An Introduction to Applied Cognitive Psychology - David Groome, Anthony Esgate, Michael W. Eysenck (2016)

Chapter 8. Witness interviews and crime investigation

Rebecca Milne and Ray Bull


Cognitive psychology has a long history of research impacting upon the real world. One key area that demonstrates this influence is criminal investigation. This chapter will endeavour to outline some ways in which research has transformed how the police go about their everyday jobs, from the attaining of initial accounts at the scene, to conducting investigative interviews, to trying to detect deceit. Identification procedures will only be touched upon, as this is covered in Chapter 3.

First, it is imperative to define what a criminal investigation is. What first comes to mind is the stereotype fed by the media, of glamorous detectives using their intuition, gut feeling and experience, tracking down the baddies, all within a few hours, and everything is solved – success. If only it were that simple. In reality a criminal investigation is about hours of methodological and systematic mundane hard work, with investigators needing keen attention to detail and good communication skills (Innes, 2003; O’Neill and Milne, 2014). After locking down a scene and handing it over to the scene of crime officers, to collect, analyse and examine forensic evidence, investigators have to initially gather information from a number of sources: from the initial call to the emergency services, from witnesses/bystanders at the scene, from victims of crime, from potential suspects, from house-to house inquiries, and experts, and so on. However, what can be seen is that central to a large proportion of these tasks is the fundamental skill of gathering information from human memory. This information is then fed into the investigative process, where open-minded investigators have to make decisions based on the accounts gained from people’s memory.

As we have said before (Milne and Bull, 2006), the key aim of an investigation is to establish the answer to two primary questions: (i) ‘What happened?’ (if anything did happen) and (ii) ‘Who did it?’ Investigators themselves (Kebbell and Milne, 1998) have noted that central to finding answers to these core investigative questions is the ability to gain accurate and fulsome information from key criminal justice players: witnesses (including professionals – e.g. police witnesses), victims and suspects. In the US, Sanders (1986) asked police officers: ‘What is the central and most important feature of criminal investigations?’ The majority replied ‘witnesses’. A similar view is held in the UK where Kebbell and Milne (1998) asked 159 serving police officers for their perceptions of the utility of witnesses within investigations. It was found that witnesses/victims were perceived usually to provide the central leads in cases. Investigators also frequently have little or no other forensically relevant information to guide an investigation (especially in the investigation of sexual offences, where it tends to be ‘word against word’). Therefore the primary source of information and evidence for the investigator tends to be that which is gathered from the witness/victim. Thus, information gained from witnesses/victims often forms the cornerstone of an investigation (Milne and Bull, 1999, 2006; Ridley et al., 2013). However, it has also been noted by officers that witnesses rarely provided as many person descriptions as the investigator would like (Brown et al., 2008).


Figure 8.1

Source: copyright LukaTDB/

Thus, the enormity of this task cannot be underestimated. Cognitive psychology has repeatedly demonstrated since the 1970s (Clifford and Bull, 1978; Loftus, 1979) that witness testimony is fragile, malleable, constructed and incomplete. Therefore, the procedures used to elicit information from this delicate repository have to take into account all imperfections of human memory to ensure that the information gained is of good quality and quantity. Thus, the information that investigators henceforth utilise can be relied upon to make informed investigative decisions. This in turn will hopefully prevent miscarriages of justice occurring, as it has been widely documented (e.g. Poyser and Milne, 2011; Milne et al., 2010) that one of the core reasons underpinning many past miscarriages of justice worldwide concerns the accuracy of testimony from witnesses. Thus, as psychologists, we need to ensure that there is the smallest error possible within the procedures that law enforcement personnel use to investigate crime.

When training investigators it is often useful to use the analogy of what happens at a real crime scene to explain the fragility of human memory to practitioners. When a crime has been uncovered and has started to be investigated, one of the first things that happens is that there should be a number of procedures and protocols set in motion for the preservation of the crime scene(s). Similarly, memory needs to be treated in the same way, as it also is a ‘crime scene’ needing to be protected. Unfortunately, unlike physical trace evidence, witness evidence does not yet have the same rigorous procedures (informed by memory experts) to provide the optimal ways for collecting, preserving, recording and interpreting such evidence (Wells and Loftus, 2003), as contaminated witness memory can also have devastating consequences on the investigation and subsequent court procedures, and could result (and has done) in many miscarriages of justice (Savage and Milne, 2007).

Given that information is the currency of the criminal justice process, it is common sense to assume that the more information (i.e. quantity factor) gleaned that is of good quality, the more likely a crime will be solved and successfully prosecuted (Milne and Bull, 2006; Wilcock et al., 2008). It then follows that the way in which the information is gathered within an investigation is key to successful crime solving. The methods used to find answers to the crucial questions (What happened? Who did it?) will in essence determine the quantity and quality of the information gained, and the success of the case as a whole. This chapter will now start traversing through what we (Milne and Bull, 1999) have termed the obstacle course that the information about the incident has to endure, which involves imperfect witness memory processes (Kebbell and Wagstaff, 1999) and the difficulties associated with interviewing and identification procedures. The criminal justice system as a whole needs to be aware of the imperfections in the witness testimony process (Wilcock et al., 2008). Indeed, it is therefore imperative for investigators in the first instance (but anyone involved in the criminal justice system, from call handlers to judges) to learn how easily they can influence what interviewees/witnesses report (Milne and Bull, 2006; Davies et al., 2016). The first sub-section of this chapter examines the initial stages of the investigation: the call to the emergency services and attaining the first account at the scene.


