The fundamental premise for generativist linguists is that the diversity of languages springs from a common cognitive structure, latent in all humans, called Universal Grammar (UG). The assumption of this core cognitive construct, from which language emerges and by which language structures are constrained, is closely tied to assumptions that the structural framework of language is innate and unlearnable, that the processing capacities necessary for language are unique to humans, and that language is inherently a cognitively isolated module and therefore a computationally unique aspect of human cognition. This black box approach to a fundamental question in cognitive science ignores the potential of an evolutionary approach to cognitive processing, including such an approach to language. It also trains linguists to avoid consideration of evolutionary biology, and animal research in general. Evans & Levinson (E&L) address the first limitation head on but fall prey to the second. The good news is that it is a modest and easy extension of their work to complete the revolution.
By shifting the terms of scientific discourse from the unsupported and untestable assumption of a common core set of linguistic universals, a sort of mystical construction, to the actual distinctive behaviors of language populations, E&L move the study of language in the direction of the research methods of the experimental sciences and away from enclosed personal belief systems. The idea of studying real behavior as opposed to a platonic idealization also brings with it the obligation to consider real mechanisms that can explain that behavior. This calls into question the longstanding barrier of the competence/performance distinction to empirical study of language use and processing. This also brings with it a different standard for language research and what should count as evidence in understanding language processes.
This is particularly important in considering the issue of the human uniqueness of language. Hauser et al. (Reference Hauser, Chomsky and Fitch2002) have claimed that the single property of recursion as applied to vocalizations is the uniquely human computational capacity of the faculty of language. To address this point, E&L discuss the definition of recursion at length, from recursively enumerable patterns to nested recursion of patterns, and they allude to the computational differences among these. However, it is important to have clarity regarding what capacity is being described. If recursive syntactic processing in humans is a concept tractable to experimental science, then recursion refers not to the concept of putatively infinite (or, in any case, exceedingly long) patterns that no human produces or ever produced, as best as we know, but to an aspect of the computational complexity of pattern processes in actual human vocalizations. The real functional limits on what humans do produce and can perceive matter, and these should not be disregarded under the rubric of performance limitations independent of linguistic competence. Research with animals is irrefutably essential to address any claims of human uniqueness. In psychology and biology a single counterexample may not destroy a theory as it does in physics, but a whole class or classes of counterexamples can do so.
E&L refer to a study by Fitch and Hauser (Reference Fitch and Hauser2004) that reports the “impressive learning powers over FSGs [finite state grammars]” displayed by cotton-top tamarins, in spite of their failure to “grasp the patterning in PSG-generated sequences” (i.e., phrase-structure grammar generated sequences) (see target article, sect. 6, para. 3). Here E&L fall prey to becoming the “outsiders … [heeding] … the articulate envoys from the universalising generativist” (sect. 1, para. 4). E&L accept a null (PSG) result reported in tamarin study with species-irrelevant signals (human speech syllables), no clear learning procedure (mere exposure to stimuli), and with possible biases by additional cues (male/female voicing distinctions available for the FSG stimuli). Indeed, it might not have even been clear to the tamarins that there was a task since they were not behaviorally shaped to make a particular response.
The experimental paradigm that Fitch and Hauser (Reference Fitch and Hauser2004) used to distinguish tamarin failure of perception at the PSG level became broadly controversial only after it was used to demonstrate that operantly trained European starlings are able to learn both the FSG and PSG sequences (Gentner et al. Reference Gentner, Fenn, Margoliash and Nusbaum2006). Claims and counterclaims surrounding these results have appeared elsewhere and need not be repeated here. Nevertheless, one clear distinction of the starling results stands out in comparison to the tamarin studies. Whether one accepts or rejects the claims regarding recursive processing in assessing the learning of the PSG sequences, the starlings displayed far more complex vocal processing behavior than was displayed by tamarins. Our point is that animals that engage in vocal sensorimotor learning (starlings are songbirds and are excellent mimics) also demonstrate heightened vocal perceptual skills. Understanding the functional relationship between perceptual and motor skills, and the functional utility of such abilities in the daily life and social needs of an organism – including humans – is critical to assessing biological and computational capacities. This fact alone importantly informs the program of E&L, but about it they are mute.
E&L appropriately reach towards cladistic methods developed by evolutionary biologists to help sort out relations among human languages. They do not reach far enough. There are also examples of rapid changes in learned vocal patterns within nonhuman species based on cultural transmission of vocal patterns – famously studied in the songbirds (e.g., Marler & Tamura Reference Marler and Tamura1964). These studies have stimulated assessment of different learning models and extensive analysis of learning mechanisms. Applying this knowledge must be part of E&L's program, but they failed to recognize it as such.
As long as linguistics will be tied up in generativist nativism, animal studies won't matter and will contribute little. If linguistics truly becomes an experimental science, then animal studies are informative, and a synthesis of animal and human work becomes a useful goal to try to conceive.
E&L have staked out an important position: that there is an ethology of language, that it is available for study, and that it is essential that models of language use and mechanisms of language be inexorably related to actual language usage across humans. We agree, and maintain that this program is fundamentally rooted in experimental science and evolutionary biology. As such, it can be deeply informed by biological work on animal communication systems. Linguistics, and psychology, are subsets of organismal biology.
