Eliciting Speech Reduction in the Laboratory: Assessment of a New Experimental Method

The aim of this study was the development and assessment of a novel method for recording speech produced under varying degrees of articulatory precision in the laboratory. Three conditions were designed to elicit reduced, citation, and careful speaking styles. In the first condition, talkers read sentences while simultaneously carrying out a distractor task. In the second condition, talkers simply read the same set of sentences in the absence of any other task. In the third condition, talkers read the sentences but were frequently prompted by the computer to re-read the last sentence more carefully. Recordings from six talkers, three female and three male, were made and measurements were carried out along a number of dimensions that have been used in the literature to distinguish careful from reduced speech or that have been linked to differences in intelligibility between talkers. Using acoustic analysis, measures of word duration, sentence duration, overall sentence energy, word energy, sentence energy range, word energy range, pitch range of the sentence, and two measures of F1-F2 vowel space dispersion were obtained for all six talkers. Overall, we observed consistent effects of speaking mode across most talkers for many of the measured dimensions. Individually, there was variability between talkers as to which dimensions showed the greatest differences between conditions. The overall pattern of results indicates that these experimental procedures can be used to elicit varying degrees of articulatory control in a highly controlled laboratory environment.