Acoustic Effects of Authentic and Acted Distress on Fundamental Frequency and Vowel Quality

This study investigates variation in fundamental frequency and vowel quality from speech produced in distress and non-distress conditions by victims of genuine attacks in emergency service forensic recordings and by trained actors. Recordings from two criminal cases involving violent attack were compared with recordings from four actors reenacting the forensic scenarios. Non-distress speech was available from the victims by chance, and from the actors by design. Findings show that: a) F0 increases and becomes more variable in distress conditions by both actors and victims; b) all speakers show a target undershoot of /i:/ realisations, though for actors this is characterized by a more open production (increase in F1) whereas in victims a retraction (a decrease in F2) is observed; and c) /ə/ productions appear more complex and harder to interpret, nevertheless two tentative trends one in the form of a peripheralisation of the nuclei and the other in the form of a centralisation of the offglides may be observed across speakers.