symbols have only a func- time with the filming and simultaneous replaying on a tional value”. screen. The experimenter D. J. POVINELLI (in FERRARI 1998) 4 For the function of “property” and the “principle of prop- concludes that the chimpanzees would not infer “that’s me” erty” among longtail macaques and chimpanzees see KUM- but possibly only “that’s like me”. Again, the upshot seems MER (1982). to be (like SAVAGE-RUMBAUGH’s result) that these (adult) pri- 5 We know that chimpanzees and orangutans (as well as mates (chimpanzees and bonobos) are stuck at a develop- bonobos) are indeed able to recognize themselves on-time mental stage compared to a human child at 2.5 or 3 years.

Bibliography Langer, S. K. (1942) Philosophy in an New Key. A Study in the Symbolism of Reason, Rite and Art. Harvard University Becker, P.-R. (1993) Werkzeuggebrauch im Tierreich. Hirzel: Press: Cambridge MA (quoted from 19823 ed.). Stuttgart. Lenk, H. (1983) Wie philosophisch ist die Anthropologie? In: Boesch, C. (1991) Teaching among wild chimpanzees. Ani- Frey, G./Zelger, J. (eds) Der Mensch und die Wissenschaft mal Behaviour 41: 530–532. vom Menschen. Bd. 1. Solaris: Innsbruck, pp. 145–187 (re- Boesch, C./Boesch, H. (1990) Tool Use and Tool Making in print in Lenk 1986, pp. 131ff, and partly in Lenk 1987, pp. Wild Chimpanzees. Folia Primatol 54: 86–99. 112ff). Boesch, C./Boesch, H. (1991) Dim Forest, Bright Chimps. Lenk, H. (1986) Zwischen Wissenschaftstheorie und Sozial- Natural History 9, Vol. 100: 50–56. wissenschaft. Suhrkamp: Frankfurt. Byrne, R. (1995) The Thinking Ape. Evolutionary Origins of Lenk, H. (1987) Zwischen Sozialpsychologie und Sozial- Intelligence. Oxford University Press: Oxford, New York. philosophie. Suhrkamp: Frankfurt. Cassirer, E. (1923) Philosophy of Symbolic Forms. 3 vols. Yale Lenk, H. (1989) Anthropologie nach vorne: Bausteine der University Press: New Haven (quoted after German trans- philosophischen Anthropologie als einer Interpretations- lation, Philosophie der symbolischen Formen 1956, 2nd und Konstruktdisziplin. In: Bonk, S./Lazzari, A. (eds) Ideen ed, 1977, 7th ed (II), 1964, 4th ed (III)). zu einer integralen Anthropologie. Festschrift Mácha. Mu- Cassirer, E. (1944) Essay on Man. Yale University Press: New nich, pp. 89–100. Haven. (2nd German translation: Frankfurt 1990). Morris, C. W. (1938) Foundations of the Theory of Signs. Uni- Cheney, D. L./Seyfarth, R. M. (1990) How Monkeys see the versity of Chicago Press: Chicago. World. Inside the Mind of Another Species. Chicago Uni- Plessner, H. (1928) Die Stufen des Organischen und der Men- versity Press: Chicago, London. sch. Berlin, de Gruyter: New York. (quoted from the 1975 De Waal, F. (1983) Unsere haarigen Vettern. Neueste Erfahr- ed). ungen mit Schimpansen. Harnack: München. Poehlmann, E. (1970) Der Mensch—das Mängelwesen? Ar- De Waal, F. (1990) Peacemaking among Primates. Harvard chiv für Kulturgeschichte 52: 297ff. University Press: Cambridge MA. Reinhardt, K. (1953) Poseidonius. In: Real-Enzyklopädie der De Waal, F. (1995) Die Bonobos und ihre weiblich bestimmte Altertumswissenschaften, vol. 22, pp. 19ff. Gemeinschaft. In: Spektrum der Wissenschaft, Mai: 76–83. Savage-Rumbaugh, S./Lewin, R. (1994) Kanzi: The Ape at the De Waal, F. (1998) Die Bonobos. Birkhäuser: Basel. Brink of the Human Mind. Wiley: New York. Ferrari, M. (ed) (1998) Self-awareness. Guilford: New York. Savage-Rumbaugh, E. S./Shanker, S. G./Taylor, T. J. (1998) Fouts, R./Mills, S. T. (1997) Next of Kin: What Chimpanzees Apes, Language and the Human Mind. Oxford University Have Taught Me About Who We Are. Morrow: New York. Press: New York, Oxford. Gehlen, A. (1962) Der Mensch: Seine Natur und seine Stel- Sommer, V. (1992) Lob der Lüge. Täuschung und Selbst- lung in der Welt. Athenäum Bonn (Originally published betrug bei Tier und Mensch. Beck: Munich. 1940). Strauss, J./Gothals, G. R. (1991/92) The Self. Springer: Heidel- Griffin, G. R. (1984) Animal Thinking. Harvard University berg, New York. Press: Cambridge MA. Sugiyama, Y. (1981) The Brush-Stick of Chimpanzees found Kummer, H. (1982) Social knowledge in free-ranging pri- in South-West Cameroon and their Cultural Characteris- mates. In: Griffin D. (ed) Animal Mind—Human Mind. tics. Primates 26 (quoted after Becker 1993). Springer: Berlin, pp. 113–130.

Evolution and Cognition ❘ 203 ❘ 1999, Vol. 5, No. 2 Theodore R. Schatzki

Humanism on the Ropes?

