Vlam Corner Vowel Stimuli

Speech signals vary substantially in a number of their key properties, with the variability deriving from, among other things, talkers’ age and gender. Speech processing requires resolution of this variation, necessitating interpretation of age and gender information in the signal. In some signals, the age and gender are not clear from acoustic information alone. In these cases there may be substantial individual variation in judgments of age and gender. In this study, we examine the interplay between the interpretation of age and gender across language communities. VLAM CORNER VOWEL STIMULI The Variable Linear Articulatory Model (VLAM, Boë and Maeda, 1997) is a computational model of the articulatory system and its speech production capacities. Midsagittal representations, such as those depicted in Figure 1 (bottom), are wrought by configuring “articulatory blocks” (Maeda, 1990, 1991) corresponding to jaw height, tongue body position, tongue dorsum position, tongue apex position, lip protrusion, lip height, and larynx height. The VLAM is age-varying and capable of representing vocal tract lengths ranging from those of infants to young adults, calibrated to age based on Beck (1996). Given an age in years, the set of all articulatory configurations of the VLAM at that age that do not result in occlusion of the oral cavity yield a corresponding maximal vowel space (Boë et al., 1989) for that age. Corner vowel stimuli ([i], [u], [a]) were generated by the VLAM set at seven different ages: 6 months, 2, 4, 5, 10, 16, and 21 years, and are indexed numerically as 1, 6, and 36, respectively, in each of the maximal vowel spaces pictured in Figure 1 (top). For each age, the corner vowel stimuli were part of a set of 38 “prototype” vowel stimuli that were presented for a vowel categorization task (Ménard et al., 2009; Munson et al., 2010).