Journal of Applied Developmental Psychology 31 (2010) 145154

Contents lists available at ScienceDirect

Journal of Applied Developmental Psychology

Long-term stability of young children's eyewitness accuracy, suggestibility, and resistance to misinformation
Elyse Brauch Lehman , Marcia J. McKinley 1, David W. Thompson 2, Ann Marie Leonard, Julie I. Liebman, Danielle D. Rothrock
Department of Psychology (MSN-3F5), George Mason University, 4400 University Drive, Fairfax, VA 22030, USA

a r t i c l e

i n f o

a b s t r a c t
Forty 4-year-olds and 39 6-year-olds participated in a modi ed misinformation-effects paradigm. At time 1 they reviewed a story and some of the children were asked questions about it in either recall or recognition format. Three weeks later they were given misinformation about some of the story events. The following week they were asked the original questions. Two years later the procedure was repeated with a different story for 31 of the children. Although 4-year-olds overtly disagreed more times than the older children did when misinformation was initially presented, this resistance did not affect their accuracy or suggestibility scores. The 6-year-olds became more resistant to the suggestive effects of misinformation when they were given an immediate recall test or when given the opportunity to disagree with misinformation. Signi cant testretest correlations occurred over a two year period for both story accuracy and one of the suggestibility scores. 2009 Elsevier Inc. All rights reserved.

Article history: Received 9 September 2005 Received in revised form 22 October 2009 Accepted 22 November 2009 Available online 20 January 2010 Keywords: Eyewitness accuracy Suggestibility Resistance ef cacy Fuzzy trace theory Individual differences

Introduction It has been well established that young children's, especially preschoolers', accounts of an event are more likely than older children's and adults' accounts to be affected by false information that is explicitly suggested to them (Bruck & Ceci, 1999; Howe, 2000). For example, if after witnessing an event young children are given misinformation about it (e.g., Remember when the doctor examined your eyes?), their subsequent reports of the event are more likely to include the (mis)information that the doctor examined my eyes when, in fact, he did not. Nevertheless, although there are consistent age group differences in suggestibility, clinicians and researchers alike have observed that some very young children are capable of providing accurate accounts of events even in the face of suggestive questioning, whereas some older children are not (Baxter, 1990; Bruck & Ceci, 1999; Geddie, Fradin, & Beer, 2000). If one goal of research is to be able to predict which children will succumb to suggestions and which will not, then this individual variation within a given age group needs to be better understood.

Individual-difference predictors of children's suggestibility Although the study of individual-difference predictors of children's suggestibility has increased dramatically in the last 10 years, general conclusions about the characteristics of children whose reports are likely to be in uenced by suggestive interviewing have been dif cult to draw. To a certain extent, the source of the dif culty is both conceptual and methodological. Suggestibility is multidimensional (Ceci & Bruck, 1993), and various components of suggestibility have been investigated including: (1) interrogative suggestibility (assenting to misleading questions), (2) misinformation effects (incorporating misinformation into later reports), (3) source misattribution (failing to remember the source of the misinformation), and (4) false-event creation (constructing a narrative of an event that never occurred) (Bruck & Melnyk, 2004; Ornstein & Elischberger, 2004). Because these components differ in many ways, including the type and timing of suggestions (Lee, 2004), the predictors of suggestibility may vary across components. For example, Bruck and Melnyk (2004) proposed that psycho-social factors may predict interrogative suggestibility and cognitive factors predict misinformation effects and source misattribution. Likewise, suggestibility may be due to different mechanisms at different ages (Bruck & Melnyk, 2004; Chae & Ceci, 2005; Holliday, Reyna, & Hayes, 2002). Sourcemonitoring skills, for example, may play a larger role in young children's suggestibility whereas resolution of con icting information, insight into a questioner's motives, and self-ef cacy may be more involved in the suggestibility of older children (Bruck & Melnyk, 2004; Mazzoni, 1998).

Corresponding author. Fax: +1 703 993 1359. E-mail address: elehman@gmu.edu (E.B. Lehman). 1 Current address: Department of Psychology, Mount Saint Mary's University, 16300 Old Emmitsburg Road, Emmitsburg, MD 21727, USA. 2 Current address: Fairfax County Public Schools, West eld High School, 4700 Stonecroft Blvd., Chantilly, VA 20151-1716, USA. 0193-3973/$ see front matter 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.appdev.2009.11.007

146

E.B. Lehman et al. / Journal of Applied Developmental Psychology 31 (2010) 145154

In 2004 Bruck and Melnyk published a review of the literature on individual differences in suggestibility. Across 69 studies they looked for evidence of relationships between all four components of suggestibility and three categories of possible predictors: (1) demographic (socioeconomic status and sex), (2) cognitive (intelligence, language, memory, theory of mind, executive functioning, distractibility, and creativity), and (3) psycho-social (social engagement, selfconcept/self-ef cacy, stress/emotional arousal/state anxiety, maternal attachment styles, parentchild relationship, parenting styles, temperament, and mental health). For some of the variables no relationship with suggestibility could be found. For others, the results were inconsistent. The most promising predictors appeared to be creativity, language ability, self-concept/self-ef cacy, the parent child relationship, and maternal romantic attachment. Children who were at risk for being suggestible were more imaginative (e.g., ClarkeStewart, Malloy, & Allhusen, 2004) and had less advanced language skills (e.g., Clarke-Stewart et al., 2004), poorer self-concept/selfef cacy (e.g., Davis & Bottoms, 2002), less supportive relationships with fathers or mothers (e.g., Clarke-Stewart et al., 2004), and mothers who were insecurely attached in their romantic relationships (e.g., Goodman, Quas, Batterman-Faunce, Riddlesberger, & Kuhn, 1997; Quas et al., 1999).

Resistance to misleading questions as a predictor of misinformation effects The present studies focused on children's initial responses to an experimenter's presentation of misinformation as a potential source of individual differences in suggestibility as assessed by children's incorporation of the misinformation into later reports. In the standard misinformation effects paradigm, participants experience an event, subsequently receive misleading information, and are then tested for their memory of the event at a later date (Loftus, Donders, Hoffman, & Schooler, 1989). Misinformation is typically presented by incorporating it into subsidiary clauses of questions about non-critical aspects of the event in order to discourage participants from questioning the veracity of the information (Lee, 2004). For example, in our study, after children have heard a story about a girl named Casey who read books about butter ies to her stuffed animals, misinformation is introduced during subsequent questioning in the following way: Remember when Casey read books to her mother Well, when Casey read books to her mother, was her favorite book about birds? We were especially interested in those children who disagreed with the misinformation by saying No, she didn't read to her mother out loud, stating the correct information, or shaking their heads no. Would they be more or less likely to incorporate the misinformation into the later test session? This type of response has been called denial (Gilstrap & Papierno, 2004), disagreement (Hunt & Borgida, 2001), and resistance (Davis & Bottoms, 2002; Zaragoza, Payment, Ackil, Drivdahl, & Beck, 2001). In spite of a methodology that was designed speci cally to minimize disagreements (see Lee, 2004), we expected that there would be enough variability in children's resistance to misinformation to warrant its consideration as a predictor of suggestibility. Although the incidence of explicit disagreement is typically not reported in misinformation effects paradigms, perhaps because it is assumed to be small, it has been evaluated in studies of children's responses to forensic interviewers' contradictions of children's earlier statements (a form of suggestive questioning called modi cations). In these studies child witnesses have been found to challenge the incorrect statements about one third of the time: 36% (Hunt & Borgida, 2001), 36% (Roberts & Lamb, 1999), and 21% (Walker & Hunt, 1998). Hunt and Borgida found no difference in the incidence of disagreements between preschoolers (35 years of age) and elementary school children (911 years of age).