The majority of victims in the UK (and several other countries) report their crime by telephone to a centralised police call-handling centre (Ambler and Milne, 2006, in prep.). The role of a call handler is primarily (i) to promptly determine the nature of the call, (ii) to assign a level of urgency to the call and (iii) to determine the most appropriate response to it. To do this, the call handler questions/interviews the caller. The effectiveness of this process in terms of the quality of the information obtained and the way in which the information is elicited is paramount, as it is this initial interaction that determines the initial response (e.g. correctly dispatching an emergency response to save lives), and it is information elicited at this stage that commences many criminal investigations (Wilcock et al., 2008). When examining investigative processes, it can be seen that information gleaned from witnesses/victims in the first instance often governs the initial direction of the investigation, helping the investigator(s) outline avenues of exploration and lines of enquiry to be pursued, and helps to identify possible suspects (Milne and Bull, 2006). In addition, information gleaned at this stage is likely to be the most detailed and accurate because of the relatively short time between experiencing the incident and recalling it (Gabbert et al., 2016). However, it must be borne in mind at this stage that information has often to be gleaned quickly from highly traumatised individuals who may be in circumstances that are not conducive to memory retrieval (e.g. chaotic crime scenes).

Unfortunately, to date there is limited knowledge as to what occurs at this vital stage of the investigation process. Ambler and Milne (2006) conducted one of the first examinations of how information was garnered from memory at this phase of the investigation process and found that the manner in which the information was elicited by the call handlers was highly interviewer driven and included many leading questions. Similarly, Leeney and Muller-Johnson (2011) examined forty emergency calls to a police force in the UK and found that 11.5 per cent of all questions asked were categorised as unproductive and included suggestive questions that introduced information to the callers that they had not mentioned themselves. A wealth of research in cognitive psychology has shown the dramatic contamination effects that leading questions can have on memory (for a recent review, see Frenda et al., 2011). This undoubtedly causes concerns when viewing memory as an uncontaminated crime scene that should be protected. Such questioning can thus contaminate memory, which in turn may influence the decision making of the call handler, and subsequent investigator.

One solution to this real-world problem has been the development of a call-handling interview protocol that aims to gain good-quality information quickly, but without marring memory. Pescod et al. (2013) created a call handler free-recall instruction which included the ‘report everything’ instruction from the cognitive interview (CI; see later for a full description of this technique). This instruction simply allowed the caller to control the flow of recall and instructed them to tell everything without any editing. This experimental condition was compared against a control condition, which mirrored the types of question that call handlers typically used, i.e. the five ‘WH’ question types of (i) who, (ii) what, (iii) when, (iv) where and (v) why – specific closed questions. It was found that the ‘report everything’ instruction elicited significantly more information than the typical control procedure. Interestingly, this instruction also elicited significantly more correct person description details. Research has consistently demonstrated that witnesses have difficulty reporting details of offenders (Brown et al., 2008; Leeney and Muller-Johnson, 2011), and descriptions are often vague and as a result can apply to many people within the vicinity of the crime (e.g. Fahsing et al., 2004).

Archival studies have examined the content of person descriptions in real cases and found that, in general, witnesses usually give between seven and nine pieces of person information (e.g. Van Koppen and Lochun, 1997), and that these details tend to include more general characteristics, giving a more general impression of the person (e.g. gender, age, race, height and build). Such impressions are not extremely helpful for directly locating and identifying a specific individual, but are useful for narrowing down potential suspects within an enquiry (Wilcock et al., 2008). Thus any procedure that helps witnesses retrieve and report person description accurately is very welcome.

It has also been found that witnesses actually use stereotypes to interpret incidents and predict behaviour. This is especially the case when the event happens quickly and unexpectedly (like a crime) and when the witness’s attention is divided (Macrae and Bodenhausen, 2001), for example doing another task (e.g. driving). Davies and Patel (2005) found that within a driving scenario, the make of a car, the gender of the driver and car colour were all linked with perceived aggressiveness of the driver. For example, it was found that young males, driving red or black Ford Escorts or BMWs, were judged to be more aggressive. Thus, stereotypes can have an influence on the initial interpretation and subsequent memory (by witnesses) of an incident (a top-down process – Bartlett, 1932), and probably on the subsequent memory of what occurred (at interview; the reconstructive nature of memory – Tuckey and Brewer, 2003a), and the final understanding of the event in a courtroom (e.g. by a jury or judges in countries that don’t have juries). (For more on stereotypes, see the relevant chapter in Wilcock et al., 2008).


The next stage in the process usually involves frontline officers attending the scene. Their first task is to ensure that the incident itself has been resolved (i.e. the safety of individuals has been ensured and the emergency is over). Second, the officers embark on their investigative role, gathering information/evidence. Gathering information from witnesses and victims at the scene is a direct source of information to guide the ongoing investigation. However, there has been limited field research as to the nature of interviewing at this early stage of the investigative process. What is known is that the majority of such interactions tend to be conducted by officers who have very limited policing experience and training (Dando et al., 2008). Indeed, Dando et al. (2008) found that officers reported feeling inadequately trained, under pressure and generally ill-equipped to conduct effective interviews. In addition, follow-up work (Dando et al., 2009a, 2009b) found that newly trained officers in fact did have some way to go before fully mastering the skills necessary to interview appropriately. As a result of this research in England and Wales, a number of initiatives have been developed by psychologists to help these novice interviewers at this stage of the investigation.


Figure 8.2

Source: copyright Corepics VOF/

One such initiative is the Self-Administered Interview (SAI; Gabbert et al., 2009). This was specifically developed to overcome challenges at the scene of the crime, particularly when there are (i) multiple witnesses and when there is (ii) a lack of resources in terms of time, expertise and enough personnel to conduct such a volume of interviews (Gabbert et al., 2016). The SAI takes the form of a standardised interview protocol consisting of instructions and questions to enable a witness to provide their own account of events. Research has shown that the SAI elicits significantly more accurate and detailed information than a simple free-recall request. In addition, it helps to inoculate the original memory against forgetting and potentially distorting post-event information (Gabbert et al., 2009; Gabbert et al., 2012; Gabbert et al., 2016; Gawrylowicz et al., 2013; Hope et al., 2014).

An innovative stream of research that is currently under way and that focuses upon gathering information at a crime scene shows how researchers have capitalised on the recent use of technology in everyday policing and concerns the use of body-worn video cameras (BWV) in the UK. In the UK, budget constraints and recent government policies have accelerated this change and have resulted in new methods being adopted by police forces (Dalton and Ang, 2014). Nevertheless, this has allowed researchers for the first time to consistently view what actually happens at a crime scene. A number of researchers (e.g. Dalton, Gabbert, Hope, Milne, McGregor, Ellis and LaRooy) are currently developing new methods in order to train officers to interact appropriately at the scene to gather good quality and quantity of information.

