Deciding whether two pictures depict the same unknown person is surprisingly difficult. The work presented here explored individual differences in the task and potential methods of training people to do the task better. A set of stimuli were developed that include age changes as well as photographic differences. A cross section of people were tested to identify people particularly skilled at face matching. The skilled matchers were interviewed and their eye movements were measured. We compared a training method where those eye movements were used to model effective inspection of the faces with more traditional training methods. Learning was tested immediately after the training, a week later and three weeks later. Effectiveness of training was tested using faces out of the same corpus and faces from a different, easier, corpus. Results showed that inspection modelling was most effective at improving perforrmance in the easy corpus. No training improved comparisons that were difficult. There were negligible differences in eye movements between those with high accuracy and those with average accuracy. Demographic characteristics that were related to skill will be presented, as will the kinds of strategies used.
Meeting abstract presented at VSS 2014