Can fMRI inform cognitive models of language?

The ultimate goal of language research is to understand the nature of the representations and computations that underlie language comprehension and production. Cognitive scientists are often skeptical about the usefulness of tools like fMRI for informing theories of language comprehension and production (e.g., Coltheart, 2006). We are sympathetic to this skepticism. However, although a fair number of fMRI studies exist where the main research question seems to be "where does mental process X happen in the brain", there are plenty of fMRI investigations that have taught us a great deal about human cognitive architecture (see e.g., Kanwisher, 2010, for a review of some such studies). In fact, for addressing certain core questions (e.g., do two mental processes rely on the same cognitive and neural mechanisms?), fMRI is arguably the best method available.

Understanding the degree to which a brain region is specialized for a particular mental process is a key way to narrow down the range of possible hypotheses about the function(s) of the region. We can ask this question for mental processes within the domain of language (e.g., does the retrieval of individual word meanings rely on the same brain region(s) as processes required for constructing complex meanings from individual words?), or across domains (e.g., does linguistic processing rely on the same brain region(s) as musical processing or general working memory?). In our ongoing work, we are characterizing each core high-level language region in terms of the linguistic manipulations it is sensitive to as well as in terms of its engagement in various nonlinguistic functions. We can then start formulating testable hypotheses about the computations that each of these regions may support.

Of course, the high-level language system does not operate in isolation. It interacts with speech perception and speech articulation regions, with brain regions that support social cognition, with domain-general brain regions of the fronto-parietal network that support goal-directed behavior, and with brain regions that store abstract conceptual knowledge, among others. To understand certain properties of the language system it will therefore be critical to study the interaction of language regions with other cognitive and neural systems.