Memon, A. Cronin, O, Eaves, R. and Bull, R. (1996) An empirical test of the `mnemonics components' of the Cognitive Interview. In: G.M. Davies, S. Lloyd-Bostock, M. McMurran and C. Wilson (eds.) Psychology. Psychology and Law: Advances in Research. Berlin: De Gruyter.

AN EMPIRICAL TEST OF THE 'MNEMONIC COMPONENTS' OF THE COGNITIVE INTERVIEW

Amina Memon, Orla Cronin, Richard Eaves
University of Southampton

Ray Bull
Portsmouth University

Abstract

This paper reports some preliminary research in which the various mnemonic components of the Cognitive Interview were isolated and compared with an instruction to try harder. The purpose of this was to control for possible motivational effects. Adult and children (aged 5- 9 years) were asked to recall details of a staged event under one of four instruction conditions. In the first condition, they were encouraged to reinstate context; in the second condition to change perspective; in the third condition to report the event on reverse order and in the fourth condition to try harder. There were no significant differences in correct recall or errors as a function of instruction condition. An analysis of different types of information reported in the interviews suggested some types of information were more frequently reported. Accuracy rates were high and while the older children recalled more information correctly they were no more accurate than the younger children. The implications of the findings are discussed with reference to some of the theoretical and methodological issues raised by research on cognitive interviewing.

INTRODUCTION

Experimental studies of human memory have clearly demonstrated its limited capacity to store information, its reconstructive nature and its deterioration over time (e.g. Kail, 1979). It is also well established that the recall of information from memory is influenced by the strategies used to gain access to that information (Ornstein, Medlin, Stone, & Naus, 1985). This knowledge has in recent years, been applied to the practical problem of remembering in the context of eyewitness memory. The cognitive interview (CI) was developed by Geiselman and Fisher as a procedure for increasing the amount of correct information recalled by a witness (Geiselman, Fisher, MacKinnon & Holland, 1985). CI was presented in the form of four 'mnemonic' strategies and was derived from pooling research on retrieval pathways, encoding specificity, and schema related recall. The success of the CI relative to a standard police interview has been attributed to the cognitive, memory enhancing components of the procedure (e.g. Geiselman, 1988). In the 'enhanced' CI (Fisher and Geiselman, 1992) a further increase in recall is attributed to other, more 'social' aspects of the interview setting and procedure.

The aim of this paper is to evaluate experimentally the effectiveness of the cognitive components of the CI procedure using adult and child witnesses to staged events. Empirical support for the technique was generated by several studies of eyewitness memory performed by Fisher, Geiselman and colleagues which have been replicated by researchers in Germany. (See Memon & Köhnken, 1992 for a recent review).

This paper reports some preliminary research in which the various mnemonic components of the CI were isolated and compared with a control group and in which we attempt to control for possible motivational elements induced by the CI (e.g to try harder). This paper also examines age differences in recall of event information and the effectiveness of CI instructions with child witnesses. This follows an earlier study with six year olds where the full CI was no more effective than a good standard interview (Memon, Cronin, Eaves & Bull, 1992). The Cognitive Interview. The original CI procedure (e.g. Geiselman et al., 1985; 1986) comprises four 'cognitive' (mnemonic) techniques. The CI-trained interviewer encourages the interviewees/witnesses to use these techniques to help them remember. The CI techniques are derived from two principles regarding the structure of memory. Firstly, according to Tulving (1974), multiple retrieval pathways exist to a given encoded memory of an event and so a variety of retrieval cues or strategies are needed to maximise the chances of a complete recall. Two such strategies are the 'change order' and 'change perspective' instructions which encourage interviewees to recount events in a variety of orders, and from a variety of perspectives, respectively. These techniques have been shown to be effective in eliciting extra details (Loftus & Fathi, 1985; Anderson and Pichert, 1978) and may work in part by reducing the extent to which prior knowledge, expectations and schema influence and/or limit recall (Norman and Bobrow, 1978).

The second principle adopted in the CI is Tulving & Thomson's Encoding Specificity theory (1973). A memory trace is made up of several features; to maximise recall, an effective retrieval cue needs to comprise as many of these features as possible. Context provides cues which increase feature overlap between initial witnessing and subsequent retrieval contexts (cf. Flexser & Tulving, 1978). Context reinstatement involves emotional elements ('How were you feeling at the time?'), which may work via state-dependent effects, perceptual features ('Put yourself back at the scene of the crime and picture the room, how did it smell, what could you hear?'), and sequencing elements ('What were you doing at the time?').