Consistent with Lee and Bussey (1999), we also expected that children would be less likely to incorporate resisted suggestions into their later reports. This hypothesis is based partly on the fact that several cognitive and psycho-social underpinnings of disagreement are known to be related to suggestibility. Presumably memory and metacognitive monitoring skills such as remembering and differentiating correct and incorrect information (Schooler and Loftus' (1993) discrepancy detection) as well as perceived self-ef cacy for resisting an interviewer's suggestions (Davis and Bottoms's (2002) resistance ef cacy) are involved. One of the most consistent ndings in the suggestibility literature is that children with better memory for an event show less suggestibility for that speci c event (Bruck & Melnyk, 2004). Furthermore, Davis and Bottoms (2002) found that 7-year-old, but not 6-year-old, children who felt con dent about telling an interviewer that the interviewer was wrong succumbed to fewer misleading questions. Resistance ef cacy in the Davis and Bottoms' study was measured with a separate Resistance Ef cacy Scale (e.g., What if Rich (the interviewer) is wrong about something that happened, and you know he's wrong. How easy or hard will it be for you to tell him he's wrong?). Studies that have investigated overt disagreement within the actual interview have found that disagreement when misinformation is rst presented leads to lower levels of later suggestibility for adults (Liebman et al. (2002) in the misinformation effects paradigm used in the present study; Zaragoza et al. (2001) in a study in which undergraduates were forced to provide misinformation about the events they had seen). The only developmental study to investigate this hypothesis found that interviewers' contradictions of children's earlier statements were less likely to be included in the later reports of preschoolers who had disagreed with the modi cations than in the reports of preschoolers who had not resisted (Hunt & Borgida, 2001). Because the elementary school children in Hunt and Borgida's study rarely incorporated information from interviewers' modi cations, the relation could not be evaluated in this age group. Testretest reliability Two years after we had studied children in a resistance to misinformation paradigm, there was an opportunity to retest some of the children who had participated in the rst study. This allowed us to investigate the stability of memory accuracy and suggestibility. The goal of predicting which children will succumb to suggestions and which will not in fact assumes that there is at least some minimum level of stability over time. Based on the ndings of the Munich Longitudinal Study on the Genesis of Individual Differences (LOGIC) which assessed children's memory performance through a 10-year period, we expected relatively high correlations for accuracy in our recall task over the 2-year period (Schneider & Weinert, 1995). Unfortunately, perhaps because American psychologists have focused on the situational determinants of suggestibility, there is little discussion in this literature about the long-term stability of suggestibility (Endres, 1997). Some evidence for stability exists in the form of testretest reliability for instruments developed to measure suggestibility within the European individual difference tradition. For example, testretest reliability for several suggestibility scores on parallel forms of the Gudjonsson Suggestibility Scale (GSS; Gudjonsson, 1997) over a 1-week to 8-month period ranged from .55 to .83 (Gudjonsson, 1997; Merckelbach, Muris, Wessel, & van Koppen, 1998). Roebers and Schneider (2002) found similarly high testretest correlations for 6- to 10-year-old children's responses to misleading questions given to different events presented 3 weeks apart. Nevertheless, because the GSS measures suggestibility within a single session, a procedure that is believed to assess immediate misinformation acceptance (Schooler & Loftus, 1993), we still do not know about the stability of suggestibility that is measured in a paradigm that is more likely to assess delayed misinformation retrieval, i.e., one in

E.B. Lehman et al. / Journal of Applied Developmental Psychology 31 (2010) 145154

147

which the inaccurate information is incorporated into subsequent memory reports (Schooler & Loftus, 1993). Information about the long-term stability of suggestibility over several years is also rare. The present studies Children's eyewitness accuracy, suggestibility, and resistance to misinformation were assessed over several weeks with a misinformation effects paradigm. In this paradigm, participants experience an event, subsequently receive misleading information, and are then tested for their memory of the event at a later date (Loftus et al., 1989). At the rst testing the children were 4 and 6 years of age. Two years later, when the children were 6 and 8 years of age, a new story was presented using the same procedure, and testretest reliability for accuracy and suggestibility was determined. We were particularly interested in whether children who showed high levels of suggestibility at one point in time also showed high levels 2 years later. The design of the studies allowed us to also investigate the effects of question format (recall vs. recognition) and timing of initial questioning (at time 1 or later) on the accuracy, suggestibility, and resistance to misinformation of young children. Previous research with elementary school age children suggests that asking neutral questions (i.e., those that neither suggest nor mislead) in recall format immediately after acquisition of material may protect children from the effects of future suggestive interviews (Warren & Lane, 1995). However, given Brainerd and Reyna's (1996) nding that initial recognition tests elevated both true memory responses and false memory responses on later tests for both 5- and 8-year-olds, the protective effect of an immediate interview may not occur when questions are asked in recognition format. Finally, because we were interested in exploring age differences in the way in which misinformation and accurate information are organized in memory, children in the present studies were given the opportunity to provide multiple answers to the questions asked at time 3. It was hoped that use of this methodology would shed some light on the important issue of the number of representations of an event that can exist in memory. Trace alteration models of memory interference (e.g., Loftus, 1975, 1979), for example, propose that only one trace exists at the point of retrieval (i.e., misinformation alters the one and only representation of the original event). On the other hand, both trace-strength models such as fuzzy trace theory (e.g., Brainerd & Reyna, 1998) and retrieval interference models that explain the misinformation effect in terms of retrieval failure (e.g., Bekerian & Bowers, 1983; Morton, 1991) propose that both the original event and postevent memory traces coexist at the point of retrieval (Holliday et al., 2002). Study 1 The main purpose of the rst study was to examine children's resistance to misinformation and to investigate how this kind of resistance affects subsequent memory accuracy and suggestibility. Four- and 6-year-olds participated in the study. Method Participants Participants included 79 children enrolled in four preschool and daycare programs in a metropolitan area. Letters describing the study were sent home with all 4- and 6-year-old children in these schools. Parents who wished to have their children included in the study signed and returned a consent form. The children with parental permission were then read a description of study procedures and asked to write their name (or to make an X) on an assent form if they agreed to participate. All children who met these two criteria were included in the study.

The sample was primarily White and consisted of children from lower to upper middle class homes. There were 40 4-year-olds (15 males and 25 females, M = 4.51 years) and 39 6-year-olds (12 males and 27 females, M = 6.40 years). Question format and time 1 questioning were counterbalanced in each age group. Materials For this study we created a visual stimulus using 21 pictures from a children's picture book. To ensure that the participants had not had prior exposure to this stimulus, we chose a picture book that was currently out-of-print and not available in the local public libraries. The selected illustrations were bound in a notebook. In addition, we wrote a narrative that corresponded to these illustrations. In this narrative a young girl (Casey) nds a fairy with a broken wing and xes the wing so that the fairy can y home. This narrative was printed on separate pages that were not included in the notebook and were not shown to the participants. Procedure For the present study we modi ed a standard misinformation effects paradigm to better re ect the aims of our study. The standard misinformation effects paradigm involves a three-step procedure (Loftus et al., 1989; Titcomb & Reyna, 1995). In the rst step, participants experience an event. In the second step, they receive misinformation about some aspects of that event and receive correct information about other aspects. In the third step, they are tested for their memory of the original event. Researchers then compare participants' memory for misinformed events with their memory for events for which correct information was presented. A signi cant difference between the two is called the misinformation effect. As Table 1 shows, our procedure is similar to the standard misinformation paradigm in the rst and third testing sessions. However, it differs in the second session. During the rst testing session, all children viewed 21 pictures from a children's storybook while the experimenter read the corresponding narrative aloud to the child. Immediately after this, approximately half of the children in each age group were asked 20 questions about the events of the story. The other half were asked no questions. Half of the questioned children were given questions in recall format (e.g., Who did Casey read books to?) and half were asked recognition questions which required only yes or no answers (e.g., Did Casey read books to her best friend?). Thus, there were four conditions in each age group,
Table 1 Summary of the modi ed misinformation effects paradigm used in Studies 1 and 2. Week 1 All children were told a story while looking at a picture book. Half of the children then answered 20 questions about the story. Half of these children answered questions in recall format (e.g., Who did Casey read books to?). The other half of these children answered questions in recognition format (e.g., Did Casey read books to her best friend?). In the example the correct answer is stuffed animals. Week 4 All children received misinformation about 10 of the time 1 questions in either recall or recognition format. The misinformation was embedded within questions about aspects of the story that were not included in the time 1 testing. In the example below, the misinformation is mother. When Casey read books to her mother, what was her favorite book about? (recall format) When Casey read books to her mother, was her favorite book about birds? (recognition format) Week 5 All children were asked to tell the story. They were then asked the 20 time 1 questions in either recall or recognition format (recognition questions included 3 choices: correct, incorrect, and misinformation). Half of the 20 questions were about aspects of the story for which misinformation had been presented at time 2 (the misinformation questions). The other 10 questions were called nonmisinformation questions. They had been asked at time 1 only. Following each question, children were asked whether there was any other answer they wanted to give.

