Dunning and Stern (1994). Distinguishing accurate from inaccurate eyewitness identifications via inquiries about decision processes

Dunning, D. & Stern, L. B. (1994). Distinguishing accurate from inaccurate eyewitness identifications via inquiries about decision processes. Journal of Personality and Social Psychology, 67, 818-835. Background Eyewitness testimony is important. In 1987 there were 78,000 trials decided based primarily on eyewitness identifications. Dunning and Stern want to find a way of postdicting eyewitness accuracy.

They base their work on Source monitoring theory. Source monitoring theory claims, among other things, that we decide if a memory is based on perception or imagination by evaluating the degree to which it contains vivid perceptual details. Accurate eyewitness IDs should involve the retrieval of perceptual details that overlap with the suspect's features.

Their basic position is that this match of perceptual details will cause accurate witnesses to make identifications automatically without having to engage in stratetic processes. Accurate eyewitnesses will thus have difficulty saying how they reached their decision and may simply say the face "Just popped out at me." Additionally because automatic processes are faster than strategic ones accurate eyewitnesses should make their decisions quickly. Inaccurate witnesses will tend to use strategic processes include relative judgment strategies (i.e. process of elimination).

Study 1

Methods

Subjects viewed a film that included a 34 second segment showing a theft of some money. They were then interviewed about the film for between 5 and 10 minutes. Afterwards they were shown a 5 member target present simultaneous photospread. Subjects were told that the perpetrator may or may not be present in the photospread and that they didn't need to pick anyone if they didn't think he was there. Think out loud protocols were collected as they made their decision.

Subjects then completed a questionnaire:

Confidence rating
8 alternative forced choice question about their decision processes
4 alternative forced choice question about the role played by the nonchosen pictures
Indicated whether the pictures or their memory had a greater influence on their choice

Studies 2-4 Study 2: Identical to Study 1 except that subjects were not told that the perpetrator may not be in the lineup.

Study 3: Manipulated response bias by telling some subject they had to be at least 70% certain before they made an identification and by telling other subjects that they had to pick someone regardless of how certain they were.

Study 4: Included both target present and target absent lineups.

Results Accurate witnesses making positive identifications were more likely to automatically recognize the culprit while inaccurate witnesses were more likely to use process of elimination and compare and contrast stategies. Consistent with this inaccurate witnesses were more likely to say that the non-chosen photos impacted their decisions.

Accurate witnesses made their identifications more quickly and were more confident in their responses than inaccurate witnesses. However differences between accurate and inaccurate witnesses persisted even when confidence was regressed out.

Study 5 Subjects were given questionnaires from the previous studies and asked to determine which witnesses were accurate and which were inaccurate. Some subjects were informed of what to look for, while others were not informed. Subjects had some ability to distinguish between the accurate and inaccurate witnesses and informed subjects did a little better than uninformed subjects. This advantage was limited to rejecting the inaccurate witnesses and did not show up for correctly accepting the accurate witnesses. Conclusions Dunning and Stern were successful in finding a number of factors that can be used to postdict the accuracy of eyewitnesses. In particular their results suggest that inaccurate witnesses fall make on strategic processes like comparing and contrasting the photos or the process of elimination. Accurate witnesses respond automatically, reporting things like, the correct picture just jumped out at them.

The authors were suprised that confidence did such a good job predicting accuracy, given that the confidence accuracy correlation has sometimes been found to be weak or non-existent. However Dunning and Stern primarily looked at cases where people made identifications and some work though suggests that the CA relationship is stronger amongst "choosers" than "nonchoosers". The authors also speculate that the concurrent verbal protocols may have allowed subjects to better calibrate their confidence judgments in much the same way that Kassin's retrospective self-awareness effect occurs.


University of Arkansas	Department of Psychology	Graduate Program in Experimental Psychology
Lampinen Lab	False Memory Reading Group	False Memory Reading Group Summer 2003