World Without Weight: Perspectives on an Alien Mind . Oxford University Press , U. K. , 2012 , 353 pp .,
Can chimpanzees understand the concept of weight? Based on data from over 30 previously unpublished experiments, Daniel Povinelli et al.'s answer in World Without Weight is “no.” The project is offered as the “final test” of their Relational Reinterpretation Hypothesis (RRH), which states that humans are alone in their ability to understand the world in terms of unobservable causes. Nonhuman animals can only learn about the world in terms of its first-order observable similarities and contingencies. If our nearest living relative cannot grasp the fundamental concept of weight, then other animals are unlikely to master any other theoretical relationships.
Povinelli practices a distinctive methodology in comparative psychology, attempting to sharply demarcate the limits of animal cognitive abilities by testing them on a series of increasingly difficult tasks and interpreting failure as the lack of some critical cognitive competency. For example, although chimpanzees have been said to select stone tools for particular tasks on the basis of weight (Sakura and Matsuzawa 1991; Schrauf and Call 2011), Povinelli dismisses such results as due to sensory-motor representations of effort-to-lift rather than a theory of weight as a hidden force, because chimpanzees in his experiments had difficulty planning where to sort objects by weight before hefting them. He similarly challenges inferences drawn from other studies (e.g., Hanus & Call 2008), often with additional data from more diagnostic (but also inferentially more complex) experiments.
Povinelli's approach has often be the subject of criticism (e.g., Allen, 2002; Whiten, 2001), with concerns centered on the repeated use of the small “Project Megan” troop (see pp. 62–64) at the New Iberia Research Center, because chimpanzees from other groups have passed some of the same tests (Bulloch, Boysen, and Furlong 2008; Kaminski, Call, and Tomasello 2004) that Povinelli's animals failed (Povinelli and Eddy 1996; Reaux, Theall, and Povinelli 1999). Where in earlier work Povinelli focused on demonstrating that many positive results in animal cognition could more parsimoniously be explained in terms of classical or instrumental conditioning, he now acknowledges that animal performance across a range of tasks demonstrates remarkable cognitive flexibility. He now ties his argument for human cognitive uniqueness to Newell's Physical Symbol System hypothesis (1980), holding that only humans form abstract, second-order representations of underlying causes, which allows them to transfer knowledge from one task to perceptually dissimilar ones. Do chimpanzees have such an understanding of weight?
Povinelli et al. (Experiments 1–8) first explore what chimpanzees can do, such as sort objects by the effort required to lift them, or calibrate lift forces to object sizes before actually hefting them. They argue that sensory-motor representations of weight account for this type of problem solving. They then turn to evidence for what chimpanzees cannot do (Experiments 9–29). They did not exhibit evidence of transfer of knowledge about object weight from one context to another, whereas 5-year-old human children succeed on several of the same tasks (Experiments 30–32).
In their recent work, Povinelli et al. attempt to move debate from “cognition vs. association” to “relational reinterpretation vs. everything else.” However, they fail to adequately justify their own exacting standard for a “genuine” understanding of weight, namely that it be fully abstract, nonperceptual, and domain general. Even adult humans only approximate this ideal, as Povinelli himself concedes (Penn et al. 2008). If we reject this standard, Povinelli's experiments offer a complex and multifaceted picture of chimpanzee's understanding of weight: they can sort objects by effort-to-lift, program lift forces before lifting, learn to sort objects by size and color cues correlated with weight, and reason about concomitants of weight such as deformation, support, and balance. Chimpanzees indeed fail on many other weight-relevant tasks, but it may be more interesting to explore their reasons for error than to describe their failures in terms of an absence of a domain-general ability that even humans do not fully possess (Seed et al. 2012). Indeed, the greatest contribution of WwW is its detailed description of the patterns of learning and error observed in troop Megan across a wide variety of tasks. The patterns remain puzzling. Why do some chimpanzees master a sorting task so much faster than others? Why is initial success often followed by a decline in performance, before a gradual increase back up to competency?
Looking to the future, Povinelli calls for the comparative exploration of a “‘causal workspace’ within which various weight-related representations can do their work” (p. 292). The goal of such a “causal workspace theory” would be to explain why certain species and individuals find some problems easier or harder than others. By developing such models of cognition, we can ask evolutionarily relevant questions about the cost/benefit tradeoffs offered by various representational strategies in different environments. Especially interesting to this project would be to compare the types and patterns of errors made by humans and animals instead of searching for tasks that humans can solve but animals cannot. Although comparative psychology is still populating its stock of phenomena to be explained, Povinelli et al. set out a goal at which the field should aim.