ArticlePDF Available

Toward a model of false recall: Experimental manipulation of encoding context and the collection of verbal reports

Authors:

Abstract and Figures

The likelihood of false recall in the Deese-Roediger-McDermott (DRM) paradigm was shown to depend on encoding context in two experiments. When fillers had been preselected to decrease the likelihood of encoding the critical lure's semantic features, false recall was virtually eliminated. However, when the same words were presented rearranged in different presentation orders, levels of false recall that were found in earlier DRM studies (Robinson & Roediger, 1997) were replicated. The role of encoding processes in the DRM paradigm was further explored with additional participants completing the experiment while thinking aloud. During encoding of word lists, participants verbalized semantic elaboration of the critical lure while studying the word lists. A path analysis demonstrated that participants' verbalization of critical lures during encoding reliably predicted their level of false recall.
Content may be subject to copyright.
Copyright 2001 Psychonomic Society, Inc. 806
Memory & Cognitio n
2001, 29 (6), 806-819
It is well known that human memory is fallible; how-
ever, Roediger and McDermott (1995) demonstrated lab-
oratory conditions that evoke high levels of false recall in
a list-learning paradigm. For example, when participants
study lists of words, such as thread, pin, eye, sewing, and
so on, during subsequent recall they also tend to remem-
ber the associated critical lure needle, although this word
had never been presented to them. This general paradigm
was first developed by Deese (1959) and is currently re-
ferred to as the Deese–Roediger–McDermott (DRM) par-
adigm. In a recent extensive review, Roediger, McDermott,
and Robinson (1998) stated that while previous researchers
have noted the robust and reliable nature of this false mem-
ory effect, including the large number of variables that af-
fect false recall, they acknowledge that the processes me-
diating the effect are still poorly understood. The goal of
the present set of experiments was to increase our knowl-
edge of the cognitive processes that mediate encoding and
retrieval of words that have never been presented. In this
study, we explored two different approaches. First, we pro-
posed new experimental manipulations that reduce the
likelihood of false recall. Second, we examined the cog-
nitive processes associated with false recall by collecting
concurrent verbal reports during encoding and recall of
list items. Before turning to a description of our experi-
ments, we will first review the body of relevant f indings
that will constrain any model of the DRM effect.
Phenomenological Aspects of False Recall
When questioned after the completion of recall in the
DRM paradigm, participants report high levels of confi-
dence in their memory for having “heard” the nonpresented
items (Roediger & McDermott, 1995, Experiment 1) and
recount various details regarding the occurrence of the
nonpresented lure during encoding (Roediger & McDer-
mott, 1995, Experiment 2). Norman and Schacter (1997)
have also found that individuals reported remembering
semantic information about the nonpresented items. For
example, in retrospective reports of nonpresented lures
recognized as “old,” participants reported the following:
“needle: this word came in the same list as thread”; “music:
I thought about music I heard this morning”; and “thief:
hea rd word ‘stop,’ thought of screami ng ‘st op thief ’”
(quoted in Norman & Schacter, 1997, p. 841).
McDermott (1997) argued that critical lures consciously
come to mind during the original encoding of list items,
because she fou nd evidence for perceptual p riming of
critical lures during study of the DRM word lists. Her in-
terpretation is supported by Roediger and McDermott’s
The authors express t heir gr atitude to C harles Brainerd , Kathleen
McDermott , and Mary Susan Weldon for their useful comments, sug-
gestions, and recommendat ions on earlier versions of this article.
Thanks also to Richard Wagner for his statistical consultat ions. The au-
thors also thank Colleen Kelley and Rolf Zwaan for their constructive
comments early in the course of this p roject . Special thanks to Ro x-
ana Gonzalez, Melissa Christle, and Ryan Heinl for their assistance in
the data collection phase of these experiments and in the transcriptions
of verbal reports. Correspondence should be addressed to K. A. Good-
win, D epartmen t of Psychology, Kalamazoo College, 1200 Academy
St., Kalamazoo, MI 49006 (e-mail: kgoodwin@kzo o.edu).
Toward a model of false recall:
Experimental manipulation of encoding context
and the collection of verbal reports
KERRI A. GOODWIN
Kalamazoo College, Kalamazoo, Michigan
and
CHRISTIAN A. MEISSNER and K. ANDERS ERICSSON
Florida State University, Tallahassee, Florida
The likelihood of false recall in the Deese–Roediger–McDermott (DRM) paradigm was shown to de-
pend on encoding context in two experiments. When fillers had been preselected to decrease the like-
lihood of encoding the critical lure’s semantic features, false recall was virtually eliminated. However,
when the same words were presented rearranged in different presentation orders, levels of false recall
that were found in earlier DRM studies (Robinson & Roediger, 1997) were replicated. The role of en-
coding processes in the DRM paradigm was further explored with additional participants completing
the experiment while thinking aloud. During encoding of word lists, participants verbalized semantic
elaboration of the critical lure while studying the word lists. A path analysis demonstrated that partic-
ipants’ verbalization of critical lures during encoding reliably predicted their level of false recall.
FALSE RECALL AND VERBAL REPORTS 807
(1995) earlier findings of high levels of “remember” judg-
ments in the false recognition of critical lures (see also
Payne, Elie, Blackwell, & Neuschatz, 1996; Schacter, Ver-
faellie, & Pradere, 1996). Taken together, these phenom-
enological findings suggest that memory associated with
the DRM effect is base d on memory traces generated
prior to the recall phase.
Encoding Manipulations in the DRM Paradigm
Oth er res earchers have successful ly influenced the
likelihood of false recall and recognition by manipulat-
ing the conditions of encoding. For example, Robinson
and Roediger (1997) found that the probability of false re-
call was increased when the number of list items associ-
ated with the critical lure was increased. Read (1996) in-
structed participants to think about and rehearse one list of
words that would help them in a later test of word mean-
ings (referred to as elaborative rehearsal, p. 107). These
semantic encoding instructions led to increases in false re-
call of the critical lure sleep relative to instructions to en-
code list items in their serial order. Furthermore, McDer-
mott (1996) and Toglia, Neuschatz, and Goodwin (1999)
found that by presenting all list words associated with
the critical lure together in blocked group s p roduced a
higher probability of false recall than when all words were
randomly distributed across the list. Finally, in an experi-
ment in which depth of processing was manipulated,
Toglia et al. found a higher likelihood of false recall when
participants encoded semantic properties of the presented
words than when they encoded surface properties of the
presented words. Thus, the effect of an experimental ma-
nipulation of encoding context on false recall suggests that
the processes at encoding contribute to access and later
false recall of critical lures.
Toward a Mo del of False Recall
Our analysis of the research literature suggests that the
phenomenon of false recall involves memory traces gen-
erated during the encoding of list words. Consistent with
participants’ phenomenological reports of the semantic
attributes of falsely recognized items (Norman & Schacter,
1997), we propose that critical lures are explicitly accessed
during the encoding of items in the list when the words are
semantically processed (Toglia et al., 1999). Fur ther-
more, the probability of their retrieval increases when
several associates of the lure are presented together, a
procedure that appears to encourage the semantic group-
ing and elaboration of list items (cf. the blocked condition
in McDermott, 1996). Thus, during recall, participants
may be unable to distinguish between semantic elabora-
tions associated with presented words and those associ-
ated with the critical lure. Consequently, participants are
more likely to falsely recall the critical lure as a result of
previously accessing it during encoding.
Robinson and Roediger (1997) found that increasing
the number of list associates in the DRM lists led to
higher levels of false recall of critical lures. We hypoth-
esized that the contextual organization of list structure
could be changed without altering the number of list
items in order to influence the likelihood that the critical
lure is accessed during encoding and falsely recalled. To
test these predictions, we designed lists in which the en-
coding context of presented words was systematically var-
ied. Similar to Robinson and Roediger, we constructed
lists of eight associates and eight fillers. However, our
filler items were specifically chosen to provide contextual
associations that were unrelated to the overall list theme
or critical lure. For example, in the soft list (in which soft
was the critical lure), the associate hard was followed by
the filler hat, the associate pillow was followed by the
filler case, and the associate loud was followed by the
filler speaker. Subsequently, five list structures contain-
ing the same words in differing orders were created as a
manipulation of contextual organization of the lists:
(1) eig hts, 8 associates and 8 fillers; (2) fours, 4 associ-
ates, 8 fillers, and 4 associates; (3) twos, 2 associates and
2 fillers, alternating; (4) ones, 1 associate and 1 filler un-
related to previous associate, alternating; and (5) the list
str uc ture contextually u nrelated (CU) to the lure (see
Table 1).
Table 1
Exa mp le of so ft List in Eights, Fours, Twos, On es, and Cont extually Un related (C U) L ist Struct ures
8-Related, 8-Unrelated 4-Related, 4-Unrelated 2-Related , 2-Unrelated 1-Related, 1-Unrelated CU
hard hard hard hard hard
light light light tone hat
pillow pillow tone l ight light
plush plush trade trade bulb
loud ton e pillow pillow pil low
cotton trad e plush g in case
fur gin gin plush p lush
touch spea ker speaker speaker garden
hat garden l oud l oud loud
bulb case cotto n garden spea ker
case bulb garden cotton cotton
garden hat case case g in
speaker lou d fur fur fur
gin cotton touch bu lb trade
trade fur bulb touch tou ch
tone tou ch hat hat tone
Note—Unrel ated filler items are given in italics.
808 GOODWIN, MEISSNER, AND ERICSSON
In the eights, fours, twos, and ones list structures, the
filler items were positioned such that they were not in
proximity to their related associate. Thus, semantic rela-
tions between the list associates were likely to be evoked,
including generation of the critical lure. In the CU list
structure, each associate was immediately followed by
(or paired with) its related filler item, which was predicted
to minimize activation of the overall list theme and criti-
cal lure during encoding.
