Mental context reinstatement or drawing: Which better enhances children's recall of witnessed events and protects against suggestive questions?

Abstract

The aim of this experiment was to examine the effectiveness of two techniques in enhancing children's recall of an event that they experienced approximately a week earlier. Younger (5–6 years) and older (8–9 years) children were interviewed about a magic show event in one of three conditions. Before recalling the event, some children were instructed to mentally reinstate the context of the event (MCR group), others were asked to draw the context of the event (DCR group), and others received no reinstatement instructions (NCR). Results showed that these instructions had no impact on children's free recall or responses to open‐ended prompts. However, reinstatement instructions impacted children's responses to suggestive questions: those in the DCR group gave more accurate responses than those in the NCR group. These findings provide preliminary support for the use of drawing as a potentially protective exercise that lessens the impact of biased questions with child witnesses.

Key words:

• child witnesses
• event recall
• drawing
• mental context reinstatement

In interview situations, young children typically provide fewer details in their free recall accounts compared with older children and adults (Lamb, Hershkowitz, Orbach, & Esplin, 2008; Powell, Fisher, & Wright, 2005). Nevertheless, children's free recall accounts elicited via open‐ended questions (e.g., ‘Tell me what happened’) are more accurate than their responses to specific questions (e.g., ‘Did you see a red car?’; Bruck, Melnyk, & Ceci, 2000; Lamb et al., 2008). One reason for this lack of detail in free recall may be due to children's limited ability to spontaneously generate their own internal retrieval cues, which helps them recall information (Katz & Hershkowitz, 2010; Roediger & Gallo, 2002). Context reinstatement is one technique investigated in the laboratory and used in the field to facilitate the retrieval of a full and accurate account. Theoretically based on Tulving and Thomson's (1973) ‘encoding specificity principle’, it is posited that memory recall is facilitated by reinstating the contextual features that were present at the time of encoding. The greater the similarity between features that were present at encoding and the memory cues given at retrieval, the better the recall of an event. Thus, according to this principle, the choice of memory cues is paramount in optimising the recall of a witness.

Mental Context Reinstatement

Typically, reinstatement of the context has been achieved by the interviewer providing a series of verbal prompts (i.e., memory cues) to encourage a witness to bring to mind environmental particulars (e.g., who was there, what the weather was like) and the personal circumstances (e.g., what they were thinking, what they were feeling) that were present when they experienced the to‐be‐remembered event (Fisher & Geiselman, 1992). Known as mental context reinstatement (MCR), this instruction is provided directly before the witness is asked to recount what happened. However, research examining MCR has revealed mixed findings regarding its impact on children's recall. Some studies have found a beneficial effect on children's memory recall (i.e., recalling more event correct details, without compromising accuracy), compared with a control group (Bowen & Howie, 2002; Dietze, Powell, & Thomson, 2010, 2012; Dietze & Thomson, 1993; Hayes & Delamothe, 1997; McCauley & Fisher, 1995). For example, one study demonstrated that children's recall in response to specific questions about an event was enhanced when they reinstated the context—either mentally or ‘out loud'—than when they did not (Dietze et al., 2010). However, other studies have not found a beneficial effect on witness recall (Hershkowitz, Orbach, Lamb, Sternberg, & Horowitz, 2001; Memon, Wark, Bull, & Köhnken, 1997; Milne & Bull, 2002). For example, it was found that a MCR instruction was no more effective in enhancing children's recall of an event than a simple instruction to ‘try harder’ (Milne & Bull, 2002).

One difference across studies may be the varying level of child‐appropriate verbal prompts chosen to reinstate the context. Supplying non‐suggestive and personally relevant MCR verbal prompts to witnesses is a noted difficulty (Dando, Wilcock, & Milne, 2009). One of the reasons for this difficulty is that contextual cues are idiosyncratic—what may prove a powerful cue to one witness may have no such effect with another (Wesson & Salmon, 2001). Another reason that makes it difficult to choose appropriate prompts is that interviewing multiple witnesses for the one crime opens up the possibility for the inadvertent cross‐pollination of (unsubstantiated and suggestive) information gleaned from one interview into the MCR verbal prompts in another (Dando et al., 2009). Further—and more particular to interviewing young children—investigators face the additional challenge of keeping in mind that the salience of particular contextual cues is different for young children compared with adults (see Ackerman, 1985; Nelson, 1990). Theoretically, the implication is that memory retrieval is not optimal because the provided verbal cues do not maximise the activation of the representation of the event in memory.

