Ingo Brigandt

Gestalt Experiments and Inductive Observations

Konrad Lorenz’s early Epistemological Writings and the Methods of Classical Ethology

thology brought Abstract scientific knowledge. Ge- E some crucial insights stalt perception and in- and perspectives to the During the 1940s Konrad LORENZ formulated his ear- duction are the main ele- study of behavior, in par- ly epistemological views, focusing on the cognitive ments of LORENZ’s ticular the idea that be- mechanisms induction and Gestalt perception. After epistemology. These ideas havior can be studied the war he used this philosophical framework to de- were originally formu- within a comparative– fend the approach of classical ethology against other lated during the second evolutionary framework approaches to animal behavior. The present paper ex- world war (see for in- by means of homologiz- amines the relationship between LORENZ’s ethological stance the Russian Manu- ing components of be- methodology and his philosophy of science and knowl- script, LORENZ 1948). In havioral patterns and by edge. The main aim of LORENZ’s post-war epistemo- the post-war period causal analysis of behav- logical writings is to provide an epistemological and LORENZ emphasized his ior components and their cognitive theory of observation in order to defend the epistemological ideas as integration. Early ethol- observational approach of classical ethology against part of an attempt to jus- ogy is well-known for its the view that it was no rigorous science. tify his ethological ap- extensive use of qualita- proach and his views tive observations of ani- Key words about the role of observa- mals under their natural tion and experiment. The Konrad LORENZ, Gestalt perception, observation, conditions. These obser- aim of the present paper ethology. vations are combined is to call attention to with experiments that try LORENZ’s philosophical to analyze behavioral patterns and establish specific account and how it integrates with his biological claims about animal behavior. Nowadays, there is methodology. still disagreement about the significance of observa- tion and experiments and their relation. ALLEN Observation and Experiment (forthcoming) points out that in debates about the interpretation of animal cognition the interpreta- LORENZ’s account of the general character of obser- tions of those ethologists who have actually spent vation and experiment remains unchanged time watching the animals are favored by some biol- throughout his writings. The specific relationship ogists. In fact, some practitioners of cognitive ethol- between these two aspects is exhibited by LORENZ’s ogy feel that experiments at best just confirm what early biological work as well as his later explicit one already knows, while others are more skeptical methodological account of them (LORENZ 1935; about interpretations of observations without ex- LORENZ/TINBERGEN 1938; LORENZ 1948, 1981). periments. LORENZ emphasizes four distinctive features of clas- As a major representative of ethology Konrad sical observational practice: observations are to be LORENZ not only shaped its approach and made the carried out in a hypothesis-free way; the approach methods and theories of ethology known to the bi- consists in doing qualitative observations rather ological community and the interested public, he than quantitative measurements; animals are to be also embedded his defense of the ethological and observed in their natural environment; and obser- comparative approach in a philosophical theory of vations include all features of the organisms and

Evolution and Cognition ❘ 157 ❘ 2003, Vol. 9, No. 2 Ingo Brigandt their environment and rely on a large observational tative measurements and recording them, observa- basis. One should not think that ethology (in partic- tions are recorded using written descriptions (1935, ular the more developed ethology of the post-war pp112ff), drawings and photographs (LORENZ/TIN- period) always proceeded in this fashion. Instead, BERGEN 1938). This qualitative approach of classical LORENZ’s bold claims about how ethology is done ethology has to be defended against an understand- are based on some of his opinions as to how science ing that only allows for quantitative measurement should be done or how a science should develop and statistical evaluation as a means of scientific ob- (see BURKHARDT 1981). In any case, the way classical jectivity (1958, pp246, 256; 1959, p281; 1963a, p1; ethology was done was noticeably different from 1981, pp40ff, 68ff). Despite LORENZ’s emphasis on other contemporary approaches to animal behavior qualitative observation, post-war ethology did very such as comparative psychology or later on sociobi- well make use of quantitative measurements, but ology. In fact, LORENZ encountered methodological connected them with qualitative observations critique from both groups. (SCHLEIDT/SCHLEIDT 1958 is an example). LORENZ defends the ethological approach of gath- A further feature of the ethological way of observ- ering observational data without having a hypothe- ing is to systematically study animals in their natu- sis in mind against the view that scientific accounts ral ecological environment. Animals are kept “in an must be based on a theory (1948, p216; unless oth- environment as close as possible to their natural erwise indicated, references refer to a publication of habitats, for the purpose of general biological and LORENZ). LORENZ thinks that it is not only possible to specifically ethological observations” (1935, pp108– make observations devoid of a theoretical frame- 109). Keeping animals under laboratory conditions work, he also points to the possibility of confirma- suffers from the drawback that the organisms do not tion bias as a disadvantage of basing one’s observa- exhibit their natural adaptive behavior patterns and tional approach on a hypothesis (1948, pp31, 71). thus functional and artifactual behavior are likely to Instead of forming theoretical considerations at an be confused. This approach of observing animals early stage of inquiry, scientific inquiry has to start under natural conditions continues the “amateur- with “observation pure and simple” (1958, p246). ism” of the forerunners of classical ethology (1981, LORENZ calls this “presuppositionless observation” p47). However, for LORENZ the optimal method is (1981, p47) or often “unbiased observation” (1958, not to study animals in the field. For in this case, it is p250; see also 1948, p213; 1950 p131; 1959, p281). difficult and extremely time-consuming to be able Now, even amateur naturalists who dedicated all to observe the relevant behavior patterns (1948, their time to field observations of animals did not p222). The chapter “Animal Keeping as a Research necessarily refrain from interpreting their observa- Method” of the Russian Manuscript gives a detailed tions and relate them to biological theories. Ed- account of how the study of animal behavior has to mund Selous, for instance, used his extensive obser- be organized (1948, pp221ff). Several animals of a vations to justify theories of sexual selection. In the species have to be kept so that they live in proximity case of LORENZ, the theoretical assumption that be- to the researchers and in a controllable environ- havioral characters can be used to characterize taxo- ment. Apart from conducting observations in that nomic groups and reconstruct phylogenies guided way, animal keeping makes it possible to conduct his observational practice from very early on. But experiments under otherwise natural conditions LORENZ is right insofar as before 1935 he and TINBER- (1948, p222). LORENZ’s practice of raising and keep- GEN completely lacked an interpretative framework, ing animals was quite similar to the practice of his unlike most people working in animal psychology. mentor Oskar HEINROTH or the American zoologist After the war ethology became more theoretical, but Charles Otis WHITMAN. WHITMAN’s and LORENZ’s compared to other approaches such as comparative practice of animal keeping enabled them to simulta- psychology or sociobiology ethology was not that neously observe the behavior of several closely re- much an enterprise driven by explicitly formulated lated species, which is hard to achieve in the field. theories and hypothesis that were put to empirical But LORENZ’s approach differed from the early field test. In particular sociobiology took quantitative hy- naturalist in the British tradition (Selous and potheses based on models from population genetics Howard) or in the Dutch tradition (RÖELL 2000). In as the starting point and used observations prima- particular Niko TINBERGEN heavily relied on field ob- rily to test these hypotheses. servations (and experiments) instead of animal In addition, early ethology relied on qualitative keeping (TINBERGEN 1932, 1935, 1951), while observations of behavior. Instead of making quanti- LORENZ viewed field work as a control for observa-

