did not ask attorneys and judges to predict the outcomes of laboratory experi-
ments. Instead, we examined their courtroom behaviors (attorneys ’ arguments
and judges ’ instructions) and converted the apparent logic underlying these
behaviors into specifi c predictions. That is, people who engage in the described
courtroom behaviors should make the following predictions about the out-
comes of controlled, laboratory experiments. We refer to the implicit theory
that underlies these courtroom arguments and jury instructions as the
126
Handbook of Psychology of Investigative Interviewing
‘ Courtroom Theory ’ of memory. What, specifi cally, are the predictions of the
Courtroom Theory?
1. Consistency of recollection is a direct indicator of the quality of a witness ’ s
memory. Therefore, factors that infl uence consistency of recollection
should have the same effect on accuracy of recollection. Experimental
factors that increase (decrease) consistency should increase (decrease)
accuracy and vice versa. No experimental factors should dissociate (have
different effects on) consistency and accuracy.
2. Contradictory statements should be inaccurate – at least as compared to
consistent statements.
3. Witnesses who make many contradictory statements should be consider-
ably less accurate overall (across their entire testimony) than witnesses who
make few or no contradictory statements. Statistically, there should be a
strong, negative correlation between amount of inconsistency in a wit-
ness ’ s testimony and the overall accuracy of the witness ’ s testimony.
4. Reminiscence should occur infrequently. Moreover, explaining reminis-
cence requires an extraordinary (non - cognitive) mechanism, such as police
informing witnesses about crime details.
5. Reminiscent statements should be inaccurate – at least as compared to
consistent statements.
6. Witnesses who make many reminiscent statements should be considerably
less accurate overall (across their entire testimony) than witnesses who
make few or no reminiscent statements. Statistically, there should be a
strong, negative correlation between amount of reminiscence in a witness ’ s
testimony and the overall accuracy of the witness ’ s testimony.
As opposed to the Courtroom Theory of memory, how would cognitive
psychology account for these witness behaviors?
Cognitive Theory
Two principles of cognitive theory are responsible for the major distinctions
between the predictions of the Courtroom Theory and Cognitive Theory: the
importance of retrieval processes; and the independence of components. We
describe these principles briefl y.
Retrieval p rocesses
Recollection refl ects not only the contents of the memory store but also the
process of retrieval (Tulving, 1983 ). If the retrieval processes applied on two
occasions differ, the recollections will differ, even if the contents of memory
do not change. The retrieval process is partially determined by the specifi c
The Relation between Consistency and Accuracy of Eyewitness Testimony 127
question that is asked. Thus, if the question that is posed to a witness changes
from one interview to another, the witness ’ s recollections may change. In
general, the more different are the retrieval cues (questions) across interviews,
the more dissimilar will be the recollections on the two interviews. Reminiscence
may occur, therefore, if a retrieval cue is present on the second interview, but
not on the fi rst interview. The amount of reminiscence should refl ect the
amount of dissimilarity between the retrieval cues (questions) on two
interviews.
Independence of c omponents
Complex events are made up of many components, each of which is processed
somewhat independently of the others (Fisher, Phillips & Krioukova, 2000 ;
Mitchell, Haw & Fisher, 2003 ). Therefore, if a witness fails to recall one
component of a crime, or even if she misperceives or mistakenly recalls one
component of the crime, she may still perceive or correctly recall other com-
ponents of the crime.
Based on these two principles, cognitive psychology predicts the following
behaviours by witnesses who are interviewed repeatedly:
1. Some mental processes underpin measures of both consistency and accu-
racy. For instance, the quality of the memory trace should infl uence both
consistency and accuracy. Better encoded events should be recalled more
consistently and also more accurately than poorly encoded events. By
comparison, other mental processes do not underpin both consistency
and accuracy: Either they have opposite effects on consistency and
accuracy or they infl uence one measure but not the other. For instance,
the similarity of the retrieval cues used across two interviews should infl u-
ence consistency of recollection, but not accuracy of recollection. We
should, therefore, expect that some experimental manipulations will have
similar effects on consistency and accuracy, whereas other manipulations
will have different effects on consistency and accuracy (experimental
dissociation).
2. As noted, better encoded events should be recalled more consistently and
more accurately than poorly encoded events. We should also expect that
encouraging witnesses to guess will lead to less consistent and less accurate
responses than instructing witnesses to be certain before volunteering a
response. Both of these propositions predict that consistent recollections
will be more accurate than inconsistent recollections.
3. Each component of a complex event is processed independently of the
other components. Therefore, accuracy of recalling some components of
a complex event may not necessarily predict how accurately witnesses recall
the other components. Witnesses who make many contradictory state-
ments may be inaccurate on those specifi c statements; however, they may
be accurate on the rest of their testimony. That is, the correlation between
128
Handbook of Psychology of Investigative Interviewing
amount of inconsistency and overall accuracy of a witness ’ s testimony may
be relatively weak.