Unfortunately, it is difficult to examine differences in the way MCR was delivered across studies as not all provide detailed descriptions of the instructions (e.g., Memon et al., 1997; Milne & Bull, 2002). However, that there are differences in memory recall depending on the way MCR was delivered might partly explain the differences in the number of correctly recalled details across studies. To give an example, participants provided more correctly recalled details when MCR instruction was delivered by a 3‐min pre‐recorded spiel (Hammond et al., 2006) than when it was delivered by five standardised directions (Dietze & Thomson, 1993). It may be that in comparison to the latter study, the former study supplied more contextual cues, thereby optimising feature overlap.

A more pragmatic concern regarding the use of the MCR is that in the course of their duties, police officers (in England and Wales) generally do not implement the MCR well or consistently (Dando, Wilcock, & Milne, 2008; Kebbell, Milne, & Wagstaff, 1999). For instance, one recent study examined how well novice police who had recently been trained in the national Initial Police Learning and Development Programme were able to utilise the Cognitive Interview (CI; Dando et al., 2009), of which MCR is a key component. Police conducted mock interviews and were then rated on their application of each of the CI components. The results indicated that no officer conducted the CI in its entirety, and only 6% of officers utilised MCR (with 20% attempting to use it). One suggested reason for the failure of police to apply the MCR is because it is time‐consuming and difficult to deliver (Dando et al., 2008; Kebbell et al., 1999). As these concerns are focused on the delivery of MCR cues (as verbal prompts supplied by the interviewer), a reasonable strategy then is to bypass the interviewer and find a way for witnesses to self‐generate their own (potentially more salient) context reinstatement cues. Directing witnesses to draw as much as they can remember about the event prior to eliciting a free narrative account may be one such technique.

Drawing Context Reinstatement

It has been well established that the use of drawing enhances children's recall when they are directed to simultaneously ‘draw and tell’, compared with a ‘tell only’ comparison group. In the seminal study, 5‐ to 6‐year‐old children were interviewed about what they could remember about a fire station tour they experienced 24-hr or 1 month earlier (Butler, Gross, & Hayne, 1995). Half the children were asked: ‘Tell me anything you can remember about when you got the medal’. The other half was instructed to ‘[d]raw me anything you can remember about when you got the medal’. Children who were able to draw during the interview reported double the amount of information compared with their counterparts, without a reduction in accuracy at either retention time.

Subsequent research has demonstrated that drawing while telling enhances children's recall when interviewed after a delay of 6 months and when re‐interviewed 1 year later (Gross & Hayne, 1999). It also enhances recall for various events, including emotional experiences and negative family experiences (Gross & Hayne, 1998; Lev‐Wiesel & Liraz, 2007; Salmon, Roncolato, & Gleitzman, 2003; Wesson & Salmon, 2001). A recent study found the beneficial effect of drawing for children extended from ages 5 through to 12 years (Patterson & Hayne, 2011). One of the reasons proposed for why drawing enhances event recall is that it encourages people to generate their own retrieval cues about the to‐be‐remembered event (Butler et al., 1995; Gross & Hayne, 1998; Wesson & Salmon, 2001).

Two recent studies have extended the aforementioned research by demonstrating that drawing prior to—rather than simultaneously with—the elicitation of a free narrative account and questioning enhanced witness memory (Dando et al., 2009; Katz & Hershkowitz, 2010). In a study conducted with undergraduates, participants viewed a brief crime film and were allocated to one of three interview conditions: no context reinstatement (NCR), MCR, or a ‘sketch’ context reinstatement (known here as a DCR). When interviewed 1 day later, those who were directed to draw the to‐be‐remembered event recalled a similar amount of correct information to those who were given the traditional verbal MCR instructions; both types of context reinstatement interviews were superior to an NCR interview (Dando et al., 2009). Further, those in the drawing condition made fewer confabulations (i.e., reporting details that were not present or events that did not occur) in the free narrative and questioning phases compared with the MCR and NCR conditions. Dando et al. suggested that the comparatively higher number of confabulations made by witnesses in the MRC interview during the free narrative phase occurred due to interviewer interference arising from less effective or inappropriate verbal prompts.

In a second study, children aged 4–14 years who were alleged victims of sexual abuse were interviewed by experienced youth investigative interviewers about their experience (Katz & Hershkowitz, 2010). Because the researchers were obliged to ensure children's accounts were not compromised by the experimental drawing instruction (which may adversely affect the investigative process), children's free narrative accounts were first exhausted via a highly structured protocol. After children were randomly assigned to a drawing condition (where children were given 7–10-min to draw what happened) or no drawing condition (where children were given 7–10-min to play or rest), they were interviewed again. Children in the drawing condition provided richer accounts than the children in the no‐drawing control group. Overall, this research suggests that drawing before providing an account may increase children's recall.