Evolution and Cognition ❘ 158 ❘ 2003, Vol. 9, No. 2 Gestalt Experiments and Inductive Observations

tions gained from animals kept in semi-natural con- Charles Otis WHITMAN, for instance, kept detailed ditions (1981, pp47ff). Thus TINBERGEN represented notes of his extensive observations, and carried out the ecological dimension of ethology more fully observations for years before he published his re- than LORENZ did.1 Despite the fact that LORENZ was sults (LILLIE 1911). Like LORENZ, he emphasized that not a field naturalist, his approach was quite dis- a detailed and extensive knowledge of the behav- tinct from the American tradition of animal psy- ioral repertoire of a species is crucial to avoid false chology and animal behavior studies. LORENZ often interpretations and conclusions (WHITMAN 1899a, labeled the American scientists ‘behaviorists’, de- 1899b). spite the fact that scientists such as John WATSON or The role of experiments within ethology is nicely Robert YERKES had quite different views of animal illustrated by LORENZ’s and TINBERGEN’s classical minds and behavior. But apart from theoretical dif- study of the egg-rolling behavior of the Greylag ferences, the American tradition was definitely goose (1938). (TINBERGEN was in fact the person who characterized by laboratory studies. Observations was more skilled in devising and conducting exper- and experiments under laboratory or controlled iments. Compare TINBERGEN 1932, 1935, TINBERGEN/ conditions dominated (WATSON 1914; YERKES 1925), KUENEN 1939 with LORENZ’s more observational and field studies were an exception (e.g., WATSON 1908 theoretical papers LORENZ 1932, 1935, 1937.) On or YERKES’ student NISSEN 1931). This is due to the LORENZ’s account, experiments are to be preceded fact that in the United Studies animal behavior was not only temporally, but also methodologically and studied at universities and institutes from early on, logically by observations that are not goal-directed. while some of the forerunners of European ethology For an experiment only makes sense if the natural were amateur naturalists. (In the United States, units and their interaction are known to a sufficient there were a few important field naturalist among degree (1981, p53). An encompassing observational the zoologists, rather than the psychologists. But basis is necessary to have a grasp of the structure even zoologists primarily studied behavior in labo- and function of the system under consideration. ratories; DEWSBURY 1988.) Knowledge about the structure of organisms and Ethological observations do not simply focus on the function of behavior is needed for having a bio- specific aspects of animal behavior. Instead, accord- logically meaningful classification of the parts of ing to LORENZ’s ideal ethological analysis start with the system studied. These elements identified by ob- considering all features of an individual including servation can one after the other analyzed in more its natural environment (1981, pp52f). Observation detail and substantiated by experiments designed has to give an overview of the complete behavioral for addressing such particular questions (1935, system of a species (1948, pp221f). Only in this pp110f). Thus, in the ethological approach, the ex- manner it is possible to obtain knowledge about the periment is necessarily posterior to observation. relevant substructures of the system, its functions LORENZ is for instance impressed by the detailed and relations to other organisms and the environ- observational work of the ornithologist Margaret M. ment (1950, p120). LORENZ calls this approach that NICE, and he states that one should always have tries to take all features into account an “analysis on such a detailed knowledge of an animal before one a broad front”. It is justified by the fact that the ob- starts to conduct experiments. ject under study is a complex entity whose parts in- “Ich war begeistert von der Forschungsweise teract mutually. The chapter “The Organism as an dieser Dame. So müßte man eben jedes Vieh kennen, Entity and Analysis on a Broad Front” of the Russian bevor man mit ihm ‘Versuche’ anzustellen beginnt.” Manuscript (1948, pp137ff) gives a discussion and (LORENZ, letter to STRESEMANN, Blatt 94–97).2 defense of this method (see also 1958, p249). To The purpose of experiments is to analyze the sub- achieve this general overview of the whole behav- components of an overall behavioral pattern (1935, ioral complex of a species it is necessary to have a pp105, 233; 1937, p293). The experiment enables large observational basis that often may require one to keep the constitutive elements of behavior years to obtain (1935, p109; 1948, p222; 1981, p48). apart, and to study a largely dissociated component Even before LORENZ, the naturalist Edmund Selous in detail. A certain experimental design allows the claimed that in his field notes he really wrote down researcher to answer specific questions about the everything he saw (BURKHARDT, forthcoming). function of behavioral component and the causal While this is literally impossible, in particular the relationship between the components. Dummy ex- early practitioners in ethology are characterized by periments serve to determine the specific stimulus spending most of their time observing animals. that triggers an innate releasing mechanism and

Evolution and Cognition ❘ 159 ❘ 2003, Vol. 9, No. 2 Ingo Brigandt thus help to determine whether this stimulus is oretically fundamental results. Induction provides learned or innate (1935, p143, p228; LORENZ/TINBER- good evidence for the results obtained from data, GEN 1938, p333). Other experiments figure out but it is unlikely that induction arrives at scientifi- whether a stimulus or drive is of external or internal cally relevant conclusions that might be substanti- origin (1950, pp135f). A type of experiment that ated by an adequate inductive basis (1948, p65). was very important for classical ethology is the dep- A central view of LORENZ on empirical science and rivation experiment. It consists of depriving ani- research is the idea that inductive research proceeds mals of certain environmental stimuli during their in three stages. The terms used to describe this cate- development. This is to show that behavioral com- gorization (the idiographic, the systematic and the ponents which nonetheless are exhibited after the nomothetic stage) are adapted from the philoso- deprivation period are not learned but must be in- pher Wilhelm WINDELBAND (1894). An early detailed nate (1965, pp83ff; 1981, pp57ff). LORENZ’s mentor account of this philosophy of science is given in the HEINROTH made extensive use of this technique by chapter “Induction” of the Russian Manuscript hand-rearing birds in relative isolation to discern (1948, pp28ff) and it is maintained throughout their instinctive behavior features (HEINROTH/HEIN- LORENZ’s writings (1958, p251; 1959, p283; 1963a, ROTH 1924–1933). p2). The first phase of inductive science is the idio- graphic stage. It consist of gathering data, observa- Induction and Gestalt Perception tions, and descriptions in a loose manner. The aim is to obtain an inductive basis that is as large as pos- Induction, on LORENZ’s account, is a rational cogni- sible. The systematic stage is the next step. It pro- tive process, which means for LORENZ that induc- ceeds by analyzing the data, comparing the various tive reasoning is a conscious process (1948, p30). bits of evidence, finding interrelations among the More precisely, a rational cognitive process is char- single items, and categorizing them. In the first two acterized by the fact that the individual is aware of stages, the inductive basis gets assembled in a hy- the assumptions made and in particular can verify pothesis-free and theory-independent fashion the steps of reasoning. Throughout LORENZ’s writ- (1950, p129). The final phase is the nomothetic ings, induction is conceptually clearly contrasted stage, which consist in formulating laws and gen- with non-rational processes such as intuition and eral principles supported by the evidence. It is char- Gestalt perception (1948, p55; 1950, p166; 1954, acterized by a process of abstraction (1948, p30). p198; 1959, pp302, 306; 1981, pp43f). Rational Whereas the idiographic and the systematic stage cognitive processes are in principle open to intro- cannot by kept clearly apart, the nomothetic phase spection and the individual has control over the can be more sharply separated from the foregoing way of reasoning, which contrasts them to subcon- stages (1948, p29). Observation, description, com- scious processes (1959, pp283, 312f; 1981, p41). A parison, and systematization are “indispensable general feature of induction is that it abstracts reg- steps that must have been gone through before the ularities and principles from knowledge about sin- first attempt at the abstraction of natural laws, of gle facts (1954, p198). The distinctive feature of in- nomothesis in WINDELBAND’s classical term, is un- ductive science is that it aims at proving assertions dertaken” (1958, p262). The reliability of the results by collected evidence (1948, p54). The last word on obtained by induction is proportional to the broad- verifying the correctness of a scientific hypothesis ness and the scope of the inductive basis (1948, p31; has to be achieved by quantitative analysis (1958, 1950, p194; 1959, pp302, 315). LORENZ claims that p246). early ethology developed just in this three stage The advantage of induction stems from the fact fashion (1950, p131; 1948). However, one needs to that it is a rational cognitive process. The inductive keep in mind that such statements are due to the base is explicitly known, the single steps in induc- fact that LORENZ thought that any real science has to tive reasoning can be analyzed and verified. In par- emerge in this fashion. BURKHARDT (1981) makes ticular the way of reasoning that led to the conclu- clear that some of LORENZ’s historical statement sion can be communicated to other persons in a should not be viewed as an attempt to reconstruct manner that allows for critical assessment of the sci- the history of ethology, but they stem from LORENZ’s entific claim. The bigger inductive basis, the better attempt to create ethology as a discipline, including the scientific result is substantiated (1948, p64). The a defense of ethology as a discipline that developed main disadvantage of induction consists in the dif- the way a real science ought to develop on LORENZ’s ficulty of anticipating empirically important or the- view.3

