Alan Turing began his essay “Computing Machinery and Intelligence” by posing the question, “Can machines think?”1 He quickly abandoned this formulation in favor of a more clearly defined problem. In their common usage, Turing observed, the terms machine and think are too ambiguous to form the basis for any serious consideration. “The original question,” he later remarked, is “too meaningless to deserve discussion.”2
Robert Berwick and Noam Chomsky’s Why Only Us seeks to address an entirely different problem, the evolution of language. Nonetheless, the question at the heart of their book is problematic in similar ways. “Who has language?” Or equivalently, “Are humans the only ones who possess language?” As Turing demonstrated, and shall be argued here, when confronted with difficulties of this nature, a fresh approach to the problem is needed.
The Book and the Field
A new book by Chomsky is always a fresh opportunity to reflect on some of the deepest and most difficult themes touching on human nature. Even if one is not moved by the particular direction favored, there can be little doubt that such a direction can serve to articulate one’s position better. Why Only Us was co-authored with Berwick, known for his computational models in language acquisition, language processing, and language change. It is easy to recognize the spirit of Chomsky, but it is Berwick’s writing style and tone that are most prominent.
Before addressing the arguments presented in Why Only Us, it should be pointed out that the book really ought to have been edited more carefully. The number of references mentioned in the text but missing from the bibliography and vice versa is not insignificant. This is unfortunate because many readers and especially students will be eager to familiarize themselves with Berwick and Chomsky’s latest ideas. More experienced readers, on the other hand, will find that there is very little new here. Virtually all these claims have been made previously.3 Those hoping for a synthesis will also be left disappointed.
Why Only Us consists of four loosely connected, but repetitive chapters. The question alluded to in the title is only addressed in three paragraphs on its final two pages.4 En route to those concluding paragraphs, the authors present a rich array of observations about the development of linguistics, evolution, and language. Although always interesting, their relevancy is often questionable.5
The intended audience for the book is difficult to discern. In places, the authors seem minded to set the popular science press straight on topics such as language genes or Neanderthal speech. But non-specialists will find the technical sections too opaque. Elsewhere, the authors appear to be addressing non-linguists working on language evolution. In the case of evolutionary biologists, I doubt they will enjoy being told about evolutionary theory in the opening chapter. They know better. Linguists, on the other hand, may well benefit from a primer, but will find that chapter way above their heads. Too basic for some, too technical for others, and too sketchy for all concerned.
Contrary to the picture presented in Why Only Us, the field of language evolution and cognitive biology is blossoming. Students and newcomers could be forgiven for assuming from the book that, aside from the authors and their close collaborators, everyone else is wrong and deeply confused about language evolution. Nothing could be further from the truth. Significant empirical progress is being made. Consider
- the accumulating body of work on FOXP2 and its interactome;6
- the newly discovered convergence among vocal learners on modifications of specific axon-guidance molecules to establish key neural circuits;7
- the insights gathered from songbirds and the prospects for research into vocal learning bats;8 and
- the growing appreciation of cognitive sophistication in the great apes.9
This list is by no means exhaustive.10 There is also research identifying mechanistic links between genes and the establishment of neural circuits that make learned, cortically controlled vocalizations possible—bridging the gap between our own and other species at the phonological level.11 Research into our closest living relatives has also revealed a much closer filiation in the domains of semantics and pragmatics. Signs of a vibrant research community can also be found in recent publications such as the Oxford Handbook of Language Evolution, and the Psychonomic Bulletin & Review special issue edited by W. Tecumseh Fitch.12
Making Language Evolution a Mystery
Berwick and Chomsky (B&C) co-authored a 2014 paper entitled “The Mystery of Language Evolution.”13 The spirit of that paper is very much present in Why Only Us. I would argue that the statement “only us” is largely responsible for making the question of language evolution a mystery.14
It makes sense to regard the study of language, and human cognition more generally, at least in part as a matter of biology. This is, in fact, the most important lesson that Chomsky and neurologist Eric Lenneberg have offered us. Nothing in biology, as anyone familiar with the work of Theodosius Dobzhansky will attest, makes sense except in the light of evolution. Nothing in evolution, Charles Darwin argued, makes sense except in the light of descent with modification. Modification opens the door to novelty, but it is always appended to descent—novelty without fundamental discontinuity.
