Studies by Pinker (1979; 1984) have argued that lexical categories cannot be acquired on the basis of distributional information. However, more recent studies (e.g. Cartwright & Brent, 1997; Mintz, 2003; Redington et al., 1998) have presented new distributional approaches to lexical category acquisition and argued that such approaches overcome the problems with distributional analysis identified by Pinker. The current article provides a detailed critical assessment of the new distributional approaches to lexical category acquisition by focussing on two such approaches - the generalized minimal pair approach of Cartwright and Brent (1997) and the frequent frame approach of Mintz (2003). The article explores ten problems with such models such as their tendency to lump and split lexical categories, their habit of assigning certain classes of words to the wrong lexical categories and their inability to deal with crosslinguistic data. It is concluded that Pinker (1979; 1984) was correct to be pessimistic about the prospects for acquiring lexical categories on the basis of distributional information. Furthermore, it is argued that such a conclusion places language acquisition research in a quandary since it suggests that there is no form of information that children could use to acquire lexical categories.