Evolution and Cognition ❘ 160 ❘ 2003, Vol. 9, No. 2 Gestalt Experiments and Inductive Observations

A peculiar feature of LORENZ’s epistemology is Ge- nism exhibits analogies to reasoning (1959, p302; stalt perception, a cognitive process that is of para- 1981, pp42f). The process starts by using incoming mount importance for the post-war defense of his data as ‘evidential basis’, it draws ‘inferences’ from approach. The notion of Gestalt perception is this data and comes to a ‘conclusion’. As this proce- present in LORENZ’s early ethological writings, but as dure must make certain implicit ‘assumptions’ as to an epistemological idea explicitly formulated for how objects in the external world behave, this kind the first time in early philosophical works such as of perception can be fooled by experimentally creat- the Russian Manuscript and defended throughout ing a situation that is unlikely to occur under nor- LORENZ’s philosophical, methodological, and bio- mal conditions and violates these ‘assumptions’ logical writings (1943, 1948, 1950, 1958, 1959, (1959, pp297ff; 1981, p41). In the preceding de- 1963a, 1963b, 1977, 1981). The most elaborated ac- scription of a ratiomorphic process the use of terms count is given in the article Gestalt perception as a refering to rational thought in scare quotes indi- source of scientific knowledge (1959). In Germany and cates the analogy with rational cognitive processes. Austria Gestalt psychology was an influential school The ratiomorphic cognitive apparatus—which is before the second world war (ASH 1998), and so evolved—has built-in ‘assumptions’ and ‘hypothe- LORENZ could make use of ideas available from the ses’ about how the external world looks like. As German-speaking psychological tradition. In fact, LORENZ assumes that Gestalt perception is impor- as a student LORENZ took classes and interacted with tant for hypothesis-free observation, these built-in the Vienna psychologist Karl BÜHLER and his assis- implicit ‘hypotheses’ need to be kept apart from the tant Egon BRUNSWIK, both of whom studied the psy- hypothesis that a scientists consciously entertains. chology of perception. On LORENZ account, the dis- Both in rational and ratiomorphic mechanisms the tinctive feature of Gestalt perception is that it is a reliability of the inference is proportional to the ratiomorphic process. LORENZ borrows this term in broadness of the inductive basis (1959, p315). the post-war period from BRUNSWIK (1952, 1955), LORENZ repeatedly mentions that Gestalt perception but the idea was present in BRUNSWIK’s post-war is analogous to rational thought in that it makes in- work even though he did not use this particular ferences and conclusions, albeit in an unconscious term for it (see BRUNSWIK 1934). Before using BRUN- manner (1948, pp57, 59; 1958, p253; 1959, p283). SWIK’s term LORENZ makes reference to Hermann His discussion of constancy mechanisms shows that von HELMHOLTZ’s concept of unconscious inference the kind of inference that is made is similar to ratio- (1948, p57; see HELMHOLTZ 1925, vol. 3.) Talking nal abstraction, which is an important part of induc- about ratiomorphic mechanisms means that we are tion (see also 1958, p252). Form constancy is a dealing with a non-rational, subconscious process, mechanism of pattern recognition; from various that nonetheless exhibits strong analogies to ratio- bits of incoming stimuli the features enabling the nal thought—in particular induction—with respect recognition of objects are mediated. At a few places to how it operates (1948, p55; 1959, pp296, 302; LORENZ explicitly states that Gestalt perception is 1981, p41). For LORENZ Gestalt perception as a spe- analogous to rational abstraction (1951, p173; cial kind of perception is obviously a neurophysio- 1958, p252; 1959, pp283, 304). logical mechanism. LORENZ uses form constancy as Form constancy is a simple example of Gestalt an example to illustrate a simple type of Gestalt per- perception, but there are more complicated types of ception (1959, pp302f). When an observer sees an it. Gestalt perception is able to extract similarities object from different sides (e.g., because the object and regularities out of data sets consisting of miscel- moves and turns), the object is recognized as the laneous items. For instance, it is able to recognize same entity with the same form even though the objects and individuals given by perceptual infor- image on the retina changes continuously. The hu- mation. Moreover, it can detect natural kinds and man perceptual apparatus is able to extract from the categories that exists in nature. By means of Gestalt different sensory data the relevant information that perception different individuals are perceived to be- makes one see the same object at different in- long to the same species, genus, or family (1959, stances. This process obviously works uncon- pp306f). Gestalt perception is an important tool in sciously. The individual is not aware of the steps systematics. This is of fundamental importance for taken by the nervous system to process informa- LORENZ because he views biological features as taxo- tion; only the result (e.g., the recognition of the ob- nomic characters that can be homologized. Ethol- ject) is mediated to consciousness (1959, p296; ogy as a comparative-phylogenetic approach to be- 1981, p43). Nonetheless, this subconscious mecha- havior can makes fruitful use of Gestalt perception.

Evolution and Cognition ❘ 161 ❘ 2003, Vol. 9, No. 2 Ingo Brigandt

In Gestalt perception “many elements are always processes. BRUNSWIK assumes that there is contin- combined into one entity” (1948, p58). This is an is- uum between measurement and perception, so that sue where the idea of the quality of a Gestalt is measurement is just a limiting case of perception stressed by LORENZ. On his account a Gestalt quality (1934, p9). LORENZ, in contrast, seems to make a is a type, something beyond concrete individuals sharper distinction between rational and ratiomor- (1959, pp306ff). Finally, this cognitive process can phic processes. even be used to discover empirical principles and The cognitive capacities that are important for law-like relations out of a large amount of informa- LORENZ are induction and Gestalt perception. In tion containing otherwise irrelevant data (1948, this sense, LORENZ’s epistemology includes nothing p64; 1959, pp282f, 310). This makes Gestalt percep- but these two features. This is clear from the Russian tion a powerful cognitive mechanism. In short, Ge- Manuscript (1948) and from Gestalt perception… stalt perception detects natural units and unex- (1959). LORENZ’s more general evolutionary episte- pected principles and lawful regularities. For this mology (see for instance LORENZ 1977) is well- reason, in the Russian Manuscript this latter complex known, so one needs to give some remarks about function of Gestalt perception is called intuition the relationship between evolution and the already (1948, pp30, 54). discussed cognitive processes. In fact, LORENZ for- Gestalt perception is claimed to be important for mulated some of his ideas about evolutionary epis- all three stages of empirical research, as it seems be- temology before giving his first detailed account of cause Gestalt perception takes in the data from the induction and Gestalt perception in the Russian first two stages (operating independent of hypothe- Manuscript (see LORENZ 1941, 1943). For LORENZ it is ses) and creates the abstractions and general princi- clear that our perceptive apparatus, including Ge- ples that are characteristic for the third, the nomo- stalt perception, and its capacities to generate reli- thetic stage (1948, p30; 1963a, p7). In addition, able knowledge is the product of evolution (1958, LORENZ says that not only research based on rational pp252f; 1959, p289; 1963a, p6). However, LORENZ induction proceeds according to the three stage does not give an elaborated account of the evolu- model, but that Gestalt perception itself implicitly tion of the human ratiomorphic apparatus. Gestalt proceeds in this fashion (1959, p283). perception is discussed in section 7.2 of Behind the Comparing LORENZ’s views on Gestalt perception Mirror (1977), but the discussion is very short (6 with the theory of the early Egon BRUNSWIK, by pages) and does not go much beyond LORENZ’s whom LORENZ was probably influenced, reveals usual remarks. Gestalt perception is not only some commonalities.4 BRUNSWIK emphasizes the evolved, but there are learned aspects of it (1951, fact that perception is analogous to reasoning pp166ff; 1977, p216). Gestalt perception needs (1934, pp2, 50, 127), but in contrast to reasoning it training and it is differently developed in different is immediate, i.e., only the result of the perceptive individuals (1959, p313; 1973, p8). So LORENZ could process is mediated to consciousness, but interme- have given an account of how the innate and diate steps are not (p1). Like LORENZ, BRUNSWIK as- learned aspects of Gestalt perception integrate. In sumes that perception is directed at gaining knowl- addition, not only ratiomorphic, but also rational edge about the objective properties of objects (1934, cognitive processes are evolved and part of LORENZ’s p.V; BRUNSWIK 1937), which was an important issue evolutionary epistemology. But LORENZ does not for Karl BÜHLER’s general approach, too. Even give a detailed account of how rational and ratio- though perception can be fooled under experimen- morphic processes integrate and how the distinc- tal conditions, BRUNSWIK states that under normal tion between these different types of cognitive pro- conditions perception is very reliable. In fact, he cesses came about in the course of evolution. compares perception with instinct/conditioning. Independent of evolutionary issues, in the post-war Instincts and conditioned behavior operate reliably period LORENZ stresses the impotence of Gestalt per- in standard cases, but in contrast to insight they are ception as a cognitive mechanism. However, he inflexible so that they cannot adapt do new situa- does not offer a development and elaboration of his tions. Perception has a performance that is similar ideas and does not refer to the psychological litera- to instincts or conditioned behavior (1934, ture (in particular new results) on perception. pp114ff). While LORENZ might agree with this anal- Rather, LORENZ repeats what are in his view the ba- ogy he does not make use of it, but contrasts ratio- sic features and faculties of Gestalt perception and morphic processes—which are usually reliable but sometimes mentions a few classical ideas of Gestalt cannot adapt to new situations—just with rational psychologists.