The perspective, order and context instructions are directly related to memory theory. The status of the fourth - an instruction to the witness to report every detail irrespective of perceived importance or triviality as a mnemonic technique is not as clear. In practice, this instruction is given in conjunction with the instructions that precede each free recall by the witness, and is a means of enhancing the communication by the witness of the information elicited by whichever of the other techniques is being used at the time. Thus, use of this instruction recognises that, from a forensic point of view, even seemingly trivial facts can be important. For the purposes of this study, the 'report every detail' instruction will not be considered as a separately assessable mnemonic technique, primarily because alone it does not instruct the witness to make a recall attempt, but instead is given as a supplementary instruction to those that do (Eaves, 1992).

The CI was developed, tested and refined in a series of studies employing trained interviewers (police and students), simulated incidents and police training films (e.g. Geiselman et al., 1985), and, more recently, in field studies using a small sample of detectives and witnesses to real events in Florida (Fisher, Geiselman & Amador, 1989) and in the U.K. (George & Clifford, 1991). Later studies of the CI have largely used the 'enhanced' version which includes some general principles for improving communication and facilitating the effective use of context reinstatement.

Acquisition of mnemonic competency A review of the effects of rehearsal on memory retrieval across various groups shows that older children are more likely to engage in active rehearsal than younger ones and are more efficient in the implementation of mnemonic strategies (e.g. Ornstein et al., 1985). However, there is also evidence to suggest that if children are directed to relevant cues and sufficiently prompted to use cues, ages differences in recall performance are eliminated (Kobasigawa, 1974, Pipe, Gee and Wilson, 1993).

This evidence suggests the cognitive interview, if appropriately applied, may facilitate recall in child witnesses. The CI has been modified in recent years and tested using children as witnesses to staged events. Memon and Köhnken (1992) review the recent research, a sample of which will be outlined here. Studies by Geiselman and colleagues (Geiselman & Padilla, 1988; Saywitz, Geiselman & Bornstein, 1992) suggest that CI (the original procedure) is as effective with children as it is with adults, particularly if children are familiarised with the CI before being questioned. However, Köhnken, Finger, Nitschke and Hofer (1992) report a substantial increase in confabulated details when the enhanced version of CI was used with children.

In an earlier study, we compared the original CI with an interview that was identical to it in all respects apart from the mnemonic techniques (this latter interview being called the standard interview or SI). We found no significant differences between CI and SI in a between subject or within subject comparison (Memon et al., 1992). There were qualitative differences in the interview transcripts which suggested that not all the child interviewees understood and could use the CI techniques (cf. Cronin, Memon, Eaves, Küpper & Bull, 1992). Interviewers showed some individual differences in their application of the various CI instructions. This was also noted in the Saywitz et al (1992) study, where an overall positive effect of CI was recorded. On the basis of this earlier research we refined some of the CI instructions and incorporated a practice session in our design to encourage the children in this study to actively use the CI techniques. Before presenting the research, the methodological issues pertaining to the cognitive interview literature will be summarised. Previous Research on CI: Methodological Problems _ In earlier studies of CI (e.g. Geiselman et al., 1985) the experimenters set out to evaluate the effectiveness of the CI by comparing it with a 'Standard Interview' (SI). Usually these SI's were modelled upon the police investigative interviewing practices used by officers untrained in CI techniques. Some of the officers had never received interview training of any description. The success of the CI over the SI has often been attributed to its mnemonic components. However, in the absence of a suitable control group (that is, interviews identical to the CI save for its special techniques) it is not possible to say why the CI is more effective.

The scoring procedure (particularly the definition of a 'correct detail' and an 'incorrect' and 'confabulated' detail ) has not been fully discussed in any of the previous publications. Some studies report an increase in confabulations when CI is used. Köhnken and Brockmann (1988) report an increase in confabulations with CI, but it is not clear what constitutes a confabulated detail.

