Telling self-defining memories: An acoustic study of natural emotional speech productions

Vocal cues in emotion encoding are rarely studied based on real-life, naturalistic emotional speech. In the present study, 20 speakers (10 male, 10 female) aged 25 to 35 were recorded while orally telling 5 successive self-defining autobiographic memories (SDM). By definition, this task is highly emotional, although emotional load and emotion regulation are expected to vary across SDM. Seven acoustic parameters were extracted: MeanF0, MedianFo, StandardDeviationF0, MinF0, MaxF0, Duration and SpeechRate. All SDM were manually transcribed, then their emotional lexicon was analysed using Emotaix. First, speech productions were examined in reference with SDM characteristics (specificity, integrative meaning and affective valence) as determined by 3 independent investigators. Results showed that overall the speech parameters did not change over the time course of the experiment, or as a function of integrative meaning. Specific memories were recounted at a higher speech rate and at greater length than non specific ones. SDM with positive affective valence were shorter and included less variability in fundamental frequency than negative SDM. Second, emotionally-charged (positive vs. negative; high vs. low arousal) vs. emotionally-neutral utterances as to Emotaix classification were compared over all SDM. Only a few significant effects were observed, which led us to discuss the role of emotion regulation in the SDM task.