Another initiative was developed by Dando et al. (2009b) that involves utilising a sketch to enable and promote the retrieval of information from memory. What must be borne in mind is that over 70 per cent of ‘volume crime’ is allocated to the uniformed ‘mainstream’ police officers who, on average, spend 20 per cent of their day-to-day duty time investigating it (ACPO, 2004; Hewitt, 2001). In addition, these officers have limited training, heavy workloads and time constraints, and are inexperienced. Thus there was a need to develop a protocol that adhered to best practice, was evidence based, drawing upon theoretical principles concerning the retrieval of information from memory, but also was simple to learn, practise and implement in the real world.

The use of a sketch draws upon ‘context effects’ and the ‘encoding specificity’ principle of memory (Tulving and Thomson, 1973). It has long been known that it is sometimes easier to recall information if you are in the same place or context as that in which the encoding of the information initially took place (Estes, 1972). This helps us to explain why we are sometimes overcome with a surge of memories about our past when we visit a place we once were, e.g. a hospital or school. The context in which the event was encoded is itself thought by some to be one of the most powerful retrieval aids, and there is a large body of experimental research examining this issue (see Pansky et al., 2005 for a review). ‘Crimewatch’ reconstruction is attempting to reinstate the context in which the event took place in order to jog people’s memories of the event.

We found that novice police interviewers often reported spontaneously asking witnesses to draw sketches as they recalled what had happened, in order to help memory (Dando et al., 2008). Two initial studies were conducted to examine the utility of sketching for memory retrieval in a frontline interviewing context. The first (Dando et al., 2009a) tested the mnemonic properties of sketching for enhancing recall about an eyewitness event. Utilising a typical eye witness paradigm, sixty student witnesses viewed a crime film, and were interviewed 24 hours later. The sketching group were asked to draw a detailed sketch or plan of the event they had seen; they were encouraged to furnish it with detail, and to recall as they drew. Overall memorial performance revealed the sketching technique to be more effective than a control group, who were simply asked to recall the event. The second study incorporated the sketching technique into a full frontline interviewing procedure (Dando et al., 2009b), and employing a mock eyewitness procedure it was found to facilitate memory compared with an appropriate control. More field research is now required to test the validity of this technique in practice.


Figure 8.3

Source: copyright Sergey Sarychev/

Regardless of the quality of the interviewing procedure, the resultant recall is dependent upon the conditions of the original encoding of the incident. There are many uncontrollable factors at the crime scene itself thought to affect subsequent recall, and these are not mutually exclusive. Instead, each factor will combine (or not) depending on each individual crime scenario. For example, the longer the actual event lasts, the greater the amount of likely recall, as more information can be encoded during that time (Fahsing et al., 2004; Yarmey et al., 2002). As the number of unknown perpetrators involved in the crime increases, the ability of the witness to describe each individual diminishes (Fahsing et al., 2004). This is due to the resources of attention and encoding being divided between the individual perpetrators and thus less information about each is encoded into memory. During a crime it would be impossible to attend to everything. Thus, witnesses are selective about what they attend to. As a consequence, much of what is in our environment never enters memory and so will not be available for later retrieval (Wilcock et al., 2008). Selective attention is driven by a ‘top-down’ process and depends on a person’s knowledge, expectancy, attitudes, past experience, interests and training, among other factors, and what that particular person judges as the most important information at that point in time (Milne and Bull, 1999). If five different people witness an event, five somewhat different versions of the event will result (Fruzzetti et al., 1992).

A witness’s level of stress at the crime scene is also thought to have a bearing on how much is encoded and then later retrieved (e.g. at interview). Research examining real-life witnesses seems to show that emotional events appear to be well retained and reported as long as the memories are elicited through appropriate interviewing procedures (Peace and Porter, 2004) and the witness is emotionally ready to talk about the distressing event. For example, one of the first studies examining this was conducted by Yuille and Cutshall (1986), who found that witnesses to a homicide who indicated high stress levels had a mean recall accuracy of 93 per cent when interviewed by the police 2 days after the event (i.e. most of what they reported was correct, though of course they only recalled a small proportion of the crime). Four to five months later, researchers interviewed these witnesses again (though only thirteen out of twenty-one agreed to participate at this time), and found that even at this delay interval the witness reports had an average accuracy rate of 88 per cent, though they did make some errors (almost two hundred between them). Yuille and Cutshall (1989) proposed that the reason why people have good memories for emotionally charged events is because they have had practice in accessing and rehearsing these memories through ‘remarking’ on the incident to themselves and others, and so such ‘remarking’ could be encouraged so long as it does not introduce errors.

Another classic factor associated with stress and remembering is what has been termed the ‘weapon-focus’ effect. That is, the presence of a weapon at the crime scene may reduce the amount of correct recall overall reported by a witness, especially recall about the perpetrator holding the weapon (e.g. Kramer et al., 1990; Loftus et al., 1987) and other ‘important’ aspects of the crime. This effect is thought to occur because the witness who is experiencing the stressful event may respond by narrowing the scope of his/her attention to those aspects producing the stressful effects (i.e. the weapon). This in turn may reduce the amount of information about other aspects of the scene which are encoded and available for subsequent reporting (cue utilisation theory; Easterbrook, 1959). Steblay (1992) conducted a meta-analysis of twelve studies examining weapon focus, which involved nineteen experiments. Six of these experiments found significant differences in the expected direction (i.e. less recall when a weapon is present). However, the remaining thirteen did not. This was in part due to the fact that recognition procedures (for example, an identification parade scenario) resulted in small weapon-focus effects, whereas much greater weapon-focus effects were found for recall tasks (for example, during an interview) (Milne and Bull, 1999; Wilcock et al., 2008).

