Police Victim and Witness Interviewing in a Northern Canadian Territory: Measuring Perceptions and Practice

Police victim and witness interviewing in a northern Canadian territory with a predominantly Indigenous population was examined across two studies. In study 1, an Internet survey about interview training, practices and cross-cultural issues was completed by serving police officers (N = 37). In study 2, transcripts of interviews with Indigenous adult victims and witnesses (N = 20) were coded for the presence of various interviewing practices (e.g. question types, interruptions, talking time). Survey results showed that most officers were untrained in scientific-interviewing protocols but were aware of the general practices that constitute a competent interview (e.g. building rapport, requesting a free narrative). Most respondents indicated that cultural differences impact their interviewing style. Results of the transcript analysis showed that officers violated the 80/20 talking rule in 90% of the interviews and unproductive question types (e.g. closed yes/no) were used often. All interviews contained a request for a full account, most interviews contained elements of active listening and few interruptions were observed. These findings are discussed with reference to how interviewing and cross-cultural communication training could help police organizations who serve Indigenous populations. Future research should consider whether established international best practices for interviewing are effective in settings with Indigenous victims and witnesses.


Introduction
Research has shown that police officers tend to ask a flurry of questions and steer the discussion when gathering information from victims or witnesses (Eades 2010;Heydon 2005;Thornborrow 2002). The natural inclination of police interviewers is to seek an efficient outcome, and they may approach an interview feeling confident that they already know what happened (Griffiths et al. 2011;Shepherd 1988). As a result, interviews conducted by police officers with limited training are perfunctory exercises, demonstrating low skill and unreasonable levels of conversational control (Clifford and George 1996;Fisher et al. 1987;Griffiths et al. 2011;MacDonald et al. 2017;Wright and Alison 2004). To counteract these tendencies, an evidence-based best practice model, the cognitive interview [CI] , was developed and shown to be highly effective (Kohnken et al. 1999). The CI forms a substantial part of the UK's national investigative interviewing model (i.e. PEACE; an acronym that stands for Planning, Engage and explain, Account, Closure and Evaluation, for a detailed explanation, see Clarke and Milne 2001) and the UK Achieving Best Evidence [ABE] guidance (Ministry of Justice [MOJ] 2011; see also Davies et al. 2016 for the interviewing of vulnerable groups). The CI is also taught and practiced internationally, including some parts of Canada (Seiden 2016), and is considered the 'gold standard' of adult interviewing protocols. It remains to be seen; however, if anything, much has improved across Canada in the way of uptake of CI principles since Snook, Eastwood, Stinson, and Tedeschini (2010) described the state of police interviewing in Canada as inadequatelacking in effective training, consistent oversight and standard practices.
As mentioned, the CI is widely considered best practice for interviewing adults. Fisher and Geiselman (Geiselman et al. 1984) developed the CI as a witness-led model for use with cooperative interviewees and is underpinned by the ethos of relinquishing control of information flow to the interviewee . The CI originally consisted of four memory aids (i.e. report everything, mental reinstatement of context, recall in different orders and report from a different perspective; see Fisher and Geiselman 1992, for a full description). The use of multiple retrieval cues was shown to help interviewees access more information about a witnessed event (Fisher 2010). Specifically, if one mnemonic is unable to elicit enough detail, interviewers can adapt to the needs of the witness and use other mnemonics (Fisher et al. 2011). The first two mnemonics in the original CI are based on the encoding specificity principle (Fisher and Geiselman 1992;Tulving and Thomson 1973). The report everything instruction encourages interviewees to report everything that they experienced, without interruption or time constraints and no pressure to have an organized account, complete thoughts, or to only provide relevant information. The mental reinstatement of context instruction asks the interviewee to recreate the personal and environmental features of the experience, which assists the interviewee to spontaneously remember details and access more context, and in turn, recall more details (Dando and Milne 2009). The other two mnemonics use different pathways to access encoded memories, by asking the interviewee to recall the memory in reverse chronology or narrate the incident from the perspective of another participant . In the CI protocol, obtaining the free account is the crucial first step in the memory retrieval process as it provides the people, locations, actions and times that can then be probed further. After that initial narrative, the interviewer ensures that any questioning follows the order of the witness's mental record of the experience (Fisher 2010), and as such, the CI is fundamentally a witness-led process. As research progressed, Fisher and Geiselman enhanced the CI by incorporating theories of social dynamics (e.g. rapport building) and communication (e.g. transference of control; Fisher et al. 1990;Fisher and Geiselman 1992).