148

E.B. Lehman et al. / Journal of Applied Developmental Psychology 31 (2010) 145154

differentiated by whether or not the children were questioned at session 1 and the question format (recall or recognition). These conditions are summarized in Table 1. In session 2, which occurred approximately 3 weeks after session 1, all children were given misinformation about 10 of the pictures within the context of answering further questions about the story. Children in the two recall conditions of the study were asked recall questions and children in the two recognition conditions were asked recognition questions. For example, participants were asked either Remember when Casey read books to her mother.Well, when Casey read books to her mother, what was her favorite book about? (recall format) or Remember when Casey read books to her mother.Well, when Casey read books to her mother, was her favorite book about birds? (recognition format). In this question, the misinformation is mother because in the original story Casey read books to her stuffed animals. The latter part of the time 2 questions (e.g., what was her favorite book about?) had not been part of the time 1 questions. All 10 of the questions asked during session 2 contained misinformation about aspects of the story that had been tested in session 1. Thus, this part of the procedure differed from the usual standard misinformation paradigm in that no session 2 questions included accurate information about aspects of the story which had been tested in session 1 and were tested later in session 3. (In the standard misinformation paradigm procedure, such questions are often referred to as control questions.). The reason for the change in procedure was that we were not primarily concerned with demonstrating the misinformation effect, the existence of which has previously been widely reported (cf. Payne, Toglia, & Anastasi, 1994). Rather, our focus was on investigating children's reaction (and especially their resistance) to misinformation. As a result, we increased the likelihood that children would notice and respond to the misinformation by including misinformation in every question. This procedure meant that there were no control questions that could be used to demonstrate a misinformation effect. The effect that this procedural change had on our measurement of suggestibility will be discussed below. In the third session, approximately 1 week after session 2, all children were rst asked to tell the Casey story and were then asked the 20 questions from session 1. Of the 20 questions asked, 10 questions (termed misinformation questions) tested children's memory for the misinformation given in session 2. These questions were used to calculate suggestibility scores. The remaining 10 questions were called nonmisinformation questions. They provided only a measure of accuracy after a 4 week delay. As in session 1, approximately half of the children were asked recall questions, which were identical to those in session 1. The remaining participants were asked recognition questions which were changed to include three choices: the correct one (i.e., from the story), the misinformation (or, in nonmisinformation questions, an incorrect one), and an incorrect one. For example, one question was In the story, did Casey read books to her stuffed animals, her mother, or her best friend? In this example the correct answer was stuffed animals and the misinformation presented at session 2 was her mother. Test format (i.e., recall vs. recognition) remained the same across the study for each child. Following each question, children in the recall condition who answered don't know were asked the recognition question. In addition, all children were asked if there was any other answer they wanted to give. Responses to all questions were recorded on scoring sheets.

Accuracy scores were obtained for the 20 questions in session 1, the 10 questions in session 2, the 10 nonmisinformation questions in session 3, and the 10 misinformation questions in session 3. For each question, the scores were 2, 1, or 0. Two points were awarded for a correct answer, 0 points for an incorrect answer, and 1 point for a partially correct answer to a recall question. For example, for the question In the story, who did Casey read books to?, 2 points would be awarded for the verbatim correct answer of her stuffed animals, 0 points for the completely incorrect answer of her best friend, and 1 point for the partially correct answer her play animals. Because the number of questions in each session was not the same (with 20 questions in the rst and third sessions and 10 questions in the second), accuracy is reported as the proportion of total possible scores at each time (i.e., 40 at sessions 1 and 2, and 20 at session 2). Suggestibility in the misinformation effects paradigm is typically measured by comparing accuracy on items for which misinformation is provided and accuracy on items for which misinformation is not provided. However, because of the design of our experiment, this comparison was not possible. Rather, in our design, we calculated ve measures of suggestibility. Two of these suggestibility scores involve measuring the number of answers in session 3 that include the previously presented misinformation. Speci cally, we counted the number of session 3 misinformation questions that were answered with misinformation as the rst response. In addition, we calculated a second suggestibility score by counting the number of session 3 misinformation questions that were answered with misinformation at any step in the questioning. These measures are similar to measures such as the shift items in interrogative suggestibility scales such as the GSS (Gudjonsson, 1997) and the Video Suggestibility Scale for Children (Video SSC; Scullin & Ceci, 2001) in that they count the number of answers that have been changed. The remaining 3 suggestibility scores included an assessment of whether only misinformation or a blend of misinformation and correct/incorrect responses were given. These three scores were calculated by dividing the number of trials on which only the misinformation choice was selected (or a blend of misinformation and correct information or a blend of misinformation and incorrect information) by the total number of trials on which misinformation was selected at any step in the questioning. Finally, we measured children's resistance to misinformation by counting the number of session 2 questions for which children overtly disagreed with the misinformation either by shaking their heads, saying no, or giving the correct information. Results and discussion Accuracy Table 2 presents the mean proportion of questions correctly answered during each of the three testing sessions. A 2(age) 2 (question format) analysis of variance (ANOVA) on the time 1 proportions for the children who were questioned at time 1 produced main effects of age and question format. At time 1, 6-year-olds were more accurate than 4-year-olds (80.1% vs. 73.5%), F(1, 36) = 6.35, p b .05, and accuracy was higher for recognition than for recall (84.3% vs. 67.9%), F(1, 36) = 30.26, p b .01. A 2(age) 2(question format) x 2(time 1 questioning) ANOVA on the proportions at time 2 indicated that accuracy was again higher for recognition than for recall (78.5% vs. 40.1%), F(1, 71) = 138.19, p b .01. There were no signi cant effects of age or whether questions had been asked at time 1. The same ANOVA on the proportions at time 3 for the nonmisinformation questions (i.e., the 10 time 1 items that were not tested at time 2) showed that older children again remembered more than younger children did (56.5% vs. 48.4%), F(1, 71) = 6.13, p b .05. Additionally, there were signi cant main effects of question format, F(1, 71) = 59.68, p b .01, and time 1 questioning, F(1, 71) = 11.39, p b .01, and a

Scoring The data were independently coded and scored by two research assistants, with discrepancies resolved by consensus. The misinformation effects paradigm resulted in 4 accuracy scores, 5 suggestibility scores, and a resistance to misinformation score.

E.B. Lehman et al. / Journal of Applied Developmental Psychology 31 (2010) 145154 Table 2 Mean proportions of questions answered correctly (standard deviations in parentheses). 4-year-olds Recall Time 1 Time 1 questions No time 1 questions Time 2 Time 1 questions No time 1 questions Time 3 nonmisinformation Time 1 questions No time 1 questions Time 3 misinformation Time 1 questions No time 1questions .628 (.097) .410 (.213) .328 (.087) .425 (.196) .211 (.082) .475 (.144) .211 (.156) Recognition .825 (.116) .775 (.160) .767 (.122) .658 (.215) .589 (.093) .608 (.173) .644 (.174) 6-year-olds Recall .736 (.074) .422 (.148) .440 (.158) .533 (.137) .315 (.165) .522 (.075) .320 (.157) Recognition .867 (.071) .800 (.071) .800 (.134) .711 (.117) .700 (.228) .700 (.150) .582 (.125) At step 1 Time 1 questions No time 1 questions At any step Time 1 questions No time 1 questions

149

Table 3 Mean number of time 3 questions answered with misinformation (standard deviations in parentheses). 4-year-olds Recall 1.40 (0.84) 1.00 (0.87) 2.00 (0.94) 2.11 (1.27) Recognition 2.42 (1.00) 2.22 (1.39) 2.75 (1.14) 2.33 (1.58) 6-year-olds Recall 1.11 (1.05) 1.80 (1.62) 1.56 (1.33) 2.60 (1.90) Recognition 2.11 (1.05) 2.91 (1.30) 2.44 (0.88) 3.45 (1.86)

Note. n = 19 (4-year-olds recall), 21 (4-year-olds recognition), 19 (6-year-olds recall), and 20 (4-year-olds recognition).

