Describing faces from memory: Accuracy and effects on subsequent recognition performance

The present research examines: (a) the accuracy of three face description methods, and (b) the effects of post-exposure description and imaging activities on subsequent face recognition performance. Participants viewed a sequence of six target photographs, and after each, performed one of three description tasks: generated their own set of descriptors, checked-off descriptors from a pre-existing list, or rated the same set of descriptors on bipolar scales. Other participants performed a distractor (control) activity. Additionally, participants were either told or noltold to image the targets while they simultaneously performed the description tasks. Results showed that the checklist task lowered subsequent recognition performance compared to the generate task. Imaging with the generate task facilitated recognition, but imaging with the checklist and rating tasks degraded recognition. The generate task produced the highest quality descriptions as determined by other participants' performance in matching the descriptions to face photographs. 1be checklist decrement is discussed in terms of memorial confusion initiated by the presence of irrelevant face cues. These results indicate that descriptor generation is the preferred method of collecting eyewitness' face descriptions. INTRODUCfION Eyewitnesses, having viewed an individual involved in a crime, are frequently asked by police investigators to participate in various memory tests, e.g., examining a lineup or a mugfile. In the intervening period prior to these tests, several things might happen: witnesses may rehearse or hold an image of the face in their mind and/or they might participate in other memory tests such as giving a verbal description or helping to produce a sketch or composite. The present research examines three types of verbal description instruments with respect to the quality of descriptions that they produce, and whether these verbal description methods (and holding an image of the target face) affects subsequent recognition performance. Why would different methods of eliciting verbal descriptions be of interest? Witnesses generally do not give many descriptors when describing faces, and the terms that they produce are frequently not very specific (e.g., 'medium nose' or 'thin lips'). These general terms are frequently inadequate to do an effective search for the culprit. One possible way to elicit better descriptions is to provide a list of possible adjective descriptors which they can select from-as opposed to generating the descriptors themselves. There has not been much empirical work directly comparing different description techniques. One exception is by Goulding (1971) who had police officers make cued or free recall descriptions of target faces. Free recall produced better quality descriptions than cued recall. In the present study, this effect is re-examined and includes ratings as another kind of cued method. Another issue is whether the description task influences subsequent recognition. Previous research on this topic is equivocal. Verbal description has been reported to degrade, facilitate, and have no effect on subsequent recognition performance. However, virtually all of the previous studies used different methods of eliciting the descriptions. The exception is Wogalter (1991) who compared two description