Two meta-analyses have demonstrated that the CI is more effective (i.e. eliciting complete and accurate information) than a standard interview (i.e. a control interview with an untrained interviewer). The effectiveness of the CI has been demonstrated using populations of various ages and abilities, including the older adult (Mello and Fisher 1996;Prescott et al. 2011), manual labourers in Brazil (Stein and Memon 2006), adults with intellectual disabilities ) and children with intellectual disabilities (Milne and Bull 1996). The first meta-analysis, as carried out by Kohnken et al. (1999), reviewed findings from 42 experiments from a 30-year period and found an overall large effect for increasing the recall of accurate information (d = 0.87). In 2010, Memon, Meissner and Fraser conducted another meta-analysis, including a further 20 experiments published since the 1999 analysis, and found an even larger effect size for the ability of the CI to elicit accurate information (d = 1.20).
The aforementioned meta-analyses excluded field studies because accuracy rates are difficult to ascertain in an analysis of real-world interviews (Memon et al. 2010). Field studies using actual police interviews are necessary to get the full picture of the use of the interviewing practices in actual investigations (Wright and Alison 2004). The limited amount of research examining real-life witness interviews has found that some elements of the CI are used less frequently, particularly the different order and different perspective mnemonics (Kebbell et al. 1999). Although not all police organizations train their members in the CI, it is possible to extrapolate general best practices from the CI protocol, even without using the more specific techniques such as asking the interviewee to recall the incident in a different order. Following these best practices, all interviews should feature an invitation to provide a free recall account, with ample time provided for the interviewee to fully explore their memory uninterruptedly, aided by helpful and productive questions from the interviewer (Griffiths and Milne 2005). Positive features of a good interview should also be present, such as building rapport, which is not only chat to put an interviewee at ease, but a process. As described by Tickle-Degnen and Rosenthal (1990), the process of developing rapport continues throughout the interview by fostering mutual attention, positive relations and coordination for a smooth and harmonious interaction (for more information on recent investigative interviewing rapport research, see Nunan et al. 2020). Other positive features of an interview include letting people complete answers to follow-up questions without interruptions, using question types that do not limit or taint information, namely productive question types that allow people to speak freely, and not talking too much, such as open-ended invitations.
In Canada, field studies on general police victim and witness interviewing have found low levels of skill in these fundamental elements of interview practice. For example, Wright and Alison (2004) found that Canadian interviewers (N = 19 interviews) asked few open-ended questions and controlled interviews by using main closed yes/no questions and repeatedly interrupting witnesses. Interviewers talked, on average, 33% of the time, which is higher than the recommended 20% (Milne 2016;Shepherd 2007). Snook and Keating (2011) replicated Wright and Alison's study with a larger number of Canadian interviews (N = 90) and identified three main areas of interviewer deficiency: talking too much, interrupting the interviewee and using too many unproductive question types. Interviewers over-talked in 89% of the interviews for an average of 36% of the time, and approximately 35% of questions asked were closed yes/no. In another study, MacDonald, Snook and Milne (2017) looked at 80 Canadian interviews with PEACE-trained and untrained interviewers and found that 70% of all interviewers asked for a full account, and that trained interviewers used more desirable practices. Specifically, they found that there was a large improvement in the number of interviewing behaviours comprising the engage and explain phase (d = 1.65), a moderate improvement in the account phase (d = 0.54) and a large improvement in the closure phase (d = 0.90). Trained interviewers also asked more open-ended questions and fewer leading questions. In general, the aforementioned results suggest that Canadian police officers rarely follow best practices, but that training does have a positive impact on interviewer behaviours.
More importantly for this research, there is a discernible gap in the police interviewing literature. Some studies have looked at cultural considerations for suspect interviews (Beune et al. 2011), but there are few studies of cross-cultural interviewing of witnesses and victims by police, even though it is well known that culture has an effect in both communication and memory (Anakwah et al. 2020;Hope and Gabbert 2019). In addition, although Indigenous people are over-represented as victims of crime (Scrim 2016), little research has been conducted in Canada on the needs and experiences of Indigenous victims and witnesses in court or police settings, or how cultural norms and perceptions affect the police interview process. Research on such cross-cultural issues in Australia, particularly by Eades (2004Eades ( , 2010, suggests that cultural differences and power differentials put Indigenous interviewees in danger of having their stories misunderstood and that allowances need to be made for differing worldviews (Powell 2000a(Powell , 2000bRoberts 2007).