We proposed that the presentation of fillers that guide
semantic encoding away from the critical lure (CU list
structure) would reduce the frequency of explicit access
to the critical lure and subsequently reduce the likelihood
of false recall. McDermott (1996) and Toglia et al. (1999)
found that randomly ordered list items reduced false re-
call relative to lists that were blocked by theme. In our
experiments, all lists were blocked by theme, and only in
the CU list structure condition were items contextually
manipulated in an effort to reduce false recall in the face
of the list theme. The other list conditions assessed the
effect of physical proximity of the list words associated
with the critical lure. Here, we hypothesized that the en-
coding of the critical lure would more likely occur in lists
that were organized around the critical theme, specifi-
cally when themed items were in cl ose proximity (cf.
Robinson & Roediger, 1997). Therefore, themed items
in close proximity to each other (as in the eights list
structure) would lead to a higher likelihood of false re-
call. In a companion experiment, we collected “think
aloud” protocols from a small number of participants in
an effort to trace their encoding and retrieval processes.
We hypothesized that false recall in the standard DRM
conditions is related to verbalized encodings involving
the lure generated during the presen tation of the lists.
In sum, we have proposed a two-pronged approach to
the study of false recall using the DRM paradigm. First,
in two experiments, we experimentally manipulated the
encoding context of list items in an attempt to eliminate
false recall. Second, using the procedures from these ex-
periments, we further explored the encoding and re-
trieval processes involved in the generation and false re-
call of critical lures with the collection of verbal reports.
Furthermore, we explored path models of false recall, in
which verbalized encoding processes predict the likeli-
hood of false recall.
EXPERIMENT 1
The purpose of Experiment 1 was to investigate the
likelihood of false recall when contextual organization
of list structure was systematically manipulated. To reit-
erate, if false recall is due to individualselaborative en-
coding of list words and critical lures, as indicated from
levels of processing approaches to the DRM effect (Toglia
et al., 1999), then the effects of experimentally induced
contextual organization of list words should differentially
influence the probability of false recall. Specifically, it
was predicted that the CU list structure would lead to a
lower rate of false recall than would all other list structures,
since this list structure was designed to induce an encod-
ing context unrelated to the critical lure. Additionally, it
was predicted that the CU list structure would lead to the
highest level of correct recall relative to all other list struc-
tures, because the structural organization of the CU list
should lend itself to more distinctive encodings of the list
items as opposed to more elaborative, error-prone encod-
ings of a list theme. Furthermore, as words associated to
the lure become distributed across list structures (eights,
fours, twos, ones), it was hypothesized that probabilities
of true and false recall should decrease.
Method
Participants. A total of 64 undergraduates participated in the
experiment to meet course requirements for an introductory psy-
chology course. The participants were tested in a computer lab at
computer terminals. They were tested in groups ranging in size
from 10 to 20 participants.
Materia ls . Ten 16-item word lists were created for use in the ex-
periment, each list containing the f irst 8 highly associated words
drawn from the original lists prepared by Roediger and McDermott
(1995). Selected lists included bread, chair, doctor, foot, mountain,
needle, rough, sleep, soft, and sweet. Eight filler words were cho-
sen for each list by the experimenters in the formation of the CU list
structure. Fillers were chosen with the aid of an Oxford Wordfinder
(Reader’s Digest Editorial Staff, 1996) due to their association or
relation with a selected word from the original 8 associates, but not
to the overall theme or critical lure. For example, the associate hard
was paired with the f iller hat, which has little or no relation to the
critical lure soft. Following the selection of all f illers, lists were or-
dered according to the f ive structures outlined earlier, each main-
taining the same sets of words across the manipulation (see Table 1).
Four random orders for the presentation of list structure and four
random orders of list themes (e.g., sleep, chair , soft) were created,
after which the total presentation was counterbalanced to control
for order effects.1Two additional lists were prepared as practice
lists to familia rize the participants with the experimental proce-
dures. Unlike the 10 target lists, the training lists contained words
unrelated to each other or to an overall list theme.
Design and Procedure . The present experiment was designed
as a replication and extension of Robinson and Roediger’s (1997)
study and used the same procedures whenever possible. The par-
ticipants were presented with 12 lists of 16 words each. List struc-
ture was manipulated as a within-subjects variable; the participants
were presented two lists from each of the five list structures result-
ing in 10 experimental lists. Two practice lists were first presented
to introduce the participants to the memory task and the computer-
style presentation and were not included in the data analysis.
Instructions appeared on a computer screen and were read aloud
by the experimenter. The participants were instructed that the ex-
periment involved memory for lists of words and that they should
pay close attention to the presented words for a subsequent mem-
ory test. All word lists and instructions were presented on Power
Macintosh 7200/120 personal computers. Programs were timed to
present each word for 2 sec, followed by a blank screen for 2 sec
until the next word was presented. During a 30-sec delay from the
initial presentation of each list, several three-digit multiplication
problems were displayed for the participants to solve. After the
multiplication problems, the participants were instructed to recall
the list words and were instructed not to guess. All participants at-
tempted to recall each list for 1.5 min. The participants recorded
FALSE RECALL AND VERBAL REPORTS 809
their recall responses in booklets that were provided by the experi-
menters. This procedure was repeated for all 12 lists.
Resu lts
Within-subjects analyses of variance (ANOVAs), with
list structure as a repeated measure, were used to analyze
both true and false recall performance. Figure 1 displays
the proportion of true and false recall as a function of list
stru ctu re.
False recall. List structure significan tly i nfluenced
the likelihood of false recall of critical lures [F(4,252) =
3.42, MSe= 0.07, p< .05]. Consistent with our predic-
tio ns, the CU list st ructure led to th e lowest levels of
false recall relative to all other list structures, as illustrated
in Figure 1. Planned pairwise comparisons conf irmed that
false recall for the CU list structure (M= .09, SD = .19)
was reliably lower than false recall in all other list struc-
tures, including the ones condition (M= .19, SD = .27),
which closely matched the structure of the CU list (all
ps < .05). Additionally, all list structures, except the CU
list, produced false recall rates comparable to Robinson
and Roediger’s (1997) 9-associates, 6-fillers condition
(M= .19). A separate within-subjects ANOVA, exclud-
ing the CU list structure, was conducted to determi ne
whether false recall decreased when list associates were
distributed across the list. However, the analysis did not
produce a significant effect of the distribution of list as-
sociates across list structure as originally predicted, and,
therefore, no linear trend across list structures was ob-
served.
Correct recall. The main effect of list structure was
reliable for correct recall o f li st words (see Figure 1)
[F(4,252) = 8.83, MSe= 0.008, p< .001]. As predicted,
pairwise comparisons demonstrated that the proportion
of correct recall was highest for the CU list structure (M=
.63, SD = .15) relative to all other list structures (all ps <
.05). When the data for the CU list structure were ex-
cluded, another within-subjects ANOVA still revealed a
significant main effect for list structure [F(3,189) = 3.12,
MSe= 0.007, p< .05]. To test our hypothesis that correct
recall would decrease as the words associated to the lure
were distributed across list structures, we conducted a
trend analysis with linear, quadratic, and cubic trends.
Only the linear trend for eights, fours, twos, and ones list
structures was reliably related to correct recall [F(1,189) =
7.11 , MSe= 0.008 p< .05].
Fig ure 1 . M ean proportions of t rue an d fa lse recall as a function o f list stru ctu re for the sequen tia l presenta -
tion of list items for Expe rim ents 1 and 2 and the verbal prot oco l sam ple .
810 GOODWIN, MEISSNER, AND ERICSSON
Summa ry. The proportions of both true and false re-
call were significantly influenced by manipulations of
the contextual organization of list structure, particularly
with regard to the CU list structure. All list structures
(except the CU list structure) led to representative levels
of false recall with means ranging from .19 to .24, com-
parable to Robinson and Roediger’s findings for 9 items
and 6 f illers (M= .19). The CU list structure that con-
tained filler items selected to bias the semantic encoding
of list items away from the critical lure produced reliably
lower levels of false recall than all other list types, in-
cluding the ones list structure, which had similar spacing
of list associates. The elevated level of correct recall for
the CU list structure demonstrates that the reduced level
of false recall is not an artifact of poor encoding of list
items in memory but suggests a more qualitative change
in the semantic encoding of the list items. The theoretical
implications of a dissociation between true and false re-
call in the CU list structure will be discussed later in the
context of theoretical explanations for false recall.
EXPERIMENT 2
The results of E xp eriment 1 demonstrated that it is
possible to manipulate t he enc od ing c ontext of words
and dramatically reduce the frequency of false recall. As
part of our second approach to study the encoding and re-
trieval processes of the DRM effect, we collected addi-
tional think-aloud protocols in Experiment 1 (to be dis-
cussed in more detail later). These protocols showed that
the participants often elaborated on list items as soon as
the words appeared on the computer screen and began to
encode associations to the presented words before the
next item appeared on the screen. As a result, some par-
ticipants verbalized elaborate associations to the critical
lure before the biasing filler item in the CU condition ap-
peared on the screen and had the opportunity to influence
the memory encoding of words associated to the lure. To
further control contextual encoding of list items in the CU
list structure, we presented list associates (e.g., hard) to-
gether with the biasing filler item (e.g., hat) in a paired
presentation format. All other list structures were also
presented in a paired presentation format. Again, we pre-
dicted that the CU list structure would lead to the lowest
level of false recall and the highest level of correct recall
relative to all other list structure conditions. Furthermore,
we predicted that, as list items become distributed across
the eights, fours, twos, and ones list structures, false recall
and correct recall should decrease. In an attempt to trace
encoding and retrieval processes in the DRM paradigm as
in Experiment 1, additional participants were asked to
“think aloud” using the procedure in Experiment 1. These
results will be discussed in a separate section.
Method
Participants . Sixty-four undergraduates participated in the ex-
periment to meet course requirements for an introductory psychol-
ogy course. They were tested in a computer lab at computer termi-
nals in groups ranging in size from 10 to 20 participants.