Suggestibility

Although drawing increases the number and accuracy of details that witnesses report with open‐ended prompts, it is not clear whether it protects witnesses against suggestive questioning. To date, studies that have examined drawing and suggested information have utilised a paradigm where children were instructed to draw event details that did not occur (Bruck et al., 2000; Gross, Hayne, & Poole, 2006; Strange, Garry, & Sutherland, 2003). These studies have demonstrated that children generate more false memories for these details when they draw them than when they do not. However, it has not yet been investigated whether simply asking children to draw everything that they can remember about the event—without directing them to draw suggested details—also increases the reporting of incorrect information when they are later asked suggestive questions.

Turning our attention briefly back to MCR, no study has examined whether MCR alone protects against later exposure to suggestive information. However, witnesses interviewed by the CI protocol (with MCR as a key element) were less likely to accept suggestive information into their account when it was presented after the interview, compared with a control interview (Memon, Zaragoza, Clifford, & Kidd, 2010; Milne & Bull, 2003). Explaining the beneficial effect of the CI, Memon et al. (2010) proposed that the stronger and more complete memory of the event elicited by the CI led witnesses to be more confident about their account. Consequently, it was postulated that witnesses viewed the suggested information with suspicion and were better able to resist it than those in the control interview condition.

It is possible that drawing should protect children's memories against suggested information in a similar way, that is through enhancing recall of the original event. Further, given that a drawing is an external and physically available reference of what they originally recalled, it is feasible that drawing may better able to protect children against suggestive questions than MCR.

The aim of the current study was to compare two memory techniques in younger and older children—DCR and MCR—to determine their impact on children's recall and ability to resist suggestive questions. It was hypothesised that both the DCR and MCR should enhance children's recall more than NCR. It was further predicted that DCR should enhance recall more than MCR due to the self‐generation of more personally relevant retrieval cues. Also, it was hypothesised that DCR should better protect children against suggestive questions compared with MCR, which in turn would be more protective than NCR.

Method

Participants

One hundred forty‐one children were recruited from four primary schools in Melbourne. There were 70 younger children aged 5–6 years (M = 5.7, SD = 0.4) (35 females, 37 males) and 71 older children aged 8–9 years (M = 8.7, SD = 0.3) (41 females, 30 males).

Design

We used a 3 (condition: DCR, MCR, NCR) × 2 (age: younger, older) × 2 (biased questions: true, false) mixed design. The first two factors were manipulated between subjects, and the last factor was manipulated within subjects. The dependent variables were correct, incorrect, and confabulated event details elicited about the magic show. Accuracy (correct details/correct, incorrect, and confabulated details) and completeness (correct details/total number of possible correct details) percentages were also calculated.

Materials and procedure

Ethics approval was obtained from the university committee, as well as the Department of Education and Early Childhood Development, and the Catholic Education Office (Melbourne). Upon agreement from the principals of each of the schools to participate in the study, letters were sent home via children's homeroom teacher to obtain parental consent.

Magic show event

Children participated in a 20‐min scripted magic show that was performed by one of the researchers at the children's school; each grade level participated in the magic show separately. To ensure that the content of the magic show was appropriate for children over both the younger and the older age groups, the ‘magician’ stated at the outset that she had only been performing magic for a short time and that she needed feedback (which they presented by clapping after each trick) about whether her tricks would be suitable for an audience of kindergarten children. Briefly, the magic show consisted of the magician: introducing herself, conducting a short puppet skit, involving the children in a warm‐up activity and choosing an assistant, ‘disappearing’ a lollipop, making a drink ‘magically’ appear, and making lipgloss ‘magically’ appear. The lipgloss was then passed around on ear buds for children to put on their lips or smell, and last, children were given a sticker. Teachers were asked to refrain from talking about the magic show with the children prior to the magic show and were reminded again directly after the show was concluded.

Draw‐a‐Person Test—Quantitative Scoring System (DAP) (Naglieri, 1988)

Given the inclusion of a drawing element in the present study, the DAP was used to measure children's drawing ability. The DAP is a 15‐min assessment that requires children to carefully draw a man, a woman and themselves. Each drawing is coded on a 64‐point scale made up of 14 criteria areas. Within these criteria, points can be awarded for presence, extra details, and proportion (where relevant). The three drawings are then summed to generate a total DAP score.