Evolution and Cognition ❘ 162 ❘ 2003, Vol. 9, No. 2 Gestalt Experiments and Inductive Observations

Mapping Gestalt Perception/Induction fication is obtained through induction. So far the iden- onto Observation/Experiment tification of this epistemic relationship between Ge- stalt perception and induction in LORENZ’s work has The first issue to be discussed is the question of in mainly been based on the Russian Manuscript. what manner the epistemological apparatus (ratio- Whereas these two different roles are very explicit in morphic Gestalt perception and rational induction) this piece of work, some passages of the later article relates to the ethological methodology (observation Gestalt perception… give the same picture (1959, and experiment) as LORENZ viewed it. Some of p316). LORENZ’s writings (especially the Russian Manuscript) On the one hand, LORENZ three stage model of sci- suggest an epistemic asymmetry between Gestalt ence sounds like a sort of naïve Baconianism, ac- perception and induction and attribute different cording to which one first needs to collect in an un- roles in scientific research to these mechanisms, so biased and theory-independent manner as much that a straightforward Gestalt perception–observa- data as possible and only then one can generalize tion and induction–experiment correspondence laws and theories from it. On the other hand, the seems plausible. idea of Gestalt perception proposing hypothesis to The different epistemic roles of induction and be rationally tested might incline one to assume that Gestalt perception are due to the fact that the LORENZ has a hypothetico–deductive account of sci- former is a rational process, but the latter is not. In- ence. Neither is really the case. LORENZ definitely is duction is a tool for scientific analysis and for the not a Popperian falsificationist. On his account, hy- confirmation of scientific claims. It is a cognitive pothesis cannot only be disconfirmed, but also con- process that aims at objective knowledge. The Ge- firmed by evidence. LORENZ accepts induction, in stalt, on the other hand, is a “purely subjective phe- fact, he states that one can “use the breadth of the nomenon” (1948, p140).5 The result of this percep- inductive basis to assess with genuine mathematical tive cognitive process is mediated as a whole to accuracy a probability value for the correctness of subjective experience. Without its parts being open the result. Where that breadth is sufficient, the prob- to analysis, the Gestalt is seen and the received en- ability is so close to certainty that we can confidently tity accepted (1948, p58). This epistemic difference equate the two.” (1948, p64) This makes him sound is in particular shown by the fact that a result ob- somewhat like a proponent of the modern Bayesian tained by Gestalt perception is refuted by inductive approach to theory confirmation by evidence. How- evidence contradicting that result (1948, p68). The ever, Bayesianism is just about confirmation and strength of induction is its accountability, while Ge- does not tell us how to come up with theories to be stalt perception is characterized by its incorrigibility verified. The same applies for a naïve Baconian ap- and unaccountability (1948, p64). For this reason, proach, which cannot make sure that anything the- Gestalt perception and induction play a different oretically useful emerges from unbiased and unfo- role in scientific research. Whereas induction is cussed data gathering. The discovery of theories was shortsighted and therefore not a good tool for scien- somewhat neglected by traditional philosophy of tific discovery (1948, p65), Gestalt perception is able science, but LORENZ addresses this issue by pointing to see unexpected regularities and unforeseen law- to Gestalt perception, which is supposed to deliver fulness (1948, p63; 1959, p282; 1981, p44). It is a unexpected principles and hypotheses. This is good “hunch generator” (1963a, p7). As only Ge- maybe the most fruitful aspect of LORENZ’s epistemo- stalt perception is really able to fulfill this function, logical perspective (besides the fact that he makes its role in scientific research is to lead the way of dis- use of an evolutionary framework). LORENZ proposes covery (1948, p56, 64; 1963a, p8). Gestalt percep- processes that generate theories as well as confirm tion is only a means of discovery. A detected princi- them. Rational and ratiomorphic processes are con- ple has to be substantiated by induction, it is the job sidered psychologically founded mechanisms, of rational processes to confirm scientific claims. which in combination address the two most impor- “Therefore, only induction can increase the reli- tant aspects about scientific rationality—discovery ability of a result and only induction can validate a and justification of theories. result that has been obtained exclusively through As discussed in the preceding section, experiment intuition!” (1948, p65) are methodologically and logically posterior to ob- This yields the following account. The (subjective) servation. Useful experiments can only be con- discovery of scientific principles is achieved by Gestalt ducted after a good deal of observation has been un- perception, while the (objective) confirmation and justi- dertaken. In fact, observation gives a meaningful