Ma terials, Design, an d Procedure. The word lists and ran-
domization schedule used were the same as in Experiment 1, with
the exception that list items were paired rather than presented se-
quentially. Word pairs were presented for 4 sec, followed by a 4-sec
pause before another word pair was shown. This timing was equiv-
alent to the total presentation time for the same two words presented
sequentially in Experiment 1. Again, list structure was manipulated
as a within-subjects factor, in which each participant was exposed
to each of the five list structures on two occasions. All other aspects
of the procedure remained the same as in Experiment 1.
Results
Within-subjects ANOVAs, with list structure as a re-
peated measure, were used to analyze both true and false
recall performance.
False recall. As shown in Figu re 1, resul ts demon -
strated that list structure significantly influenced the like-
lihood of false recall of critical lures [F(4,252) = 8.32 ,
MSe= 0.05, p< .001]. Consistent with our original hy-
potheses, pairwise comparisons indicated that the likeli-
hood of false recall was significantly reduced, if not vir-
tually eliminated, in the CU list structure (M= .02, SD =
.09) relative to all other list structures (all ps < .01). These
findings support the original hypothesis that the CU list
structure should lead to a decrease in false recall of criti-
cal lures. Consistent with Experiment 1, all list structures
except the CU list structure produced false recall rates
comparable to those observed by Robinson and Roediger
(1997) for their 9-associates, 6-fillers condition (M= .19).
To test our prediction that false recall would decrease
as list associ ates became distri buted across list stru c-
tures, we conducted a separate within-subjects ANOVA
with the CU list structure excluded. There was a signifi-
cant main effect for list structure [F(3,189) = 2.70, MSe=
0.06, p< .05]. A trend analysis for linear, quadratic, and
cub ic trends showed that only the linear trend for t he
eights, fours, twos, and ones conditions was significant
[F(1,189)= 7.44, MSe= 0.06, p< .01]. Hence, false recall
decreased as list associates were distributed across list
structure conditions.
Correct recall. As shown in Figure 1, results also in-
dicated a significant main effect for list structure for cor-
rect recall of list words [F(4,252) = 26.05, MSe= 0.007,
p< .001]. As predicted, pairwise comparisons demon-
strated that the proportion of correct recall was highest
for the CU list structure (M= .68, SD = .13), which was
reliably higher than all other list structures (all ps < .05)
including the eights condition (M= .62, SD = .13;
p< .05).
When the data from the CU list structure were excluded,
a separate within-subjects ANOVA revealed a significant
main effect for li st str ucture [F(3,189) = 9.01, MSe=
0.007, p< .001]. As predicted, there was a reliable lin-
ear trend, which demonstrated that, as list items became
distributed across eights, fours, twos, and ones list struc-
tures, correct recall decreased [F(1,189) = 18.73, MSe=
0.007, p< .001]. However, this was further qualified by
a significant cubic trend [F(1,189) = 8.14, MSe= 0.004,
p< .01].
Summ ary. As in Experiment 1, the proportions of both
true and false recall were significantly influenced by ma-
FALSE RECALL AND VERBAL REPORTS 811
nipulations to the contextual organization of list items,
par ticularly wi th regard to the CU list structure. F ur-
thermore, with a paired presentation of list items, there
were decreases in both true and false recall with the dis-
tribution of list items across list structures. Again, a dis-
sociation between true and false recall was demonstrated
in the CU list structure. Namely, for the CU list structure,
correct recall exceeded that of the eights list structure, yet
false recall was virtually eliminated. Again, these data
suggest a qualitative change in the semantic encoding of
list items, particularly in the CU list structure condition,
which will be discussed later.
VERBAL PROTO COL EX PERIMENT
In Experiments 1 and 2, we found that experimental
manipulations designed to decrease explicit access to the
critical lure during the encoding of list items reduced the
level of false recall. In our second approach to the study
of false rec all in t he DRM paradigm, we assessed the
cognitive processes associated with false recall by in-
structing different samples of participants to “think
aloud” during both the study and recall of the presented
lists using the experimental procedures from either Ex-
periment 1 or Experiment2. If the participants explicitly
generated the critical lure during study by producing en-
codings of list items based on associations to it, we would
expect that participants would be more likely to verbalize
the lexical term or some semantic referent as part of their
concurrent verbalizations. The think-aloud participants
studied list items with the same presentation rates and
formats as in Experiments 1 and 2, with the additional in-
struction to verbalize their spontaneous thoughts during
the encoding and recall of list items. Given that the think-
aloud participants were instructed to verbalize their
thoughts in real time, one cannot assume that all thoughts
are verbalized. The potential incompleteness of verbal
reports limits theoretical claims to the structure and fre-
quency of the verbalized thoughts. Thus, we only make
inferences of encoding processes based on the thoughts
that the participants verbalized during encoding.
We proposed that the think-aloud protocols obtained
during the participants’ study would reflect some of the
encoding and retrieval processes that eventually result in
false recall of critical, nonpresented lures. Furthermore,
assessment of on-line verbal protocols during encoding of
word lists in the DRM paradigm would provide rich data
for predicting false recall. Roediger et al. (1998) noted
that the DRM effect may occur because “subjects are con-
sciously, effortfully processing and rehearsing these non-
presented items during the study phase” (p. 212). There-
fore, the present investigation assessed three aspects of
the think-aloud protocols during encoding: (1) type of
encoding strategy, (2) whether or not the critical lure was
verbalized during encoding, and (3) the number of unique
contexts in which the critical lure was verbalized. At re-
call, think-aloud protocols were also coded for (4) whether
or not the critical lure was verbalized during recall and
(5) whether or not the critical lure was falsely recalled. Ap-
pendix A provides a brief summary of each coding vari-
able. Two of the more subjective coding categories (en-
coding strategy and contextual variability) are discussed
in more detail below.
Two major types of verbal reports are expected to occur
during the study of list items in any memory experiment,
including the DRM paradigm: semantic elaboration and
verbatim rehearsal. For example, when participants think
about the presented words for the purpose of memorizing
them, they may create a complex context that links the
concepts in a meaningful manner (semantic elaboration),
or they may merely repeat the words to themselves with-
out any additional semantic information (verbatim re-
hearsal). Unfortunately, it is possible that participants en-
code words semantically yet still only verbalize the words
verbatim. On the other hand, if the participants merely re-
hearsed the items verbatim, they would not verbally re-
port approp ria te contexts for the words. Fol lowing the
practice of coding verbal reports, we distinguished be-
tween two encoding strategies used by the participants.
We defined the category of semantic elaboration as ver-
balizations involving the on-line elaboration of list items
in which information beyond or other than the list items
themselves was present (e.g., “putting thread on th e
sharp pin”). Although the verbalization of the critical lure
during encoding may constitute a semantic elaboration in
and of itself, we examined the overall encoding strategy
used by the participants. We defined the category of ver-
batim rehearsal as verbalizations of list words without
information other than the list items themselves (e.g.,
thread, pin, sharp”). Additionally, each time the critical
lure is verbalized during encoding, it is possible that it may
emerge in a variety of different contexts. For example, the
critical lure (e.g., needle) could be verbalized in associa-
tion with several different list items (e.g., “needle and
thread” and “sharp point on the needle”) or with novel se-
mantic information (e.g., “stepped on a needle yesterday”).
Given our assumption that false recall is primarily a
consequence of associations made between list items and
critical lures generated durin g the presentation of list
items, we developed the following hypotheses for the ver-
bal reports. First, individuals who more frequently en-
gage in semantic elaboration of list items (i.e., verbal re-
ports reflect semantic elaboration) should be more likely
to explicitly access and verbalize the critical lure during
encoding than should individuals whose verbal reports
primarily consist of verbatim rehearsal of list items. Sec-
ond, verbalization of the critical lure during encoding
would suggest the association of the lure to one or more
of the list items and thus predict a higher probability of
access to the lure (as reflected by its verbalization) dur-
ing retrieval. Consequently, the likelihood of false recall
of the lure should increase on the basis of the degree of se-
mantic elaboration and number of different contexts cre-
ated during encoding.
In our protocol experiment, there were 16 participants
who thought aloud under the conditions of Experiment 1
812 GOODWIN, MEISSNER, AND ERICSSON
and an additional 16 participants who thought aloud under
the conditions of Experiment 2. The focus of our verbal-
report experiment was to assess the relationship between
enc od ing processes during study of list items and the
likelihood of false recall. This analysis of the production
of false recall stands in contrast to our manipulations in
Experiments 1 and 2, which attempted to reduce the level
of false recall across list structure conditions. Given that
we were successful in replicating the results of our list
structure manipulations across the first two experiments,
our aim in the present experiment was only to provide
data with which to examine variables that might predict
generation of the critical lure during encoding and re-
trieval processes. Therefore, we were willing to accept
that the estimated power to reproduce both of the effects
of the experimental manipulation on false and correct re-
call was low (17% and 61%, resp ec tively, fo r Experi-
ment 1; 14% and 60%, respectively, for Experiment 2).
Given that our predictions for the relation between ver-
balized encoding processes and false recall were equiva-
lent for the methods used in Experiments 1 and 2, we com-
bined the two data sets to increase power.
Method
Participan ts . A total of 32 introductory psychology students
participated in the protocol experiment and received class credit for
their participation. In contrast to the group administration used in
Experiments 1 and 2, all participants were tested individually under
the direct supervision of an experimenter.
Materials and Procedure. The materials and procedures were
the same as in Experiments 1 and 2. Sixteen participants received
the sequential presentation of list items as in Experiment 1, and 16
participants received the paired presentation of list items as in Ex-
periment 2. The rate of presentation of words and other constraints
on recall were identical to those of Experiments 1 and 2. All par-
ticipants were instructed to “think aloud” or to verbalize their
thoughts during the presentation of the word lists as well as during
recall (Ericsson & Simon, 1993). More specifically, the participants
were instructed:
As you are studying and recalling the list words, we would like for you
to “think aloud” by vocalizing your thoughts. This process is quite sim-
ilar to the exper ience of “talking to yourself and s aying ever ything
that comes to mind as you work on the tasks. You should not attempt to
explain your thinking, but rather to vocalize your thoughts as they occur
to you. Try to do your best to simply verbalize your thoughts as they
occur.
