APPLIED COGNITIVE PSYCHOLOGY
Appl. Cognit. Psychol. 16: 1±16 (2002)
DOI: 10.1002/acp.789
Automatic and Intentional Processes in Children's
Recognition Memory: The Reversed
Misinformation Effect
ROBYN E. HOLLIDAY1* and BRETT K. HAYES2
1
2
University of Kent at Canterbury, UK
University of New South Wales, Australia
SUMMARY
This study investigated the contribution of automatic and intentional memory processes to 5- and
6-year-old children's suggestible responses in a reversed misinformation paradigm. The temporal
order of the conventional eyewitness paradigm was altered such that children were initially presented
with a pre-event narrative containing misinformation that was either read to them or was selfgenerated in response to semantic and linguistic cues, and the following day were presented with a
witnessed event in the form of a picture story. Children then completed a standard forced-choice
recognition memory test under two instruction conditions. In the inclusion condition children were
reminded about the presentations of the pre-event narrative and the original story and asked to chose
the witnessed event item. In the exclusion condition children were instructed to exclude pre-event
suggestions. Suggestibility effects were found with the magnitude of such effects differentially
affected by the encoding of misleading suggestions and test instructions. In the exclusion condition,
children were more likely to correctly reject suggestions that were `self-generated'. Both automaticity and intentional recollection contributed to children's suggestible responding. Copyright #
2002 John Wiley & Sons, Ltd.
Although it is acknowledged that both social and memory-based factors are implicated in
children's suggestibility, the nature of the underlying mechanisms responsible for these
effects continue to be debated (Bruck and Ceci, 1999; Bruck et al., 1997; Ceci et al.,
1998). Speci®cally, the fate of memories of the witnessed event has yet to be resolved.
Loftus and her colleagues (e.g. Loftus et al., 1978) have proposed a strong version of a
trace alteration account of suggestibility which holds that misled participants report
suggested details at test because their memories for the original or witnessed event details
are altered or updated by the post-event suggestions. More moderate versions of the trace
alteration model hold the view that the presentation of misinformation does not necessarily
overwrite memories for witnessed events. The partial degradation hypothesis, for example
(Belli and Loftus, 1996), argues that misleading suggestions render certain features of the
original memory less accessible.
Retrieval interference models argue that misinformation effects are due to competition
between original and post-event traces at the point of retrieval (Bekerian and Bowers,
1983; Chandler, 1991; Christiaansen and Ochalek, 1983). One in¯uential version of the
*Correspondence to: Dr Robyn E. Holliday, Department of Psychology, Keynes College, University of Kent,
Canterbury, UK CT2 7NP, UK. E-mail: [email protected]
Copyright # 2002 John Wiley & Sons, Ltd.
2
R. E. Holliday and B. K. Hayes
retrieval interference approach proposes that both the original and the post-event
misinformation traces are retained but that access to the original trace is `blocked' by
the more recently encoded misinformation. That is, when multiple traces for an item are
created, retrieval is biased in favor of the most recently presented trace (cf. Ayers and
Reder, 1998; Morton et al., 1985). This blocking hypothesis correctly predicts that when
children are presented with a three-stage misinformation paradigm patterned after Loftus
et al. (1978) in which they ®rst view an event and are subsequently presented with
misleading details, they are likely to choose suggested items over witnessed event items in
a forced-choice recognition test (e.g. Ceci et al., 1987).
There are, however, reasons to question the assumption that suggestibility in children
arises from post-event alteration of memories or the more recent presentation of
misleading information. Two adult studies (Lindsay and Johnson, 1989; Rantzen and
Markham, 1992) have demonstrated robust misinformation effects in a reversed misinformation paradigm in which misleading suggestions were presented prior to the event
that was to be remembered. Such effects can be seen as a special case of `proactive
interference' in which old information interferes with the learning of new information (e.g.
Underwood, 1957; Underwood and Postman, 1960) and pose serious problems for models
of suggestibility that assume that the reporting of witnessed information is biased in favour
of more recently presented material.
In order to provide a direct test of this assumption with reference to suggestibility in
children we employed a reversed misinformation design with children aged between 5 and
6 years. If suggestibility in children is due solely to alteration of the initial memory trace
by the subsequent presentation of new details there should be no indication of a
suggestibility effect when the misinformation is presented before the picture story
information. If, as assumed by the trace alteration (Loftus et al., 1978) and certain
retrieval competition theories (e.g. Morton et al., 1985), children are biased to report the
most recently presented information, and that retrieval of this information blocks access to
previously witnessed material, then the misinformation effect should also be eliminated in
the reversed design.
Alternately, a number of theoretical approaches would predict that the reversal of
the conventional order of presentation of event and suggested information might
still produce suggestibility in children. According to trace-strength accounts (e.g. Brainerd
and Reyna, 1998; Holliday et al., 1999), the likelihood of ®nding a misinformation
effect will be dependent on the relative strengths of original and misled information at the
point of retrieval. Within this framework the relative recency of presented material is only
one of many factors that may affect its trace strength. If memories for suggestions are
suf®ciently salient there is no reason why, within a reversed misinformation paradigm,
they cannot compete with more recently presented event material and produce a
misinformation effect.
Source-monitoring accounts (e.g. Johnson et al., 1993) assume the co-existence of
the original event and the post-event misinformation memory traces. According to
such accounts participants make source misattribution errors such that they
mistakenly attribute the source of their memories to the post-event misinformation
instead of to the original event. Source misattributions can occur irrespective of
the presentation order of the original and misleading details because both types of
information remain in memory (Lindsay, 1994). Consequently, the effect should still be
evident when the misinformation is presented before the witnessed event (Lindsay and
Johnson, 1989).
Copyright # 2002 John Wiley & Sons, Ltd.
Appl. Cognit. Psychol. 16: 1±16 (2002)
Memory processes and children's suggestibility
3
A number of researchers (e.g. McCloskey and Zaragoza, 1985; Newcombe and Siegal,
1997; Zaragoza, 1991) have proposed that suggestibility effects obtained in the standard
testing paradigm may be in¯uenced by social demand factors and/or response biases. The
social demand hypothesis proposes that a misled participant who remembers both the
original and the misleading details may incorrectly select the misled item because he or
she perceives the researcher as a credible information source, or he or she may wish to be
viewed favorably by the researcher and acquiesces with the suggestion (McCloskey and
Zaragoza, 1985). The response bias argument asserts that in the standard misinformation
paradigm many misled participants forget or fail to encode the original event details but do
remember the more recently presented misleading details. Notably, in most previous
investigations (e.g. Ceci et al., 1987; Zaragoza, 1991) the social demand and response bias
hypotheses make identical predictions regarding the presence or absence of a misinformation effect. In a reversed misinformation design, however, the two accounts make quite
different predictions. If children report misinformation due to compliance with the perceived wishes of the experimenter (Zaragoza, 1991), or to misinterpretation of the intent of
the questioner (Newcombe and Siegal, 1997) even though they remember the original
event details, then that they would still be more likely to select the misled item than the
original item. In contrast, if misled participants are biased relative to controls towards
selecting the more recently presented misled item then no misinformation effect should be
found using a reversed misinformation design.
