Some studies of stroke patients with semantic deficits have found no effect of word frequency on semantic tasks, as well as inconsistent performance across items and tasks. A deficit in semantic control has been suggested as the source of the deficit - i.e., an inability to focus on semantic features appropriate to the task. In the present study, two stroke patients performed significantly better in single-distractor versions (low semantic control) than multiple-distractor versions of semantic tasks (high semantic control) of comprehension tasks, which appears consistent with the semantic control hypothesis. On the other hand, two aphasic patients showed substantially better performance for auditory than visual presentation of words in comprehension tasks – a finding that is not expected on the basis of semantic control. Experiment 1 evaluated whether performance on a multiple-distractor comprehension task could be predicted solely on the basis of performance on a single-distractor version using Luce’s choice axiom. Single distractor performance significantly predicted performance and no convincing evidence was obtained for a role for semantic control. Experiment 2, which examined the modality effect, showed that for one of the patients, worse performance with auditory presentation was most likely due to rapid decay of phonological representations. For the other, worse performance was most likely due to a disruption to phonological representations of words or to their connection to semantic representations. In all, the results suggest that word comprehension deficits in aphasia can result from a variety of sources and not all are due to semantic control deficits.