The two of us wrote Anti- Abstract There is something, as Oedipus together. Since indicated, to this injunc- each of us was several, there This essay examines what of humanism remains tion. Today we know too was already quite a crowd. when one incisive “posthumanist” critique of the much about people to (DELEUZE/GUATTARI 1987, uniqueness of human beings is digested. This critique, leave the free agent of p3) issuing from a stream of contemporary thought that Western lore unim- can be labeled “social multiplicity theory,” prolifer- peached as either a defini- HIS MANNER OF DIS- ates agency while also dissolving the human variety. tive portrait of human be- T course is familiarly Part one describes the social multiplicity theories of ing or the unique locus of plausible. At least since Michel FOUCAULT, Gilles DELEUZE and Félix the do-ing called FREUD it is common stock GUATTARI, and Bruno LATOUR and Michel CALLON. “agency”. But does aban- that conscious agents do Parts two and three examine two of their critiques of doning the majesty and not fully master the whys humanism. One denies the uniqueness of humans as uniqueness of humans en- and wherefores of their agents by analyzing actors as anything to which do- tail the leveling of entities activity. Some aspects of ing is attributed. The second analyzes human agents and the evisceration of a the psyche escape self- themselves as multiplicities of subagents. I contend distinctly human manner monitoring and can even that the linguistic multiplication of agency does not of shaping the world? Can oppose consciously hewn challenge the distinctiveness of the human sort, and there be a chastened and designs. Many thinkers that the claim about subagencies does not undermine modest humanism that today avow, as a result, a the integrity of human agency. The conclusion up- retains a distinctive place multiplication of the holds the viability of a humanism sensitive to the infor humans in the cosmos agencies that bear on sights of multiplicity theory. instead of centering the human conduct. cosmos around human At the same time, the Key words beings? sentences, however hu- The current essay urges Humanism, agency, human agency, types of agents, morous, do not seem a positive answer. It inter- doing, social multiplicity. quite plausible. Anyone rogates three mutually-re- who has tried knows the inforcing contemporary tribulations of coauthorship. If two can be trouble, “posthumanist” accounts of agency, which share a crowd poses a real problem. If there actually had above all one crucial characteristic: They are social been a crowd, Anti-Oedipus probably would not have theoretical harbingers and practitioners of an inci- been written. So there really wasn’t a crowd. And we sive metaphysics of multiplicity that discerns disper- don’t think twice before thanking just two individ- sion where sameness was once assumed. The ways, uals for their efforts. consequently, they multiply agency and dissolve the Yet, would it be right to assume that the only human variety are part and parcel of much larger and agency responsible for the book is the conscious portentous pictures. As a result, the subtext of this intentional agency of its two named authors? Many essay is, whether an ontology of multiplicity toler- contemporary writers assert that myriad other en- ates an updated and clearer-sighted humanism. tities are actors, among them nonhumans, parts of humans, and collections of humans. Denying this, 1. Social ontological multiplicity it is said, is anthropocentric. Worse, it is humanism, that roundly denounced modern ideology that One significant development in 20th-century phi- treats human beings as something unique and spe- losophy has been a challenge to the age-old domina- cial. tion of sameness and unity. In a variety of guises,

Evolution and Cognition ❘ 204 ❘ 1999, Vol. 5, No. 2 Humanism on the Ropes? philosophers have questioned whether the entities dispersion as events and in the occurrence that is with which we deal, the worlds in which we live, and proper to them. Before approaching, with any degree we ourselves including our actions, thoughts, and of certainty, a science, or novels, or political concepts are marked by well-defined and stable speeches, or the oeuvre of an author, or even a single essences, substances, meanings, and space-times. book, the material with which one is dealing is, in Such long-standing Greek figures as essence and its raw, neutral state, a population of events in the accident, species and genre, the one and the many, space of discourse in general” (FOUCAULT 1972, and form and matter have suffered withering pp26–27). assaults; assaults challenging, inter alia, their reduc- On this gambit, the being of a discourse lies in tionist penchant, the well-definedness and stability whatever connects dispersed statements either into of the entities they invoke, their adequacy to the the discourses from which FOUCAULT begins or into diversity of things, and their capacities to capture hitherto unrecognized unities. FOUCAULT explains becoming and to do justice to the particulars sub- that what he once thought effected unification can- sumed under them. Perhaps the most renowned and not do so, namely, relations among or the objects, influential instance of such critique is Ludwig WIT- concepts, theories (etc.) referred to and utilized in TGENSTEIN’s conception of family resemblances: statements. For the relations, objects, and concepts Instances of any natural language concept that is not tied to the statements of any conventionally recog- introduced through explicit definition fail to share nized discourse form irreducible dispersions. Just as any specific property—a unifying essence—and are on WITTGENSTEIN’s analysis the instances of a con- instead linked by an array of crisscrossing similari- cept do not share a common property, so, too, are ties that yield overall “family resemblances” among discourses not demarcated by common objects and them. Where traditional philosophy sees sameness concepts. FOUCAULT (1972, p37) then wonders and continuity, WITTGENSTEIN sees multiplicity. whether these dispersions themselves might hold Instances of a concept form a dispersion, a distribu- the key to a solution: tion of similar but ultimately disparate phenomena “Hence the idea of describing these dispersions that do not as a group share anything in common. themselves; of discovering whether, between these In the 1960s and 70s, such thinkers as Jacques elements, … one cannot discern a regularity: an or- DERRIDA, Michel FOUCAULT, and Thomas KUHN of- der in their successive appearance, correlations in fered analyses of concepts and meaning parallel to their simultaneity, assignable positions in a com- or supportive of WITTGENSTEIN’s. In the meantime, mon space, a reciprocal functioning, linked and hi- moreover, a growing line of thought has expanded erarchized transformations.” the notion of dispersion into a potent social ontol- The being of a discourse is tied to regularities in its ogy. FOUCAULT’s The Archaeology of Knowledge was objects and concepts; more precisely, it is tied to the ground-breaking in this regard. The book’s leading rules, defined by FOUCAULT as the conditions to question runs, What is the being of a discourse? In which dispersions submit, that govern these regular- pursuit of an answer, FOUCAULT begins by provision- ities. A discourse, consequently, is a multiplicity of ally accepting such conventionally designated dis- statements whose objects, concepts, relations etc. courses as those of natural history, political econ- form rule-governed dispersions. It is, in short, a con- omy, and medicine. He rejects, however, the stellation of dispersions: of statements, objects, con- phenomena traditionally countenanced as unifying cepts, theories, relations, subject positions, and and demarcating these discourses and instead pur- rules. sues a radical strategy. Much as HUSSERL in his epoché A second landmark in the development of what places the reality of the external world at bay, FOU- can be called “social multiplicity theory” is DELEUZE CAULT suspends the assumption that discourses are and GUATTARI’s A Thousand Plateaus. This work pro- unified wholes. He dissolves their alleged common- claims multiplicity the central ontological concept. alities and continuities, erases the boundaries that It is also the key phenomenon opaque to “arbores- seemly demarcate them, and resolves discourses as a cent” thought, that dominant form of thought that set into a field of naked discursive events: a space of comprehends a set of phenomena by relating it to statements per se devoid of relations, principles, or one or more common and often underlying sources, contexts to connect, group, or explicate them. essences, elements, or reference points (DELEUZE/ “Once these immediate forms of continuity are GUATTARI 1987, p4) In misconstruing dispersion by suspended, an entire field is set free… [It] is made up reconstructing it as a ramification of a common es- of the totality of all effective statements … in their sence or source, arborescent thinking reduces it to