In addition to using the reversed design to distinguish between a number of competing
accounts of children's suggestibility, the current study sought to explore the memory
processes which underlie suggestibility in children. Rather surprisingly, although the
preponderance of evidence suggests that memory processes are an important factor in
many instances of suggestibility (Ceci and Bruck, 1993), only scant attention has been
given to identifying the basic memory processes involved in the act of recognition and how
these processes contribute to suggestibility effects in children. Most notably, in recent
years there has been increased interest in dual-process models of retrieval (e.g. Brainerd
et al., 1999; Jacoby, 1991) that propose that there are two distinct components to
recognition memory, recollection and automaticity. Jacoby (1991), for example, developed a process dissociation procedure as a way of separating the relative contributions of
intentional recollection and automatic memory processes to recognition memory performance. For Jacoby (1991), `recollection' is de®ned as `consciously controlled, intentional
use of memory' (p. 516). The process of `automaticity', in contrast, is de®ned as
`relatively automatic . . . faster, less effortful, and less reliant on intention' (p. 516).
Intentional control is operationally de®ned `as the difference between performance
when a person is trying to as compared with trying not to use information from some
particular source' (p. 527). Hence, the process dissociation procedure employs two tasks,
one in which automatic and intentional processes work together, an `inclusion' condition,
and one in which the two processes oppose each other, an `exclusion' condition. Support
for the independent contribution of these two processes to performance on tests of
recognition has been provided by a number of researchers (see Jacoby, 1998; Reingold and
Toth, 1996, for reviews).
From an applied perspective, if suggestibility effects in children are strongly in¯uenced
by automatic memory processes then such effects may be dif®cult to prevent. On the other
hand, if suggestibility effects are due to children intentionally reporting misleading
information then such effects may be minimized by techniques such as warning children
to disregard leading questions.
Copyright # 2002 John Wiley & Sons, Ltd.
Appl. Cognit. Psychol. 16: 1±16 (2002)
4
R. E. Holliday and B. K. Hayes
The process dissociation technique was adapted for use in the study of children's
suggestibility by Holliday and Hayes (2000). Five- and 8-year-old children were presented
with a picture story followed by misleading post-event details. The level of intentional
processing of these misleading details was manipulated by having some suggested items
read aloud to participants while others were self-generated in response to semantic and
perceptual cues. In an inclusion recognition test children were instructed to make a
recognition response whenever they identi®ed test items from either of the previous phases
while those in an exclusion test condition were instructed to exclude post-event suggestions and respond only if they remembered the information from the original picture story.
It was found that the probability of incorrectly recognising suggested items was, for the
most part, determined by automatic memory processes. The contribution of intentional
recollection to children's suggestibility depended on how the misleading suggestions were
encoded such that estimates of intentional recollection were larger for misleading
suggestions that were self-generated than for suggestions that were read aloud.
The use of this process dissociation technique in conjunction with the reversed
misinformation paradigm in the current study served to further re®ne and clarify the
conclusions that could be drawn about the causes of suggestibility in children. As noted
above, a positive ®nding of suggestibility in the reversed design could be explained by a
number of competing accounts: social demand, trace competition, or source monitoring.
These approaches can be further differentiated according to their predictions about the
respective roles of automatic and intentional processing in suggestibility. The social
demand hypothesis suggests that children intentionally deliberate over the reporting of
witnessed or suggested details, and cues taken from the local social context in¯uence ®nal
recognition judgments. Such an approach suggests a strong role for intentional processes
in the reporting of suggested information.
In contrast, certain memory trace-strength accounts, most notably fuzzy-trace theory
(Brainerd and Reyna, 1998), propose that misinformation can be reported either intentionally on the basis of verbatim memory for an item's surface features, or automatically on the basis of gist memories (Brainerd and Reyna, 1998). Moreover,
manipulations that increase the ability to preserve details in storage (e.g. self-generation
of suggestions) will also increase the intentional recollection component. Such an account
suggests that both automatic and intentional processes will be implicated in reversed
misinformation effects.
The predictions of source monitoring accounts regarding the roles of automatic and
intentional processing are a little more dif®cult to derive since different versions of this
model have given different emphasis to the role of intentional processing in source
misattribution (cf. Johnson et al., 1993; Lindsay, 1994). All source-monitoring models do
agree, however, that source misattributions can arise from the automatic processing of test
items whereby participants make recognition judgments based largely on the familiarity of
the test item. Such an account would suggest that if a reversed information effect is
detected then such an effect would be strongly in¯uenced by automatic memory processes.
Hence, following Holliday and Hayes (2000) we applied the process dissociation
equations adapted from Jacoby (1991) to the children's choice of suggested items in a
forced-choice recognition procedure. Only one age group was targeted because in our
previous work we found no evidence of age-related changes in levels of suggestibility
between 5- and 8-year-olds. Moreover, the younger age group closely matches the age of
children used in a large proportion of previous investigations of children's suggestibility
(e.g. Cassel and Bjorklund, 1995; Ceci et al., 1987). It was assumed that in the inclusion
Copyright # 2002 John Wiley & Sons, Ltd.
Appl. Cognit. Psychol. 16: 1±16 (2002)
Memory processes and children's suggestibility
5
test condition children could select a picture depicting a misleading suggestion on the
basis of intentional recollection (R) alone, either because they remember the suggested
item as having been presented, or because of compliance with the perceived wishes of
the adult experimenter (Zaragoza, 1991). On the other hand, children could select a picture
depicting a misleading suggestion by automatic processes (A) alone on the basis of the
item's familiarity, without any intentional recollection. In contrast, in the exclusion
condition, children will select a picture depicting a misleading suggestion only if it is
recalled automatically (Jacoby et al., 1993), and if they forget that the suggestion item was
presented in the pre-event narrative but believe that the suggestion was included in the
original event (Lindsay et al., 1995).