Evolution and Cognition ❘ 163 ❘ 2003, Vol. 9, No. 2 Ingo Brigandt account of the structure and function of the ob- “It is quite definitely a simplification of this kind served system and experiments can only then verify to represent the interaction between the various details about the components of the behavior of a cognitive processes, as I have done above, as if there species and analyze their causal interaction. In an were always a distinct separation between the prior analogous manner, Gestalt perception has to be em- discovery of an inherent principle through ratio- ployed first to detect interesting principles that are morphic processes and its subsequent verification to be confirmed by induction. This suggests that the through rational processes.” (1959, p320) two step procedure Gestalt perception–induction Quantification, which belongs to inductive anal- can be mapped onto the observation–experiment ysis, is claimed to be dependent on Gestalt percep- procedure in a manner such that both correspond tion (1958, p256; 1959, p320). In fact, Gestalt per- to each other. This interpretation means that obser- ception is a precondition for rational reasoning at vation in the ethological approach is largely driven all (1948, p30). On the other hand, “rational, quan- by the cognitive process of Gestalt perception, while tifying, statistical and surveying pre-treatment is mainly the cognitive function of induction is used necessary to permit Gestalt-formation” (1959, in the experimental phase. The fact that the experi- p320). There is a necessary cooperation of Gestalt ments fulfill the analytic demands of inductive sci- perception and analysis/experiment (1981, p54). ence became clear in my exposition of this method. Rational and ratiomorphic processes are strongly As will be discussed in more detail in the next sec- entangled and the different steps of inductive sci- tion, Gestalt perception actually corresponds in sev- ence need Gestalt perception (1948, pp28ff). In the eral respects to observation. Indeed, a main func- Russian Manuscript there are passages that some- tion of LORENZ’s account of the importance of times suggest a strict separation and at other places Gestalt perception is to defend his observational ap- an interdependency of both types of processes. Ge- proach. stalt perception… proceeds by first suggesting a clear However, both in the Russian Manuscript and in distinction between rational and ratiomorphic Gestalt perception… there are passages that do not fit functions, which is then explicitly relativized (com- the simple interpretation given so far. The relation pare pp304–319 with 319–322). between Gestalt perception and induction is more Thus LORENZ states that the relation between ra- complex. First, according to the references given tional and ratiomorphic processes is complex and above Gestalt perception and induction have an that both processes need to be highly entangled in epistemically different status insofar as only the lat- the effective generation of knowledge. Neverthe- ter provides objective knowledge and in the case of less, he does not explain sufficiently how he views conflict the result obtained by a rational process is the interplay between these two processes in detail. to be preferred. Nonetheless, Gestalt perception and (In Gestalt perception… he basically devotes two induction are not completely different with respect pages to this issues; see 1959, pp320f). The difficulty to their objectivity. In particular when defending is that LORENZ tries to combine rational and ratio- the qualitative approach of ethology against views morphic processes (which are of a different nature that consider only quantitative measurements as on his account) without a clear account of how they objective, LORENZ’s points out that all cognitive pro- combine in practice. Despite these complexities, on cesses are in a sense subjective (1959, p320). In addi- my interpretation LORENZ still sees a parallel be- tion, perception tells us about the properties of ob- tween the ethological methodology (observation jects in the external reality (1959, p301). In fact, it is and experiment) and the philosophical epistemol- the only source of knowledge about the reality sur- ogy (Gestalt perception and induction). Not only rounding us (1981, p41). Under standard condi- are Gestalt perception and induction strongly en- tion—given that the ‘evidential basis’ is right—re- tangled, observation and experiment also inter- sults obtained by Gestalt perception are true (1948, grade. Both are needed for ethology and in practice p62). Thus Gestalt perception is reliable, despite the there is an interplay between them. Observation fact that its results need to be rationally verified by usually precedes experiment and suggests relevant induction. experimental questions. Correspondingly, Gestalt Second, in Gestalt perception… LORENZ explicitly perception leads the way of scientific investigation states that the ratiomorphic mechanism of Gestalt by proposing hypotheses that need to be confirmed perception and the rational process of induction by means of induction. LORENZ makes clear that Ge- cannot be sharply separated and that their func- stalt perception is crucial for observation, whereas tions (discovery/verification) intergrade. experiments fulfill the role of scientific induction.

Evolution and Cognition ❘ 164 ❘ 2003, Vol. 9, No. 2 Gestalt Experiments and Inductive Observations

Lorenz’s Justificatory Efforts Entirety ^ Gestalt perception

A good deal of LORENZ’s writings are devoted to the LORENZ’s uses this second complex—the ideas about defense of his views on biology and scientific knowl- entirety—to justify the need for Gestalt perception edge. The next question with respect to the relation- as a cognitive tool. Gestalt perception is ideal for ship of observation, experiment, Gestalt perception, studying complex systems and the nervous system/ and induction is how these items justify each other. behavior are especially complex (1959, p283; 1981, For instance, does LORENZ’s position on how obser- p46). One of the main steps in dealing with an or- vations are to be made imply the need for Gestalt ganic entirety is to get an overview of its parts. The perception as a cognitive mechanism used in etho- best way to do so is to make use of Gestalt percep- logical research, or rather are philosophical remarks tion (1981, p47). The entirety approach makes it on Gestalt perception intended to justify the specific necessary to study the function of a system in its ethological observational approach? In LORENZ’s context as well, and Gestalt perception is a good writings two groups of ideas can be identified that tool for this. In addition, Gestalt perception is a cog- function as justificatory primitive principles. This nitive capacity that is ideal for a comparative ap- means that they are ideas which play an important proach. As already outlined, it is able to detect ge- role in justifying other items of the framework, but neric kinds and natural units (1948, pp60f; 1959, which are not themselves justified by other ideas. A pp306ff; 1981, pp45f). In general, Gestalt psychol- group of primitive principles is largely taken for ogy was a very useful approach for LORENZ’s perspec- granted or justified internally by claims from the tive. It was a tradition that—for the most part— same group of ideas. (Figure 1 gives a rough illustra- aimed at a rigorous and scientific causal-explana- tion of the justificatory relationships existing be- tory framework, in accordance with LORENZ empha- tween the different items of LORENZ’s theoretical ac- sis on causal-analytical thinking in science as op- count.) posed to psychological–teleological speculations The first group of ideas are LORENZ’s views on in- (1942). But Gestalt psychology was not just reduc- duction, which have already been discussed. These tionistic and instead provided an objective account are his general ideas on science and rational reason- of holistic features inherent in the perception of the ing, including rational induction as a cognitive Gestalt as a quality. The perception of some objects mechanism and in particular LORENZ’s views of how is different from the collection of individual percep- empirical science works, i.e., the distinction be- tions; the Gestalt is a unity. Thus Gestalt psychology tween the idiographic, systematic, and nomothetic allowed LORENZ to maintain his rigorous–scientific stages of science. (There are obviously other views of attitude and emphasize the fact that biological ob- science that do not assume a theory-free idiographic jects also need to be understood and studied as com- stage.) The second complex of primitive princi- plex wholes (see 1948, 1951). ples—which despite its paramount importance for LORENZ has not yet been discussed—are related to Entirety ^ observation LORENZ’s understanding of “Ganzheit”, often trans- lated as entirety or (systemic) entity (see 1948, The ideas about entirety also justify observation as pp137ff; 1950, pp120f; 1958, pp248f; 1959, pp281f; LORENZ understands this method. The recognition 1981, pp36ff). The main ideas in this complex are of an individual as an organismic entity that has a claimed to stem from the recognition that biology functional context and a history implies the need deals with complex wholes. A systemic entity or an for studying healthy animals in their natural envi- entirety is a system that has subsystems that mutu- ronment (1948, pp213, 221, 223; 1981, p40). An- ally interact which each other. Every part depends other point constitutive of LORENZ’s observational on the other parts. For LORENZ this implies that it is approach is the fact that the researcher has to be fa- indispensable to study the complex system as a miliar with all details of the behavioral patterns of whole. In the case of biological entities this means an organism. Again, the systemic entity approach that one has to take both the structure and the func- makes this obvious (1948, p215; 1981, p38). In tions of the entirety and its subcomponents into other words, recognizing the demands of an analy- consideration. Research on organisms includes the sis on a broad front in the case of organisms and study of their overall behavior and their environ- their behavior means that observations have to be ment. LORENZ calls this approach ‘analysis on a performed (at least in some aspects) in the manner broad front’.6 of classical ethology.