Interview

Children were interviewed once about what they could remember about the magic show that they had participated in 7–10 days earlier. Each child was escorted to a quiet room that was different from where the magic show was staged and interviewed according to their assigned condition. Children were pseudo‐randomly allocated to an interview condition to ensure that age and gender were represented roughly equally across cells. All interviews were conducted by the first author. Training consisted of the interviewer becoming familiar with the CI protocol, as described by Fisher and Geiselman (1992) and Saywitz, Geiselman, and Bornstein (1992), and investigative interviewing best practices (Powell, 2002; Powell et al., 2005). At a practical level, multiple opportunities were provided to practice interviewing (in the form of mock interviews), and extensive feedback was provided by an expert in the field. Overall, the interviewer undertook 25-hr of training.

All three interview conditions (DCR, MCR, and NCR) had five common phases (see Table 1). In Phase 1, children were informed that the interviewer was ignorant about the content of the magic show; they were instructed to concentrate and not make anything up. Next, children assigned to the MCR or DCR interview conditions were given specific instructions. Those assigned to the MCR condition heard the following instruction (adapted from Geiselman et al., 1984 and Saywitz et al., 1992):

I want you to think back and picture the magician and the magic show as if you were there right now. Think about what it was like there on that day. Think about what you can see there. Think about what you can smell there. Think about what you can hear there. Think about what you were feeling that day. Think about what you were thinking during the magic show. Picture any other people there.

Table 1. The five phases of the interviews

Phase 1: Greet and interview preparation

● I need for you to tell me about when the lady came and did the magic show

● I'm going to ask you to tell me about it, then I am going to ask you some questions

● I wasn't there that day so I don't know what happened

● Please concentrate

● Don't make anything up

Phase 2: Initiate a free narrative account

● Tell me everything that happened in the magic show; start from the very beginning

● … What else can you tell me about the magic show?

Phase 3: Extended open‐ended prompts based on free narrative (x 3)

● You mentioned that the magician. … Tell me everything that happened in the part. …

● … What else happened in the part where you saw …?

● … And then what happened?

● … Tell me more about the part where you saw. … I need to hear every little detail.

● … What happened then?

Phase 4: True/false biased specific questions

● Eight pre‐set true‐ and false‐biased questions

Phase 5: Closure

● Thanks and praise

● Questions?

Children assigned to the DCR condition were given a sheet of paper and a pencil, and heard the following instruction (adapted from Butler et al., 1995):

I want you to think back and picture the magician and the magic show as if you were there right now. Now I want you to draw as many things that you can about the magician and the magic show. You have only 5-min to do it, so don't worry about being too neat.

The time limit was established to approximate a comparable time frame to the MCR interview. In both conditions, children were not required to speak out loud while they were reinstating the context or drawing. Children in the NCR condition did not receive any instruction before Phase 2.

In Phase 2, a free narrative account was elicited. All children received the same three prompts (detailed in Table 1). In accordance with best practices, only minimal encouragers were given (e.g., ‘uh huh’, ‘mmm’, and head nodding) (Powell et al., 2005). In Phase 3, the children were probed for further details on three events they mentioned previously during their free narrative. Each of the three events was probed by three open‐ended prompts, with the first open‐ended prompt (listed at the top of the Phase 3 section in Table 1) posed to every child. The interviewer then chose two other open‐ended prompts from a predetermined list, composed of two breadth and two depth open‐ended prompts (e.g., … and then what happened?; … tell me more about the part where you saw. … I need to hear every little detail). This was to afford the interviewer some flexibility in the selection of the most appropriate prompt depending on the flow of the child's account while maintaining a level of experimental control.

In Phase 4, children were asked a predetermined list of true and false‐biased questions about eight target items. Four question formats were used, which varied according to whether they presumed certain information to be true and whether they provided broader or more specific suggestions. Each question type was repeated twice, once with a true‐biased version of the target item and once with a false‐biased version of the target item (e.g., Did the magician have a favourite lollipop? or Did the magician have a favourite ribbon?). The questions were presented in random order. In Phase 5, children were thanked for their time.

