Back to Blog
Evidence Based Investigative Interviewing
01/05/2021
Heather Prince
The adoption of evidence-based policing has had a profound effect on many aspects of policing, but there remains one glaring omission: training and practice in the investigative interviewing of suspects.
Currently, police in Australia and New Zealand are implementing adapted versions of the UK PEACE based training model, for both interviews with suspects and witnesses. The legal requirements underpinning the UK version of PEACE (both legislation and policy) differ markedly from those in Australasia (e.g., right to silence, presence of legal representatives), and consequently, the model needs some adaptation, particularly for the questioning of suspects.
PEACE was created as an alternative to the models of interrogation used in the United States such as the Reid model (www.reid.com). While it is something of a simplification, US-based models are seen as accusatorial, centered on confessions, whilst the UK-model is seen as information gathering, centered on gathering evidence.
Setting aside such conceptual issues, we should turn our attention to whether or not either model is actually effective (we will return to the issue of defining “effective” below). This is where matters become extremely concerning. Simply put, there is remarkably little evidence that either technique is effective and, furthermore, that evidence-base is largely anecdotal.
One problem with anecdotes as the basis for policymaking is that one can easily cherry-pick examples to support one’s favored conclusion. For example, there are numerous examples (see https://innocenceproject.org/) showing that the Reid technique, is too effective at eliciting confessions, as a large number of innocent suspects have been persuaded to offer false confessions. One could then find examples showing that the PEACE model is not effective enough, and guilty suspects have been able to evade detection (e.g., the borderline sycophantic interviews by Sussex Police with sex-offender Jimmy Savile).
Pseudoscientific Policing
A reliance on anecdotes as a source of evidence is one of the surefire signs that a claim is pseudoscientific. A pseudoscience is a series of beliefs or practices that are mistakenly seen as scientific. A pseudoscience might look like a science, but it isn’t. Distinguishing between a science and a pseudoscience is often difficult, most obviously because pseudosciences borrow the language of real sciences.
For those unfamiliar with the criteria for distinguishing between science and pseudoscience, a good primer has been prepared by the Australia New Zealand Policing Advisory Agency [ANZPAA; 1]. That primer begins by highlighting the difference between scientific methods and o pinions. The scientific method is the process taken to generate a result, such as the steps taken to generate a DNA profile. An opinion would be the cognitive interpretation of that result, such as a statement that two profiles ‘match’. It is thus possible for a scientific method to be sound, but the opinion arising from it may not be.
Any opinion based on a flawed method will in most cases only be ‘correct’ by chance (or as the old saying goes, even a broken clock is correct twice a day).
ANZPAA goes on to state that “In order to adequately identify the underpinning science of a forensic discipline, each technique must be broken down into its individual elements. These individual elements may refer to an assumption or process which underpins the entirety of a method or may be related to the formation of an opinion that arises from the results of any testing or comparison performed” (p. 6). It should be possible to identify and to then evaluate each individual element of a science.
In a separate and equally worthwhile publication, ANZPAA [2] outline the importance of making claims “testable and specific”. A “testable claim is one that sets an expectation that can be met or not met”; a “specific claim is one for which an empirical study can be designed, without the need for too many different tests and measures” (p.7). Let’s illustrate this with an example.
A claim that a new interview training system works lacks specificity. Parameters for assessing such a claim need to be more precise, and as a consequence, testable. For example, as a starting point, a trainer may claim that an interviewing technique increases the number of pieces of information that are provided by suspects. This is a sound beginning but there’s still some work to do: What exactly is meant by “pieces of information”? Similarly, if we are trying to determine the efficacy of a new technique we need to be clear about what we are comparing it to. For example, Cunningham [3] studied the implementation of PEACE-based training in New Zealand (a self-report study, not an evaluation per se). The study found that both the trained and the untrained officers were aware of the core components of PEACE, which means that the research findings say very little (good or bad) about the efficacy of PEACE training.
Developing an Evidence Base for Interview Training
The need to isolate and test the elements of a claim is an important step in verifying the validity of a science. Here’s how that might work with evaluating interview training courses.