Note. n = 19 (4-year-olds recall), 21 (4-year-olds recognition), 19 (6-year-olds recall), and 20 (4-year-olds recognition).

signi cant interaction between question format and time 1 questions, F(1, 71) = 5.36, p b .05. Simple effects analyses on this interaction revealed that the difference between the effects of questioning (i.e., greater accuracy at time 3 when questions were initially asked at time 1 than when they were not) occurred for recall but not for recognition, F(1, 71) = 15.64, p b .001 and F b 1, respectively. When the accuracy proportions at time 3 for the misinformation questions (i.e., the 10 items that had received misinformation at time 2) were analyzed, there were again signi cant main effects of question format, F(1, 71) = 56.49, p b .01, and time 1 questioning, F(1, 71) = 16.77, p b .01, and a signi cant interaction between question format and time 1 questions, F(1, 71) = 8.23, p b .01. As in the analysis of the nonmisinformation time 3 questions, simple effects analyses on this interaction revealed that the greater accuracy at time 3 when questions were initially asked at time 1 than when they were not occurred for recall but not for recognition, F(1, 71) = 25.13, p b .01 and F b 1. Neither the main effect of age nor any interactions with it were signi cant in the original ANOVA. Finally, in order to investigate the effects of time on accuracy, a 2(age) 2(question format) 3(time of testing) ANOVA was conducted on the proportions of questions correctly answered. The analysis, which included only children who were questioned at time 1, produced signi cant main effects of question format and time, F(1, 36) = 48.68, p b .001 and F(2, 72) = 28.22, p b .001, respectively, as well as a signi cant interaction between question format and time, F(2, 72) = 8.53, p b .001. NewmanKeuls analyses (p b .05) revealed that, for recall, accuracy decreased from time 1 to time 2, but showed no further drop from time 2 to time 3. For recognition, on the other hand, the major decrease in accuracy occurred between time 2 and time 3. The difference from the rst to the second testing for recognition was not reliable. Suggestibility Table 3 presents the mean number of time 3 questions answered with misinformation (maximum score = 10). Two measures of suggestibility are given. The rst includes misinformation given as the rst response (i.e., at step 1). The second includes misinformation given at any step in the questioning. The results were similar for the 2 measures. The mean number of time 3 questions to which misinformation was given at the initial step was 1.91. The standard deviation of 1.28 and the range of 0 to 5 questions for misinformation given rst indicate that there was a considerable amount of variation across children in the number of questions to which misinformation was given as the rst response. A 2(age) 2(question format) 2(time 1 questioning) ANOVA indicated that more misinformation was given for recognition than for recall questions (2.44 vs. 1.33), F(1, 71) = 16.87, p b .01. There were no signi cant overall differences on this

measure between the two age groups (1.80 vs. 2.03 for 4- and 6-yearolds, respectively), although an age time 1 questioning interaction did emerge, F(1, 71) = 3.87, p b .05. Simple effects analyses on this interaction revealed that for 6-year-olds, as expected, suggestibility increased when no questions were asked at time 1, F(1, 71) = 4.33, p b .05. However, for 4-year-olds, suggestibility was not signi cantly related to time 1 questioning, F b 1. The mean number of time 3 questions to which misinformation was given at any step in the questioning was 2.44 (SD = 1.46, range = 0 to 7). A 3-factor ANOVA (age question format time 1 questioning) again showed the main effect of question format to be signi cant (recall: 2.08, recognition: 2.78), F(1,71) = 4.48, p b .05. The age time 1 questioning interaction, although only a trend, could be interpreted in the same way as the interaction from the analysis performed on the step 1 measure, F(1, 71) = 3.38, p = .07. That is, 6-year-olds, but not 4-year-olds bene ted from answering questions immediately after hearing the story. Finally, in order to answer questions about age differences in how misinformation and accurate information are organized in memory, the proportions of time 3 questions answered with misinformation that contained misinformation only, both misinformation and correct information, or misinformation and incorrect information were calculated. Fig. 1 presents these proportions. Three 2(age) 2(question format) 2(time 1 questioning) ANOVAs were run, one on each type of response. These analyses showed that older children were more likely than younger children to give responses that included both misinformation and the correct information, F(1, 63) = 4.11, p b .05. The decrease with age in responding with only misinformation suggested by the gure was not reliable, F(1, 63) = 1.75, ns. However, in this analysis on misinformation only proportions, the main effect of question format was signi cant, F(1,63) = 3.72, p = .05. Giving only misinformation was more likely in recognition than in recall (M = .64 and .50, respectively).

Fig. 1. Mean proportion of time 3 questions answered with misinformation that contain only misinformation or a blend of misinformation and correct/incorrect information. The proportions are based on the 14 4-year-olds and 16 6-year-olds who accepted misinformation.

150

E.B. Lehman et al. / Journal of Applied Developmental Psychology 31 (2010) 145154 Table 5 Effects of resistance to misinformation on accuracy and suggestibility. 4-year-olds Disagree Recall Accuracy Time 1a Time 2a Time 3 nonmisinformationa Time 3 misinformationa Suggestibility At step 1b At any stepb Recognition Accuracy Time 1a Time 2a Time 3 nonmisinformationa Time 3 misinformationa Suggestibility At step 1b At any stepb
a b

Resistance to misinformation Fifty-three percent of 4-year-olds (10 in the recall condition and 11 in recognition) and 41% of 6-year-olds (8 in the recall condition and 8 in recognition) resisted at least once when misinformation was presented at time 2 either by saying no or by shaking their heads no. Table 4 presents the mean number of trials in which children disagreed with the experimenter's presentation of misinformation at time 2. A 2(age) 2(question format) 2 (time 1 questioning) ANOVA on these resistance scores indicated that the 4-year-olds disagreed more times than the 6-year-olds did, F(1, 71) = 5.03, p b .05 (M = 1.44 and 0.72, respectively). Question format and whether or not questions were asked at time 1 had no effect on either resistance to misinformation score. Effect on accuracy. Table 5 presents data on the effect of resistance to misinformation on accuracy and suggestibility. The comparison is between children who resisted at least once when misinformation was presented at time 2 with those who never exhibited any indication of resistance at time 2. Means for accuracy re ect the proportion of trials for which the answers were correct. Means for suggestibility are the number of trials for which children answered with the misinformation. There was no difference in accuracy at time 1 between children who resisted and those who did not (M = 0.79 and 0.75, respectively). In a 2(age) 2(question format) 2(resistance) ANOVA, neither the main effect of resistance nor any interaction with it was signi cant. Thus, resistance to misinformation at time 2 was not simply the result of higher accuracy at time 1. Resistance to misinformation also had no general effect on accuracy at later testings. In three 2(age) 2(question format) 2(time 1 questioning) 2(resistance) ANOVAs (at time 2, time 3 on the nonmisinformation questions, and time 3 on the misinformation questions), the main effect of resistance was never signi cant. In these analyses, the one interaction that reached signi cance was the 3-way interaction between age, question format, and resistance in the time 3 misinformation questions analysis, F(1, 63) = 5.34, p b .05. Simple effects analyses showed that there was little effect of resistance for 4-year-olds, either for recall or recognition. For 6-year-olds, recall accuracy on the misinformation questions at time 3 was surprisingly lower for those children who resisted misinformation at time 2 than for those who did not resist, F(1, 63) = 4.79, p b .05, whereas recognition accuracy on these questions was higher for 6-year-olds who had resisted than for those who had not, F(1, 63) = 6.53, p b .05. Effect on suggestibility. A 2(age) 2(question format) 2(time 1 questioning) 2(resistance) ANOVA on the number of time 3 misinformation questions to which participants gave misinformation at any step in the questioning produced a main effect of resistance and an interaction between age and resistance, F(1, 63) = 4.68, p b .05 and F(1, 63) = 4.40, p b .05. Simple effects analyses showed that resisting when misinformation was presented at time 2 had no effect on the amount of misinformation given at time 3 for 4-year-olds, F b 1 (M = 2.28 and 2.30 for resistance and no resistance, respectively). However, for 6-year-olds, resisting misinformation at time 2 signi cantly reduced the amount of misinformation given at time 3, F(1, 63) =
Table 4 Mean number of time 2 trials on which children resisted the misinformation provided by the experimenter. 4-year-olds Recall Time 1 questions No time 1 questions Total 0.80 1.56 1.16 Recognition 1.67 1.78 1.71 Recall 0.33 1.20 0.79 6-year-olds Recognition 0.89 0.46 0.65

6-year-olds Disagree Do Not Disagree

Do Not Disagree

.654 .345 .360 .395 1.30 2.20

.610 .400 .284 .300 1.10 1.90

.800 .428 .367 .339 0.75 1.25

.718 .435 .465 .485 2.48 2.93

.843 .782 .637 .618 2.14 2.36

.800 .760 .620 .630 2.50 2.70

.888 .813 .775 .725 2.00 2.25

.850 .800 .667 .575 2.81 3.37

Proportion. Number of trials.