The participants were then given a series of warm -up tasks to acquaint
them with thinking aloud. If the participants stopped talking during the
warm-up tasks or during the experiment for more than 10 sec, they were
instructed by the experimenter to “keep talking.” The participants’ ver-
bal reports were tape-recorded and transcribed verbatim.
Two raters were presented with coding instructions and were
asked to code each of the 320 protocols. The encoding and retrieval
portions of each protocol were separated and coded blindly, such
that the raters could not draw inferences when coding the retrieval
section based on their knowledge of the encoding verbalizations
(and vice versa). A third rater coded each of the protocols in the
same manner, such that potential disagreements between the two
initial raters could be resolved. All raters were blind to the nature
and hypotheses of the experiment. Agreement rates between coders
ranged from .89 to 1.0. Kappa coefficients, which assess propor-
tionate agreement rates between coders by taking into account
chance agreement, were within the good to excellent range of
between-coder agreement for all coding categories (Cooper &
Hedges, 1994). Both agreement rates and Kappa coefficients for all
coded variables are found in Appendix A.
Results
Because we successfully replicated our experimental
findings across the first two experiments, recall perfor-
mance data for sequential and paired presentations of list
items were combined. Within-subjects ANOVAs, with
list structure as a repeated measure, were used to assess
bot h true a nd false re call performance (see Fi gu re 1).
Consistent wi th power calculat ions (36% fo r the com-
bined sample of 32 participants), list structure failed to re-
liably influence false recall [F(4,124)= 1.18, MSe= 0.05,
n.s.]. However, the average frequ ency of false recall was
17%, which is not reliably different fro m that found in
Robinson and Roediger’s (1997) 9-associates, 6-fillers
condition.
As expected from our power calculations for correct
recall (99% for the combined sample), there was a sig-
nificant main effect for list structure for correct recall
[F(4, 124) = 8.70, MSe= 0.007 , p< .001 ]. Consisten t
with Experiments 1 and 2, pairwise comparisons indi-
cated that correct recall in the CU list structure (M= .70,
SD = .14) was significantly higher than that in all other
list structures (ps < .01). When the data from the CU list
structure were excluded, there was no main effect for list
structure [F(3,93) = 1.43, n.s.], and no further trend analy-
ses were conducted.
Given that our analyses of the verbal reports focused
on the production of false recall, the most important
finding was that the proportion of false recall for the ver-
bal report sample was roughly comparable to the average
probability of false recall for Experiments 1 and 2 com-
bined. Specifically, the average probabilities for false
recall for the CU list structures were 7.8% for the verbal-
report sample and 5.1% for Experiments 1 and 2 com-
bined. For the other lists, the average probabilities of false
recall were 14.5% for the verbal-report sample and 19.2%
for Experiments 1 and 2 combined. Likewise, for correct
recall, the corresponding probabilities of correct recall
for the CU list structure and all other list structures were
60.1% and 69.8% for the verbal-report sample and 57.0%
and 65.5% for Experiments 1 and 2 combined. There-
fore, we further analyzed verbal reports collected for the
relationships between encoding processes and the likeli-
hood of false recall, given our experimental manipulations.
Path M odels of Encoding Proc esses
in False Recall
Sample verbal protocols for encoding and recall can
be found in AppendixB. Two modeling approaches were
taken to analyze the protocol data. First, we used a trial-
based approach, in which we estimated relationships
among coded variables across all list learning instances or
across all list learning trials (320 list protocols; Model 1
in Figure 2). From our process model of false recall we
hypothesized a sequential process in which one observ-
FALSE RECALL AND VERBAL REPORTS 813
able state, such as verbalization of the lure, would influ-
ence subsequent states and, eventually, the observable out-
come of false recall. However, this trial-based approach
did not control for the potential dependence between at-
tempts by the same participants, as a path analysis treats
each instance as an independent observation. Therefore,
this analysis based on list-learning trials cannot be used
to infer generalizability across a pattern of results to the
population of participants. To adequately test for the gen-
eralizability of the relationship between the variables, we
subsequently report the results of a second path model
(Models 2 and 2-R in Figure 2), in which each participant
contributes a single estimate for each variable—namely,
the average across the 10 list-learning trials. In a final
analysis, we assessed the merits of a process model by ex-
amining for a given participant whether the probability
of false recall on a given memory trial is higher when the
same participant has verbalized the lure during encoding.
Our path models were developed using the following
fou r c oding variables : (1) type of en coding st rategy,
(2) the encoding factor, which combined the categories
of whether or not the critical lure was verbalized during
encoding and the number of unique contexts in which the
critical lure was verbalized, (3) whether or not the critical
lure was verbalized during recall, and (4) whether or not
the critical lure was falsely recalled. Predictions were de-
rived a priori on the basis of a simplistic process model
of false recall. Specifically, use of a semantic elaboration
strategy at encoding was predicted to increase the likeli-
hood of the verbalization of the critical lure combined with
the number of unique encoding contexts of the lure (en-
coding factor). Furthermore, the encoding factor was hy-
Fig ure 2. Pat h models of false recall based on verba l prot ocol data . Recta ngu lar va riab les represen t ob -
served measu res ta ken from the pa rtici pa nts’ ver bal reports. Circu lar variables (encodin g factor) repre-
sent the combined i nfluence of th e two enco din g variables—n am ely, the nu mber o f u ni que co ntexts in
which the lure was verb aliz ed an d wh ethe r or n ot the lure w as verbal ized during encoding.
814 GOODWIN, MEISSNER, AND ERICSSON
pothesized to predict the formation of memory traces in-
volving the critical lure and, consequently, verbalization
of the lure at retrieval and subsequent false recall.2
Mode l 1: Tria l-based a pproac h. Pearson correla-
tions for the trial-based approach (n= 320) are presented
in Table 2. As displayed, correlations supported the pre-
vious hypotheses with regard to magnitude and appropri-
ate direction. Model 1 was developed using the four cod-
ing variables to more directly assess the hypothesized
relations between variables. The results of our instance-
based approach to modeling the verbal protocol variables
are displayed in Figure 2. The goodness-of-fit test, re-
ported as a chi-square statistic,3demonstrated good model
fit to the data [c2(3) = 7.11, p= .07]. The normed fit index
(NFI) and the comparative fit index (CFI) also demon-
strated good fit (.98 and .99, respectively). Furthermore,
no standardized residuals exceeded the .10 criterion for
concern. All model coefficients were statistically signif-
icant (Zs > 1.96, p< .05), and relationships between the
variables were in the predicted direction. Specif ically, the
model suggests that when the participants used an en-
coding strategy that involved semantic elaboration of list
items, they were more likely to verbalize it during encod-
ing and in a variety of different contexts. When the par-
ticipantsverbalized the lure at encoding, they were more
likely to verbalize it at retrieval, which was associated
with a higher level of false recall. In order to test the gen-
eralizability of this pattern of results, we analyzed the
average probability of the outcomes for each participant.
Models 2 /2-R: Participant-based approach. Pear-
son correlations for participant-based approaches are dis-
played in Table 3. Again, correlations were significant
and in the appropriate direction, consistent with our orig-
inal predictions. A second path model (Model 2) was used
to more directly assess the hypothesized relations between
variables. Figure 2 provides the results of the analysis of
the independent data for the second path model. The
goodness-of-fit test for this model demonstrated poor fit
to the data [c2(3) = 8.56, p< .05]. The NFI and CFI in-
dices also demonstrated relatively poor fit (.81 and .86,
respectively). Nevertheless, all of the model coefficients
were statistically significant (Zs > 1.96, ps < .05), and
relationships between the variables were in the predicted
direction. Discrepancies between the observed and repro-
duced covariances resulted in two standardized residuals
of .27 and .21, exceeding the .10 criterion for concern. A
Table 2
Cor relatio ns (Trial-Based App roach) Am on g Verbal Protoco l Coding Categor ies
Acros s Sequen tia l and Paired Presentatio n o f List Items (n= 320 )
Overall Lure Verbalized Number of Con texts Lure Verbalized Lure
Encodi ng Strategy at Encodi ng Lure Was Stated at Retrieval Falsely Recalled
Overall
Encoding Strategy .262** .239** .172* .077
Lure Verbalized
at Encoding .881*** .404*** .319**
Number of Contexts
Lure was Stated .430* ** .381* **
Lure Verbalized
at Retrieval .767* **
Lure Falsely Recalled
Encoding Factor .236** .960*** .956*** .402*** .349**
Note—“Encoding Factor” represents the combin ed influence of the “Lure Verbalized at Encodi ng” and the “Nu mber of
Contexts Lure Was Stated” coding categories. *p< .05. **p< .01 . ***p< .001.
Table 3
Correl atio ns (Participa nt-Based Approach) Am on g Verbal Pro toco l Codin g Cat egories
Across Sequen tia l and Paired Presentatio n of List Items (n= 32 )
Overall Lure Verbalized Number of Con texts Lure Verbalized Lure
Encodin g Strategy at E ncoding Lure Was Stated at R etrieval Falsely Recalled
Overall
Encoding Strategy .411* .403* .371* .196
Lure Verbalized
at Encoding .954** .458** .472**
Number of Contexts
Lure Was Stated .473** .528**
Lure Verbalized
at Retrieval .784**
Lure Falsely Recalled
Encoding Factor .360* .930** .941** .348 .443*
Note—“Encoding Factor” represents the combin ed influence of the “Lure Verbalized at Encodi ng” and the “Nu mber of
Contexts Lure Was Stated” coding categories. *p< .05. **p< .01 .
FALSE RECALL AND VERBAL REPORTS 815
revised model (Model 2-R in Figure 2) was constructed
in which two paths were added, representing the large
residuals in the initial model.
The goodness-of-fit test for Model 2-R resulted in a
fail-to-reject decision, indicating good model fit [c2(1) =
1.4 4, n.s.], and a significant improvement in f it when
compared with the initial hypothesized model [c2(2) =
7.1 2, p< .05]. The NFI and CFI indices also demo n-
strated good model fit (.97 and .99, respectively). Three
of five of the model coefficients were statistically signif-
icant as noted in Figure 2 (Zs > 1.96, p< .05), and rela-
tionships between the variables were in the predicted di-
rection. The revised model suggests that the participants
who frequently used an encoding strategy that involved
semantic elaboration of list items were more likely to ver-
balize the critical lure during encoding. Most interest-
ingly, the participants with a higher propensity to ver-
balize a critical lure at encoding were more likely to later
exhibit false recall.
