View in Plain Text


How do computerized reports fit into custody examinations, reports, and testimony?


Alex B. Caldwell, Ph.D.


How do computerized reports fit into the examination process?


The computer-generated MMPI-2 report basically describes the patterns of behavior that are characteristic of those who obtain similar profiles. What reactions, what sensitivities, what internal issues, what external interpersonal conflicts, etc., are likely? This is, of course, actuarial hypothesis generation: it alerts the clinician to what to look for, perhaps what to give weight to even if the examinee minimizes the problem. For example, it may alert the clinician as to what may be presented as a relatively superficial problem that could cover over other more uncomfortable issues.


All such statements are probabilistic even though it is not possible to set universal numerical probability values on each statement. Such values would fluctuate far too much when the MMPI-2 is used with quite different populations. My own solution to this is simply to graduate the overall level with such phrases as, "in most cases," "in many cases," "in some cases," "in a few cases," etc. An example of "a few cases" would be covered over paranoid trends in a profile that is not usually marked by paranoid thinking, but there are signs that this may be an exception. Thus the clinician is alerted to take note if clinically there are such signs.


Distinguishing the Actuarial Function from the Clinical Function The actuarial task is to offer relative baselines for various behaviors. The item responses are entered into a complex computer program that is unaffected and unbiased by any information about the issues being considered or by any information gained by the examiner. The output then becomes an array of hypotheses to which the examiner may want to attend. If a particular report statement happens not to fit and one needs to explain the probabilistic nature of "actuarial" to the court, then one might use the following as an example. The actuarial function is like a professional actuary tabulating the driving records of adolescents with versus without driver training. Everyone can recognize that there are exceptions (i.e., wrong predictions) - some with driver training are still poor drivers, and some without are nevertheless good drivers. But the point is that, on the average, one group has a different record from the other, and the size of this predictable difference becomes an element in setting their insurance rates. That an individual prediction in an actuarial MMPI-2 report does not fit does not take anything away from the fit of the other predictions, given that a good preponderance do fit. So far, Meehl’s prediction has proven right: the whole body of research on statistical vs. clinical prediction remains an amazing 100% in favor of statistical/actuarial predictions as equal to or exceeding clinical judgment. The clinical function is to accumulate all the available information that one can obtain that is relevant to the determination to be made by the trier of fact. An important part of this can be the testing of the hypotheses based on what has been observed with similar MMPI-2 results. The probabilities are hardly 1.0, so to become practically meaningful they must be verified via interviews, observations, other records, etc. This process is, of course, vulnerable to accusations of bias and selectivity. But as noted above, the actuarial predictions are generated solely from the individual’s item responses and such demographics as age, gender, marital status, or years of education: the computer-generated actuarial characteristics cannot be biased by any clinical information about the person. Thus, whenever the objective predictions are clinically documented to be accurate, they clearly were not originated by observer bias; this strongly supports the objectivity of the examiner. One of my refrains is that to focus on the convergence of the clinical and the actuarial data can enable the most clearly objective and least challengeably biased presentation of one’s opinions and recommendations. A friend recently had an opposing psychologist witness assert that he had no need for nor use of computer-generated reports because "they do not take the person’s circumstances into account." This is as sensible as saying that a car was poorly designed because it cannot fly, i.e., that it cannot do something for which it was in no way designed. This deceiving dodge was straightforwardly explained to the court.


How are computer-generated reports to be used in the preparation of a custody examination report? How the information from the computer-generated or actuarial report (CBTI or Computer Based Test Interpretation) is to be integrated into the final report is ambiguous. Specific rules or even recommendations have never been formally specified, and consequently individual practices vary considerably. I am not presently aware of any courts having taken any precedent-setting actions on this issue. I only occasionally see the final clinical reports that have made significant use of my (or other) CBTI reports, so my comments are in part based on feedback from practicing examiners. In my awareness, this use has ranged from paraphrasing, to copying a few words, to entering whole intact paragraphs, or to appending the entire CBTI. I do have a concern that copying extended passages, especially whole paragraphs or more with no recognition of the source, might be found misleading by members of the court. I am aware that some professionals may not want to indicate the CBTI source out of concern that they will be forced to produce the entire CBTI report. There may be statements in it which they do not want to be forced to explain or defend, e.g., a serious diagnosis listed in my "Diagnostic Impressions" section or a diagnostically serious discussion elsewhere in my report (see discussion below). Ideally, this should largely be a false fear as I will discuss, but aggressive cross examining attorneys can find ways to make what should be straightforward become tortuous if not torturous. I believe my responsibility is to provide the most accurate, complete, and useful test analyses that I can. The refinement of the material in my interpretive system has proven a lifelong task. For many obvious reasons it is not realistic or even possible for me to "police" how my reports get used beyond doing my utmost to be sure that the clients to whom we send reports are appropriately licensed professionals. The policing of abusive uses must be done by state professional ethics agencies, the A.P.A., or the courts in which they appear.