4. Reminiscence should be a common experience. Furthermore, the amount
of reminiscence should be explained easily by conventional cognitive
theory, which relates reminiscence to changes in retrieval cues from one
interview to another.
5. Reminiscent statements may or may not be accurate. Whether reminiscent
responses are accurate or not depends on a variety of factors, such as the
nature of the question that is asked: Open - ended questions or probes (e.g.,
Describe his face) should yield more accurate responses than closed ques-
tions (e.g., What color were his eyes?).
6. For the same reason as indicated in point 3 above (the independence of
components), witnesses who make many reminiscent statements should
not necessarily be inaccurate on the other (non - reminiscent) statements
of their testimony. The correlation between amount of reminiscence and
overall accuracy may be relatively we
ak.
Experimental t esting
We report an overview of the results from 19 experiments to assess the predic-
tions of the Courtroom and Cognitive theories. Each of the experiments
conformed to the following general procedure. Witnesses (typically college
students, but the same patterns of results also obtained for others) either
watched a videotape of a simulated crime (robbery or homicide) or observed
a live, innocuous event or a staged confrontation between two people. The
witnesses were then tested formally (paper
- and - pencil test) or, as in most
experiments, participated in face - to - face interviews to assess their memories of
the event. Most of the witnesses were tested twice. The tests or interviews
occurred either shortly after observing the event (within 30 minutes) or after
a delay of up to two weeks. The interview questions or probes were either
open - ended (e.g., Describe the robber) or were closed. There were three kinds
of closed questions: cued recall (e.g., What color were the robber ’ s eyes?),
multiple choice (What color was the robber
’ s eyes: green, blue, black, or
brown?) and True/False (The robber ’ s eyes were green: true or false?). The
witnesses were sometimes encouraged to be very certain before volunteering
an answer, sometimes encouraged to guess, and sometimes not given any
explicit instructions about certainty.
We compared the witness statements across the two interviews and cate-
gorized them as one of four types: Consistent (same answer at Time 1 and
Time 2, e.g., robber was a white male at Time 1, and robber was a white male
at Time 2), contradiction (contradictory answers at Time 1 ( clean - shaven ) and
Time 2 ( bearded) ), reminiscent (no answer at Time 1, but witness provides
an answer at Time 2 ( red shirt ), and forgotten (witness provides an answer at
Time 1 ( baseball hat ) but does not answer at Time 2. We then calculated the
The Relation between Consistency and Accuracy of Eyewitness Testimony 129
accuracy of each of the four response categories (consistent, contradiction,
reminiscent, and forgotten) in addition to the accuracy of the entire testi-
mony. Accuracy was calculated by dividing the number of correct statements
by the total number of statements. For instance, if a witness made eight
correct statements out of ten total statements, then her accuracy rate was 0.8
(8/10).
The results of these experiments are organized around the predictions of
the ‘ Courtroom ’ and ‘ Cognitive ’ Theories.
Experimental d issociations: c ommon v ersus u nique m ental p rocesses. Some
experimental manipulations had the same effects on consistency and accuracy
of testimony, whereas other manipulations had different effects on the two
measures. When witnesses were instructed to be certain before volunteering
answers, they were both more consistent and more accurate than when
instructed to guess if unsure (Phillips, Fisher & Krioukova, 1999 ). Varying
the format of the question (open - ended vs. closed) also had parallel effects on
consistency and accuracy: Open
- ended questions yielded more consistent
responses and also more accurate answers than closed questions (Fisher &
Patterson, 2004 ). Other manipulations, however, had different, and some-
times opposite, effects on consistency and accuracy of recollection. Specifi cally,
witnesses were less consistent but more accurate when tested shortly after the
critical event (within 30 minutes) than when tested after two weeks (Fisher,
Schreiber, Burguera & Alvarez, 2003 ). That is, delaying the tests increased
consistency but decreased accuracy. This experimental dissociation suggests
that consistency and accuracy may refl ect different underlying mechanisms
(Tulving, 1985 ), in opposition to the Courtroom Theory.
Accuracy of c ontradictions. In all of our experiments, the accuracy rate of
contradictory answers was low (Brock, Fisher
& Cutler,
1999 ; Fisher
&
Patterson, 2004 ; Gilbert & Fisher, 2006 ; ). For instance, in Gilbert & Fisher,
the accuracy rate of contradictory statements was only 0.49; by comparison,
the accuracy rate of consistent answers was almost perfect (0.95). Both the
Courtroom and Cognitive theories correctly predicted the low accuracy of
contradictory statements.
