494 Opinion TRENDS in Cognitive Sciences Vol.7 No.11 November 2003

Primate cognition: from ‘what now?’ to ‘what if ?’

Louise Barrett1, Peter Henzi2 and Robin Dunbar1

1School of Biological Sciences, University of Liverpool, Biosciences Building, Crown St, Liverpool, UK L69 7ZB 2Behavioural Ecology Research Group, University of Natal, 4041 Durban, South Africa

The ‘social brain’ hypothesis has had a major impact on unique features of human cognition, such as language the study of comparative cognition. However, despite a and metacognition [11]. strong sense, gained from both experimental and Here, we present a new hypothesis to explain the observational work, that monkeys and differ from differences between monkeys and apes, based on an each other, we are still no closer to understanding understanding of the ecological constraints under which exactly how they differ. We hypothesize that the dis- these animals operate. Specifically, we argue that the persed social systems characteristic of societies dispersed social systems (‘fission–fusion’ societies) in explains why monkeys and apes should differ cogni- which most (if not all) of the hominoids (apes and humans) tively. The increased cognitive control and analogical live can be viewed as cognitively more demanding social reasoning ability needed to cope with life in dispersed ‘market places’ than monkey groups. The demands of societies also suggests a possible route for human cog- navigating a more complex social landscape thus con- nitive . This hypothesis is supported by beha- stituted a unique selection pressure among the ancestral vioural and neurobiological data, but we need more of apes for increased and cognitive abilities. This both if we are to fully understand how our cou- ecologically-valid approach will, we hope, enable research- sins see the world. ers to uncover the cognitive differences that have, so far, been concealed. have unusually large brains for body size. It is now widely recognized that this is mainly associated with The social market place their distinctive social skills (the ‘Machiavellian intelli- To date, the social brain hypothesis has focused on a general notion of social complexity (often indexed by group gence’ [1] or ‘social brain’ hypothesis’ [2,3]). Within the size) as a driving force in brain evolution [3] and has not primate array, however, there are significant differences in attempted to explain how the particular characteristics of competence – most strikingly between monkeys and apes species’ social structure are related to cognitive capacity. – both in general cognitive abilities and, more importantly, However, recent empirical work on non-human primates in social behaviour and sociocognitive skills. These [12–14], along with theoretical speculations on humans contrasts seem to parallel differences between species in [15], suggests that applying a ‘biological markets’ [16] key aspects of brain evolution [4–6]. framework to studies of primate sociality might help point However, despite extensive psychological experimen- the way to more informative experimental studies of tation, the exact nature of the cognitive differences monkey and ape cognition. The theory of biological between monkeys and apes are unclear [7], and the degree markets views animals as traders engaged in a mutually to which apes are capable of any form of metarepresenta- beneficial exchange of commodities. Individuals within a tional (whether socially-based or not) remains biological market attempt to maximize ‘profit’ (in terms of contentious [8,9]. We suggest that this is because the tests fitness) by selecting social partners that offer the best have not been exhaustive and/or the right kinds of tests value, with an exchange rate set by the supply and demand have not been performed. The latter could be because of the commodity in question. much of the work on social cognition has been anthropo- Unlike standard optimality models of decision making, centric, seeking to discover which human cognitive market models are inherently dynamic, recognizing that characteristics are shared by our simian cousins [10]. individuals’ decisions vary across time according to local Thesameistruefromaneurobiologicalperspective, circumstance. They are perfect for the analysis of social where the use of monkeys as a model species has decision-making because the social world is also inher- meant, again, that similarities between non-human ently dynamic: each individual has its own particular set of primates and humans have been highlighted at the desires and goals that might not coincide with those of expense of distinctive differences. It is only recently others and decisions therefore reflect an ongoing process of that researchers have begun to probe for neurobiolo- negotiation. gical differences that might be linked to the apparently Studies of baboons (Papio hamadryas ursinus) [12–13] and chimpanzees (Pan troglodytes) [14] have shown that Corresponding author: Louise Barrett ([email protected]). animals trade grooming as a ‘service’, either for its own http://tics.trends.com 1364-6613/$ - see front matter q 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.tics.2003.09.005 Opinion TRENDS in Cognitive Sciences Vol.7 No.11 November 2003 495 utilitarian value or for its value equivalent; moreover, true market effects can be found, with the duration of grooming 1000 (the ‘price paid’) varying in relation to the supply of particular commodities [13]. Viewing primate groups as market places allows us to refine the social brain 100 hypothesis because we can identify the key differences between ape and monkey market places and thus predict consequent differences in cognitive skills. 