MEMORY, 2006, 14 (1), 27±36

Remembering a criminal conversation: Beyond

eyewitness testimony

Laura Campos
University of Granada, Spain

MarõÂa L. Alonso-Quecuty
University of La Laguna, Spain

Unlike the important body of work on eyewitness memory, little research has been done on the accuracy
and completeness of ``earwitness'' memory for conversations. The present research examined the effects
of mode of presentation (audiovisual/ auditory-only) on witnesses' free recall for utterances in a criminal
conversation at different retention intervals (immediate/delayed) within a single experiment. Different
forms of correct recall (verbatim/gist) of the verbal information as well as different types of errors (dis-
tortions/fabrications) were also examined. It was predicted that participants in the audiovisual modality
would provide more correct information, and fewer errors than participants in the auditory-only modality.
Participants' recall was predicted to be impaired over time, dropping to a greater extent after a delay in
the auditory-only modality. Results confirmed these hypotheses. Interpretations of the overall findings
are offered within the context of dual-coding theory, and within the theoretical frameworks of source
monitoring and fuzzy-trace theory.

Most research on witness memory is concerned
with eyewitnesses (see Wells & Olsen, 2003, for a
review). Empirical research on earwitnesses' tes-
timony is much more limited (see Yarmey, 1995,
for a review). However, for some crimes, such as
those committed over the telephone (e.g., obscene
phone calls) or crimes committed in darkness
(e.g., hooded rape), the most important piece of
evidence is the auditory information.
Earwitness research has focused primarily on
voice identification (e.g., Cook & Wilding, 2001;
Yarmey, Yarmey, Yarmey, & Parliament, 2001),
including some research on memory for different
kinds of nonverbal auditory stimuli, such as gun-
shots (e.g., Huss & Weaver, 1996). Nevertheless,
there are particular crimes that are committed
verbally through conversation; that is, ``language
crimes'' (verbal harassment at the workplace,
bribery, etc.). In these cases the central point is not
who did it, but witnesses' memory of what was said
(i.e., verbal content). In the legal context, witness
statements about criminal conversations are
essential because they are frequently the only
available evidence (Pezdek & Prull, 1993).
Observers who hear a criminal conversation
may be either earwitnesses who can both hear and
see the conversational event, or earwitnesses (e.g.,
blind witnesses) who can only listen to that con-

Correspondence should be addressed to Maria L. Alonso-Quecuty, Department of Cognitive Psychology, Faculty of Psychology,
University of La Laguna, Campus de Guajara, s/n 38205 La Laguna (Tenerife), Spain. Email: mlalonso@ull.es
This research was supported by Grant BSO2002-03811 from the Spanish Ministerio de Ciencia y TecnologõÂa (MCYT) and Grant
COF2003/014 from the Canary Government to both authors. Some of the data from this article were previously presented at the Fifth
Biennial Conference of the Society for Applied Research in Memory and Cognition (SARMAC) in Aberdeen (July 2003). We would
like to express our deep gratitude to Professor Margarita Diges from the Universidad AutoÂnoma de Madrid (Spain) for her special
assistance. We also thank Dan Wright, James Ost, and an anonymous reviewer for their helpful comments on an earlier draft of this
paper.

