We described a patient with a dramatic deficit of both word comprehension and naming but with good preservation of visual pictorial semantics. On word-picture matching, his performances were slightly better than expected based on the observed lexical semantic disorder; in addition, the patient, who maintained good preservation of his underlying phonology, showed a tendency to point to the picture phonologically related to the target. In order to interpret these data, we advanced the hypothesis that the patient, in spite of his virtually complete inability to name, would be able, in a word-picture matching task, to "covertly" (i.e., preverbally) retrieve the name from the picture and to use this name to attempt a match with the phonological form of the stimulus word. This mechanism, that we called "phonological" comprehension, would allow the identification of the correct target and would explain the choice of the phonologically related foil that was sometimes selected.