4.61, p b .05 (M = 1.75 and 3.15, respectively). This result occurred whether the questions were in recall or recognition format and whether or not questions were asked at time 1. Because the same outcome occurred when the ANOVA was run on the number of trials in which misinformation was given as the rst response (instead of at any step in the questioning), the details of that analysis will not be given here. Study 2 The purpose of Study 2 was to investigate the stability of memory accuracy and suggestibility. A subset of the children who participated in Study 1 were available to participate in a new misinformation effects paradigm study 2 years later and comparisons were made with their earlier performance. Method Participants Parents of the children who had participated in Study 1 and who were still living in the area were contacted by phone and asked if they would consent to have their children participate in Study 2. Thirtyone families from the original sample agreed to do so. There were no obvious differences between the samples in the two studies. Fourteen 6-year-olds (5 males and 9 females) and 17 8-year-olds (6 males and 11 females) were in Study 2. All four combinations of question format and time 1 questioning were represented in each age group. Children were assigned to the same condition in which they had participated in Study 1. Procedure The procedure for Study 2 was identical to the procedure for Study 1, except that a new story written speci cally for this study was used for the original event. The new story, called Henry's Bag, describes two best friends, Henry and Katie. Henry has a magic bag that helps him to do wonderful things. When he goes away on vacation, Katie keeps the bag for him and also learns to do wonderful things. Examples of the 20 session 1 questions are: What kind of animal appeared when Katie opened the bag? (recall format), and Did a bunny appear when Katie opened the bag? (recognition format). The time 1 manipulation of questioning (present or not present) was continued in Study 2 in order to maintain children in the same condition across the two studies. In session 2, which was

Note. n = 19 (4-year-olds recall), 21 (4-year-olds recognition), 19 (6-year-olds recall), and 20 (4-year-olds recognition).

E.B. Lehman et al. / Journal of Applied Developmental Psychology 31 (2010) 145154

151

approximately 3 weeks after session 1, misinformation was introduced to all children for 10 of the session 1 questions within the context of asking new questions about the story. An example of a session 2 question in recognition format is Remember when the dove ew out of the bag.Well, when the dove ew out of the bag, did Katie say to it, Hello. It's nice to meet you?. In this question, the misinformation is dove because in the original story a bunny appeared when Katie opened the bag. The matching recognition question for this item in session 3 is: What kind of animal appeared when Katie opened the bag? Was it a mouse, a bunny, or a dove? (correct response: bunny, misinformation response: dove, incorrect response: mouse). Results and discussion Descriptive statistics for Study 2 sample Table 6 presents the means and standard deviations for all accuracy, suggestibility, and resistance to misinformation scores. Time 1 accuracy scores were not analyzed because of the small number of children who were questioned at this time. All other scores were analyzed with 2(age) 2(question format) ANOVAs. Accuracy. When the ANOVA was performed on the proportion of questions correctly answered, the main effect of question format was signi cant in all three analyses, F(1,27) = 38.64, p b .01 (time 2), F(1, 26) = 33.35, p b .01 (time 3 nonmisinformation questions), and F(1, 26) = 3.94, p = .05 (time 3 misinformation questions). As expected, accuracy was greater for recognition than for recall. The main effect of age on accuracy was signi cant only for the time 3 misinformation questions, F(1, 26) = 6.54, p b .05. In this analysis, 8-year-olds were more accurate than 6-year-olds (57.2% vs. 39.3%, respectively). The interaction between age and question format was not signi cant in any of these analyses. Suggestibility. Main effects of age and question format only appeared in the analysis of the rst suggestibility score. That is, overall, misinformation given as the rst response decreased with age, F(1, 26) = 9.31, p b .01, and was more prominent for recognition than for recall questions, F(1, 26) = 20.69, p b .01. More importantly, however, there was a signi cant interaction between age and question format that occurred both when misinformation was given as the rst response, F(1, 26) = 15.17, p b .01, and when misinformation was given at any step in the questioning, F(1, 26) = 15.38, p b .01. Simple effects analyses of this interaction indicated that a decrease in suggestibility with age happened when the test was in recognition format, F(1, 9) = 15.35, p b .01 (misinformation at step 1) and F(1, 9) = 16.85, p b .01
Table 6 Descriptive statistics for Study 2: Means and standard deviations (in parentheses). 6-year-olds Recall Accuracy Time 1a Time 2a Time 3 nonmisinformationa Time 3 misinformationa Suggestibility At step 1b At any stepb Misinformation onlya Misinformation and correcta Resistanceb
a b

(misinformation at any step). When the test was in recall format, on the other hand, there was no signi cant difference between 6- and 8-year-olds for misinformation given at step 1, F b 1. The means were 1.56 and 1.90, respectively. For misinformation recalled at any step, however, there was a trend for an increase with age, F(1, 17) = 3.48, p = .08. The means were 2.78 and 4.10 for 6- and 8-year-olds, respectively. There were no signi cant effects in the analyses of the two blending scores. Resistance to misinformation. There were no signi cant differences in age or question format in the ANOVA on the number of trials in which misinformation was denied at time 2. The means for age were .71 and 1.19 for 6- and 8-year-olds, respectively, and .79 and 1.27 for recall and recognition, respectively. The number of participants was inadequate to evaluate the effects of resistance to misinformation on accuracy and suggestibility. Comparison of testing sessions two years apart Table 7 presents the Pearson correlation coef cients (2-tailed) for accuracy, suggestibility, and resistance to misinformation scores between testing sessions 2 years apart. Also given are the partial correlation coef cients with the effect of age removed. The table indicates that there was considerable stability for the accuracy of scores over a 2-year period. That is, children who accurately remembered a story at one age were also likely to remember another story well 2 years later. There was also some evidence of stability for suggestibility. Signi cant correlation coef cients occurred when the suggestibility score was misinformation given as the rst response. Resistance to misinformation scores across the 2-year interval, on the other hand, was not related. General discussion The present study was designed to answer three questions about children's suggestibility. (1) How do memory question format and timing of initial questioning affect the emergence of misinformation effects? (2) How is resistance to misinformation related to age and suggestibility? (3) How stable are memory accuracy and suggestibility over a 2 year period? The results have implications for models of the memory processes underlying misinformation effects and suggest ways in which children's eyewitness reports may be improved. With respect to the rst question, the results of these studies showed that varying the way in which misinformation effects are measured affects both accuracy and suggestibility. In Study 1, for both 4- and 6-year-olds, questions in a recall format reduced the amount of misinformation that was given at a later time whereas questions in a recognition format produced both greater accuracy and greater misinformation effects. These results support those of Roberts, Lamb, and Sternberg (1999) and Schwartz-Kenney and Goodman
Table 7 Correlation coef cients for accuracy, suggestibility, and resistance to misinformation scores between testing sessions two years apart. n Accuracy Time 1 Time 2 Time 3 nonmisinformation Time 3 misinformation Suggestibility At step 1 At any step Misinformation only Misinformation & correct Proneness to disagree
*

8-year-olds Recall .775 (.066) .450 (.178) .175 (.204) .460 (.227) Recognition .870 (.104) .729 (.138) .633 (.151) .683 (.172)

Recognition .800 (.100) .700 (.158) .580 (.303) .420 (.084)

.775 (.106) .267 (.132) .139 (.171) .367 (.173)

Zero-order r .19 .57** .76*** .63*** .48** .14 .06 .01 .16

Partial r (age removed) .12 .56** .78*** .59*** .53** .15 .06 .02 .13

1.56 (1.13) 2.78 (1.30) .361 (.356) .380 (.428) 0.89 (1.36)

5.00 (0.71) 5.60 (1.14) .656 (.153) .165 (.105) 0.40 (0.89)

1.90 (0.88) 4.10 (1.73) .498 (.336) .323 (.236) 0.70 (1.34)

2.17 (1.47) 2.67 (1.21) .542 (.459) .417 (.492) 2.00 (2.76)

13 31 30 30 30 30 28 28 30

Proportion. Number of trials.

p b .05. **p b .01.