# 2006 Psychology Press Ltd

http://www.psypress.com/memory DOI:10.1080/09658210444000476
28 CAMPOS AND ALONSO-QUECUTY
versation. Does the memory for conversations
differ depending on the modality of experience?
This study aims to answer this question.
In previous studies of memory for conversa-
tions, the participants witness the conversational
event either in an audiovisual modality, through a
videotape as a general rule (Bates, Masling &
Kintsch, 1978; MacWhinney, Keenan & Reinke,
1982), or in an auditory-only modality through an
audiotape (Hjelmquist, 1989; Jarvella & Collas,
1974; Pezdek & Prull, 1993). Surprisingly, little
research on free recall for conversations deals
with both the auditory-only and the audiovisual
modalities in the same design. In one of these few
studies, Stafford, Waldron, and Infield (1989)
found no significant differences in immediate free
recall for utterances as a function of the mode of
presentation (audiovisual/auditory-only). Other
studies examining the effects of audiovisual and
auditory-only modalities on memory have shown
the same results when immediate free recall for
the verbal information conveyed through a target
material (e.g., magazine programmes, party poli-
tical broadcasts) has been tested (e.g., Furnham,
Proctor, & Gunter, 1988; Gunter, Furnham, &
Leese, 1986).
In previous witness research, however, it has
been found that memory for an event witnessed
audiovisually may differ from memory for the
same event when it is presented in an auditory-
only modality, as was shown in an interesting
study carried out by Toglia, Shlechter, and Che-
valier (1992). Toglia et al. examined memory for
an event involving an argument between two
students that was experienced either in an audio-
visual modality, in which participants witnessed
the staged live event (Experiment 1), or in an
auditory-only modality, in which participants lis-
tened to a written version of the same event that
was read aloud to them (Experiment 2). Retention
over time was also tested. In both modalities, it
was found that participants recalled more correct
information related to actions and conversation
immediately than 2 weeks later. Moreover, parti-
cipants in the auditory-only modality showed a
greater decrement in memory performance
between the immediate and delayed tests com-
pared to the decrement in performance showed
over time by participants in the audiovisual
modality.
Unfortunately, Toglia et al. reported no data on
possible differences between the audiovisual
modality and the auditory-only modality in
memory performance, as they did not examine the
effects of mode of presentation on memory for the
event in the same design. As the findings obtained
by Toglia et al. suggest, it might be expected that
witnesses in the audiovisual modality would, in
general, show better memory for conversation, as
well as less forgetting over time, compared to
witnesses in the auditory-only modality. Accord-
ing to the dual-coding theory (Paivio, 1971, 1975,
1986; Paivio & Csapo, 1969), dually coded items
will be remembered better than unitarily coded
items, thanks to richer encoding-elaboration pro-
cesses. The nonverbal (imaginal) trace compo-
nent, at least if it is visual, leads to better memory
performance than the verbal component in their
additive combination when items are encoded
dually. Consequently, it is likely that the same
target memory can be retrieved from either code,
compared to the case where only one of the codes
has been stored (Paivio, 1986).
Another important issue in research on recall
for conversations in forensic context is concerned
with different forms of retention of the verbal
information. That is, when people remember
conversations, what do they remember? Previous
research on memory for conversations shows that
recall protocols contain largely gist memory, and
that verbatim memory for utterances, therefore, is
unusual when retention is measured through a
free recall (as opposed to a recognition) test
(Miller, deWinstanley, & Carey, 1996; Ross &
Sicoly, 1979; Stafford, Burggraf, & Sharkey, 1987;
Stafford & Daly, 1984; Stafford et al., 1989).
Furthermore, empirical studies of memory for
sentences, models of discourse processing, and
models of memory claim that memory for the
semantic content (gist) of to-be-remembered
(TBR) material is better retained over time than
memory for the actual words and syntactic form
(verbatim wording) (see e.g., Brewer & Naka-
mura, 1984; Gernsbacher, 1985; Kintsch, Welsch,
Schmalhofer, & Zimny, 1990; Sachs, 1967). Par-
allel findings have been observed in previous
studies of memory for conversations (e.g.,
Hjelmquist & Gidlund, 1985; Stafford et al., 1987).
More recently, the topic of verbatim and gist
memory has received extensive treatment by
Brainerd and Reyna in their Fuzzy-Trace Theory
(FFT), which provides a relevant approach to
interpret memory performance (Brainerd &
Reyna, 2002; Brainerd, Reyna, & Brandse, 1995;
Reyna, 2000; Reyna & Brainerd, 1995; Reyna &
Kiernan, 1994). The FTT states that during
encoding of TBR material, verbatim and gist tra-
ces are formed in parallel. The verbatim trace

