Testimonies are an important part of the legal process, at two main levels: They can not only contribute to resolving a police inquiry by providing investigators with vital information about the criminal event, but also provide judges and juries with evidence that will influence their decisions about the guilt of an accused person (Loftus, 1975).
Despite these important facts, those responsible for obtaining testimonies (police officers, examining magistrates, etc.) are not routinely trained to conduct interviews and often have no knowledge of inter- viewing techniques that might optimize the retrieval of information from a witness or a victim (Fisher, Geiselman, & Raymond, 1987; Ginet, Py, Joguet-Recordon, & Gendre, 1998; Memon, Holley, Milne, Köhnken, & Bull, 1994). In an attempt to solve this problem, an American research team came up with a realistic interview technique in 1984: the Cognitive Interview (Geiselman, Fisher, Firstenberg, Hutton, Sullivan, Avetissian, & Prosk, 1984).
The Cognitive Interview requires giving four instructions that the witness is asked to follow throughout the interview.
1 / Return to both the environmental and the emotional context of the scene of the crime (mental reinstatement of context).
2 / Recall the maximum amount of information, even if it appears to have little relevance or is accorded a lower level of confidence (hypermnesia).
3 / Recount the scene in a different chronological order, for example, from the end to the beginning (change of narrative order).
4 / Recount the scene from a different perspective, for example, by telling it from the point of view of another person who was involved (change of perspective).
The first two instructions were drawn mainly from Tulving’s model of encoding specificity (1983) and the idea of bringing the context to bear. In particular, they make use of the principle that information is more likely to be recalled if the circumstances in which the retrieval takes place are similar to those present during encoding, especially when it comes to contextual elements (the witness’s physical environment and mood). In support of this hypothesis, much research has shown that simply putting the participant back in the encoding context during an experiment enhances performance, irrespective of the environmental setting (Godden & Baddeley, 1975), the internal context such as mood (Bower, Gilligan, & Monteiro, 1981; Teasdale & Fogarty, 1979), or the person’s emotional (Macht, Spear, & Levis, 1977) or psychophysiological state (Eich, Weingartner, Stillman, & Gillin, 1975; Goodwin, Powell, Bremer, Hoine, & Stern, 1969). A comparable level of success has also been demonstrated using both mental and physical reinstatement of context (Malpass & Devine, 1981; Smith, 1979).
The instruction to return to the original context is aimed at putting the witness’s mind back into contextual conditions that are as close as possible to those prevailing at the moment of the event. In addition to having a beneficial effect on the interviewee’s memory, mental recontextualization offers obvious practical benefits over physical recontextualization  It should be noted that « reconstruction », which...: Returning to the actual scene takes more time and can prove counterproductive if the environment has changed since the event.
The hypermnesia instruction is designed to also assist in the recovery of the encoding and retrieval contexts. Details of minor importance tend to act as powerful retrieval cues which allow interviewees to gain access to their memories with greater ease. The hypermnesia instruction has the additional advantage of preventing witnesses from failing to report details which they think might be irrelevant but which could prove to be highly important within the framework of the inquiry. (Witnesses are not qualified to judge what may or may be not be pertinent from a legal point of view.) In short, this instruction forces interviewees to lower their degree of self-confidence. Given that it has not been possible to establish a relationship between confidence and accuracy (Bertone, Mélen, Py, & Somat, 1995; Wells & Murray, 1984), the fact of compelling interviewees to provide information which is accorded a lower level of confidence does not usually result in an increase in errors. Moreover, when they are not sure of a certain point, witnesses spontaneously mention this fact.