The success of the Model 2-R raises doubts about our
proposed process model. If the participants were equally
likely to falsely recall the lure on a memory trial whether
or not they had verbalized the lure on that trial, then an
account based on propensities for various activities would
provide a more parsimonious account than our process
model. To evaluate that possibility, we calculated phi co-
efficients between verbalization of lure on trial and false
recall for each of the 19 participants who met the neces-
sary conditions (i.e., having at least one instance of false
recall and one instance of verbalization of lure). The av-
erage phi coefficient for these participants was .45, which
was s ignificantly g reater than ze ro [t(18) = 6.42, p<
.001]. The reliable relationship between verbalization of
the lure and subsequent false recall within participants
cannot be explained by differences between individuals
and, thus, uniquely supports our process model.
Summa ry. From a detailed analysis of verbalized en-
coding processes, we hypothesized a process model
(Model 1) that reliably predicts false recall in the DRM
paradigm. A similar path model (Model 2-R) was found
to predict the level of false recall for individuals on the
basis of their frequencies of use of encoding strategies
and verbalizations of the lure across trials. A final analy-
sis found support for our process model by showing a re-
liable relation between verbalization of the lure and sub-
sequent false recall on the same trials for each participant.
The findings from our verbal reports are thus consistent
with findings that participantsgeneration of specific se-
mantic associates to specific list items will increase false
memory (see Brainerd, Reyna, & Mojardin, 1999, for a
review). In the following section, we will discuss the im-
plications of our experimental manipulations and our path
model for theories and approaches to the study of false
recall.
DISCUSSION
In our study, two types of evidence demonstrate that
encoding processes dramatically influence the likelihood
of false recall in the DRM paradigm. First, in an experi-
mental manipulation of encoding context, false recall
was virtually eliminated when list items were selected to
induce encodings unrelated to the critical lure. Further-
more, this reduction in false recall was not an artifact of
poor recall, because the recall of correct words from the
CU list was reliably better than that of the standard DRM
lists. Second, we collected concurrent verbal reports from
a separate set of participants studying the same types of
word lists to assess the cognitive processes during en-
coding that were associated with recall of the lure at test.
We found that the participants verbally reported the crit-
ical lure during the encoding of the word lists, especially
when the verbalized encodings of list items corresponded
to phrases that cont ai ned s emantic elaboration of list
items. Overall, verbalization of the critical lure during
encoding of a particular word list predicted an increased
likelihood of false recall. We will first discuss our ex-
perimental findings in the context of current theories of
false recall in the DRM paradigm, including Underwood’s
(1965) theory of implicit associative response (IAR) and
fuzzy-trace theory (FTT; Reyna, 1998; Reyna & Brainerd,
1995; Reyna & Lloyd, 1998). Subsequently, we discuss the
implications of our verbal report models in providing a
more complete account of the processes underlying false
recall.
Reduction of False Recall and
Dissociation of True and False Recall
We found that false recall was dramatically reduced in
lists with items that were organized to guide semantic
encoding away from the critical lure. According to Un-
derwood’s (1965) IAR ap proach an d ot her theories of
spreading activation (e.g., Collins & Loftus, 1975), the
same number of semantically related list items, regard-
less of organization within a list, should show the same
level of activation of the critical lure and, thus, the same
likelihood of false recall across list structures. In our ex-
periments, our lists contained the same number of seman-
tically related list items in all list structure conditions; how-
ever, we found an increase in false recall when semantic
associates were more closely grouped together. Most in-
terestingly, the CU list structure is structurally identical
to the ones list structure, with the exception that the filler
items are arranged randomly in the ones list structure, yet
the likelihood of false is lower in the CU list structure.
Mere association of semantically related list items, as
proposed by spreading activation theories, does not fully
account for our findings, and, instead, the contextual or-
ganization of list items influences the likelihood of false
recall in the DRM paradigm.
One intriguing finding was that, within our CU list
structure condition, we found the highest levels of cor-
rect recall and the lowest levels of false recall, suggesting
a dissociation between true and false recall. When the en-
codings have been biased against such semantic relations,
as in the CU list structure conditions, false recall is vir-
tually eliminated (cf. Experiments 1 and 2). If this ma-
nipulation had reduced overall semantic processing of list
816 GOODWIN, MEISSNER, AND ERICSSON
items, then fewer semantic connections to the critical lure
and lower levels of false recall would occur, consistent
with Toglia et al.’s (1999) notion of “more is less” in the
DRM paradigm.4However, our experiments show that
this metaphor must be constrained to specific types of se-
mantic processing. Specifically, when semantic process-
ing is directed away from the lure or list theme, then false
recall of the critical lure will be reduced, even though
overall memory for list words is more accurate. The con-
textual organization of the CU list structure provides ev-
idence for just this type of semantic processing; overall
memory for list words is highly accurate, yet false recall
of critical lures is diminished.
The dissociation between true and false recall ob-
served in the CU list structure may be explained by FTT.
FTT proposes that, when individuals encode information,
they form two types of independent memory representa-
tions: verbatim representations (which include specific
perceptual features) and gist representations (which in-
clude semantic information but lack perceptual details)
(Reyna, 1998; Reyna & Brainerd, 1995; Reyna & Lloyd,
1998). In FTT, false recall of critical lures in the DRM
paradigm is a consequence of dual-retrieval mechanisms.
One retrieval mechanism involves direct access to ver-
batim traces of list items and is primarily responsible for
correct recall. The other mechanism reconstructs items
by processing gist representations and is primarily re-
sponsible for false recall.5During the study of word lists
from the DRM paradigm, semantic associations to the
critical lure are generated and encoded into the gist rep-
resentation. During subsequent recall of the list, the crit-
ical lure is retrieved by cues from the gist representation
(Reyna, 1998; Reyna & Lloyd, 1998). FTT would predict
that a gist representation of the DRM list theme is not
formed in the CU list structure condition, and, therefore,
verbatim traces of list items are accurately retrieved and
the critical lure is not. However, this account is not par-
ticularly convincing, because FTT would also predict that
activation of the gist representation would occur equally
in all other list structures, leading to equally high levels
of false recall. We found a decrease in false recall when
semantic associates were spread throughout the list.
Clearly, the contextual organization of semantically as-
sociated list items dramatically influences the likelihood
of both true and false recall. Our encoding manipulation
demonstrates the importance of encoding processes in
the DRM paradigm and has highlighted the need for cur-
rent theories to account for the contextual influence of list
items in false recall. To better understand the processes
und erly ing false recall in the DRM paradigm, we also
asked a set of participants to provide think-aloud proto-
cols during the study and recall of word lists. We now turn
to our verbal-report findings and their implications for
theories of false recall in the DRM paradigm.
Verbal Reports a nd Theories of False Recall
Our introduction of verbal-report data on the encoding
and retrieval of word lists has forced us to be more spe-
cific about the nature of different types of encoding pro-
cesses and their relationship to false recall. According to
our path models of false recall, during the presentation
and stu dy of list items, the gene rated encodings were
more likely to have been semantically related to the crit-
ical lure when false recall was observed. Our encoding
factor included whether or not the lure was stated during
encoding and showed that the frequency of verbalization
was di rectly related to the likel ihood that the lure was
falsely recalled. Below, we attempt to further examine our
verbal protocols for the different types of semantic en-
codings generated (either automatically or deliberately)
during encoding. In so doing, we believe that it may be
possible to better characterize the gist representation
proposed by FTT (Reyna, 1998; Reyna & Lloyd, 1998),
as well as to assess the specific theoretical mechanisms
that mediate access to the lure during encoding, such as
with Underwood’s (1965) IAR.
FTT describes encoding processes based on inferences
from retrieval measures, whereas our verbal reports pro-
vide concurrent accounts of encoding and retrieval pro-
cesses generated by participants in the DRM paradigm.
With regard to FTT, our verbal reports provide an oppor-
tunity to explicate some of the particular processes and
representations that reflect the formation and resultant
cueing of the gist representation. For example, when the
critical lure is verbalized during encoding, we can infer
that the gist representation has been cued and memory
traces involving associations to the lure are likely to have
been formed. Furthermore, repeated verbalization of the
lure during encoding may be considered repeated cuing
or activation of the gist representation, which is hypoth-
esized to increase false recall. This is consistent with our
encoding factor and its ability to reliably predict verbal-
ization of the lure at recall and false recall (Model 2-R).
Consistent with Collins and Loftus’s (1975) theory of
spreading activation, the critical lure is often related by
mere association to an encoded word. For example, in
the chair protocol described in Appendix B, the first ver-
balization of the lure chair may be retrieved due to word
associates on the list (in particular, the list item recliner):
“you can have a reclining CHAIR at your desk.” Consistent
with Underwood’s (1965) IAR account, we can see the
strategic and controlled use of the lure chair in the par-
ticipants rehearsal of list items and the lure: “sit, CH AIR,
recliner.” Likewise, a similar pattern may be observed in
the needle protocol described in Appendix B: Automatic
spreading activation of the presented list items sewing,
cook, and sharp may have led to the activation of needle
(“sharp is the NEEDLE”), followed by subsequent, deliber-
ate use of the lure in combination with list items (“point
on the NEEDLEand prick your finger with the NEEDLE
that you are sewing”).