represents the item's surface form whereas the gist
trace represents the semantic, relational, and ela-
borative properties of the stimulus. This theory
posits that people retrieve dissociated verbatim
and gist traces, with correct recollection being
supported by both kinds of traces (Brainerd &
Reyna, 2002). According to the FTT, forgetting is
assumed to be the gradual fragmentation of traces,
which is higher for verbatim (as opposed to gist)
representations. Therefore, people are more
likely to rely on verbatim representations imme-
diately after the TBR material is presented, but
shift to gist after a delay (Reyna & Kiernan, 1994).
The FTT also offers a framework in which to
interpret a variety of memory errors such as fab-
rications or illusory memories (Toglia, Neuschatz,
& Goodwin, 1999). This theory assumes that these
kind of false memories occur because gist is
accessed, rather than verbatim memory, and fac-
tors that influence the accessibility of gist repre-
sentations (e.g., when memory testing is delayed
for several days and participants are instructed to
rely on the meaning content of experience) have
predictable effects on memory performance
(Reyna & Kiernan, 1994). Accordingly, several
experiments have shown high levels of false
memories persistence (e.g., Brainerd et al., 1995;
Toglia et al., 1999).
In addition to fabrications (i.e., recollecting
events that never happened), false memories may
also take the form of remembering an event in a
distorted manner (e.g., incorrect recall of some
detail of the original event) (Toglia et al., 1999).
Previous research suggests that these two different
forms of false memories may represent different
psychological phenomena. Gudjonsson and Clare
(1995) found that distortions and fabrications did
not correlate significantly with each other and
loaded on separate factors, so that they recom-
mend considering them separately in research
involving both types of false memories.1 In line
with FTT, more distortions should be expected
when memory is tested immediately compared to
when memory is tested after a delay, as individuals
are more likely to rely on verbatim traces imme-
diately after the TBR material is presented.
Another theoretical approach to false mem-
ories is Johnson and colleagues' source-monitor-
ing framework (SMF) (Johnson, 1988; Johnson,
Hashtroudi, & Lindsay, 1993; Lindsay & Johnson,
1
In the present study, distortions and fabrications were
considered separately since, in a forensic investigation, the two
may have quite different implications (Milne & Bull, 2002).
BEYOND EYEWITNESS TESTIMONY 29
1991; Mitchell & Johnson, 2000), which is an
extension of the reality-monitoring (RM) model
proposed by Johnson and Raye (1981). According
to this theory, false memories are a result of
source misattribution errors, which are less likely
to occur when the event memory is richly detailed
and its attributes are distinctive, uniquely char-
acteristic of its source (Johnson et al., 1993). After
a delay, the likelihood of source errors arising
increases, as the perceptual characteristics of the
externally generated memories fade (e.g., Frost,
2000; Underwood & Pezdek, 1998). From this
view, it should be expected that participants in the
audiovisual modality would make fewer fabrica-
tions than participants in the auditory-only mod-
ality, because of richer perceptual features at
input when a conversational event is both seen
and heard compared to when the conversational
event is only heard. Accordingly, fewer fabrica-
tions in the audiovisual modality compared to the
auditory-only modality should also be expected
over time.
The purpose of the present research was to
study the effects of both mode of presentation
(audiovisual/auditory-only) and retention interval
(immediate/delayed) on witnesses' free recall for
utterances in a criminal conversation within a
single experiment. Our hypotheses can be sum-
marised as follows: (1) Participants in the audio-
visual modality will provide overall more correct
information, and fewer errors than participants in
the auditory-only modality; (2) Participants in the
audiovisual modality will show less forgetting of
correct information, and fewer errors over time
compared to participants in the auditory-only
modality; (3) Participants overall will show
superior gist over verbatim recall; (4) Participants'
verbatim and gist recall will be impaired over
time; (5) Forgetting of correct information over
time will be higher for verbatim details than for
gist; and (6) Participants will make fewer
distortions over time, whereas the amount of
fabrications will tend to be persistent.
METHOD
Participants
A total of 80 undergraduate students from the
University of La Laguna in Tenerife volunteered
to participate in the experiment. There were 64
females and 16 males. Their ages ranged from 20
to 22 years (mean 20.52).
30 CAMPOS AND ALONSO-QUECUTY
Materials
A videotape in which two actors staged a mock
dyadic criminal conversation was recorded on the
Campus de Guajara square of the University of La
Laguna. The event under discussion was a planned
theft. The recording lasted 15 minutes approxi-
mately. This length provided the conversation
with an episodic structure (i.e., consisting of
introduction, core, and ending) such as that used
in other researches (e.g., Bates, Kintsch, Fletcher,
& Giuliani, 1980; Bates et al., 1978; Miller et al.,
1996). In order to ensure high ecological validity,
the conversation was recorded as if it had been
obtained by chance by one of the security cameras
located in the campus. It was recorded at night in
black and white format and with environmental
sound.
Design
The design was 2 6 2 factorial. The first factor,
mode of presentation, was between subjects and
consisted of two groups: (1) audiovisual and (2)
auditory-only. The second factor, retention inter-
val, was also between subjects and consisted of
two groups: (1) immediate and (2) delayed.
Procedure
Participants were randomly assigned to four
treatment combinations as a function of Mode of
Presentation (audiovisual/auditory-only) and
Retention Interval (immediate/ delayed). Partici-
pants in group 1 (n = 20) received the audiovisual
modality and were tested immediately; partici-
pants in group 2 (n = 20) received the audiovisual
modality and were tested 4 days later; participants
in group 3 (n = 20) received the auditory-only
modality and were tested immediately; and
participants in group 4 (n = 20) received the
auditory-only modality and were tested 4 days
later.
All participants went through three experi-
mental stages. In the first phase, participants in the
audiovisual condition were told that they were
going to watch a dyadic conversation and that they
had to pay as much attention as possible to it, as if
they were real eyewitnesses. Then they watched
the videotaped conversation in small groups
comprising around 10 persons. Participants in the
auditory-only condition received the same
instructions, with the exception that they were
going to listen to an audiotaped conversation
(instead of watching it).
In the second phase, all participants carried out
a distractor task for 15 minutes, consisting of
crossword puzzles (HernaÂndez-Fernaud &
Alonso-Quecuty, 1997). Afterwards, participants
were asked to free recall the target conversation,
which they had to complete in writing during an
unlimited time. Participants in groups 1 and 3 did
an immediate free-recall task; participants in
groups 2 and 4 did it 4 days later.
Finally, all participants were asked questions
concerning their comprehension of the conversa-
tion (e.g., what the speakers were going to do after
the conversation took place), and whether they
had any previous knowledge of the target mate-
rial, in order to eliminate those participants who
might have been familiar with such conversation.
All participants showed that they had understood
the conversation and were naõÈve concerning the
purpose of the experiment.
Scoring procedure
The videotape was transcribed. Next, utterances
were split up into idea units. An idea unit was
defined as the smallest utterance that contained a
subject and predicate (Miller & deWinstanley,
2002; Miller et al., 1996). The final coding script
consisted of 374 idea units. Recall protocols were
similarly split up into recall units.
Later, each recall unit was compared to the
original idea units and coded as being verbatim
recall, gist recall, distortion, or fabrication. Each
recall unit was scored by giving 1 point. Recall
protocols were coded by two independent judges,
who were doctoral students previously trained in
this task. They split each recall protocol up into
recall units and then coded each one within the
following dependent measures:
(1) Verbatim recall: Accurate recall of an ori-
ginal entire idea unit included in the conversation.
(2) Gist recall: Accurate recall of the kernel of
meaning from an original entire idea unit of which
wording differs from the one included in the
conversation; that is, a paraphrase of the original
idea unit (e.g., recalling the utterance ``don't fail
in your task'' instead of the original idea unit
``don't make any mistake when you do the job'').
Gist recall also included recall units that sum-
marised two or more idea units accurately (e.g.,
 BEYOND EYEWITNESS TESTIMONY 31

