In this paper a method for the generation of test vocabularies is introduced, basically aimed for the assessment of the intelligibility of speech output systems. The approach is to generate application-oriented vocabularies which mirror the characteristics of the vocabulary typical of the application field. Mostly mono-and poly-syllabic nonwords are generated that follow the phonemotactic and morphotactic rules of the target language, here of the German language. Statistic data that have an impact on the assessment procedure - and thus on the test results - are taken into account. In the context of this paper the generation procedure of a test vocabulary which is typical of an address-optimized synthesis system is given.