Categories in the lexicon* JAN DON Abstract This article argues against the hypothesis that roots are stored in the lexi- con without categorial specification, such as noun and verb, as proposed in Marantz (1997, 2001). On the basis of evidence from Dutch, we show that certain generalizations and rules cannot be expressed without having roots that are lexically specified for their category. Furthermore, we show that the arguments put forward by Barner and Bale (2002) for categorial un- derspecification are not valid with respect to the data from Dutch. Finally, following Kiparsky (1997), we show that analyses of denominal verbs in English that embrace the categorial underspecification hypothesis run into serious problems. We conclude that roots are stored in the lexicon with a categorial specification. Introduction This article argues against the hypothesis, proposed by Marantz (1997, 2001) and supported in Barner and Bale (2002) that lexical categories are syntactically determined rather than the result of a lexical specification of roots. Their idea is that roots are stored in the lexicon without such cate- gorial specification and that only after insertion in syntax, the category of a word is contextually determined. This paper presents several new arguments against this position, and also reexamines several arguments that have been put forward in the literature against specific proposals for the analysis of so-called denominal verbs, as in Hale and Keyser (1993). We will show that these arguments also apply to Marantz’ proposal for a category-less lexicon. The central topic around which these arguments cluster is zero deri- vation, that is, the phenomenon that words without any phonological change may occur as members of di¤erent syntactic categories. In a Linguistics 42–5 (2004), 931–956 0024–3949/04/0042–0931 6 Walter de Gruyter