***

p b .001.

152

E.B. Lehman et al. / Journal of Applied Developmental Psychology 31 (2010) 145154

(1999) who found that misleading information was rarely reported in free recall by preschoolers or 6-year-olds. They are also in agreement with Brainerd and Reyna's (1996) nding that recognition tests both inoculate true memories and create false memories. It is interesting that misinformation effects were, however, suggested in Study 2 for the recall of 8-year-olds. Roberts and Blade (1998) and Schwartz-Kenney and Goodman (1999) also reported that, on a free recall task, misinformation effects were more likely to be found for older children (i.e., 9-year-olds). The effect, they suggested, may be due to source-monitoring confusions that result from older children's greater retrieval of both accurate and inaccurate information. In effect, children cannot remember which one was part of the original event. The nding in our study that older children were more likely than young children to give responses that included both misinformation and correct information supports this sourcemonitoring account of misinformation effects (see also Ackil & Zaragoza, 1995; Lindsay & Johnson, 1989). It has also been proposed that the timing of initial questioning plays a role in determining the size of misinformation effects in recall. In particular, Warren and Lane (1995) showed that asking neutral questions in recall format immediately after presentation of material protected elementary school age children from the effects of future suggestive interviews. This result was also found in our Study 1. Questioning in recall format immediately after the presentation of the story (i.e., at time 1) increased accuracy at time 3 for both 4- and 6-year-olds and reduced suggestibility for 6-year-olds. In contrast, the protective effect of an immediate interview did not occur for recognition. Questioning in recognition format at time 1 had no reliable effect on time 3 accuracy, although it did reduce suggestibility somewhat for 6-year-olds. These results have implications for memory interference explanations of the misinformation effect, i.e., for theories making the assumption that misinformation interferes with the storage and/or retrieval of event details (see Holliday et al., 2002 for a review). First, our nding that children's responses sometimes included both misinformation and correct information indicates that a strong version of trace alteration models is unlikely. The trace alteration model proposes that new information alters or overwrites the old, leaving only one trace at retrieval (e.g., Loftus, 1975, 1979). Instead, our data provide support for trace-strength and retrieval interference models that propose the coexistence of original event and postevent memory traces at the point of retrieval (e.g., Bekerian & Bowers, 1983; Brainerd & Reyna, 1998). Second, our data on recall/recognition differences appear to be best explained by trace-strength models, e.g. trace-integrity theory (Howe & Brainerd, 1989) and fuzzy trace theory (Brainerd & Reyna, 1998). These models predict that the magnitude of the suggestibility effect depends on conditions that affect the relative strength of original and postevent memory traces. Based on trace-integrity theory, the protective effect of an immediate interview with recall questions would presumably stem from the strengthening of the original memory trace during the immediate interview (Howe, 1991; Marche & Howe, 1995). With recognition (or yes/no) questions, on the other hand, a protective effect of immediate questioning would not occur if the original memory trace is weakened, or at least not strengthened, by the introduction of alternative choices (e.g., recall question: Who did Casey read books to? vs. recognition question: Did Casey read books to her best friend?). The misinformation given at time 2 (i.e., Casey read books to her mother) would then be more likely to be recognized at time 3 because, as the most recently presented piece of information, it has the strongest memory trace or if, as in fuzzy trace theory, children are more likely to respond on the basis of familiarity, a more automatic memory process, than on the basis of intentional recollection. Support for trace-strength models is also provided by our nding that giving only misinformation as the answer to a question (instead of blends of misinformation and accurate/inaccurate information) occurred more frequently when the

questioning was done in recognition format and when the children were younger. The second research question asked how resistance to misinformation was related to age and how it affected memory accuracy and suggestibility. Our nding that more than half of the 4-year-olds and almost half of the 6-year-olds overtly disagreed at least once when misinformation was presented at time 2 indicated that there was suf cient variance to answer this question. The analyses documented that although 4-year-olds were more likely than 6-year-olds to resist misinformation when it was presented, there was little effect of resistance on accuracy or suggestibility in the 4-year-old age group. For 6-year-olds, on the other hand, resisting misinformation when it was rst presented (i.e., at time 2) signi cantly reduced the amount of misinformation given and increased recognition accuracy at time 3, while it also reduced recall accuracy at time 3. These results contrast with Hunt and Borgida's (2001) report of no difference in the incidence of disagreements between preschoolers and 9- to 11-yearold children and agree with the results of studies showing lower levels of suggestibility when adults overtly disagree with misinformation (Liebman et al., 2002; Zaragoza et al., 2001). Note that although the overall correction levels in our study were relatively low, they were in line with those reported in earlier research (e.g., Roberts & Lamb, 1999). Most likely, linguistic, psycho-social, and cognitive factors in uence both the likelihood that children will disagree and, if they do, the likelihood that disagreement will have an effect on accuracy and/or suggestibility. As noted by Brady, Poole, Warren, and Jones (1999), there are many reasons why children might say no to adult's questions, including the general tendency to say no (i.e., naysaying). Naysaying in and of itself would not be expected to affect accuracy or suggestibility. A more interesting explanation for why 4-year-olds were more likely to resist than were 6-year-olds is that the age effect re ects older children's higher levels of conformity to authority. Preschool and elementary school teachers typically value obedience and discourage assertiveness (Fagot, 1985). Perhaps the older children's longer exposure to school environments has reduced their willingness to question authority. Our nding of age differences in resisting misinformation does, in fact, t with earlier studies documenting increases with age in conformity to authority, especially in situations that are ambiguous, e.g. when the mismatch between the group's choice and one's own is hard to detect (Costanzo & Shaw, 1966; Hoving, Hamm, & Galvin, 1969). Disagreement probably also re ects resistance ef cacy, i.e., perceived self-ef cacy for resisting an interviewer's suggestions (Davis & Bottoms, 2002, p. 188). As proposed by Davis and Bottoms, a child who feels con dent about telling adults that they are wrong should succumb to fewer misleading questions. This effect was observed for older but not younger children in Davis and Bottoms's study in which resistance ef cacy was measured with a separate scale. Our nding that actual resistance at time 2 reduced the amount of misinformation given at time 3, again only for the older children, supports Bruck and Melnyk's (2004) suggestion that self-ef cacy, a psycho-social factor, plays a larger role in the suggestibility of older children (see also Mazzoni, 1998). Young children's positive, but sometimes unrealistic, sense of self-ef cacy (i.e., their belief that they can do more than they actually can) (Bjorklund, Gaultney, & Green, 1993) may limit the in uence of self-ef cacy on suggestibility in this age group. As discussed in the Introduction, cognitive factors are also involved both in the likelihood that children will resist and in the effect of resistance on accuracy and/or suggestibility. For example, unless a discrepancy is noted between the event information and the misinformation, overt resistance is unlikely. Furthermore, the effect of resistance may depend on children's ability to handle two different beliefs or representations for one target. It is well known that young children have

E.B. Lehman et al. / Journal of Applied Developmental Psychology 31 (2010) 145154

