Individual differences in cortisol stress response predict increases in voice pitch during exam stress

- The finding that voice pitch increases under stress is often not replicated.
- We tested whether this is tied to variation in the severity of the stress response.
- Voice recordings and saliva were collected at baseline and during exam stress.
- Increases in cortisol levels predicted voice pitch under stress but not at baseline.
- Researchers examining speech under stress should control for the stress response.

Despite a long history of empirical research, the potential vocal markers of stress remain unclear. Previous studies examining speech under stress most consistently report an increase in voice pitch (the acoustic correlate of fundamental frequency, F0), however numerous studies have failed to replicate this finding. In the present study we tested the prediction that these inconsistencies are tied to variation in the severity of the stress response, wherein voice changes may be observed predominantly among individuals who show a cortisol stress response (i.e., an increase in free cortisol levels) above a critical threshold. Voice recordings and saliva samples were collected from university psychology students at baseline and again immediately prior to an oral examination. Voice recordings included both read and spontaneous speech, from which we measured mean, minimum, maximum, and the standard deviation in F0. We observed an increase in mean and minimum F0 under stress in both read and spontaneous speech, whereas maximum F0 and its standard deviation showed no systematic changes under stress. Our results confirmed that free cortisol levels increased by an average of 74% (ranging from 0 to 270%) under stress. Critically, increases in cortisol concentrations significantly predicted increases in mean F0 under stress for both speech types, but did not predict variation in F0 at baseline. On average, stress-induced increases in voice pitch occurred only when free cortisol levels more than doubled their baseline concentrations. Our results suggest that researchers examining speech under stress should control for individual differences in the magnitude of the stress response.