An analysis of strategies in remembering and generating faces



Journal Title

Journal ISSN

Volume Title



This thesis explored the strategies people use in carrying out a facial recall task. Two major analytic efforts are presented. The first analysis examined strategies "witness" subjects reported at completion of the facial recall task. These strategies recounted the techniques witness subjects had used to help them remember the "target" subjects. A nominal cluster analytic technique based on the unconstrained method of sorting objects was developed. This technique applied to 27 sorters' categorizations of 121 witness responses was used to identify two possible organizations of witness strategies for preparing to carry out a facial recall task. The first organization identifies three strategy tjHpes, and the second organization collapses two of these three leaving two types. An analysis of thezfit between the obtained cluster organization and the individual sorters' categorization indicated that the three strategy types accounted for 54% of the variance across sorters. Further analysis showed the relationship between strategy type and quality of facial image to be significant though small. Strategy type accounted for five percent of the variance in image quality. The second analysis examined properties of the time and order offeature description during the image generation process. Particular emphasis was placed on identifying device dependent and device independentcomponents of image generation using the Identi-kit or the sketch technique. Results indicate that these devices can be described by different information processing characteristics. An analysis of certain molarproperties of image generation suggests that witnesses using the sketch technique provide more information about the target subject than witnesses using the Identi-kit. More fine grained analysis suggests that for the Identi-kit witnesses loss of information about the target is linearly related, to the time of description and memory load; i.e., number of features not yet described since beginning description for users of the Identi-kit. No comparable linear relationship was found for users of the sketch technique.