Patrick Hanks sees linguistic approaches to word meaning as divided between two unattractive extremes. Generative theories, such as were pioneered by Katz and Fodor (1963) and pursued recently e.g. by Wierzbicka (1996), attempt to capture meanings with an apparatus of quasi-mathematical rules and universal semantic primitives which is unequal to reflecting the messy realities revealed by empirical corpus studies. On the other hand, the doctrine of linguistic creativity advanced by Sampson (1980, 2001) is unduly defeatist in denying the possibility of scientific analysis. Hanks argues that theoretical linguistics and practical lexicography should both embrace an intermediate position which distinguishes between high-frequency “norms” of usage and rare “exploitations”. This allows linguists and lexicographers to produce scientific lexical description while nevertheless acknowledging messy variability.