As we recognise people we know over many years, their faces can change, sometimes profoundly, and yet we continue to recognise them with ease. How do we update our representations over time? We present four pre-registered experiments to examine this. In Experiment 1, using likeness ratings and speeded name verification, fans of a long-running TV soap opera demonstrated that their representations of the characters' faces were weighted towards their most recent encounters - when the characters were oldest. While we initially hypothesised that this was due to recency, Experiment 2 showed this not to be the case. When new participants were taught these characters either in chronological or reverse-chronological order they all demonstrated representations weighted towards the characters at their oldest ages, regardless of the order in which they had encountered them. We ruled out potentially artefactual explanations using statistical analysis of the images themselves and, in Experiment 3, restricted learning sets. A further, final experiment showed that our results are unlikely to be fully explained by perceived distinctiveness of the stimuli. We conclude that the processes involved in developing representations for familiar people are more sophisticated than previously thought, incorporating real-world constraints, including natural chronology.
Laurence et al. (Tue,) studied this question.