Hair today, gone tomorrow: holistic processing of facial-composite images

Abstract

Faces are processed as complete entities, with individual features and their interrelations interacting perceptually with each other. These holistic face-processing effects have been observed for photographs of faces and, more recently, for facial composites, pictures of faces normally constructed from memory. In the current work, we explore the impact of reducing the likeness of a single exterior feature, hair, on naming of facial composites. Our expectation, that such hair manipulations would severely interfere with (holistic) face-recognition, was supported. In the first experiment, using identities of celebrities and footballers, hair was exchanged between composites constructed using one of three types of face-production system: feature, EvoFIT and sketch. Correct naming of composites with manipulated hair was greatly impaired relative to veridical composites; this effect applied equally to the three systems tested. In Experiment 2, using composites of celebrities, changes made to either hair length or style substantially-decreased correct naming, while changes in shade led to more-subtle effects. Two such changes applied to the same composite-face led to worst naming: best naming emerged for composites presented with hair from the actual target-photograph. Across both experiments, reducing the likeness of a composite’s hair promoted worse naming than naming of the internal-features region—the part usually responsible for familiar-face recognition. It was also found that good-quality internal features lessened the negative impact of changes made to hair, and vice versa. The research underscores the importance of rendering accurate hair in composites, and a plausible mechanism was proposed for increasing the efficacy of this facial feature.