Wells,
G.L., Olson, E.A. & Charman, S.D. (2003). Distorted retrospective
eyewitness reports as a function of feedback and delay. Journal of
Experimental Psychology: Applied, 9, 42-52.
In
this article Wells and colleagues are looking at a phenomenon they call postidentification
feedback (PIF). The basic idea is
that if a witness makes an identification and the police provide the witness
with feedback (e.g. "Good job, you got 'im") that feedback may alter
the witness's confidence and their memory for the witnessing conditions (e.g.
how good of a view they had, etc.).
This
is a robust phenomenon that has been demonstrated in a number of different
studies now (cites and links). Wells and colleagues think of this in terms of a
reconstructive memory process. They
argue that people typically don't have very good memories for what their
confidence level was at some previous time (i.e. when they made the ID) or the conditions
under which they witnessed the even. So when asked about these issues on a
questionnaire or at trial they use external sources of information, like the
feedback they got from the investigator, to infer what their confidence must
have been (e.g. "The investigator said I got the right one, I must have
been pretty confident.")
In
the article I'm summarizing Wells et al. wanted to know if the PIF effect still
occurs when longer delays are used.
There are two issues. (1) Does
feedback have to be given right after the identification for the PIF effect to
occur? (2) Does the questionnaire that
asks about confidence and so on have to been given right after the feedback?
The Study
Here's
the deal:
(1)
Participants saw a short video about a "rooftop bomber".
(2)
Participants shown a target absent six-pack.
(3)
Participants picked the person they believed to be the culprit (all
participants ended up picking someone, thus making a false identification).
(4)
They were given confirming, disconfirming or no feedback.
(5) They answered a set of questions:
·
·
·
Confidence
at the time of the identification: A single question, e.g. "At the time you
identified the person from the photo lineup, how certain were you that the
person you identified was the person you saw in the video?"
·
·
·
"Primary
Measures": These are the questions they
used in previous PIF studies. There
were nine of them altogether and they asked how clear their memory was at the
time of the identification, how difficult the decision was, and how good the
witnessing conditions were when viewing the original video.
·
·
·
Decision
processes:
Seven questions similar to the types of questions Dunning has used to try to
get at whether people are using absolute (just popped out) or relative decision
processes (compare and contrast, process of elimination).
·
·
·
Influence of
feedback: Single
question about whether feedback influenced answers to these questions.
The
experimenters varied the delay between the identification and the feedback and
the delay between the identification and the final questionnaire. In both of the feedback conditions
(Confirming/Disconfirming) the delays were set up in one of three ways.
Immediate/Immediate: The feedback is given
right after the ID and the witnessing conditions questionnaire is given in the
same session.
Immediate/Delay: The feedback is given right after the ID but
the witnessing conditions questionnaire is given 48 hrs later.
Delay/Delay: The feedback is given 48 hrs after the ID, and the
witnessing conditions questionnaire is given during the same session as the
feedback.
Of
course, in the no feedback condition, there are only two possibilities
(Immediate, Delay) resulting in a total of 8 conditions.
Major Findings
Participants
gave higher estimates of how confident they were at the time of the
identification when they received confirming feedback. Disconfirming had no effect on their
confidence estimates. They found the
same basic pattern for their "primary measures". In both cases, the PIF effect didn't seem to
be influenced in any significant way by delay.
Nor did the PIF effect depend on whether the participants believed that
the feedback influenced them.
They
asked a number of questions about the decision processes used in making their
selection. People who received
confirming feedback were more likely to say they recognized the culprit but
couldn't say why, that the photo just "popped out" at them, and that
the photo matched the image in their mind's eye. Disconfirming feedback had no significant effect, and the PIF
effect wasn't influenced by delay.
The Conclusions Drawn By
Wells et al.
As
we saw PIF didn't seem to depend on delay.
The authors acknowledge though that (1) its possible that the null
effect of delay reflected power limitations and (2) actual criminal cases may
have delays that are longer than 48 hours.
At the very least though, Wells et al. have shown that you can get big
PIF effects at moderately long delays.
They
also pointed out that, unlike their previous studies, the effect of
disconfirming feedback wasn't significant.
They say that disconfirming feedback may just have a weaker effect and
they point out analogous findings in the hindsight bias literature.
So
the applied lesson they draw is that giving a witness feedback about their
identification causes problems even when there is a delay. They make two recommendations to try to
ameliorate this problem. First, they
argue that lineups should be presented in a double blind manner. Second, they argue that a measure of
confidence should be taken right after the identification and before the
witness is told anything about the accuracy of their selection.
The
theoretical lesson they draw is that people don't have direct access to the
information being asked about in the postidentification questionnaire. Instead, Wells et al. believe that people
reconstruct answers to these questions.