Evolution and Cognition ❘ 165 ❘ 2003, Vol. 9, No. 2 Ingo Brigandt

Induction ^ gestalt perception gous relation that experiment is subordinated to ob- servation might suggest that the conduct of The preceding discussion has already pointed to the experiments justifies the need for observations. fact that LORENZ’s views on induction justify the use Some of LORENZ’s remarks to the effect that experi- of Gestalt perception. Induction has the function of ments do not make sense without observation seem confirming scientific claims. But as it is short- to amount to a direct justification of observation by sighted, Gestalt perception as a mechanism that experiment. However, it hardly amounts to a justifi- finds unforeseen lawfulness is needed to carry out cation of the specific characteristics of LORENZ’s ob- successful inductive science. Gestalt perception servational approach. I prefer the interpretation that leads induction the way (1948, pp56, 64). LORENZ’s the need for experiments mainly justifies the use of account of how the rational and ratiomorphic pro- Gestalt perception, which in turn calls for observa- cesses operate makes clear that induction can hardly tion (as we will se below). In other words, the neces- do without Gestalt perception. sity of experiments justifies the specific aspects of the observational approach only indirectly, namely in- Induction ^ observation sofar as Gestalt perception is the tool for conducting observations. The use of Gestalt perception follows The claims about induction justify directly the ap- from the need of doing experiments because experi- proach to observation. LORENZ emphasizes that the ments are useful only if the substructures and func- nomothetic stage can only be reached after the hy- tions of a system are known to a certain degree pothesis-free idiographic phase (1948, p216; 1950, (1981, pp53, 65). Gestalt perception is the best cog- p129). The crucial function of the idiographic stage nitive mechanism to tackle this question (1981, is to provide an inductive basis that is as broad as p47). It can break down a complex system into possible including all relevant features (1948, p213; meaningful parts on which an experiment can focus. 1950, p131). New explanatory principles can only be abstracted from a sufficiently broad base (1958, Gestalt perception ^ observation p250). Furthermore, the premature formation of a hypothesis can distort the inductive basis. When a The following quotations illustrate how self-evident researcher has a specific theory in mind it is likely it is for Konrad LORENZ that observation and Gestalt that the data of the idiographic stage are biasedly perception go together: evaluated so that the scientist sees his or her own “… in the observation of complex animal behav- hypothesis supported (1948, pp31, 71). This gives a iour patterns, one can literally see the same process justification for the idea that observation has to pro- thousands of times without noticing the inherent vide a large body of data, that it includes all features principle until—quite abruptly—on the following of the organism and its environment, and in partic- occasion the Gestalt is distinguished from the back- ular that it has to be conducted in hypothesis-free ground of accidental features …” (1959, p306) way (1959, p281). “At the other extreme are the died-in-the-wool behaviourists who deny that Gestalt perception— Induction ^ experiment and thus observation of organisms in their natural environment—has any value or even scientific char- Induction also implies the need for experiments. An acter.” (1959, p319) important part of scientific induction is to provide In fact, the use of Gestalt perception actually jus- an account of the details of the object under study, tifies the ethological approach to observation. First, to give a causal analysis, as well as to confirm hy- the effectiveness of Gestalt perception in finding potheses. As has already been discussed the main important regularities and principles is propor- function of ethological experiments is to establish tional to the amount of data of which Gestalt per- claims about the specific components of behavior ception can make use (1948, p63; 1959, p305; 1981, and to examine their (causal) relationship. p46). An advantage of this ratiomorphic process is its retentive memory and the large amount of data it Experiment ^ Gestalt perception can take into account (1959, pp309, 314f; 1981, p44). Gestalt formation can suddenly occur after a I argued above that the inductive approach justifies long period of data collection (1959, p306; 1981, the use of Gestalt perception because successful in- p45). This lends support to the habit of observing duction presupposes Gestalt perception. The analo- animals over a long span of time (1959, p315; 1981,

Evolution and Cognition ❘ 166 ❘ 2003, Vol. 9, No. 2 Gestalt Experiments and Inductive Observations

p45). In addition, Gestalt perception deals with the Entirety Gestalt Induction whole (1948, p139; 1959, p306; 1981, p46). It is able Perception to get an overview of a multitude of aspects of a sys- tem and can extract a lawful relationship out of it (1948, p64). For this reason, effective use of Gestalt Ethological perception can be made if all behavioral patterns of Approach to Experiment organisms are observed in their natural context. As Observation a perceptual mechanism Gestalt perception has ABThe necessity of B is conceptually contained in A; qualitative information as input. This justifies the accepting A means also doing B qualitative approach to observation (1948, p64). Fi- ABB fulfills ideally the demands of A nally, Gestalt perception does not need a hypothesis ABThe cognitive mechanism A is a powerful tool for (1948, p63). It works best when the researcher is re- obtaining knowledge from the biological method B laxed and contemplates his object of study while unconsciously collecting data (1959, p316; 1981, Figure 1: Overview of LORENZ’s justificatory system. pp45f). In fact, rational reasoning about details of the object negatively influenced the operation of Gestalt perception (1959, pp314, 317). In this case, ment relation, it is not clear in detail what the actual the features of Gestalt perception show that obser- role of rational processes in observation is and to vation has to be conducted without rational influ- which extent Gestalt perception is also important ences based on theories or hypothesis. To sum up, for ethological experiments. LORENZ tries to make clear that the distinctive as- The result of my reconstruction of LORENZ’s justi- pects of the ethological observational approach is ficatory relationships between the ideas of entirety, chosen in accordance with a cognitive enterprise Gestalt perception, induction, observation and ex- that relies on Gestalt perception. Observations yield periment are illustrated in Figure 1.7 LORENZ’s work effectively knowledge because this powerful ratio- exhibits a justificatory system in which all of his rel- morphic mechanism is employed. The idea that Ge- evant ideas are logically related. What becomes stalt perception is predominantly used for ethologi- clear is the fact that the ideas about entirety/sys- cal observations (as well as for systematics) can be temic entity (Ganzheit) and the claims about induc- found throughout LORENZ’s writings (1948, pp61f, tion and inductive science function as primitive 213ff; 1954, pp197ff; 1958, pp251ff, 278; 1959, principles. They are not justified by other items, but pp306ff, 311, 315; 1963a, pp8f; 1981, pp40ff). both are used to justify the need for Gestalt percep- tion as a cognitive tool and the necessity of observa- Conclusion tions in an ethological manner. Furthermore, three different items lend support to the use of Gestalt I have argued that on LORENZ’s view the observa- perception as a cognitive tool: the views about en- tion–experiment relation is largely analogous to the tirety, induction, and experiments. It is not very sur- Gestalt perception–induction relation. Experiment prising that Gestalt perception is justified from dif- needs foregoing observation so that it can be con- ferent perspectives. While this cognitive ducted meaningfully. Similarly, induction needs mechanism is very important for LORENZ’s episte- Gestalt perception that leads induction the way. Ob- mology, it is hardly endorsed by other post-war biol- servation and Gestalt perception discover new prin- ogists. On LORENZ’s account Gestalt perception has ciples, whereas experiment and induction have to very strong capacities, e.g., finding scientific hy- give an analytic confirmation. Gestalt perception is potheses. LORENZ needs to justify the use and need the crucial cognitive tool for ethological observa- of this non-rational (albeit ratiomorphic) process tion, while rational induction is important for ex- that he uses to stress as a tool for scientific inquiry. perimental analysis and confirmation. LORENZ Finally, my analysis shows that the primary aim makes clear that the relation between Gestalt per- of LORENZ’s justificatory efforts is the ethological ap- ception and induction is very complex. However, he proach to observation. Even though LORENZ associ- does not give a concrete account of the interplay be- ates Gestalt perception with observation, it is not tween these two cognitive mechanisms and their re- the case that the specific features about the observa- lation to the biological approach. For this reason, tional approach are used to justify the use of Gestalt while the Gestalt perception–induction relation can perception. Rather, the ideas about Gestalt percep- be largely mapped onto the observation–experi- tion justify the way observations were carried out by