Darwin’s explanatory logic conflicts with assertions that some species are unique. No matter how modest we rate the linguistic abilities of non-humans to be, and no matter how rich and complex we judge our own, if one is interested in the question of “why?” there must be a path from them to us.
To begin with the question “why only us?” is to start on the wrong foot.
According to B&C, human language can be reduced to a single unique trait: “the ability to construct a digitally infinite array of hierarchically structured expressions with determinate interpretations at the interfaces with other organic systems.”15 Marc Hauser, Chomsky, and Fitch termed this recursion in a notoriously polysemous essay for Science.16 Published in 2002, that essay provides the backbone for the position outlined in Why Only Us. Following Chomsky, linguists now refer to the fundamental operation of language as Merge.
This basic property of human language, as B&C describe it, is unique to anatomically modern humans. It did not necessarily emerge at exactly the same time as modern human anatomy, around two hundred thousand years ago. But it had emerged prior to the last exodus of modern humans from Africa, roughly eighty thousand years ago.17 This explains its species-uniform character.18 The possibility that archaic Homo possessed the basic property is ruled out.
B&C argue that language, reduced to the basic property, emerged during a brief window of time. They describe the basic property as simple, perfect, and “something like a snowflake.”19 It is the sort of property one would be reluctant to consign to the bricolage of evolution. The safest way to prevent natural selection from getting its hyperactive hands on the basic property is by not giving it enough time to do so—whence the insistence on a sudden and recent emergence.
A Shaky Argument
Whether the basic property is as neat, elegant, and minimal as B&C assert is a topic for another time. While such a property unquestionably exists, their characterization conceals numerous layers of complexity. The term Merge makes it seem atomic, but there is a lot of room at the bottom of atoms.20 There are other, equally specific basic properties. Linguists could provide a list of domain-specific properties that cannot, it seems, be reduced to Merge.21
If Merge is more complex than it sounds, and if there is more than Merge to human language, the question arises whether there is sufficient time for it all to be put together in the window proposed by B&C. We can’t know for sure. This is because, in part, biologists are not yet in a position to determine the speed at which evolution could assemble the building blocks. Precise details of what this integration would amount to are also sketchy at best.
The complexity of life is such that one can readily find examples of quick and dramatic changes over a few generations alongside much slower rates of change.22 We simply have no idea whether one hundred thousand years is an appropriate amount of time. If language is more complex, B&C may not be allowing enough time. If they are too generous, natural selection is then left with time on its hands to tinker with Merge.
What had to evolve was a particular brain configuration that made language acquisition possible. One might reasonably assume this arose in part via coordinated genetic mutations. Without knowing what these mutations were, how they were coordinated, and how complex the resulting configuration needed to be for implementation of the basic property—or, more plausibly, the basic properties—the question is unanswerable.
Progress on this issue requires engagement with ongoing work in neurogenetics. On its own, a general computational description just won’t do. A description couched in neural terms is needed, which must then be related to genes. There is no direct link between genes and Merge. The critical link between molecular biology and cognition is, of course, the brain.
It is not language that emerged, it is a language-ready brain.
On the issue of neural implementation, B&C simply state that the problem is, “little understood.”23 “[R]ecent empirical evidence,” they add, “suggests that this could be compatible with some ‘slight rewiring of the brain,’ as we have put it elsewhere.” Few neuroscientists would agree with them, because few neuroscientists reduce language to the basic property, and those who do so may not treat the basic property as atomic.
B&C review neurolinguist Angela Friederici’s work stressing the importance of the white matter fiber tracts connecting Broca’s and Wernicke’s areas of the brain; some of these have undergone expansion in our lineage.24 Knowing where it may be useful to begin looking does not explain how or why this aspect of the brain does what it is claimed to do.25
As Lenneberg once remarked, “nothing is gained by labeling the propensity for language as biological unless we can use this insight for new research directions—unless more specific correlates can be uncovered.”26 That is, unless we link the mental, the neural, and, eventually, the molecular.
