Sentence comprehension is a complex process involving at least a grammatical processor and a procedural component that supports language computations. One type of cerebral architecture that may underlie sentence processing is a network of distributed brain regions. We report two experiments designed to evaluate the cognitive and physiological substrate of sentence processing diaculties in nondemented patients with Parkinson's disease (PD). In the first experiment, patients answered simple questions about sentences that varied in their computational demands. Group and individual patient analyses indicated that PD patients are significantly compromised on this task, and that their difficulties become more prominent as the computational demands of the sentences increase. We manipulated the set of sentences to stress performance aspects of sentence processing. PD patients were compromised in their ability to detect errors in the presence and nature of a sentence's grammatical morphemes, suggesting a deficit in selective attention, but their ability to answer questions about a sentence was not afFected by short-term memory factors. In the second experiment, positron emission tomography was used to correlate this pattern of sentence comprehension impairment with regional cerebral glucose metabolism (rCMRgl) obtained at rest in a representative subset of these PD patients. Grammatical comprehension and attention in sentence processing correlated significantly with mesial frontal rCMRgl. Regression analyses confirmed the central role of left mesial frontal cortex, and identified a subsidiary role for left caudate in overall sentence comprehension, for left dorsolateral frontal cortex in grammatical processing, and for bilateral dorsolateral frontal cortex in attending to the presence of grammatical features. We conclude that compromised mesial frontal functioning underlies in part the sentence processing deficit of these patients, and these data illustrate one method for mapping portions of a sentence processing mechanism onto a distributed cerebral architecture.