Our verbal protocols contained instances in which the
participants sometimes verbalized the critical lure in dif-
ferent semantic contexts during encoding. These in-
stances are interesting theoretically for FTT, because they
imply the existence of several gist representations. Dur-
FALSE RECALL AND VERBAL REPORTS 817
ing retrieval, it may become difficult to distinguish which
words and concepts had been generated during encoding
and which had actually been presented as words on the
list. We believe that this source confusion involving mul-
tiple, overlapping semantic encodings in memory is con-
sistent with findings of increased levels of false recall
with repeated memory testing (Roediger, 1996) and main-
tained levels of false recall after delayed testing (Payne
& Elie, 1998). This finding might also account for the
difficulties of attributing the source of a retrieved memory
trace of the critical lure (Jacoby, Kelley, & Dywan, 1989;
Johnson, Hashtroudi, & Lindsay, 1993).
Taken together, our verbal protocols appear to provide
an insightful tracing of the processes underlying false re-
call. Althoug h some support for the various theories may
be observed across instances, the more formal path mod-
els have allowed us to explicate certain relationships as
predictive of subsequent false recall. We believe that these
relationships may help to clarify existing theories (i.e.,
FTT) by forcing theorists to provide a more detailed ac-
count of the proposed mechanisms underlying encoding
and recall in the DRM paradigm. Our verbal reports also
strongly suggest that the particular semantic encodings
generated are quite variable and diverse across partici-
pants. We believe that future studies of the DRM para-
digm will need to experimentally control encoding pro-
cesses during the study of list items in order to further
constrain subsequent retrieval processes and the likeli-
hood of false recall. When experimental manipulations
are coupled with process-tracing methods, such as verbal
reports, our understanding of this complex phenomenon
of false recall will be significantly advanced.
Future Directions
We believe that any theoretical framework that would
be sufficient to explain the full range of findings relevant
to the DRM paradigm will have to be quite complex and
will have to accommodate many different types of en-
coding processes, as well as various kinds of retrieval and
monitoring p rocesses. Our mani pulation of c ontextual
encoding demonstrated unique effects both in reducing
false recall and in illustrating a dissociation between true
recal l an d false recall. By ex am ining th e ve rb ally re-
ported thoughts of the participants engaged in the DRM
task, we have also attempted to identify more precisely
several of the mechanisms involved in false recall. De-
spi te our experimental manipulatio n th at reduced the
overall level of false recall and the limitation of the in-
completeness of verbal reports (cf. Ericsson & Simon,
1993), it appears that various aspects of the aforemen-
tioned theories are at least partially consistent with ver-
bal reports collected in our experiments. However, a
complete model of the processes mediating false recall
will require a more explicit account of the mechanisms
involved, including an ability to generate testable pre-
dictions and to support recent findings in the DRM par-
adigm. Here, we have attempted to demonstrate the use-
ful ne ss o f coupl ing experimental and pro ce ss-tracing
methods in critically examining current theories. Future
research might benefit from furthering such an approach,
leading to a more complete account of the encoding and
retrieval processes underlying false recall.
REFERENCES
Bentler , P. M. (1995). EQS: Structural equations program manual.
Los Angeles: BMPD Statistical Soft ware.
Brainerd, C. J., Reyna, V. F., & Mojardin, A. H. (1999). Conjoint
recognition. Psychological Review, 106, 160-179.
Collins, A.M., & Loftus, E. F. (1975). A spreading-activation theory
of semantic memory. Psychologica l Review, 82, 407-428.
Cooper, H., & Hedges, L. V. (1994). The h andb ook of research syn-
thesis. New York: Russell Sage Foundation.
Deese, J. (1959 ). On the predict ion of o ccurren ce of particular verb al
intru sions in immediate recall. Jou rnal of Experimenta l Psycholog y,
83, 17-22.
Ericsson, K. A., & Simon, H. A. (1993). Protoco l analysis: Verbal re-
ports as data (rev. ed.). Camb ridge, MA: MIT Press.
Jacoby, L. L., Kelley, C. M., & Dywan, J. (1989). Memory attribu-
tions. In H. L. Roediger III & F. I. M. Craik (Eds.), Varieties of mem-
ory and consciousness: Essays in honour of Endel Tulving (pp. 391-
422). Hillsdale, NJ: Erlb aum.
Johnson, M. K., Hashtroudi, S., & Lindsay, S. (1993). Source mon-
itorin g. Psychological Bulletin, 114, 3-28.
McDermott, K. B. (1996). The persistence of false memories in list re-
call. Journa l of Memory & Language, 35, 212-230.
McDermott, K. B. (1997). Priming on perceptual implicit memory
tests can b e achieved thro ugh presentation of asso ciates. Psycho-
nomic Bull etin & Review, 4, 582-586.
Norma n, K. A., & Sc hacter, D. L. (19 97). Fal se recog nition in
younger and older adults: Exploring the characteristics of illusory
memories. Memory & Cogn ition, 25, 838-848.
Payne, D. G., & Elie, C. J. (1998, November). Repeated list presenta-
tion reduces false memories for pictures and words. Paper presented
at the annual meeting of the Psychon omic Society, Dallas.
Payne, D. G., Elie, C. J., Blackwell, J. M., & Neuschatz, J. S. (1996).
Memory illusions: Recalling, recognizing, and recollecting events that
never occurred. Journal of Memory & Language, 35, 261-285.
Read, J. D. (1996). From a passin g thought to a false memory in 2 min-
utes: Confusing real and illusory events. Psychonomic Bulletin & Re-
view, 3, 105-111.
Reader’s Digest Editorial Staff (Ed.) (19 96). Reader’s Digest Ox-
ford c omplete wordf inder. Pleasantv ille: Reader’s Dig est Associ a-
tion, Inc.
Rey na, V. F. (199 8). Fuzzy-t race theory and false memory. In M. J.
Inton s-Peterson and D. L. Best (Eds.), Challenges and controversies
in appli ed cognition: Memory disto rtions and their prevention
(pp. 15-27). Mahwah, NJ: Erl baum.
Rey na, V. F., & Brainerd, C. J. (1995). Fuzzy-trace theory: An interim
synthesi s. Learning & Individu al Differences, 7, 1-75.
Rey na, V. F., & Lloyd, F. J. (1998, November). Fuzzy-trace theory and
source-moni toring explanations of false memories. Paper presented
at the annual meeting of the Psychon omic Society, Dallas.
Robinson, K. J., & Ro ediger, H. L., III (1997). Associative processes
in false recall and false recognition. Psychological Scien ce, 8, 231-
237.
Roediger, H. L., III (1 996). Memory illusion s. Journal of Memory &
Language, 35, 76-100.
Roediger, H. L., III, & McDermott, K. B. (1995). Creating false mem-
ories: Remembering words not presented in lists. Journal of Experi-
mental Psychol ogy: Learning , Memory, & Cognition, 21, 803-814.
Roediger, H. L., III, McDermott, K. B., & Robinson, K. J. (1998 ).
The role of associative processes in creating false memories. In M. A.
Conway, S. E. Gathercol e, & C. Cornoldi (Eds.), Theories of mem-
ory: Volu me II (pp. 187-245). London: Psychology Press.
Schact er, D. L., Verfael lie, M., & Pradere, D. (1996). The neu-
ropsy chology of memory illus ions : False recall and recognition in
amnesic patients. Journal of Memory & La nguag e, 35, 319-334.
818 GOODWIN, MEISSNER, AND ERICSSON
Toglia, M. P., Neuschatz, J. S., & Goodwin, K. A. (1999). Recall ac-
curacy an d illusory memories: When more is less. Memory, 7, 233-
256.
Underwood, B. J. (1965). False recognition produced by implicit ver-
bal respon ses. Journal of Experimental Psychology, 70, 122-129.
NOTE S
1. This proced ure resulted in a 4 (random order of list structu re) 34
(random or der of list themes) 32 (wit hin-subjects factor of two expo-
sures of list structure) randomization schedule in which 2 participants
were run in each cell of the design. Within each cell , the participants
were randomly exposed to each list structur e (eights, fours, twos, ones,
and CU list s) on two occasions, resulting in 10 studied lists. Addition-
ally, the schedule ensured that themed lists (e.g., sleep, chair, mountain,
etc.) were randomly presented in each list structur e across participan ts.
2. Due to the high degree of correlation between verbalization of the
critical lure at encod ing and the number of u nique contexts generated
(r= .954, p< .01), these variables were combined to yield a single fac-
tor score using principle components analysis. Although we believe
these facto rs to be rather distin ct constructs, their hig h degree of relat-
edness would have caused a degree of suppression or shifting in the
model weights, thus masking their true relation with the outcome vari-
ables. The resultant factor score that combined the two variables
demonstr ated a v ery simi lar pattern of correlation with the other vari-
ables in the model, includ ing overall encodin g strategy (r= .360, p<
.05) and false r ecall (r= .443, p< .05).
3. In discussing the results of a path model, a goodness-of-f it test, re-
ported as a chi-square statistic, is typically used to indicate how well the
sample variance– covariance matrix is reprodu ced as a function of the
parameter estimates of the proposed model. An obtained chi-square
value with a pr obabi lity level greater than a standard cut-off, su ch as
.05, indicates adequate fit of the model to the sample. In contrast, a sta-
tistical ly si gnif icant chi-square value implies rej ection of th e mo del.
Because th e chi-squ are statisti c is sensitive to sample size, we also re-
port the r esults of two adjunct indicators of model fit: the normed f it
index (NFI) and the comparative f it index (CFI). Both ind icators can
range from 0 to 1, with values of .90 or greater indicatin g good model
fit (Bentler, 1995).
4. The “more is less” metaphor was used to illust rate the finding that
conditi ons that increase the likelih ood of cor rect recall (e.g. , semantic
processing and blocked present ation) simultaneous ly increase false re-
call. Thus, “more” studied and nonstudied items are remembered, such
that overall remembering is “less” accurate.
5. It should be noted that because there are strong interitem semantic
relations made between encoded list items in the DRM paradigm , later
list items are likely to provoke retrieval of earlier list items, whi ch may
contrib ute greatly to th e formation of the gist representation .
APPENDIX A
Protocol Coding Categories, Category Descriptions, Agreement Rates (AR), and Kappa Coefficients
Observed for Sequential and Paired Presentations in the Verbal Protocol Experiment
AR Kappa
Description Sequential Paired Sequential Paired
Encoding
Overall encoding strategy A general assessment of the overall encoding .963 .937 .925 .872
strategy used by subjects, either semantic
elaboration, in which subjects went beyond
the literal presentation of list items, or
verbatim rehearsal, which included only
literal organization of list items and lures
Statement of lure Yes or No (was the lure stated during
encod ing?)
