Similarity effects on recognition : testing the diagnostic feature-detection hypothesis

Diagnostic feature-detection (DFD) hypothesis proposes that the simultaneous presentation of faces (e.g., in police lineups) enhances identification accuracy by allowing shared features to be discounted and diagnostic features to be detected. However, no direct tests of this theory have been conducted in face recognition tasks. We directly tested the DFD hypothesis. 60 participants (42 women; aged 18 to 45 years) completed a facial recognition task. Memory for faces was tested in three task conditions: a yes/no (Y/N) task, in which one face was presented at a time; a two alternative forced choice (2AFC) task, in which participants had to choose one of two faces, presented simultaneously; a two alternative open choice (2AOC) task, in which two faces were presented simultaneously, with a 'not present' option. Difficulty was manipulated through two levels of target-lure similarity (high, low). Discrimination (d') was calculated and compared between the high and low similarity trials within each condition. Contrary to the predictions of DFD, we found significant differences in discrimination between high and low similarity trials in the 2AFC and 2AOC conditions. Unexpectedly, the manipulation of difficulty did not affect performance in the Y/N condition. We found no support for the DFD hypothesis.