A vast amount of data has been gathered in the wake of the genomic revolution. To exploit this data, detailed linking hypotheses are necessary. Here there are, as David Poeppel and David Embick have noted, two options: bridge the gap between mind and brain, or accept that the two domains are incommensurable, and there is nothing more to say.27 The first makes problems like language evolution tractable, while the second confines them to the realm of mysteries.
Lessons from Other Disciplines
Generative linguists are fond of the notion that linguistics is really biology at a suitable level of abstraction. To genuinely address the why questions, biology at a much more rigorous level is required. This involves linguists moving beyond their comfort zone and almost exclusive focus on the study of linguistic expressions. Since these are not found in extinct humans, or in living non-human animals, it is all too easy to declare that it is only us. This reflects a failure to engage with fields whose primary focus is not linguistic expressions, but brain rhythms or protein–protein interactions.28
In Why Only Us, the authors fail to give proper credit to the many serious attempts to bridge the gap between mind and brain, and, in particular, ongoing work with animal models.29 If we are the only linguistic creatures, it follows that there is little to be learned from animal models. This includes the songbirds, whose capacities are dismissed by B&C: “birdsong is only a model for speech, if that—not language.”30 This is an area where work linking genetics and neuroscience has been progressing rapidly. There is little mention of this in Why Only Us.
One way to date the emergence of the language faculty, according to B&C, is the glimpses of cognitive and behavioral modernity offered by the appearance of the first unambiguously symbolic artifacts in the fossil record.31 B&C cite the shell ornaments, use of pigments, and geometric engravings from Blombos Cave in South Africa, dated to around one hundred thousand years ago.32
An older version of this argument asserted that, based on an explosion of cultural artifacts in Europe forty thousand years ago, cognitive modernity emerged after the exodus from Africa. This notion was refuted by the paleoanthropologists Sally McBrearty and Alison Brooks, among others, nearly twenty years ago.33 Since then, evidence has steadily accumulated in favor of a much earlier date for the emergence of symbolic artifacts, up to five hundred thousand years ago34—not only in our species, but also in archaic Homo.35
This puts B&C in an odd position. At variance, once more, with the research community, they are left with two options. Either their date for the emergence of language needs to be pushed back, potentially granting linguistic capacity to archaic Homo, or they need to distinguish between kinds of symbolic artifacts, finding ways to reliably associate only more recent artifacts with our species. No criteria are provided for making such a distinction.
A causal link between artifact complexity and the language faculty is nothing more than a plausible argument that makes sense to our modern eyes. Past attempts by linguists to make such arguments have been quickly dropped when an artifact believed to be of modern origin has been reassigned to an extinct species, or another living species has been found capable of producing it.36
Despite an apparent lack of technocultural sophistication, communities such as the Pirahã are deemed by (generative) linguists to possess a fully modern mind and brain. If we are to make judgments on this basis, how can we be so sure that archaic Homo did not have a fully modern, but perhaps similarly underexploited, mind, brain, and language faculty?
Distinctions versus Dichotomies
The questions asked in Why Only Us echo those asked by Hauser, Chomsky, and Fitch in “The Faculty of Language: What Is It, Who Has It, and How Did It Evolve?”37 The difference is that in their new book, B&C answer the questions that they originally posed:38
- “What” boils down to the Basic Property of human language…
- “Who” is us—anatomically modern humans…
- “Where” and “When” point to… southern Africa roughly 200,000 years ago, but prior to the last African exodus approximately 60,000 years ago.
- “How” is the neural implementation of the Basic Property…
- “Why” is language’s use for internal thought…
The question of “Why?” is the only one missing from the earlier essay.
