To the editors:

Juan Uriagereka sets up his review of Angela Friederici’s book with a clear message: “In Language in Our Brain, we are dealing with an obviously important book.” Approaching the topic from his vantage point, leading minimalist that he is, Uriagereka puts his theoretical finger on the empirical wound opened by this line of work: Is Merge really in the brain or is it kept in the mind?

I have two remarks to raise here. One concerns what Uriagereka does say in his review and the other what he does not.

Searching for the Mind in the Brain

Before zooming in on the localization of Merge—“the essential operation of Chomsky’s minimalism, because it is the simplest way of putting linguistic items together”—Uriagereka puts Friederici’s work on the neurobiology of language into historical perspective. He ties Santiago Ramón y Cajal’s contiguous neuron theory to the computational theory of mind, which is in no small part due to Noam Chomsky’s groundbreaking work in linguistics, and the often ignored insights into cognitive sciences that stem from it. The computational theory of mind, Uriagereka reminds the reader, allows us to explain three core properties of human language: compositionality, representation, and recursion.

Uriagereka certainly has his doubts about whether what underlies Language in Our Brain is really borne out: “the expectation, or the hope, that a division of labor into phonetics, morphology, syntax, semantics, and pragmatics more or less corresponds to the tasks the brain executes.” While admitting that a “functional language network is, no doubt, playing some kind of role in such processes,” he wonders, “whether the activity that imaging techniques reveal when our brain entertains these symbolic dependencies involves the grammarian’s Merge, or something else entirely.”

In this context, Uriagereka addresses the competence versus performance dichotomy, which is entirely absent from Friederici’s book, but of fundamental importance when attempting to relate neuroimaging activities to purported symbolic manipulations. “Present-day observational technology,” Uriagereka explains, “does not seem capable of teasing apart these different components of syntax at work, so it seems to me premature to claim that the observables localize Merge.”

At the bottom of such deductions lies the observation made by David Poeppel and David Embick that “[t]he formal structures of linguistics and neurophysiology are disjoint”—the so-called granularity mismatch problem.

Linguistic and neuroscientific studies of language operate with objects of different granularity. In particular, linguistic computation involves a number of fine-grained distinctions and explicit computational operations. Neuroscientific approaches to language operate in terms of broader conceptual distinctions.¹

Coupled with the ontological incommensurability problem, “the primitives of linguistic theory cannot be reduced to the fundamental units currently identified by neuroscience.”²

Uriagereka devotes the remainder of his review to skeptical ruminations about whether neuroimaging can tell us as much about Merge as Friederici reports. Even Norbert Hornstein, another staunch defender of theoretical syntax, has made the point that “[w]hatever operations grammarians propose must ultimately be embedded in brain circuitry,” that is, “brain structures that compute simple operations.”³ Hornstein adds: “A useful step in the direction of bridging the granularity problem would be for grammatical accounts to ‘make use of computational analyses that appeal to generic computational subroutines.’”⁴

One might wonder, as Evelina Leivada has suggested to me,⁵ whether Uriagereka does justice to Friederici when he writes that the book “does not attempt explicitly to ask such questions.” The field, it seems, still has not decided whether we should be talking about granularity mismatches—largely due to exposition and a matter of identifying the correct primitives—or dealing with deeply rooted incommensurability. No matter how we refine our definitions and no matter if we select the best of our primitives—i.e., the most promising ones for interdisciplinary explorations—we will still be stuck with the mismatch. Embick and Poeppel have not provided an answer, and neither has Chomsky. Do we really want this question to be first addressed outside the circle of linguists?

A decade ago, Hornstein tackled the issue within a formal approach. Taking his cue from Poeppel’s work, he laid out a proposal “to look for basic operations plausibly dischargeable by simple and general neural circuits in terms of which the laws of grammar can be coded.” This is an approach that fits “rather neatly with the view of the Minimalist Program.”⁶ According to Hornstein, we need “to find a class of very basic primitive operations that plausibly underlie linguistic computations[,] potential candidates for the primitive operations that might be reasonable building blocks for neural circuits.”⁷

This may be so, and Friederici’s research may just contribute evidence for identifying the relevant neural circuits. Still, Uriagereka is right to point out some of the difficulties involved with inferring the proposed functional language network from syntactic applications of Merge in phrase structure, such as the constituents that eventually yield “The man sailed the ship,” and even in word formation, as in transform–(a)tion–al–ize. Perhaps another relevant domain might be compounding. While Friederici reports her research team’s previous work on compounds, she restricts herself to discussing the acoustic differences relevant for the lexical access of wheel as a single word versus wheel in the compound wheelchair; she does not discuss the required application of Merge to combine the two.⁸

As a theoretician, Uriagereka has good reasons to restrict his discussion of Language in Our Brain to the first half of Friederici’s book: Part I is concerned with language functions in the brain, while Part II examines structural and functional language networks.