Interview coding

All interviews were audio‐recorded and transcribed verbatim. To code the transcripts, a scoring template was developed based on pre‐existing techniques (Gross & Hayne, 1998; Memon et al., 1996). More specifically, one coder generated a comprehensive list of the key concepts from the script of the magic show (parsed as meaningful simple sentences or clauses) that if recalled were scored as one point each. Extra relevant details that were intrinsically embedded in the key concepts were also identified, which made it possible to score at a greater level of specificity, over and above the key concept (e.g., ‘Puts up the poster (1) on the whiteboard’ (1) ‘Chose/had a helper (1) name’ (1)). Next, the scoring template was trialled on 15 randomly selected transcripts with a second coder, who was familiar with the magic show but uninvolved in the present study, for the purpose of progressively fine‐tuning the template. Any discrepancies were discussed and resolved. The final coding system comprised 119 key concepts and 81 associated extra details, totalling 200 event details.

Each interview phase (free recall, extended open‐ended prompts) was coded separately. Every piece of information recalled by the children was coded as correct, incorrect (e.g. ‘The magician's favourite lollipop was strawberry'—where it was actually banana), or confabulated (reporting a detail that did not appear or occur; e.g. ‘The magician made baby butterflies appear’). Errors of distortions (i.e., incorrectly recalled information) and confabulation were coded separately as there is some evidence that they arise from different underlying psychological mechanisms (Gudjonsson & Clare, 1995). Information that was repeated, subjective (e.g. ‘I really enjoyed the magic show’), or was unrelated to the magic show (e.g. ‘My dad can do a magic card trick’) was not coded.

Children's responses to the true‐/false‐biased specific questions were coded as one of the following: correct, incorrect, or ‘don't know’. Inter‐rater reliability was obtained by two coders independently marking 20% of the same transcripts (that were chosen at random). Agreement was then calculated by intraclass correlation for each of the dependent variables in the free recall and the open‐ended question phase of the interview. Inter‐rater reliability was very good to excellent (.87–.99).

Results

Preliminary analyses

Context reinstatement instruction duration

The DCR instruction took about twice as long as the MCR instruction (M = 4.3-min, SD = 1.03 vs M = 2.02-min; SD = .29). Only 60% of children in the DCR condition drew for the fully allotted 5-min, and only two children said that they would have liked more time. This gives some indication that a 5‐min time limit was sufficient for the children.

Interview duration

To determine whether the duration of the interviews was the same for each of the three conditions, we measured interview duration from when the free narrative account began until the end of the interview. The results of a 2 (age) × 3 (condition) Analysis of variance (ANOVA) revealed a significant effect for condition, F(2, 135) = 4.68, p = .011, η_p² = .065. Follow‐up post hoc Tukey comparisons indicated that the MCR interview and the DCR interview were significantly longer (M = 12.60-min, 95% CI [11.74, 13.49], and M = 12.55-min, 95% CI [11.65, 13.44], respectively) than the NCR interview (M = 10.88-min, 95% CI [9.96, 11.78]). There was no main effect for age, F(1, 135) = 1.60, p = .208, and no significant interaction, F(2, 135) = 0.13, p = .882. Given that the difference in interview length may have influenced participants' responses in the free recall and open‐ended prompt stages of the study, it was added as a covariate for the main analyses.

Drawing ability

To examine children's drawing abilities, their total DAP scores were submitted to a 2 (age) × 3 (condition) between‐groups ANOVA. As expected, older children had higher drawing scores (M = 122.79, SD = 23.79, 95% CI [117.45, 127.81]) than younger children (M = 66.89, SD = 18.71, 95% CI [61.67, 71.89]), F(1, 132) = 230.47, p < .001, = η_p² = .63. Importantly, there was no difference in children's drawing ability between the different interview conditions, F(2, 132) = 0.59, p = .587.

Free recall and open‐ended prompts

Next, we turn to our main analyses. The mean number of correct, incorrect, and confabulated event details recalled for each age group summed over the free narrative and extended open‐ended prompt phase is displayed in Table 2. To determine whether the MCR and DCR enhanced children's recall of the magic show, the completeness and accuracy of their recall were examined. Overall, the younger children recalled approximately 15% of the details from the event; 90% of these details were accurate. The older children recalled 27% of the details from the event; their accuracy was 88%.

