Dissociation between recognition and detection advantage for facial expressions: a meta-analysis

Happy facial expressions are recognized faster and more accurately than other expressions in categorization tasks, whereas detection in visual search tasks is widely believed to be faster for angry than happy faces. We used meta-analytic techniques for resolving this categorization versus detection advantage discrepancy for positive versus negative facial expressions. Effect sizes were computed on the basis of the r statistic for a total of 34 recognition studies with 3,561 participants and 37 visual search studies with 2,455 participants, yielding a total of 41 effect sizes for recognition accuracy, 25 for recognition speed, and 125 for visual search speed. Random effects meta-analysis was conducted to estimate effect sizes at population level. For recognition tasks, an advantage in recognition accuracy and speed for happy expressions was found for all stimulus types. In contrast, for visual search tasks, moderator analysis revealed that a happy face detection advantage was restricted to photographic faces, whereas a clear angry face advantage was found for schematic and "smiley" faces. Robust detection advantage for nonhappy faces was observed even when stimulus emotionality was distorted by inversion or rearrangement of the facial features, suggesting that visual features primarily drive the search. We conclude that the recognition advantage for happy faces is a genuine phenomenon related to processing of facial expression category and affective valence. In contrast, detection advantages toward either happy (photographic stimuli) or nonhappy (schematic) faces is contingent on visual stimulus features rather than facial expression, and may not involve categorical or affective processing.