Evolution and Cognition ❘ 205 ❘ 1999, Vol. 5, No. 2 Theodore R. Schatzki

variation of the same. For DELEUZE and GUATTARI’s zations of statements) that occupy a territory and “rhizomatic” thought, by contrast, multiplicity is a embrace individuals, groups, or things that flee its substantive, indeed THE substantive; it is not an ad- orders and thereby threaten it with disintegration. jective of unity. Any social formation is thus a multiplicity of multi- “…this substantive … was created precisely in or- plicities: of contents, expressions, vanishing points, der to escape dialectics, to succeed in conceiving the and lines of segmentation that define territoriality. multiple in the pure state, to cease treating it as a A third key conception of social multiplicity is the numerical fragment of a lost Unity or Totality or as theory of the actor-network principally developed the organic element of a Unity or Totality yet to by Michel CALLON and Bruno LATOUR. This theory come…” (DELEUZE/GUATTARI 1987, p32) contends that the primary medium of a person’s life Instead of illusorily seeking the unity of disper- is a multileveled and interconnected array of net- sion, rhizomatic thinking maps its dimensions. Uni- works, each an association of heterogeneous types of ties exist for it only as either the singularity of words phenomena. A student, for instance, is embedded in that designate dispersed phenomena (à la WITTGEN- networks that embrace professors, classrooms, ex- STEIN) or the singularity of self-conscious subjects aminations, bicycles, apartments, pets, desires, hair- who stand over against the world: styles, sleepless nights, fast food, peer pressure, and “The notion of unity appears only when there is the like. The network array of this sort in which a a power takeover in the multiplicity by the signifier student is suspended defines his or her situations, or a corresponding subjectification proceeding… identities, allies, and possibilities. Furthermore, such Unity always operates in an empty dimension sup- traditional social theoretic phenomena as power, or- plementary to that of the system considered (over- ganizations, and society are mere side effects of net- coding).” (DELEUZE/GUATTARI 1987, p8) works and do not contribute to the dynamic reality In contrast to wholes or totalities, whose unity is of network building and dissolution. Processes pecu- an additament to the things unified, multiplicities liar to networks are instead what is responsible for have no dimensions over and beyond their lines their evolution, for example, association and black (components); their components fully fill, or oc- boxing (treating networks as units). cupy, their dimensions by way of constituting them. Networks comprise two primary sorts of multi- Multiplicities, consequently, cannot be treated as plicity. First, the types of material out of which they emanations, versions, or determinations of some are composed are heterogeneous. In opposition to common point, source, or entity (overcoding). They practically all social thought, actor-network theory are acentered and irreducible distributions. refuses to identify basic elements of social life. Just This conception of multiplicity informs DELEUZE as multiplicities for DELEUZE and GUATTARI, in occu- and GUATTARI’s general metaphysics. A central fea- pying all their dimensions, proscribe reductive ture of reality on their view is pure process-becoming overcoding of their components, so, too, is for ac- multiplicity made up of continuums of intensities, tor-network theory the multiplicity of network ma- conjunctions of flows, and emissions of “particles– terials a positive reality irreducible to a list of pri- signs” (in their terms: the plane of consistency, or mary stuff. Taking a vacation at a distant lake, for the body without organs). Out of this plenum coa- instance, can involve a network combining com- lesce various strata, on each of which these dynamic puters, alloys, jet engines, research departments, dispersions receive determination in a “double artic- market studies, advertisements, welcoming host- ulation” of contents and expressions. On the anthro- esses, natives, fishing boats, luggage, and so on in- pomorphic stratum, contents are things done and terminably (see CALLON 1991, p139). Attempts to expressions things said. On any stratum, moreover, reduce such burgeoning lists to categorical dimen- the articulation of contents and expressions is ef- sions falsify that as which entities enter and func- fected in and through multiple assemblages, each of tion in networks. which (1) combines mutually presupposing con- Actor-network theory further maintains that each tents and expressions, (2) occupies a territory, and entity in a network is itself a multiplicity that is (3) contains “lines of flight,” which are phenomena viewed as a unit by other network individuals and already fleeing and thus breaking down the assem- entities. A computer, a research department, an ad- blage. On the anthropomorphic stratum, assem- vertisement, and a welcoming hostess, for instance, blages are social formations: configurations of (a) are all multiplicities that are taken to be single enti- power regimes (organizations of activities and bod- ties by the users, travellers, readers, bosses, and em- ies–things) and (b) regimes of enunciation (organi- ployees who network with them. Only on such oc-

Evolution and Cognition ❘ 206 ❘ 1999, Vol. 5, No. 2 Humanism on the Ropes? casions as breakdown, deliberate scrutiny, or change among subject positions that are furnished in dis- in interest do the multiplicities involved come to courses. (For present purposes, think of these posi- attention and thereby become network components tions as both roles and occupations such as mother in their own right. According to actor-network the- and banker as well as ways of being such as listener ory, consequently, networks embrace heteroge- and speaker.) Subjects are “dispersed” among such neous materials that are themselves networks of en- positions in that their erstwhile self-sameness is tities. replaced by a multitude of discontinuous positions. FOUCAULT, DELEUZE and GUATTARI, and LATOUR Part of the import of the notion of a discourse posi- and CALLON all claim that dispersion is central to the tion, moreover, is that what is entailed in occupying social, and further concur that social phenomena are one is specified by the discourse providing it. When configurations of heterogeneous multiplicities. Refusing someone, for instance, occupies the position to reduce variation and difference to sameness, they “author” that is found in certain modern discourses, insist that social formations can be grasped only via that the name of the author performs a classifica- maps of their heterogeneous constitutive multiplic- tory function vis–à–vis texts and that to be the ities, not by reference to origins, single basic phe- author of a text is to own it, to have a “deep motive” nomena, or common reference points. In arguing for producing it, or to be the source of the expres- thus they echo WITTGENSTEIN’s contention that un- sion manifested in it—these matters are specified by derstanding a concept requires an overview of the the discourses concerned and do not result from use of words, not, as traditionally supposed, a grasp anything a particular author is or does (see FOU- of essence. CAULT 1984). Now, social thought has long deemed agency cen- FOUCAULT’s analysis is usually taken to epitomize tral to social affairs. Since multiplicity, for the above contemporary “decenterings” or “fragmentations” theorists, marks everything the social comprises, of the subject. It is important, consequently, to agency enters its province where it is proliferated in point out that it fragments only one dimension of two basic ways. Following FOUCAULT’s conception of the subject, the discursive. Consider the following the dispersion of the subject into multiple discourse sentence: “Thus conceived, discourse is not the ma- positions, one line of thought, found above all in jestically unfolding manifestation of a thinking, actor-network theory, cleaves the occurrence and knowing, speaking subject, but … a totality, in nature of agency to the multiplicity of attributions which the dispersion of the subject and his discon- thereof. A second line of thought, found both in tinuity with himself may be determined.” (FOU- DELEUZE and GUATTARI and actor-network theory and CAULT 1972, p55) Despite appearances, this sen- again presaged in FOUCAULT’s treatment of the sub- tence does not aver that the subject is nothing but a ject, analyzes agents themselves as multiplicities. discontinuous pastiche of positions. For the subject Both lines, by scattering agency into either the va- involved, the subject spoken of throughout The Ar- garies of discourse or the heterogeneities of net- chaeology, is a living, bodied, and acting person who works, license and encourage the attribution of is never simply whatever disparate positions it as- agency to nonhuman entities, not just living organ- sumes when speaking, writing, listening, watching isms but artifacts and inorganic things as well. Com- and so on. Evidence for this interpretation is found, bined, moreover, they transform all entities into po- inter alia, in the fact that throughout the book, tential agents, while also tying human activity to the without adequate explication, FOUCAULT calls dis- determination and support of subhuman and non- courses “discursive practices”. It is only on pages human thing–agents. Herein lies the central chal- 208–9 that the reason for this expression becomes lenge of multiplicity thinking to those humanisms clear: A discourse’s statements are things people do, that trumpet the uniqueness of humans as doers, i.e., actions. So the rules that govern a discourse’s effectors, and instigators. dispersed objects, concepts, positions etc. are in fact rules that govern discursive activity, although they 2. Discourse multiplicity are “not so much limitations imposed on the initiative of subjects as the field in which that initiation The discourse multiplicity approach to agency take is articulated, … rules that it puts into operation, … its lead from FOUCAULT’s conception of discourse relations that provide it with a support… [Mine] is positions. In The Archaeology of Knowledge, FOU- an attempt to reveal discursive practices in their CAULT counters the once dominant idea, that sub- complexity and density; to show that to speak is to jects are the source of discourse, by dispersing them do something.” (FOUCAULT 1972, p209; cf. p200)