The last two instructions are based on another concept: the script (Schank & Abelson, 1977). The instructions to change the order and perspective are designed to make use of certain effects brought about by the influence of script-based knowledge on recall. More precisely, the order-change instruction deactivates the script initiated by the critical event by removing the retrieval cue of sequentiality (Bekerian & Bower, 1983). In this case, the witness should no longer be able to make use of this type of knowledge to retrieve information from memory. Details that are inconsistent with the script should thus be rendered more accessible, and consistent information not contained in the scene itself should be less easily evoked in error (logical inferences, confabulations). These hypotheses were confirmed in an experiment carried out by Geiselman and Callot (1990). Another advantage of the change-of-order instruction is that it makes use of recency effects (Whitten & Leonard, 1981). The most recent details, that is, those the witness is asked to recall first if the instructions are followed, are the easiest to remember. These details may therefore function as retrieval cues and enhance the recall of chronologically recent information, which in turn, functions as a cue and facilitates the recall of earlier details, and so on. Through a sort of “snowball effect”, overall recall should improve.
The perspective-change instruction leads witnesses to activate other scripts or schemas spontaneously associated with the scene (Anderson & Pichert, 1978). This should help them recall details that are consistent with the newly adopted point of view (and which could correspond to elements that are inconsistent with the initial script and were not remembered spontaneously).
Since 1984, cognitive interviewing has been studied in about fifty experiments and has proven to be a very effective technique. By effective, we mean that a significant increase in the quantity of correct information remembered by the witness is achieved without a comparable increase in the number of errors, or at least, without any deterioration in the accuracy level of the testimonies.
Cognitive interviews seem to work equally well on ordinary adults (Aschermann, Mantwill, & Köhnken, 1991; Geiselman et al., 1984; Geiselman, Fisher, MacKinnon, & Holland, 1986) as on specific populations with memory deficiencies and greater suggestibility, such as children (Geiselman & Padilla, 1988; McCauley & Fisher, 1995; Memon, Wark, Holley, Bull, & Köhnken, 1996; Saywitz, Geiselman, & Bornstein, 1992), the elderly (Mello & Fisher, 1996), and the mentally handicapped (Brown & Geiselman, 1990; Milne, Clare, & Bull, 1999).
Although this interview technique was shown to be immediately effective in the laboratory, it was not tested for use in the field by the concerned investigators (mostly police officers). The first experiments carried out in collaboration with law-enforcement officials showed that the technique did not present any major difficulties as far as its application was concerned (i.e., the training period was short, the instructions were well applied by the investigators, little increase was observed in the length of the interviews, etc.), and that it could be used on a population of witnesses of a variety of socioeconomic groups and ages (Geiselman, Fisher, MacKinnon, & Holland, 1985; Geiselman, Fisher, MacKinnon, & Holland, 1986, Experiment 1). Fisher, Geiselman, and Amador (1989) made further progress in validating the technique in the field by undertaking an experiment in a real situation. Sixteen experienced police inspectors from Miami, Florida conducted two interviews on each of 47 witnesses or victims of shoplifting or mugging. Between the two interviews, seven inspectors were trained in theories of memory and in the cognitive interview technique. The other nine officers formed the control group. The researchers measured both the increase in facts elicited in the second interview compared with the first, and the number of facts elicited by the police officers trained in cognitive interviewing in comparison with the performance of officers who had not been trained. The results indicated an overall benefit of the cognitive interview: The trained investigators elicited 47 % more facts relating to matters already examined, whereas the benefit from the second interview proved negligible for the officers who had not been trained in cognitive interviewing.
This study is of considerable practical interest, even if it poses several methodological problems. Indeed, apart from the facts that it is difficult to define “elicited facts” and that we do not have reliable information concerning the accuracy of the testimonies other than the high level of consistency in the information given by different witnesses to the same offence (this was true for half of the cases in the study), the fact that the officers in the control group were not trained may have affected their motivation in ways discovered in early studies of human relations (Mayo, 1945; Roethlisberger & Dickson, 1939). Furthermore, there is no data relating to the investigators’ understanding of the different techniques involved in the cognitive interview and of the possible problems the witnesses might have had in applying the instructions.
