Mental time travel and eyewitness recall 1 Running Head: Mental time travel and eyewitness recall Mental Reinstatement of Context: Do individual differences in mental time travel and eyewitness occupation influence eyewitness performance over different delay intervals?

The Cognitive Interview is a memory-enhancing interview protocol designed to optimize the access and retrieval of eyewitness memories. Its Mental Reinstatement of Context (MRC) component requires interviewees to mentally reconstruct the crime event they witnessed. Individual differences in mental time travel (MTT) relate to the extent to which a person mentally re-experiences personal events from his or her past. Individual differences in MTT have been found to predict correct recall of a simulated crime event under immediate MRC recall conditions. To explore the relationship between MTT and performance under MRC conditions further, the present study presented a simulated crime video to 30 police officers and 26 members of the public. Eyewitness recall was tested under MRC conditions either immediately or one week later. Participants’ general MTT and also MTT relating specifically to the crime video itself was measured via self-report. Less correct information and more confabulations were produced after one week but delay had no effect on the amount of incorrect information reported. No difference in recall was found between police officers and members of the public. Better quality MTT relating to the crime video was found to be a positive predictor of the amount of information correctly recalled under immediate conditions but not after one week. General MTT scores did not predict correct recall under either delay condition. Interviewers need to be aware that, due to individual differences, some witnesses may perform better under the MRC component than others.


Introduction
Eyewitnesses often hold key information about the events that they have seen. Indeed, an eyewitness may be the only source of information available to investigators to identify and bring to justice those responsible for perpetrating a crime. The Cognitive Interview (Geiselman et al., 1984) is a commonly used interviewing technique designed to facilitate the recall of eyewitnesses. The current study explored eyewitness recall under its Mental Reinstatement of Context (MRC) component. Individual differences in the extent to which individuals mentally re-experience personally experienced past events, known as mental time travel (MTT; e.g., Tulving, 2002), have been found to influence the accuracy of eyewitness recall under MRC conditions when participants are questioned after a short delay (Smith-Spark, Bartimus & Wilcock, 2017). The current research built on this work; firstly, by testing eyewitness recall under MRC conditions either immediately after witnessing a simulated crime event or after a delay of one week and, secondly, by seeing whether MTT for the crime event itself, as well as general levels of MTT (as tested previously by Smith-Spark et al.), would predict eyewitness performance. More generally, it extended past research on the Cognitive Interview by having a serving police officer (the first author, KB) administering the tasks, testing police officers as well as university students and members of the public (thus obtaining a broader sample than just students), and using a simulated crime video filmed from a first-person eyewitness perspective rather than a third-person perspective (thus adding to the verisimilitude of the event witnessed).