Thus for practitioners a real conundrum concerns the following: when is it best to interview a traumatised individual? Investigatively it is often necessary to interview an individual immediately, for example in order to apprehend an offender running rampage with a firearm. The speed of gathering accurate information is often crucial. Thus the aim of an initial account is to gather as much information as is required to steer an ongoing investigation. At a later date the witness may be involved in a full investigative interview, which aims to gain a full and faithful account to help investigators make informed decisions within the ongoing enquiry and to serve as evidence in any subsequent court proceedings. When that interview should be conducted is crucial.


Figure 8.4

Source: copyright Fer Gregory/

Memory research has shown that the quantity of witness accounts, however, is affected by time (quality of the account tends to remain stable – see ‘remarkable memories’ earlier, if there is no contamination). The delay between encoding and retrieval is critical; as this period increases, the amount of recall systematically decreases (Ebbinghaus, 1913/1885; Tuckey and Brewer, 2003b) and the memory becomes more gist-like over time (e.g. Goldsmith et al., 2005). For example, archival studies found that witnesses give fewer descriptions after longer delays (Van Koppen and Lochun, 1997). In addition, as the delay increases, the window for memory contamination also increases, as witnesses have a greater opportunity to encounter post-event information (e.g. social media).

Thus, frontline interviewing needs to be quick and effective (Dando et al., 2009a, 2009b). Unfortunately, as a result officers tend to resort to quick-fire questioning, peppered with inappropriate question types (e.g. leading questions), and often interrupt the witness to obtain the information they require (e.g. Wright and Alison, 2004; Snook and Keating, 2010; MacDonald et al., in press). This inappropriate questioning could potentially contaminate the all-important fragile witness memory. Furthermore, in an attempt to resolve the situation and elicit information quickly, officers seem to be breaching a core principle of crime scene management, where most police protocols outline that witnesses to an incident should be separated from each other as soon as possible and interviewed individually (Dalton et al., in prep.). Warnick and Sanders (1980) examined the influence of group discussion of a previously viewed event on individual witness memory. Interestingly, it was found that individuals who had participated in the group discussion had superior memory with regard to both accuracy and completeness. Similarly, Yarmey and Morris (1998) had a group provide a description together, which was found to be more complete than one provided by an individual. This is what is termed collaborative or pooled recall. The downside to this, however, is what is termed the ‘conformity effect’. For example, Gabbert et al. (2003) created a situation in which witnesses viewed events differing in several key features. Witnesses were then asked to discuss the event with each other before providing independent descriptions. It was found that 71 per cent of participants incorporated erroneous details provided by a co-witness. As no two witnesses will ever have the exact same memory for an incident, pooling memory in this way will contaminate memory and will raise doubt as to the integrity of the account (see Gabbert and Hope, 2013). Thus, there is a fine line between interviewing someone as quickly as possible and taking into account the circumstances surrounding the event, both situational (e.g. location, multiple witnesses) and witness factors such as trauma and intoxication (see Gabbert et al., 2016 for a discussion of witness factors).


The next stage in the investigation process typically is the investigative interview, which comes in many forms depending on a number of factors such as the severity of the case, resources available and time available. In the ideal world, for maximum memory retrieval a witness should be interviewed in a neutral environment, free from distractions, with well-trained interviewers who can spend as much time with the interviewee as is needed. However, the real world is far from ideal. This chapter so far has outlined the real-world problems at the scene and attaining initial accounts and how psychologists have utilised and are attempting to utilise research to create evidence-based memory tools. The interview at this stage differs from an initial account as its aim is to garner information that answers not only the core investigative questions to help investigative decision makers (‘what happened?’ and ‘who did it?’), but also ‘who did what? And why?’ The ‘why’ helps legal decision makers in assessing grounds for culpability (Gabbert et al., 2016).


Figure 8.5

Source: copyright Kzenon /

Witnesses can be interviewed multiply across an investigation, by a call handler at the outset, at the scene, and then in an investigative interview. Indeed, Brown et al. (2008) asked officers how frequently witnesses were interviewed within an investigation for person information, and 82 per cent indicated more than once (range from two to six times). Repeated interviews can actually help to preserve memory and result in an increased number of details remembered about the incident and the persons involved (e.g. Hashtroudi et al., 1994; Hope et al., 2014). However, this initially depends on how the information was elicited across the successive recall attempts. If these attempts are conducted appropriately, using predominately free-recall tasks and open-ended questions, there is less opportunity for the interviewer to influence unduly what the interviewee says. However, if the interviews involve a predominantly interviewer-driven questioning style, with large numbers of specific closed questions and leading questions, then multiple interviews can render memory increasingly less accurate and more contaminated across each recall attempt/interview (e.g. Meissner, 2002). Indeed, a good-quality first interview can help to inoculate a witness’s memory against subsequent suggestion (e.g. Memon et al., 2010b; Gabbert et al., 2012). Thus, interviewing procedures need to ensure that memory is protected as far as is possible across each recall attempt.

The cognitive interview was a technique developed in the 1980s by two American psychologists, Ed Geiselman and Ron Fisher, in order to provide investigators with a tool to help them elicit a greater amount of accurate information from witnesses and victims of crime but without reducing the quality (i.e. accuracy) of information gained. The original CI comprised a set of four instructions given by the interviewer to the interviewee (Fisher et al., 1989), which were as follows:

(i)    the report everything instruction;

(ii)   the mental reinstatement of context;

(iii)  the recalling of events in a variety of different orders;

(iv)  the change perspective technique.

Each of the four techniques was developed from extant research and theory concerning the retrieval of information from memory. (For a fuller description of the CI and the enhanced cognitive interview (ECI), see Fisher and Geiselman, 1992; Milne and Bull, 1999.) The report everything instruction encourages interviewees to report everything they remember without any editing, even if the interviewees think the details are not investigatively important/trivial or cannot remember completely a particular aspect of the event (Milne, 2004). The mental reinstatement of context technique emanates from the research demonstrating that context can have a powerful effect on memory (see earlier discussion on sketching). The context reinstatement instruction asks interviewees to reconstruct in their minds the context, both physical (environmental) and personal (for example, how they felt at the time), of the witnessed event (Milne and Bull, 1999). Once interviewees have (using free report) recounted the event in their own order, the interviewer can encourage the interviewee to recall the event using a variety of different orders; for example, from the end to the beginning of the event (i.e. reverse-order recall) and/or working backward and forward in time from the most memorable aspect of the event. This technique attempts to counter the reconstructive nature of memory, where an event that is being remembered is influenced by a witness’s prior knowledge, expectations, and the employment of scripts (e.g. what typically happens in an armed robbery is often gained from media representations) (Holst and Pezdek, 1992). The final instruction, the change perspective instruction, asks the interviewee to recall the event from a different perspective. For example, asking the interviewee to focus on one individual in the event in their memory, put a ‘spotlight’ on them and report all that they could remember about that individual.