This last problem was pointed out in a study by some English and German researchers (Memon et al., 1994) following their failure to replicate the effects of the cognitive interview technique when used by police officers. In their study, 38 police officers with an average of 10 years of job experience participated in a training session that lasted four hours. The entire group was made aware of the general principles of communication, good and bad questioning techniques, and the effects of different question formats (open, closed, leading questions, etc.). Then half of the group was trained in the cognitive interview technique, while the other half studied the principles of communication and the problem of question wording in greater detail. Sample situations were set up in both groups. Immediately after the training session, the police officers were asked to interview witnesses of a staged event (a simulated armed robbery in the parking lot of the police training school a few hours earlier). The results did not indicate any benefits of cognitive interviewing over standard interviewing as far as the quantity of correct information remembered by the witnesses was concerned. There was also no difference between the two groups in the number of errors and confabulations. Moreover, it appeared as though the cognitive interview training had little effect on the interview procedures used by the police officers: There was no difference between the two groups with respect to the number and type of questions asked (interviews remained largely unstructured and dominated by short, closed questions asked abruptly and at a rapid pace), and the cognitive interview instructions were infrequently used. Furthermore, for the most part, the techniques were not really presented by the investigators in such a way that the witnesses could understand what was expected of them. These results run counter to those obtained by Fisher et al. (1989) and led the English-German team to conclude that their attempt to implement cognitive interview techniques in the field had failed.
The study reported in the present article was aimed at field testing a French version of the cognitive interview. This version puts particular emphasis on the use of free recall by the witness, and on the separation of the different cognitive interview instructions. The witness performs an initial free recall while applying the hypermnesia and reinstatement-of-context instructions. This is followed by a second free recall while applying the change-of-order instruction, and a final recall while applying the change-of-perspective instruction (Ginet, 1998; Ginet & Py, 1996; Py, Ginet, Desperies, & Cathey, 1997). Free recall has the advantage of being rich in information; it also limits the amount of influence the investigator has on the interviewee (Fisher et al., 1987). Bear in mind, however, that strictly speaking, this study is not an experiment in an actual situation, in the sense that it was initially necessary a) to establish minimal control conditions for factors that could affect recall independently of the effect of the cognitive interview (encoding conditions, violence of the event, duration of the detention period, etc.), and above all b) to use measures likely to precisely assess the effectiveness of the technique. In particular, in order to overcome the limitations of Fisher et al.’s (1989) study, it had to be possible to gauge the veracity of the remarks made by the witnesses. This entailed working with a critical event that could be assessed (in this case, films) and a population that was not directly involved in the event (students).
The study had several objectives: 1 / Determine how easily cognitive interviewing techniques were assimilated by investigators trained for only a short period of time. 2 / Assess its suitability in the field while keeping in mind the requirements of day-to-day operations in a police station (limited time and space, etc.). 3 / Assess the effectiveness of the cognitive interview not only through complete testimonies (all information that became apparent in the course of the interview), but also through the written record of the witness’s statement. This last point seemed particularly important from a practical standpoint, since according to Napoleonic law in France, the interview write-up is the only record of the witness’s version of the events made available to the attorney and judge. There is no guarantee that the benefits brought about by the use of the cognitive interview will necessarily be transferred to the witness’s statement. In fact, this written record necessarily involves a certain amount of amendment, interpretation, and memorization on the part of the investigator, and this alone could negate the benefits brought about by cognitive interviewing (oversights, distortions, and condensation of the witness’s account).
In this study, three types of interview procedures were compared: 1 / standard interview procedures usually practiced by police officers; 2 / procedures commonly used by investigators who are aware of the techniques and pitfalls of interactive interviewing, including the importance of free recall, zeroing-in, question wording, etc., which is an improvement over the standard police interview (hereafter called structured interview); and 3/ the cognitive interview. The first two conditions were designed to enable comparison with the cognitive interview on a group in which the only difference would reside in the way in which the instructions were applied and to control potential motivational factors related to participation in a training session.
It was predicted that the cognitive interview would lead to a significant increase in the amount of correct information compared with a standard interview or a structured interview, without a concomitant rise in errors or confabulations. This benefit should be apparent both for the information provided in the course of the interview and for that produced in the written record of the witness’s testimony.