Evolution and Cognition ❘ 207 ❘ 1999, Vol. 5, No. 2 Theodore R. Schatzki

Subject positions, in short, are statuses people is relative to who describes the network together assume in their capacity as discoursing actors, as with the discursive conventions he or she cites in producers and consumers of statements, a capacity doing so. that embraces perception. Since they are actors, For actor-network theory, consequently, the or- however, they are always more than discoursers dering of networks around agents and nonagents is alone. As detailed in FOUCAULT’s subsequent writ- an effect of imputations of these statuses. On this ings, and as already suggested by the acknowledg- view, both the character of agency and what qualifies ment in The Archaeology that the discoursing sub- as an agent or nonagent depends on the imputations ject is a perceiver, people are also (1) performers of involved. For John LAW, an actor-network theorist nondiscursive actions and (2) bodily creatures pos- under the influence of FOUCAULT, moreover, such im- sessing and possessed by physiology, anatomy, sen- putations come organized in the form of discourses. sations, and pleasures. It is true that these later writ- This entails that what qualifies as an agent derives ings analyze body and nondiscursive activity as from discourse. For example, whereas in some dis- socially molded phenomena, just as The Archaeol- courses events result from indomitable structural ogy analyzes discourse positions as similarly forces, in others even the same events arise from the shaped. The social entities involved, however, are efforts of individuals, or the struggles of collectives. no longer discourses but biodisciplinary appara- In LAW’s (1994, p74) words, “…it is possible to im- tuses, which comprise nondiscursive activities and pute several modes of ordering to the talk and actions architectures in addition to discourses (see, e.g., of managers. And I’m saying that people are written FOUCAULT 1980). into them in varying degree… So I am saying that Actor-network theory applies discursive disper- agents are effects which are generated by such modes sion to agency. It treats “actor” (or “agent”) as a of ordering. The subject has been decentred.” discursive status, though not so much one that dis- Construing an agent as anything to which do-ing coursers occupy as one that is attributed to entities is attributed obviously abets the idea that nonhu- discursively. Bruno LATOUR, for example, analyzes man entities are agents. Graphite rods, computers, actors as actants, where an actant, roughly speak- alloys, power plants, black holes, tornados, expert ing, is anything said (in a story) to do something. systems, people, dogs, scallops, viruses, jungles, you Anything that as narrated has effects, brings some- name it—it qualifies as an agent if there exists a dis- thing about, impacts this or that, or makes a differ- course in which it is said to do something. ence in some way, counts as an actor. LATOUR (1991, In one regard, I believe, this position is unassail- p122) more picturesquely defines an actor as a sta- able. Doing is the core feature of agency. Indeed, bilized “list of answers to trials [of strength]”: a sta- agency is roughly equivalent to doing. To say that Y bilized concatenation of performances, occurring is attributable to the agency of X is to say that X in different agonistic states of affairs (each embrac- either did Y or did something that directly effected ing multiple actions), that is attached to a name and Y. If Y is an action, the agent is whatever did it, that thereby attributed to a substance. is, carried it out; if Y is a change in the world, such Meanwhile, Michel CALLON (1991, p140) writes as a shovel’s change of location or the collapse of a that an actor is anything able to associate texts, hu- toy house, the agent is whatever did something that mans, nonhumans, and money in a network. Peo- directly brought it about, for example, the child who ple obviously qualify. In her work, for instance, a moved the shovel or the fan that blew on the house.1 nuclear power plant technician might tie together So all actor-network theory does is correctly note technical manuals, fellow employees, graphite that doing is attributed to many sorts of entity. How- rods, control panels, poker winnings, and decon- ever, this theory plays cavalierly with distinctions in tamination chambers. The graphite rods, however, which types of do-ing (agency) are appropriately at- might associate such entities, too, by thwarting pre- tributed to different sorts of entity, despite its predi- scribed procedures, cracking control panel hous- lection for multiplicity. ings, drawing technicians into the control room, Compare, for instance, attributions of agency to a causing one of them to drop her winnings, and forc- person, a dog, and a tornado. When we say of a per- ing workers into decontamination chambers. For son that she did something, say, cut down a tree, CALLON, who or what the agents are in a given net- more than likely we are saying that she did so inten- work is a matter of who or what is credited with tionally (i.e., aimed at a result brought about). Many associating entities. Consequently, what is taken as of the actions she performs intentionally, moreover, associator (actor) or associated in a given network will be carried out deliberately and as a result of plan-

Evolution and Cognition ❘ 208 ❘ 1999, Vol. 5, No. 2 Humanism on the Ropes?