recalling the utterance ``I need you to give me TABLE 1

more information about my mission'' instead of Mean of verbatim and gist recall as a function of mode of
presentation and retention interval
the original utterances ``I don't have enough
information about the target student'' and ``Please Retention interval
tell me more about the magic key'', separately).
These were less precise and more global state- Type of correct recall Immediate Delayed Mean
ments than a paraphrase of an original idea unit
Verbatim
(Miller et al., 1996). Also, these summary state- Audiovisual 0.35 (0.58) 0.10 (0.30) 0.22 (0.48)
ments could be not directly related to any one idea Auditory-only 0.80 (1.36) 0.00 (0.00) 0.40 (1.03)
unit but rather captured the global situation being
Gist
described and permitted the scoring of a ``deeper'' Audiovisual 14.90 (6.15) 12.45 (6.56) 13.68 (6.39)
level of gist than was allowed by one-to-one idea Auditory-only 14.90 (7.15) 6.10 (3.87) 10.50 (7.22)
unit matching (Stafford & Daly, 1984; Stafford et
al., 1987, 1989). Standard deviations are presented in parentheses.
(3) Distortion: Recall unit that was in some
way discrepant with the original idea unit
0.57) than in the delayed test (mean 0.05). Indeed,
expressed in the conversation; that is, incorrect
participants' gist recall was greater when they
recall of some detail of an original utterance in the
were tested immediately (mean 14.90) than after a
conversation (e.g., recalling the distorted utter-
delay (mean 9.28). There was also a main effect of
ance ``we have already got three magic keys''
mode of presentation on gist recall. Participants in
instead of the original utterance ``we have already
the audiovisual modality recalled the gist of more
got two magic keys'').
idea units (mean 13.68) than did participants in
(4) Fabrication: Recall of something not
the auditory-only modality (mean 10.50). Mode of
expressed in the conversation (e.g., recalling that
presentation main effect was moderated by a sig-
one of the speakers asked the other for some
nificant mode of presentation 6 retention interval
money, when this utterance was not expressed in
interaction. Simple effects analyses revealed that
the conversation).
participants' gist recall was greater in the audio-
visual modality (mean 12.45) than in the auditory-
RESULTS
TABLE 2
Correct recall Mode of presentation 6 retention interval 6 type of correct
recall ANOVAs performed on mean verbatim and gist correct
A series of 2 (mode of presentation: audiovisual/ recall scores
auditory-only) 6 2 (retention interval: immedi-
Source df F o2
ate/delayed) 6 2 (type of correct recall: verbatim/
gist) analyses of variance (ANOVAs) with the Between subjects (verbatim recall)
first and second factors between subjects and the Mode of presentation (MP) 1 1.06 .0008
third factor within subjects were performed on Retention interval (RI) 1 9.62** .09
mean verbatim and gist correct recall scores. The MP 6 RI 1 2.64 .02
S within-group error 76 (0.57)
descriptive statistics for these data are displayed in
Table 1. Between subjects (gist recall)
Mode of presentation (MP) 1 5.48* .05
As expected, there was a main effect of type of
Retention interval (RI) 1 17.20*** .17
correct recall (see Table 2). Participants' gist MP 6 RI 1 5.48* .05
recall was greater (mean 12.09) than their verba- S within-group error 76 (36.79)
tim recall (mean 0.31). All the interactions were Within subjects
significant. Simple effects analyses revealed fur- Type of correct recall (CR) 1 320.88*** .67
ther that participants' gist recall was greater than CR 6 MP 1 6.49** .03
their verbatim recall in all treatment combina- CR 6 RI 1 15.05*** .08
tions. CR 6 MP 6 RI 1 4.87* .02
CR 6 S within-group error 76 (17.28)
As also shown in Table 2, there was a main
effect of retention interval on both verbatim and Values in parentheses represent mean square errors. S =
gist recall. As predicted, participants' verbatim subjects.
recall was greater in the immediate test (mean * Significant at p < .05; ** p < .01; *** p < .001.
32 CAMPOS AND ALONSO-QUECUTY
only modality (mean 6.10) when they were tested
after a delay, F(1, 77) = 9.05, MSE = 44.53, p =
.004, o2 = .09. Nevertheless, no significant differ-
ences were found in immediate gist recall as a
function of the mode of presentation. As pre-
dicted, a significant reduction in gist recall over
time was observed, but solely in the auditory-only
modality (mean 14.90 versus mean 6.10), F(1, 77)
= 19.89, MSE = 38.93, p < .001, o2 = .19, which
suggests that gist recall is more sensitive to the
4-day delay in this mode of presentation, whereas
gist recall remains relatively stable from the
immediate retention interval (mean 14.90) to the
delayed interval (mean 12.45) in the audiovisual
modality (see Figure 1).
Errors
A series of 2 (mode of presentation: audiovisual/
auditory-only) 6 2 (retention interval: immedi-
ate/delayed) 6 2 (type of error: distortion/fabri-
cation) ANOVAs were performed on mean
distortions and fabrications scores, with the first
and second factors between subjects and the third
factor within subjects. The descriptive statistics for
these data are displayed in Table 3.
A main effect of type of error was obtained (see
Table 4). Participants overall made more fabri-
cations (mean 1.98) than distortions (mean 1.39).
Both two-way interactions were significant. Sim-
ple effects analyses revealed that superior amount
of fabrications (mean 2.53) over distortions (mean
Figure 1. Mode of presentation 6 retention interval interaction on gist recall.
1.25) occurred in the auditory-only (as opposed to
the audiovisual) modality, F(1, 78) = 17.76, MSE =
1.83, p < .001, o2 = .09. Indeed, participants
showed a superior amount of fabrications (mean
2.07) over distortions (mean 1.00) in the delayed
(as opposed to the immediate) test, F(1, 78) =
11.85, MSE = 1.95, p = .001, o2 = .06.
As also shown in Table 4, there was a main
effect of retention interval on distortions. As
expected, participants tested immediately made
more distortions (mean 1.78) than did participants
tested after a delay (mean 1.00). There was also a
main effect of mode of presentation on fabrica-
tions. Also in accordance with our predictions,
participants in the auditory-only modality made
more fabrications (mean 2.53) than did partici-
pants in the audiovisual modality (mean 1.42).
DISCUSSION
In this paper we have explored the effects that
mode of presentation and delay have on wit-
nesses' recall for utterances of a criminal con-
versation. Different forms of accurate recall
(verbatim/gist) of the verbal information as well as
different forms of false memories (distortions/
fabrications) have been examined.
The results showed that conversational recall
was not very good, consistent with those obtained
in previous research on memory for conversations
in which free recall (as opposed to recognition)
tests have been used to examine retention (e.g.,
 BEYOND EYEWITNESS TESTIMONY 33

