INTRODUCTION FOR COLLEAGUES
TO THE ANALYTIC PROCESS SCALES (APS)
by Sherwood Waldron Jr., Robert D. Scharf, Stephen K. Firestein, Marianne Goldberger & Anna Burton
As psychoanalytic psychotherapists, we often see with our own patients the extensive benefits of a carefully applied psychoanalytic approach to patients' difficulties. Not only are symptoms alleviated, but the quality of work and personal relationships is affected favorably, often for a lifetime. Our clinical impressions are confirmed by the only comprehensive scientific study of psychoanalytic treatments available: in the Menninger Psychotherapy Research Project more than 80% of the patients had improved by 2 years after their intensive treatment. Of these, there was an average improvement in quality-of-life of 37%, based upon the most comprehensive and well established measure of health-sickness available!(1)
However, these impressive results do not demonstrate directly a relationship between the skill of the psychoanalytic approach, the patient's contribution to the work, and benefit. In other words, the links between processes and outcomes have not been established.
The most daunting difficulty in establishing the links between processes of psychodynamic therapy and outcome has been that of evaluating the nature and quality of the observed process, in a way that reflects central clinical dimensions. Borkovec and Miranda pointed out in a review recently published by NIMH, "Despite initial attempts for some types of therapy, there is no valid way to measure quality for any therapy technique."(2) This situation reflects a number of weaknesses in methodology in studies attempting to explore the efficacy of psychoanalytic treatment . Past studies have often examined analytical samples too brief for deep understanding, have drawn upon the work of inexperienced clinicians, or have used inexperienced analysts, or non-analysts, to evaluate the data. Their rating scales and the psychoanalytic concepts underlying them tended to be too abstract, oversimplifed or ambiguous. And such scales as have been attempted have usually not been developed with extended input from experienced psychoanalytic clinicians, to ensure the final scales reflect closely the clinical understanding of the process that they were intended to measure.
We started our research group in 1985 in order to develop reliable methods for systematic studies designed to avoid the problems encountered by previous investigators. We study only the work of highly experienced analysts and use highly experienced raters, choose central, unambiguous, experience-near process features of both patient and analyst, and we define our variables in the language of the clinical surface. Careful study of individual cases has been our approach throughout. Recognizing that this could be a long and possibly fruitless undertaking; we nevertheless felt that its potential for effecting dramatic improvements in clinical practice far outweighs the demands on our time and the risk of failure.
Our research group is directed by Sherwood Waldron Jr., Robert D. Scharf, associate director, Stephen K. Firestein, and Anna Burton. Each is a full-time practitioner with more than twenty-five years of clinical experience. In developing our instrument and methodology, we have used recorded cases provided by Lester Luborsky of the Philadelphia analytic study group, whose work presaged our own, as well as sessions from the case of Mrs. C., a psychoanalytic treatment that has been widely studied by other researchers. We have also studied recorded analyses from the archives of the Psychoanalytic Research Consortium, a nonprofit organization under our direction, to collect such recordings, preserve them under protected conditions, and make them available to qualified researchers.
FEATURES OF THE ANALYTIC PROCESS SCALES (APS)
We constructed variables to measure central psychoanalytic features of patients' communications and analysts' interventions.(3) Ratings are only performed after acquiring familiarity with the case by studying several sessions immediately before the one to be rated. Each feature is rated on a five-point scale in which zero reflects the absence of a feature and four its strong presence. Fourteen patient variables assess the degree to which patients' contributions help the analyst understand them, the degree to which patients are self reflective and express feelings, both in regard to the analyst and analytic situation, and in regard to the rest of their lives. The degree to which patients identify their own contribution to their problems in life is assessed. Developmental references and expression of problems with self esteem are rated for the patients, while parallel variables assess the degree to which the analyst approaches the same issues. Similarly, there are several variables which assess separately the quality of work accomplished by the patient and by the analyst. In addition, seven of the eighteen analyst scales allow us to evaluate what we call core analytic activity, including the types of interventions made, and the degree to which interventions address the patient's conflicts, reactions to the analyst, and resistances. Although our principal outlook is the conflict model of psychoanalytic theory, we have tried to make the APS as broad in its conceptualization as possible: for example, we evaluate the analyst's approach to problems of self-esteem, a particular concern of self psychology. (See pictographs in the results section below and Appendix 1 for detailed description of the APS rating scales.)
