Training methods for facial image comparison: a literature review

Anne Hillstrom, James Sauer, Lorraine Hope

Research output: Working paper

84 Downloads (Pure)

Abstract

This literature review was commissioned to explore the psychological literature relating to facial image comparison with a particular emphasis on whether individuals can be trained to improve performance on this task. Surprisingly few studies have addressed this question directly. As a consequence, this review has been extended to cover training of face recognition and training of different kinds of perceptual comparisons where we are of the opinion that the methodologies or findings of such studies are informative. The majority of studies of face processing have examined face recognition, which relies heavily on memory. This may be memory for a face that was learned recently (e.g. minutes or hours previously) or for a face learned longer ago, perhaps after many exposures (e.g. friends, family members, celebrities). Successful face recognition, irrespective of the type of face, relies on the ability to retrieve the to-berecognised face from long-term memory. This memory is then compared to the physically present image to reach a recognition decision. In contrast, in face matching task two physical representations of a face (live, photographs, movies) are compared and so long-term memory is not involved. Because the comparison is between two present stimuli rather than between a present stimulus and a memory, one might expect that face matching, even if not an easy task, would be easier to do and easier to learn than face recognition. In support of this, there is evidence that judgment tasks where a presented stimulus must be judged by a remembered standard are generally more cognitively demanding than judgments that require comparing two presented stimuli Davies & Parasuraman, 1982; Parasuraman & Davies, 1977; Warm and Dember, 1998). Is there enough overlap between face recognition and matching that it is useful to look at the literature recognition? No study has directly compared face recognition and face matching, so we turn to research in which people decided whether two non-face stimuli were the same or different. In these studies, accuracy of comparison is not always better when the comparator is present than when it is remembered. Further, all perceptual factors that were found to affect comparisons of simultaneously presented objects also affected comparisons of successively presented objects in qualitatively the same way. Those studies involved judgments about colour (Newhall, Burnham & Clark, 1957; Romero, Hita & Del Barco, 1986), and shape (Larsen, McIlhagga & Bundesen, 1999; Lawson, Bülthoff & Dumbell, 2003; Quinlan, 1995). Although one must be cautious in generalising from studies of object processing to studies of face processing (see, e.g., section comparing face processing to object processing), from these kinds of studies there is no evidence to suggest that there are qualitative differences in the perceptual aspects of how recognition and matching are done. As a result, this review will include studies of face recognition skill as well as face matching skill. The distinction between face recognition involving memory and face matching not involving memory is clouded in many recognition studies which require observers to decide which of many presented faces matches a remembered face (e.g., eyewitness studies). And of course there are other forensic face-matching tasks that will require comparison to both presented and remembered comparators (e.g., deciding whether any person in a video showing a crowd is the target person). For this reason, too, we choose to include studies of face recognition as well as face matching in our review.
Original languageEnglish
PublisherStationery Office
Number of pages68
Publication statusPublished - Mar 2011

Fingerprint

Dive into the research topics of 'Training methods for facial image comparison: a literature review'. Together they form a unique fingerprint.

Cite this