10 Mean group size Mean group A market for brain power In primate market places, individuals must track the price 1 0.1 1 10 of commodities and respond flexibly to changes in supply Neocortex ratio and demand. This agrees with the ‘Machiavellian intelli- gence’ hypothesis, in that sociality is assumed to drive TRENDS in Cognitive Sciences brain evolution. It differs, however, by assuming that Figure 1. Evidence for grades within the primate brain suggests that there may be animals have not been selected to cope with increasingly quantitative, or even qualitative, differences between monkey and ape social cog- elaborate strategies and counter-strategies, the goals of nition. When a measure of social complexity (in this case, group size) is plotted which are to ‘outwit’ the competition. Instead, it assumes against a measure of relative brain size (in this case, the ratio of neocortex volume to the volume of the rest of the brain), it is apparent that hominoids (apes and that the evolution of brain size and structure have been humans, orange squares) require more neural capacity to support a given group driven by a need to track fluctuations in commodity value. size than monkeys (pink circles) do, who in turn require more than prosimians Monitoring the market place is intrinsically complex (blue triangles). Each datapoint is a single genus. Humans are represented by the top right-hand square. because the value of a particular partner is contingent on the value of others. Each of these values can shift with whole. By contrast, the nature of spider monkey fission– changes in reproductive state, health, dominance and fusion is not well studied and could differ from apes in ongoing social behaviour. Those who are good value today important ways. If, however, their behaviour is truly ape- might not be so tomorrow. This constant state of flux like, then we have an ideal test case: we would predict that means that keeping tabs on the social market is very they manifest ape-like cognitive abilities – something that different from the other kinds of contingent monitoring is, perhaps, hinted at by the fact that they are one of the that primates must do, such as tracking fluctuations in most encephalized of the South American monkeys. fruit availability [17]. Fruits, unlike conspecifics, do not In fission–fusion societies, individuals see each other make decisions of their own in response to primate only at infrequent intervals, often weeks apart, yet each behaviour. This inherent contingency in primate market recognizes the members of its community and is capable of places thus selects for a ‘knowledge-based’ [18], rather maintaining long-standing relationships. In such systems, than strictly cue-based, understanding of others, as individuals must be able to represent mentally individuals evidence from monkeys and, to an even greater degree, that are not present and to retain and manipulate apes has shown [18]. Nevertheless, there remains a information about them for substantial periods of time, cognitive difference between monkeys and apes that, whereas there is no such pressure for these abilities to although poorly characterized, is apparent when compar- evolve in spatially and temporally stable monkey groups ing their performance on psychological tests [7,19–20].It where animals are only out of view for at most a few hours. is also apparent that there are distinct grades within the The computational load required to monitor and track brain data that point specifically to increased cognitive changes in a dispersed ape market place is therefore powers among the apes [3,6,21] (Figure 1). We suggest that significantly greater than when members of a group see these emerged because of a key difference in the market each other every day [6]. A salient point to note in this places of these two taxonomic groups. context is that humans also live in dispersed societies Whereas monkeys are all highly gregarious and live in based on fission–fusion dynamics [25]. cohesive groups in which individuals encounter every member of their group every day, the apes – in particular, Monitoring space and time the chimpanzee and orang-utan (Pongo pygmaeus) – live More importantly, in terms of monitoring the state of their in more fragmented societies as a result of food compe- social market place, animals need to recognize that the tition, which forces females to forage in small parties or absence of a particular individual can affect the relative alone [22]. The apparent exceptions to this – group-living value of those that are present, and can reliably predict the gorillas (Gorilla gorilla) and fission–fusion spider mon- occurrence of certain behaviours in others in just the same keys (Ateles spp.) – are less convincing