Although Indigenous groups in Canada are not homogenous, there are common elements of Indigenous cultures that are very different from non-Indigenous Canadian society. In particular, traditional Indigenous concepts of justice are holistic, concerning families, lands, harm prevention and healing; relational and collective rather than individualistic, which contrasts with the acceptance of the non-Indigenous population that the practice of justice is driven and accomplished by the Canadian government (Arnakak 2000;Corntassel et al. 2009;Nunavut Tunngavik Incorporated [NTI] 2014; Royal Commission on Aboriginal Peoples [RCAP] 1996;Ross 2008). Justice Murray Sinclair (1994) observed that efforts in Canada have concentrated on instructing Indigenous participants to conform to the justice system, rather than the reverse. The Royal Commission on Aboriginal Peoples (1996) recommended separate court systems to meet Indigenous needs, but despite scattered pilot projects, this recommendation remains largely unfulfilled (Monchalin 2010).
The broadly collective nature of Indigenous cultures has been acknowledged not only in Canada, but around the world (Holder and Corntassel 2002). The consequences of a collectivist culture to interpersonal interactions, which include police interviews, encompass both communication and memory/attention differences. For example, people from a more collectivist culture, where the social aspect of life is more important than individual actions (Lord and Cowan 2011), will tend to communicate in more indirect or non-verbal ways (Hofstede 2001). With relevance to interviewing, culture will determine how a narrative is structured (Fivush et al. 1995) and which interpersonal cues a person notices and is impacted by, for example: mirroring body language, eye contact and how close the interviewer sits (Abbe and Brandon 2013;Sanchez-Burks et al. 2009). Individuals from collectivist cultures may also pay attention to, and subsequently remember, different aspects of an event than a person from a more individualistic culture and give narrative accounts with less personal details and more general descriptions of routines and social activities (for a more detailed explanation of the collectivist/individualist continuum, see Conway and Jobson 2012). Recent interviewing research (Anakwah et al. 2020) comparing interviews with African and European eyewitnesses suggests that it may be necessary, to achieve the level of detail required for a police interview, for the interviewer to request additional information from witnesses from collective cultures. As such, there is value in examining how cross-cultural police victim and witness interviews could be carried out with sensitivity and efficacy in Canada and around the world.

The Current Research
The aim of this research was to examine both the selfreported perceptions and the real-world practice of interviewing by Canadian police officers in a cross-cultural environment. A survey was administered to police officers, serving in small Northern Canadian settlements with a predominantly Indigenous population, to assess their level of training and their understanding of, and experiences with interviewing victims and witnesses. The survey was intended to determine whether the officers were aware of, and felt that they used, the best practice techniques available to them in their interviewing. We also analyzed victim and witness interviews from the same police organization, looking for the use of these techniques, to ascertain gaps in perception and the reality of current practice, while exploring the challenges of cross-cultural interviews. When considered together, the two studies take a preliminary look at the usefulness of the underlying principles of the CI for the cross-cultural interviewing of Indigenous victims and witnesses.

Method
An online survey was designed and administered to assess victim and witness interview training, practices and perceptions of police officers.
Sample. With permission from the participating police organization, an invitation email was sent to 55 police officers in the target communities. Informed consent was obtained from all individual participants in the survey. Thirty-seven officers responded, giving a response rate of 69%. Two surveys were discarded as only the first two questions were answered. Officers (n = 35) estimated that they interviewed, on average, 9.14 victims and witnesses per month (SD = 7.18). Of the 34 officers who responded to the demographic questions, 91% were men. Nearly half of officers (n = 16, 46%) were between 25 and 34 years of age, 49% (n = 17) had between 4 and 8 years of service and 63% were constables (n = 22; the most junior rank in this police service). Seventy-four percent (n = 26) of officers indicated that they had received training in investigative interviewing while serving as police officers, and 34% (n = 12) reported receiving CI training.
