Semmler, C.,
Brewer, N., & Wells, G. L. (2004). Effects of postidentification feedback on eyewitness identification
and nonidentification confidence.
Journal of Applied Psychology, 89,
334-346.
Eyewitness confidence is considered by
most people, especially jurors, to be an appropriate measure of accuracy. To a juror, the more confident an eyewitness
is the more believable they are.
Experimental studies and forensic exonerations have shown that many
wrongful convictions result from mistaken identifications. It is important we understand how eyewitness
confidence is influenced. The two
experiments here addressed several issues that have not been examined. It is not known if the postidentification
feedback effect exists beyond the target-absent lineup with biased
instructions. Experiment 1 employs
target-absent lineups only and Experiment 2 employs target-present lineups
only. The authors hypothesized that the
PIF effect would occur under unbiased instructions. The purpose was not to see if the effect was
stronger or weaker with unbiased instructions but if it occurred at all. Another issue was whether the confidence
inflation happened for all of the possible identification responses in
target-absent and target-present lineups.
The authors’ hypothesis was that the effect would be greater for
mistaken identifications than correct rejections in target-absent lineups. For target-present lineups, the authors
predicted that the effect would be greater for incorrect identifications and
incorrect rejections than in correct identifications. The third issue that was addressed was the
effect on retrospective and current confidence and the fourth issue was the
exploration of feedback on the confidence-accuracy relationship.
Experiment 1
208 participants were told they would
watch a short film and to pay close attention to the people in the film as they
would be asked questions about them at a later time. After viewing the video of a thief entering a
restaurant and stealing a credit card left on a counter, the participants were
seated in individual cubicles and worked on picture puzzles for 15
minutes. After selecting a suspect from
the lineup, participants in the feedback condition were prompted with a screen
that read, “This study has now had a total of 87 participants, 84 of them made
the same decision as you!” Participants
in the no feedback condition were told to wait while the next screen was
loading. All participants were then asked to rate their retrospective
confidence and their current confidence on an 11-point scale.
Results
A high
proportion of participants correctly rejected the target-absent lineup and
there were no identification rate differences in the feedback and no-feedback
conditions. A 2x2 between groups ANOVA
on retrospective confidence showed a significant main effect on feedback. The feedback x identification response effect
was not significant. There was a
significant identification response main effect that reflected lower confidence
levels for mistaken identifications than for correct rejections. The PIF effect occurred with a target-absent
lineup, which is consistent with previous research and it occurred for both
confidence measures, both identification response categories and with unbiased
instructions.
Experiment 2
The
procedures in Experiment 2 were the same as Experiment 1 except they used a target-present lineup.
Results
2x3
between-groups ANOVA on retrospective confidence found a main effect on
feedback. There was not a significant
feedback x identification response interaction and there was a main effect of
identification response on confidence.
An ANOVA on current confidence also had a main effect for feedback and
again the feedback x identification response interaction was not
significant.
General Discussion
There
were four major findings in this study:
Also, the
confidence-accuracy correlation was unaffected by feedback, using the O/U
statistic showed an increase in overconfidence (both current and retrospective)
in the feedback condition. The reference
point for confidence may control the magnitude of the confidence
inflation. There are several
implications to be taken from this study.
The fact that confidence inflation occurs even when witnesses are told
the suspect might not be in the lineup and occurred in all response categories
poses a problem. Also, the existence of
inflation effects when people pick incorrect fillers or incorrectly reject
lineups poses another problem. And
finally, this study points out that retrospective versus current confidence
influences the confidence inflation but there is no standard way in the real
world to ask these questions.