The Cognitive Interview
The Cognitive Interview was developed by Geiselman et al. (1984; see also Fisher & Geiselman, 1992) as a memory-enhancing interview protocol designed to aid the elicitation of crime event information from eyewitnesses and cooperative suspects. The Cognitive Interview Mental time travel and eyewitness recall 4 has been found to elicit greater recall accuracy and completeness of reporting compared with a standard interview (Fisher, Geiselman & Amador, 1989;Geiselman, Fisher, MacKinnon & Holland, 1986;Stein & Memon, 2006). Its effectiveness as an interviewing tool has been demonstrated over different eyewitness age groups, in different countries, and in the field (e.g., Fisher et al., 1989;Paulo, Albuquerque & Bull, 2013;Stein & Memon, 2006;Verkampt & Ginet, 2009).
The Cognitive Interview utilizes different interviewing components to maximize eyewitness recall. At a theoretical level, it is underpinned by Bower's (1967) Multiple Trace Theory. This theory argues that different features make up a memory trace and that, as a result, multiple retrieval paths could be followed to access the same encoded event. Each Cognitive Interview component, therefore, triggers a different facet of memory recall to maximize the information accessed at retrieval. In this way, the interviewee is provided with alternative recall opportunities should one technique prove to be unsuccessful in obtaining information useful to the investigation (Westera, Kebbell & Milne, 2011). The four components of the Cognitive Interview are Report Everything, Change Perspective, Change Order, and Mental Reinstatement of Context (MRC). Report Everything instructs the eyewitness to report all the details that they can, even if they do not think them important. This component helps to prevent witnesses from withholding information that they do not consider relevant (Koriat & Goldsmith, 1996). Change Order requires the eyewitness to recall the event in a different order, such as in reverse chronological order, while Change Perspective asks the eyewitness to try to recall the event from an alternative perspective, such as from the perspective of another person present (Geiselman, Fisher, MacKinnon & Holland, 1985). The MRC component employs cues to prompt the eyewitness mentally to recreate the physical environment and emotions that were present at the Mental time travel and eyewitness recall 5 time of the witnessed event, effectively imagining themselves back at the crime scene, before being asked to recall their memory of that event (Davis, McMahon & Greenwood, 2005;Geiselman et al., 1986).
The MRC component was the focus of the current study. This component is aimed at increasing the overlap of features between encoding and retrieval cues (Geiselman et al., 1986) and is widely regarded as one of the most effective components of the Cognitive Interview (Dando, Wilcock & Milne, 2008). It is based on the Encoding Specificity Hypothesis (Tulving & Thomson, 1973) which argues that recall can be improved by providing cues to reinstate the context at the time that an event was encoded. The interviewer provides a series of short verbal prompts to the witness, such as "Think about that day", "What was the weather like?", "Who had you seen or spoken to that day", "Think about the room you were in", "Try and picture the room in your mind", "Did you smell anything in that room?", with sufficient time being allowed after each prompt to enable the eyewitness to re-create the event mentally. perspective, was shown to the participants. To avoid physical reinstatement of context from the surrounding environment, the participants were moved to a different room before being asked to Mental time travel and eyewitness recall 6 write down personal memories from specified time-periods after being given a cue word (Crovitz & Schiffman, 1974 Johnson, Foley, Suengas & Raye, 1988) for each recalled memory. The 12-item MCQ requires respondents to rate the extent to which they re-experienced a personally lived event along a range of phenomenological dimensions and is commonly used as a measure of MTT (e.g., Arnold, McDermott & Szpunar, 2011). An overall mean MCQ score was generated to provide a measure of each participant's general quality of MTT. Following this, the participants were allocated to one of the two Cognitive Interview conditions and either the MRC or the Report Everything instructions were presented. After receiving these instructions, the participants were given three minutes to write down what they could remember of the crime video. Individual differences in mental time travel, as measured by MCQ scores, were found to be positive predictors of both correct and incorrect eyewitness recall using the MRC component of the Cognitive Interview. However, no predictive relationship between MCQ scores and eyewitness performance was observed in the Report Everything condition, despite equivalent levels of accuracy being found between the two conditions. Thus to explore this relationship in more depth. To this end, the current study measured the influence of individual differences in MTT on recall of a simulated crime event video under MRC conditions after either a minimal delay or a delay of one week. Both general levels of MTT (as employed by Smith-Spark et al., 2017) and MTT relating specifically to the recall of the crime event video were investigated. Memon, Meissner, and Fraser's (2010) meta-analysis indicated that little research has been conducted on the Cognitive Interview over differing delay intervals. In general terms, the delay incurred between the encoding and recall of information has a detrimental effect on memory. As more time elapses since encoding the information, the memory deteriorates and becomes less retrievable (e.g., Turtle & Yuille, 1994). Furthermore, finer-grained details are lost at a faster rate than basic information (Goldsmith, Koriat & Pansky, 2005). Memon et al.'s metaanalysis found that the duration of the delay between encoding and recall reduced the effect size for correct recall using the Cognitive Interview, while the effect size for confabulations increased. However, the advantage of the Cognitive Interview over a control structured interview was still sizeable.
As well as adding to the literature on the relationship between MTT and eyewitness performance under MRC conditions over differing delay intervals, the study also explored the effect of participant occupation on recall, comparing the eyewitness performance of law enforcement professionals with members of the public. As highlighted by Memon et al. (2010), only a small percentage of studies have assessed the eyewitness recall performance of law enforcement professionals under the Cognitive Interview. There have been mixed findings when the eyewitness memory of police officers has been compared with that of members of the public.
Some studies have found that police officers recalled significantly higher quantities of correct Mental time travel and eyewitness recall 8 information than members of the public, with no increase in incorrect information (e.g., Christianson, Karlsson & Persson, 1998;Lindholm, Christianson & Karlsson, 1997). However, other studies have reported no difference in recall accuracy between police and civilian eyewitnesses (e.g., Stanny & Johnson, 2000). Given the nature of their role and its concomitant exposure as eyewitnesses to situations which require later reporting, more research is needed to understand the eyewitness recall performance of law enforcement professionals. Therefore, the current study compared the performance of serving police officers with non-police participants (a group consisting of members of the public and university students) to explore the influence of It was predicted that the quantity of correct information would decrease after a one-week delay, while the quantity of incorrect recall and confabulations would increase. These predictions were based on studies investigating memory decay (e.g., Gabbert, Memon & Allan, 2003;Gabbert, Memon, Allan & Wright, 2004;Goldsmith et al., 2005;Turtle & Yuille, 1994). Given the equivocal findings regarding the relative eyewitness recall performance of police officers compared with non-police participants (e.g., Christianson, Karlsson & Persson, 1998;Lindholm, Christianson & Karlsson, 1997;Stanny & Johnson, 2000), it was an open question as to whether differences would be found between the two participant groups or whether there would be an interaction between occupation group and delay interval. From Smith-Spark et al.'s (2017) findings, it was hypothesized that individual differences in MTT would be a significant predictor Mental time travel and eyewitness recall 9 of the quantity of correct and incorrect information recalled under the immediate recall condition. Given the MRC's reported effectiveness over longer time intervals (see Memon et al., 2010, for a review) and the predictive relationship found between MTT and recall under the MRC component (Smith-Spark et al., 2017), it was expected that a similar relationship would emerge for the one-week delay condition. It was predicted that MTT scores relating to the crime event video itself would more closely predict recall than more general MTT scores derived from events unrelated to the crime video. Firstly, a 2 x 2 between-subjects design was used to investigate the effects of occupation group and delay condition on, firstly, MCQ scores and, secondly, on eyewitness recall. The factors were occupation group (levels: police, non-police) and delay condition (levels: immediate recall, delayed recall). The dependent variables were the number of bits of information correctly recalled, the number of bits of information incorrectly recalled, and the number of confabulations. Incorrect details were defined as errors of detail (e.g., an eyewitness stating that the colour of a person's hat was red when in fact it was black). Confabulations were commission errors (Memon et al., 2010), such as an eyewitness describing a hat worn by a suspect when, in fact, no hat was present.
Multiple regression was employed to determine the extent to which MCQ scores (both general and specifically for the crime event video) predicted eyewitness performance under the MRC component of the Cognitive Interview.