Subsequently, the originators (Fisher et al., 1987) found that real-life police interviewing of witnesses lacked much that the psychology of interpersonal communication deemed important, including the fact that anxious and inarticulate witnesses often seem unsure of their role in the interview (Fisher et al., 1990). The ECI was therefore developed and represents an allegiance between two fields of study, namely cognition and communication (Fisher et al., 2011). The ECI constitutes a number of phases providing the interview with structure: rapport building; explaining the aims of the interview (including report everything and mental context reinstatement); questioning and probing topics using mental imagery and witness-compatible questioning (asking the questions in the order of the witness’s mental representation of memory); varied retrieval (including use of senses, different order of recall and change perspectives); and closure (see Fisher and Geiselman, 1992; Fisher et al., 2010; Dando et al., in press; Milne and Bull, 1999 for a full description of each of the phases).

Over a hundred empirical studies of the effectiveness of the CI have now been published, and the vast majority of these have found that the CI/ECI elicits more correct information than a comparison interview. However, the CI has also been found sometimes to increase slightly the reporting of incorrect details (for meta-analyses of CI/ECI studies see Köhnken et al., 1999; Memon et al., 2010a). However, the accuracy of the information (proportion of correct details relative to the total amount of details reported) obtained with CI/ECI and with comparison interviews is usually almost identical (e.g. average accuracy is 85 per cent for the CI and 82 per cent for the comparison interview; Köhnken et al., 1999). The increase in correct recall with the CI has also been found with different types of interviewees; that is, adults in the general population, vulnerable populations (e.g. adults with learning disability; Milne et al., 2002 – see later in this chapter for more on interviewing vulnerable groups), and adults with low socio-economic background (Stein and Memon, 2006). In addition, it is believed to promote therapeutic jurisprudence (Fisher and Geiselman, 2010). Kebbell et al. (1999) also found that police officers perceived the CI to be useful, but officers noted that time was a major problem in applying this technique in the field. Thus shortened and less complex versions have been developed where time is of the essence (see earlier in this chapter).

There is no doubt that the CI is a successful forensic tool due to the fact that it enhances memory for accurate investigation of relevant detail. As information is at the heart of establishing the answers to the two core investigative questions, the CI should be one of the most prominent tools in any investigator’s armoury for combating crime. Indeed, many law enforcement organisations around the world over the past 20 years have themselves seen the promise and practical potential of the CI and have adopted it as one of the main interviewing frameworks (e.g. the UK, Norway, Australia and New Zealand; See Fisher et al., 2011).

Nevertheless, what research has started to show is that the CI in its entirety is rarely implemented. Early studies examining the applicability of the CI to the field were very promising in that it was demonstrated that police officers could be readily trained in a very short period of time and that the resultant interview behaviour was improved substantially (Fisher et al., 1989; Clifford and George, 1996). However, it was not until the police in England and Wales developed a national approach to interviewing, due to much public outcry as a result of miscarriage of justice cases which had poor interviewing at the heart of the acquittals (see Poyser and Milne, 2011 for a review), that the CI was for the first time adopted by a police organisation nationwide. The government and police response in the UK was to professionalise its police force, and the investigative interviewing ethos and PEACE approach to interviewing was established (see Milne et al., 2007; Griffiths and Milne, 2005). Within the PEACE framework of communication (where P stands for planning and preparation, E for engage with and explain the interview process to the interviewee, A for gaining an account, C for closure of the interview and E for evaluation of the information attained and interviewer skill level), two models were adopted by the British police service as a whole; (i) the CI and (ii) conversation management (see Shepherd and Griffiths, 2013). Thus for the first time worldwide a whole country was to train all operational officers (N = 127,000) in the use of the CI (using police officers as trainers). This was an immense step in the evolution of the CI from the laboratory to the field.

Research examining the CI’s perceived practical utility found that police officers (both experienced and frontliners) on the whole noted the CI to be a worthwhile approach, though some techniques were preferred and used more frequently than others (e.g. the report everything instruction; Kebbell et al., 1999; Dando et al., 2009a). Similarly, studies examining real-life witness interviews also found that the CI techniques were used sparingly, if at all (Clarke and Milne, 2001; Griffiths et al., 2011; Wright and Alison, 2004). There seem to be barriers to the CI being utilised to its full potential in the real world. It is believed that these barriers fall within three key areas; (i) knowledge of police trainers, (ii) the CI is not being seen as a flexible tool and (iii) the CI needs to be tailored according to the circumstances. The latter reason has already been fully dealt with within this chapter.

Psychological knowledge of police trainers is a concern; i.e. who is training the police trainers in the first place, where are they attaining their knowledge and skills to interview cognitively? In essence, trainers are being expected to be pseudo-psychologists, often overnight. Training also needs to be tailored to each trainee group independently and, for any good transference of interview skills into the workplace, trainee interviewers need to practise the interviewing skill repeatedly within the training framework (Wright and Powell, 2007). In addition, the trainees need on-the-spot constructive feedback within the training environment.