153

dif culty considering two dimensions of a perceptual array at the same time (e.g., Piaget & Inhelder, 1969) or two different representations for one object (e.g., DeLoache, 1987; Flavell, Green, & Flavell, 1986; Wimmer & Perner, 1983). Indeed, Welch-Ross, Diecidue and Miller (1997) and Templeton and Wilcox (2000) have demonstrated that suggestibility is negatively related to children's understanding of con icting mental representations. Extending this argument to overt resistance, it may be that denying the misinformation leads preschoolers to attend unilaterally, i.e., to focus solely on the misinformation or on the correct information, thus producing an encoding that contains either one or the other piece of information, whereas disagreement in older children leads to the encoding of both, along with information about source. Source information may be especially strong for older children who have disagreed with the discrepancy because their challenge to the source of the misinformation has made source more distinctive and, therefore, more memorable (see Howe, 2000 for a discussion of the role of distinctiveness in memory). Whether overt resistance simply enhances memory for the resistance or actually produces change in the event memory representation itself is an important question for future research (see Zaragoza et al., 2001 for a similar recommendation). Our nding that resistance to misinformation differentially affected the accuracy of responding to recall and recognition test questions at time 3 (with increased recognition accuracy but reduced recall accuracy) supports the idea that memory changes are involved. Although the number of children who could be located 2 years later and who agreed to be retested was smaller than desired, there was still enough power to begin to answer the third research question, i.e., about stability over time. As expected, the results con rmed the strong testretest correlations found for story recall accuracy by Schneider and Weinert (1995) and by Roebers and Schneider (2002). Of greater interest was our nding that giving misinformation as a rst response, i.e., one of the suggestibility scores, was relatively stable over a 2-year interval. Roebers and Schneider reported that elementary school age children who accept the false suggestions in one interview are also likely to do so in a similar interview situation 3 weeks later. Furthermore, testretest reliability in adults for suggestibility scores on parallel forms of the GSS given up to 8 months apart have been shown to be high (Gudjonsson, 1997). Our study is the rst, however, to investigate children's acceptance of misinformation about two events that occur 2 years apart. These reports documenting stability over time provide support for research whose goal is to understand variables that will identify children who are more likely to be misled by suggestive questioning. An approach recognizing that different mechanisms may be predictive at different ages should be especially useful (Bruck & Melnyk, 2004; Chae & Ceci, 2005; Holliday et al., 2002). Resistance to misinformation, it appears, is more involved in the suggestibility of older children. The results, however, also advise caution in an interpretation of suggestibility as a trait-like construct (see Roebers and Schneider, 2002 for a similar interpretation of their results). We found little stability across the 2-year interval for resistance to misinformation or for giving misinformation at any step in the questioning, another measure of suggestibility. It may be that stronger testretest correlations for resistance to misinformation would occur for older children after greater exposure to teacher expectations about appropriate school behavior. Future research could easily test this hypothesis. The most important goals in future research, however, should be replicating our nding of stability for giving misinformation as a rst response and investigating why this measure of suggestibility is stable over time and others apparently are not. A potentially interesting mechanism to explore in this regard is the psycho-social variable of conformity. Finally, the results of this investigation have implications for children's participation in the legal system. They suggest that training and interview techniques that reduce the risk of suggestibility in older

children may not be as effective for preschoolers. In our study, for 4year-olds, neither time 1 questioning nor disagreeing when misinformation was presented at time 2 had any effect on the amount of misinformation given at time 3. Others have also reported that techniques such as informing children that it is appropriate to say I don't know are not very effective with younger children (e.g., Memon & Vartoukian, 1996). The positive news from this study is that by 6 years of age children do become more resistant to the suggestive effects of misinformation when interviews contain certain elements, such as a recall test that occurs immediately after an event or when children are given the opportunity to disagree with the misinformation. As suggested by Davis and Bottoms (2002), interviewers could even use the results of an assessment of children's perceived resistance ef cacy prior to the actual interview to plan individualized interventions that would increase the likelihood of children disagreeing when they know the interviewers' statements are wrong. Disagreeing when misinformation is presented may act as a self-warning and may serve the same function as being warned in advance by an adult about the possibility of tricky questions, another technique known to increase accuracy and reduce suggestibility in school age children (see Warren, HulseTrotter, & Tubbs, 1991). In summary, the results of our study replicate earlier work demonstrating that there is a reduced incidence of reporting misinformation about an event when a recall rather than a recognition test is used, especially if questions in recall format are asked immediately after presentation of the event. Our study also extends the small literature on resistance to misinformation by showing in a misinformation effects paradigm that 6-year-old children who disagree with misinformation when it is rst presented produce lower levels of suggestibility on a later test. Arguably, however, our most important nding is that stability exists over a two year interval for at least one measure of suggestibility. Taken together, these ndings emphasize the importance of the measures used when testing for the in uence of misinformation, suggest that age differences exist in the mechanisms that underlie suggestibility, and encourage research efforts to understand individual variation in suggestibility within a given age group. Acknowledgements This work was supported in part by a Research/Educational Development Support Program award from the Of ce of the Provost at George Mason University and by Graduate Research Assistant awards from the College of Arts and Sciences at George Mason University. Portions of this work were presented at the meeting of the Eastern Psychological Association, Washington, D.C., April 1997, and at the Conference on Human Development, Mobile, AL, March 1998. References
Ackil, J. K., & Zaragoza, M. S. (1995). Developmental differences in eyewitness suggestibility and memory for source. Journal of Experimental Child Psychology, 60, 5783. Baxter, J. S. (1990). The suggestibility of child witnesses: A review. Applied Cognitive Psychology, 4, 393407. Bekerian, D. A., & Bowers, J. M. (1983). Eyewitness testimony: Were we misled? Journal of Experimental Psychology: Learning, Memory and Cognition, 9, 139145. Bjorklund, D. F., Gaultney, J. F., & Green, B. L. (1993). I watch, therefore I can do.: The development of meta-imitation during the preschool years and the advantage of optimism about one's imitative skills. In M. L. Howe & R. Pasnak (Eds.), Emerging themes in cognitive development: Vol. 2. Competencies (pp. 79102). New York: Springer-Verlag. Brady, M. S., Poole, D. A., Warren, A. R., & Jones, H. R. (1999). Young children's responses to yesno questions: Patterns and problems. Applied Developmental Science, 3, 4757. Brainerd, C. J., & Reyna, V. F. (1996). Mere memory testing creates false memories in children. Developmental Psychology, 32, 467478. Brainerd, C. J., & Reyna, V. F. (1998). Fuzzy-trace theory and children's false memories. Journal of Experimental Child Psychology, 71, 81129.