Participants
Fifty-six adults (31 females, 25 males, mean age = 38 years, SD = 12, range = 46) took part. Of the participants, 30 were police officers and 26 were either university students or members of the public. The participants were assigned randomly to one of two recall conditions, in which they were tested for their memory of a simulated crime event either immediately or after a delay of one week. Group characteristics are displayed in Table 1. The participants were tested either individually or in groups (when time constraints and resource limitations prevented individual testing). Group testing usually involved two participants but one group consisted of six participants and another group consisted of seven. Similar variation in the size of the groups tested has been reported previously by Smith-Spark et al. (2017). No inducement or rewards were offered for participation.

TABLE 1 ABOUT HERE
A one-way unrelated ANOVA indicated that there was a significant difference in age between the occupation groups, F(3, 52) = 4.12, MSE = 113.60, p = .011, η 2 p = .19. Post-hoc comparisons indicated that the non-police immediate recall group was significantly older than the police delayed recall group (p = .046), the police immediate recall group (p = .015), and the non-police delayed recall group (p = .032). No other comparisons were significant (p = 1.00).
There was no significant association between gender and delayed recall condition, χ 2 (1, N = 56) < 1, p = .453. However, there was a significant association between gender and occupation group, χ 2 (1, N = 56) = 9.134, p = .003, such that there were more males (N = 19) than females Cue word recall Using the Crovitz-Schiffman technique (Crovitz & Schiffman, 1974), the participants were asked to recall past, personally lived events from three different temporal locations. The timeframes in question were one day ago, one week ago, and one month ago. Two cue words were provided by the researcher for each timeframe for two separate recall attempts. This gave a total of six events to recall. The same cue words were employed as those used by Smith-Spark et al. (2017), namely "garden" and "kitchen". These words were matched for age of acquisition and imageability using ratings taken from Bird, Franklin, and Howard's (2001) database and were also matched for Celex Word Frequency (Baayen, Piepenbrook & van Rijn, 1993). The timeframe for the memory to be recalled and the cue word order was counterbalanced across conditions. Order 1 required memories to be recalled in the order of one day ago through to one month ago, while Order 2 required memories to be recalled in the reverse order to Order 1. The instructions were presented as a computer slide show as well as verbally.

