The responses of subjects in six face recognition memory studies were analyzed to determine whether false alarms (F As) ~ distractors were randomly distributed. Evidence from every study strongly supports the concluslOn that F As are nonrandom. For reasons still unclear some faces are often mistakenly selected as targets, others are never selected. Possible explanatio~s of these findings are discussed.In laboratory studies, observers' recognition memory performance for faces following a single brief exposure is typically characterized by fairly large numbers of correct identifications (hits), smaller but consistent numbers of errors of commission (false alarms), and errors of omission (misses). It is more than likely that both kinds of errors occur in criminal investigations, but for obvious reasons false alarms (FAs)-misidentifying innocent individuals-are more newsworthy when discovered and much more troublesome than misses. Nonsystematic observations by at least two investigators (Cross, Cross, & Daly, 1971; Goldstein, in press) suggest that some distractor faces "attract" more than their statistical share of F As. The single purpose of this research was to discover whether F As are distributed randomly among distractors in recognition memory experiments.Both theoretical and practical advantages would be derived from more detailed knowledge regarding responses to visual distractors. If, in fact, occasional distract or stimuli attract erroneous responses in a nonrandom fashion, and this kind of response bias could be shown to occur with some generality, it might afford a method which could be used to discover what is the "content" of the stored information carried by a subject who had viewed target faces in an earlier study session. That is, if F As are not distributed randomly, then faces erroneously selected as targets may upon appropriate analysis tell us something about the stimulus the subject was "looking for" in the test session. Moreover, nonrandom responses to distractors might offer a solution to the puzzling fact that on occasion several eyewitnesses in a criminal case have identified the same innocent person as the culprit (Wall, 1965).
METHODFive investigations were conducted using standard procedures for testing recognition memory . Racially mixed target faces were presented rust in a study session, then in an immediately following test session interspersed among distractor faces. In both sessions, faces were presented singly; in the test trials the subject had to decide for each stimulus whether it was a target face (old) or a distractor face (new). Face stimuli were achromatic full-face portraits obtained from high school yearbooks and placed on slides. All photographs were selected to reduce nonfacial cues for recognition. Subjects were white college students. Number of targets, number of distractors, length of study-test interval, and task instructions were varied to assess their effect on the distributions of F As. These procedural variations will be described below, where appropriate. All ...