Research involving lexical components of both text-to-speech and recognition systems is of vital importance for their total performance. Lexicons vary in content dependent on the theory behind the systems and their proposed use. For a number of years, we have been working with lexica for different European languages as part of our multi-language text-to-speech project. However, the need for more detailed lexical descriptions is increasing in our projects. We have chosen to base our current efforts on a morphological component, developed at the University of Helsinki. This component, based on many different kinds of corpora, has a very good coverage of running text. The morph dictionary has been transcribed with the help of the KTH text-to-speech system. A partial morphological analysis of native names has proven to be efficient ito predict the pronunciation of names in a telephone directory. We will also report on a project where we have applied knowledge about lexical structure to facilitate lexical search for motorically disabled or non-speaking computer users.