Objective.-The recovery of language function seen in poststroke aphasia is significantly related to the severity of the initial neurologic deficit. However, a great deal of variability still remains in the improvement that is actually achieved. To detect predictors that will help in explaining this variability, positron emission tomography (PET) and fludeoxyglucose F 18 (18-F-fluorodeoxyglucose) were used and the results were analyzed by stepwise regression. Design.-Using PET imaging after injection of fludeoxyglucose F 18, the regional changes in glucose metabolism in 26 patients at a period of 12 to 18 days following an ischemic stroke involving the left middle cerebral artery were examined. A second PET examination was performed on 17 of our 26 patients who were able to perform speech activation exercises. All patients received an initial and a 4-month follow-up evaluation of languange performance. Setting.-During the two PET studies and the first language assessment, the patients were hospitalized in a neurologic clinic. The follow-up evaluation of language performance was performed when the patients were ambulatory. Patients.-Twenty-six patients (10 women, 16 men; aged 38 to 77 years; mean+/-SD, 60+/-9.2 years) were selected in the study. Their aphasias were of various types and of varying severity ranging from mild impairment to severe global aphasia. Main Outcome Measures.-For the stepwise regression analysis of variables, the following variables were analyzed in resting and activation PET to explain residual variance from the first to the second Token Test: regional cerebral metabolic rate for glucose of infarct and mirror region, left and right cerebral and cerebellar hemispheres, left and right Broca's area, left and right Wernicke's area, and left and right temporoparietal cortex. Results.-As was expected, early and late Token Tests exhibit a high correlation (.85). The stepwise regression analysis shows that only the left cerebral hemisphere glucose value of the resting PET had significant effect on the residual variance of the Token Test regression. Regional metabolic rates during speech activation had the largest contribution to a significant recovery from aphasia. The infarct area and its corresponding mirror region, the left Broca's area, and the entire left cerebral hemisphere accounted for 80% of the residual variance. Conclusions.-These results emphasize not only the application of PET activation studies in the prediction of a tissue's potential reserve capacity but also the importance of left hemisphere integrity in the recovery of functional language.
