How we are not equally competent for discriminating acted from spontaneous expressive speech

Abstract

This paper reports how acted vs. spontaneous expressive speech can be discriminated by human listeners, with various performances depending on the listener (in line with preliminary results for amusement by [3]). The perceptive material was taken from the Sound Teacher/E-Wiz corpus [1], for 4 French-speaking actors trapped in spontaneous expressive monoword utterances, and then acting immediately after, in an acting protocol supposed to be a very convenient for them. Pairs of acted vs. spontaneous stimuli, expressing affective states related to anxiety, irritation and satisfaction, were rated by 33 native French listeners in audio-only, visualonly and audiovisual conditions. In visual-only condition, 70% of listeners were able to identify acted vs. spontaneous pairs over chance level, for 78% in audio-only condition and up to 85% in audio-visual condition. Globally, a highly significant subject effect confirms the hypothesis of a varied affective competence for separating involuntary vs. simulated affects [2]. One feature used by listeners in the acoustic task of discrimination can be the perceived emotional intensity, in accordance with the measurement of this intensity level for the same stimuli from a previous perception experiment by Laukka and al.[9].