Extensive clinical and imaging research has characterized the neural networks mediating the adaptive distribution of spatial attention. In everyday behavior, the distribution of attention is guided not only by extrapersonal targets but also by mental representations of their spatial layout. We used eventrelated functional magnetic resonance imaging to identify the neural system involved in directing attention to locations in arrays held as mental representations, and to compare it with the system for directing spatial attention to locations in the external world. We found that these two crucial aspects of spatial cognition are subserved by extensively overlapping networks. However, we also found that a region of right parietal cortex selectively participated in orienting attention to the extrapersonal space, whereas several frontal lobe regions selectively participated in orienting attention within on-line mental representations.

The meaning of a sentence differs from the sum of the meanings of its constituents. Left anterior temporal cortex responds to sentences more strongly than to unconnected words. We hypothesized that the anterior temporal response to sentences is due to this difference in meaning (compositional semantics). Using positron emission tomography (PET), we studied four experimental conditions (2 × 2 factorial design): In one condition, subjects read normal sentences. In a second condition, they read grammatically correct sentences containing numerous semantic violations (semantically random sentences). In a third condition, we scrambled the word order within the normal sentences, and, in a fourth condition, the word order was scrambled within the semantically random sentences. The left anterior temporal pole responded strongly to sentences compared to scrambled versions of sentences. A similar although weaker response occurred in the left anterior superior temporal sulcus and the left posterior middle temporal gyrus. A subset of voxels within the left anterior temporal pole responded more to semantically random sentences and their scrambled versions than to normal sentences and the corresponding scrambled versions (main effect of semantic randomness). Finally, the grammatical and the semantic factor interacted in a subset of voxels within the anterior temporal pole: Activity was higher when subjects read normal sentences compared to their scrambled versions but not for semantically random sentences compared to their corresponding scrambled versions. The effects of grammar and meaning and, most importantly, the interaction between grammatical and semantic factors are compatible with the hypothesis that the left anterior temporal pole contributes to the composition of sentence meaning.