In order to capture the back-and-forth flow of the process between patient and analyst, we divide sessions into segments (principally defined by a significant change of speaker or psychological topic) and rate each of these individually. The segmenting process has been reliable, once segmenters study our segmenting instructions. We are currently developing a full segmenting manual with examples so that researchers in other centers may reliably apply our segmenting process to their own materials. Most of our segments consist of patient speech, rated on the 14 patient variables, or analyst speech, rated on the 18 analyst variables. There are some segments, where the back-and-forth exchange between patient and analyst is rapid, which are designated joint segments, and rated on both patient and analyst scales. Applying these scales to successive patient and analyst segments throughout a session affords a special opportunity to study the interactional aspects of the work, specifically the relationships between what the patient communicates, how the analyst intervenes and how the patient responds to the intervention. We can determine which qualities of the patient process are enhanced by the various interventions over a series of segments, and how the patient's responsiveness and productivity serve to shape the analyst's interventions. Accurate rating of these typical psychoanalytic interactions has scarcely been accomplished by previous investigators. Because our scales focus on issues which analysts think about and use in their daily clinical work, they have a strong face validity; that is, expert colleagues who have studied our instrument believe that the definitions of the clinical features we use validly represent what they claim to describe. Almost all previous measures of therapy have been much less closely related to clinical psychoanalytic concepts. They have been more ambiguous, less carefully chosen, and less persuasively validated.
Perhaps the most substantial contribution of our research group has been the APS Coding Manual, a seventy-six page document in which each variable is defined and illustrated. Every illustration consists of a short unfolding clinical example from a single patient, showing manifestations of the variable at the 0, 2, and 4 levels, and is designed to be easy to read and remember. The development of the manual has entailed years of examining the scoring differences between our group members on each variable for each segment of many rated hours in order to clarify the meanings of the concepts we use, develop ordinary language to convey them, and shape the clinical illustrations to match our intentions for each level of manifestation of each variable. As a result, the inter-rater reliability of our scores has steadily improved, so that we have now achieved an entirely satisfactory level of reliability on virtually all of our scales--a landmark accomplishment (see Appendix 2 for description and list of reliability scores for each variable). Furthermore, we have found that experienced analysts can achieve this same reliability without extensive training, simply by studying the manual as they perform ratings.
The relationship between process and outcome is the single most important aspect of psychoanalytic and psychotherapeutic work in assessing its efficacy and consequent social value, yet the one that has been the least well researched because of its technical difficulty. Our variables have an outcome dimension built in, since they measure how well the patient is functioning in the work with the analyst throughout the treatment. Positive changes in functioning, directly observed in the analytic situation, may be expected to reflect improved functioning in general, although the degree of relationship has not previously been established. We plan, in addition, to apply several other existing measures of psychological function to the recorded material from the beginning and end of treatment. These outcome assessments are being developed by a subgroup of the Collaborative Analytic Multisite Program. These investigators, Harold Koenigsburg, Lee Brauer, Laura Pontrelli and Jennifer Stuart, have already obtained reliability in assessing outcome, using well established measures applied to recorded material.
The APS is intended to serve as a narrow-angle focused instrument to study processes of psychoanalytic work segment by segment in sessions. We are also using the APS to evaluate the overall process by scoring each entire session as a unit. This wider-angle focus permits study of a much more extensive sample of sessions for the same investment of rater time. We have already found preliminarily a good degree of reliability among junior clinicians using the instrument in this fashion, so further investigation of whole-session evaluations is warranted. Then we can determine how reliable the APS is depending upon the level of experience of the rater, and the degree of detailed versus more global assessment. This will provide the basis for establishing the range of studies in which the APS can be useful.