Finally, recent research suggests the CI technique results in substantial increases in correct information recalled with nine to ten year olds (Köhnken et al., 1992). However, in the same study there was a corresponding increase in confabulated details. In this particular study, the CI interviewers asked twice as many questions as the SI interviewers and errors fell largely into the person and action description category. This work suggests it is important to monitor the type of errors that may occur when CI techniques are applied.

The justification for including a 'try harder 'control group. The observed success of the CI may be due to relatively more recall attempts than are generated by the SI (cf. Scrivner and Safer, 1988). In addition, motivation may be higher in the CI groups merely because of the complex instructions they receive. A 'try harder' control group was therefore used in the studies presented below.

THE PILOT STUDY

The pilot study (Eaves, 1992) examined the recall- enhancing properties of each of the mnemonic components of the original CI by studying them individually, comparing each one with the other two and the control group. The data from this study suggested that when the mnemonic techniques of the cognitive interview are isolated and tested individually, their so called mnemonic effects are no more apparent than the motivating effects of the 'try harder' instruction. The design employed in the pilot study was retained in subsequent studies of CI in order to examine the effectiveness of the individual CI instructions (the change context, change perspective and change order) compared with a try harder control using larger samples of college students (experiment one) and children of different ages (experiment two). In each study interviewer differences were minimised by use of a fixed interview protocol in each condition.

EXPERIMENT ONE: THE TAUNTON COLLEGE STUDY

There were 68 participants in this study: all were biology students (aged 17 and 18 years) attending a local sixth form college. The participants were randomly distributed across conditions: try harder (N=14), context reinstatement (N=18), change perspective (N=17) and change order (N=19).

The staged event involved a postgraduate student from the university (a stranger to the class) who interrupted a lesson in order to request the students' help in completing a questionnaire. Interviews took place 10-14 days after the event. Each of three interviewers followed a clear interview protocol for each condition. This comprised an initial free recall in which subjects were specifically asked to give as much information as possible about the person and event, followed by one of the four instruction conditions: (i) Try harder; (ii) Change Perspective (iii) Reinstate Context; (iv) Change Order. Full details of the instructions are available from the first author. In each condition, the first recall consisted of a narrative account of the event. If, on completion, the participants had neglected to make any reference to a description of the man, of what he did, of what he said, or of what happened to him, they were prompted once for each of these elements. This ensured that neglecting to mention an element would not be misidentified as inability to recall anything about it

Coding and scoring of data. There were three coders and scorers. Each coder went through a sample of transcripts and assigned information to various categories. At this point accuracy of details were not assessed. Inter-coder discrepancies were discussed by working through all the codes given for each transcript and coding rules were modified once a consensus had been reached.

The coding scheme that was used was determined to some extent by the details of the to-be-remembered event. For example, the central feature of the event was the 'questionnaire' that participants were asked to complete. Thus we had a category where information about the questionnaire was placed. The other categories were: person descriptions, person action, location of persons and objects, temporal information and subjective information (information that was uncheckable including impressions of the event and person). Errors were coded as incorrect details (e.g. a blue shirt reported when the shirt was green) or confabulations (e.g. jacket reported when there was no jacket). Errors were coded and scored in the same way as correct details. For example, "he wore a green shirt and carried a black briefcase" was coded as a person descriptor with correct details assigned the following scores: one point for green and one point for shirt and errors scored as follows: one error score for black and one for briefcase.

Results.

A one way ANOVA showed no significant differences across conditions in free recall of correct or incorrect details (all F's<2). The number of errors in the free recall accounts were small (mean error rate =0.6%). All but two accounts were free from confabulations. A series of one way ANOVAs were then performed on the pooled accuracy data (i.e free recall score combined with additional details elicited following instructions). There were no significant differences between conditions across any of the comparisons (all Fs<1). As indicated in table one, the means tended to favour the 'try harder' control group with the context reinstatement condition producing the least amount of overall correct information. An examination of types of information elicited under each condition showed some interesting trends. The amount of location information differed across conditions F ( 3, 64)=7.71, p<.01, with more details about locations in the try harder as compared to context conditions (Fisher PLSD=.589, p<.05). Furthermore, there were more details in the change perspective conditions as compared to the change context conditions ( Fisher PLSD =.559, P<.05). The change order condition elicited more temporal details as compared to the change perspective condition (Fisher PLSD= .552, p<.05). These trends will be discussed in the conclusion section of the paper.