Let’s assume that an interviewing training course consists of five distinct elements. Each element is taught across one of five days. It should be possible to test the benefits of teaching each of those elements (or days) individually. This might show that elements one, two, and three have a clear impact on subsequent behavior (for the sake of simplicity, we will say that interviewing performance improves on some measure); element four has no impact on performance; and element five appears to be rather counter-productive in that performance declines.
The implication for training might be that courses should be cut from five days to three (a 40% saving in costs!) without any detriment in performance outcomes. If anything, overall performance would increase as it is not dragged down by the poor scores from days four and five.
If this example seems far-fetched, it should be noted that early tests of two major components of PEACE, the cognitive interview and conversation management, showed the following: When taught in isolation, the cognitive interview improved performance; when conversation management was taught in isolation it had no effect on performance; when both the cognitive interview and conversation management were taught together, there was no effect on performance. In short, the conversation management training did nothing on its own, and when combined with an otherwise effective technique, undermined it [e.g., 4].
If a training course is only evaluated as a whole, then all five elements are evaluated cumulatively. This might show that there is an improvement in performance (relative to a control group who have not received such training), but it would not be clear which components of the training have produced that result, and it would entirely fail to identify the fact that element five is counterproductive. Many widely cited papers arguing for the overall merits of the PEACE model clearly fall into this trap.
Evidence for the cognitive interview, of which there is an abundance, is cited as proof of the overall PEACE model. Logically, that evidence only validates (at best) half of the PEACE model. The practice of citing evidence out of context and the co-opting of largely irrelevant evidence from other domains pervades the literature base on interviewing. Proponents of both the Reid and PEACE models use this academic sleight-of-hand on a regular basis.
Those who argue for the conversation management technique do so on the belief that it is effective, not because of any evidence base. To be clear, there is no compelling evidence base that conversation management is effective, in fact, there is barely even an evidence base. Establishing a science of interviewing is complicated as it requires us to isolate the elements of effective interviewing. It is further complicated by the selection of an appropriate outcome measure. By way of comparison, the efficacy of a new medicine can be established by an objective recording of the numbers of people who live and die when that medicine is or isn’t used. This would be a fairly objective and valid outcome measure.
Establishing an objective and valid outcome for interviewing is more complex. What is a good outcome? A confession is one possible outcome measure (it would count as being reasonably objective), but if it is a false confession then it would not be a valid measure.
There are several alternate outcomes, such as numbers of successful prosecutions, or public perceptions of policing, but as important as these outcomes might be, they do not validate the science underpinning interviewing. It is entirely possible that some other factor might explain the seemingly positive findings. For example, it may be that the introduction of mandatory video recording of interviews changes interviewer behavior for the better, not the training in interviewing.
Towards a Science of Investigative Interviewing of Suspects
When the cognitive interview was first tested the training was rudimentary. The creative team behind the cognitive interview, Geiselman, Fisher [5], provided interviewers with a short, written description of the four components of the technique (combined to a total of 202 words) and little else. In later writings, they would describe the early training as lasting about 15 minutes. Despite this, these simple instructions yielded significant improvements in the recall performance of those interviewed using this technique.
The four components of the cognitive interview have since been extensively studied, resulting in some modifications to the procedure, but the core ideas first put forward in 1985 have endured remarkably well. Analysis has validated the procedure and there is now a strong scientific basis for the questioning of witnesses. A similar set of core principles is required for the interviewing of suspects. Assuming one wanted to test the PEACE model for the interviewing of suspects, it should be possible to evaluate each individual component (e.g., from phase one: a study on preparation and planning). To date, there have only been two evaluations of PEACE training with police officers. The first was conducted shortly after PEACE training began [6], the second a few years later [7]. Subsequent evaluations of PEACE have either used non-police samples (e.g., benefits center investigators) or lacked a relevant comparison group. Neither of the original two police studies provides a compelling case for the adoption of PEACE (nor do the non-police studies). To recap, the evidence-base for PEACE’s procedures for interviewing suspects is small; it is from one country; it is at least twenty years old; and it does not make a clear case that the training ‘works’.