Table 2. Correct, incorrect, and confabulated details reported during free recall and in response to open‐ended prompts

	NCR	MCR	DCR
Interview condition	Younger children (N = 70)
	M (SD)	M (SD)	M (SD)
Correct	27.77 (11.45)	30.88 (11.42)	29.68 (8.61)
95% CI	[25.73, 36.02]	[25.76, 35.02]	[24.03, 34.10]
Incorrect	2.68 (2.06)	3.73 (3.31)	4.00 (2.56)
95% CI	[1.95, 4.60]	[2.44, 4.83]	[2.58, 5.18]
Confabulated	1.05 (1.53)	1.57 (2.04)	1.50 (1.57)
95% CI	[0.42, 2.01]	[0.84, 2.27]	[0.69, 2.25]
	Older children (N = 71)
	M (SD)	M (SD)	M (SD)
Correct	54.60 (16.26)	52.39 (15.37)	55.88 (14.39)
95% CI	[50.91, 60.79]	[45.76, 55.67]	[49.87, 59.35]
Incorrect	6.65 (3.88)	7.65 (3.37)	6.48 (3.85)
95% CI	[5.61, 8.17]	[6.05, 8.61]	[5.01, 7.46]
Confabulated	2.13 (2.20)	1.96 (1.92)	1.64 (1.77)
95% CI	[1.43, 2.96]	[1.10, 2.63]	[0.84, 2.31]

Notes. Standard deviations appear in parentheses. NCR = no context reinstatement; MCR = mental context reinstatement; DCR = drawing context reinstatement.

To examine the impact of interview condition on completeness, a 2 (age) × 3 (condition) ANOVA was conducted on the proportion of details that children recalled. Older children recalled more details (M = 27.17%, SD = 7.59, 95% CI [25.50, 28.32]) than younger children (M = 14.76%, SD = 5.28, 95% CI [13.64, 16.50]); in other words, there was a significant main effect for age, F(1, 134) = 135.97, p < .001, η_p² = .504. There was no significant main effect for interview, F(2, 135) = 0.61, p = .542. Participants recalled a similar percentage of details regardless of interview condition (NCR: M = 20.74%, SD = 9.73, 95% CI [19.93, 23.55]; MCR: M = 20.49%, SD = 8.57, 95% CI [18.60, 22.1]; DCR: M = 21.81%, SD = 8.90, 95% CI [19.26, 22.74]). The interaction between age and condition was also not significant, F(2, 135) = 0.69, p = .505.

To examine the impact of interview condition on overall accuracy, a 2 (age) × 3 (condition) ANOVA was conducted. There was no difference in accuracy for older (M = 88.72%, SD = 5.33, 95% CI [87.20, 90.41]) and younger children (M = 89.34%, SD = 8.08, 95% CI [87.63, 90.90]); that is, there was no significant main effect for age, F(1, 134) = 0.18, p = .675. There was also no significant main effect for interview, F(2, 134) = 0.50, p = .607. Accuracy did not depend on interview condition (NCR: M = 89.93%, SD = 6.72, 95% CI [87.66, 91.70]; MCR: M = 88.15%, SD = 7.79, 95% CI [86.30, 90.24]; DCR: M = 89.07%, SD = 5.77, 95% CI [87.25, 91.10]). Finally, the interaction between age and condition was also not significant, F(2, 134) = 1.14, p = .324.

It is possible that the non‐significant effect of interview condition on children's recall was due to asking an identical number of follow‐up open‐ended prompts in each of the interview conditions (meaning that it was possible that not all the content was followed up from information elicited in the free recall phase, creating a ceiling effect). Therefore, a 2 (age) × 3 (condition) ANOVA was conducted on the number of correctly recalled detail in the free recall phase. A non‐significant result was found F(1, 135) = .254, p = .776.

Taken together, these results suggest that while older children reported more details than younger children, their accounts were similarly accurate. Interview condition had no impact: regardless of whether they were interviewed using NCR, MCR, or DCR, children had similar levels of completeness and accuracy. Next, we determined whether this pattern of results was the same for children's responses to the suggestive questions.

Drawing ability and memory performance

To investigate the relationship between drawing ability and event recall—that is, whether children's drawing ability limited the beneficial effect of DCR—six Pearson product‐moment correlations (two‐tailed) were calculated between the total DAP scores and total correct details for each interview and each age group. No relationships between drawing scores and correct details were found for older children (NCR: r = −.11, p = .612; MCR: r = .11 p = .626; DCR: r = .17, p = .428). For younger children, an unexpected significant relationship was found between drawing ability and correct details for all interview conditions (NCR: r = .58, p = .001; MCR, r = .61, p = .001; DCR, r = .63, p = .001). As the relationship between drawing ability and event recall was found across all interview conditions and not in DCR alone, it suggests that drawing ability does not limit the effect of DCR on event recall, as found by Butler et al. (1995). Rather, younger children's (but not older children's) scores on the DAP may represent some general ability at a particular developmental stage, and not drawing ability alone.