Memory Characteristics Questionnaire
The participants were asked to complete a modified MCQ (Johnson et al., 1988) after each event that they were required to recall. The participants were instructed to answer 12 questions. Each question required the participants to rate on a one to seven scale the extent to based on the recall of the simulated crime video (referred to as video MCQ hereafter) was also obtained following the same procedure as described above.

Procedure
Ethical approval was granted by the relevant research ethics committee at the authors' host institution. Informed consent was obtained from the participants prior to testing.
Initially, the participants were informed that they would watch a mock-crime video and would be asked questions about it later. After viewing the video, the participants were taken to a different room to complete the remainder of the tests. This was done to ensure that physical reinstatement of context would not influence recall.
The cue word recall task was then completed. The experimenter read out the adapted Crovitz-Schiffman instructions. The participants were asked to use the first memory that came to mind. They were informed that it was not important for the memory necessarily to be related to the cue word as this had been provided simply to assist them. However, they were reminded that it was very important to ensure that the memory which they wrote about was from the correct timeframe. For each recalled event, the cue word and timeframe appeared on a computer slide.
The participants were given three minutes to write down their memory for that cue word and timeframe in their answer books, before being asked to cease writing. The slide containing the cue word and the timeframe remained visible to the participants until recall of that memory had been completed. After the three minutes had elapsed, the participants were asked to complete an MCQ relating to their recall of the event about which they had written. The Crovitz-Schiffman Mental time travel and eyewitness recall 14 cue word recall task and completion of the associated MCQ was completed twice for each timeframe (once for each cue word), giving a total of six recalled events.
The participants' memory for the simulated crime event was then tested. In the immediate recall condition, the participants were tested straight after the completion of the six Crovitz-Schiffman and MCQ trials. The recall of the participants assigned to the delayed recall condition was tested one week later. They were asked not to discuss the contents of the video with anyone else in the meantime.
Regardless of the condition to which they had been assigned, all the participants received the same MRC instructions prior to recall and their memory for the simulated crime event was tested following the same procedure. The instructions presented a range of cues to help place the participants back at the time that they watched the video. The participants were asked to think back to the simulated crime that they watched as though they were trying to remember something that they had lost and were trying to remember where they had last seen it. They were then asked to think about what they had been doing just before watching the crime video, then to think about the room in which they had viewed it (thinking about its appearance, any smells, any noteworthy items present, the physical layout of the room, and where they sat to watch the screen). Following this, they were asked to think whether there was anyone else present in the room and where they were positioned. Once the participants had built up a good mental image of the room, they were then asked to write down as much information as they could recall about the mock crime event video in the answer book provided. They were told that it was important not to guess details or make them up and that it was fine to say if they did not know a detail. The participants were allowed five minutes to complete this phase before completing the MCQ Mental time travel and eyewitness recall 15 relating to the crime video. The answer books were then collected and the participants were debriefed.

Comparison of scores under individual and group testing conditions
To determine whether any group differences in scores were evident between participants tested individually compared with those tested in groups, independent-samples t-tests were run on both MCQ scores and eyewitness recall performance.
There was no significant difference between the mean cue word MCQ scores of the

Relationships between MCQ scores and eyewitness memory performance
Analyses were performed on the amount of information correctly produced by the overall sample of 56 participants and by the immediate and delayed recall conditions separately.
Analyses were not performed on the incorrectly recalled information or the confabulation data due to the low number of non-zero values obtained.
The mean amount of information correctly recalled by the 56 participants was 23.04 (SD = 9.08). The mean cue word MCQ score was 4.86 (SD = 0.64) and the video MCQ score was 4.60 (SD = 0.84). Mean cue word MCQ score and video MCQ score correlated significantly, r = .511, p < .001. There was no significant correlation between mean cue word MCQ score and the amount of information correctly recalled, r = .126, p = .177. However, there was a very significant correlation between video MCQ score and the amount of information correctly recalled, r = .365, p = .003.
The multiple correlation between the predictor variables and the amount of information correctly recalled was .372. The regression model accounted for 11% of the variance in correct recall (adjusted-R 2 ) and the model significantly predicted the amount of information correctly recalled, F(2, 53) = 4.25, p = .019. When the predictor variables were considered individually, video MCQ score was found to be a significant positive predictor of the amount of information correctly recalled, Standardized β = .41, t = 2.74, p = .008, while mean cue word MCQ score did not predict correct recall significantly, Standardized β = -.08, t < 1, p = .584. The positive association between video MCQ score and the amount of information correctly recalled is shown in Figure 1.

FIGURE 1 ABOUT HERE
To explore the contribution of MTT to eyewitness memory at a finer-grained level across the different delay intervals, regression analyses were also performed separately on the amount of information correctly recalled by the immediate recall and delayed recall conditions.

Immediate recall condition
Mental time travel and eyewitness recall 20 The mean amount of information correctly recalled by the 26 participants in the immediate recall condition was 26.81 (SD = 7.13). The mean cue word MCQ score was 4.87 (SD = 0.51) and the video MCQ score was 4.91 (SD = 0.86). Mean cue word MCQ score and video MCQ score correlated significantly, r = .459, p = .009. There was no significant correlation between mean cue word MCQ score and the amount of information correctly recalled, r = .100, p = .314. However, there was a very significant correlation between video MCQ score and the amount of information correctly recalled, r = .459, p = .009.
The multiple correlation between the predictor variables and the amount of information correctly recalled was .484. The regression model accounted for 17% of the variance in correct recall (adjusted-R 2 ) and the model significantly predicted the amount of information correctly recalled, F(2, 23) = 3.51, p = .047. When the predictor variables were considered individually, video MCQ score was found to be a significant positive predictor of the amount of information correctly recalled, Standardized β = .53, t = 2.59, p = .016, while mean cue word MCQ score did not predict correct recall significantly, Standardized β = -.15, t < 1, p = .487. The relationship between video MCQ score and the amount of information correctly recalled is shown in Figure   2.