Conclusion
Much of what passes as interviewing science is in fact, pseudoscience. Many of the elements in training courses are ineffective (producing no discernible improvements) and some are actually detrimental to performance. Talking to people, be it a victim, witness or suspect is a core part of policing. Significant progress has been made on the interviewing of victims and witnesses; it is now time to establish a science of interviewing suspects. This requires an evidence-base. This will mean opening up police investigative processes to both internal and external scrutiny, and therein lies the real problem.
References:
1. Australia New Zealand Policing Advisory Agency, A guideline to forensic fundamentals. 2016, ANZPAA.
2. Australia New Zealand Policing Advisory Agency, Empirical study design in forensic science. 2019, ANZPAA.
3.Cunningham, S., Evaluation of the implementation of investigative interviewing training and assessment (Level 1) final report, P.N.H. Organisational Assurance Group, Editor. 2010, New Zealand Police.
4.Clifford, B.R. and R. George, A field evaluation of training in three methods of witness/victim investigative interviewing. Psychology, Crime & Law, 1996. 2(3): p. 231-248.
5.Geiselman, R.E., et al., Eyewitness memory enhancement in the police interview: Cognitive retrieval mnemonics versus hypnosis. 1985. 70: p. 401-412.
6.McGurk, B.J., M.J. Carr, and D. McGurk, Investigative interviewing courses for police officers: An evaluation. Police Research Series Paper 4. 1993, London: Police Research Group, The Home Office.
7.Clarke, C. and R. Milne, National evaluation of the PEACE Investigative interviewing course. 2001, London: Home Office.
Dr Stephen Moston was the author of the Metropolitan Police-funded report Investigative Interviewing, which apart from introducing the concept of “investigative interviewing”, contained the blueprint for the development of the PEACE model. He has subsequently published widely on police interviewing techniques and suspect behavior. He is currently the Director of Forensii (www.forensii.com), an educational company specializing in legal and forensic psychological training.
Heather Prince
The adoption of evidence-based policing has had a profound effect on many aspects of policing, but there remains one glaring omission: training and practice in the investigative interviewing of suspects.
Currently, police in Australia and New Zealand are implementing adapted versions of the UK PEACE based training model, for both interviews with suspects and witnesses. The legal requirements underpinning the UK version of PEACE (both legislation and policy) differ markedly from those in Australasia (e.g., right to silence, presence of legal representatives), and consequently, the model needs some adaptation, particularly for the questioning of suspects.
PEACE was created as an alternative to the models of interrogation used in the United States such as the Reid model (www.reid.com). While it is something of a simplification, US-based models are seen as accusatorial, centered on confessions, whilst the UK-model is seen as information gathering, centered on gathering evidence.
Setting aside such conceptual issues, we should turn our attention to whether or not either model is actually effective (we will return to the issue of defining “effective” below). This is where matters become extremely concerning. Simply put, there is remarkably little evidence that either technique is effective and, furthermore, that evidence-base is largely anecdotal.
One problem with anecdotes as the basis for policymaking is that one can easily cherry-pick examples to support one’s favored conclusion. For example, there are numerous examples (see https://innocenceproject.org/) showing that the Reid technique, is too effective at eliciting confessions, as a large number of innocent suspects have been persuaded to offer false confessions. One could then find examples showing that the PEACE model is not effective enough, and guilty suspects have been able to evade detection (e.g., the borderline sycophantic interviews by Sussex Police with sex-offender Jimmy Savile).
Pseudoscientific Policing
A reliance on anecdotes as a source of evidence is one of the surefire signs that a claim is pseudoscientific. A pseudoscience is a series of beliefs or practices that are mistakenly seen as scientific. A pseudoscience might look like a science, but it isn’t. Distinguishing between a science and a pseudoscience is often difficult, most obviously because pseudosciences borrow the language of real sciences.
