Naming manipulable objects: anatomy of a category specific effect in left temporal tumours.

Whether semantic knowledge is categorically organized or is based in an undifferentiated distributed network within the temporal lobes or it is at least partially organized in property-based networks is still an open issue. With a naming task involving living and nonliving entities, the latter divided according to degree of manipulability, we studied a group of 30 tumour patients with either right, left anterior or left posterior temporal lobes' lesions and a herpes simplex encephalitis patient (MU). Both cross-subject and cross-stimulus analyses were conducted. Left hemisphere patients were overall worse than both right hemisphere patients and controls in the naming task. They moreover named nonliving items worse than living. This effect was larger in left posterior temporal than both right temporal and also left anterior temporal patients and significant both at a cross-subject and cross-stimulus levels of analysis. In addition the left posterior temporal group had more difficulties with highly manipulable objects than left anterior temporal patients, but the effect was significant only on a cross-subject analysis. VLSM lesion analysis revealed that the area most critically associated with the larger naming deficit for manipulable objects was the posterior superior portion of the left temporal lobe, particularly the posterior middle temporal gyrus. These results support a 'property-based networks' account of semantic knowledge rather than an 'undifferentiated network' account. For manipulable objects, this would be a posterior-temporal/inferior-parietal left hemisphere "action/manipulation-property-based" network related to the dorsal pathways which is thought to be important in action control, as suggested by neuroimaging results.