PRELIMINARY RESULTS USING THE APS
The results presented here are preliminary, since they come from sessions studied over the course of the years of the development of the APS manual, and were themselves the basis of many changes in the manual. Nonetheless, many of the findings accord with our clinical impressions, and may be substantiated by future studies.
Several trained raters each scored the APS on early, middle, and late sessions for two patients described in this paper as AA and CC, and two late sessions for a third patient, BB. The averages of these scores are what we examine here. Brief descriptions of the patients and their treatments will help to make the subsequent discussion clearer.
Patient AA was a woman with agoraphobia who did quite well symptomatically in the course of her three hundred-plus hour analysis, and continued her treatment twice weekly as termination approached. While clinically she was much improved, her analyst felt that further work would have been beneficial. Our raters agreed, and thought that the interaction between AA and her analyst was negatively influenced by the analyst's imposing presence, towards which she seemed unusually compliant, and upon the analyst's emphasis on analyzing transference in ways which were often not meaningful to the patient.
Patient BB was a married man who experienced some relationship problems, and had left a previous long analysis because he felt too distant from the first analyst. The two-year treatment of just under 400 hours with the second analyst touched on important dynamics and themes of the ways he kept himself at a distance, and resulted in positive changes in his sessions and life.
Patient CC had more serious relationship difficulties, including staying distant from women. An initially positive reaction to the present analyst gradually gave way to a growing sense of alienation and a feeling that he was not being understood. The analyst seemed to become increasingly frustrated, and more or less blamed CC for his failing marriage and the faltering treatment, which ended after about 660 hours. As you will see, the analyst received high scores for confrontation and on how much feeling he expressed. Our impression was that the confrontation became increasingly hostile.
1. Central aspects of the treatment process.
What central aspects serve to characterize the work by patient and analyst? In order to clarify and simplify our analysis of the data, it was helpful to group the twenty-five variables obtained from the APS rating scales into just a few broad categories without losing much of their distinctive richness.
Factor analysis is a method of finding out how the scores on individual variables (such as our patient and analyst variables) move up and down together. We factor-analyzed the data for each patient-analyst pair separately, and found a substantial overlap in the factors from one patient-analyst pair to the next. The study of these factors helped us to group the variables into three larger clusters for the patient and three for the analyst. These clusters, described in the two pictographs on the next two pages, characterize global aspects of the treatment process, at least in this sample.
Patient's Work Quality includes the APS patient variables measuring overall productivity, the psychological continuity of communications with previous issues, and the response to the analyst's interventions. Patient Participation-Collaboration includes the variables which measure the clarity with which experiences of conflict are conveyed, the degree of conveyed feelings, self-reflection, awareness of contributing to one's own suffering, references to development, and communication of issues involving self-esteem problems. Transference Expression includes the three variables which assess how well the patient conveys experiences, expresses feelings, and self-reflects, all specifically about the analyst and the analytic situation.
The analyst clusters are parallel to the three patient clusters. The first, Intervention Quality, is composed of only two variables: how well the analyst follows the patient's productions, and overall goodness of intervention. The second analyst cluster is called Core Analytic Activity. This includes variables measuring the degree to which the analyst clarifies and interprets, and how much the analyst focuses on conflict, resistance, and transference. Also included is a focus on self esteem issues, and the expression of a developmental perspective. The third analyst cluster is Affective Involvement: how confronting, expressive of feeling and supportive is the analyst.(4)
THE ANALYST PROCESS SCALES (APS)
clustered patient variables
To what degree does the patient:
Since this preliminary work, two patient variables have been added. Patient's expression of feeling is further subdivided into a variable assessing expression of romantic or sexual issues, and another variable assessing the expression of aggressive or hostile issues.
THE ANALYST PROCESS SCALES (APS)
clustered analyst variables
To what degree does the analyst:
Since this preliminary work, four analyst variables have been added. The analyst addressing conflict is further divided into the analyst addressing 1) romantic or sexual issues, or 2) aggressive or hostile ones. The analyst expressing feelings is further divided into 1) amicable feelings and 2) hostile feelings. These additional variables permit specification of fluctuations between patient and analyst as well as specification of the focus of analytic work, in relation to the patient expressing romantic or sexual concerns, versus aggressive or hostile ones.