Overall encoding A general assessment of the encoding .959 1.00 .958 1.00
strategy of the lure strategy of verbalizations of the lure, either
semantic elaboration or verbatim rehearsal
Number of different The number of different, unique encoding .889 1.00 .727 1.00
contexts of lure contexts in which the lure was stated
(if mentioned more
than once)
Retrieval
Statement of lure Yes or No (was lure stated during retrieval?)
False recall of lure Yes or No (was lure falsely recalled?)
FALSE RECALL AND VERBAL REPORTS 819
APPENDIX B
Verbal Protocol Samples
A typical protocol from the eights condition is provided below (italicized items during encoding are presented list
words read by the participants from the computer, and italicized items during retrieval are recalled words as they were
written by the participants).
Encoding
table\ I like to sit on top of tables\ I’m more comfortable\ tables have four legs\ you sit on them\ on table there
are actual seats to sit on\ not at the table\ couch\ I like to sit on the couch\ it’s more comfortable than the table\
you sit at the desk\ which is a type of table\ a recliner\ you can have a reclining CHAIR at your desk \ sofas the
same thing as a couch\ the contents of the desk\ which has a recliner\ my desk is very tight full of stuff\ I put
my hosieryon top of the table\ hosiery belt\ these words don’t relate\ I like to eat potatoes at my table\ sit CHAIR
recliner\ on my desk there’s a computer\ on my desk at the computer Jeff likes to look at the football scores\ on
my bed\ my bed is right next to the desk\
Here, we see that the participant verbalized the critical lure chair twice within the context of recliner. Interest-
ingly, while this individual engages primarily in a semantic encoding strategy, she also engages in rehearsal of the
lure.
Retrieval
table\ sit\ CHAIR\ couch\ sofa\ bed\ desk\ recliner\ footba ll\ computer\ um\ on the table I like to sit on the table\
it was comfortable\ I like to sit on the CHAIR\ I prefer to sit on the couch\ than on the table\ another word for
couch is a sofa\ next to my desk is the bed\ on the desk is the computer\ the desk has a recliner sometimes\ the
desk is tight full of stuff \ on the computer my boyfriend likes to do his football scores\ um\ let’s see\ what were
the words that did not fit with this\ that were not part of the list\ that’s about all I can remember\ sofa\ couch\
next to the desk is the bed\ on the desk is tight full of stuff\
During recall, she falsely recalls chair, and does so with sit, one of the words that she rehearsed with chair dur-
ing encoding.
The following protocol example is more representative of most encoding protocols in which semantic elaboration
occurs and the critical lure is stated.
Encoding
thread \ maneuver\ pin and thread\ head \ a pin head\ maneuver and thread\ eye\ there is eyes on your head\ pin
plus thread\ ball \ maneuver\
sewing\ pin, thread, ball\ oh gosh\
cook\ sharp\ sharp is the NE EDLE\ shooter\ I don’t
know\ point on the NEED LE\ sewing with the finger\ prick your finger with the NEEDLE that you are sewing\
thorn\ prick your finger with a thorn\ use a thimble so you don’t prick your finger\ with a thorn\ collection\ ma-
neuver\
In this encoding protocol, we again see the verbalization of the critical lure within a semantic elaboration encod-
ing strategy. Again, needle is stated in a variety of different contexts (e.g., with sharp, point, and prick).
Retrieval
there’s maneuver\ and there’s a thimble\ so you don’t have a thorn\ in you finger\ sewing with a NEEDLE and a
pin\ maneuver\ and something else\ was sewing\ pin\ maneuver\ something else\ thimble for thorn\
prick I think\
sewing the NEEDLE with a pin\ and\ something more about sewing\ and\ sewing, NEEDL E, pin\ I don’t know\
In the corresponding retrieval protocol,
needle is falsely recalled and is mentioned with related list items, but not
necessarily with the same list items in which it was encoded.
(Manuscrip t received May 3, 1999;
revision accepted for publi cation June 4, 200 1.)

Supplementary resource (1)

... However, global list structure is important, as it can orient processing towards noticing the relation among items within a list. For example, associative lists blocked by theme increase true and false memories relative to random list presentation (Brainerd et al., 2003;McDermott, 1996;Toglia et al., 1999), and intermixing unrelated filler items within blocked lists reduces false memories (Goodwin et al., 2001). Because local gist processing is equated across list format (i.e., same number of semantically related list items), this suggests that random presentation disrupts global gist processing (Goodwin et al., 2001;Lampinen et al., 2006). ...
... For example, associative lists blocked by theme increase true and false memories relative to random list presentation (Brainerd et al., 2003;McDermott, 1996;Toglia et al., 1999), and intermixing unrelated filler items within blocked lists reduces false memories (Goodwin et al., 2001). Because local gist processing is equated across list format (i.e., same number of semantically related list items), this suggests that random presentation disrupts global gist processing (Goodwin et al., 2001;Lampinen et al., 2006). Instructions that encourage relational processing among items within a list also increase false memories relative to instructions that focus on item-specific features of each list item (Mccabe et al., 2004), and distinctive processing at encoding (e.g., unique fonts) has been shown to reduce false memories (Arndt & Reder, 2003). ...
... Although semantic processing has been shown to increase memory errors in the DRM paradigm (e.g., Goodwin et al., 2001;Toglia et al., 1999), providing contextual information that differentiates items at encoding or presenting encoding instructions that direct attention to differences among stimuli can reduce false memories (e.g., Arndt & Reder, 2003;Thomas & Sommers, 2005). Based on the mechanisms described by AMT and FTT, individual sentences that converge on the theme of a critical lure (local gist) are a necessary prerequisite for the creation of false memories. ...
Article
Full-text available
Although false memories have largely been examined with the Deese-Roediger-McDermott (DRM) paradigm, little research has focused on the semantic context in which associates are encoded. Across three experiments, we varied semantic context during a sentence-processing task with DRM associates embedded within sentences. More meaningful sentences resulted in greater memory errors (Experiment 1). Furthermore, providing contextual information to discriminate old from new items did not reduce false alarms relative to encoding words in isolation when sentences converged on the meaning of the critical lure (Experiment 2), and actually increased memory errors (Experiment 3). These results suggest that semantic context that allows for meaningful relational processing of items within-lists and that converge on the semantic meaning of the critical lure increases the likelihood that the list theme is identified, resulting in more errors at test.
... If the critical lure is rehearsed as other list words are during learning, its RT profile should be the same as or similar to that of a studied word. In an attempt to answer the question of whether subjects rehearse the critical lure, Goodwin, Meissner, and Ericsson (2001) and Goodwin (2007, Experiment 1) conducted a "thinking aloud" protocol experiment and revealed that subjects indeed often rehearsed the critical lure during encoding and that the level of rehearsing the critical lure was highly correlated with the likelihood of falsely recalling the critical lure later. If the thematic word is rehearsed during the encoding stage as list words, it should be searched for in a Sternberg task just as a word in the memory set and therefore yield a linearly increasing RT/MSS function. ...
... For example, they gave high confidence ratings and "remember" judgment, and even attributed list serial positions to the critical lure (Lampinen, Meier, Arnal, & Leding, 2005;Lampinen et al., 1999;Payne et al., 1996;Roediger & McDermott, 1995;Tulving, 1985). It has been suggested that the critical lure might be generated during encoding and rehearsed with the other list words (Cramer, 1970;Goodwin, 2007;Goodwin et al., 2001;Rhodes & Anastasi, 2000;Roediger, Balota, et al., 2001;Underwood, 1965). In addition, there were conceptual and perceptual priming effects of the semantic associates on the critical lure, suggesting that subjects might be conscious of that word (McDermott, 1997; see also Seamon, Luo, & Gallo, 1998). ...
Article
Full-text available
In this study, the number of semantic associates in Deese–Roediger–McDermott (DRM) lists was varied from 4 to 14 in a modified Sternberg paradigm. The false alarm (FA) and correct rejection (CR) reaction time (RT)/memory-set size (MSS) functions of critical lures showed a cross-over interaction at approximately MSS 7, suggesting a reversal of the relative dominance between these two responses to the critical lure at this point and also indicating the location of the boundary between the sub- and supraspan MSS. For the subspan lists, FA to critical lures was slower than CR, suggesting a slow, strategic mechanism driving the false memory. Conversely, for the supraspan lists, critical lure FA was faster than its CR, suggesting a spontaneous mechanism driving the false memory. Results of two experiments showed that an automatic, fast, and a slow, controlled process could be error-prone or error-corrective, depending on the length of the DRM memory list. Thus there is a dual retrieval process in false memory as in true memory. The findings can be explained by both the activation/monitoring and the fuzzy-trace theories.
... If the critical lure is rehearsed as other list words are during learning, its RT profile should be the same as or similar to that of a studied word. In an attempt to answer the question of whether subjects rehearse the critical lure, Goodwin, Meissner, and Ericsson (2001) and Goodwin (2007, Experiment 1) conducted a "thinking aloud" protocol experiment and revealed that subjects indeed often rehearsed the critical lure during encoding and that the level of rehearsing the critical lure was highly correlated with the likelihood of falsely recalling the critical lure later. If the thematic word is rehearsed during the encoding stage as list words, it should be searched for in a Sternberg task just as a word in the memory set and therefore yield a linearly increasing RT/MSS function. ...
... For example, they gave high confidence ratings and "remember" judgment, and even attributed list serial positions to the critical lure (Lampinen, Meier, Arnal, & Leding, 2005;Lampinen et al., 1999;Payne et al., 1996;Roediger & McDermott, 1995;Tulving, 1985). It has been suggested that the critical lure might be generated during encoding and rehearsed with the other list words (Cramer, 1970;Goodwin, 2007;Goodwin et al., 2001;Rhodes & Anastasi, 2000;Roediger, Balota, et al., 2001;Underwood, 1965). In addition, there were conceptual and perceptual priming effects of the semantic associates on the critical lure, suggesting that subjects might be conscious of that word (McDermott, 1997; see also Seamon, Luo, & Gallo, 1998). ...