Finally, the CI needs to be seen as a flexible tool. In the past, the CI has been presented during training as a structured rigid protocol, which must be followed exactly. On the contrary, the CI is a set of tools in an interviewer’s ‘tool-belt’ that can be applied and matched according to numerous investigative factors: (i) the interviewee type; (ii) interview location; (iii) the interviewer skill level (see Griffiths and Milne, 2005 for a full discussion of the tiered approach to interview training used in the UK, which develops an interviewer’s skill level across the officer’s career span and the complexity of the case; volume versus serious crime); (iv) investigative need; and (v) how the interview is being recorded. For example, if the interview is not being electronically recorded and the record of the interview is dependent on the interviewer’s own memory (i.e. in a handwritten statement format), then passing full control of the interview to the interviewee is very difficult, because usually in such situations interviewers will always be in control as they will be taking numerous notes to help their own memories of what is being said (see Milne and Bull, 2006; Westera et al., 2011 for investigative and evidential problems with handwritten statements). Thus, the CI is not a one-size-fits-all approach (Fisher et al., 2011).

So far we have seen how fragile the human memory process is, and if interview procedures are not optimum then miscarriages of justice can occur. When examining such cases there is a golden thread running through them (e.g. Savage and Milne, 2007), in that they often involve those deemed vulnerable. It is to this group we now turn.


In our 1999 book, which explained how cognitive psychology had up to that time made a major contribution to investigative interviewing, we concluded the chapter on vulnerable people by lamenting that little research had been conducted on how best to improve such interviewing and saying that ‘Something must be done about this’ (Milne and Bull, 1999, p. 128). This is especially important because vulnerable people (e.g. all children and those adults with what used to be called ‘mental handicap’ or ‘mental retardation’) seem to be at greater risk of victimisation. Today, it may seem obvious that police officers should be able to obtain accurate and detailed information from such witnesses. However, until fairly recently, investigative/police organisations had not devoted time and effort to assisting their officers to be effective at this critical task, possibly because until around 25 years ago the discipline of greatest relevance, that is, psychology, had little to say on this topic because relevant research had not yet been conducted.

Fortunately, since that time a considerable amount of cognitive psychology research has taken place, some of which will be mentioned below. Such research has recently been informing ‘good practice’ in the applied/real world in the form of practitioner guidance, e.g. in the UK, see the extensive guidance document ‘Achieving Best Evidence In Criminal Proceedings’, published by the government in England and Wales, which has a focus on vulnerable people (

However, see Davies et al. (in press) for more on ‘Achieving Best Evidence’ and some of the difficulties interviewers experience in implementing aspects of the relevant training. Of course, people only need training if they do not already routinely do what the training requires; however, this necessitates that they change their beliefs and customary behaviour, which people typically have difficulty doing – but cognitive and social psychology can help explain why this is so (e.g. Ajzen, 2002).

In 2010, the second author of the present chapter published an overview of some cognitive aspects regarding the interviewing of vulnerable people, including children (Bull, 2010). Here we will mention some other topics.

Way back in 1976, a government-requested report called for a substantial increase in the then rather paltry research literature that sought to explain how honest witnesses can often make mistakes when trying to identify/provide information about crime perpetrators (Devlin, 1976). Soon after that, in 1978, the second author of the current chapter and a colleague published a book that overviewed the then available research and theories in cognitive (and social) psychology (Clifford and Bull, 1978). Since then, hundreds, maybe thousands, of research articles have been published on witness psychology, which are are far too numerous to be reviewed here. (For reviews, see Chapter 3 in the present book; Lindsay et al., 2007; Wilcock et al., 2008.) Instead, in the limited space available, we will briefly mention some work on applying cognitive psychology to assist a group of people who find the making of correct identification choices particularly difficult – the elderly.


In the ‘developed’ world people are living much longer than even a few decades ago, such that an increasingly large part of the population lives way beyond the typical retirement age. It is well known that with increasing age people’s cognitive abilities tend to decline, for some at a fast rate. This might help explain why researchers have found that older adults can perform at an equivalent level to younger adults in terms of identifying the perpetrator from lineups or photo-spreads in which the perpetrator is actually present, but be significantly more error prone when the perpetrator happens to be absent (which does occur in real-life investigations). In light of this, cognitive psychologists in various countries have sought to apply what we now know to improving the performance of older witnesses regarding criminal identification (e.g. Wilcock and Bull, 2014).

Ten years ago, Wilcock et al. (2005) noted that one factor that is known to reliably reduce the rate of false identifications on perpetrator-absent lineups is the issuing of instructions informing the witness that the perpetrator ‘may or may not be’ in the lineup (a compulsory police practice in some countries, such as England and Wales, and a recommendation by the Attorney General in the USA – see Wells et al., 2000). However, in her earlier research Wilcock had noticed that less than half of the elderly participants were able to recall such an instruction correctly, and that there was a significant effect of memory for this instruction on lineup accuracy, with participants who failed to remember the instructions demonstrating poorer performance (i.e. making false identifications). The literature on ageing and memory offers a number of explanations as to why the elderly may be less likely to remember these instructions. There may be problems with encoding; due to older adults’ reduced attentional resources, they may be less likely to engage in elaborate encoding of instructions. Also, concentrating on the lineup may reduce the processing resources available to spontaneously recall the instructions, which are designed to ‘counter’ the common belief that because the police have a lineup/photo-spread, the perpetrator is very likely to be in it. Therefore, she designed some experiments (see Wilcock and Bull, 2014) to try to improve performance. She found that older adults benefited from pre-lineup questions and a practice lineup prior to viewing a perpetrator-absent lineup, in that they made significantly fewer false identifications and more correct rejections compared with participants in the control (non-practice) group. Thus, these innovations reduced the suggestibility of perpetrator-absent lineups. (The practice lineup consisted of famous female faces, and witnesses were asked to identify the Queen’s face, which was not present – all participants performed correctly.)

Another applied procedure that psychology has found to benefit elderly participants is the CI (see Section 8.4 above on the CI). Holliday et al. (2012) examined the effect of a suitably modified form of the CI on older adults’ recall of a staged crime, followed the next day by misinformation presented in a written summary. They were then interviewed with either a modified CI or a control interview. The modified CI elicited more correct details and improved overall accuracy compared with a control interview. Importantly, older adults who were interviewed in a modified cognitive interview were not susceptible to the misinformation.