than might seem at way that the presence of that individual does. They also first sight. Among gorillas, group living might be a need to be able to track and update any changes observed relatively recent adaptation in response to infanticide by in the interactions of others when particular animals come adult males [23]. More importantly, our conception of into contact after a period of absence. Although all gorilla society has been determined entirely by a single anthropoid primates are capable of tracking ‘third-party (ecologically rather atypical) population (the Virunga relations’ (relationships between two other animals with- mountain gorillas); a fission–fusion pattern of sociality out reference to self) [7], the ability to generate a causal has been reported from western lowland gorilla popu- understanding of such behavioural interactions in the lations [24] and may in fact be more typical of gorillas as a absence of certain individuals would seem to be much more http://tics.trends.com 496 Opinion TRENDS in Cognitive Sciences Vol.7 No.11 November 2003 demanding cognitively. Thus, the key to social survival in Implications for primate cognition dispersed societies is the ability to work with a social world Having identified the likely ecological pressure respon- that is partially virtual, rather than purely physically sible for generating a more complex social environment for instantiated. The fact that chimpanzees are apparently apes compared with monkeys, more precise predictions able to represent the relative spatial locations of stone about how cognitive abilities might differ between these tools (hammer stones and anvils) and to use this groups can be formulated. We suggest that four cognitive information in a flexible manner [26] supports the notion mechanisms are crucial to the development of high-level for a similar capacity in the social domain. Intriguingly, sociocognitive abilities. These are: (i) the ability to reason the finding that scrub jays also show such abilities [27] causally, (ii) the ability to reason analogically, (iii) the suggests that an episodic-like (the ability to ability to exert cognitive control over neural pathways to represent ‘what, where, when and who’) is a precursor to generate and assess alternative options, and (iv) the more highly developed aspects of cognition, rather than ability to formulate these options into alternative the outcome of these processes. sequences of future actions and to select between them. As abundant research has demonstrated, humans are Analogical abilities adept at all four of these processes. Apes are known to Boroditsky has argued that, in humans, the sense of time show higher performance than monkeys in both causal emerges via a metaphorical analogy from a sense of space [35–36] and analogical reasoning tasks [37], but there [28]. This suggestion provides the basis for a promising have been, as yet, few attempts to test for differences in the explanation for cognitive evolution among the hominoids: latter two abilities. Anecdotal evidence suggests that a sense of space could, with sufficient additional cognitive monkeys show extreme ‘myopia for the future’ [38] and control, be used to develop an extended sense of time, thus we predict that, compared with apes, monkeys will show enabling animals to predict future states of the market no evidence of generating ‘contingency plans’ for future place, as well as track current changes, in a very effective events. way. If the ability to engage in analogical reasoning scaffolded this development, as Boroditsky suggests [28], then a plausible suggestion would be that, combined with Box 1. A neurological basis for social cognition the ability to reflect on a virtual world, an understanding of future time would have led to the emergence of an ability In humans, many of our more advanced cognitive skills are to sequence social events temporally into causal chains, associated with improved executive function [7] and the expansion of the frontal lobe [39], particularly the lateral pre-frontal area. Miller allowing animals to appreciate that behaviour in the and Cohen [40] have argued that this is related to the need for greater present can have significant consequences beyond the ‘cognitive control’ – the ability to take charge of one’s actions and immediate future. Although others have suggested that direct them towards future, unseen goals. Their model thus fits access to this fourth dimension, time, is the key to beautifully with our ideas concerning the selection pressures acting on hominoid brains. They argue that the PFC effectively constructs a understanding the extraordinary evolutionary trajectory map of the neural pathways needed to solve a task, thereby allowing of humans [29–30], we can add an important new an organism to escape reality and gain greater control of its actions dimension by identifying the likely ecological context in by forming abstract representations that are not linked to specific which such an ability was selected. associations, but which correspond to overarching principles and concepts. As both the frontal lobes [39] and the Such a stance can perhaps give some insight into how [41,42] of the macaque (Macaca sp.) are