The patient and analyst quality clusters are obviously similar and may be considered as a pair. There is also a clear parallel between what may be called patient and analyst participation clusters -- Participation-Collaboration for the patient and Core Analytic Activity for the analyst. These two clusters then make a second pair. Finally the patient's Transference Expression cluster is related conceptually to the analyst's Affective Involvement cluster, in that each cluster reflects specific aspects of the degree of manifest involvement with the other participant. These can be paired as the two involvement clusters. We will see later on that the scores for some of these pairs of patient and analyst clusters tend to vary together, session by session, suggesting that the descriptive similarities between them correspond to deeper relationships. As may be seen from this description, we have accomplished the goal, on a preliminary basis, of reducing our data to a simpler structure.
2. Differences in the work of each patient-analyst pair.
What is the relationship between the nature and quality of the work by one patient-analyst pair as compared to the work by other patient-analyst pairs? In other words, how consistent is the work by each patient-analyst pair, in contrast to that of others, across the phases of treatment? Figure 1 presents patient and analyst quality scores from each session of each patient-analyst pair. The first three groups of bars represent the average of the segment scores from an early, middle, and late session for patient AA. Then follow the average scores for two successive sessions from late in the treatment of patient BB, whose earlier hours were not recorded. The last three groups are average quality scores from an early, middle, and late session from patient CC. The patient clusters have been placed side by side with the analyst clusters: the first, lighter bar of each group represents the patient's mean score; the second, darker bar the analyst's.
This chart indicates that the patterns are distinctive for each patient-analyst pair, and an analysis of variance confirms this. The quality of patient AA's activity steadily increases to a level about fifty percent higher than her initial level, while her analyst intervention quality is at about the same level throughout. Because our data for patient BB is limited to two late sessions, we can observe only that the levels for the patient and analyst are both significantly higher than the patient and analyst levels for the other two patients, an observation which is consistent with our clinical impression that this was generally the most productive of the three treatments. Not surprisingly,
the picture for the rather troubled analysis of patient CC is quite different. The quality of work by both patient and analyst declines by approximately half from the initial values to the end values (5).
The individual variable scores for each patient-analyst pair demonstrate the discriminating power of the APS: for almost every variable, the levels achieved by both patient and analyst are different for each pair from one or both of the other pairs(6). Thus the different scores distinguish between treatments on central psychoanalytic dimensions(7).
Turning to an examination of specific aspects of the scores obtained by each patient-analyst pair, we find an interesting correlation(8) between types of intervention and the quality of the treatment as a whole. We have already described how the quality of the interventions by BB in this sample was substantially higher than for AA, which in turn was substantially higher than the average level for CC. Nearly half the interventions for pair BB have at least some score for both interpretation and clarification, whereas this is true for less than a quarter of the late interventions for pair AA, and of none of the late interventions for pair CC. In other words, interpretations were given with much more of a clarifying component when the case went well than when it went less well. This in turn might well reflect the way the analysts engage patients in a collaborative look at what was going on in more successful treatments, sharing with the patient the basis for conclusions drawn.
Our second finding relates to the aims of the analyst's intervention. Interventions addressing transferential aspects were common: 88% of late interventions for BB, 78% for AA and 40% for CC had at least some substantial component of addressing transference. This demonstrates that these 3 processes fulfilled one of Freud's two criteria in regard to what constitutes a psychoanalytic approach. However, when we examine the degree of analysis of resistance, which is Freud's other criterion, and which we define as approaching defensive aspects of the patients' reactions within the session, we find very large differences between cases. When the case was going better there was much more analysis of resistance than when it was going less well: 64% of late interventions for BB had at least some component of addressing defenses, compared to 28% for AA and none for CC, the case which did the least well on other grounds. The striking differences in whether resistances were addressed demonstrates how the aim of a given analyst can be quite different from that of other analysts, in the broad framework of a psychoanalytic enterprise.