TABLE 1: Experiment 1: Mean number of details recalled in each condition


CONDITION

TRY HARDER
CONTEXT
PERSPECTIVE
ORDER
total correct
31.43
25.72
30.76
29.63
error
2.86
2.22
2.64
2.05
confabulation
0.21
0.17
0.06
0.05
action
8.29
8.22
9.35
9.84
description
7.14
6.17
6.41
7.12
location
1.00
0.28
1.41
0.32
questionnaire
8.21
4.16
5.35
4.89
temporal
1.79
2.17
2.23
3.79
other
5.00
4.72
5.94
3.68

There were no significant interviewer differences in total information, total correct information or errors (all Fs<1).

The main conclusion drawn from this study was that the CI instructions have no general effect over and above the motivational element to try harder. Since the interviewers correctly followed the interview protocols each time, such a finding cannot be attributed to the failure to instruct interviewees in the use of the technique. It may be the case however, that the interviewees were not following the instructions. When participants were asked informally at the end of the study which techniques they had used the only one mentioned was context reinstatement. It may be the case that our interviewees had not understood the instructions sufficiently. In the next study we included a 'practice session' to give interviewees an opportunity to try out the techniques. As discussed above, there is some evidence to suggest that a practice session may maximise any beneficial effects of the CI (Saywitz et al., 1992).

EXPERIMENT 2: BEVOIS STUDY

The participants were 68 children aged five to nine years divided on the basis of school year group. The youngest year group were five to six years (N=17), the intermediate group were six to seven years (N=19), and the oldest age group were eight to nine years (N=31). There were an approximately equal number of children in each of the four conditions.

The event took the form of two strangers taking part in a school assembly by reading and acting out a poem. The event lasted about 10 minutes and memory was tested approximately a week later. As in the previous study there were three interviewers (all were female) and an interview protocol was strictly adhered to by each interviewer. Each interview began with a practice session in which children were asked to describe what they did when they brushed their teeth in the morning. In most cases the children gave a brief account without any prompting. The interviewer then asked them to give a second account but this time it was preceded by the specific instruction to try harder, to change perspective, to change context or to change order. The younger children (five to six year olds) required more prompting in the last three conditions as they did not readily understand the instructions. Our observations during the practice session suggested that with guidance and prompting children became familiar and felt comfortable with the techniques. However, it was not possible to conclude from this that the children understood and could apply the techniques when asked to describe the event (this point is taken up in discussion).

The categories into which recall data could be placed were determined by the nature of the event (see Table two for a list of categories) and we followed the procedure for coding and scoring outlined earlier.

A three (age) by four (condition) analysis of variance on total correct information (free recall and cued recall data pooled) showed a significant main effect of age F(2,55)= 5.19, P<.01 with the eight year olds producing significantly more correct information than the five year olds (Scheffé's S= 6.23, P<.01). The intermediate age group (six and seven years) did not differ significantly from either of the other two groups (cf. Davies, Tarrant, & Flin, 1989). There were no significant differences in errors or accuracy (number of items correctly recalled as a proportion of total number of items) as a function of age or condition (all Fs <1). It is interesting to note that the younger children recalled less information but were no less accurate (cf. Poole and White, 1992, Saywitz et al., 1992).

An examination of age differences in total correct information as a function of information type showed significant effects for descriptive details F (2, 55)=3.57, p<.05. with the eight years olds producing more information within this category than the five year olds (Scheffe's S=4.55, P<.05). There was also a significant age effect on subjective information F (2,55)=4.43, p<.05 with the eight year olds producing more information in this category than the 5 year olds (Scheffé's S= 1.42, p<.05). Given the small frequency of errors within each category, the analysis of error type has to be interpreted with caution.

TABLE 2: Experiment 2: Mean number of details recalled in each condition


CONDITION

TRY HARDER
CONTEXT
PERSPECTIVE
ORDER
total correct
19.00
16.05
12.50
19.06
error
5.33
4.31
3.37
6.83
confabulation
0.67
0.00
0.68
0.22
action
1.73
1.37
1.12
0.89
description
13.6
11.63
9.00
13.78
location
0.07
0.21
0.12
0.11
questionnaire
1.80
1.63
1.12
1.33
temporal
0.67
0.21
0.00
0.33
other
1.00
1.00
1.00
1.61

There was no significant main effect of condition on total correct information F(3,55)=2.20, P>.05 or errors. There were significant differences between conditions in total information (total correct plus errors F (3, 64) =2.85, P<.05) with the try harder and change order conditions producing more information than the change perspective instruction (Fisher PLSD= 7.78 and 7.43 respectively, p<.05). However, more errors occurred in these conditions, although as indicated earlier there was no significant main effect of condition on error means (F(3,64)=2.33, p=.08). There were no significant interactions between age and condition on any of the measures (all Fs <1).