The probabilities of children selecting test pictures that corresponded with the misled
details on the basis of recollection were calculated by subtracting the probabilities of
selecting such test pictures in the Exclusion condition from the probabilities of selecting
such test pictures in the Inclusion condition:
R ˆ P…misled item j Inclusion†
P…misled item j Exclusion†
…1†
The probabilities of children selecting test pictures that corresponded with the misled
details on the basis of automatic processes were calculated by:
A ˆ P…misled item j Exclusion†=…1
R†
…2†
Following the procedure developed by Holliday and Hayes (2000), some suggested
items were read out aloud to children in a fashion which closely parallels the way in which
misleading narratives have been presented in many previous developmental studies of
suggestibility (e.g. Ceci et al., 1987; Zaragoza, 1991). Other misleading details were
generated by the children themselves in response to semantic and linguistic cues.
Following the ®ndings of Holliday and Hayes (2000), we expected that recollection
estimates would be larger for misled-generated items than for misled-read items.
METHOD
Participants
Fifty-nine 5- and 6-year-olds (M ˆ 5 years, 11 months, SD ˆ 4 months, Range: 5 years, 4
months ± 6 years, 9 months), including 27 males and 32 females participated in this study.
All children attended public primary schools in predominantly middle-class areas of the
New South Wales Central Coast, Australia. Children participated only if parental consent
had been granted.
Materials
A 1200-word story was written which related the fortunes of an old lady who found a
peanut and proceeded to make herself rich and famous selling peanut butter. Thirty
8'' 8'' coloured drawings were prepared to illustrate events in this story and were
presented on laminated cardboard. For 12 of these target pictures two alternatives
were constructed, one depicted the misleading detail and one depicted a novel alternative.
Copyright # 2002 John Wiley & Sons, Ltd.
Appl. Cognit. Psychol. 16: 1±16 (2002)
6
R. E. Holliday and B. K. Hayes
The three versions of the target pictures differed only in terms of the targeted detail. For
example, for Item 3 the picture story (version 1) showed Billy and Miss Peabody sitting at
a table eating eggs for breakfast. In the pre-event narrative the misled alternative that was
read or self-generated was cereal and the novel alternative was toast. Two practice
pictures, one shown prior to the picture story and the other prior to the pre-event narrative,
were used to assess understanding of the test instructions. Three versions of the picture
story were constructed to control for possible differences in the dif®culty of stimuli. For
example, for Item 3, one third of the children saw `eggs' in the picture story, one third saw
`cereal', and one third saw `toast'. There were six pre-event narratives for each version of
the original story (i.e. 18 versions) that consisted of a summary of the picture story and
contained two misleading items that were `read', two misleading items that were
`generated', and neutral information about the other two picture items (control items).
The assignment of items to neutral and misled roles was counterbalanced across
participants. For example, a child who saw `eggs' in the picture story (version 1) was
misled with `cereal' in the pre-event narrative. A child who saw `cereal' in the picture
story (version 2) was misled with `toast' in the pre-event narrative. A child who saw `toast'
in the picture story (version 3) was misled with `eggs' in the pre-event narrative. An
example of a pre-event narrative and a list of all experimental items is presented in the
Appendix.
Procedure
All children were tested in a quiet room at their school. On the ®rst day, the
®rst experimenter commenced by introducing a clown picture (`Bozo' Practice picture 1,
misled) to individual children and then read one of the 18 versions of the preevent narratives that contained six items; two misled-read, two misled-generate, and
two in which neutral information (control items) was presented. No pictures were
presented in this phase. All the suggested items were consistent with the syntactic and
semantic context of the original story. The narrative including all neutral and misled-read
items was read aloud to the participants. For example, for the misled±read item `cereal',
the experimenter read aloud `Billy was eating cereal with Miss Peabody when he noticed
she looked like a witch'. For the misled±generate items, the words preceding the item
were read and the experimenter then gave the hint for the suggested word containing a
semantic and letter cue. For example, for the misled±generate item `cereal', the
experimenter read aloud: `Billy was eating' [stop]; the experimenter stopped at this point
and provided the ®rst hint, `this is what you put milk on and eat for breakfast, it starts with
c . . . '. If the child was unable to produce the word the experimenter gave a second hint
containing additional semantic information, for example, `Weetbix, Corn¯akes, and Rice
Bubbles are all these kinds of things, they are c . . . '; the experimenter then continued to
read the rest of the sentence aloud: `with Miss Peabody when he noticed she looked like a
witch.' In the event the word was not produced after the second hint the experimenter
supplied it.
One day later, a second experimenter began the session by introducing a new clown
picture (`Koko' the clown Practice picture 2, control) to groups of 7±10 children. Children
were informed that the clown was going to help the experimenter tell them another story.
The experimenter then read the 1200-word event narrative and showed the picture
corresponding to the story details (each viewed for approximately one minute), with the
experimenter pointing to each item.
Copyright # 2002 John Wiley & Sons, Ltd.
Appl. Cognit. Psychol. 16: 1±16 (2002)
Memory processes and children's suggestibility
7
After a short delay (15±30 minutes), the second experimenter administered two
recognition tests to individual children, an inclusion test followed by an exclusion test
(cf. Jacoby et al., 1993) with exactly the same items tested in each test condition. The
administration of these two tests was not counterbalanced because in pilot testing it was
found that the children had considerable dif®culty following the inclusion instructions
when these were presented after exclusion instructions (i.e. after they had been told that
Experimenter 1 had made some mistakes in their narrative). Moreover, it was thought that
if the exclusion instructions were presented ®rst then this would be likely to increase the
intentional monitoring of all subsequent test items, including those presented in the
inclusion test. This would have the effect of arti®cially lowering the proportion of correct
responses in the inclusion condition. Each test contained six target items that were paired
with four pictures of items on which misleading suggestions were given at Phase 1 (two
misled±read and two misled±generate), and two misled alternatives that had not been
presented previously (control items). For example, for the misled±read item `cereal', a
child was presented with a picture of Billy eating cereal with Miss Peabody and the picture
of Billy eating eggs with Miss Peabody that was shown in the picture story. Pairs of
pictures were placed directly in front of each child and the left±right positioning was
randomized across items.