In their 2002 paper, Hauser, Chomsky, and Fitch distinguished between a faculty of language in the broad sense (FLB), and a faculty of language in the narrow sense (FLN): “FLB includes a sensory-motor system, a conceptual-intentional system, and the computational mechanisms for recursion … FLN only includes recursion and is the only uniquely human component of the faculty of language.”39 “[W]e leave these questions open,” they add, “restricting attention to FLN as just defined but leaving the possibility of a more inclusive definition open to further empirical research.”40 No longer considered a possibility, in Why Only Us it has become the only option—the basic property.41
The approach of dividing up the problem into FLB and FLN has, over time, ossified into a scenario in which the evolution of the basic property has become a keystone added to a preexisting structure. This is a very asymmetric way of thinking, but nonetheless consistent with other Chomskyan distinctions: core versus periphery and competence versus performance. It’s basic, or not. It’s there, or it’s not. It’s ours, or it’s not.
An alternative perspective much more in line with how evolutionary biologists understand novelties is the mosaic view. No single piece of a mosaic is more important than another, and while they may have distinct origins, it is only when they are assembled that a pattern emerges. Mental properties can, in my view, be considered in similar terms. The contours of their patterns may not necessarily respect the borders and pedigrees of the individual pieces.
Mosaic patterns are like the paintings of Giuseppe Arcimboldo in which fruit and vegetables organized in a particular way acquire novel functions: part of a pear, for example, might become a nose. The language faculty, I would argue, can be thought of the same way. It may be the case that no component of language is unique, or unique to humans. What is unique might simply be the particular way in which non-linguistic, evolutionarily old structures, observable in other species, are assembled. Assembly may occur gradually with parts put to use for nonlinguistic purposes, only to be later recycled. In contrast to B&C’s vision, components of language would have a much more generic character, consistent with the nature of neural circuits. This approach fits with the bottom-up comparative cognition program advocated by the ethologists Frans de Waal and Pier Francesco Ferrari:
Over the last few decades, comparative cognitive research has focused on the pinnacles of mental evolution, asking all-or-nothing questions such as which animals (if any) possess a theory of mind, culture, linguistic abilities, future planning, and so on. … A dramatic change in focus now seems to be under way, however, with increased appreciation that the basic building blocks of cognition might be shared across a wide range of species.42
The top-down, all-or-nothing perspective is enshrined in B&C’s title. It is this stance that separates them from the rest of the research community, which, without rejecting useful distinctions, has moved away from stark dichotomies.
The “only us” perspective has been around for a very long time, and the arguments offered by its proponents have remained remarkably consistent.43 But if they did not convince then, why would they convince now? Though much has been learned since Steven Pinker and Paul Bloom wrote their famous essay on language evolution almost thirty years ago,44 the facts of language evolution remain fully consistent with a normal Darwinian view.
Language is special, but not that special; all creatures have special abilities.
The field of linguistics has tried to move away, at least in spirit, if not in practice, from what the linguists Barbara Scholz and Geoffrey Pullum once termed exuberant nativism.45 In Chomskyan terms, this would be an endowment so rich, complex, and language-specific that it makes little biological sense. Generative linguists still adhere to what might be termed exceptional nativism: language is so special as to require a special explanation. As an alternative, I favor normal nativism, and I am far from alone in this respect.
It could be, of course, that Berwick and Chomsky are right. But the premises they adopt make the problem even harder than it already is. A pluralistic approach would require
- recognizing the mosaic character of language;
- insisting on making properties as basic as possible;
- recognizing the centrality of animal models;
- being open to the possibility that extinct archaic Homo had a fairly elaborate linguistic system;
- viewing cultural transmission as a major factor in linguistic complexity; and
- readily admitting that while the problem is hard it is not hopeless.
To conclude, a final remark on the first word in the title of the book. On the question of why, the authors write that language is used for internal thought and acts as “the cognitive glue that binds together other perceptual and information-processing cognitive systems.”46 They are not the first to make this argument.
In a book that often appeals to Lenneberg, I missed the references to the ethology literature that he would have liked. Nikolaas Tinbergen’s famous article on the four senses of why-questions in ethology continues to animate research in modern cognitive biology.46 Although Chomsky has drawn upon on the insights of ethologists elsewhere, it is disappointing that in this book Tinbergen’s influence has faded.
As Turing might have observed, Why Only Us is the wrong question.