significantly smaller than apes and humans differ from each other as well. Given those of humans and apes (in the case of prefrontal cortex, 11% of further increases in brain size within the hominid line total cortical volume in macaque, compared with 17% and 36% for (whether socially-driven or not), an initial ability to chimpanzees and humans, respectively [41]), it is likely that monkeys represent the self and others in different spatio-temporal will be more limited in this respect than both apes and humans (and by the same token, that apes will be more limited than humans) [6]. locations could have scaffolded the development of higher However, the PFC is not the only element crucial for achieving metacognitive skills, leading eventually to full-blown cognitive control. Allocation of control might also depend on the ‘theory of ’ skills in modern humans [31]. Given anterior cingulate cortex (ACC), an area variously associated with this, we prefer to view ToM (and associated levels of error detection, response selection and conflict monitoring [43–45]. By detecting conflict signals, the ACC is able to signal that additional advanced social cognition) not as a phenomenon in itself, control needs to be allocated to a task. It has also been suggested that but as a specialized, socially-oriented form of analogical the upgrading of the ACC would have enabled the construction of a reasoning that arises as an emergent property of powerful ‘troop in the head’ because it is also linked to generating a sense of executive functioning (assisted, no doubt, by the represen- self in relation to others [46]. In this respect, it is intriguing that spindle cells, a class of large projection neurons found principally in tational properties of language). The core difference the ACC region, are found only in apes and humans, not monkeys between humans and other apes might, then, simply [47]. Allman et al. [48] have speculated that these are involved in reflect the number of alternative future scenarios that can coordinating widely distributed neural activity involving emotion be managed and compared simultaneously while keeping and cognition, again fitting well with our speculations on the need for the overall goal ‘in mind’ (a facility associated with the greater attention-to-action and cognitive control within a dispersed ape market place. The role of cerebellar–neocortical circuits in lateral pre-frontal cortex [32] and also with Brodman’s coordinating cognitive functioning might also be crucial in this area 32 [33] – intriguingly, the latter has been identified respect [49]; recent work has identified an expansion of the by fMRI studies as an area of unique activation during the cerebellum across the primate order, suggesting that evolution has execution of ToM tasks [34]) and the time depth to which acted on distributed systems of functionally related components in just this way [50]. these can be projected. http://tics.trends.com Opinion TRENDS in Cognitive Sciences Vol.7 No.11 November 2003 497

The significant experimental evidence in support of its 21 Kudo, H. and Dunbar, R.I.M. (2001) Neocortex size and social network basic elements [35–37] gives us some confidence in size in primates. Anim. Behav. 62, 711–722 developing our hypothesis that the difference between 22 Wrangham, R.W. (1980) On the evolution of ape social systems. Soc. Sci. Inform. 18, 335–368 monkeys and apes stems from the time frame over which 23 Harcourt, A.H. and Greenberg, H. (2001) Do gorilla females join males they can monitor their social market places. However, to avoid infanticide? Anim. Behav. 62, 905–915 owing to the gap in our knowledge of the neurobiological 24 Parnell, R.J. (2002) Group size and structure in western lowland differences between monkeys and apes, we can be less sure gorillas (Gorilla gorilla gorilla) at Mbeli Bai, Republic of Congo. Am. about how these cognitive differences map onto brain J. Primatol. 56, 193–206 25 Dunbar, R.I.M. (1993) Coevolution of neocortex size, group size and structure and function. Although the functions of those language in humans. Behav. Brain Sci. 16, 681–735 brain areas that show expansion across the primates 26 Boesch, C. and Boesch, H. (1984) Mental map in wild chimpanzees: an support our proposal, most of these data are from humans analysis of hammer transports for nut cracking. Primates 25, 160–170 (Box 1). However, we hope that a new perspective will 27 Clayton, N.S. and Dickinson, A.D. (1998) What, where and when? encourage both cognitive psychologists and neurobiol- Evidence for episodic-like memory during cache recovery in scrub jays. ogists to probe more deeply into how monkey and ape Nature 395, 272–274 28 Boroditsky, L. (2000) Metaphoric structuring: understanding time brains differ from each other, both structurally and through spatial metaphors. Cognition 75, 1–28 functionally. 29 Suddendorf, T. and Corballis, M.C. (1997) Mental time travel and the evolution of the human mind. Genet. Soc. Gen. Psychol. Monogr. 123, Acknowledgements 133–167 We are grateful to Filippo Aureli and Rob Barton for very helpful 30 Tulving, E. (2002) Episodic memory: from mind to brain. Annu. Rev. discussion of these ideas and to the anonymous referees for their Psychol. 