In both inclusion and exclusion test conditions children were told that they would be
shown some pictures and that their task was to chose the ones that had been presented in
event story. Before testing commenced, however, children in the two conditions were
given different instructions about the role of the information presented in the pre-event
narrative. In the inclusion condition children were primed to respond to the test using
information from either the pre-event or event phases by being asked to `Remember the
picture story about Miss Peabody and the story Miss Selmes read you yesterday?' The
practice clown pictures were then shown to ascertain whether the child understood
the instructions that they could include information from the pre-event narrative and the
picture story: `Here are two practice pictures Ð say YES if you remember the picture, say
NO if you don't remember the picture.' A participant was judged to have understood the
test instructions if they responded `yes' to both the `Bozo' picture shown by the ®rst
experimenter (pre-event narrative) and to the `Koko' picture shown by the second
experimenter (picture story). If a child failed this test on the ®rst try the experimenter
repeated the instructions and administered the practice pictures again until the child
demonstrated an understanding of the instructions. If a child repeatedly failed this task
then testing was terminated. Children were then asked to choose the picture that they
remembered seeing from the story: `Point to the picture in the story.'
In the exclusion condition, children were informed that the ®rst experimenter had made
some mistakes with the pre-event narrative and were instructed to forget about what they
had been told by this experimenter and not to choose a picture corresponding to details the
®rst experimenter had read them: `Remember the story about Miss Peabody? Remember
the story with Miss Selmes yesterday? Well, you know what? I think Miss Selmes made
some mistakes. I think she read you the wrong story. I think you should forget all about
what she told you. If you see a picture of something Miss Selmes told you about I want you
to forget it because it is wrong.'
Before commencing the test the practice clown pictures were shown to determine
whether the child understood the instructions to disregard the pre-event narrative
presented by the ®rst experimenter: `Now here is a practice picture (Koko) from the
second story. Did I show you this picture today? (If the child says YES proceed and show
Copyright # 2002 John Wiley & Sons, Ltd.
Appl. Cognit. Psychol. 16: 1±16 (2002)
8
R. E. Holliday and B. K. Hayes
practice picture Bozo). If Miss Selmes showed you this picture (i.e. Bozo) I want you to
say NO you haven't seen it before because it was wrong. Did Miss Selmes show you this
picture yesterday?' Participants were judged to have understood the test instructions if
they responded `no' to the `Bozo' picture shown by the ®rst experimenter (pre-event
narrative) and `yes' to the `Koko' picture shown by the second experimenter (picture
story). If a child failed this test on the ®rst try the experimenter repeated the instructions
and administered the practice pictures again until the child demonstrated an understanding
of the instructions. If a child repeatedly failed this task then testing was terminated. The
remainder of the test instructions followed those in the inclusion condition: `Point to the
picture in the story.'
RESULTS
In both instruction conditions the proportion of correct responses of the Phase 2
pictures was calculated for control, misled-read, and misled-generate items. All
analyses were performed after collapsing across version and narrative type, as a
preliminary analysis found no reliable effects of these factors on response accuracy.
Table 1 illustrates the mean proportion correct in each experimental condition. A 2(test
type) 3(item type) analysis of variance with repeated measures on both factors was
performed on these data. Two planned contrasts on the item factor were used comparing
control items with misled±read items and misled±generate items, respectively. Bonferroni
adjustments were used to control the familywise error rate at ˆ 0.05 for this analysis (cf.
Hall and Bird, 1975). The signi®cance level for Tukey's post-hoc tests was set at p<0.05 in
all cases.
A signi®cant main effect for test condition was found across item type, F(1, 58) ˆ 7.23,
MSE ˆ 0.34, p ˆ 0.01, with children more likely to choose the original item in the
exclusion test condition (M ˆ 0.74) than in the inclusion test condition (M ˆ 0.68).
There was a signi®cant effect for the contrast comparing recognition accuracy on
control and misled±read items, F(1, 58) ˆ 6.76, MSE ˆ 1.30, p ˆ 0.02. Across test
conditions, children were more accurate on control items (M ˆ 0.79) than on misled±
read items (M ˆ 0.65). Across test conditions, recognition accuracy was also found to be
superior on control (M ˆ 0.79) as compared to misled±generate item types (M ˆ 0.67),
F(1, 58) ˆ 5.79, MSE ˆ 0.95, p ˆ 0.03. This latter contrast, however, interacted signi®cantly with test condition, F(1, 58) ˆ 7.99, MSE ˆ 0.42, p ˆ 0.02. Tukey's post-hoc tests
con®rmed that the difference between control and misled±generate items was signi®cant
in the inclusion test condition but not in the exclusion test condition.
Table 1. Mean proportion recognition (and standard deviations) as a function of experimental
condition
Correct recognition
Control
Misled±read
Misled±generate
Incorrect recognitiona
Inclusion
Exclusion
Inclusion
Exclusion
0.79 (0.32)
0.64 (0.37)
0.58 (0.35)
0.80 (0.26)
0.64 (0.36)
0.75 (0.30)
0.31 (0.21)
0.40 (0.24)
0.44 (0.24)
0.30 (0.17)
0.40 (0.23)
0.33 (0.20)
a
Snodgrass and Corwin's (1988) correction method was applied to individual participant's proportions incorrect
response on the misled±read, misled±generate items and control items.
Copyright # 2002 John Wiley & Sons, Ltd.
Appl. Cognit. Psychol. 16: 1±16 (2002)
Memory processes and children's suggestibility
9
Estimates of recollection and automaticity
The process dissociation equations (1) and (2) were used to calculate separate estimates of
the contribution of the processes of recollection and automaticity to responding to misled±
read and misled±generate item types. Individual proportions of incorrect responses for
misled±read and misled±generate items were ®rst corrected using Snodgrass and
Corwin's (1988) procedure (cf. Hayes and Hennessy, 1996) to eliminate the problem of
responses of one or zero. In this procedure 0.5 is added to individual proportions incorrect
which are then divided by n ‡ 1, where n is the number of misled±read or misled±generate
items. Table 1 shows the corrected proportions of errors (i.e. incorrect recognition) to
suggested items. Equations (1) and (2) were then applied to the corrected response rates to
calculate the probabilities of responding on the bases of recollection and automaticity.
A one-way repeated measures analysis of variance was performed to examine the
effects of encoding condition on the recollection estimates. There was a signi®cant main
effect of encoding, F(1, 58) ˆ 11.84, MSE ˆ 0.37, p ˆ 0.01. Recollection estimates for
misled±generate items (M ˆ 0.11) were larger than for misled-read items (M ˆ 0.00).
An analogous one-way repeated-measures analysis of variance was carried out on the
automaticity estimates. No signi®cant differences between automaticity estimates for the
misled±read (M ˆ 0.40) and misled±generate (M ˆ 0.38) conditions were found.