TABLE 3 As predicted, gist recall was greater than ver-

Mean of distortions and fabrications as a function of mode of batim recall. The results also revealed that
presentation and retention interval
superior gist recall over verbatim recall occurred
Retention interval in both immediate and delayed tests. Participants'
memories were not word-for-word reproductions,
Type of error Immediate Delayed Mean as an average of only 0.31 idea units recalled
verbatim indicates (Hjelmquist & Gidlund, 1985).
Distortion
Audiovisual 1.85 (1.08) 1.20 (1.19) 1.52 (1.17)
Such a low level of verbatim recall might be due to
Auditory-only 1.70 (1.86) 0.80 (0.76) 1.25 (1.48) a floor effect; however, this might also be due to
the fact that participants overall tend to focus on
Fabrication
Audiovisual 1.45 (1.79) 1.40 (1.35) 1.42 (1.56) the kernel of meaning rather than specific details
Auditory-only 2.30 (2.75) 2.75 (1.80) 2.53 (2.30) (e.g., Stafford et al., 1987; Stafford & Daly, 1984).
According to Fuzzy-Trace Theory (FTT)
Standard deviations are presented in parentheses. (Brainerd & Reyna, 2002; Reyna, 2000; Reyna &
Brainerd, 1995; Reyna & Kiernan, 1994), indivi-
duals retrieve dissociated verbatim and gist traces,
Miller et al., 1996; Ross & Sicoly, 1979; Stafford & with accurate memories being supported by both
Daly, 1984). As predicted, there was forgetting
kinds of traces. Gist retrieval is favoured when gist
over time: participants recalled more correct
traces are strong relative to verbatim traces, as
information immediately than after the 4-day
when individuals are instructed to rely on the
delay, consistent with findings in previous studies
meaning content of experience (Brainerd &
of recall for conversations (Hjelmquist & Gidlund,
Reyna, 2002). From this view, participants in the
1985; Stafford et al., 1987). Although participants'
present study were likely to rely mainly on gist
verbatim and gist recall were both greater in the
traces, rather than verbatim traces, as intentional
immediate test than in the delayed test, extremely memory instructions were given to them (Toglia
poor verbatim recall found in the delayed test (only
et al., 1999). Conditions of test might also bias
0.05 idea units on average) suggests that verbatim
towards gist retrieval, as the participants were
traces became inaccessible more rapidly than gist
asked to free recall the target conversation, where
traces (Reyna & Brainerd, 1995).
naturally no instructions to recall the verbal
information verbatim nor to respond to specific
TABLE 4
questions were given (Lindsay & Johnson, 1991).
Mode of presentation 6 retention interval 6 type of error Participants made more fabrications (i.e., recall
ANOVAs performed on mean distortions and fabrications of something not expressed in the conversation)
scores than distortions (i.e., incorrect recall of some
detail of an original utterance in the conversa-
Source df F o2
tion), but only after the 4-day delay. According to
Between subjects (distortions) FTT, false memories vary directly with the rela-
Mode of presentation (MP) 1 0.90 .001 tive accessibility of gist versus verbatim repre-
Retention interval (RI) 1 7.18** .07 sentations, and individuals are more likely to rely
MP 6 RI 1 0.18 .01 on gist (as opposed to verbatim representations)
S within-group error 76 (1.67)
after a delay (Reyna & Brainerd, 1995). An
Between subjects (fabrications) interesting finding was that distortions and fabri-
Mode of presentation (MP) 1 6.09** .06
cations rates differed over time. As predicted,
Retention interval (RI) 1 0.20 .009
MP 6 RI 1 0.31 .008 participants made more distortions in the
S within-group error 76 (3.97) immediate test than in the delayed test. No sig-
Within subjects
nificant differences were found, however, in the
Type of error (E) 1 7.96** .04 amount of fabrications as a function of the
E 6 MP 1 10.89*** .06 retention interval. Given that forgetting is
E 6 RI 1 5.48* .03 assumed to be the gradual fragmentation of traces,
E 6 MP 6 RI 1 0.81 .001 which is higher for verbatim (as opposed to gist)
E 6 X within-group error 76 (1.73)
representations (Reyna, 2000), one would expect
Values in parentheses represent mean square errors. S = that participants would make more distortions
subjects. immediately after the target conversation was
* Significant at p < .05; ** p < .01; *** p < .001. presented than after the 4-day delay. On the other
34 CAMPOS AND ALONSO-QUECUTY
hand, the relative stability of fabrications over
time is also consistent with FTT, which predicts
that illusory memories can be highly persistent, as
they arise from particularly stable memory
representations (gist traces) (Brainerd et al., 1995;
Toglia et al., 1999).
In this study, participants witnessed the con-
versational event either in an audiovisual mod-
