This paper proposes an understanding of agrammatism from the perspective of a recent usage-based theory of grammatical status, the ProGram theory (Boye and Harder, 2012). According to this theory, grammatical elements have two central properties: they are by convention discursively secondary (i.e. attentional background) and dependent on combination with a host item. The paper first surveys studies of agrammatic speech which, based on or reconsidered in relation to the above-mentioned criteria, show that the usage-based theory makes correct predictions about the behaviour of linguistic elements in agrammatic speech. Subsequently, the paper outlines and discusses two hypotheses about the mechanism behind agrammatism that can be derived from each of the two central properties of grammatical items. According to the prominence hypothesis, agrammatism is due to insufficient overall processing resources; this leads to a prioritization of lexical over grammatical expressions because the latter, being discursively secondary, can be dispensed with for communicative purposes. According to the dependence hypothesis, agrammatism results from an impaired capacity for combining or unifying simple elements into complex wholes: This impairment affects grammatical elements in particular, because these are dependent on (combination with) host items.