A major assumption underlying the process dissociation procedure is that base rates for
responding `old' are equal across test conditions (Jacoby et al., 1993). In order to check the
equivalence of false alarm rates across inclusion and exclusion conditions, participants'
proportion of incorrect responses to items on which no suggestions were given were
examined (see Table 1). A 2(test: inclusion, exclusion) within-subjects analysis of variance
was performed. No main effect for test condition was found, F(1, 58) ˆ 0.05, MSE ˆ 0.00,
p ˆ 0.99. Children were no more likely to choose the novel alternative on control items in
the inclusion test condition (M ˆ 0.31) than in the exclusion test condition (M ˆ 0.30). This
suggests that the important process dissociation assumption of equivalent base rates for
inclusion and exclusion tests (cf. Jacoby et al., 1993) was satis®ed for this data set.
DISCUSSION
We employed a reversed misinformation design to examine suggestibility in 5-to 6-year
old children. Evidence of a suggestibility effect was found such that children were less
likely to choose the correct target detail on items for which they had received misleading
suggestions than for control items. Signi®cantly, this effect was obtained when misleading
suggestions preceded the presentation of event information.
This ®nding serious poses problems for a strong version of the trace alteration
hypothesis (e.g. Loftus et al., 1978, 1985), a retrieval interference account in terms of
blocking (e.g. Bekerian and Bowers, 1983; Morton et al., 1985), and a response bias
account (e.g. McCloskey and Zaragoza, 1985). A strong trace alteration account, for
example, argues that the new information (i.e. the picture story details) updates and
replaces the previously presented information (i.e. the pre-event summary) resulting in
permanent erasure from storage. The blocking hypothesis assumes that both the picture
story and misinformation memory traces are held in storage, but that retrieval of the most
recently presented information blocks access to the earlier presented information (Ayers
and Reder, 1998). Similarly, response bias accounts (e.g. McCloskey and Zaragoza, 1985)
Copyright # 2002 John Wiley & Sons, Ltd.
Appl. Cognit. Psychol. 16: 1±16 (2002)
10
R. E. Holliday and B. K. Hayes
argue that children given a standard test are biased toward selecting the more recently
presented information. According to each of these models, the reversal of the standard
order of presentation of witnessed and suggested information should have eliminated the
misinformation effect.
The ®nding that suggestibility levels were differentially affected by the way that
suggestions were encoded and by test instructions raises further problems for trace
alteration accounts and blocking models of retrieval interference. In the inclusion
condition, children were equally likely to correctly reject `read' and `self-generated'
suggestions. Suggestibility effects for misled items that were read to the children were also
invariant across the inclusion and exclusion instruction conditions. In the exclusion
condition, however, children were more likely to correctly reject misled details that
were self-generated. Generating a suggestion led to a reduction in misinformation levels
under exclusion instruction conditions. The fact that children can exclude misleading
suggestions that are self-generated, even when such suggestions are presented prior to the
picture story details, is further evidence that new information neither overwrites the old
information (Loftus et al., 1985), nor blocks its retrieval (Morton et al., 1985).
Notably, however, the exclusion instructions did not affect misinformation acceptance
when the suggested details were read aloud. This interaction between the read/generate
encoding of misleading items and the instructions given at retrieval indicates that
the degree to which memory interference is implicated in suggestibility varies according
to the speci®c encoding-retrieval conditions that obtain. When suggestions are encoded
elaborately as in the generate condition, children are more likely to be conscious or aware
of these suggested memories and hence be able to exclude them when given appropriate
retrieval instructions. Moreover, generating suggestions prior to the presentation of the
picture story details is likely to increase awareness at encoding of the pictures of the
difference between the suggested and picture story details. When, however, the suggestions were simply read aloud to children, it appears a more automatic (and less aware)
form of processing of these suggestions was invoked such that, even when exclusion
instructions were given these suggestions still had a negative effect on recognition
performance.
The obtained process dissociation estimates provide further support for this account. In
accordance with the prediction, estimates of intentional recollection were larger for
misled-generate items than for misled±read items. Indeed, there was no evidence that
intentional recollection contributed to children's suggestible responding when such
suggestions were simply read aloud. It was also found that the contribution of automatic
processes to suggestibility did not vary as a function of encoding procedure in that
automatic processes were found to make a consistently strong contribution to children's
suggestibility responses regardless of the way in which the suggestions were encoded.
However, the ®nding of an increase in the intentional recollection component to children's
suggestible responses that were self-generated demonstrates that the processes underlying
trace competition at retrieval do not operate entirely automatically and outside of
awareness as proposed by the trace alteration (e.g. Loftus, 1997) and trace competition
models (e.g. Morton, 1991).
The ®nding that the reversed misinformation effect was the result of both automatic and
intentional memory processes further narrows the ®eld of viable explanations of children's
suggestibility. The social demand hypothesis (McCloskey and Zaragoza, 1985), for
example, can explain the ®nding of a reversed misinformation effect but is seriously
challenged by the fact that process dissociation estimates of automatic processes indicated
Copyright # 2002 John Wiley & Sons, Ltd.
Appl. Cognit. Psychol. 16: 1±16 (2002)
Memory processes and children's suggestibility
11
that such processes were strongly implicated in the incorrect selection of suggested details
in both read and generate conditions.
In contrast, the current results can be explained by certain memory trace-strength
accounts (e.g. Brainerd and Reyna, 1998; Marche, 1999). Most notably, fuzzy-trace theory
(Brainerd and Reyna, 1998) allows for the possibility of proactive suggestibility such that
strong memory traces for suggestions may interfere with memory for subsequently
witnessed events. Moreover, this approach assumes a dual-process model of misinformation effects arising from the processing of gist and/or verbatim traces which is consistent
with the current ®nding of suggestions mediated by both automatic and intentional
processes. The ®nding of an observed increase in `intentional' suggestibility following
self-generation of suggestions could also be explained in terms of a strengthening of the
integrity of verbatim traces of suggested information in this encoding condition.
Our ®nding of a suggestibility effect in a reversed misinformation design is also
consistent with a source-monitoring account (e.g. Johnson et al., 1993) that holds that
misleading suggestions are reported at test when individuals misattribute the source of
their memories to the misinformation instead of to the original event. According to the
source-monitoring hypothesis, the post-event misinformation and the original event traces
co-exist in memory, and hence, participants are just as likely to select the misinformation
when it is presented before the original event as when it is presented after the original
event (Lindsay, 1994). The fact that children's suggestible responses were predominantly
in¯uenced by automatic memory processes also accords with Johnson et al.'s (1993)
proposal that on the standard test, participants adopt an undifferentiated response criterion
of familiarity and misattribute the source of their memories for the original event to the
misinformation.