The data from this experiment demonstrate that the cognitive interview technique was successfully used in the field by investigators likely to employ this tool in their work. This technique led to a substantial gain in the amount of correct information remembered by the witnesses in the course of the interview as compared to a standard (+ 27 %) or structured (+ 36 %) approach. The gain was also apparent in the write-ups of the witnesses’testimonies when comparing the cognitive interviews with the structured ones, although the difference between the cognitive and standard interviews was only tendential in the written statements. The reformulation of the interviewee’s account by the investigator and the writing up of the report did not constitute a major obstacle to the use of the cognitive interviewing instructions, as Köhnken, Thürer, and Zoberbier demonstrated in 1995. However, as in their study, the present data indicated a reduction in the benefit of the cognitive interview when write-ups rather than tape recordings of the interviews were analyzed. Several interpretations seem likely. The fact that we presented the experiment as an attempt to demonstrate the greater efficacy of new interview techniques (i.e., the cognitive interview and the structured interview), compared with the usual procedures, may have encouraged the investigators to defend their usual practices. In this case, they would have tended to produce a more systematic write-up of the witness’s account in the standard interview. However, another interpretation cannot be readily ruled out: The increase in information in the cognitive interview might include details pertinent to the inquiry, but also some proportion of unimportant details which the investigator did not deem worth recording. A third interpretation would be that police officers have an a priori conception of what the write-up of a testimony should look like (Fisher et al., 1987; Ginet et al., 1998), in terms of style (especially judicial and legal terminology), background (order of events, description of persons involved, etc.), and length. This preconceived idea may lead them to “smooth out” the information contained in the declarations.
This study also allowed us to ascertain that, as a whole, the effect of the type of interview was more clearly in favor of cognitive interviewing when it was compared with the structured interview than when compared with the standard police interview. The standard interview corresponds to the police officers’usual way of proceeding, but the other two types of interviewing require modifications of their daily job practices. From the methodological standpoint, then, the basis for a legitimate comparison of the effects of the cognitive interview seems to be the structured interview. The fact that the benefits of the cognitive interview were clearer when compared to the structured interview rather than the standard interview is therefore not surprising. More surprising was the ineffectiveness of the structured interview compared with the standard police interview. Structured interviewing is supposed to correct the main shortcomings of the usual procedures used by interviewing investigators: Greater reliance upon free recall (which was achieved here), attention paid to how the questions were worded (this proved more difficult), etc. It is reasonable to suggest that having to modify their usual practices is an annoyance for police officers, at least in the short term. This was true here not only of the structured interview, but also of the cognitive interview.
Although, generally speaking, the structured interview provided the best basis for the favorable assessment of cognitive interviewing, the standard interview revealed the cognitive interview’s potential ability to reduce errors and confabulations. In fact, the technique exhibited the following three qualitatively beneficial characteristics:
1 / Emphasis was placed on free recall, particularly in the French version of the cognitive interview (information remembered spontaneously is generally more accurate than information provided in response to questions: Lipton, 1977).
2 / As a direct consequence of the importance attached to free recall, the investigator asked fewer questions (compared with the structured interview) and this limited the police officer’s influence on the witness, especially in the case of leading questions. Note, however, that no difference was found in the number of questions of this kind asked before and after training. Although investigators were made aware of the negative effect of badly worded questions, they did not succeed in avoiding them. This result is consistent with the findings obtained by Fisher-Holst and Pezdek (1992), who demonstrated that leading questions could have a powerful impact on the recall of participants even if they were forewarned of the risks of being influenced by such questions. According to the results of our study, it seems that the investigators were not aware that asking questions of a leading nature might influence the interviewee (e.g., “Was the girl young?”, “Was it a big car?”, “Did that car have four doors?”, “Did he tear her clothes, rip them off?”).
