Perceived speech tempo and vowel space size

Since the seminal work of Lindblom (1963) a wealth of studies have investigated undershoot and the factors governing its occurrence and variability. However, although sex-specific differences in reduction patterns have been found in adults (Byrd 1994; Whiteside 1996), as well as in children (McCormack & Knighton 1996), possible sex-specific differences in formant transitions and in particular undershoot have received little attention. In a large acoustic study of German vowels using data from the Kiel Corpus of Read and Spontaneous Speech (IPDS 1994, Simpson 1998), a number of unexpected sex-specific differences in correlations between the first two vowel formants and vowel duration were identified. In particular, several statistically significant correlations between acoustic vowel quality and duration present in male vowel categories were absent in their female congeners. The absence of a significant correlation between F1 or F2 and duration suggests that the acoustic target of a vowel has been attained within the range of durational variability. Two main explanations are possible for this. First, female speakers exhibit faster post-and pre-consonantal formant transitions allowing them to attain vocalic target positions earlier than male speakers. The second possibility is that there is no difference in the dynamics of formal transitions, but female vowels are in general longer, hence increasing the chance that the vowel target will be attained by the temporal midpoint of the vowel. Indeed, longer durations and slower speaking rates have been found in female speech (Byrd 1994, Whiteside 1996, Simpson & Ericsdotter 2003). In contrast, a common stereotype is that females speak faster than males. A possible explanation for this mismatch might be the relation between perceived speech tempo and vowel space size (which is larger in female speakers). A listening experiment was conducted which set out to test whether differences in the size of speakers' acoustic vowel spaces correlates with perceived tempo. In other words, is a speaker traversing a large acoustic vowel space within the same time-frame as a speaker with a smaller acoustic vowel space perceived as speaking at a faster tempo? It has been found that a stimulus with a moving f0-contour is perceived as faster than the same stimulus with a monotonous contour (Lehiste 1976, Kohler, 1986). If a dynamic f0 contour triggers the perception of a faster speaking rate, then a larger acoustic vowel space might have the same effect. Stimuli from 5 female German speakers with varying vowel space sizes served as input …