Many fields of language and speech research need a tool to get full information about phenomena that occur at the lexical level of language. To this end, a countrywide accessible database is being built with lexical information of the Dutch and English languages. CELEX is the co-operative effort of the University of Nijmegen, the Max-Planck-Institute for Psycholinguistics, the Institute for Dutch Lexicology, the Institute for Perception Research, and the Dr. Neher Laboratories of the Dutch Telecommunication Company. The design of a coherent lexical database with generally usable information about, amongst other things, morphology and syntax, leads to a challenging conflict. On the one hand, the objective of CELEX demands that the representation of all occurring phenomena should be independent - as far as possible -of specific language theories. On the other hand, the value of a lexical database is to a great extent dependent on the relational structure, which is based precisely on linguistic theories. In other words, a lexical database, which makes it possible to test theories, is itself a necessary tool for the construction of a lexical database. The pragmatic approach of the CELEX-project means that countless applications will be possible, but choices will have to be made. One application will be to research the optimal structure of theoretically-founded lexica and lexical databases.