During 1997-2001 an Estonian text-to-speech synthesiser was developed. Implementation of text-to-speech synthesis, however, is still limited by the monotony and the poor coherence of the output speech. The prosody of interrogative sentences has proved a particularly challenging modelling problem. The present study deals with the modelling of the fundamental frequency of kas-initial general questions. Such questions make up 22.5% of the 9000-strong sample of interrogative sentences derived from the Tartu University Corpus of Estonian Written Texts. Of this sample 40 relevant questions were recorded as pronounced by four voices, two of them male and two female. The analysis of the recordings revealed three basic contours that can be evoked by specific combinations of formalisable conditions. The report addresses the modelling of those conditions for the synthesiser.