154

E.B. Lehman et al. / Journal of Applied Developmental Psychology 31 (2010) 145154 Marche, T. A., & Howe, M. L. (1995). Preschoolers report misinformation despite accurate memory. Developmental Psychology, 31, 554567. Mazzoni, G. (1998). Memory suggestibility and metacognition in child eyewitness testimony: The roles of source monitoring and self-ef cacy. European Journal of Psychology of Education, 13, 4360. Memon, A., & Vartoukian, R. (1996). The effects of repeated questioning on young children's eyewitness testimony. British Journal of Psychology, 87, 403415. Merckelbach, H., Muris, P., Wessel, I., & van Koppen, P. J. (1998). The Gudjonsson Suggestibility Scale (GSS): Further data on its reliability, validity, and metacognition correlates. Social Behavior and Personality, 26, 203210. Morton, J. (1991). Cognitive pathologies of memory: A headed records analysis. In W. Kessen A. Ortonly & F. Craik (Eds.), Memories, thoughts, and emotions: Essays in honor of George Mandler (pp. 199210). Hillsdale, NJ: Erlbaum. Ornstein, P. A., & Elischberger, H. B. (2004). Studies of suggestibility: Some observations and suggestions. Applied Cognitive Psychology, 18, 11291141. Payne, D. G., Toglia, M. P., & Anastasi, J. S. (1994). Recognition performance level and the magnitude of the misinformation effect in eyewitness memory. Psychonomic Bulletin & Review, 1, 376382. Piaget, J., & Inhelder, B. (1969). The Psychology of the Child. New York: Basic. Quas, J. A., Goodman, G. S., Bidrose, S., Pipe, M. -E., Craw, S., & Ablin, D. S. (1999). Emotion and memory: Children's long-term remembering, forgetting, and suggestibility. Journal of Experimental Child Psychology, 72, 235270. Roberts, K. P., & Blade, M. (1998). The effects of interacting with events on children's eyewitness memory and source. Applied Cognitive Psychology, 12, 498503. Roberts, K. P., & Lamb, M. E. (1999). Children's responses when interviewers distort details during investigative interviews. Legal and Criminological Psychology, 4, 2331. Roberts, K. P., Lamb, M. E., & Sternberg, K. J. (1999). Effects of the timing of postevent information on preschoolers' memories of an event. Applied Cognitive Psychology, 13, 541559. Roebers, C. M., & Schneider, W. (2002). Stability and consistency of children's event recall. Cognitive Development, 17, 10851103. Schneider, W., & Weinert, F. I. (1995). Memory development during early and middle childhood: Findings from the Munich Logitudinal Study (LOGIC). In F. E. Weinert & W. Schneider (Eds.), Memory performance and competencies: Issues in growth and development (pp. 263279). Hillsdale, NJ: Lawrence Erlbaum. Schooler, J. W., & Loftus, E. F. (1993). Multiple mechanisms mediate individual differences in eyewitness accuracy and suggestibility. In J. M. Puckett & H. W. Reese (Eds.), Mechanisms of everyday cognition (pp. 177203). Hillsdale, NJ: Lawrence Erlbaum. Schwartz-Kenney, B. M., & Goodman, G. S. (1999). Children's memory of a naturalistic event following misinformation. Applied Developmental Science, 3, 3446. Scullin, M. H., & Ceci, S. J. (2001). A suggestibility scale for children. Personality and Individual Differences, 30, 843856. Templeton, L. M., & Wilcox, S. A. (2000). A tale of two representations: The misinformation effect and children's developing theory of mind. Child Development, 71, 402416. Titcomb, A. L., & Reyna, V. F. (1995). Memory interference and misinformation effects. In F. N. Dempster & C. J. Brainerd (Eds.), Interference and inhibition in cognition (pp. 263294). San Diego, CA: Academic Press. Walker, N. E., & Hunt, J. S. (1998). Interviewing child victim-witnesses: How you ask is what you get. In C. P. Thompson D. Herrmann J. D. Read D. Bruce D. Payne & M. P. Toglia (Eds.), Eyewitness memory: Theoretical and applied perspectives (pp. 5587). Mahwah, NJ: Lawrence Erlbaum. Warren, A. R., Hulse-Trotter, K., & Tubbs, E. C. (1991). Inducing resistance to suggestibility in children. Law and Human Behavior, 15, 273285. Warren, A. R., & Lane, P. (1995). Effects of timing and type of questioning on eyewitness accuracy and suggestibility. In S. E. Hobfoll M. S. Zaragoza J. R. Graham G. C. N. Hall R. Hirschman & Y. S. Ben-Porath (Eds.), Applied Psychology: Individual, Social, and Community Issues: Vol. 1. Memory and testimony in the child witness (pp. 4460). Thousand Oaks, CA: Sage. Welch-Ross, M. K., Diecidue, K., & Miller, S. A. (1997). Young children's understanding of con icting mental representation predicts suggestibility. Developmental Psychology, 33, 4353. Wimmer, H., & Perner, J. (1983). Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children's understanding of deception. Cognition, 13, 103128. Zaragoza, M. S., Payment, K. E., Ackil, J. K., Drivdahl, S. N., & Beck, M. (2001). Interviewing witnesses: Forced confabulation and con rmatory feedback increase false memories. Psychological Science, 12, 473477.

Bruck, M., & Ceci, S. J. (1999). The suggestibility of children's memory. Annual Review of Psychology, 50, 419439. Bruck, M., & Melnyk, L. (2004). Individual differences in children's suggestibility: A review and synthesis. Applied Cognitive Psychology, 18, 947996. Ceci, S. J., & Bruck, M. (1993). Suggestibility of the child witness: A historical review and synthesis. Psychological Bulletin, 113, 403439. Chae, Y., & Ceci, S. J. (2005). Individual differences in children's recall and suggestibility: The effect of intelligence, temperament, and self-perceptions. Applied Cognitive Psychology, 19, 383407. Clarke-Stewart, K. A., Malloy, L. C., & Allhusen, V. D. (2004). Verbal ability, self-control, and close relationships with parents protect children against misleading suggestions. Applied Cognitive Psychology, 18, 10371058. Costanzo, P. R., & Shaw, M. E. (1966). Conformity as a function of age level. Child Development, 37, 967975. Davis, S. L., & Bottoms, B. L. (2002). Effects of social support on children's eyewitness reports: A test of the underlying mechanism. Law and Human Behavior, 26, 185215. DeLoache, J. S. (1987). Rapid change in the symbolic functioning of very young children. Science, 238, 15561557. Endres, J. (1997). The suggestibility of the child witness: The role of individual differences and their assessment. The Journal of Credibility Assessment and Witness Psychology, 1, 4467. Fagot, B. I. (1985). Stages in thinking about early sex role development. Developmental Review, 5, 8398. Flavell, J. H., Green, F. L., & Flavell, E. R. (1986). Development of knowledge about the appearance-reality distinction. Monographs of the Society for Research in Child Development, 51(Serial No. 212). Geddie, L., Fradin, S., & Beer, J. (2000). Child characteristics which impact accuracy of recall and suggestibility in preschoolers: Is age the best predictor? Child Abuse and Neglect, 24, 223235. Gilstrap, L. L., & Papierno, P. B. (2004). Is the cart pushing the horse? The effects of child characteristics on children's and adults' interview behaviours. Applied Cognitive Psychology, 18, 10591078. Goodman, G. S., Quas, J. A., Batterman-Faunce, J. M., Riddlesberger, M. M., & Kuhn, J. (1997). Children's reactions to and memory for a stressful event: In uences of age, anatomical dolls, knowledge, and parental attachment. Applied Developmental Science, 1, 5475. Gudjonsson, G. H. (1997). The Gudjonsson Suggestibility Scales Manual. Hove, East Sussex: Psychology Press. Holliday, R. E., Reyna, V. F., & Hayes, B. K. (2002). Memory processes underlying misinformation effects in child witnesses. Developmental Review, 22, 3777. Hoving, K. L., Hamm, N., & Galvin, P. (1969). Social in uence as a function of stimulus ambiguity at three age levels. Developmental Psychology, 1, 631636. Howe, M. L. (1991). Misleading children's story recall: Forgetting and reminiscence of the facts. Developmental Psychology, 27, 746762. Howe, M. L. (2000). The Fate of Early Memories: Developmental Science and the Retention of Childhood Experiences. Washington, DC: American Psychological Association Press. Howe, M. L., & Brainerd, C. J. (1989). Development of children's long-term retention. Developmental Review, 9, 301340. Hunt, J. S., & Borgida, E. (2001). Is that what I said?: Witnesses' responses to interviewer modi cations. Law and Human Behavior, 25, 583603. Lee, K. (2004). Age, neuropsychological, and social cognitive measures as predictors of individual differences in susceptibility to the misinformation effect. Applied Cognitive Psychology, 18, 9971019. Lee, K., & Bussey, K. (1999). The effects of misleading and inconsistent postevent information on children's recollections of criterion-learned information. Journal of Experimental Child Psychology, 73, 161182. Liebman, J. I., McKinley-Pace, M. J., Leonard, A. M., Sheesley, L. A., Gallant, C. L., Renkey, M. E., et al. (2002). Cognitive and psychosocial correlates of adults' eyewitness accuracy and suggestibility. Personality and Individual Differences, 33, 4966. Lindsay, D. S., & Johnson, M. K. (1989). The eyewitness suggestibility effect and memory for source. Memory and Cognition, 17, 349358. Loftus, E. F. (1975). Leading questions and the eyewitness report. Cognitive Psychology, 7, 560572. Loftus, E. F. (1979). The malleability of human memory. American Scientist, 67, 312320. Loftus, E. F., Donders, K., Hoffman, H. G., & Schooler, J. W. (1989). Creating new memories that are quickly accessed and con dently held. Memory and Cognition, 17, 607616.