Ronald P. Fisher (1996) Misconceptions in Design and Analysis of. Psycoloquy: 7(35) Witness Memory (12)

Volume: 7 (next, prev) Issue: 35 (next, prev) Article: 12 (next prev first) Alternate versions: ASCII Summary
Topic:
Article:
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 7(35): Misconceptions in Design and Analysis of

MISCONCEPTIONS IN DESIGN AND ANALYSIS OF
RESEARCH WITH THE COGNITIVE INTERVIEW
Commentary on Memon & Stevenage on Witness-Memory

Ronald P. Fisher
Department of Psychology
Florida International University
North Miami, FL 33181, USA

FISHERR@SERVAX.FIU.EDU

Abstract

This commentary addresses two methodological and interpretational errors found in Memon & Stevenage's (1996) target article and some of the commentaries on the Cognitive Interview (CI). First, the Structured Interview is not an appropriate control group for the CI. Second, the CI does not reduce the accuracy of eyewitness recall, as was claimed; accuracy rate is as high or higher for the CI than for control interviews.

Keywords

Cognitive interview, errors, eyewitness memory, facilitated recall, police procedures, questioning, recovered memories, structured interview.

I. WHAT IS THE APPROPRIATE CONTROL GROUP?

1. One of two different control groups is typically used in Cognitive Interview (CI) research: either a Standard Interview or a Structured Interview. The Standard Interview is the interview procedure currently used by practitioners in a given target domain, for example, police investigators for eyewitness research. The Structured Interview, according to Memon & Stevenage (M&S) (1996), is a variant of the CI: it is the CI minus a few specific techniques (e.g., context reinstatement, request for detailed information).

2. What are the goals of the two control groups and how effectively do they accomplish their goals? The Standard Interview is used to test the purely applied goal of the CI: to increase witness recollection over and above what is elicited by police interviewers. Naturally, when the CI is applied to other domains of interviewing (e.g., accident investigation, epidemiology), the control group reflects the standard interview practice for that particular domain (Brock & Fisher, 1994; Fisher, Falkner, Trevissan, McCauley & Mello, 1996). As M&S (par. 12) state "practitioners ...need to know whether [the CI] produces gains relative to the techniques THEY CURRENTLY USE" (emphasis added).

3. The Structured Interview is not used by any particular group of interviewers for real-world investigations. As such, comparisons between it and the CI do not reveal whether the CI will be an improvement over the interviewing technique currently used by any target group of interviewers.

4. The rationale for using the Structured Interview, according to M&S, is that differences between it and the CI permit a theoretical analysis of the CI. M&S claim that the Structured Interview differs from the CI only in terms of the "cognitive retrieval strategies." Other non-retrieval aspects of the task (e.g., communication) presumably are equivalent for the CI and Structured Interview. Neither M&S nor any other researchers who use the Structured Interview provide any conceptual analysis to support this claim; they simply state it by fiat. Just the opposite, Fisher & Geiselman (1992) argue that many CI techniques (including those that differentiate between the CI and the Structured Interview) affect both retrieval and non-retrieval components of the task (e.g., knowledge representation, communication, social dynamics). At the present time, there is no compelling argument to support the claim that differences between the CI and the Structured Interview reflect any particular underlying psychological processes. As such, use of the Structured Interview does not lead to any theoretical insights into the workings of the CI.

5. I conclude, therefore, that because the Structured Interview (a) is not used by any specific target group of interviewers in the real world (par. 3) and (b) provides no theoretical insights into the workings of the CI (par. 4), it is not an appropriate control. One needs to create a control group that is either (a) better grounded in a rigorous theoretical analysis of the CI, or (b) a technique that is currently used by practitioners in the target domain of interviewing (Kebbell & Wagstaff, 1996).

II. RECALL ACCURACY

6. The claim has been made in the target article (M&S) and repeated elsewhere (Higham & Roberts, 1996) that the CI leads to less accurate recollection than do the Structured or Standard Interviews. As a consequence of this unfortunate error, doubts have been raised about the CI's use in real-world investigations.

7. These claims are based on an incorrect analysis of the results, namely, looking at the wrong statistic (number of incorrect recollections). M&S correctly state that, at least for some recent tests, the revised CI (Fisher & Geiselman, 1992) elicits more errors than does the comparison technique (although note Geiselman, 1996). This is true; it is simply irrelevant for measuring recall accuracy.

8. The concept of recall accuracy is generally taken to mean something like: To what degree does the witness's recollection correspond to what actually happened? Recall accuracy is conventionally measured as the number of errors DIVIDED BY THE TOTAL NUMBER OF RESPONSES (Koriat & Goldsmith, 1994). This normalizing procedure, dividing by total number of responses, is implemented to equate for differences in total number of responses that are generated across different conditions. Identical normalizing functions exist in almost all domains. In signal detection theory, for example, proneness to error is measured as the False Alarm RATE (number of false alarms divided by the total number of catch trials). A car's efficiency is measured as miles PER gallon; a baseball player's effectiveness as number of hits PER total number of at bats; a country's illiteracy as the number of illiterates PER total population. Recollection accuracy is no different than any of these other measures: error RATE (number of errors divided by total number of responses) is the appropriate measure.