For those unfamiliar with the criteria for distinguishing between science and pseudoscience, a good primer has been prepared by the Australia New Zealand Policing Advisory Agency [ANZPAA; 1]. That primer begins by highlighting the difference between scientific methods and o pinions. The scientific method is the process taken to generate a result, such as the steps taken to generate a DNA profile. An opinion would be the cognitive interpretation of that result, such as a statement that two profiles ‘match’. It is thus possible for a scientific method to be sound, but the opinion arising from it may not be.
Any opinion based on a flawed method will in most cases only be ‘correct’ by chance (or as the old saying goes, even a broken clock is correct twice a day).
ANZPAA goes on to state that “In order to adequately identify the underpinning science of a forensic discipline, each technique must be broken down into its individual elements. These individual elements may refer to an assumption or process which underpins the entirety of a method or may be related to the formation of an opinion that arises from the results of any testing or comparison performed” (p. 6). It should be possible to identify and to then evaluate each individual element of a science.
In a separate and equally worthwhile publication, ANZPAA [2] outline the importance of making claims “testable and specific”. A “testable claim is one that sets an expectation that can be met or not met”; a “specific claim is one for which an empirical study can be designed, without the need for too many different tests and measures” (p.7). Let’s illustrate this with an example.
A claim that a new interview training system works lacks specificity. Parameters for assessing such a claim need to be more precise, and as a consequence, testable. For example, as a starting point, a trainer may claim that an interviewing technique increases the number of pieces of information that are provided by suspects. This is a sound beginning but there’s still some work to do: What exactly is meant by “pieces of information”? Similarly, if we are trying to determine the efficacy of a new technique we need to be clear about what we are comparing it to. For example, Cunningham [3] studied the implementation of PEACE-based training in New Zealand (a self-report study, not an evaluation per se). The study found that both the trained and the untrained officers were aware of the core components of PEACE, which means that the research findings say very little (good or bad) about the efficacy of PEACE training.
Developing an Evidence Base for Interview Training
The need to isolate and test the elements of a claim is an important step in verifying the validity of a science. Here’s how that might work with evaluating interview training courses.
Let’s assume that an interviewing training course consists of five distinct elements. Each element is taught across one of five days. It should be possible to test the benefits of teaching each of those elements (or days) individually. This might show that elements one, two, and three have a clear impact on subsequent behavior (for the sake of simplicity, we will say that interviewing performance improves on some measure); element four has no impact on performance; and element five appears to be rather counter-productive in that performance declines.
The implication for training might be that courses should be cut from five days to three (a 40% saving in costs!) without any detriment in performance outcomes. If anything, overall performance would increase as it is not dragged down by the poor scores from days four and five.
If this example seems far-fetched, it should be noted that early tests of two major components of PEACE, the cognitive interview and conversation management, showed the following: When taught in isolation, the cognitive interview improved performance; when conversation management was taught in isolation it had no effect on performance; when both the cognitive interview and conversation management were taught together, there was no effect on performance. In short, the conversation management training did nothing on its own, and when combined with an otherwise effective technique, undermined it [e.g., 4].
If a training course is only evaluated as a whole, then all five elements are evaluated cumulatively. This might show that there is an improvement in performance (relative to a control group who have not received such training), but it would not be clear which components of the training have produced that result, and it would entirely fail to identify the fact that element five is counterproductive. Many widely cited papers arguing for the overall merits of the PEACE model clearly fall into this trap.
Evidence for the cognitive interview, of which there is an abundance, is cited as proof of the overall PEACE model. Logically, that evidence only validates (at best) half of the PEACE model. The practice of citing evidence out of context and the co-opting of largely irrelevant evidence from other domains pervades the literature base on interviewing. Proponents of both the Reid and PEACE models use this academic sleight-of-hand on a regular basis.
Those who argue for the conversation management technique do so on the belief that it is effective, not because of any evidence base. To be clear, there is no compelling evidence base that conversation management is effective, in fact, there is barely even an evidence base. Establishing a science of interviewing is complicated as it requires us to isolate the elements of effective interviewing. It is further complicated by the selection of an appropriate outcome measure. By way of comparison, the efficacy of a new medicine can be established by an objective recording of the numbers of people who live and die when that medicine is or isn’t used. This would be a fairly objective and valid outcome measure.
