Peter E. Morris

TL;DR: A straightforward guide to understanding, selecting, calculating, and interpreting effect sizes for many types of data and to methods for calculating effect size confidence intervals and power analysis is provided.

Abstract: RESPONSE BIAS WITH PROTOTYPIC FACES Kennech R. Laughery and Dean G. Jensen Department of Psychology, Rice University Houston, Texas, U.S.A. Michael S. Wogalcer Department of Psychology, University of Richmond Richmond, Virginia, U.S.A. Three experiments employed a facial recognition task where target faces did not appear in a five or six-item test sec. The test set consisted of a prototype face that differed by one feature from each of the distractors, which in turn differed by two features from each other. Subjects ranked the faces on the likelihood they were a target face. Results showed the prototype was ranked significantly above chance, indicating the procedure resulted in a response bias. The findings have implications for lineup and photospread construction. This paper presents the results of three experiments that addressed the issue of response bias in a facial recognition task. This issue concerns the situation in which the recognition test set contains a face that is a prototype of the others in the set. For example, suppose the set is made up of faces each of which is a variation on one particular face, the prototype, that is also included in the set. One might expect that such a set would result in a response bias favoring the identification of the prototype face. There is some support for this prototype notion in a study by Solso and McCarthy (1981). Using a recognition memory paradigm, they constructed distractor faces from the features of faces that hed been presented. Their subjects were more confident in recognizing (incorrectly} the distractor faces than the faces they had seen. Additionally, Wogalter and Jensen (1986) have demonstrated a bias towards a prototype in a recognition task using nonfacial stimuli. The practical issue with which this work is concerned is fairness, or its opposite -bias, in law enforcement lineups and photospreads. A biased lineup is one where persons who were not witnesses to a crime are IIIOre likely than chance to pick the suspect, Malpass and Devine (1983) and Wells (1978) have discussed lineup bias and have noted that the suspect must not be distinctive in comparison with other members (the distractors). Malpass and Devine (1983) reported an experiment in which similarity between a suspect photograph and the other photospread members was manipulated. Their results showed an increase in fairness with increasing suspect-distractor similarity. Hence, in constructing lineups or photospreads, law enforcement agencies would be advised : o s~iec c ii neup ~r pno cospread d i scractors who are similar co the susp ect. But such an a ppro ach creates th e possibilit y of bias due to th~ suspect bei ng a pr ot otype of the lineu p or phocospread faces. That is, che suspect may have more features in common with th e di s cr ac tors chan the di stract o rs 9hare with each other. In this reg ar d the suspect may be distinctive, and in situation s where the sus pec c i s not the target per son (criminal ) , th e Lineup or phocos pread may fail a cru c ial cr it erion in that the likelihood of the suspecc being chose n is greater than chance. The · pr es en t experiments employ a recognition paradigm. In Experiment l subjects saw a s ingle target face before examining a gr oup of photographs -a photospread. ln Experiments 2 and 3, sub je ct s saw a large number of faces, and then examined a series of phoco spre ads. ln Experiment l the target face appeared in some of the photospre ads, wher eas in Experiments 2 and 3 it did not. Each of che phot ospreads was made up of a prototype face (not the targ~t ) and dist r actors that were more similar to th e prototype face than th e y were to each othe r . The hypothesis is that the likelihood is greater than chance that the prototype will be identified as a target fa ce. In additi on t o differences in some procedural details, th e experiments al so differed in the stimulus materials used: identi-Kit faces, Mac -a-Mug Pro faces and photographs of real faces. A sample Mac-a-Mug photospread is shown in Figure l.