To summarise the data from Experiment two, there were no significant differences between the try harder control and mnemonic instruction conditions in total correct or incorrect information elicited. The results showed the youngest age group produced less correct information than the two older age groups but did not produce more errors and were no less accurate. In the absence of significant interactions between age and condition it is not possible to comment on age differences in performance under each condition. There are however, some notable trends in the means which warrant attention in further studies: (1) Inspection of the means for total correct information suggest the youngest age group perform less well under the context conditions as compared to the eight year olds which is consistent with the recent findings of Pipe et al., (1993) and (2) the youngest age group perform less well under the perspective conditions (cf Saywitz, et al., 1992) and we argue below that this may be due in part to the difficulty the children have understanding this instruction.

In the three studies there were no significant between group differences in the effects of CI instructions as compared to a try harder control group. This leads to the conclusion that the memory enhancing effects of CI reported in earlier studies may be due to increased interviewer and interviewee motivation to work harder. We are undertaking further research to explore this hypothesis in more depth. One factor relating to use of the CI that has been overlooked in earlier research is that it relies upon the co-operation of the interviewer and witness. The way in which the CI instructions are communicated and the extent to which the witness understands them is therefore critical in determining the success or failure of the procedure. For example, with respect to the change perspective instruction we found that interviewees were more likely to understand what we meant if we asked them to adopt a different location perspective rather than a different person perspective (cf. Saywitz et al., 1992).

The necessity to prompt with specific questions poses an additional problem in interpreting the effects of CI on recall performance: namely, that any increases in information obtained with CI instructions may be an artefact of the language that has to be used to fulfil the instructions given, and not a product of memory improvement. For example, in one study (Memon et al., 1992) we found the CI increased the reporting of details about the location of various people and objects (cf. Smith, 1979) as compared to a similar interview where the special CI instructions were not applied. It was also noted in one of the studies reported in the present paper (Experiment 2) that the 'change order' instruction tended to produce more information about the timing of events. Such details may of course be relevant in some cases (cf. Fisher and Geiselman, 1992).

The difficulty in communicating the CI instructions to the witness is one problem. Another is that it is not clear to what extent the success of CI resides in the motivation of the CI interviewer to produce a better interview and more information. The CI clearly takes more time and effort to perform than what has been described as a standard interview (George, 1992, Saywitz et al., 1992). The major way in which the CI components may differ from the 'try again' instruction is in the 'novel' way they request the interviewee to recall further information. Although more recent studies of CI (e.g. Köhnken et al., 1992) have tried to ensure their control groups received equivalent levels of training to cognitive interviewers the motivational element has, prior to our studies, not been sufficiently addressed in earlier studies of CI. Future studies of CI need to include a try harder group as an additional control group and compare this group's performance with that of the so called standard interview group.

Physical context cues and props may have powerful effects on recall (See Gee and Pipe, 1993 for a recent review). Yet, in the present studies we found no effects of context. Gee and Pipe argue that young children (six year olds) are less likely to use spontaneously generated cues such as imagery, and the re-enactment of an event using cues and props makes communication easier than merely reinstating the environmental context. They also show that the types of cues used and the method of presenting them are important determinants of the amount and accuracy of information reported. This reinforces the point that was made earlier about the communication of CI techniques.

Because of its potential application in the field of police interviewing, it is essential to scrutinise very carefully the experimental data related to CI. An ESRC funded study is currently underway to look more closely at the ways in which differences in interviewer style, interviewer non-verbal behaviour and witness motivation may influence the potency of any effects obtained with the CI.

Acknowledgements

We are grateful to the staff and pupils of Taunton College and Bevois Town School in Hampshire for their participation in this research. We are also grateful to Trudy Wright for her help with data collection and to Sigi Sporer for his comments on an earlier draft of the paper.