Establishing an objective and valid outcome for interviewing is more complex. What is a good outcome? A confession is one possible outcome measure (it would count as being reasonably objective), but if it is a false confession then it would not be a valid measure.
There are several alternate outcomes, such as numbers of successful prosecutions, or public perceptions of policing, but as important as these outcomes might be, they do not validate the science underpinning interviewing. It is entirely possible that some other factor might explain the seemingly positive findings. For example, it may be that the introduction of mandatory video recording of interviews changes interviewer behavior for the better, not the training in interviewing.
Towards a Science of Investigative Interviewing of Suspects
When the cognitive interview was first tested the training was rudimentary. The creative team behind the cognitive interview, Geiselman, Fisher [5], provided interviewers with a short, written description of the four components of the technique (combined to a total of 202 words) and little else. In later writings, they would describe the early training as lasting about 15 minutes. Despite this, these simple instructions yielded significant improvements in the recall performance of those interviewed using this technique.
The four components of the cognitive interview have since been extensively studied, resulting in some modifications to the procedure, but the core ideas first put forward in 1985 have endured remarkably well. Analysis has validated the procedure and there is now a strong scientific basis for the questioning of witnesses. A similar set of core principles is required for the interviewing of suspects. Assuming one wanted to test the PEACE model for the interviewing of suspects, it should be possible to evaluate each individual component (e.g., from phase one: a study on preparation and planning). To date, there have only been two evaluations of PEACE training with police officers. The first was conducted shortly after PEACE training began [6], the second a few years later [7]. Subsequent evaluations of PEACE have either used non-police samples (e.g., benefits center investigators) or lacked a relevant comparison group. Neither of the original two police studies provides a compelling case for the adoption of PEACE (nor do the non-police studies). To recap, the evidence-base for PEACE’s procedures for interviewing suspects is small; it is from one country; it is at least twenty years old; and it does not make a clear case that the training ‘works’.
Conclusion
Much of what passes as interviewing science is in fact, pseudoscience. Many of the elements in training courses are ineffective (producing no discernible improvements) and some are actually detrimental to performance. Talking to people, be it a victim, witness or suspect is a core part of policing. Significant progress has been made on the interviewing of victims and witnesses; it is now time to establish a science of interviewing suspects. This requires an evidence-base. This will mean opening up police investigative processes to both internal and external scrutiny, and therein lies the real problem.
References:
1. Australia New Zealand Policing Advisory Agency, A guideline to forensic fundamentals. 2016, ANZPAA.
2. Australia New Zealand Policing Advisory Agency, Empirical study design in forensic science. 2019, ANZPAA.
3.Cunningham, S., Evaluation of the implementation of investigative interviewing training and assessment (Level 1) final report, P.N.H. Organisational Assurance Group, Editor. 2010, New Zealand Police.
4.Clifford, B.R. and R. George, A field evaluation of training in three methods of witness/victim investigative interviewing. Psychology, Crime & Law, 1996. 2(3): p. 231-248.
5.Geiselman, R.E., et al., Eyewitness memory enhancement in the police interview: Cognitive retrieval mnemonics versus hypnosis. 1985. 70: p. 401-412.
6.McGurk, B.J., M.J. Carr, and D. McGurk, Investigative interviewing courses for police officers: An evaluation. Police Research Series Paper 4. 1993, London: Police Research Group, The Home Office.
7.Clarke, C. and R. Milne, National evaluation of the PEACE Investigative interviewing course. 2001, London: Home Office.
Dr Stephen Moston was the author of the Metropolitan Police-funded report Investigative Interviewing, which apart from introducing the concept of “investigative interviewing”, contained the blueprint for the development of the PEACE model. He has subsequently published widely on police interviewing techniques and suspect behavior. He is currently the Director of Forensii (www.forensii.com), an educational company specializing in legal and forensic psychological training.