ning.2 And in many cases where she does something to overcome an absolutist distinction between “the unintentionally (e.g., exhausts the family), her do- language of things in themselves” and “the language ing so will be a product of things she did intention- of scientists among themselves,” that is, a hard and ally and maybe also as a result of planning (e.g., re- fast distinction between nonhumans and humans. lentlessly searching for just the right ax). By contrast, Indeed, they do not really mean, they write, to grant when we say of a dog that it did something (e.g., intentionality to things. The point of extending the asked for more grapes), we only sometimes mean language is methodological—to force recognition of that it did so intentionally. Dogs often act without the interminglings of humans and nonhumans and intention, and more or less nothing they do is done the contributions nonhumans make to the networks either deliberately or as a result of planning. Saying of which humans are part. CALLON/LATOUR (1992, of the tornado, finally, that it flattened a seven mile p354) add that they are only trying to develop a long path does not entail intentionality, let alone “symmetric metalanguage” for the description of deliberation and planing. It implies only that a me- humans and nonhumans and cannot be held re- teorological phenomenon caused by violently col- sponsible if no “unbiased” vocabulary exists. liding air masses had certain physical effects. Agency An unbiased, symmetrical vocabulary does exist, in this case is physical causality and nothing more. however: the language of doing, applicable without If doubts linger about the significance of these dif- prejudice to humans and nonhumans alike. If, more- ferences, consider the very different measures that over, the issue is one of recognizing the contribu- are taken against marauding people, dogs, and tor- tions of nonhumans to the networks of which hu- nados, or the very different techniques that must be mans are part, the vocabulary of doing is sufficient. employed to “enroll” these different entities into There is no need to add to this simple common vo- networks. cabulary further terms that apply paradigmatically These points, I assume, are obvious. Yet, actor- to humans. In fact, doing this is likely to undercut network theorists insist upon ascribing the paradig- the goal of sorting out the contributions that differ- matic human type of do-ing, intentional agency, to ent entities make to networks. Using a vocabulary a wider variety of nonhuman entities than is stan- that connotes human ways of being obfuscates the dard practice. At one point in his famous article overwhelming likelihood that entities act “categori- about the scallops of St. Brieuc Bay, for instance, CAL- cally” differently, that is, that the natures of their LON (1986, pp219–220) describes them as “dissi- doings and contributions vary; and getting these dif- dents”. A trio of scientists had concluded that scallop ferent contributions right will involve differential larvae would anchor themselves to collectors im- applications of terms, indeed, probably some of the mersed in the Bay. Many larvae, however, fail to do differences leveled by LATOUR and CALLON’s overly so, thus “betraying” their “enrollment” into net- thick symmetrical metalanguage. In addition, justi- works of scientists, fishermen, tides, parasites, local fying the extension to nonhumans of terms paradig- politicians, and so on as previously “negotiated” matically applied to humans on the grounds that with the scientists. Descriptions such as this, how- this overcomes absolutist distinctions between hu- ever jarring, raise a familiar and important issue: To mans and nonhumans ignores that any such distinc- which nonhuman entities is intentionality correctly tion is already blurred in extant practices, for in- ascribed? Today we know that animals boast varying stance, by the attribution of intentionality to degrees of intentionality and that machines might chimpanzees and dogs. Attributing intentionality to soon qualify as intentional beings. But scallops? scallops in particular and nonhumans in general is Unfortunately, actor-network theorists nowhere unnecessary for the purpose of counteracting theo- argue that certain entities hitherto denied purpose ries that consign intentionality to humans alone. and will, for example, scallops and tornados, have In short, it is one thing to say on linguistic or other been wronged. Without offering arguments ger- grounds that graphite rods, computers, scallops, mane to particular cases, they simply extend a way chimpanzees, and humans are all agents, i.e., do-ers. of talking paradigmatically applied to human beings It is quite another to attribute intentionality to to creatures immensely different from the paradigm them. Certain distinctions among entities must be case. (Notice that my complaint is not that actor- respected; and not every word used for humans network theory anthropomorphizes, nor that an- should be applied to nonhumans indiscriminately— thropomorphisms are illicit.) As justification, CAL- or at all (cf. CALLON/LATOUR 1992, p353). LON/LATOUR (1992) instead claim that expansive ex- To approach this point from a slightly different tension of the language of intentionality is required angle, consider Steve WOOLGAR’s attempt to defend

Evolution and Cognition ❘ 209 ❘ 1999, Vol. 5, No. 2 Theodore R. Schatzki the attribution of activity and mentality to ma- neurological, genetic, hormonal, cognitive, and chines. In response to the objection that such talk is conative assemblages. More expansively, individual just metaphorical, WOOLGAR (1991, p91) remarks entitites such as human beings, organisms, and so- that the interesting question is what entitles us to cial formations are “molar” phenomena that arise attribute intentionality to humans in the first place. from the segmental organization of molecular mul- The answer is that nothing “entitles” us. The lan- tiplicities. The psyche, personality, and self-con- guage of mentality is today paradigmatically a dis- sciousness that help mark a human being, for in- course about human beings, whose extension to stance, arise from the segmental organization of nonhumans reflects the degrees to which the activ- physiological, neurological, genetic, hormonal, ities and constructions of these entities resemble cognitive, and conative assemblages. A human be- those of humans.3 The more creatures approximate ing, consequently, is an integral organized molar us, the easier and more comprehensive the attribu- assemblage of molecular multiplicities. tion to them of actions and mental conditions. This “It will be noted that names … function as com- means, incidentally, that the distinction between mon nouns ensuring the unification of an aggregate humans and nonhumans defended here is not an a they subsume. The proper name can be nothing priori one. It is based, on the one hand, on analyses more than an extreme case of the common noun, of the human form of life which theoreticians find containing its already domesticated multiplicity themselves already carrying on once they attain the within itself and linking it to a being or object pos- sophistication to analyze it and, on the other, on ited as unique.” (DELEUZE/GUATTARI 1987, p27) observed differences between the character of that The molecular, it might be added, is the home of form of life and the lives of other species. Conse- desire; its components are “originally” structured as quently, even though conceptions of human kin- a manifold of “desiring machines”. Individuals, ship to other entities vary over sociohistorical space– moreover, are molar desiring machines that invest time, and even though “primitive” peoples attrib- their physical, organic, and social environments. uted intentionality to a wider variety of phenomena Disclosing this picture’s psychoanalytic prove- than modern Westerners do, WOOLGAR, in the con- nance, DELEUZE/GUATTARI (1983, p289) call the en- temporary scientific context in which he writes, ille- ergy of desiring machines “libido” and label the the- gitimately shifts the burden of proof. Detractors of ory of “the machinic connections of a molecular actor-network’s profligacy need not defend the ap- order, their distribution into maps of intensity on plicability of the language of mentality/activity to the giant molecule of the body without organs, and humans—it is already ubiquitously applied to them, the statistical accumulations that form and select and it is here that it acquires its meaning. Rather, the larger aggregates” the “theory of schizophre- actor-network theory must argue for the extension nia”. They also tie this psychoanalytic conception of this language to nonhumans on a case by case to contemporary cell microbiology in treating DNA- basis; it is not enough that this extension criss- coded proteins as the “unconscious” components of crosses certain misguided absolutist ontological di- molecular desiring machines (see DELEUZE/GUATTARI vides. 1983, p290). DELEUZE and GUATTARI never focus on agency per 3. Network multiplicity se. Given their (1987, pp256, 260) definitions of lat- itude and longitude as, roughly, the particle aggre- Actor-network theory does not simply fracture gates belonging to an entity and the affects (in agency into diverse occurrences of doing. It also SPINOZA’s sense) of which that entity is capable, it treats agents themselves as heterogeneous net- makes sense to construe the agency of a human be- works. Agency, consequently, is an unstable and ing (or of any other entity)—like his or her psyche, possibly disunified feature of whole networks that self-consciousness, and personality—as a molar depends on the operations of their parts. phenomenon. Accordingly, stable and unified indi- Before considering these claims, consider the par- vidual agency rests on the successful and persisting allel views in DELEUZE and GUATTARI. (The following organization of micromultiplicities, the ordering of combines ideas found in DELEUZE/GUATTARI 1983; a “city” of molecular desiring machines. Agency, 1987.) For these theorists, an individual entity (or- however, is a property not just of individuals, but ganism, person, machine, social formation) is a also of the micro desiring machines whose com- multiplicity of molecular assemblages. A human be- bined action gives rise to individuals’ agency. In- ing, for instance, is a multitude of physiological, deed, molar agency is at once something people