A group of vulnerable people who in the past have been denied full access to justice are those with autism spectrum disorder (ASD), partly because their verbal free recall of an event has often been found not to be as efficient as that of ordinary people. In a pioneering study, Mattison et al. (in press) examined the possibility that the provision of visual rather than verbal cues might assist, but they were wise enough not to employ visual cues provided by the interviewer (that could be inappropriate because in real-life crime investigations, interviewers can rarely be sure that they already know which cues could be valid rather than be suggestive or misleading). Instead, they asked children who had witnessed an event to provide their own visual cues by drawing the event while verbally recalling. Mattison et al. noted that drawing could be particularly beneficial to people with ASD because their memory/cognition is known to rely more on visual processing than is the case for ordinary people. In their experiment they found that children with ASD spontaneously recalled less of a video-recorded, staged shop theft than did ordinary children. However, although being asked to draw/sketch did assist ordinary children to recall around 18 per cent more correct information than the ordinary children not asked to draw, sketching for children with ASD improved correct recall by around 50 per cent – bringing it close to the level of performance of ordinary children not asked to draw. (For more on ordinary children recalling while drawing, see Gross et al., 2009; and for adults doing this see earlier in the present chapter.) Also, among the ASD children, those who drew more items were able to verbally recall more. Furthermore, for the proportion of recall that was correct (i.e. accuracy), whereas without sketching the ASD children performed more poorly than the ordinary children (77 and 90 per cent, respectively), sketching substantially improved ASD children’s accuracy (to 94 per cent – with ordinary children’s sketching accuracy also being 94 per cent).

One of the many possible reasons for the success of such sketching may relate to children (and adults) with ASD seeming not to prefer to look often at people with whom they interact. Perhaps closing their eyes while being interviewed might also be useful. Earlier in this chapter we overviewed some of the important research on the effectiveness of the CI. For over 20 years, one of the recommendations of the founders of the CI (Fisher and Geiselman, 1992) has been to invite interviewees to close their eyes while trying to recall fully.

In recent years, an as yet small number of experiments have found that asking people to close their eyes when verbally recalling events in interviews usually improves re-call, probably because this restricts the concomitant attending to and processing of the available visual information while trying to verbally recall (partly from visual memory) (e.g. Perfect et al., 2008). However, it seems that few crime witnesses spontaneously close their eyes when recalling an event in interviews (Vredeveldt et al., in press). In light of this, and in collaboration with the South African Police Service, Vredeveldt and her colleagues asked some police interviewers in the facial identification unit to invite witnesses to serious crimes (e.g. murder, rape, robbery) to close their eyes while being interviewed, which might seem an unnatural thing to do. Tran scripts (i.e. written records) of what these real-life witnesses recalled were assessed for their investigative/forensic relevance by a retired senior police officer who was unaware of which witnesses had closed their eyes. The accounts from the eye-closure witnesses were given higher scores by this assessor than those from the non-eye-closure witnesses. However, such a difference was not found in the assessments made by Vredeveldt (who is a research psychologist of high quality, but not an experienced police officer). Similarly, when three other researchers scored the transcripts for the number of items of information recalled, no effect of eye closure was found. Thus, to date, ‘the jury is out’ on whether eye closure will assist real-life witnesses, and research needs to examine the use of this technique with people with ASD.


Figure 8.6

Source: copyright: racorn/

Regarding juries/courts, published research from the late 1980s onwards (e.g. Flin et al., 1993; Kebbell et al., 2004; Zajac, 2009) demonstrated that lawyers were often unable to question vulnerable people in ways that could be understood. To remedy this, the use of intermediaries in court and in prior interviews by lawyers and/or the police was formally introduced a few years ago in England and Wales, these being among the first countries in the world to have a formal, government-sponsored scheme, though the use of intermediaries was pioneered years before in South Africa. In their (mandatory) training, intermediaries receive relevant comprehensive information gleaned from cognitive psychology research on such topics as question types, attention span (including concentration), use of abstract/concrete words, maximum number of words to use in a question, compliance and suggestibility, temporal concepts (before/after/last), spatial concepts (in/on/over/under/in front/behind) and narrative sequencing.

In 2010, the second author of this chapter noted that psychological research had discovered much that can inform good interviewing and that in some countries investigators (e.g. the police) have attempted to improve their practice accordingly. However, up to then, cognitive psychology had rarely been applied to the interviewing/questioning of witnesses in court proceedings. Research in England (by Kebbell et al., 2004), in Australia (by Cashmore and Trimboli, 2005) and in New Zealand (by Zajac et al., 2003; Zajac and Hayne, 2003, 2006) had made it clear that most of those professionals who question witnesses (including vulnerable ones) during court proceedings largely seem to be unaware of the constructive findings of relevant psychological research and to be rather poor at this task. Recently, cognitive psychology has contributed to seeking to improve this situation by way of a crucial online initiative called ‘The Advocates Gateway’, where advice can be found to help lawyers become more skilled (see:

In this chapter overviewing some of the most important applied contributions from cognitive psychology, we have not had the opportunity to mention such contributions from social psychology, even though in applied/real-world settings these two aspects of psychology are frequently intertwined; for example, recalling may well be facilitated by an effective rapport between the interviewer and the interviewee. (For more on that topic, see Hershkowitz et al., 2015; Walsh and Bull, 2012.) Similarly, seeking ways to improve the effectiveness of ways of validly determining if a possible victim, witness or suspect is telling the truth seems nowadays to be benefiting from modern research in both cognitive and social psychology. Regarding the applying of cognitive psychology, the notion of ethically putting cognitive load on possibly lying interviewees has become an important topic. Unfortunately, within an investigation people lie, and it is necessary for investigators to be able to distinguish between liars and truth-tellers.


Until fairly recently, research on people’s ability to determine whether they were being lied to had found that most people, including professionals such as police officers, were at or close to chance level (for a comprehensive overview, see Vrij, 2008). One reason for this was that people seemed to believe (i) that lying would be more emotional than truth-telling and (ii) that such emotion would be revealed by observable behaviour. However, research has found both of these beliefs to be not well founded, and that training based on such beliefs is rarely effective (Bull, 2004). Instead, the focus nowadays has moved on to the notions that (i) lying may usually be more cognitively demanding than truthtelling and (ii) cues to deception may be more likely to occur in what people say as opposed to how they behave (i.e. what is termed non-verbal behaviour).