Article
Full-text available
In this study, the number of semantic associates in Deese-Roediger-McDermott (DRM) lists was varied from 4 to 14 in a modified Sternberg paradigm. The false alarm (FA) and correct rejection (CR) reaction time (RT)/memory-set size (MSS) functions of critical lures showed a cross-over interaction at approximately MSS 7, suggesting a reversal of the relative dominance between these two responses to the critical lure at this point and also indicating the location of the boundary between the sub- and supraspan MSS. For the subspan lists, FA to critical lures was slower than CR, suggesting a slow, strategic mechanism driving the false memory. Conversely, for the supraspan lists, critical lure FA was faster than its CR, suggesting a spontaneous mechanism driving the false memory. Results of two experiments showed that an automatic, fast, and a slow, controlled process could be error-prone or error-corrective, depending on the length of the DRM memory list. Thus there is a dual retrieval process in false memory as in true memory. The findings can be explained by both the activation/monitoring and the fuzzy-trace theories.
... One possibility could be to employ an overt rehearsal condition where participants are asked to "think out loud" and say any words or thoughts that come to mind when they are studying the items (Rundus, 1971;Rundas & Atkinson, 1970). Overt rehearsal has been used in DRM studies to explore whether participants consciously think of critical lures during encoding (Goodwin et al., 2001;Marsh & Bower, 2004;Seamon et al., 2002) and has also been used to examine rehearsal strategies in the context of directed forgetting (Hourihan, 2021). If overt rehearsal reveals that during R-cues, participants frequently generate category labels or make explicit connections with other category members, this would favour Montagliani and Hockley's account, especially if the frequency with which category labels are generated is found to be correlated with false recognition for R-category items. ...
Article
Full-text available
Montagliani and Hockley (2019) presented evidence that item-method directed forgetting not only leads to worse recognition of forget-cued targets than remember-cued targets but also better rejection of foils associated with forget-cued targets than remember-cued targets. Based on that result, they proposed that participants elaboratively encode more category-level information about R-cued targets. We present a retrieval-based explanation of the result within an instance-based memory model. The model imports word representations from two distributional semantic models, latent semantic analysis (LSA) and random permutation model (RPM), into an instance-based model of memory, MINERVA 2. The model reproduced Montagliani and Hockley's results without requiring assumptions about elaborated encoding of category-level information at study. The simulations demonstrate that whereas Montagliani and Hockley's findings are consistent with an account grounded in elaborated encoding of words at study, the results do not force that conclusion. Instead, better encoding of remember-cued targets at study establishes the conditions for retrieval-time effects at test to produce a corresponding influence on false recognition for category-related foils. Our model can be used as a formal tool to think about and study the incidental consequences of item directed forgetting in recognition memory. (PsycInfo Database Record (c) 2022 APA, all rights reserved).
... Surprisingly, the enhancement of false memories under verbal generation is not reduced when it is associated with the imagination act, as is the case in Foley's works. However, this outcome is consistent with Marsh and Bower (2004) and other authors (e.g., Goodwin, Meissner, & Ericsson, 2001;Seamon et al., 2002) who reported that verbal generation inflates false memory rates. ...
Article
This study examined the possibility that moderators of false memories in the Deese-Roediger-McDermott (DRM) paradigm affect the occurrence of false memories in the misinformation paradigm. More precisely, the purpose was to determine to what extent an imaging instruction modulates false memories in the DRM and misinformation paradigms. A sample of young adults was assigned to the DRM or the misinformation tasks, either in control conditions or in conditions including an imaging instruction. Findings revealed that an imaging instruction decreases false memories in DRM whereas there is no evidence about imaging effects in the misinformation task. These observations tally with previous studies, reporting a weak or no correlation between false memories in these paradigms, and are discussed in the light of current theories.
... Lure activation is thought to be automatic and preserved in old age ( Gallo & Roediger, 2003;Koutstaal, 2003), but can also be influenced by processing at encoding. In young adults, false recognition is lower when list items have unique as opposed to shared study context, in terms of either list blocking ( Goodwin, Meissner & Ericsson, 2001) or font ( Arndt & Reder, 2003). Such contextual effects may be greater in older adults, as reflected in greater impact of blocking or task manipulations ( Taconnat et al., 2006;Thomas & Sommers, 2005;Tun et al., 1998;see also Thomas, Bonura & Taylor, 2012). ...
Article
Full-text available
This study investigated semantic and perceptual influences on false recognition in older and young adults in a variant on the Deese-Roediger-McDermott paradigm. In two experiments, participants encoded intermixed sets of semantically associated words, and sets of unrelated words. Each set was presented in a shared distinctive font. Older adults were no more likely to falsely recognize semantically associated lure words compared to unrelated lures also presented in studied fonts. However, they showed an increase in false recognition of lures which were related to studied items only by a shared font. This increased false recognition was associated with recollective experience. The data show that older adults do not always rely more on prior knowledge in episodic memory tasks. They converge with other findings suggesting that older adults may also be more prone to perceptually-driven errors.
Article
In the Deese-Roediger-McDermott-Read-Solso (DRMRS) procedure, which has recently enjoyed widespread use, participants try to remember lists of items constructed around critical themes that are not presented. The purpose of this quantitative review was to estimate the extent to which these themes are falsely recalled (Critical Intrusions) and to compare this error rate with Correct Recall and with false recall of other words that were not presented (Noncritical Intrusions). Based on 111 estimates, the mean rate of Critical Intrusions was .374 (95% confidence interval of .344 to .403), which was a very large effect (d=2.40). Critical Intrusions were less frequent than Correct Recall (.572) but more frequent than Noncritical Intrusions (.078). Critical Intrusions were lower (.318) for people told to be confident (not to guess) than for people not so cautioned (.428), a medium to large effect size (d=0.69). Effects of other variables are summarized, and the theoretical and applied implications of these results are discussed.
Article
Full-text available
The process-dissociation model has stimulated important advances in the study of dual-process conceptions of memory. The authors review some limiting properties of that model and consider the degree of support for its parent theory (the recollection–familiarity distinction). A 2nd-generation model (conjoint recognition) is proposed that addresses these limitations and supplies additional capabilities, such as goodness-of-fit tests, the ability to measure dual processes for false-memory responses, and statistical procedures for testing within- and between-conditions hypotheses about its parameters. The conjoint-recognition model also implements an alternative theoretical interpretation (the identity–similarity distinction of fuzzy-trace theory). Worked applications to data are provided.
Article
Full-text available
Two experiments (modeled after J. Deese's 1959 study) revealed remarkable levels of false recall and false recognition in a list learning paradigm. In Experiment 1, subjects studied lists of 12 words (e.g., bed, rest, awake ); each list was composed of associates of 1 nonpresented word (e.g., sleep). On immediate free recall tests, the nonpresented associates were recalled 40% of the time and were later recognized with high confidence. In Experiment 2, a false recall rate of 55% was obtained with an expanded set of lists, and on a later recognition test, subjects produced false alarms to these items at a rate comparable to the hit rate. The act of recall enhanced later remembering of both studied and nonstudied material. The results reveal a powerful illusion of memory: People remember events that never happened.
Article
Full-text available
The frequency with which subjects erroneously included a nonpresented word in their recall of a study list was explored in two experiments. The intrusion error was recalled by as many as 80% of the subjects, and when it was perceived to have been presented early in the study list, it was assigned confidence ratings and phenomenological retrieval characteristics equivalent to those for presented words. As a result, subjects were often unable to discriminate memories of real study words from their memories of a related but nonpresented word. Manipulations of encoding, but not of retrieval, conditions altered both the frequencies of illusory memories and their metamemorial characteristics. The results and paradigm are discussed in terms of their relevance to the "memory-recovery" debate.
Book
Since the publication of Ericsson and Simon's ground-breaking work in the early 1980s, verbal data has been used increasingly to study cognitive processes in many areas of psychology, and concurrent and retrospective verbal reports are now generally accepted as important sources of data on subjects' cognitive processes in specific tasks. In this revised edition of the book that first put protocol analysis on firm theoretical ground, the authors review major advances in verbal reports over the past decade, including new evidence on how giving verbal reports affects subjects' cognitive processes, and on the validity and completeness of such reports. In a substantial new preface Ericsson and Simon summarize the central issues covered in the book and provide an updated version of their information-processing model, which explains verbalization and verbal reports. They describe new studies on the effects of verbalization, interpreting the results of these studies and showing how their theory can be extended to account for them. Next, they address the issue of completeness of verbally reported information, reviewing the new evidence in three particularly active task domains. They conclude by citing recent contributions to the techniques for encoding protocols, raising general issues, and proposing directions for future research. All references and indexes have been updated. Bradford Books imprint
Article
Roediger and McDermott (1995) recently re-introduced a paradigm to study the creation of false memories. Subjects hear short lists of related words (e.g., thread, pin, eye, sewing, etc.), all of which are associates of a critical nonpresented word (e.g., needle); on a free recall test given immediately after list presentation, subjects often erroneously recall the critical nonpresented word. The experiments reported here explore (a) the effect of test delay on false recall and (b) whether multiple study/test opportunities reduce the proportion of critical items erroneously recalled. In Experiment 1, introduction of a 2-day delay between study and test produced probabilities of false recall that exceeded those of veridical recall. In addition, prior testing of the list enhanced false recall, much like testing enhances later recall of studied items (the testing effect). In Experiment 2, an attempt was made to reduce or eliminate the false recall effect by using a multitrial study/test procedure. Although subjects were able to reduce the proportion of critical nonpresented words erroneously recalled, they were unable to eliminate the false recall effect, even after 5 study-test trials. An interaction occurred between accurate and false recall as a function of retention interval: after a one-day delay, false recall levels rose, whereas accurate recall decreased. Results of both experiments demonstrate the persistence of this memory illusion.