Materials and Procedure. The first section of the survey pertained to interview training, asking whether respondents had received training in interviewing victims and witnesses, other than the basic training provided during the standard 6-month recruit pre-engagement training. Next, participants were asked a set of 32 questions similar to those used by Kebbell, Milne and Wagstaff (1999) and adapted by Dando, Wilcock and Milne (2008), with 5-point scales asking officers to report how often they used a range of 13 interview techniques (Never, Rarely, Sometimes, Very Often, Always) and how useful they found them (Very Ineffective, Ineffective, Moderately Effective, Effective, Very Effective), and indicate how they would use the following techniques: 1. Building rapport-make conversation to put the interviewee at ease, achieve harmony throughout interview, 2. Explaining the process-inform the witness of the procedure for the interview (Recording? Notes? Purpose of the interview?), 3. Asking for a full account-encourage a free narrative with an initial open-ended question (e.g. tell me what happened), 4. Telling the victim/witness to take as much time as needed (take your time, no rush, if you need to stop and think…), 5. "Tell everything"-ask witness to give as much detail as possible about the event, even if it seems irrelevant or trivial or only partly remembered, 6. Concentrate/think hard-ask witness to concentrate and try to recall more details, 7. Mentally reinstate context-ask witness/victim to recreate the scene/context of the event in his or her mindboth the environment and the emotions involved, 8. Asking questions in the same order-as information remembered by the witness/victim (no topic-hopping), 9. Re-tell in different order-ask witness to recount incident in a different order (e.g. reverse) 10. Re-tell from another perspective-ask witness/victim to re-tell the story from the perspective of someone else (a bystander, another participant in the event), 11. Draw a sketch-ask the witness/victim to draw a picture to illustrate any part of the interview (for example: a room/house plan, a diagram or a map). 12. Demonstrating active listening-use approving noises/ small words (mmhm, ok, yeah, oh) to reassure the witness without stopping the flow of talk, 13. Uninterrupted-allow interviewee to speak without interruption until they are finished their full account of the incident.
Short answer questions about officers' perceptions and experiences conducting interviews were also asked, including whether cultural differences had an effect on the course of interviews, and if so, how. Participants were also asked to provide demographic information.

Results
The perceived frequency of use and effectiveness of interviewing components is shown in Table 1. Asking for a full account and not interrupting were reported to be the top two most frequently used techniques, followed by instructing the interviewee to take their time, tell everything and rapport building, respectively. The components that were perceived to be used less often, in order of frequency, were explaining the process, witness compatible questioning and asking the witness to think hard. Few officers reported using CI mnemonics (e.g. asking for an account of what occurred in a different order or from a different perspective), using witness-compatible questioning or asking victims or witnesses to sketch their account. In terms of perceived effectiveness, respondents, on average, indicated that building rapport was the most effective interviewing component, followed by sketching, asking for a full account, telling people to take their time, asking questions in the same order, explaining the process and using the tell everything mnemonic, respectively.
When aligning perception of use and effectiveness, it was found that building rapport was seen as both being effective and used often. In response to a question about how they would build rapport, officers described being calm, polite, compassionate, encouraging and friendly. Fourteen officers (40%) indicated that they initiated icebreaking conversation about common interests. They also discussed ensuring the interviewee was comfortable, for example, by offering refreshment or tissues. Three-quarters of officers indicated that they always get a full account but just over one-eighth of participants indicated that it was very effective. The difficulty in obtaining full accounts was attributed to problems with language barriers and communication, and this was noted by 50% of officers (n = 15). Other explanations included memory issues due to intoxication at the time of the incident (n = 11), cultural differences such as reluctance to incriminate others who are friends and family in a small Table 1 Perceived frequency of use and effectiveness of interview components (N = 35) compared with interview analysis scores (N = 20) Notes: Not all respondents answered these questions: *(n = 26), **(n = 18), ***(n = 12) a No usage or effectiveness questions were asked for active listening b No usage question was asked for uninterrupted, which was also not scored in the transcript analysis community (n = 9) or a lack of trust in the police (n = 3).
In addition, nearly half of the officers indicated that they always get the victim and witness to report everything but only approximately 10% of the officers indicated that it was a very effective technique. Also, of note was the finding that approximately two-thirds of officers indicated that sketching was an effective tool but that it was used infrequently. In general, these findings suggest that there is some disparity between what officers viewed as being effective and their actual use in practice. One potential explanation for this disparity may be contextual; that is, some practices such as rapport building can be used in nearly all situations but other situations may not call for the use of strategies (e.g. recall in different order) even if they are thought to be effective. It may also be the case that officers are not trained on various techniques and are not comfortable using them. Further research is required to determine why such disparity may exist.
Officers' answers to the question about the challenges of their current position in Northern Canada and whether cultural differences impacted the course of their interviews are shown in Table 2. As can be seen, officers did not report much confidence that they were obtaining full accounts, with most feeling that they only 'sometimes' (50%) or 'rarely' (21%) got the whole story from interviewees. Cultural difficulties were perceived to impact the course of interviews 'sometimes' by 47% of respondents and 'very often' by 35%. Despite these misgivings, officers indicated that their training had been mostly adequate, with 60% agreeing, and 14% strongly agreeing, that it had prepared them for the challenges of their posts. In addition, many officers ('very often'-35%; 'always'-26%) indicated that they had enough time to conduct interviews.
