Invariances in the acoustic expression of emotion during speech.

An experiment was designed to test whether different individuals produce similar voice patterns when they read the same emotional passage. Quantitative scoring criteria were developed that reflect the extent to which different individuals consistently produce similar constellations of acoustic attributes in response to the same emotional context. The scoring procedure was applied to the voice tracks of standard utterances produced by 11 subjects reading 10 different emotionally evocative scripts. The results supported the hypothesis that different individuals produce standard acoustic configurations to express emotions. Because acoustic properties reflecting contrastive stress consistently varied with emotional context over syntactically and semantically identical utterances, some factor related to emotional context other than syntax or semantics must account for the variations. An evolutionary argument that emotion communication can be seen as intention communication is presented to account for these variations. Implications for theories of emotions and of intentional generative semantics are discussed.