Confidence judgements for detecting and discriminating changes in emotional facial expressions

The ability to reflect on our own performance when we have to judge subtle changes in facial expressions might have a strong influence on our future social communication. It is still unknown to what extent introspection can access the accuracy of these judgements. Here, we aimed to investigate how confidence judgments map onto performance for detecting and discriminating emotional facial expressions. We sequentially presented two faces with the same identity, but with two different morphed levels of intensity of the same emotional expression, one corresponding to a nominal ‘pedestal’ intensity (0% for detection, or 75% for discrimination), and the other corresponding to the pedestal plus one of seven intensity increments. At the end of each trial, we asked 98 participants to report an intensity judgement (‘which face had the stronger expression intensity?’) and also a confidence judgement (‘how confident are you that your intensity judgement was correct?’). On average, intensity thresholds (our performance measure) for detection and discrimination did not differ, corresponding to about 18% intensity increment. However, the psychometric function was shallower for discrimination, implying that participants could discriminate subtle intensity increments better than they could detect them. Confidence judgements revealed that participants hugely overestimated their ability to perceive small intensity increments, for which their performance was in fact at chance. This overconfidence was more pronounced for intensity discrimination than for detection, probably because, in the former, but not in the latter condition, the pedestal reduces uncertainty about which expression is being presented on a given trial. Moreover, the difference between confidence and performance thresholds correlated negatively with performance thresholds, indicating that poor performers were more aware of how they performed than good performers. This is the opposite pattern of results to that observed in the ‘Dunning-Kruger’ effect and it might extend to other domains of sensory discrimination.