Delayed recall condition
The mean amount of information correctly recalled by the 30 participants in the immediate recall condition was 19.77 (SD = 9.41). The mean cue word MCQ score was 4.86 (SD = 0.74) and the video MCQ score was 4.34 (SD = 0.75). There was a highly significant correlation between mean cue word MCQ score and video MCQ score, r = .625, p < .001. The mean cue word MCQ score and the amount of information correctly recalled did not correlate Mental time travel and eyewitness recall 21 significantly, r = .149, p = .216, nor did video MCQ score and the amount of information correctly recalled correlate significantly, r = .131, p = .246.
The multiple correlation between the predictor variables and the amount of information correctly recalled was .156. The adjusted-R 2 value was -.05 and the model did not significantly predict the amount of information correctly recalled, F(2, 27) < 1, p = .716. Neither video MCQ score, Standardized β = .06, t < 1, p = .802, nor mean cue word MCQ score, Standardized β = .11, t < 1, p = .654, were significant predictors of the amount of information correctly recalled.
While non-significant, the relationship between video MCQ score and the amount of information correctly recalled is shown in Figure 3 for comparison with the immediate recall condition. between the police and non-police participant groups. Each of these findings will now be considered in more detail.
As predicted, the participants in the delayed recall condition provided significantly less correct information than the participants in the immediate recall condition. These findings are consistent with the previous literature in showing that the quality of eyewitness memory decreases over time (e.g., Goldsmith et al., 2005;Turtle & Yuille, 1994). No difference in the amount of incorrect information produced was found between the two delay conditions. More confabulations were found to be produced when recall was delayed than if recall took place immediately. The latter finding is consistent with Memon et al.'s (2010) meta-analysis which found that significantly more confabulations were produced under the Cognitive Interview when there was a delay between encoding and recall. However, the quantity of confabulations and incorrect details was very small in the current study. This may be because the participants were reminded not to guess or make things up and were also told that if they did not know a detail then that was acceptable. Thus, a report option was given when they received the MRC instructions and this has been found to contribute towards greater recall accuracy (Koriat & Goldsmith, 1996) and reduce incorrect information.
Mental time travel and eyewitness recall 23 No effect of participant occupation on eyewitness performance was found in the current study. As noted in the Introduction, findings in this area have been equivocal but the present results are broadly consistent with Stanny and Johnson (2000) who reported no significant difference between police officers and members of the public in the amount of information correctly recalled. However, they are contrary to the findings of Christianson et al. (1998) and Lindholm et al. (1997) who found that police participants provided significantly more correct recall than non-police participants but did not produce an increased amount of incorrect detail.
This section will turn now to the relationship between MTT and eyewitness recall under MRC conditions. Video MCQ score was found to be a significant predictor of correct recall overall and in the immediate recall condition. This finding is consistent with Smith-Spark et al. (2017)  The mock crime video used in the current study was filmed from a first-person perspective and the participants viewing the video would, thus, have encoded the event from the point of view of being an eyewitness to it. Perhaps unsurprisingly, therefore, given that the crime event and any phenomenological aspects attached to it would be fresh in the participants' minds, a significantly higher video MCQ score was observed in the immediate condition than in the delayed recall condition. In contrast, the mean cue word MCQ score, representing general levels of MTT, was significantly higher than the video MCQ score overall but did not differ between the two delay conditions. This difference in MCQ scores may reflect the much greater phenomenological experience attached to the recall of lived events compared with the recall of a filmed event. Experience of the phenomenological dimensions of memories probed by the MCQ, therefore, may be limited when watching a video, possibly reducing the number or quality of memory cues available (such as smell, taste, and sense of embodiment). This may explain why the predictive relationship was not found over the longer delay interval. Wright and Holliday (2007) suggest that live events provide a broader variety of available memory cues. While the use of a video filmed from a first-person perspective in the current study appears to have moderated this limitation for the immediate recall condition, the use of a first-person perspective video may still not be sufficient in the delayed recall condition to provide the cues necessary for MTT to occur to a meaningful extent. Although Smith (1988) suggested that MRC increased accessibility to memory especially after a delay, MRC may not be effective if the number and strength of cues are limited from the outset by the medium through which the eyewitness experienced the event in question. In support of this argument, participants actively involved in an event have been found to remember more than those not directly involved (Baker-Ward, Hess & Flanagan, 1990). Furthermore, live scenarios and experimental set-ups requiring participant involvement have demonstrated greater effects of the Cognitive Interview (Köhnken et al., 1999) The Crovitz-Schiffman word-cue recall task required each participant to recall a total of six events across three timeframes. It is possible that recalling six events may have become monotonous for the participants and, therefore, may have affected how much care they took to score each MCQ. Set against this objection, there was a significant difference in MCQ scores between the timeframes, with the event from one day ago having the highest MCQ score and the event from one month ago having the lowest MCQ score. This would suggest that the participants were engaging well with the task and putting careful, considered thought into their responses. Moreover, the pattern of MCQ scores fits comfortably with MTT theory. Roberts and Feeney (2009) have argued for a bi-cone distribution, with diminishing quality of MTT reported as the temporal distance from the present moment increases. In addition, the order of testing was counterbalanced to protect against potential order effects and, as reported in the Results, no difference between the two orders of presentation was found. No significant difference was found in any MCQ score between the police and non-police participants. Furthermore, there was no significant difference in mean cue word MCQ score between the two delay conditions.
In conclusion, this study explored different aspects of eyewitness recall under the MRC component of the Cognitive Interview. Participant occupation was not found to affect recall. A longer delay between witnessing the crime event and recalling it resulted in a lower amount of Mental time travel and eyewitness recall 26 correct information being reported and a greater number of confabulations being produced but did not lead to more incorrect information being generated. Further to this, it was found that a mock crime video shot from a first-person perspective led to a predictive relationship between MTT and the amount of correct information generated under immediate recall conditions using the MRC component of the Cognitive Interview. Although no significant predictive relationship was found between MTT and eyewitness performance in the delayed recall condition, this may be a result of the methodological limitations identified previously. That the success of the MRC component of the Cognitive Interview can be influenced by individual differences is likely to be of interest to law enforcement personnel and should inform their approach to interviewing eyewitnesses as some people will therefore respond more favourably to the component than others.