ality or in an auditory-only modality. As
predicted, participants in the audiovisual modality
showed better recall of the conversation overall
compared to participants in the auditory-only
modality. Participants in the audiovisual modality
recalled the gist of more idea units of the con-
versation, and made fewer fabrications compared
to participants in the auditory-only modality.
However, no significant differences were found in
verbatim recall, nor in the amount of distortions as
a function of the mode of presentation of the
conversation. As was mentioned above, the pro-
blem here might be a floor effect. But it might also
evidence that the typical manner of remembering
a conversation is in gist when a free recall is used
to test retention (Hjelmquist & Gidlund, 1985;
Miller et al., 1996; Stafford et al., 1987; Stafford &
Daly, 1984).
The mode of presentation main effect on gist
recall was moderated by a significant interaction
with retention interval, which showed that super-
ior gist recall in the audiovisual modality over the
auditory-only modality occurred only after the 4-
day delay. Consistent with previous research on
immediate free recall for conversations (e.g.,
Stafford et al., 1989) as well as with other studies
examining the effects of audiovisual and auditory-
only modalities on memory (e.g., Furnham et al.,
1988; Gunter et al., 1986), no significant differ-
ences were found in immediate gist recall as a
function of the mode of presentation of the con-
versation. Because gist (as opposed to verbatim)
memories become inaccessible more slowly
(Kintsch et al., 1990; Reyna & Brainerd, 1995;
Sachs, 1967; Stafford et al., 1987), it may be that an
immediate (as opposed to a delayed) test is not
sensitive enough to detect any modality effects on
gist recall.
One of the more noteworthy findings of this
study was that, as predicted, participants in the
audiovisual modality showed little forgetting of
correct information over time compared to parti-
cipants in the auditory-only modality. It appears,
therefore, that there are different rates of forget-
ting for witnessing a conversational event audio-
visually and witnessing it in an auditory-only
mode (Toglia et al., 1992). One possible explana-
tion focuses on differences in encoding elabora-
tion processes. In accordance with the dual-coding
theory (Paivio, 1971, 1975, 1986), dually coded
items will be remembered better than unitarily
coded items. From this view, participants who saw
and heard the conversational event were likely to
encode it dually in both an imaginal and a verbal
code, whereas participants in the auditory-only
modality were likely to represent it primarily in a
verbal code, although some referential (imaginal)
coding could also occur. It is likely, therefore, that
the same target memory could be retrieved from
either code in the audiovisual modality, compared
to one code in the auditory-only modality.
Overall better recall of the conversation in the
audiovisual modality compared to the auditory-
only modality was bolstered by the fabrication
data. As predicted, participants in the audiovisual
modality made fewer fabrications compared to
participants in the auditory-only modality. This
result may be explained in line with the source-
monitoring framework (SMF) (Johnson et al.,
1993; Johnson & Raye, 1981). According to the
SMF, false memories arise due to source-mon-
itoring errors, which are less likely to occur when
the mental event is richly detailed and its attri-
butes are distinctive, unique characteristics of its
source (Johnson et al., 1993). In line with this, to
the extent that there are more external perceptual
features at input, it should make source errors less
likely. In the audiovisual modality, it is obvious
that there were richer perceptual features at input,
as there was additional perceptual (visual) infor-
mation (e.g., the environmental setting, actions,
clothing and physical characteristics of partici-
pants, nonverbal information), which was not
presented in the auditory-only modality. Partici-
pants in the audiovisual modality, therefore, were
likely to make fewer fabrications compared to
participants in the auditory-only modality.
The fact that participants in the auditory-only
modality showed a greater fabrication rate over
time compared to participants in the audiovisual
modality is also consistent with the SMF. This
theory predicts that because perceptual char-
acteristics of the externally generated memories
fade over time, the likelihood of making source
errors increases (e.g., Frost, 2000; Underwood &
Pezdek, 1998). In accordance with this, as fewer
perceptual details concerning the conversational
event should be encoded at input in the auditory-
only modality compared to the audiovisual mod-
ality, it should make source errors more likely

