The claim made in this paper is that various aspects of this analysis are unmotivated. It is argued that the standard analysis is the consequence of a theory in which grammatical functions are represented in terms of constituent structure, and that an approach in which grammatical functions are modeled as a distinct dimension of linguistic structure is better able to account for both the grammatical functions and the constituency. On the other hand, it supports the hypothesis that the structure of the action nominal in Hebrew includes both verbal and nominal projections. In LFG terms, the Hebrew action nominal is what Bresnan (1997) calls a “mixed category,” and it is shown how Bresnan’s theory accounts for the mixed properties of the action nominal without derivations and empty categories. It is also argued that the lexicalist implementation of the NP-over-VP analysis of mixed categories is superior to the derivational implementation in the analysis of Hebrew action nominals, in that it provides an account of both the pattern with accusative "objects" and the so-called "passive" pattern.