Speech timing patterning as an indicator of discourse and syntactic boundaries

Although the perceptual reality of speech rhythm has not been unambiguously observed in acoustic correlates, speech has always been considered as rhythmic one way or another. This study looks at speech rhythm in spontaneous speech and its relationship to discourse and syntactic units. Two four-frame comic strips were used to elicit speech, and syllable onsets were measured from waveform/spectrogram displays. Inter-speaker variability due to different speaking styles is eliminated by zscore normalization. The monologues were transcribed and segmented into Discourse Segment Units according to a discourse structure model [1]. Syntactic units such as clauses and phrases were also identified. Results showed that isochrony is generally preserved until the pre-boundary syllable. Preboundary syllables are characterized by a significant lengthening effect, which differs by boundary types. Lengthening before discourse units was longer than that tied syntactic units such as clauses or phrases.