over time in the auditory-only modality because
the memory for the conversational event loses its
perceptual details as time progresses.
The results of the present research can be
summarised as follows. First, when participants
are asked to free recall a conversation they tend to
remember the gist, showing little verbatim mem-
ory of what was said. Second, the retention
interval between witnessing a conversation and
later recall test impairs witnesses' performance.
Finally, the extent to which retention interval
impairs memory performance depends on the
mode of presentation (audiovisual/auditory-only)
of the conversation. The findings of the present
research provide evidence that participants in the
auditory-only modality made more fabrications
and showed a greater forgetting of accurate
information over time compared to participants in
the audiovisual modality. This issue is important
for the psychology of testimony because, in some
cases, a criminal conversation takes place under
circumstances in which the participants in the
conversation cannot be seen (e.g., the area is in
darkness), but what they say is heard. Further,
witnesses do not usually make statements imme-
diately (at the scene), but after a delay (at the
police station or in court) following the occur-
rence of the criminal conversation. In this sense,
accounts of ``earwitnesses'' who could only listen
to a criminal conversation should be treated with
extreme caution in court.
Manuscript received 7 July 2003
Manuscript accepted 22 November 2004
PrEview proof published online 5 April 2005
REFERENCES
Bates, E., Kintsch, W., Fletcher, C. R., & Giuliani, V.
(1980). The role of pronominalization and ellipsis in
texts: Some memory experiments. Journal of
Experimental Psychology: Human Learning and
Memory, 6, 671±691.
Bates, E., Masling, M., & Kintsch, W. (1978). Recog-
nition memory for aspects of dialogue. Journal of
Experimental Psychology: Human Learning and
Memory, 4(3), 187±197.
Brainerd, C. J., & Reyna, V. F. (2002). Fuzzy-trace
theory and false memory. Current Directions in
Psychological Science, 11(5), 164±169.
Brainerd, C. J., Reyna, V. F., & Brandse, E. (1995). Are
children's false memories more persistent than their
true memories? Psychological Science, 6, 359±364.
Brewer, W. F., & Nakamura, G. V. (1984). The nature
and functions of schemas. In R. S. Wyer Jr. & T. K.
Srull (Eds.), Handbook of social cognition (pp. 119±
160). Hillsdale, NJ: Lawrence Erlbaum Associates Inc.
BEYOND EYEWITNESS TESTIMONY 35
Cook, S., & Wilding, J. (2001). Earwitness testimony:
Effects of exposure and attention on the face over-
shadowing effect. British Journal of Psychology, 92,
617±629.
Frost, P. J. (2000). The quality of false memory over
time: Is memory for misinformation ``remembered''
or ``known''? Psychonomic Bulletin and Review,
7(3), 531±536.
Furnham, A., Proctor, E., & Gunter, B. (1988). Memory
for material presented in the media: The superiority
of written communication. Psychological Reports,
63, 935±938.
Gernsbacher, M. A. (1985). Surface information loss in
comprehension. Cognitive Psychology, 17, 324±363.
Gudjonsson, G. H., & Clare, I. C. H. (1995). The rela-
tionship between confabulation and intellectual
ability, memory, interrogative suggestibility and
acquiescence. Personality and Individual Differ-
ences, 19(3), 333±338.
Gunter, B., Furnham, A., & Leese, J. (1986). Memory
for information from a party political broadcast as a
function of the channel of communication. Social
Behaviour, 1, 135±142.
HernaÂndez-Fernaud, E., & Alonso-Quecuty, M. L.
(1997). The cognitive interview and lie detection: A
new magnifying glass for Sherlock Holmes? Applied
Cognitive Psychology, 11, 55±68.
Hjelmquist, E. (1989). Recognition memory for utter-
ances in conversation. Scandinavian Journal of Psy-
chology, 30, 168±176.
Hjelmquist, E., & Gidlund, A. (1985). Free recall of
conversations. Text, 5(3), 169±185.
Huss, M. T., & Weaver, K. A. (1996). Effect of mod-
ality in earwitness identification: Memory for verbal
and nonverbal stimuli presented in two contexts.
The Journal of General Psychology, 123(4),
277±287.
Jarvella, R. J., & Collas, J. G. (1974). Memory for the
intentions of sentences. Memory and Cognition, 2(1),
185±188.
Johnson, M. K. (1988). Discriminating the origin of
