We wholly endorse Penn et al.'s unsentimental critique of the current zeitgeist in comparative cognition. One area they gloss over in their review is the natural communication systems of primates and other nonhumans. This omission is unfortunate because, following Don Griffin's spirited prescriptions (e.g., Griffin Reference Griffin1978), studies of animal communication in particular have been a key driver of the field's blinkered focus on only continuity between animal and human minds – the focus that Penn et al. are now at pains to exorcise. A paradigmatic example is research on vervet monkey communication by Seyfarth et al. (Reference Seyfarth, Cheney and Marler1980), which documented a small repertoire of alarm calls specific to different kinds of predator that prompt functionally different escape responses. The apparent referential quality of these alarm calls encouraged comparison to the symbolic properties of human words and provided a timely confirmation of Griffin's instinct that communication behaviors offered privileged insight into animal minds and that continuity with human minds would be revealed in proportion to their language-like qualities. The vervet monkey work thus catalyzed a generation of research aimed at identifying additional language-like phenomena in nonhuman communications. It also inspired research on continuity in other cognitive domains (e.g., transitivity, numerosity, causality, etc.) on the assumption that the manifest continuity between animals and humans in our own most vaunted ability – language – must foreshadow continuity in almost everything else; hence, the inference that some (perhaps many) nonhumans might be teetering on the brink of humanity.
Subsequent research on intentionality (qua theory of mind, ToM) seriously undercut these inferences. Additional field studies of communication showed that, contra language, primate vocalizations were not produced flexibly to contact and manipulate the intentional states of others. Indeed, signalers proved remarkably oblivious to the audience to which their vocalizations were “addressed.” Laboratory studies of ToM pointed to the same social-cognitive lacuna. Results of ape language studies were corroborative: despite sometimes impressive symbol learning abilities, apes' use of artificial languages proved almost entirely instrumental and solipsistic. Because human language and its sine qua non – meaning – hinges on the intentionality of communications (mutual, implicit mental attributions; Grice Reference Grice1957), it is increasingly difficult to see primate (and other animal) communications as very much like language in this fundamental respect (Notman & Rendall Reference Notman and Rendall2005; Owren & Rendall Reference Owren and Rendall2001). Thus, as Cheney and Seyfarth themselves now argue, although the functional behavioral outcomes that animal signals support might sometimes seem superficially language-like, the underlying psychological mechanisms are profoundly different (Cheney & Seyfarth Reference Cheney and Seyfarth2005).
The upshot here for Penn et al. is that, in their efforts to correct the increasingly blinkered view that threatens comparative cognition, they have missed an opportunity to do so using results from the very research programs on communication that inspired that blinkering to begin with.
At the same time, however, we ask whether the profound functional gulf between humans and nonhumans in these and other domains need be underwritten by the fundamentally different psychologies that Penn et al. trace to our ability for “relational reinterpretation.” True, the functional limitations in nonhuman communication and intentionality that we emphasize might well reflect a common lacuna in relational reinterpretration: truly symbolic communication requires understanding not just the (observable and local) referential connection between particular symbols and their real-world referents, but also the broader (but unobservable) relational system that they instantiate; similarly, ToM requires seeing through specific (observable and local) behavior-context associations to the broader (but unobservable) relational mentalistic system that they reflect.
However, it is not clear that either ability requires any qualitatively new representational process. For example, Landauer has shown that, starting only with a large corpus of (experienced) words and the discursive contexts in which they occur, and no other pre-existing knowledge, singular value decomposition followed by dimension reduction can distill from their local (first-order) associations the latent relational dimensions (i.e., second-order associations) among them that yield functional semantic comprehension (latent semantic analysis; e.g., Landauer & Dumais Reference Landauer and Dumais1997). In principle, with sufficient and appropriate dimensions and observations, similar associative processes could produce the latent relational dimensions among individuals, behaviors, and contexts that would give the functional appearance of a theory of mind.
In other words, is it possible that relational reinterpretation is not a distinct and qualitatively novel, abstract representational process, but rather, a theoretical approximation for the under-appreciated (and still insufficiently explored) returns of (taxonomically widespread) associative processing? And is it therefore possible, at least in quotidian contexts, that humans too might have a mostly behavioristic understanding of the world that is driven primarily by perceptions of concretes, not by hypothesized unobservables, no matter how much our scientific theorizing spirits heuristic mentalistic constructs into our heads: that what is frequently seen to be necessarily analytic is in reality the result of nonanalytic cognition (Brooks Reference Brooks, Rosch and Lloyd1978)?
There are two obvious objections. The first is that, compared to the representational account, the associative/nonanalytic account would be far more memory intensive, computationally extensive, and therefore processually inefficient. To which we respond, “Just exactly what defines the human brain and what our theories of comparative cognition need to account for: a tissue profligate, massively interconnected, parallel processor.” A second objection is that the associative/nonanalytic account does not actually produce semantic knowledge or a mentalistic understanding of others, only the functional appearance of them. To which we respond, “And what else is there?”
Ultimately, then, we completely agree with Penn et al. that the current zeitgeist in comparative cognition is wrong; however, the mistake maybe lies not in emphasizing mental continuity, but rather in the kind of mental continuity emphasized. Animals and humans are probably similar: however, similar not because animals are regularly doing cognitively sophisticated things, but because humans are probably doing cognitively rather mundane things more often than we think. This assertion does not mean that there are not qualitative phenomenological differences between humans and animals. There are, and Penn et al. rightly foreground these in their opening statements. But, these phenomenological novelties need not be underwritten by a qualitatively different psychology. Rather these functional novelties might emerge naturally from humans' exaggerated tendency to put things in the world (sensu Clark Reference Clark1997) and then to re-perceive them again, repeatedly. And then also to crunch the vast perceptual datasets created from this habit with exponentially more associative power. Paradoxically, and contra part of Penn et al.'s proposal, these habits might also make us more, rather than less, perceptually-driven than nonhumans.