Contradiction as a p redictor of o verall a ccuracy. Although contradictory
statements were considerably less accurate than consistent statements, wit-
nesses who made many contradictory statements were not much less accurate
overall (all of the statements in their entire testimony) than were witnesses
who made only a few contradictory statements. Witnesses were scored in terms
of the accuracy of their overall testimony and the proportion of all statements
that were contradictory (typically, this proportion is relatively small, as wit-
nesses who take their task seriously rarely make more than a few contradic-
tions). Across the various conditions of the experiments, the correlations
130
Handbook of Psychology of Investigative Interviewing
between the proportion of contradictory statements and the accuracy of the
entire testimony was relatively low (the Pearson correlation coeffi cient was
generally between 0.00 and 0.35) (Fisher & Cutler, 1995 ; Brewer et al ., 1999 ;
Fisher & Patterson, 2004 ; Gilbert & Fisher, 2006 ). The fi nding that contra-
dictions are poor predictors of witnesses ’ overall testimonial accuracy is in
direct violation of the ‘ Courtroom ’ theory.
Reminiscence: f requency and e xtraordinary e xplanations. In our studies,
and also those of other researchers (e.g., Scrivner & Safer, 1988 ), reminiscence
was a common phenomenon (see Payne, 1987 , for a review). In Gilbert &
Fisher (2006) , 98% of witnesses who were tested twice (189 of 192) made at
least two reminiscent recollections, hardly a rare phenomenon, as suggested
by the Courtroom Theory. Furthermore, the number of reminiscent state-
ments a witness made was highly related to the dissimilarity of the questions
(retrieval cues) that were asked on the two tests. When the retrieval cues
changed from Test 1 to Test 2, witnesses made almost twice as many reminis-
cent statements (10.1) as when the same cues were given on the two tests
(6.1). As this fi nding is compatible with the Cognitive Theory, one need not
postulate extraordinary mechanisms to account for reminiscence, as the
Courtroom Theory suggests.
Accuracy of
r eminiscence. Reminiscent statements varied in accuracy
across studies, from a low of 0.66 (Brock et al ., 1999 ) to a high of 0.87
(Gilbert & Fisher, 2006 ). That reminiscent statements can be very accurate
violates the dire predictions of the Courtroom Theory, which assumes refl ex-
ively that reminiscence is problematic (see also LaRooy, Pipe
& Murray,
2005 ).
Although reminiscent statements were often accurate, they were not as
accurate as either consistent or forgotten statements. In Gilbert
& Fisher
(2006) , for example, the accuracy rates for consistent, forgotten
, and reminis-
cent, statements were, respectively, 0.95, 0.93, and 0.87 (see also Brock et al .,
1999 ; Fisher & Patterson, 2004 , for similar patterns). Reminiscent statements,
however, were much more accurate than were contradictions (0.49). At the
very least, then, we should distinguish between different kinds of inconsistency
and pay most attention to direct contradictions.
Reminiscence as a p redictor of o verall a ccuracy. The prevalence of remi-
niscent statements was not predictive of overall accuracy. Witnesses who made
more reminiscent statements were only minimally, and non - signifi cantly, less
accurate than witnesses who made fewer reminiscent statements. In Gilbert &
Fisher (2006) the Pearson correlation coeffi cient between proportion of remi-
niscent statements and overall accuracy was − 0.05. This correlation was uni-
formly low across the two tests: Proportion of reminiscent statements was
non - signifi cantly correlated with overall accuracy at Test 1 (0.03) and also at
The Relation between Consistency and Accuracy of Eyewitness Testimony 131
Test 2 ( − 0.14). Similar patterns of non - signifi cant correlations were also found
in Fisher & Patterson (2006) and Gordon & Follmer (1994) .
Relative to consistent recollections, which are by far the most common
category of response, reminiscent responses are relatively infrequent (approxi-
mately 22% of all responses in Gilbert & Fisher, 2006 ) and contradictions are
very rare (less than 1%). To increase the reliability of these infrequent events,
we therefore combined contradictory and reminiscent statements into one
score to determine if these ‘ troublesome ’ inconsistencies, when aggregated,
were more predictive of overall accuracy. Specifi cally, we examined whether
the amount of inconsistency was related to the accuracy of consistent items.
The number of inconsistencies (contradictions and reminiscences) was not at
all predictive of the accuracy of consistent items (correlation coeffi cient = − 0.06).
Even when we compared the most discrepant witnesses (those who made
12 – 18 inconsistencies) to the most consistent witnesses (0 – 6 inconsistencies),
the accuracy rates of consistent items did not differ, 0.94 vs. 0.96, respectively.
No matter how we scored the data, there was no evidence to support the
Courtroom Theory that reminiscence is predictive of inaccuracy of the overall
Handbook of Psychology of Investigative Interviewing: Current Developments and Future Directions Page 25
