A COMPARISON OF DIRECT-MAGNITUDE ESTIMATION AND EQUAL-APPEARING INTERVAL SCALING OF VOWEL ROUGHNESS

The term "vocal roughness" generally refers to a perceived voice quality that is detected, defined, and described on the basis of a listener's auditory impressions. Such listener perceptions may be quantified in research by any of several available psychophysical scaling methods.;Ten adult females sustained phonation of the vowels /a/ and /i/ in a normal voice, while simulating moderately rough phonation, and while simulating extremely rough phonation. Twenty listeners then rated the roughness of each vowel sample using DME and EAI scaling. The resulting EAI and DME ratings were then compared to study the type of perceptual continuum on which vowel roughness perceptions seem to lie.;Each vowel sample was also analyzed acoustically to obtain its narrow bandwidth (3 Hz) spectrum. The relationship between spectral noise level (SNL) measurements and vowel roughness ratings obtained by each of the rating methods studied was then examined. Additionally, the intra- and interjudge reliability of the EAI and DME roughness ratings was assessed using several statistical procedures: exact agreement, Pearson correlation, Spearman correlation, Kendall coefficient of concordance, and intraclass correlation.;The findings indicated that the relationship between DME and EAI roughness ratings was more strongly curvilinear than linear. Thus it appeared that listener perceptions of vowel roughness may be represented better by a prothetic than by a metathetic continuum. It was also found that the relationship between vowel SNL measurements and roughness ratings varied with the vowel rated and the psychophysical method used to obtain ratings. Specifically, the relationship between SNL measurements and EAI ratings was curvilinear for ratings of /a/ samples, but linear for ratings of /i/ samples.;Two commonly-used psychophysical scaling methods, direct magnitude estimation (DME) and equal appearing intervals (EAI), were applied in this study in an effort to learn whether one or the other was more appropriate for rating vowel roughness.;The listeners manifested approximately the same degree of reliability in using DME scaling to rate vowel roughness as they did in using EAI scaling. The findings suggested that the "appropriateness" of the statistical procedure selected to estimate roughness-rating reliability may be to a considerable extent dependent on the purpose for which the ratings are obtained.