How can one deal with strong clinical and diagnostic statements? The issue of strong diagnostic statements and possibly formal diagnostic entries with serious implications in CBTIs merits specific comment. In part, their presence in our narrative reports reflects the predominance of clinical cases (often psychiatric inpatients) in the evolution of MMPI and MMPI-2 interpretation as well as being the MMPI’s strongest historical area of application. Much of the original interpretive data came from such settings, although the test has been used with hundreds of thousands of individuals in non-clinical settings. In the custody examination context, for example, such diagnostic CBTI content can best be understood as a possible vulnerability. That is, it is reasonably interpreted as reflecting trends or outside potentials in the person’s makeup. With relatively unelevated profiles, this is essentially the assertion that if more adversities were to befall the person and the person’s life were to go seriously downhill, the categories mentioned would be the most likely "summary labels" as to the direction(s) in which the person’s deteriorating emotional state would evolve and be seen. Assigning a formal diagnosis is not an actuarial function; a formal diagnosis is a clinical opinion based on a hopefully wide range of information. In treatment settings, the actuarial function is to contribute to differential diagnosis by alerting the clinician as to what labels are most commonly associated with psychotherapeutic clients and psychiatric patients who obtain similar patterns on the test (see Caldwell, 1996). It can sincerely be the custody examiner’s straightforward opinion that the multiple requirements for making such a diagnosis are not met in this immediate instance--it could be possible if everything got a lot worse and fell apart for him/her, but such a more extreme state is not now the case. On this basis it would be quite legitimate to dismiss the possible identified diagnoses as largely or even entirely irrelevant for this person at this point in time. Note that whenever the pattern is within the normal range, my reports explicitly state that fact. The diagnostic statement then almost always starts with such phrasing as, "Among psychotherapy patients . . ." and is followed by a statement that the normal range profile may reflect no more than an essentially normal personality or else a situational adjustment reaction (overall a large majority of subjects who obtain normal range profiles are indeed functioning individuals). In some cases with atypical or highly defensive profiles, an additional normality-qualifying statement may comment that the profile is within the normal range but more ambiguous than most because of the degree of defensiveness. This latter in part recognizes the fact that a significant minority of psychotherapeutic client profiles (including psychiatric inpatient profiles) are nevertheless within the normal range (denial and defensiveness, milder problems that benefit from working through, lack of self-awareness, etc.).


How does the CBTI connect to the concluding opinion? Damaging one’s credibility through attributions of bias is, of course, not an infrequent effort in adversarial custody examination proceedings. My belief here is that the use of CBTIs as non-case-biased sources of information can be very helpful in anchoring one’s objectivity and credibility. By emphasizing the hypothesis-generating or "alerting" function of the CBTI as to what are likely to be problematic issues for each of the litigants as parents and in relation to each other, the examiner can start from an uninfluenced and objective basis from which to develop recommendations. The available MMPI-2 interpretive data are not readily organized for searching in depth on a codetype-by-codetype basis (beyond textbook summaries). It would take many hours for the clinician to make a thorough search of the data sources for each profile considered, and the clinician’s own search itself might be made to look selective or biased. By pointing out that MMPI-2 interpretation is a very complex undertaking, it then becomes quite reasonable to the trier of fact for the examiner to consult an expert who has spent his career working on the task. Using the Caldwell Report Custody Report (the interpersonal implications of the MMPI-2 test results) and the Caldwell Report Narrative Report (the intrapsychic processes of each individual) clearly conforms to the nature of an expert consultation. In summary, I believe the direct discussion of the convergence of the clinical data with the actuarially-generated hypotheses can add a strong element of objectivity and logical flow to the process of exploring the particular person’s characteristics as a parent as well as maritally if not more generally. My impression is that the courts typically find this objective anchor to lead to substantial increments in the credibility of the opinions and recommendations provided.




Caldwell, A. B. (1996). Forensic questions and answers on the MMPI/MMPI-2. Los Angeles: Caldwell Report.