More on temporal compensation
S. M. Ellis, Gary G. Weismer · The Journal of the Acoustical Society of America · 1978
Ohala and Lyberg [J. Acoust. Soc. Am. 59, 990–992 (1976)] have suggested that the often observed inverse relationship between the durations of adjacent segments in speech production may not reflect a true organizational process. Rather, Ohala and Lyberg argue that the inverse relationship may be a measurement artifact resulting from the fact that the adjacent segments share a measurement boundary, and that any error in the placement of this boundary “forces” an inverse relationship between the segment durations. Because the notion of temporal compensation may have utility in a model of speech production, we decided to perform a simple test in an attempt to separate the systematic influence of measurement error from any “true” covariance that might exist for the durations of adjacent segments. A single speaker recorded the utterance To catch a fish is to hook it [Wright, J. Acoust. Soc. Am. 56, 1258–1265 (1974)] one hundred times; two complete sets of spectrograms were subsequently prepared so that one set could be measured with a simultaneous boundary placement for adjacent segments, whereas the other set allowed independent placement of all segment boundaries. Results suggest that Ohala and Lyberg overestimate the influence of measurement error on measures of temporal compensation.