Evolution and Cognition ❘ 210 ❘ 1999, Vol. 5, No. 2 Humanism on the Ropes? share with organisms, machines, and social forma- Agents are not just networks. They are also “ef- tions and an epiphenomenon dependent on the or- fects” of networks. An individual cannot do any- ganization of the molecular agencies that supply thing without the support of the network compos- these molar entities with motion. ing it (organs, memories, skills, preferences, and Actor-network theorists treat entities, similarly, artificial limbs in the case of human beings; machin- as networks. As in DELEUZE and GUATTARI, they are ery, operators, fuel rods, and so on in the case of the networks to which unity is ascribed. A human be- power plant). Indeed, an individual’s activity is gen- ing, for instance, is a network of neurons, muscles, erated by the network composing it. Agency is also memories, skills, preferences, hormones etc. taken dependent, however, on the wider network(s) in as a unit. (Actor-network theory calls this telescop- which agents are embedded (chairs, computers, stu- ing “black boxing” or “punctualization”.) An actor, dents, and administrations in the case of professors; moreover, is any network to which both unity and customers, suppliers, operating companies, and sub- doing are ascribed: Taking X to be an actor is appre- soil geologies in the case of the plant). So the net- hending the network that X is as a unit and crediting works of which agency is a unity-effect come in two that unit with doing something. “[A]n agent is a flavors: compositional and embedding. And the spokesperson, a figurehead, or a more or less opaque components of networks of both sorts are them- ‘black box’ which stands for, conceals, defines, selves (potential) agents that are generated by the holds in place, mobilizes, and draws on, a set of jux- further networks whose unities they are. Inciden- taposed bits and pieces.” (LAW 1994, p101) LATOUR, tally, one might wonder where this decomposition recall, defines actors as the substances to which sta- bottoms out (or the contextualization tops out). bilized concatenations of action are ascribed. We Only LATOUR tackles the issue philosophically. Real- now see that the “substances” involved are really ity, he (1988, chapt. 1) claims, is a NIETZSCHEAN ple- networks taken as units. John LAW (1991) offers a num of incommensurable and indivisible, acting variation of this theme when he writes that any net- and acted on forces, the number and identity of work sufficiently stable to generate power effects is which can be determined only after they take the an actor.4 measure of one another in “trials of strength”. Ac- In this context, consider CALLON’s (1991, p141) cording to his account, consequently, actor-network description of nuclear power plant. This hybrid of decomposition terminates somewhere, though its “graphite rods, turbines, atoms, operators, control endpoint is a shifting one subject to redetermination boards, flashing lights, concrete slabs, and engi- in further trials and never an immutable fundament. neers [converts] files, bills, fuel, water, skills, and I wish to describe one further proposition resulting budget lines into electrons transported to consum- from the analysis of agents as networks. CALLON and ers, taxes paid to local councils, and waste prod- LAW suggest that features of the networks that under- ucts—which in turn lead to the formation of groups write (human) agents determine the types of actor of angry environmentalists.” they are. In, for instance, networks whose human If one groups together such actions as transport- components largely (1) share definitions of people, ing electrons, paying taxes, and producing waste things, and situations and (2) observe the same rules products, the actor that performs them, i.e., the sub- in both attributing identity and agency and black stance to which they are attached, is the power boxing generally, people are actors with precise ob- plant. The power plant, however, is a network of jectives and instruments, perfect but limited infor- graphite rods, turbines, operators, and the like. So mation, few choices, and no disagreements. When the actor that performs these actions is a network definitions, attributions, and black boxing diverge, treated as a unit. If, by contrast, one groups together by contrast, people are actors who pursue strategies, such actions as containing emergencies and work- negotiations, revisable projects, and varied aims. In- ing shifts or heating up and melting down, the ac- termediate cases between these extremes compel the tors involved are the operators and graphite rods application of other conceptual apparatuses to hu- respectively. Each of these agents is also, of course, man activity, for instance, procedural rationality or a network taken as a unit. Consequently, an ascrip- game theory. Hence, “…there is no theory or model tion of agency, as in DELEUZE and GUATTARI, is an of the actor, even in the plural. The actor has variable instantaneous apprehension of multiplicity. By geometry and is indissociable from the networks that considering different congeries of action, moreover, define it and that it, along with others, helps to de- agency can be seated in any component of a net- fine”. (CALLON 1991, p154) This means, I believe, that work, as well as in the network as a whole. which properties beyond doing are ascribed to some-