53, 1–25 comments on an earlier draft. 31 Dunbar, R.I.M. The social brain: mind, language and society in evolutionary perspective. Annu. Rev. Anthrop (in press) References 32 Krawczyk, D.C. (2002) Contributions of the pre-frontal cortex to the neural basis of human decision-making. Neurosci. Biobehav. Rev. 26, 1 Byrne, R.W., Whiten, A. eds (1988) Machiavellian Intelligence: Social Expertise and the Evolution of Intellect in Monkeys, Apes, and Humans 631–664 Oxford University Press 33 Koechlin, E. et al. (2000) Dissociating the role of medial and lateral 2 Barton, R.A. and Dunbar, R.I.M. (1997) Evolution of the social brain. In anterior prefrontal cortex in human planning. Proc. Natl. Acad. Sci. Machiavellian Intelligence II (Whiten, A. and Byrne, R.W., eds), U. S. A. 97, 7651–7656 pp. 240–263, Cambridge University Press 34 Frith, U. and Frith, C.D. (2003) Development and neurophysiology of 3 Dunbar, R.I.M. (1998) The social brain hypothesis. Evol. Anth. 6, mentalizing. Philos. Trans. R. Soc. Lond. B Biol. Sci. 358, 459–473 178–190 35 Limongelli, L. et al. (1995) Comprehension of cause-effect relationships 4 Rapoport, (1999) How did the evolve? Brain Res. Bull. 50, in a tool-using task by chimpanzees. J. Comp. Psychol. 109, 18–26 149–165 36 Visalberghi, E. and Limongelli, L. (1994) Lack of comprehension of 5 Enard, W. et al. (2002) Intra and interspecific variation in primate gene cause-effect relations in tool using capuchin monkeys. J. Comp. expression patterns. Science 296, 340–343 Psychol. 108, 15–22 6 Dunbar, R.I.M. (2003) Why are apes so smart? In Primate Life 37 Thompson, R.K.R. and Oden, D.L. (2000) Categorical perception and Histories (Kappeler, P. and Peirera, M., eds), pp. 285–298, MIT Press conceptual judgments by non-human primates: the paleological 7 Tomasello, M. and Call, J. (1997) Primate Social Cognition, Oxford monkey and the analogical ape. Cogn. Sci. 24, 363–396 University Press 38 Roberts, W.A. (2002) Are animals stuck in time? Psychol. Bull. 128, 8 Povinelli, D.J. and Vonk, J. (2003) Chimpanzee : suspiciously 473–489 human? Trends Cogn. Sci. 7, 157–160 39 Semendeferi, K. et al. (1997) The evolution of the frontal lobes: a 9 Tomasello, M. et al. (2003) Chimpanzees understand psychological volumetric analysis based on three-dimensional reconstructions of states – the questions is which ones and to what extent. Trends Cogn. magnetic resonance scans of human and ape brains. J. Hum. Evol. 32, Sci. 7, 153–156 375–388 10 Povinelli, D. et al. (2000) Toward a science of other minds: escaping the 40 Miller, E.K. and Cohen, J.D. (2001) An integrative theory of pre-frontal argument by analogy. Cogn. Sci. 24, 509–541 cortex function. Annu. Rev. Neurosci. 24, 167–202 11 Preuss, T. (2000) What’s human about the human brain? In The 41 Fuster, J. (1989) The Prefrontal Cortex, Raven Press New Cognitive , 2nd edn, (Gazzaniga, M., ed.), 42 Passingham, R. (1993) The Frontal Lobes and Voluntary Action, pp. 1219–1234, MIT Press Oxford University Press 12 Barrett, L. et al. (1999) Market forces predict grooming reciprocity in 43 Botvinick, M. et al. (1999) Conflict monitoring versus selection-for- female baboons. Proc. R. Soc. Lond. B. Biol. Sci. 266, 665–670 action in ACC. Nature 402, 179–181 13 Henzi, S.-P. and Barrett, L. (2002) Infants as a commodity in a baboon 44 Carter, C.S. et al. (1998) ACC, error detection and on-line monitoring of market. Anim. Behav. 63, 915–921 performance. Science 280, 747–749 14 de Waal, F.B.M. (1997) The chimpanzee service economy: food for 45 Bush, G. et al. (2000) Cognitive and emotional influences in anterior grooming. Evol. Hum. Behav. 18, 375–386 cingulate cortex. Trends Cogn. Sci. 4, 215–223 15 La Cerra, P. and Bingham, R. (1998) The adaptive nature of the human 46 Skoyles, J.R. and Sagan, D. (2002) Up from Dragons: the Evolution of neurocognitive architecture: an alternative model. Proc. Natl. Acad. Human Intelligence, McGraw-Hill Sci. U. S. A. 95, 11290–11294 47 Nimchinsky, E.A. et al. (1999) A neuronal morphologic type unique to 16 Noe¨,R.et al. (2001) Economics in Nature: Social Dilemmas, Mate humans and great apes. Proc. Natl. Acad. Sci. U. S. A. 96, 5268–5273 Choice and Biological Markets, Cambridge University Press 48 Allman, J. et al. (2001) The anterior cingulate cortex: the evolution of 17 Milton, K. (1988) Foraging behaviour and the evolution of primate an interface between cognition and emotion. Ann. N. Y. Acad. Sci. 935, intelligence. In Machiavellian Intelligence (Byrne, R. and Whiten, A., 107–117 eds), pp. 285–305, Oxford University Press 18 Call, J. (2001) Chimpanzee social cognition. Trends Cogn. Sci. 5, 49 Schmahmann, J.D. (1998) Dysmetria of thought: clinical consequences 388–393 of cerebellar dysfunction on cognition and affect. Trends Cogn. Sci. 2, 19 Hare, B. et al. (2001) Do chimpanzees know what conspecifics know? 362–371 Anim. Behav. 61, 139–151 50 Whiting, B.A. and Barton, R.A. (2003) The evolution of the cortico- 20 Hare, B. et al. (2003) Do capuchin monkeys, Cebus capella, know what cerebellar complex in primates: anatomical connections predict conspecifics do and do not see? Anim. Behav. 65, 131–142 patterns of correlated evolution. J. Hum. Evol. 44, 3–10 http://tics.trends.com