Eyewitness confidence: social influence and belief perseverance

Two studies investigated changes in eyewitness identification confidence after an identification has been made. In Experiment 1, thefts were staged 70 times for pairs of unsuspecting eyewitnesses (total n = 140). Before campus security arrived, witnesses were separated and attempted identifications of the thief from a target-absent photospread. Biased instructions were used to induce false identifications. The 96% who made false identifications were then randomly assigned to 1 of 9 conditions telling them of the alleged identification decision of their co-witness. Witnesses were told that (a) the co-witness identified the same person, (b) the co-witness identified a different but plausible other person, (c) the co-witness identified an implausibly different person, (d) the co-witness rejected the photospread, or (e) the witness was told nothing about the co-witness\u27s decision. In addition, some witnesses who were told that the co-witness identified the same person were later told that the information was incorrect and that the co-witness had actually identified a different person or were told that it was not known who the co-witness identified. As well, some witnesses who were told that the co-witness had identified a different person were later told that the information was incorrect and that the other witness had actually identified the same person or were told that it was not known who the co-witness identified. Following the co-witness information manipulation, a campus security officer questioned witnesses about their identifications of the thief. Compared to the no-information control condition, co-witness agreement produced a robust inflation of certainty whereas co-witness disagreement produced a precipitous decline in certainty. With one exception, correcting the co-witness information did not eliminate the confidence-inflating and deflating effects of the original information, indicating a strong perseverance effect. In the second experiment, subject-jurors (n = 378) viewed the testimony videotapes and evaluated the credibility of the witnesses. Subject-jurors\u27 ratings of perceived credibility were influenced by the type of information witnesses had received in a pattern that generally paralleled the results of Experiment 1