The possibility of using a logic grammar as the basis for automatically generating a part-of-speech lexicon is investigated, and an implementation of such a lexicon learner is demonstrated together with several possible improvements concerning efficiency and other factors.