Linear and nonlinear analysis of the stability of gestural organization in speech movement sequences

Recent procedures have been developed that allow the analysis of gestural stability across repetitions of phrase-length utterances by linearly normalizing aspects of the articulatory signal. This process produces an index of variability called spatiotemporal index, or STI (Smith, Goffman, Zelaznik, Ying, & McGillem, 1995). Consistent findings that different STIs underlie changes in speaking rates in normally speaking adults have been found in subsequent studies by Smith and colleagues. However some researchers have raised concerns that linearly normalized data do not adequately account for the nonlinear aspects in the articulatory signal (Lucero, Munhall, Gracco, & Ramsey, 1997). The present study compared findings from linear and nonlinear normalization procedures in the analysis of lower-lip displacement of phrase-length utterances for a group of 8 speakers and across three rates. Findings indicated that at a group level, gestural stability, although higher for each rate contingency, was similar to that found in earlier STI studies. However, variability was greater, and 4 of the 8 subjects failed to consistently demonstrate greater stability at habitual rate, followed by fast and then slow rate. A nonlinearly normalized analysis of the same data produced significantly lower stability indices, and variability was also reduced. It is argued that a nonlinear normalization procedure based on lower-lip displacement holds advantages in the analysis of phrase-length speech data over both linear and alternative nonlinear normalization techniques.