Evolution and Cognition ❘ 211 ❘ 1999, Vol. 5, No. 2 Theodore R. Schatzki thing qua agent depends on the networks within quently, are indeed assemblages-networks. But this which it is one. realization does not undermine human agency and Much of these accounts, I believe, is unproblem- should disconcert only those who metaphysically atic. Agents, to begin with, are networks to which distance will and its motivational context from ma- action is ascribed. Human beings, for instance, are terial (and social) reality. molar aggregates of molecular assemblages, net- In its haste to level humans and nonhumans, ac- works of organs, preferences, memories etc. to which tor-network theory also neglects to sort out signifi- unity and doing are ascribed. Agency, moreover, is a cant differences between compositional and embed- unity-effect generated by these networks. Just as hu- ding networks. The agency effected by networks, for man consciousness depends (presumably) on brain example, is located differently depending on the operations, self-conscious intentional agency arises network involved. Qua effect of compositional net- from the intermeshed operations of multiple bodily works it is attributed to the networks themselves as systems. Agency also indeed rests on embedding net- units, whereas qua effect of embedding networks it works. Human agency, for instance, depends on the is ascribed to components thereof. These two sorts networks of people, organisms, artifacts, and things of network also maintain different functional rela- that populate human worlds. And, yes, components tions to activity. of compositional and embedding networks are Of a compositional network it makes sense to say themselves (potential) agents that are generated by that an agent is both a network and an effect thereof. the further networks they black box: Pets, comput- An actor is its compositional network because any- ers, bodily systems, and the like do things, thereby thing is that of which it is composed. An agent is also qualify as agents, and upon closer inspection turn an effect of its compositional network since its ca- out to be networks of entities, the components of pacity to act as a single entity depends on the co- which are likewise agents. In short, when actor-net- operation of its components. An actor is not, how- work theory is treated as an account of the causal ever, its embedding networks: a professor is not her material composition of agents and the causal con- chair, computer, secretaries, and administration, ditions of agency, it is relatively uncontroversial (or just as a power plant is not its operating company, at least unexotic). nuclear fuel suppliers, and subsoil geologies. To deny Multiplicity theorists write, however, as if their this is to deny that entities have boundaries, ulti- conceptions of constitutive multiplicities captured a mately, that there are distinct entities at all. Some- deep truth that subverts human agency. Actor-net- thing’s agency, furthermore, is not generated by its work’s picture of multileveled networks connected embedding network as it is by its compositional one. via association and black boxing, for instance, asks The doings of a professor or power plant are not gen- the reader to treat humans, parts of humans, and erated or brought about by networks of chairs, sec- nonhumans alike. Much of the fanfare attending the retaries, operating companies, and subsoil geologies: “discovery” that people are assemblages or networks A professor can act (hopefully) sans her chair and arises from the thesis that human agents are com- secretary, just as a power plant can act absent its opposed of and dependent upon nonhuman and “sub- erating company (it is now employee owned). human” actors. As discussed, however, the thesis Agency is an “effect” of embedding networks in at that these other–than–human entities are actors least three other ways. First, agency presupposes cer- holds only at a high level of generality: Actor-net- tain general types of embedding network, for exam- work theory resembles an obedience trainer who re- ple, arrangements of physical things, since without marks that because any dog is a dog, all dogs are them there wouldn’t be anything to act on and the alike. Once the thinness of the claimed equivalence networks that compose agents couldn’t function or is appreciated, initially provocative formulations be- survive. Second, embedding networks circumscribe come more familiar. For instance, the proposition agency (they are not alone here). Without his com- that human agents are (effects of) networks of fur- puter the professor might not be able to finish next ther agents resolves into such claims as (1) that hu- Tuesday’s presentation, but he still can bemoan los- man actors are composed of active physical sub- ing the ability to write legible longhand. Similarly, systems that maintain causal relations among without its graphite rod suppliers the plant won’t be themselves and with the environment outside the able to supply its customers with electricity, but it skin, and (2) that what people are capable of doing might still lobby the government to ease safety re- depends in part on the people, organisms, things, strictions on the production of nuclear fuel. Agency, and artifacts around them. Human actors, conse- therefore, is an effect of embedding networks, sec-

Evolution and Cognition ❘ 212 ❘ 1999, Vol. 5, No. 2 Humanism on the Ropes? ond, because networks enable and constrain (better: 4. Humanism saved? prefigure) what agents do. An important form this effectuation takes is the enforcement of norms per- “Humanism” has no precise meaning. Generally taining to the whats and hows of action. A third way speaking, it is a cultural stance that arose in Europe embedding networks effect agency is by people im- during the 1500s and 1600s to celebrate Man as puting agency (doing) to other network elements (cf. thinker, creator, and actor. This exaltation of the section 2). human coincided with growing sentiment that Not only has actor-network theory neglected to humans are not essentially beholden to God, the sort out these differences; the further thesis that net- chain of being, or the implacable order of the uni- works determine the character of agency overex- verse. They can create and take responsibility for tends its insights. As noted, CALLON and LAW main- conceptual and material structures whose splendor tain that which features other than intentional honors their capacities and power. A pair of centu- doing mark human agency depends on the character ries passed before more militant atheisms proof the networks as part of which people proceed. claimed Man the equal or vanquisher of God as well Whereas in some situations actors must be credited as the unique source of meaning, truth, and being. with precise objectives, perfect but limited informa- Theories of social multiplicity damage humanism tion, and few choices, in others they must be im- in this general sense scarcely. They do not, for in- puted strategy, negotiation, and varied aims. Varia- stance, deny that human beings are agents, a gain- tions of this sort, however, do not entail that the saying that would indeed hollow out any human- nature of human agency is variable and that a ism. Their claim that human agency is an effect of “model” or “theory” of the actor is impossible. Pre- multiplicities at bottom means that agency arises cise versus varied objectives, perfect versus imperfect from the co-operation of the multiplicities that com- information, and few or many choices are different pose a person and is enabled and constrained by values of specific parameters of a single (broadly those embedding her. This claim does not debunk speaking rational choice) model of activity, not the agency; rather, it conceptualizes its character and constants of different ones. Moreover, the presence provenance scientifically. Like the neurophysiolog- or absence of strategies and negotiations results from ical explanation of human activity familiar since at the specific values these parameters assume in sets of least the end of the 18th-century, it challenges hu- individuals in particular interactional circum- manism only on unnecessarily reductive or deter- stances. Variability in these matters means neither ministic readings. that what a human actor is varies across networks Theories of multiplicity might also be thought to nor that a single model of agency is utopian. decapitate humanism by fragmenting the subject in I write this realizing that no extant model is suffi- the manner FOUCAULT allegedly epitomizes. As dis- ciently capacious or realistic to capture all occasions cussed, however, his dispersion of subjects among and aspects of human agency. This failure arises, discourse positions does not exhaust their being. In- however, from the patently reconstructive nature of deed, part of what remains is the performance of many models and/or the simplification inherent to nondiscursive actions. Many theorists who develop them all. It does not confirm the alleged dependency FOUCAULT’s line press, not just discursive action, but of the character of human agency on embedding its nondiscursive kin too, into the fragmented tem- contexts, even though circumstantial features of plate of subject positions or their ilk (e.g., BUTLER agency clearly are so conditioned. Although, in ad- 1993). In their hands, however, subject positions are dition, no current model covers intentional along mere scripts or bundles of norms (whose specificity with all forms of nonintentional agency, a future and malleability varies for different theorists). Such model might prove successful here. accounts emphasize the social formation of agency, At the same time, fundamental differences among not its dissolution. the agencies of humans, dogs, scallops, power sta- Of course, if the humanist subject is take to be the tions, and tornados are tied, I presume, to the differ- ethereal free agent of Western lore, humanism inent types of network composing these entities. In this deed suffers at the hands of multiplicity theory. regard, the nature of agency is indeed an effect of net- Obituaries of this agent, however, first appeared long works, and a single model of agency is implausible. ago. If, consequently, multiplicity theorists are tak- Notice that these claims align WITTGENSTEINIAN con- ing aim at this subject, they, like too many others, ceptual dispersion vis–à–vis agency with the disparate are simply re-sounding its death knoll in the face of multiplicities that compose different types of agent. dwindling efforts to save it.5