3 / Information was more accessible and recollection more detailed as a result, which made the witnesses more resistant to suggestion (Geiselman et al., 1986; Milne, Bull, Köhnken, & Memon, 1995). The structured interviews had two of these three characteristics (free recall and a small number of questions). For this reason, we should expect to see a tangible and qualitative difference when comparing the cognitive interview with the standard one. That is indeed what was found regarding the number of confabulations produced (and transcribed in the written statements) in the cognitive interviews compared with the standard ones.
Although this study has demonstrated that the cognitive interview successfully enhances the richness and quality of recollection (its use led to an increase in the accuracy level of the testimonies), it has also shown that this technique is easily assimilated by police officers even after only a short period of training. These results are in line with those obtained by Geiselman et al. (Geiselman et al., 1986), who were able to show that the cognitive interview could be very successfully employed by police officers trained in a short 20-minute session. Remember, however, that Memon et al. (1994) failed to find any effect of the technique when used by police officers who had undergone training lasting more than four hours. Perhaps the training given by Memon was ineffective. The authors in fact emphasized that the cognitive interviewing instructions were under-used by the investigators. In the current study, nothing of this kind happened: Apart from the hypermnesia instruction, the cognitive interviewing instructions were systematically followed by the police officers.
Finally, it is worth noting that the cognitive interview did not pose any major application difficulties with respect to the physical and temporal limitations involved in the day-to-day operations of a police station. Particularly, using this technique did not seem to take more time. However, it is planned for use principally in serious and exceptional circumstances. For more minor events, the technique could prove costly both for the investigator and the witness when weighed against the potential benefits. Perhaps it would be a good idea to adapt the cognitive interview to such circumstances by only employing, for example, one or two of the instructions rather than all four. However, we currently lack the necessary data for pinpointing the specific effects of each instruction in order to use them in a manner appropriate to the concerned witness (for example, it is difficult to ask victims to try and put themselves in the shoes of their attacker) and to the type of event treated in the testimony (degree of violence, level of familiarity, etc.).
In short, this study strongly indicates that the cognitive interview could be fruitfully employed in the field. At the moment, its application in preliminary investigations of criminal events in France and Belgium is rare. Among the 200 police detectives, officers, and examining magistrates we trained in investigative interview techniques in these two countries, most appear to use some aspects of the technique from time to time: free recall, hypermnesia, reinstatement of mental context, change of narration order. Two problems limit the general use of the cognitive interview, however. The first has to do with the risks incurred by the use of the change-of-perspective instruction. In fact, many law-enforcement and judicial investigators are reluctant to ask a witness to adopt the point of view of another protagonist because of the risk of confabulations this instruction could generate. Although these studies do not reinforce this viewpoint, it cannot be argued either that there is no negative effect on confabulations resulting from the change of perspective. The second problem concerns the specific nature of Napoleonic penal procedures. This inquisitorial type of procedure in fact assumes that the written statement is the sole record of the interview with a witness (the only exception in France since June 1999 being child interviews). Law-enforcement and judicial investigators in other countries with a similar penal tradition find it necessary to use a technique which includes a written statement. Consequently, our current research aims are, firstly, to replace the change of perspective with a less problematic instruction, and secondly, to incorporate the statement-writing phase into the formula.
Finally, we might add that the cognitive interview technique was developed in a judicial context, where the testimony is of major importance. Yet all evidence suggests that other social fields could benefit from the use of a tool that can enhance people’s memory. As Leplat stated (1996), the cognitive interview could easily be adapted to investigations of occupational accidents. In fact, there is no a priori difference in terms of recollection of facts by the victim between an accident at work, a road accident, and an attack. The results of a study currently in progress in a large public transport firm seem to indicate that the cognitive interview technique can greatly assist in understanding why accidents happen. Thus, this technique could also be useful in other social fields that are far removed from the law-enforcement and judicial systems, such as medical diagnostic research, or even as a “crutch” for the memory of elderly people.
Paper received: September 1999.
Accepted in modified form: February 2001.