Alternately, the ®nding of an intentional recollection component to misled-generate
items may indicate that source misattribution is not based entirely on automatic memory
processes (e.g. Lindsay, 1994). One possibility is that participants may remember that the
source of the misinformation occurred during the pre-event narrative, but also misattribute
the misinformation to the witnessed event.1 Hence, under inclusion instructions, generation of the pre-event misinformation may have led to a stronger misinformation effect
because such generation led to the encoding of a stronger trace of the item and the
inclusion instructions did not preclude the possibility that the pre-event misinformation
also occurred at the event. Following this reasoning, under exclusion instructions the
misinformation effect in the misled±generate condition would be weaker than in the
misled-read condition because generation also enhances memories for the source of
the misinformation. Hence, it is possible that the recollection parameter for misledgenerated items may have been affected by an `unintentional' process involving participants who simultaneously correctly attribute the source of the misinformation to the
pre-event but incorrectly attribute its source to the event.
Although the current ®ndings do not ®t a strong version of trace alteration or a blocking
account, we cannot entirely rule out the in¯uences of these processes on children's
suggestibility. Reviews of the existing evidence (e.g. Ceci and Bruck, 1993) point to
suggestibility in children being a phenomenon that has multiple causal determinants. By
controlling the effects of these processes in this particular paradigm we do not eliminate
the possibility that under different conditions (e.g. with different kinds of events,
misinformation, and retention intervals) such processes may play a major role. Never1
The authors would like to thank Robert Belli for this suggestion.
Copyright # 2002 John Wiley & Sons, Ltd.
Appl. Cognit. Psychol. 16: 1±16 (2002)
12
R. E. Holliday and B. K. Hayes
theless, a blocking account of retrieval interference (even if it is only one of many factors
contributing to the misinformation effect) would predict that the reversed design should
weaken or reduce the magnitude of the misinformation effect relative to the standard postevent procedure. In order to further examine this proposal, we compared the results of an
earlier study (i.e., control versus misled-read items) that used identical test materials,
procedures and study-test delays, and children of a similar age, in a conventional postevent design (Holliday and Hayes, in press) with the current results. Contrary to the
predictions from trace-alteration and blocking accounts, we found that the magnitude of
misinformation effects obtained were similar in the reversed (15%) and the conventional
paradigms (13%).
Limitations of the process dissociation procedure
Since the introduction of the process dissociation procedure there has been debate
surrounding possible violations of the assumption that the recollection and automaticity
represent independent processes (e.g. Brainerd et al., 1998; Curran and Hintzman, 1995;
Jacoby et al., 1997, 1999; Jacoby and Shrout, 1997). Curran and Hintzman (1995) argued
that the process dissociation estimates that they obtained in a series of word-stem
completion experiments provided evidence of violations of the assumption of independence of the processes of recollection and automaticity. These researchers found that
estimates of recollection increased with longer study time but that the opposite was the case
for estimates of automaticity. Signi®cant correlations were found between estimates of
recollection and automaticity that were interpreted as evidence of non-independence. In
reply, Jacoby and Shrout (1997) conducted a psychometric analysis of the effects of
independence assumption violations and demonstrated that signi®cant correlations between recollection and automaticity estimates do not necessarily re¯ect violations of the
independence assumption. Further, Jacoby (1998) pointed out that if estimates of recollection and automaticity are calculated within-subjects, and not across subjects (cf. Curran and
Hintzman, 1995), within-subjects correlations do not bias process estimates. (See Gruppuso
et al., 1997 for a similar view.) In the current research the estimates of automaticity and
recollection were calculated within-subjects, and hence, the likelihood of problematic
violations of the assumptions underlying the process dissociation procedure was reduced.
Brainerd and colleagues (e.g. Brainerd and Reyna, 1998; Brainerd et al., 1998,
1999) recently developed a multinomial `conjoint-recognition' model with the aim
of evaluating the independent contributions of intentional recollection and automatic
processes to children's recognition. The model's core assumptions are incorporated
in fuzzy trace theory's distinction of `identity-similarity' (Brainerd et al., 1999). Fuzzytrace theory holds that verbatim memories (i.e. an item's surface form) and gist
memories (i.e. an item's semantic and relational details) are stored in parallel. At test,
recognition of a verbatim memory is made on the basis of identity (cf. recollection) and
gives rise to a feeling of explicit remembering. In contrast, recognition of a gist memory is
made on the basis of the similarity of target items to stored traces (cf. familiarity)
(Brainerd et al., 1999). This model has been evaluated with adult (Brainerd et al., 1999)
and developmental data (Brainerd et al., 1998). Brainerd and Reyna (1998) have outlined a
modi®ed version of the conjoint-recognition, the conjoint-misinformation model, for use
with data obtained from misinformation studies. Future research using this model,
therefore, may supplement our understanding of the cognitive processes underlying
children's suggestibility.
Copyright # 2002 John Wiley & Sons, Ltd.
Appl. Cognit. Psychol. 16: 1±16 (2002)
Memory processes and children's suggestibility
13
In summary, the current ®ndings are inconsistent with a strong version of the trace
alteration hypothesis (e.g. Loftus et al., 1978), a retrieval interference account in terms of
blocking (e.g. Christiaansen and Ochalek, 1983; Morton et al., 1985), and social demand
and response bias hypotheses (e.g. McCloskey and Zaragoza, 1985), but are compatible
with the source-monitoring hypothesis (e.g. Johnson et al., 1993) and fuzzy trace theory
(e.g. Brainerd and Reyna, 1998). Moreover, the current ®ndings favour a dual-process
account of memory-based suggestibility in terms of changes to memory trace-strength of
the original and/or misinformation traces (e.g. Brainerd and Reyna, 1998; Ceci et al.,
1988; Holliday et al., 1999).
Our ®nding that 5-year-old children's memories of a witnessed event were deleteriously
affected by the introduction of misleading suggestions before a witnessed event indicates
that professionals should be especially careful to avoid suggestive questioning and the
induction of negative stereotypes. In a somewhat similar vein, Leichtman and Ceci (1995)
examined the effect of stereotypes on 3- to 6-year-old children's memories of a subsequent
visit by `Sam Stone'. At weekly intervals before Sam's visit, children were read stories
that portrayed Sam as a good-hearted but awkward person. Ten weeks after Sam's visit,
children were given an interview that included two leading questions about actions that
Sam did not perform (e.g. destroying a book, dirtying a teddy bear). Signi®cantly, over
30% of the children reported that Sam performed at least one of these actions. In other
words, children's false reports were based on their negative stereotypes of `Sam' induced
before he actually visited the preschool.
