Face to face with emotion: Holistic face processing is modulated by emotional state

Abstract

Negative emotions are linked with a local, rather than global, visual processing style, which may preferentially facilitate feature-based, relative to holistic, processing mechanisms. Because faces are typically processed holistically, and because social contexts are prime elicitors of emotions, we examined whether negative emotions decrease holistic processing of faces. We induced positive, negative, or neutral emotions via film clips and measured holistic processing before and after the induction: participants made judgements about cued parts of chimeric faces, and holistic processing was indexed by the interference caused by task-irrelevant face parts. Emotional state significantly modulated face-processing style, with the negative emotion induction leading to decreased holistic processing. Furthermore, self-reported change in emotional state correlated with changes in holistic processing. These results contrast with general assumptions that holistic processing of faces is automatic and immune to outside influences, and they illustrate emotion's power to modulate socially relevant aspects of visual perception.

Keywords:

• Face recognition
• Emotional state
• Global processing
• Local processing
• Holistic processing

ACKNOWLEDGEMENTS

We thank Steve Most, Isabel Gauthier, and two anonymous reviewers for helpful comments on an earlier version of this manuscript and Kuba Glazek for help with data collection.

Notes

¹In the partial composite design, the task-irrelevant face parts are always different across the images to be matched. Often, only trials where the task-relevant parts are the same are analysed. The key comparison in such analyses is between performance on trials where the task-irrelevant parts are aligned with the task-relevant part and those where they are misaligned.

²These relationships did not reach significance when the analyses were performed on data from participants in the negative or positive emotion induction groups separately (p>.05). The loss of power due to the reduced sample size and the truncated range of change in reported emotion likely contributed to this.

³Of the residents in the same zip-code region as Temple University, 84.5% report themselves to be from a non-Caucasian racial category (US Census, 2000).