On the topic of cultural differences, communication (65%, n = 19) and language barriers (48%, n = 14) were again cited. Non-verbal communication, which is important in Indigenous cultures, was cited as easily misconstrued (n = 6). Mistrust of the police was mentioned by 31% (n = 9) of the officers and attributed to past negative experiences with the police and racism, with the result characterized by one officer as "a reluctance to go to the police and not to speak as freely during an interview". Another officer commented, "Sometimes the person does not believe you understand what they are going through, and have a hard time opening up to you".
In response to a question about modifications officers may have made to their interviewing practices to accommodate victims and witnesses in the North, 39% (n = 11) talked about changes to their behaviour outside the interview room, in an effort to understand and engage with the local language and culture. One officer summed it up: "[Changes to] interviewing practices -none. I make myself visible in the community, talk to people and be friendly and open". Another tied this to interview practices, saying that being involved in the community allowed for reassurances of understanding in the interview room, such as "That I am living [here] and learning the culture everyday". Open-mindedness was discussed: "Learn [the] culture, traditions humor etc. try to understand their situation, don't judge". A few officers (n = 6) mentioned that they asked more questions. One officer noted, "[I] learned that I might have to dig a little more to get the information". Others mentioned asking more closed or leading questions, or as one put it, being "more specific with my questions than I normally would have to be".
One officer in the survey suggested intensifying the patience shown already and recognising when interviewees are thinking about their answers: "You have to be persistent and use the long pauses… if you keep asking open questions… you get a much clearer understanding of what happened." Another said, "I have learned to listen and be patient. I will mirror the witness/victim, which will allow him/her to feel more comfortable". Many of the answers given to these questions were thoughtful and demonstrated that although the officers encountered difficulties, they were optimistic that they could persevere. Table 2 Officers' perceptions of interviewing Note: The categories for superscript a are 1 = always, 2 = very often, 3 = sometimes, 4 = rarely, 5 = never. The categories for superscript b are 1 = strongly agree, 2 = agree, 3 = undecided, 4 = disagree, 5 = strongly disagree

Method
For study 2, transcripts and audio/video recordings of police interviews with Indigenous adult victims or witnesses in the period from June 2011 to September 2013 were requested. Sample. A convenience sample of 20 interviews was received (18 transcripts with accompanying audio recordings and 2 transcripts without audio) from 19 interviewers in 10 communities. All interviewers were male and non-Indigenous, and all interviewees were Indigenous. The average interview length was 16 min and 30 s (SD = 11.08), with a range from 5 to 47 min. The interviews were conducted with 12 victims and 8 witnesses; 15 of which were women. Six of the interviewees indicated that they were intoxicated by alcohol at the time of the incident being discussed. The cases included 12 common assaults, 4 sexual assaults, and one each of assault with a weapon, dangerous driving causing injury and uttering threats against a person. The interviewer and the interviewee were the only people present in all of the interviews.
Materials and Procedure. In the interview analysis, 16 interview components were coded using a coding system similar to Griffiths and Milne's (2005) Witness Interview Rating Scale. 1 This rating scale examines 107 behaviours and gives interviewers a score between 0 and 5, where 5 is the best display of a particular behaviour and 1 the weakest, with 0 scored if the behaviour is not observed at all (0 = not present-no evidence of the technique/skill; 1 = attempted ineffectively-interviewer tried without success, unskilled; 2 = attempted to some small effect-interviewer tried but abandoned before completion; 3 = carried out moderately effectively (fit for purpose)-adequate display of skills; 4 = carried out effectively-above average display of skills; 5 = carried out highly effectively-professional display of skills, clear and confident throughout).
For a point of reference, interviewers scoring 3 on an item would mean that their practices were 'fit for purpose'; that is, that they demonstrated an acceptable level of skill to accomplish the task at hand-conducting a competent interview with a victim or witness to obtain evidence for investigation and court. Interviewers scoring higher than 3 went above and beyond by demonstrating advanced skills. For example, an interview where the interviewer started asking questions in the same order as the interviewee remembered the events, but then skipped haphazardly to another point in the narrative would be scored as 2 (attempted to some small effect) for the item asking questions in the same order. An interview where the interviewer emphasized throughout the interview that the victim could take their time, that there was no rush, and they would get through it together, would be scored as 5 (carried out highly effectively) for the item telling the victim/witness to take as much time as needed. An interview that started with an open question: "Tell me what happened?", but then before the witness answered, the interviewer interjected with, "Do you remember what time it was?" would be scored as 1 (attempted ineffectively) for the item asking for a full account.
Coding of talking time, interruptions and question types was conducted by listening to the audio files (where available) and marking codes on the paper copy of the transcript. Talking time was calculated as a percentage of the total number of words spoken by each party in the interview. Interruptions by the interviewer were coded and tallied if the interviewer talked over the interviewee. The various question types coded along with definitions are shown in Table 3 (adapted from Griffiths and Milne 2005;Schreiber Compo et al. 2012;Snook and Keating 2011;Wright and Alison 2004).
