Factor Analysis of Voice Perception

Abstract

Thirty scales from the semantic differential were utilized to investigate the dimensions underlying voice perception in 10 normal male and 10 normal female speakers. Nineteen observers rated the concepts of normal and normal voice quality in addition to a sentence read by the 20 speakers. A principal components analysis with varimax rotation was performed on the data. The results revealed five factors, accounting for approximately 63% of the variance. Factor I accounted for 20% of the variance and had high loadings on such scales as passive/active, still/vibrant, and vigorous/ feeble. Factor II accounted for 18% of the variance and loaded on pure/impure, smooth/rough, and choked/free. Factor III accounted for 11% of the variance and loaded on thick/thin, dark/bright, and heavy/light. Factors IV and V each accounted for 7% of the variance and loaded on scales such as youthful/mature, new/old (Factor IV), dry/wet, and cold/ hot (Factor V). The D metric was utilized to portray the distances between the concepts normal and normal voice quality and the 20 speakers. Further analysis is currently in progress utilizing 154 additional observers.