Subjects with brain damage resulting in agrammatic aphasia frequently omit or substitute function items (function words and inflectional affixes). However, they show only mild deficits in using meaningful content words. Agrammatics' performance on comprehension tests reveals a rather complex pattern. They usually understand active sentences correctly, but perform on chance level on passives. The same contrast is observed for more complex sentence types, such as subject vs. object clefts or relatives. This complex comprehension pattern suggests that agrammatism is a syntactic disturbance that selectively affects particular sentence structures. Nevertheless, it is possible that both the production and comprehension patterns go back to an access problem for grammatical morphemes. If an agrammatic aphasic comprehends only 50% of these morphemes, she or he may well be problem-free in understanding the resulting “pruned” active sentences in which some of the grammatical morphemes are missing. In contrast, she or he is likely to arrive at the wrong meaning when confronted with “pruned” passives. This hypothesis raises the question of how fully competent speakers interpret ungrammatical “pruned” strings derived from well-formed sentences by systematically deleting function items. Experimental data demonstrate that competent speakers approximate the agrammatic comprehension pattern when being presented with “pruned” strings. This argues that agrammatism can, indeed, be viewed as a deficit in processing function words and inflectional affixes, which may manifest itself in production and/or comprehension tasks. Assuming that cortical cell assemblies with distinct topographies correspond to content words and grammatical morphemes, it is possible to explain agrammatism in terms of lesions to these neuronal networks. This neurobiological model can explain additional aspects of agrammatics' performance.