Inter-Rater Reliability. Two randomly selected interviews (10% of the sample) were coded by a second rater, to assess inter-rater reliability. A police officer with interview experience was given instruction on coding question types and assessing interview quality. The officer was then provided with best practice examples and was assisted in coding two interviews for question types and interview components, before coding two further interviews independently. Cohen's kappa for what constituted a question was 0.91, p < 0.001 [95% CI = 0.81 to 1.00] and was 0.90, p < 0.001 [95% CI = 0.82 to 0.99] for question types; these values indicate high data reliability (Landis and Koch 1977). Kappa with quadratic weights for the interview assessment was 0.83, [95% CI = 0.74 to 0.91], p < 0.001, which suggests high data reliability (Fleiss et al. 2003).

Results
In terms of social dynamics from the interview analysis, the interviewers were polite but business-like, with the average score for rapport-building being 0.6 (SD = 1.09) out of 5. No small talk was observed, although four interviewers displayed compassion, by encouraging the interviewee during the conversation and acknowledging the difficulty of discussing the disturbing events.
The average percentage of talking time for interviewers was 44.6% (SD = 14.78), ranging from 15.9 to 76.0%. All but one interviewer (95%) broke the 80/20 talking rule, with 8 interviewers (40%) speaking more than half the time. On average, a total of 1934 words were spoken by both parties during an interview. Interviewers were observed to practice active listening in the interview 1 Coding systems can be obtained from corresponding author on request. sample, with the average score being 2.90 (SD = 1.37), the best score for any interview skill and the only component that was close to being fit for purpose. This finding suggests that the interviewers over-talked, thus, not providing as much space to the interviewee to provide information as is recommended in the scientific literature. On a positive note, there was a total of 14 interruptions across all the interviews, less than one per interview, with at an average of 0.7 (SD = 0.95) interruptions.
There was also little usage of most interview components, as was shown with average scores in Table 1, thus, raising questions about how well the officers were able to gather complete accounts from the victims and witnesses. The exception to this was that all interviewers asked for a full account with an open-ended question. All interviewers (n = 20) asked an initial open-ended question to prompt a free recall and allowed interviewees time to answer, with an uninterrupted average response of 249 words (ranging from 3 to 1100 words). Many, however, qualified what they wanted to hear, for example: "Explain to me what happened yesterday from noon onwards," rather than allowing the interviewee to decide where to begin recounting the incident. Only one interviewer attempted the 'think hard' and 'mentally reinstate context' components, by encouraging a victim to concentrate, close her eyes and think about being back at the scene. The witness responded that she did not remember, and the context reinstatement attempt was abandoned at that point. Thirteen (65%) interviewers requested more detail at some point during the interview with at least one further open-ended question ("What happened then/ next?") and interviewees gave shorter responses, averaging 36 words (ranging from 0 to 289 words). The interviewers then proceeded to ask mostly probing and clarifying questions, averaging 16.65 probing and 15.55 clarifying questions per interview (see Table 4 for a breakdown of questions asked during interviews). The average number of questions in each interview was 56 (SD = 29.94). In total, 1122 questions were asked across all interviews, of which 44% were classified as appropriate. An open-ended question was asked, on average, every 6 min and 21 s; a probing question was asked every minute, and a clarifying question every minute and 10 s.
