Until fairly recently almost everything we knew about the cognitive neuroscience of language came to us from investigation of the correlations between patterns of language impairment and their associated loci of brain damage. This approach has contributed importantly to a first-pass characterization of the distribution of language functions in the brain. But its contribution has been even greater to the development of functional theories of language processing. The patterns of dissociations and associations of deficits have informed theories of various language processes from the perception and production of speech, to the organization of the lexicon, to the syntactic and semantic processes involved in sentence comprehension and production. Especially important inroads have been made in characterizing the processing machinery that makes reading and writing possible. Another fertile area of research has been the relation between cognitive and linguistic processes, such as the role of working memory in sentence processing.

Sometimes the observed dissociations and associations of deficits have been rather unexpected, forcing reconsideration of the received view in a given area of language processing. For example, it is now well established that there are patients who make semantic errors in naming pictures orally but can write the name without difficulty, and there are patients with the reverse pattern of dissociation, who make semantic errors in writing the name of an object but who can name orally without problem. These results imply, perhaps unsurprisingly, that the phonological and the orthographic forms of words are represented by distinct neural mechanisms. However, they are not so easily accommodated in theoretical frameworks that hypothesize an abstract lexical representation between the semantic and lexical form levels. Furthermore, since these modality-specific naming deficits have also been found to be restricted to words of one grammatical class or another, they place even stronger constraints on a biologically defensible theory of the lexicon. Thus, for example, it has been shown that patients may have difficulty in writing verbs but not nouns or speaking verbs but not nouns, while showing no difficulty in the other modality of output for both verbs and nouns (Caramazza & Hillis, 1991). The reverse dissociation, modality-specific selective difficulty with nouns, has also been documented. One implication of these results is that grammatical operations are carried out over modality-specific and not abstract lexical representations. This conclusion is in line with several of the chapters in this section that emphasize the “task-dependent” nature of the representations computed in the course of language processing.

Highly selective dissociations such as those mentioned here place strong constraints on theories of the functional organization of language processes. However, they have been less useful in informing theories of the brain structures that are associated with hypothesized cognitive and linguistic processes. The reason for this discrepancy may be found in the fact that these highly selective deficits are rare and that they are associated with fairly large lesions, making it extremely difficult, if not impossible, to identify the brain regions causally related to the observed dissociations.

Fortunately, the neuropsychological approach is now complemented by various neuroimaging methods that can be used to systematically assess cognitively motivated hypotheses in the intact human brain. This approach is extremely promising but is still in its early stages of development. It has proven difficult to ascribe a causal role to the many brain areas that are activated when performing a complex task. For this reason, it is important to combine the methods of neuropsychology (or transcranial magnetic stimulation), which allow stronger inferences about the causal role of a brain area in the performance of a task, with the increasingly sophisticated use of MRI, EEG, and MEG methods. The chapters in this section fully exemplify the close link between the study of language disorders and neuroimaging research that are being used to converge on a cognitive neuroscience theory of language.

The chapters in this section cover only a limited set of questions about language, reflecting the areas of greatest current interest. The topics include phonological (Hickok) and morphological processing (Shapiro and Caramazza), reading (Cohen and Dehaene), syntactic (Caplan) and semantic processing (Hagoort, Baggio, and Willems), language acquisition (Kuhl), and the genetics (Ramus and Fisher) and evolution of language (Fitch). And, as already noted, in all the chapters that focused on human research, the interplay between the results obtained from the study of language disorders and those obtained from neuroimaging research plays a central role. The result is a comprehensive view of our current understanding of the neurobiology of language.

Consider as an example chapter 54, on reading, which focuses on the word-recognition component of the reading process. Cohen and Dehaene review the various forms of reading impairments that affect some or other aspect of word recognition and conclude that the impairments can be subdivided into two sets: those that principally involve letter recognition and those that involve operations over letters, such as attention mechanisms and serial decoding. They note that the lesions associated with these two broad types of processes concern, respectively, ventral and dorsal visual pathways. They also review imaging results that converge with this interpretation of the neuropsychological evidence. The result is a fairly comprehensive if preliminary view of the neural machinery involved in visual word recognition.

Or consider the case of phonological processing. In chapter 52, Hickok also argues for a two-stream circuit for phonological processing, one dedicated to speech recognition and the other to speech production. Recognition involves primarily a circuit that includes the superior temporal sulcus bilaterally, while production involves a left-dominant frontoparietal/temporal circuit. The evidence for this distinction comes primarily from the patterns of phonological processing deficits in aphasic patients, but Hickok also reviews imaging results that converge with the neuropsychological evidence.

Despite the important developments charted in the chapters included in this section, it is clear that we are still very far from an articulated theory of the biology of language. In some respects we are really only now beginning to develop the methodological and theoretical foundations for such a theory. This fact is illustrated in the beautiful work on the genetics of language described in chapter 58 by Ramus and Fisher, where it is clear that we have only begun to scratch the surface of the many complicated factors that enter in a genetic theory. The same is true for cognitive neuroscience accounts of language, especially for the more complex functions such as syntactic processing and semantic integration. At this stage of the game we have many titillating insights but not yet articulated theories.