information. In T. F. Oltmanns & B. A. Maher
(Eds.), Delusional beliefs (pp. 34±65). New York:
Wiley.
Johnson, M. K., Hashtroudi, S., & Lindsay, D. S. (1993).
Source monitoring. Psychological Bulletin, 114(1),
3±28.
Johnson, M. K., & Raye, C. L. (1981). Reality mon-
itoring. Psychological Review, 88, 67±85.
Kintsch, W., Welsch, D., Schmalhofer, F., & Zimny, S.
(1990). Sentence memory: A theoretical analysis.
Journal of Memory and Language, 29, 133±159.
Lindsay, D. S., & Johnson, M. K. (1991). Source mon-
itoring and recognition memory. Bulletin of the
Psychonomic Society, 29, 203±205.
MacWhinney, B., Keenan, J. M., & Reinke, P. (1982).
The role of arousal in memory for conversation.
Memory and Cognition, 10(4), 308±317.
Miller, J. B., & deWinstanley, P. A. (2002). The role of
interpersonal competence in memory for conversa-
tion. Personality and Social Psychology Bulletin,
28(1), 78±89.
Miller, J. B., deWinstanley, P., & Carey, P. (1996).
Memory for conversation. Memory, 4(6), 615±631.
36 CAMPOS AND ALONSO-QUECUTY
Milne, R., & Bull, R. (2002). Back to basics: A com-
ponential analysis of the original cognitive interview
mnemonics with three age groups. Applied Cognitive
Psychology, 16, 743±753.
Mitchell, K. J., & Johnson, M. K. (2000). Source mon-
itoring: Attributing mental experiences. In E. Tul-
ving & F. I. M. Craik (Eds.), Oxford handbook of
memory (pp. 179±195). New York: Oxford Uni-
versity Press.
Paivio, A. (1971). Imagery and verbal processes. New
York: Holt, Rinehart & Winston.
Paivio, A. (1975). Coding distinctions and repetition
effects in memory. In G. H. Bower (Ed.), The psy-
chology of learning and motivation (pp. 179±214).
New York: Academic Press.
Paivio, A. (1986). Mental representations. A dual coding
approach. New York: Oxford University Press.
Paivio, A., & Csapo, K. (1969). Concrete-image and
verbal memory codes. Journal of Experimental Psy-
chology, 80, 279±285.
Pezdek, K., & Prull, M. (1993). Fallacies in memory for
conversations: Reflections on Clarence Thomas,
Anita Hill, and the like. Applied Cognitive Psychol-
ogy, 7, 299±310.
Reyna, V. F. (2000). Fuzzy-trace theory and source
monitoring. An evaluation of theory and false-
memory data. Learning and Individual Differences,
12, 163±175.
Reyna, V. F., & Brainerd, C. J. (1995). Fuzzy-trace
theory: An interim synthesis. Learning and Indivi-
dual Differences, 7(1), 1±75.
Reyna, V. F., & Kiernan, B. (1994). Development of gist
versus verbatim memory in sentence recognition:
Effects of lexical familiarity, semantic content,
encoding instructions, and retention interval.
Developmental Psychology, 30(2), 178±191.
Ross, M., & Sicoly, F. (1979). Egocentric biases in
availability and attribution. Journal of Personality
and Social Psychology, 37(3), 322±336.
Sachs, J. S. (1967). Recognition memory for syntactic
and semantic aspects of connected discourse. Per-
ception & Psychophysics, 2, 437±442.
Stafford, L., Burggraf, C. S., & Sharkey, W. F. (1987).
Conversational memory: The effects of time, recall
mode, and memory expectancies on remembrances
of natural conversations. Human Communication
Research, 14(2), 203±229.
Stafford, L., & Daly, J. A. (1984). Conversational
memory: The effects of recall mode and memory
expectancies on remembrances of natural conversa-
tions. Human Communication Research, 10,
379±402.
Stafford, L., Waldron, V. R., & Infield, L. L. (1989).
Actor±observer differences in conversational mem-
ory. Human Communication Research, 15(4),
590±611.
Toglia, M. P., Neuschatz, J. S., & Goodwin, K. A.
(1999). Recall accuracy and illusory memories:
When more is less. Memory, 7(2), 233±256.
Toglia, M. P., Shlechter, T. M., & Chevalier, D. S.
(1992). Memory for directly and indirectly experi-
enced events. Applied Cognitive Psychology, 6,
293±306.
Underwood, J., & Pezdek, K. (1998). Memory sug-
gestibility as an example of the sleeper effect. Psy-
chonomic Bulletin and Review, 5(3), 449±453.
Wells, G. L., & Olsen, E. A. (2003). Eyewitness testi-
mony. Annual Review of Psychology, 54, 277±295.
Yarmey, A. D. (1995). Earwitness speaker identifica-
tion. Psychology, Public Policy and Law, 1(4),
792±816.
Yarmey, A. D., Yarmey, A. L., Yarmey, M. J., &
Parliament, L. (2001). Commonsense beliefs and the
identification of familiar voices. Applied Cognitive
Psychology, 15(3), 283±299.