We have used on-line psycholinguistic and functional neuroimaging techniques to investigate the cognitive and neural bases of sentence-processing. In one study, PET data and reaction time data were used to determine what areas of the brain are responsible for syntactic and semantic processing of written sentences (Stromswold et al., 1996). Subjects performed 3 tasks.

In task 1, subjects judged whether center-embedded relative clauses sentences were semantically plausible (e.g., the juice that the child spilled stained the rug) or implausible (e.g., *the child that the juice spilled stained the rug).  Task 2 was identical to task 1 except subjects judged right-branching relative clause sentences (e.g., the child spilled the juice that stained the rug vs. *The juice stained the rug that spilled the child).  

karin4The only difference between task 1 and task 2 is that subjects have to "hold" the noun phrase the juice in a memory buffer longer in task 1 than in task 2.  As shown in Figure 1, the only region that was significantly more active in task 1 than task 2 was part of Broca's area (the pars opercularis, p < .008), suggesting that this region is the site of the memory buffer used to process sentences with filler-gap relationships. 

In task 3, subjects judged acceptable center-embedded and right-branching sentences, and structurally identical sentences that were unacceptable because they contained a pseudo-word (e.g., the center-embedded sentence *The juice that the child mulved stained the rug and the right-branching sentence *The child spilled the juice that mulved the rug.). Analyses of the reaction time data revealed that subjects did parse syntactic structures in task 3, but they did not semantically interpret the sentences. Only left perisylvian brain regions were significantly more activity in tasks 1 and 2 than in task 3, suggesting that these regions are involved in the semantic interpretation of sentences.

Subsequent studies have revealed similar results for female subjects, for subjects who were concurrently engaged in articulation (suggesting that the activity in pars opercularis does not merely reflect a strategy of subvocalization for the center-embedded sentences), and for subjects who judged the plausibility of auditorily presented cleft subject sentences (e.g., It was the child that spilled the juice versus *It was the juice that spilled the child) and cleft object sentences (e.g., It was the rug that the juice stained versus *It was the juice that the rug stained).  

Stromswold, K., Caplan, D., Alpert, N., & Rausch, S. 1996. Localization of syntactic comprehension by Positron Emission Tomography.  Brain & Language, 52, 452-473