Children's suggestibility

To examine the effects of the true‐ and false‐biased questions on children's memories, a 2 (age) × 3 (condition) × 2 (bias) repeated measures ANOVA was conducted on children's responses to the true‐ and false‐biased questions. There was a significant main effect for bias: children gave more correct responses to the true‐biased questions (M = 3.64, SD = 0.58, 95% CI [3.54, 3.73]) than the false‐biased questions (M = 3.13, SD = 0.96, 95% CI [2.97, 3.30]), F(1, 134) = 26.90, p < .001, η_p² = .167. There was a significant main effect for age: older children gave more correct responses to the biased questions (M = 6.96, SD = 1.28, 95% CI [6.66, 7.23]) than younger children (M = 6.47, SD = 1.14, 95% CI [6.19, 6.76]), F(1, 134) = 10.22, p = .002, η_p² = .071. Finally, there was a main effect for interview, F(2, 134) = 3.37, p = .037, η_p² = .048. Post hoc Tukey tests revealed that children gave more correct responses when interviewed using the DCR (M = 7.09, SD = 0.90, 95% CI [6.72, 7.41]) than the NCR (M = 6.47, SD = 1.46, 95% CI [6.11, 6.82]). There was no significant difference between the DCR and MCR (M = 6.59, SD = 1.22, 95% CI [6.27, 6.95]), or between the MRC and NCR. None of the interactions among age, condition, and bias were significant, all F < 0.65, all p > .524.

An analysis of children's ‘don't know’ responses revealed that there was no difference in the number of these responses provided by younger (M = 0.33, SD = 0.70, 95% CI [0.19, 0.47]) and older children (M = 0.16, SD = 0.40, 95% CI [0.02, 0.29]), F(1, 134) = 3.22, p = .075. There was also no effect for interview condition, F(2, 134) = 0.06, p = .940, and no significant interaction, F(2, 134) = 0.18, p = .839 (Table 3).

Table 3. Correct, incorrect, and ‘don't know’ responses to the biased questions

	NCR	MCR	DCR
Interview Condition	Younger children (N = 70)
	M (SD)	M (SD)	M (SD)
Correct	6.36 (1.14)	6.35 (1.29)	6.74 (0.94)
95% CI	[5.86, 6.87]	[5.88, 6.81]	[6.22, 7.23]
Incorrect	1.23 (0.97)	1.35 (1.16)	0.86 (0.64)
95% CI	[0.86, 1.60]	[1.01, 1.69]	[0.49, 1.23]
Don't know	0.36 (0.73)	0.27 (0.67)	0.36 (0.73)
95% CI	[0.12, 0.61]	[0.05, 0.50]	[0.12, 0.61]
	Older children (N = 71)
	M (SD)	M (SD)	M (SD)
Correct	6.57 (1.73)	6.87 (1.10)	7.40 (0.76)
95% CI	[6.07, 7.06]	[6.38, 7.36]	[6.93, 7.87]
Incorrect	0.96 (0.98)	0.87 (0.76)	0.44 (0.58)
95% CI	[0.59, 1.32]	[0.51, 1.23]	[0.09, 0.79]
Don't know	0.14 (0.35)	0.17 (0.39)	0.16 (0.47)
95% CI	[0.11, 0.38]	[0.06, 0.41]	[0.07, 0.39]

Notes. Standard deviations appear in parentheses. NCR = no context reinstatement; MCR = mental context reinstatement; DCR = drawing context reinstatement.

These results suggest that, overall, children responded less accurately to false‐biased questions than true‐biased questions. Similar to their responses to free recall questions and open‐ended prompts, older children were more accurate than younger children in their responses to biased questions. However, in contrast to the above findings, children interviewed with the DCR had more accurate responses to biased questions than children interviewed with the NCR.

Discussion

The results of the current study showed that the two context reinstatement techniques—MCR and DCR—did not affect children's responses to free recall or open‐ended prompts, yet interview condition affected children's accuracy when they answered suggestive questions. Those in the DCR condition who drew the event prior to recall gave more accurate answers to true‐ and false‐biased questions than those in the NCR interview condition. However, unexpectedly, no difference in the accuracy of children's responses to suggestive questions was found between the DCR and MCR conditions, or between the NCR and MCR conditions. The pattern of results is consistent with what we would expect developmentally. Older children reported more details and were more accurate than younger children; older children were less misled by the suggestive questions than younger children.