Evolution and Cognition ❘ 167 ❘ 2003, Vol. 9, No. 2 Ingo Brigandt

the founders of ethology. Indeed, LORENZ’s remarks ommendations given by reviewers about a grant ap- about Gestalt perception are intended to show that plication to the Deutsche Forschungsgemeinschaft there is an important and powerful cognitive mech- (German Research Council). anism that is able to get knowledge out of what has “The grant was approved, but not without the ad- been observed. The discussion shows that there ex- visory board adding a little benign admonition: care ist several items that are used to justify the observa- should be taken, lest the investigation lapse into tional approach, but the tenets about observation merely being descriptive, “daß die Untersuchung are not employed to justify any of the other items. nicht ins Deskriptive abgleitet.” (1973, p5). Induction and Gestalt perception justify the hy- In addition, LORENZ reacted furiously to the at- pothesis-independency of observation, and the tempt to remove Otto KOENIG as the director of the qualitativeness of the approach is substantiated by Austrian Institut für Vergleichende Verhaltensfors- Gestalt perception. The ideas about entirety justify chung (Institute for Comparative Ethology). Some the fact that organisms are to be studied in their nat- of the board members stated that the research at the ural environment, while the view that all features institute was in bad shape. The justification is re- have to be observed is supported by all three items vealing. On their account, comparative ethology is (induction, Gestalt perception, and entirety). a amateur program instead of rigorous biology—just From a historical point of view it is highly plausi- making photographs does not presuppose a univer- ble that the observational method is the main target sity degree and the crucial feature of making mea- of Konrad LORENZ’s justificatory efforts. For one of surements is not seriously pursued. the main intentions of LORENZ’s post-war defense of “In einer Sitzung des Kuratoriums für das Institut his works was to show that non-quantitative, largely für Vergleichende Verhaltensforschung haben MA- theory-independent observations are in fact scien- RINELLI und der Physiologe Prof. KMENT den Versuch tific (this is explicit in 1958, pp246, 256; 1959, p281; unternommen, Otto KOENIG hinauszuschmeißen… 1963a, p1; 1973, p1; 1981, pp40, 68ff). LORENZ felt MARINELLI: Tierhaltung sei nur ein Hobby, die Her- that his ideas on observation were not taken seri- stellung von Filmen verlange kein Hochschulni- ously by many animal psychologists and that he had veau (das Wort Hochschulniveau kommt in dem a difficult task of making clear that qualitative obser- sechs Seiten langen Protokoll der Sitzung 7× vor.) vations are a necessary step of research. Indeed, in Die vergleichende Verhaltensforschung sei ein Dil- his view the biological community as whole focused letantenprogramm. KMENT (S. 2 unten) gibt eine alone on measurement and statistical evaluation of völlig falsche Vorstellung von unserem Institut und data as the ideal of rigorous science. seiner Arbeitsweise. KMENT meint, “entscheidend sei Originally LORENZ discussed the importance of die Messung, man müsse messend vorgehen” und Gestalt perception without suggesting that his der alte Blödsinn usw.” (LORENZ, letter to STRESE- methodological approach might be in need of de- MANN, February 17, 1971; the letter to STRESEMANN fense (1948, 1951). In the 50s and 60s, however, he includes a copy of the minutes of the board meet- criticized especially behaviorist and American ani- ing, showing that these accusations were actually mal psychology in the context of defending his raised). views. LORENZ states, for instance, that the paper Ge- Thus the main target of LORENZ’s justificatory ef- stalt perception… is targeted at American animal psy- forts is his observational approach, and he felt the chologists (1958, p256). However, it is not just the need to defend his methodological views against case that LORENZ had to defend his methodological general trends and conceptions in biology and be- views against the influential tradition of behavior- havioral science. Despite the fact that LORENZ recog- ism and other theoretical approaches that had a rad- nized the importance of experiments and quantita- ically different approach. Some of the remarks in the tive techniques he viewed the scientific contempt writings of the 60s and 70s were directed at biolo- for qualitative observation as part of a generally in- gists in general, because creasing disregard for nature. LORENZ felt that his observa- Author’s address While we may wonder tional approach was consid- whether Gestalt perception is ered by many biologists (in- Ingo Brigandt, Department of History and really as important for etho- cluding German-speaking Philosophy of Science, 1017 Cathedral of logical observations as biologists) as not being really Learning, University of Pittsburgh, Pitts- LORENZ maintained, we are rigorous science. For instance, burgh, PA 15260. Email: [email protected] still not quite clear about the LORENZ complained about rec- epistemic significance and re-

Evolution and Cognition ❘ 168 ❘ 2003, Vol. 9, No. 2 Gestalt Experiments and Inductive Observations

lation of observation and experiment (ALLEN, forth- Acknowledgments coming). LORENZ’s achievement is to make clear that we need to have an account of the cognitive mech- I am indebted to Paul GRIFFITHS, Werner CALLEBAUT, anisms that are at work in observation to generate Wolfgang SCHLEIDT and Manfred WIMMER for help- scientific knowledge. ful discussions on my work.

Notes point to some useful features of LORENZ’s epistemological framework. 1 See BURKHARDT (1999) for a historical comparison of the 4 Both were of the same age and working on their Habilita- ways animals and behavior was studied (in museums, in tion theses at the University of Vienna. While BRUNSWIK zoos, and in the field) and how this related to and influ- was Karl BÜHLER’s assistant, LORENZ took his psychology enced research approaches. BURKHARDT (2003) compares classes with BÜHLER. BRUNSWIK gave LORENZ comments on the research practices of LORENZ and TINBERGEN. the Companion (LORENZ 1935) as regards issues dealing 2 The correspondence between Konrad LORENZ and Erwin with the psychology of perception (LORENZ, letter to STRE- STRESEMANN is archived at the Staatsbibliothek zu Berlin, SEMANN, November 21, 1934). See HOFER (2001) for a discus- Preußischer Kulturbesitz (Germany) as ‘Nachlaß 150 (E. sion of the personal and intellectual relationship between STRESEMANN), Kasten 40’. Any letter from LORENZ that is part BÜHLER, BRUNSWIK, and LORENZ. of this source is referred to in the paper as ‘LORENZ, letter to 5 It is not quite clear what LORENZ means by saying that the STRESEMANN’. The letter quoted from above is not dated (it Gestalt is just subjective. Gestalt psychologists often as- is probably from the end of 1933), ‘Blatt 94–97’ refers the sumed that Gestalt perception aims at objective features of sheet numbers given by the archive. the world (see ASH 1998) and LORENZ does not seem to dis- 3 Some of LORENZ’s views on epistemology and philosophy agree with this in general. of science—in particular as formulated in the Russian 6 In a letter from 1936 LORENZ complains that so many phys- Manuscript—might appear naive, given the fact that there iological chemists are incapable of viewing animals as or- were much more sophisticated approaches to these topics. ganic entireties (“Ich kenne aber so viele Physiologische At the same time LORENZ developed his philosophical Chemiker, die alle ganz unfähig sind, im Tier eine organi- views in the pre-war period in Vienna, philosophers and sche Ganzheit zu sehen”, LORENZ, letter to STRESEMANN, Oc- scientists that were members of the Vienna circle or asso- tober 4, 1936). ciated with it developed detailed accounts as to how un- 7 The use of three kinds of justificatory arrows in the diagram derstand scientific rationality and the justification of is a rough classification. A more detailed and qualified ac- empirical knowledge. LORENZ did not pay attention to count can be found in the above discussion. Note for in- these developments probably because he and the members stance that an arrow justifying the observational approach of the Vienna circle belonged to different philosophical, refers to some subset of the features peculiar to the obser- scientific, and social-political communities. I will later on vational approach.