In the forensic context, the impact of negative stereotypes on children's reports of an
event can be seen in allegations of sexual abuse made by preschool children in the Little
Rascals Day Care and the McMartin Preschool cases, and in the 1987 death row case of
Frederico Macias in Texas (see Ceci and Bruck, 1995, for an extensive review). In the
latter case, for example, six months after Macias was charged with two counts of ®rstdegree murder, a 9-year-old girl told police that she had seen the defendant on the day of
the murders with a ri¯e and blood on his clothing (Ceci and Bruck, 1995). During
interviews with police, however, it became evident that for many years the girl's mother
had warned her to keep away from this `bad' man. Hence, the child had a negative
stereotype of Macias before she allegedly witnessed the event (Leichtman and Ceci, 1995).
Several years later, and two weeks before Macias' execution, the child recanted her
testimony stating that she gave incorrect evidence because she knew Macias was a bad
man and because she wished to comply with the wishes of the adult interviewers.
Moreover, our research demonstrates that memory changes following presentation of
misleading suggestions was, for the most part, attributable to automatic processes. Such
®ndings imply that warning children to discount misinformation and pre-existing stereotypes may be insuf®cient to prevent the subsequent reporting of these suggestions. In
addition, this study suggests that interviewing approaches that encourage self-generation
of suggested details might be more detrimental to the accuracy of subsequent testimony
than are leading questions.
ACKNOWLEDGEMENTS
This research was carried out by the ®rst author in partial ful®lment of the requirements
for a PhD in Psychology at the University of Newcastle, Australia and was supported by
an Australian Postgraduate Award. We would like to thank Christine Selmes and
Copyright # 2002 John Wiley & Sons, Ltd.
Appl. Cognit. Psychol. 16: 1±16 (2002)
14
R. E. Holliday and B. K. Hayes
Bradley Lindsay for assistance in data collection and stimulus development
respectively, and the children and staff from Wyong Public School for their enthusiastic
participation.
REFERENCES
Ayers MS, Reder LM. 1998. A theoretical review of the misinformation effect: Predictions from an
activation-based memory model. Psychonomic Bulletin and Review 5: 1±21.
Bekerian DA, Bowers JM. 1983. Eyewitness testimony: Were we misled? Journal of Experimental
Psychology: Learning, Memory and Cognition 9: 139±145.
Belli RF, Loftus EF. 1996. The pliability of autobiographical memory: Misinformation and the false
memory problem. In Remembering Our Past: Studies in autobiographical memory, Rubin DC
(ed.). Cambridge University Press: New York; 159±179.
Brainerd CJ, Reyna VF. 1998. Fuzzy-trace theory and children's false memories. Journal of
Experimental Child Psychology 71: 81±129.
Brainerd CJ, Stein LM, Reyna VF. 1998. On the development of conscious and unconscious
memory. Developmental Psychology 34: 342±357.
Brainerd CJ, Reyna VF, Mojardin AH. 1999. Conjoint recognition. Psychological Review 106:
160±179.
Bruck M, Ceci SJ. 1999. The suggestibility of children's memory. Annual Review of Psychology 50:
419±439.
Bruck M, Ceci S, Melnyk L. 1997. External and internal sources of variation in the creation of false
reports in children. Learning and Individual Differences 9: 289±316.
Cassel WS, Bjorklund DF. 1995. Developmental patterns of eyewitness memory and suggestibility:
An ecologically based short-term longitudinal study. Law and Human Behavior 19:
507±532.
Ceci SJ, Bruck M. 1993. Suggestibility of the child witness: An historical review and synthesis.
Psychological Bulletin 113: 403±439.
Ceci SJ, Bruck M. 1995. Jeopardy in the Courtroom: A scienti®c analysis of children's testimony.
American Psychological Association: Washington, DC.
Ceci SJ, Crossman AM, Gilstrap LL, Scullin MH. 1998. Social and cognitive factors in children's
testimony. In Eyewitness Memory: Theoretical and applied perspectives, Thompson CP, Herrman
DJ, Read JD, Bruce D, Payne DG, Toglia MP (eds). Erlbaum: Mahwah, NJ; 15±30.
Ceci SJ, Ross DF, Toglia MP. 1987. Suggestibility of children's memory: Psycholegal implications.
Journal of Experimental Psychology: General 116: 38±49.
Ceci SJ, Toglia MP, Ross DF. 1988. On remembering . . . more or less: A tracestrength interpretation
of developmental differences in suggestibility. Journal of Experimental Psychology: General 117:
201±203.
Chandler CC. 1991. How memory for an event is in¯uenced by related events: Interference in
modi®ed recognition tests. Journal of Experimental Psychology: Learning, Memory, and
Cognition 17: 115±125.
Christiaansen RE, Ochalek K. 1983. Editing misleading information from memory: Evidence for
the coexistence of original and post-event information. Memory and Cognition 11:
467±475.
Curran T, Hintzman DL. 1995. Violations of the independence assumption in process dissociation.
Journal of Experimental Psychology: Learning, Memory, and Cognition 21: 531±547.
Gruppuso V, Lindsay DS, Kelley CM. 1997. The process-dissociation procedure and similarity:
De®ning and estimating recollection and familiarity in recognition memory. Journal of Experimental Psychology: Learning, Memory, and Cognition 23: 259±278.
Hall W, Bird KD. 1986. Simultaneous multiple comparison procedures in psychiatric research.
Australian and New Zealand Journal of Psychiatry 20: 46±54.
Hayes BK, Hennessy R. 1996. The nature and development of nonverbal implicit memory. Journal
of Experimental Child Psychology 63: 22±43.
Holliday RE, Douglas K, Hayes BK. 1999. Children's eyewitness suggestibility: Memory trace
strength revisited. Cognitive Development 14: 443±462.
Copyright # 2002 John Wiley & Sons, Ltd.
Appl. Cognit. Psychol. 16: 1±16 (2002)
Memory processes and children's suggestibility
15
Holliday RE, Hayes BK. in press. Automatic and intentional processes in children's eyewitness
suggestibility. Cognitive Development.
Holliday RE, Hayes BK. 2000. Dissociating automatic and intentional processes in children's
eyewitness memory. Journal of Experimental Child Psychology 75: 1±42.
Jacoby LL. 1991. A process dissociation framework: Separating automatic from intentional uses of
memory. Journal of Memory and Language 30: 513±541.
Jacoby LL. 1998. Invariance in automatic in¯uences of memory: Toward a user's guide for the
process dissociation procedure. Journal of Experimental Psychology: Learning, Memory, and
Cognition 24: 3±26.
