Abstract: In the past 10 years a text-to-speech system including a phonetic library, tonemodel and basic synthetic rules had been established in IAAS. At present the speech synthesis of unrestricted vocabulary has been attained,but the naturalness has to be improved.The effect of segmental and supra-segmental feature of speech upon naturalness of syntheticspeech had been studied. This study shows that the rhythm and coarticulation occupy a dominant position for improving the naturalness of synthetic speech. The tone model for synthesis of short sentence is suitable. But if the synthesis of larger linguistic unit than simple sentence is considered,the intonation must be controled carefully. In this article the subjectiveassessment method and the results of the study on the naturalness of synthetic Chinese willbe represented