6 • Great ape cognitive systems ANNE E. RUSSON Psychology Department, Glendon College of York University, Toronto INTRODUCTION CONCEPTS AND MODELS OF COGNITION This chapter considers cognition in great apes as inte- grated systems that orchestrate the many abilities that Situating great ape cognition comparatively hinges on great apes express, systems for which satisfactory char- mental processes that support symbolism, notably rep- acterizations remain elusive. In part, difficulties owe to resentation, metarepresentation, and hierarchization. research trends. Empirical studies have been guided by Weaker and stronger conceptualizations exist for each diverse and sometimes contradictory models, questions, and which is used affects assessments of great apes’ capa- measures, tasks, and living conditions. Performance bilities. levels have proven inconsistent across individuals, rear- Weak meanings of symbolism include reference by ing conditions, and testing conditions, and evidence is arbitrary convention (Peirce 1932/1960), using inter- patchy across species for virtually any facet of cogni- nal signs like mental images to stand for referents tion. Evidence on wild great apes, the most important rather than using direct sensations or motor actions, from an evolutionary perspective, is especially patchy and solving problems mentally versus experientially. In because research has favored captives; much of what the strong sense, symbolism refers to self-referring sys- is available was collected for other purposes, so it was tems wherein phenomena owe their significance and neither described nor analyzed with cognition in mind. even existence to other symbols in the system rather The issues at stake are also hard-felt ones that touch on than to sensorimotor entities (e.g., Deacon 1997; Donald the human–nonhuman boundary, so entrenched beliefs 2000; Langer 2000). Representation can refer to any infect how the literature is interpreted and even what of form of mental coding that stands for entities, perceptual it is read. included (Perner 1991; Whiten 2000) or, more strongly, Attempts have none the less been madetodevelop to recalling to mind or “re-presenting” mental codes an integrated model of great ape cognition using avail- forentities and simple object relations in the absence able evidence. They include both edited survey vol- of normal sensorimotor cues (P&M; Whiten 2000). umes (Matsuzawa 2001a; Parker, Mitchell & Miles 1999; Meanings of metarepresentation range from represent- Russon, Bard&Parker1996)andintegrative reviews, ing other representations (e.g., Leslie 1987; Matsuzawa three of the latter as major books (Byrne 1995 (RWB), 1991; Whiten & Byrne 1991) to representing a represen- Parker & McKinney 1999 (P&M); Tomasello & Call tation as a representation, i.e., an interpretation of a sit- 1997 (T&C)) and others as articles (e.g., Byrne 1997; uation (Perner 1991). Meanings of hierarchization span Suddendorf & Whiten 2001; Thompson & Oden 2000; creating new, higher-order cognitive structures from Whiten & Byrne 1991). My aim is not to analyze this lower-level ones (i.e., structures with superordinate– terrain, yet again, in detail, but to offer a compact mise subordinate features: Byrne & Byrne 1991; Case 1985; adate` to ground evolutionary reconstruction. Guiding Langer 1998) to generating cognitive structures that questions are “what, if anything, about great ape cog- show embedding (e.g., classification showing nesting of nition requires evolutionary explanations beyond those classes: Langer 1998). developed for other nonhuman primates?”, and “how is Developmentalists commonly consider weak and great ape cognition best characterized with respect to strong forms to be related in humans, as basic and evolutionary questions?” advanced ontogenetic achievements of early and later Copyright Anne Russon and David Begun 2004. 76 Great ape cognitive systems 77 childhood respectively (Table 6.1). Comparative pri- problem-specific structures – effectively eliminating mate cognition often shares this view (P&M; Whiten chances for detecting use of multiple or general purpose 2000). Insofar as symbols must be grounded in real processes. Given the lack of relevant empirical evidence, world referents at some point (Donald 1991) and weak these models remain speculative concerning great apes. symbolism is the more likely in great apes, I consider Cognitive science models portray the mind as a great ape cognition relative to weak symbolism and its device for processing, storing, integrating, and trans- associated processes (strong representation, weak hier- forming information. Some of their concepts have been archization, weak metarepresentation). The terms sym- incorporated into models of cognitive development (e.g., bolic, representation, hierarchization, and metarepre- Case 1985; Leslie 1987; Pascual-Leone 1987), others sentation henceforth refer to these meanings. have aided in detecting hierarchization in great ape cog- The models guiding empirical studies of great ape nition (Byrne & Byrne 1991; Byrne, Corp & Byrne 2001; cognition also contribute to disparities because of the Byrne & Russon 1998; P&M; Russon 1998). Limita- ways in which they shape the generation of evidence tions concern portraying cognition in static, mechanistic and the interpretive frameworks they impose. Several terms that may not apply to living beings. important models are sketched below to suggest their Models of human cognitive development have strengths and limitations for understanding great ape proven valuable for assessing primate cognition com- cognition. paratively because they provide conceptual and method- Animal models designed for nonhuman mentality ological tools for assessing non-symbolic and symbolic have been frequent frameworks for studies of great ape cognition within one unified framework and the gen- cognition. They concentrate on the non-symbolic, asso- eration and structure of cognition. Piaget’s model sup- ciative processes presumed to govern nonhuman cog- ported the first developmental studies of great ape cog- nition, for example trial-and-error experiential learn- nition; among its greatest contributions is its portrayal of ing or behavior chains. This leaves them conceptually cognition as constructed progressively during ontogeny and methodologically impoverished concerning sym- and directly affected by interaction with the environ- bolic cognition (Anderson 1996; RWB;Rumbaugh 1970; ment. Early piagetian studies focused on sensorimotor T&C), quantification and logic being important excep- (human infant) cognition, which relies on pre-symbolic tions (e.g., Boysen & Hallberg 2000; Thompson & Oden processes similar to those portrayed in animal models, so 2000), so relatively little of the evidence they have gener- similar limits apply.Recently,neo-piagetian models have ated helps determine whether great apes, or any species, supported studies of the rudimentary symbolic range are capable of symbolic cognition. (for an overview, see P&M). Models inspired by Vygot- Generality–modularity models are potentially sky, which portray socio-cultural forces like apprentice- important because they concern cognitive architecture. ship or enculturation as fundamental to cognitive devel- In this view, favored by evolutionary psychologists and opment (e.g., Donald 2000; T&C), have spawned many neo-nativists, there exist cognitive “modules,” problem- studies on social cognition and cognitive development specific cognitive structures that represent innately in great apes. Given how richly primate lives are socially specified neurological systems and operate with rela- embedded, their merits are obvious. Among these mod- tive autonomy, as well as general purpose or central els may be included models of understanding others’ processes that apply across problem types and affect mental states, or theory of mind, which some propose to system-wide properties (e.g., representation, executive underpin much cognitive progress in early human child- control structures, working memory). These models hood (e.g., Carruthers & Smith 1996). Two such models have influenced understandings of great ape cognition have been applied to great apes, both proposing that with their assumption that modular architecture char- general-purpose cognitive processes in the rudimen- acterizes nonhuman cognition and generality evolved tary symbolic range underwrite this progress (second- uniquely in humans (e.g., Mithen 1996; T&C; Tooby order representation – Leslie 1987; secondary represen- &Cosmides 1992). Little if any empirical study has tation – Perner 1991). While both offer useful tools for examined cognitive architecture in great apes, however. assessing rudimentary symbolic processes, their focus Studies of great ape cognition have typically assumed on one ability series in the social domain risks under- modularity and have aimed for clean tests of individual representing the breadth of great apes’ achievements. Table6.1. Processes and structures posited in rudimentary sybolic level cognition, in human developmental perspective Age Piaget Case1 Langer2 Leslie3 Perner4 (yr) (general) (causality) (logic-math) (Theory of Mind) (Theory of Mind) 0 Sensorimotor stage Sensorimotor stage 1st-order cognition 1st-order representation Primary representation Schemata Operational consolidation 1.5 (Inter-) relational cognition 2nd-order cognition 2nd-order representation Secondary representation 2 unifocal relational operations metarepresentation re-represent primary stage 6, symbols, representations