Jacoby LL, Begg IM, Toth JP. 1997. In defense of functional independence: Violations of
assumptions underlying the process-dissociation procedure? Journal of Experimental Psychology: Learning, Memory, and Cognition 23: 484±495.
Jacoby LL, McElree B, Trainham TN. 1999. Automatic in¯uences as accessibility bias in memory
and Stroop-like tasks: Toward a formal model. In Attention and Performance. (Vol. 17), Koriat A,
Gopher D (eds). Bradford: Cambridge, MA; 461±486.
Jacoby LL, Shrout PE. 1997. Toward a psychometric analysis of violations of the independence
assumption in process dissociation. Journal of Experimental Psychology: Learning, Memory, and
Cognition 23: 505±510.
Jacoby LL, Toth JP, Yonelinas AP. 1993. Separating conscious and unconscious in¯uences of
memory: Measuring recollection. Journal of Experimental Psychology: General 122: 139±154.
Johnson MK, Hashtroudi S, Lindsay DS. 1993. Source monitoring. Psychological Bulletin 114:
3±28.
Leichtman MS, Ceci SJ. 1995. The effects of stereotypes and suggestions on preschoolers' reports.
Developmental Psychology 31: 568±578.
Lindsay DS. 1994. Memory source monitoring and eyewitness testimony. In Adult Eyewitness
Testimony: Current trends and developments, Ross DF, Read JD, Toglia MP (eds). Cambridge
University Press: Cambridge; 27±55.
Lindsay DS, Gonzales V, Eso K. 1995. Aware and unaware uses of memories for postevent
suggestions. In Memory and Testimony in the Child Witness, Zaragoza MS, Graham JR, Hall GC,
Hirschman R, Ben-Porath Y-S (eds). Sage: Thousand Oaks, CA; 86±108.
Lindsay SD, Johnson MK. 1989. The reversed eyewitness suggestibility effect. Bulletin of the
Psychonomic Society 27: 111±113.
Loftus EL. 1997. Dispatch from the (un)civil memory wars. In Recollections of Trauma, Read JD,
Lindsay DS (eds). Plenum Press: New York; 171±198.
Loftus EF, Miller DG, Burns HJ. 1978. Semantic integration of verbal information into
visual memory. Journal of Experimental Psychology: Human Learning and Memory 4: 19±31.
Loftus EF, Schooler JW, Wagenaar WA. 1985. The fate of memory: Comment on McCloskey and
Zaragoza. Journal of Experimental Psychology: General 114: 375±380.
Marche TA. 1999. Memory strength affects reporting of misinformation. Journal of Experimental
Child Psychology 73: 45±71.
McCloskey M, Zaragoza MS. 1985. Misleading postevent information and memory for events:
Arguments and evidence against memory impairment hypotheses. Journal of Experimental
Psychology: General 114: 1±16.
Morton J. 1991. Cognitive pathologies of memory: A headed records analysis. In Memories,
Thoughts, and Emotions: Essays in honor of George Mandler, Kessen W, Ortonly A, Craik F (eds).
Erlbaum: Hillsdale, NJ; 199±210.
Morton J, Hammersley RH, Bekerian DA. 1985. Headed records: A model for memory and its
failures. Cognition 20: 1±23.
Newcombe P, Siegal M. 1997. Explicitly questioning the nature of suggestibility in preschoolers'
memory and retention. Journal of Experimental Child Psychology 67: 185±203.
Rantzen A, Markham R. 1992. The reversed eyewitness testimony design: More evidence for source
monitoring. The Journal of General Psychology 119: 37±43.
Reingold EM, Toth JP. 1996. Process dissociations versus task dissociations: A controversy
in progress. In Implicit Cognition, Underwood G (ed.). Oxford University Press: Oxford;
159±202.
Snodgrass JG, Corwin J. 1988. Pragmatics of measuring recognition memory: Applications to
dementia and amnesia. Journal of Experimental Psychology: General 117: 34±50.
Copyright # 2002 John Wiley & Sons, Ltd.
Appl. Cognit. Psychol. 16: 1±16 (2002)
16
R. E. Holliday and B. K. Hayes
Underwood BJ. 1957. Interference and forgetting. Psychological Review 64: 49±60.
Underwood BJ, Postman L. 1960. Extra-experimental sources of interference in forgetting.
Psychological Review 67: 73±95.
Zaragoza MS. 1991. Preschool children's susceptibility to memory impairment. In The Suggestibility of Children's Recollections: Implications for eyewitness testimony, Doris J (ed.). American
Psychological Association: Washington, DC; 27±46.
APPENDIX: EXAMPLE PRE-EVENT NARRATIVE
PRESENTED AT PHASE 1
Picture story version 1 pre-event narrative version 1
A peanut hit Miss Peabody on the head while she was looking in her letterbox. She picked
it up, put it away (kitchen Ð control item 1) and forgot about it. Sally was thinking about
helping her mother cook (biscuits Ð control item 2) while waiting for the bus. She got
scared when she saw the two old ladies jumping on the table. Billy was eating cereal
(misled±read item 1) with Miss Peabody when he noticed she looked like a witch. He ate
so quickly that he got a stomachache (misled±read item 2). The kids saw Zippy with his
ball (misled±generate item 1) talking to the postman while the shark stole the bicycle.
Miss Peabody and her friends made smart peanut butter in the factory and they worked
very hard and got sore feet (misled±generate item 2). She became rich and famous. When a
movie producer rang she got ready and ¯ew to America. The gorillas burst into the factory
and found Zippy and took him away. Miss Peabody raced home with her new friends.
But the supply of smart peanut butter had already run out. The kids set up a shop and sold
things to the crowd. The kangaroo-nappers fed Zippy jellybeans so that he would tell them
the secret smart peanut butter recipe. They didn't know he needed peanut butter and
special food so he could talk. A bear rescued Zippy and set him free. Everyone was so
happy to see him they forgot about the smart peanut butter and lived happily ever after.
Table A1. Summary of stimulus alternatives for target items
Item
1
2
3
4
5
6
Response alternatives
Kitchen/bedroom/bathroom
Biscuits/cakes/chocolate crackles
Eggs/cereal/toast
Headache/stomachache/sore throat
Doll/ball/truck
Hands/feet/legs
Copyright # 2002 John Wiley & Sons, Ltd.
7
8
9
10
11
12
Hat/coat/shoes
Knife/fork/spoon
Dog/cat/rabbit
Chips/lollies/drinks
Green/red/yellow
Banana/apple/orange
Appl. Cognit. Psychol. 16: 1±16 (2002)