The interview analysis showed that interviewers encountered problems with understanding the accounts of the interviewees, including differences in how the interviewees used language (n = 11). In particular, definitions of action were Table 3 Question types and associated definition a Clarification was included in inappropriate questions in this study due to the tendency of interviewees in this sample to respond to the question with a flat yes or no response, but it should be noted that this type of 'echo' questions is a difficult categorization and may actually be appropriate when used by a skilled interviewer (Oxburgh et al. 2010) Question type Definition Open-ended Invite a detailed answer, using words such as tell, describe, explain Probing Invite a more specific answer using words such as who, when, where, what, how Appropriate closed yes/no Answers are limited to yes/no. These closed yes/no are used after probing to cover investigative points of interest Inappropriate closed yes/no Answers limited to yes/no. These are asked before probing questions are finished and are not necessarily connected to any focused series of questions pertaining to a discrete topic Leading Answer is suggested in the question. For example, "He drank too much, didn't he?" Multiple Asking several questions at once without giving interviewee a chance to answer each one. For example, "Where were you? Who was with you? How did you get there?" Forced choice Specifies at least two response options and requires one of the options to be selected. For example, "Did he hit you with his left hand or his right hand? Opinion/statement The interviewer utters an opinion or a statement without any attempt to gather information. For example, "I think you remember, you just do not want to say" Clarification a Repeat something the interviewee has already stated, as a question. For example, in response to a witness stating, "I crossed the road" the interviewer says " Ok, so you crossed to the other side of the road?" Repeated Questions already asked and answered. Note that this is not coded when the interviewee did not hear or understand the question the first time it was asked not commonly understood between the parties: "pushed" and "grabbed" or "threw"; "fooling around"; "trying to"; "almost hit me"; for example, requiring multiple questions from the interviewer to clarify. Interviewers were noted to struggle in other ways with a lack of shared context (n = 8), as interviewees referred to individuals as 'he' or 'she' without specifying their names, or referred to them by previously unexplained nicknames, even when there were multiple actors in play. Also, narrative constructions were confusing to the interviewers, as interviewees described multiple events without separating them (n = 5). In addition, interviewers were observed pressing interviewees to be more specific about quantities such as distance between locations, frequency of events and numbers of objects and actions (n = 13), to confirm times or dates (n = 9) or to narrate action in chronological order (n = 5).

Discussion
Looking at social dynamics, interviewers in all three Canadian studies routinely violated the 80/20 talking rule. In the Snook and Keating (2011) study, interviewers spoke 36% of the time, in the Wright and Alison (2004) study, 33% of the time, compared with 44% in this study. Notably, the interviews in this study were characterized by a lack of interruptions by the interviewer and the demonstration of active listening. Police officers may have adapted to the slower pace of the North, but it is also possible that if interviewees talk less, there are fewer chances to interrupt. The officers in this study still talked too much and may have perceived uncomfortable silences and filled them with words. Rapport-building between the interviewer and interviewee, although rated highly for effectiveness in the survey, was not evident in the interviews. Dando, Wilcock and Milne (2008) found officers reported high usage of rapport, but lower perceived effectiveness, although with a slightly different Likert scale, in a survey of inexperienced police officers. These differences suggest that in this sample, like officers internationally, officers acknowledge the expectation that they will build rapport, but do not routinely incorporate rapport-building into interview practice. This omission is unfortunate, however, given that as little as 5 min of rapportbuilding can positively affect the recall of interviewees (Collins et al. 2002). It is possible that rapport-building took place before the interview and was not recorded, but ideally the rapport process should continue throughout the interview (Abbe and Brandon 2013). It is also possible that rapport is affected by culture (Hope and Gabbert 2019) and that officers feel a sense of stress in the cross-cultural interaction (Sanchez-Burks et al. 2009), that manifests as businesslike behaviour).
Our findings show that most interviewers were not conducting textbook 'fit for purpose' interviews, as defined in the CI literature. In examining the correspondence between what police officers in Northern Canada thought about their interviewing skills and the reality of practice, we found discrepancies, and little-to-no use of CI components, particularly the more complex ones. This is not a surprising result, however, given that many officers in the survey reported not receiving training in cognitive interviewing, or the fact that people often struggle to use CI tactics for a number of reasons (Kebbell et al. 1999). Whether or not the CI is an appropriate model for interviewing Indigenous witnesses and victims is a larger question, one which will require more investigation.
Overall, compared with the Canadian results published by Snook and Keating (2011) and Wright and Alison (2004), the interview component analysis showed some better and some worse practices, but mostly similar practices. In particular, asking for a full account was universally rated highly for usage and efficiency in our study, and this was borne out in the interview analysis. Snook and Keating (2011) found that 73% of officers requested a free narrative, whereas all the interviews in this study also featured this request. However, the fact that the officers reported not always feeling they had obtained a full account demonstrated that, for this population, asking a generic "tell me everything" may be insufficient.
We found that officers in the interview sample followed accepted protocols, at least initially, by asking an openended question, despite the fact that survey respondents reported feeling unable to consistently obtain full accounts. Question types are difficult to compare with other studies, as they are scored slightly differently in each (see Oxburgh et al. 2010 for a full explanation). The percentage of open questions asked were low, with 4.6% in this study, very close to the 5.8% found in the Snook and Keating (2011) sample. Wright and Alison (2004) found that an open-ended question was asked every 6 min and 15 s, which was also similar to the current finding at 6 min and 31 s. As has been noted, all open-ended questions are not the same (Powell and Snow 2007), and if certain types of open-ended questions do not elicit the desired answers, then the types of questions asked may need to be changed, to encourage interviewees to give more detail, as suggested by Anakwah and colleagues (2020). After the initial invitation to speak freely, continuing to ask follow-up invitations could help, as officers in the survey reported-for example: "And then what happened?" rather than falling back on direct or repetitive questioning. The observed practice of alternating clarifying and probing/closed questions is a problematic pattern and could be an interviewer habit more suited to everyday conversation, a way to play for time while formulating another question. This inappropriate questioning would certainly be an area for improvement in this force's practice. Perhaps most importantly, the finding that language difficulties, due perhaps to a lack of shared context, interfered with the course of interviews concurs with the perception of officers that they struggled with the cultural aspects of interviewing in the north.

