What makes a face photo a ‘good likeness’?

Photographs of people are commonly said to be ‘good likenesses’ or ‘poor likenesses’, and this is a concept thatwe readily understand. Despite this, there has been no systematic investigation of what makes an image a goodlikeness, or of which cognitive processes are involved in making such a judgement. In three experiments, weinvestigate likeness judgements for different types of images: natural images of film stars (Experiment 1), imagesof film stars from specific films (Experiment 2), and iconic images and face averages (Experiment 3). In all threeexperiments, participants rated images for likeness and completed speeded name verification tasks. We consistentlyshow that participants are faster to identify images which they have previously rated as a good likenesscompared to a poor likeness. We also consistently show that the more familiar we are with someone, the higherlikeness rating we give to all images of them. A key finding is that our perception of likeness is idiosyncratic(Experiments 1 and 2), and can be tied to our specific experience of each individual (Experiment 2). We arguethat likeness judgements require a comparison between the stimulus and our own representation of the person,and that this representation differs according to our prior experience with that individual. This has theoreticalimplications for our understanding of how we represent familiar people, and practical implications for how wego about selecting images for identity purposes such as photo-ID.