References the founding of ethology as a scientific discipline. In: Legakis A./Sfenthourakis S./Polymeni R./Thessalou-Legaki Allen, C. (forthcoming) Is anyone a cognitive ethologists? Bi- M. (eds) The New Panorama of Animal Evolution: Proceed- ology and Philosophy. ings of the 18th International Congress of Zoology. Pen- Ash, M. G. (1998) Gestalt Psychology in German Culture soft: Sofia, pp. 329–336. 1890–1967: Holism and the Quest for Objectivity. Cam- Burkhardt, R. W. (forthcoming) Patterns of Behavior: Konrad bridge University Press: Cambridge. Lorenz, Niko Tinbergen, and the Founding of Ethology. Brunswik, E. (1934) Wahrnehmung und Gegenstandswelt: Chicago University Press: Chicago. Grundlegung einer Psychologie vom Gegenstande her. Dewsbury, D. A. (1988) A brief history of the study of animal Franz Deuticke: Leipzig. behavior in North America. In: Bateson P. P. G./Klopfer P. Brunswik, E. (1937) Psychology as a science of objective rela- H. (eds) Perspectives in Ethology. Volume 8: Whither tions. Philosophy of Science 4: 227–260. Ethology? Plenum Press: New York, pp. 85–122. Brunswik, E. (1952) The Conceptual Framework of Psycholo- Heinroth, O./Heinroth, M. (1924–1933) Die Vögel Mitteleu- gy. Chicago University Press: Chicago. ropas, in allen Lebens- und Entwicklungsstufen photogra- Brunswik, E. (1955) ‘Ratiomorphic’ models of perception and phisch aufgenommen und in ihrem Seelenleben bei der thinking. Acta Psychologica 11: 108–109. Aufzucht vom Ei ab beobachtet. 4 volumes. Hugo Ber- Burkhardt, R. W. (1981) On the emergence of ethology as a müller: Berlin. scientific discipline. Conspectus of History 1: 62–81. Helmholtz, H. (1925) Treatise on Physiological Optics. The Burkhardt, R. W. (1999) Ethology, natural history, the life sci- Optical Society of America. Ithaca: New York. ences, and the problem of place. Journal of the History of Hofer, V. (2001) Konrad Lorenz als Schüler von Karl Bühler: Biology 32: 489–508. Diskussion der neu entdeckten Quellen zu den persönli- Burkhardt, R. W. (2003) Konrad Lorenz, Niko Tinbergen, and chen und inhaltlichen Positionen zwischen Karl, Bühler,

Evolution and Cognition ❘ 169 ❘ 2003, Vol. 9, No. 2 Ingo Brigandt

Konrad Lorenz und Egon Brunswick. Zeitgeschichte 28: Press, Cambridge MA, pp. 281–322. 135–159. Lorenz, K. (1963a) A scientist’s credo. Reprinted in: Lorenz, Lillie, F. R. (1911) Charles Otis Whitman. Journal of Morphol- K. (1971) Studies in Animal and Human Behaviour, Vol- ogy 22: xv–lxxvii. ume II. Harvard University Press, Cambridge MA, pp. 1–13. Lorenz, K. (1931) Contributions to the study of the ethology Lorenz, K. (1963b) Do animals undergo subjective experi- of social Corvidae. Reprinted in: Lorenz, K. (1970) Studies ence. Reprinted in: Lorenz, K. (1971) Studies in Animal in Animal and Human Behaviour, Volume I. Harvard Uni- and Human Behaviour, Volume II. Harvard University versity Press: Cambridge MA, pp. 1–56. Press, Cambridge MA, pp. 323–337. Lorenz, K. (1932) A consideration of the methods of identifi- Lorenz, K. (1965) Evolution and Modification of Behavior. cation of species-specific instinctive behaviour patterns. University of Chicago Press: Chicago. Reprinted in: Lorenz, K. (1970) Studies in Animal and Hu- Lorenz, K. (1973) The fashionable fallacy of dispensing with man Behaviour, Volume I. Harvard University Press, Cam- description. Die Naturwissenschaften 60: 1–9. bridge MA, pp. 57–100. Lorenz, K. (1977) Behind the Mirror. Methuen: London. Lorenz, K. (1935) Companions as factors in the bird’s envi- Lorenz, K. (1981) The Foundations of Ethology. Springer Ver- ronment: The conspecific as the eliciting factor for social lag: New York. behaviour patterns. Reprinted in: Lorenz, K. (1970) Studies Lorenz, K./Tinbergen, N. (1938) Taxis and instinctive behav- in Animal and Human Behaviour, Volume I. Harvard Uni- ior pattern in egg-rolling by the Greylag goose. Reprinted versity Press, Cambridge MA, pp. 101–258. in: Lorenz, K. (1970) Studies in Animal and Human Behav- Lorenz, K. (1937) The establishment of the instinct concept. iour, Volume I. Harvard University Press, Cambridge MA, Reprinted in: Lorenz, K. (1970) Studies in Animal and Hu- pp. 316–350. man Behaviour, Volume I. Harvard University Press, Cam- Nissen, H. W. (1931) A Field Study of the Chimpanzee: Obser- bridge MA, pp. 259–315. vations of Chimpanzee Behavior and Environment in Lorenz, K. (1941) Kants Lehre vom Apriorischen im Lichte Western French Guinea. John Hopkins Press: Baltimore. gegenwärtiger Biologie. Blätter für Deutsche Philosophie Röell, D. R. (2000) The World of Instinct: Niko Tinbergen and 15: 94–125. the Rise of Ethology in the Netherlands (1920–1950). Van Lorenz, K. (1942) Inductive and teleological psychology. Re- Gorcum: Assen. printed in: Lorenz, K. (1970) Studies in Animal and Human Schleidt, M./Schleidt, W. (1958) Kurven gleicher Lautstärke Behaviour, Volume I. Harvard University Press, Cambridge beim Truthhahn (Meleagris gallopavo). Die Naturwissen- MA, pp. 351–370. schaften 45: 119–120. Lorenz, K. (1943) Die angeborenen Formen möglicher Erfahr- Tinbergen, N. (1932) Über die Orientierung des Bienenwolfes ung. Zeitschrift für Tierpsychologie 5: 235–409. (Philanthis triangulum Fabr.). Zeitschrift für vergleichende Lorenz, K. (1948) The Natural Science of the Human Species: Physiologie 16: 305–334 An Introduction to Comparative Research. The “Russian Tinbergen, N. (1935) Über die Orientierung des Bienenwolfes manuscript” (1944–1948). Edited from the author’s post- (Philanthis triangulum Fabr.) II: Die Bienenjagd. humous works by A. von Cranach in 1996. MIT Press: Zeitschrift für vergleichende Physiologie 21: 699–716. Cambridge MA. Tinbergen, N. (1951) The Study of Instinct. Oxford Universi- Lorenz, K. (1950) Part and parcel in animal and human soci- ty Press: Oxford. eties. A methodological discussion. Reprinted in: Lorenz, Tinbergen, N./Kuenen, D. J. (1939) Über die auslösenden K. (1971) Studies in Animal and Human Behaviour, Vol- und richtungsgebenden Reizsituationen der Sperrbewe- ume II. Harvard University Press, Cambridge MA, pp. 115– gung von jungen Drosseln (Turdus m. merula L. und T. e. 195. ericetorum Turton). Zeitschrift für Tierpsychologie 3: 37– Lorenz, K. (1951) The role of Gestalt perception in animal 60. and human behaviour. In: Whyte, L. L. (ed) Aspects of Watson, J. B. (1908) The behavior of the soddy and nooty Form: A Symposium on Form in Nature and Art. Lund terns. Carnegie Publications 103: 187–255. Humphries: London, pp. 157–178. Watson, J. B. (1914) Behavior: An Introduction to Compara- Lorenz, K. (1954) Psychology and phylogeny. Reprinted in: tive Psychology. Henry Holt: New York. Lorenz, K. (1971) Studies in Animal and Human Behav- Whitman, C. O. (1899a) Animal Behavior. Reprinted from iour, Volume II. Harvard University Press, Cambridge MA, the Biological Lectures of the Marine Biological Laborato- pp. 196–245. ry, Woods Hole, 1898. Ginn: Boston. Lorenz, K. (1958) Methods of approach to the problem of be- Whitman, C. O. (1899b) Myths in animal psychology. The havior. Reprinted in: Lorenz, K. (1971) Studies in Animal Monist 9: 524–537. and Human Behaviour, Volume II. Harvard University Windelband, W. (1894) Geschichte und Naturwissenschaft. Press, Cambridge MA, pp. 246–280. Heitz und Mündel: Straßburg. Lorenz, K. (1959) Gestalt perception as a source of scientific Yerkes, R. M. (1928) The Mind of a Gorilla, Parts 1 and 2. knowledge. Reprinted in: Lorenz, K. (1971) Studies in Ani- Clark University: Worcester. mal and Human Behaviour, Volume II. Harvard University

Evolution and Cognition ❘ 170 ❘ 2003, Vol. 9, No. 2