Evolution and Cognition ❘ 213 ❘ 1999, Vol. 5, No. 2 Theodore R. Schatzki

The main challenge multiplicity theory poses to by extraterrestrials who have been acting thusly all humanism arises from its proliferation of agency. In along. Modesty precludes abjudicating whether hu- theory, of course, the idea that animals, machines, mans alone possess the self-conscious, intentional, and power plants are actors need not trouble anyone deliberative, and planning agency they display. At intent on celebrating the thinking, making, and do- the same time, we must not overlook that as far as we ing of humans. Humanism need not proclaim people do know today humans alone, and not even all of the only agents. There is no denying, however, that them, possess it fully. it has tended in this direction and thereby rendered To go one step further: Even if the above possibili- itself vulnerable to intellectual developments that ties materialize, we might not want to consign hu- multiply agency. Still, the proliferation of agency is manism to the past. We might still continue to view compatible with the distinctive and unique richness humans as special, on the grounds that they are the of the human variety—and this is distinction enough entities overwhelmingly responsible for the quality of to satisfy many humanists. So the real challenge to life they themselves enjoy or suffer. This fact, increas- humanism cannot arise from the insights that people ingly the case throughout history, shows no sign of are multiplicities and many (potentially all) nonhu- abating in the future. It is human agency that prima- mans agents. rily fashions the worlds people live in and that must It is instead posed by the possibility that entities be confronted in making those worlds better places. other than humans are the type of actor they are. (Notice that this is a causal, not an ethical point.) Perhaps dolphins act intentionally, deliberately, and I don’t know whether this “saves” humanism. It on the basis of plans. Maybe machines will confront certainly leaves room for a clairvoyant variant, one and confound us one day soon cognizant that nonhumans with such behavior. Maybe are agents, that humans are some day prosthetic and im- Author’s address multiplicities whose agency plant technology will produce rises therefrom, and that hu- Theodore R. Schatzki, Department of Philos- exotic human–machine hy- mans may not be the sole crea- ophy, University of Kentucky, Lexington, KY brids who display it. And per- tures in the cosmos capable of 40506-0027 USA. haps one day anywhere along self-conscious, intentional, Email: [email protected] the way we will be contacted deliberate, planning activity.6

Notes 5 Jean-Paul SARTRE may be metaphysical freedom’s last great defender. In his eyes (1966), pace what I wrote earlier, a 1 This sort of definition, I acknowledge without further com- human agent is no more her compositional network than ment, faces subtle dilemmas concerning omissions, the dis- she is her embedding ones—for the subject cannot be con- tinction between doing and happening, and the proper stituted by anything in the world. In later works, of course, delimitation of causal chains. See, for instance, BRAND SARTRE (e.g., 1968) tried to square free action with sociohis- (1971) and DAVIDSON (1980a, 1980b). torical contextualization. 2 For a discussion of different ways of doing something, see 6 This essay was completed while a visiting scholar at the Max AUSTIN (1970). Planck Institute for the History of Science in Berlin. I would 3 This way of thinking derives from WITTGENSTEIN (1958). like to thank the Institute and the director of Abteilung 4 The relevant effects are stockpiles of, degrees of discretion Drei, Hans-Jörg RHEINBERGER, for their support. concerning, and the particular circumstantial character of both “power-to” and “power-over”.

References Callon, M. (1991) Techno-economic networks and irrevers- ibility. In: Law, J. (ed) A Sociology of Monsters: Essays on Austin, J. (1970) Three Ways of Spilling Ink. In: Austin, J. Power, Technology, and Domination. Routledge: London, Philosophical Papers, 2nd edition. Oxford University pp. 132–161. Press: Oxford, pp. 272–287. Callon, M./Latour, B. (1992) Don’t Throw the Baby out with Brand, M. (1971) The Language of Not Doing. American the Bath School! A Reply to Collins and Yearly. In: Picker- Philosophical Quarterly 8 (1): 45–54. ing, A. (ed) Science as Practice and Culture. University of Butler, J. (1993) Bodies that Matter. Routledge: New York. Chicago Press: Chicago, pp. 343–368. Callon, M. (1986) Some elements of a sociology of transla- Davidson, D. (1980a) Agency. In: Davidson, D. (ed) Essays on tion: domestication of the scallops and the fishermen of St Actions and Events. Oxford University Press: Oxford, pp. Brieuc Bay. In: Law, J. (ed) Power, Action and Belief. Rout- 83–102. ledge and Kegan Paul: London, pp. 196–233. Davidson, D. (1980b) Intending. In: Davidson, D. (ed) Essays

Evolution and Cognition ❘ 214 ❘ 1999, Vol. 5, No. 2 Humanism on the Ropes?

on Actions and Events. Oxford University Press: Oxford, Latour, B. (1991) Technology is society made durable. In: pp. 43–62. Law, J. (ed) A Sociology of Monsters: Essays on Power, Deleuze, G./Guattari, F. (1983) Anti-Oedipus. (Translated by Technology, and Domination. Routledge: London, pp. R. Hurley, M. Seem, and H. Lane). University of Minnesota 103–131. Press: Minneapolis. Law, J. (1991) Power, discretion, and strategy. In: Law, J. (ed) Deleuze, G./Guattari, F. (1987) A Thousand Plateaus. (Trans- A Sociology of Monsters: Essays on Power, Technology, lated by Brian Massumi). University of Minnesota Press: and Domination. Routledge: London, pp. 165–191. Minneapolis. Law, J. (1994) Organizing Modernity. Blackwell: Oxford. Foucault, M. (1972) The Archaeology of Knowledge. (Trans- Sartre, J.-P. (1966) Being and Nothingness. (Translated by H. lated by A.M. Sheridan Smith). Harper and Row: New York. Barnes). Pocket Books: New York. Foucault, M. (1980) The Confession of the Flesh. In: Gordon, Sartre, J.-P. (1968) Search for a Method. (Translated by H. C. (ed) Power/Knowledge. Pantheon: New York, pp. 194– Barnes). Vintage: New York. 228. Wittgenstein, L. (1958) Philosophical Investigations. (Trans- Foucault, M. (1984) What is an Author? (Translated by J. lated by G. Anscombe). Macmillan: New York. Harari). In: Rabinow, P. (ed) The Foucault Reader. Panthe- Woolgar, S. (1991) Configuring the User: the case of usability on: New York, pp. 101–120. trials. In: Law, J. (ed) A Sociology of Monsters: Essays on Latour, B. (1988) Irreductions. In: Latour, B. The Pasteuriza- Power, Technology, and Domination. Routledge: London, tion of France. (Translated by A. Sheridan and J. Law). Har- pp. 57–99. vard University Press: Cambridge, part 2.

Evolution and Cognition ❘ 215 ❘ 1999, Vol. 5, No. 2