From Species to Species

My second remark concerns an important aspect of the book that Uriagereka does not touch on in his review. This is remarkable because Chomsky, in his foreword, highlights it as containing “Friederici’s most striking conclusions.” In particular, her findings in relation to “specific regions of Broca’s area (BA 44 and BA 45) and the white matter dorsal fiber tract that connects BA 44 to the posterior temporal cortex.”⁹ According to Friederici, “This fiber tract could be seen as the missing link which has to evolve in order to make the full language capacity possible.”¹⁰

One might be tempted to infer that Chomsky focuses on the evolutionary impact of research in the neurobiology of language so as not to get too heavily involved in the purported neurobiology of Merge. If so, Uriagereka does a great job at offering an explanation. After all, Chomsky has expressed doubts about whether linguistic theory and psychology or neuroscience should be brought closer together, even while he has acknowledged “different approaches to the study of language (vision, etc.), and intriguing problems where they intersect,” and advised researchers to pursue them.¹¹

Part IV of Language in Our Brain addresses language evolution and its neural basis. The evolutionary aspect of the work presented here is not solely confined to Friederici’s proposal that “the basic computation Merge, like other more complex syntactic processes, has a well-defined localization in the human brain.”¹² It also includes the suggestion that “[d]uring the evolution of language two crucial abilities had to evolve.” The first, sensory-motor learning involving simple rule-based sequences, can also be seen in songbirds. The second, an ability to process hierarchical structure, is not. “[I]t is conceivable,” Friederici writes, “that the ability to process structural hierarchies is what should be considered as a crucial step toward the language faculty.”¹³

This raises an interesting question, as Hornstein has pointed out to me: Does the human capacity build on the bird capacity?¹⁴ Or, as Hornstein has suggested elsewhere, is Merge a composite, part of which reflects the bird capacity to string sequences together and part of which involves the construction of complex hierarchies? Hornstein took labeling to be sufficient for the second part and assumed that stringing together labeled complexes will lead to the human capacity. Friederici seems to be proposing the same thing in her book. If so, this suggests a different evolutionary trajectory than Chomsky implies. Endless sequencing already has a touch of recursion built in, but with labeled objects, the process might yield the capacity we observe.

The emergence of the faculty of language in Homo sapiens is often referred to as Darwin’s problem or “the logical problem of language evolution.”¹⁵ It is arguably intertwined with the neurobiology of language, that is, the most ventral anterior part of BA 44 or other relevant neural circuits. “[I]f we are lucky,” Hornstein remarks, “the basic operations sought by minimalists will also help to solve the Granularity [Mismatch] Problem and Darwin’s Problem. All three are looking for the same kinds of basic operations.”¹⁶

The functional language network worked out by Friederici can also account for across-species structural differences: “macaques and chimpanzees display a strong ventral and a weak dorsal pathway, whereas humans display a strong dorsal pathway and a well-developed ventral pathway.” This variation leads her to pursue the dorsal pathway “as the crucial pathway for the language ability in human adults.”¹⁷

Among the findings highlighted by Chomsky, Friederici notes:

[T]hese phylogenetic and ontogenetic data provide compelling evidence for the specific role of BA 44 and the arcuate fasciculus connecting BA 44 and the posterior superior temporal cortex for the language faculty. These brain structures may have emerged to subserve the human capacity to process syntax. This fiber tract could be seen as the missing link that has to evolve in order to make the full human language capacity possible.¹⁸

From a minimalist vantage point, the ontological incommensurability problem may remain an obstacle for the neurobiology of Merge put forth by Friederici. But the huge steps she makes toward solving the granularity mismatch problem and Darwin’s problem are highly laudable.

Minding the Species’ Brain

Let me finish with an afterthought. Friederici begins the final chapter of Language in Our Brain with this summary:

Language is an integral part of our daily life. We use it effortlessly, and when doing so, we usually think about what we want to say rather than how to say it in a grammatically correct way. Thus we use language quite automatically. The intriguing question is what puts us humans in the position to learn and use language the way we do. I started out in the beginning of the book by saying that this is due to an innate language faculty. Over the course of the book it became clear that the language system is a complex system.¹⁹

Neither Uriagereka in his review nor this letter have touched on the book’s third topic, an insightful discussion of how language is acquired and learned. The section on language acquisition links the origin and use of language to the neurobiological underpinnings of the child’s language development period. Friederici summarizes the relevant findings as follows:

[T]he path from sensory-auditory system to motor-output system is crucial for song learning. A similar pathway from the auditory system to the motor system is already present at birth in the human infant, providing a good basis for auditory learning and babbling in humans. This pathway is the dorsal pathway connecting the posterior superior temporal gyrus and the premotor cortex (Perani et al., 2011), which must be distinguished from the dorsal pathway relevant for syntax targeting BA 44.²⁰

In her excellent doctoral dissertation, Leivada casts a critical eye over the parameter-driven view of generative grammar and proposes a bold hypothesis about why “certain domains of language seem to be particularly vulnerable to impairment, while others appear to be consistently preserved.”²¹ According to her locus preservation hypothesis, “Syntactic operations are impenetrable to variation—both across languages and across pathologies; as such, their manifestations are universally preserved across (a)typical cognitive phenotypes.”²²

Friederici’s gradual exposition of language as a complex system seems to be correct, also from the perspective of how resilient some parts of language are to induced change due to pathological reasons. The capacity of a population’s developmental process to produce an invariant phenotype regardless of genetic perturbations, termed canalization, is manifested in language as it is in other complex systems. The child’s brain undergoes a series of processes that permit neural pathways, such as those identified by Friederici, to stabilize in order to produce a uniform result in spite of genetic and environmental variations.²³ This is a fitting perspective, since, as Chomsky noted in his essay “The Galilean Challenge,” the publication of Language in Our Brain was “scheduled to coincide with the fiftieth anniversary of Eric Lenneberg’s classic study, Biological Foundations of Language.”²⁴

Studies of language within the mind, extended into the brain, and illuminating its origin in the species form the core of Friederici’s book. Uriagereka might be right that some conclusions come across as being a little premature, but that is the exciting promise of ongoing research in biolinguistics.²⁵