Exploring the basis of confidence in recognition : psychophysical approach vs. an inferential approach to ecphoric confidence ratings

We investigated whether ecphoric confidence ratings were best accounted for by psychophysical or inferential models of metacognition. 60 participants (43 female; aged 16 to 75 years), undertook a facial recognition task. Participants saw a mix of full and partial faces at both test and study (partial faces displayed the top half of the face), and provided ecphoric confidence ratings (indexing recognition without a yes/no decision) for each face at test on a coarse-grained verbal or fine-grained probabilistic scale. Inferential models of metacognition propose that additional information at test, regardless of its diagnosticity, increases confidence. Therefore, we would expect confidence to be higher in trials where participants viewed a partial face at study (TS) followed by the corresponding full face at test (FT), than when they viewed a partial face followed by the corresponding partial face (TT). Psychophysical models, in contrast, propose that confidence indexes stimulus discriminability, and should be unaffected by additional non-diagnostic information, TS/FT = TS/TT. The doubt-scaling model argues that non-diagnostic information should decrease confidence, and therefore TS/TT > TT/FT. Linear mixed effects models supported the doubt-scaling account. However, these results must be interpreted with caution given potential limitations. Scale type did not affect results