9. Using the absolute number of errors instead of the error rate to measure recall accuracy leads to the bizarre conclusion that an interview that generates 98 correct and 2 incorrect responses is more error-prone than an interview that generates 1 correct response and 1 error (Geiselman, 1996). Is anyone truly willing to make that claim? If there is any doubt, note that this logic would lead to the conclusion that a Volkswagen that used 2 gallons of fuel to travel 100 miles was less economical than a Sherman tank that used 1 gallon to travel 1 mile.

10. Imagine that two people witness a crime, and that each can provide 100 facts at 90% accuracy. Interviewing both witnesses will generate more errors (20) than interviewing only one (10). But certainly, no one would refuse to interview the second witness simply because the total number of errors will increase. The second witness's set of 100 facts is just as informative as the first witness's set of 100 facts, as both are generated with the same degree of accuracy. And it is obviously better to have 200 facts at 90% accuracy than 100 facts at 90% accuracy. The analogy to the CI is apparent: it elicits additional facts, but at the same accuracy rate. Anyone who is concerned about the additional (absolute number of) errors when using the CI would have to argue that it is preferable to interview only one of the two witnesses. Note that the logical extension of this argument is not to interview anyone at all for fear of eliciting some incorrect responses. Clearly, accuracy RATE, and not absolute number of errors is the appropriate statistic.

11. As M&S correctly note, when we examine accuracy rate, the CI is equivalent or slightly better than Standard or Structured Interviews. This pattern holds for virtually all of the published research (see the meta-analysis by Kohnken, Milne, Memon & Bull, 1994). We can safely conclude therefore that there is no cause for concern that the CI generates unduly low accuracy, or at least, lower accuracy than any other tested interview method.

12. The above argument is based on studies conducted in the laboratory. How well does it hold for interviews conducted with victims and witnesses of real crime? In one field study, the accuracy rate (as estimated from inter-witness corroboration) with standard police interviews was high (.930); with the CI accuracy was even (non-significantly) higher (.945) (Fisher, Geiselman, & Amador, 1989).

13. Simply stated, in comparison to other conventional interviewing methods, whether in the laboratory or in the field, the CI elicits more information at the same or slightly higher accuracy rates.

REFERENCES

Brock, P. & Fisher, R. (1994, March). Effectiveness of the cognitive interview in a multiple-testing situation. Paper presented at the American Psychology-Law Society, Santa Fe, NM.

Fisher, R. Falkner, K., Trevissan, M., McCauley, M. & Mello, E. (1996, June). Cognition and long-term recall of physical activity. Paper presented at the Society for Epidemiological Research, Boston.

Fisher, R.P. & Geiselman, R.E. (1992). Memory-enhancing techniques for investigative interviewing: The cognitive interview. Springfield, IL: Charles C. Thomas.

Fisher, R.P., Geiselman, R.E. & Amador, M. (1989). Field test of the cognitive interview: Enhancing the recollection of actual victims and witnesses of crime. JOURNAL OF APPLIED PSYCHOLOGY, 74, 722-727.

Geiselman, R.E. (1996). On the Use and Efficacy of the Cognitive Interview. PSYCOLOQUY 7(11) witness-memory.2.geiselman.

Higham, P.A. & Roberts, W.T. (1996). Analyzing States of Consciousness during Retrieval as a way to Improve the Cognitive Interview. PSYCOLOQUY 7(17) witness-memory.4.higham.

Kebbell, M.R. & Wagstaff, G.F. (1996). Enhancing the Practicality of the Cognitive Interview in Forensic Situations. PSYCOLOQUY 7(16) witness-memory.3.kebbell.

Kohnken, G., Milne, R., Memon, A. & Bull, R. (1994, March). Recall in Cognitive Interviews and Standard Interviews: A meta-analysis. Paper presented at the American Psychology-Law Society, Santa Fe, NM.

Koriat, A. & Goldsmith, M. (1994). Memory in Naturalistic and Laboratory Contexts: Distinguishing the Accuracy-oriented and Quantity-oriented Approaches to Memory Assessment. JOURNAL OF EXPERIMENTAL PSYCHOLOGY: GENERAL, 123, 297-315.

Memon, A. & Stevenage, S.V. (1996). Interviewing Witnesses: What Works and What Doesn't? PSYCOLOQUY 7(6) witness-memory.1.memon.


Volume: 7 (next, prev) Issue: 35 (next, prev) Article: 12 (next prev first) Alternate versions: ASCII Summary
Topic:
Article: