This paper reports on a joint project between the Royal Institute of Technology in Stockholm and the University of Iceland in Reykjavik. The project involves the adaptation to Icelandic of a (multi-lingual) text-to-speech synthesis system. The main purpose of the project has been the development of a communication aid for the handicapped population. Here, however, we concentrate on certain linguistic aspects of this project, mainly the modelling of the interaction between phonetics, phonology and morphology. Among the issues discussed are the phonetic and phonological aspects of preaspiration, devoicing of sonorants, palatalization, diphthongization and the influence of morphological boundaries on the applicability of certain phonological rules. The results presented illustrate potential of text-to-speech systems as a test-bench for linguistic studies.