One of our long term research goals is to build a synthesis model which <br>
is able to produce spontaneous speech, including disfluencies. One of <br>
the aims is to gain a better understanding of which features contribute <br>
to the impression of hesitant speech on a surface level. The current <br>
study investigates acoustic correlates to perceived hesitation based on <br>
previous work showing that pause duration and final lengthening both <br>
contribute to the perception of hesitation. It is the total duration <br>
increase that is the valid cue rather than the contribution by either <br>
factor. The present experiment using speech synthesis was designed to <br>
evaluate F0 slope and presence vs. absence of creaky voice before the <br>
inserted hesitation in addition to durational cues. The manipulations <br>
occurred in two syntactic positions, within a phrase and between two <br>
phrases, respectively. The results showed that in addition to durational <br>
increase, variation of both F0 slope and creaky voice had perceptual <br>
effects, although to a much lesser degree. The results have a bearing on <br>
efforts to model spontaneous speech including disfluencies, to be <br>
explored, for example, in spoken dialogue systems. <br>