Auditory comprehension is commonly measured with multiple choice tasks. The sensitivity of these tasks in identifying deficits, however, is limited by credit given for correct guesses by forced choice. In this study, we compare performance on the multiple choice task to an alternative word/picture verification task, in 122 subjects with acute left hemisphere stroke. As predicted, word/picture verification identifies significantly more subjects with deficits in auditory comprehension. Furthermore, the pattern of errors in the two tasks is consistent with a current distributed model of semantic memory, and inconsistent with alternative local models.