The hypothesis that closed-class items which participate in theta-role assignment are less problematic in agrammatism than items which do not (Rizzi, 1985) is put to an empirical test. Five French-speaking agrammatic patients were tested in a sentence-picture matching paradigm to probe their comprehension of sentences containing articles, which are not involved in theta-role assignment, and of sentences containing pronouns, which in the direct object position are homophonous with articles and are theta-role assignees. Gender was used as a variable to differentiate between target and distractor. The data indicate that pronouns are significantly more difficult to process than articles. This result disconfirms the claim that the availability of grammatical information encoded in closed-class items is a function of their involvement in theta-role assignment. The present study demonstrates that the ability to process gender marked articles is generally well preserved in French-speaking agrammatic patients.