Surprisingly, neither context reinstatement interview had a beneficial effect on the completeness and accuracy of younger and older children's responses to free recall and open‐ended prompts. Regardless of whether they were interviewed using the NCR, MCR, or DCR, children had similar levels of completeness and accuracy. Prior research with adults demonstrated the DCR to be as effective as MCR, with both more effective than an NCR instruction (Dando et al., 2009). Further, in their field study, Katz and Hershkowitz (2010) found that DCR, compared with NCR, significantly increased the number of details reported by alleged child sexual abuse victims. However, contrary to what was predicted, the present study did not show a beneficial effect of DCR (or a MCR) on recall compared with a control condition of NCR.

There are a number of possible reasons why MCR and DCR did not facilitate children's free recall and responses to open‐ended prompts. First, it is possible that children in the current study mentally reinstated the context of the event regardless of whether they were specifically instructed to do so. In other words, merely interviewing them about the event might have been enough for them to naturally think back to the context of the event. Thus, further instructions to reinstate the context might not have had a significant impact (see also Dietze, Sharman, Powell, & Thomson, 2013, for a discussion of this point).

Second, we did not ask children specific questions about the event, which is where some other studies have demonstrated the impact of context reinstatement. For example, Dietze et al. (2010) found that MCR instructions affected children's responses to specific cued recall questions but not their responses during free recall. They suggested that this difference between free and cued recall might have occurred because mental reinstatement enhances recall through establishing the boundaries of a search set (see also Dalton, 1993; Smith, Glenberg, & Bjork, 1978). These boundaries may enhance the activation of the memory items within the set. If the activation is low, then cued recall questions should be more effective in probing the set than less specific free recall. Thus, in this situation, cued recall questions will demonstrate more of a benefit from mental reinstatement instructions than cued recall. Similarly, in the current experiment, it is possible that free recall and open‐ended prompts did not effectively probe the memory set. As a result, we found no significant effect of mental or drawing context reinstatement.

Third, it is possible that the current study lacked the statistical power to detect any significant differences between conditions that were actually present. This explanation is unlikely, however, given that we conducted an a priori power analysis to determine an adequate sample size using G*Power (Faul, Erdfelder, Lang, & Buchner, 2007). Following Howell's (2002) recommendation of the use of prior research as a rough approximation of the expected population effect size, Katz and Hershkowitz's (2010) reported effect size for interview condition (η_p² = .11) was used, along with a significance criterion of .05 and the desired power coefficient of .80. The output indicated that a total sample size of 82 yielded power at a respectable 80.5% to detect an effect of similar size. Thus, a sample of 141 children was deemed more than adequate.

The expectation that drawing would more effectively reinstate the context compared with the interviewer supplying verbal prompts was not supported. While it was theorised that drawing would assist children generate personally relevant and thus more effective retrieval cues, it is possible that drawing did not provide enough of a scaffold to overcome deficits in the internal generation of retrieval cues. Children's limited ability to self‐generate retrieval cues, compared with adults, may explain the difference between our study and Dando et al. (2009). They found that the completeness of accounts was improved when the adult witness was able to draw prior to recalling the event.

The main implication from the current experiment is that asking younger and older child witnesses to draw an event prior to freely recalling it increases their accuracy when answering questions that are true‐ or false‐biased than where no instruction is given. The fact that no positive correlation was found between drawing ability and the amount of information recalled for only those in the drawing condition indicates that the protective function of drawing prior to recall is not limited to those who can draw well. Because we did not find that event recall was improved, compared with the control interview, our results do not support Memon et al.'s (2010) explanation that resistance to suggestive questioning lies only in improving memory, and thus confidence to question suspicious information. Rather, once children have drawn the event, they have a record of the details that they can refer to during questioning, perhaps protecting against source monitoring failure. Alternatively, an external reference to the memory of the event may reduce children's uncertainty of their recall possibly invoked by biased information, thereby reducing interviewer compliance. Yet another possible explanation of the protective properties of drawing against suggestive questioning may lie in the effectiveness of reinstating the context via drawing instead of mentally. That is, utilising a drawing to aid in responding to a prompt designed to elicit a narrow response may aid the child to be more effortlessly directive in their memory search, and thus more accurate. This is in comparison to the more effortful mental reinstatement of the contextual feature which must be done each time a question is asked to fully utilise the memory aid (while simultaneously the child is holding the given prompt in memory). This last explanation may explain why there was no difference between MCR and NCR, and also between MCR and DCR.

These protective properties of the DCR—compared with the NCR—are promising as a first step. It is possible that asking children to self‐generate retrieval cues through drawing at the beginning of the investigative process might be used to protect the child from changing their testimony in the face of repeated questions if they were allowed to refer to their drawing. Future research is needed to clarify the extent of this easily implemented, potentially protective exercise.