Limitations
Limitations of this research include the size and convenience nature of the samples for both the survey and the transcripts. The small size of the survey sample precludes any meaningful quantitative analysis other than descriptive statistics, and both samples were potentially biased, as they may include only diligent or cooperative officers and interviews from files that were actively investigated. Further, the respondents to the survey were mostly male, young and junior in service and did not report universal training in CI techniques. It would be interesting to know whether or not this sample is representative of police officers who serve in the north. Another limitation is that details other than gender were not supplied for the officers conducting the interviews in the transcript analysis, so other factors that could potentially have an effect on interview performance, such as interviewer experience, age, time in the north and training, could not be analysed. Although the rough triangulation of the survey and transcript study raises interesting points, it is not possible, overall, to generalize beyond this sample. In addition, as far as generalizability, as with other research with Indigenous people in Canada (Dylan et al. 2008), there was no way to capture the self-identification of interviewees as belonging to a particular Indigenous group, as residents of the north are mobile and diverse. However, despite these limitations, the findings echo those of other studies in Canada (Snook and Keating 2011;MacDonald et al. 2017;Wright and Alison 2004) and the real-world nature and the qualitative richness of the data obtained is valuable and acts as a starting point for future research into these topics.

Directions for Future Research
The finding that officers did not feel they obtained full accounts and struggled with communication indicates that further useful research would involve consulting victims and witnesses from Indigenous backgrounds to find out their experiences of police interviews. Indigenous interviewees could explain whether they felt respected, listened to, heard and understood and whether they told their whole stories, and if not, what the police could do differently. Also, research is needed to determine whether the CI is the best model for police interviews with Indigenous victims and witnesses. Questions raised by this research that could be explored include whether the perceptions of some of the officers that they had to ask more direct questions to obtain details was correct or whether a continued use of open-ended questions would achieve the same results, as compared with interviews with non-Indigenous victims and witnesses in Canada, and whether the mnemonics of a CI interview could be useful.
Focused research of this type could assist with the development of culturally appropriate strategies for interviewing Indigenous victims and witnesses. Indigenous trust and confidence in the police is fragile or broken in Canadian communities and must be fostered and repaired through communication (Chrismas 2012;Cao 2014), which is vital for police legitimacy, particularly in the north (Griffiths and Clark 2017

Conclusion
Broadly, our findings show that police agencies serving large Indigenous populations in Canada should receive more specialized training in how to interview and that cross-cultural police interviewing should be researched, with guidelines developed to inform policy. Cross-cultural sensitivity training, covering the idea that questioning is not a natural event for Indigenous people and that silence can be productive, could be helpful. Although cross-cultural training is not the entire answer (Powell 2000b), it could be a good start, if supplemented with encouragement to join in the life of the community, as recommended by the officers in the survey. Interestingly, the officers who reported that they found it useful to involve themselves in community events match the perception of Indigenous groups in Canada that the police enjoy better community relationships if they take part in the cultural life of the community (Jones et al. 2015). Furthermore, training for interviewers should cover ways to assess each individual interviewee's cultural background, personal needs and comfort level with language and the process itself. There is value in learning to build a rapport with each interviewee separately, keeping cultural factors in mind but not assuming stereotypes apply, and reflexively checking one's own assumptions and biases regularly.
The results of this research highlight the need to encourage police officers from all organizations internationally to familiarize themselves with different cultures, through shared experience, and to look at their own interviewing practice through a realistic and critical lens. The survey conducted here and the interview analysis show that although knowledge and awareness of best practices are high, interviews still show low levels of skill, showing that officers' perceptions and the reality of practice do not match up. Officers in these Northern communities also struggle with the cross-cultural elements of practice, which interferes with their ability to obtain full accounts for evidential purposes.

Compliance with Ethical Standards
Conflict of Interest Kate Chenier declares that she has no conflict of interest but discloses that she was employed by the police agency studied, although not in a position where supervisory or investigative duties were required. Research was performed as a student and not as an employee.
Ethical Approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed Consent Informed consent was obtained from all individual participants included in the study.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.