One of the apparently few ‘common-sense’ beliefs about cues to deception that seems to have some validity is that liars’ accounts may sometimes contain more contradictions than truth-tellers’ (Hartwig et al., 2006 – but see Vredeveldt et al., 2014). For example, Baughman (2015) found that rape allegations deemed likely to be genuine contained fewer contradictions than those deemed likely not to be genuine. However, in order to expose contradictions one has to interview in a way that allows, indeed encourages, interviewees to talk. Similarly, Hunt and Bull’s (2012) finding, that rape allegations that were likely to be genuine contained significantly more information about speech/conversation, implies that truthful interviewees need to be encouraged to provide comprehensive accounts (this may also ‘trip up’ liars).

Furthermore, interviewers/police officers revealing what they already know to interviewees during an interview has been found to affect the number of contradictions provided. For example, in a pioneering study in Sweden, Hartwig and her colleagues found that revealing such information late, rather than early, in interviews resulted in lying interviewees saying more that was not consistent with what the interviewers knew. (For an up-to-date overview of that and similar work, see Granhag and Hartwig, 2015). Research in England published in the 1990s (see Milne and Bull, 1999) had found that police interviewers in those days often revealed to suspects much or all of such information/evidence very early on in interviews, even though doing so was not required by law. (In many countries, especially in democracies, suspects do have the right to know why they are being interviewed.) However, in 2003, the second author of this chapter noted (when supervising the doctoral research of Stavroula Soukara) that in interviews in which suspects eventually shifted from denying to admitting/confessing, the interviewers at the time of these shifts were still ‘disclosing information’, which meant that they had not revealed it early on. This led on to a programme of research looking at the possible effects of such gradual disclosure, which found that this (compared with early or late disclosure) made it significantly easier to determine if interviewees were lying or telling the truth (Dando and Bull, 2011; Dando et al., 2015), possibly because gradual-disclosure interviews were rated by the interviewees as more cognitively demanding than late or early disclosure. Our research also found with children that the timing of the interviewer’s evidence/information revelation had a significant association with (i) liars mentioning during their free recall some of this information and (ii) the total number of details they mentioned in free recall (Lingwood and Bull, 2013), and was associated in liars with more within-statement inconsistency (McDougall and Bull, 2015).

Our above experiments on the effectiveness of interviewers’ gradual revelation of information suggests that this could be of applied value, possibly because it may put more cognitive load on liars (see Granhag and Hartwig, 2015) – but what about in real-life investigations? In our analysis of interviews with people suspected of benefit fraud (Walsh and Bull, 2015), we found that interviews in which the interviewers disclosed evidence/information gradually resulted in the gaining of more comprehensive accounts, thus demonstrating effectiveness in the real world.

With regard to applying the findings of cognitive psychological research to training police, Vrij et al. (in press) recently designed a workshop for experienced detectives who, before and then after the workshop, interviewed ‘mock’ suspects. The police detectives were allowed to interview in ways they felt appropriate, but were encouraged to use information provided in the workshop. After the workshop the detectives asked better questions and more successfully discriminated between lying and truth-telling suspects. However, the detectives failed to use some of the new ideas presented during the workshop, even though they thought they had done so. (In 1995, Memon, Bull and Smith found a similar thing in connection with training police interviewers to try to fully use the CI.)


In this chapter we have provided an overview of some of the contributions being made by psychologists from applying cognitive psychology to crime investigations. However, as just noted above, successful application of cognitive psychology is not achieved solely by the carrying out of research – there is also the need for successful dissemination that leads to impact in the real world. Often, the best solutions to real-world problems are down to a combination of practitioner and academic knowledge and working in partnership to produce evidence-based techniques.


•  Witness testimony is fragile, incomplete and malleable. It is therefore important to devise investigative and interviewing procedures that maximise the accuracy of the information elicited from a witness, while minimising the risk of contamination of the witness’s memory.

•  The memory of a witness needs to be treated as though it is a part of the ‘crime scene’, which must be protected and examined in a carefully planned way.

•  Frontline interviewing needs to be quick and effective. Any delay between encoding and retrieval of information will increase the amount of forgetting and will also increase the risk of contamination from other sources.

•  Witnesses are likely to experience stress during a crime, and this could reduce the amount of detail that they can recall. A related finding is the tendency of witnesses to suffer weapon focus.

•  Inappropriate questioning during a witness interview can potentially contaminate the already fragile memory of a witness to crime.

•  Witnesses to an incident or crime should be separated from each other as soon as possible and interviewed individually, to prevent cross-witness contamination.

•  The cognitive interview is a technique that helps investigators to elicit a greater amount of information from witnesses and victims of crime, without reducing the accuracy of the information gained.

•  The original CI comprised four main principles, which are to report everything, to mentally reinstate context, to recall events in different orders and to change perspective.

•  Special care needs to be taken when interviewing vulnerable witnesses, such as children, the elderly and those with impaired mental functioning.

•  Until recently, the research findings on witness testimony and interviewing have not been widely applied in the real world. However, some countries have now introduced extensive training in these approaches for police officers, and police and researchers are increasingly working in partnership to produce evidence-based techniques.


•  Fisher, R., Milne, R. and Bull, R. (2011). Interviewing cooperative witnesses. Current Directions in Psychological Science20, 16–19.

•  Granhag, P.A., Vrij, A. and Verschuere, B. (2015). Deception detection: Current challenges and cognitive approaches (pp. 231–251). Oxford: Wiley Blackwell.

•  Milne, R. and Bull, R. (1999). Investigative interviewing: Psychology and practice. Chichester: Wiley.

•  Oxburgh, G., Myklebust, T., Grant, T. and Milne, R. (eds). (in press). Communication in legal contexts: A handbook. Chichester: Wiley.

•  Wilcock, R., Bull, R. and Milne, R. (2008). Criminal identification by witnesses: Psychology and practice. Oxford: Oxford University Press.