3. Changes in functioning as treatment progresses.
. Are there changes from the beginning, to the middle, and to the end of treatment in the patient-analyst pairs? In answering this question, our data are limited to two cases, AA and CC. The changes in our variables characterize different patterns of patient-analyst work, and serve as outcome measures.
Initially, the AA and CC patient-analyst pairs were at very similar levels on almost all of the variables. However, in the middle phase, AA significantly outperformed CC on all measures of patient participation and quality. Finally, the scores for the late phase indicated that AA had improved in many ways, CC declined, and BB was generally considerably higher than both on many (but not all) variables.
Using many of our process variables to assess outcome will be an important application of the APS. We expect that the patient quality cluster, for instance, will reflect the patient's general level of coping and adaptation. Most of the patient participation-collaboration variables -- how well patients convey their inner experiences, how usefully self reflective they are, how much their expressed feelings inform their communications, how much they acknowledge and "own" their personal problems -- are all useful emotional aspects which we would hope to see functioning well as a result of treatment.(9) Self reflection and awareness of one's own problems relate as well to Fonagy and Target's concept of "mentalization", a quality with powerful prognostic significance. Similarly, the patient transference variables measure a person's ability to participate in a close cooperative relationship, and to tolerate intense immediate emotions.
There are two different ways in which the scores on our patient variables might be used to assess outcome. First, we can compare early scores to late scores and observe how much they changed. For example, patient AA's scores definitely improved, indicating that she benefited from treatment, whereas CC's scores showed a clear decline, reflecting that patient's unsatisfactory outcome (if the trends we found were confirmed in a larger sample). Secondly, we could similarly assess outcomes for specific problems (analogous to target symptoms) which the patient was experiencing at the start of treatment. For instance, a high late score on expressing feelings in analysis would represent significant improvement for a patient like BB, who initially had trouble expressing his feelings.
We are also interested in characterizing changes in the analyst's work with patients during the course of treatment. When we compare each analyst to himself at the early, middle and late phases, we discover noticeable differences in the late session. AA's analyst scored higher for analyzing defenses and conflict and lower for supportiveness in the late session than in the first two sessions; nevertheless, his scores on these variables were never very high at any phase of treatment. The best session for CC's analyst was clearly the early one. Compared to the middle and late hours, he made higher quality interventions. In the late session, his irritation with the patient was manifested in elevated scores for confrontation and the expression of feelings. Changes in the analyst's ways of working as treatments progress may turn out to be important indicators of psychoanalytic processes which are developing constructively, are stalled or are deteriorating. Also changes in the analyst's way of working over time may also turn out to indicate qualities of flexibility in analysts who are more capable, and consequently obtain better results.
4. The way interventions contribute to the patient's work, and vice-versa.
What is the relationship between the analysts' interventions and the patient's immediate responses to them? Because the APS provides scores on segments as each session unfolds, we can study the immediate effects of interventions on the patient's activity. We all realize, of course, that interventions often have delayed effects, and may not become effective unless the analyst adequately follows them up when the patient's resistances have become more evident. Nevertheless, it is reasonable to expect that, over a series of instances, the analyst's interventions often have an immediate impact.
Detecting patterns in a time series is a highly specialized kind of investigation, and one that is subject to various errors of interpretation. With this in mind, we have enlisted the aid of Patrick Shrout, who is a statistician and head of the Department of Psychology at NYU. He has helped us to analyze the data using a fairly simple approach which we hope produces valid results. We calculate the correlation between analyst scores on all our variables and the immediately following patient scores in order to arrive at an estimate of the impact of the analyst's intervention on the patient's productions. As we are equally interested in the degree to which the patient influences the analyst, we then calculate the correlation between the patient's productions and the analyst's subsequent intervention. These calculations need to be carried out patient by patient; otherwise, false conclusions may be reached.
Our analysis of our preliminary data yielded many correlations signifying an appreciable effect of one party on the other.(10) In general, we find a pattern of mutual interaction and mutual enhancement for all three patient-analyst pairs. Patient productivity usually went up when the analyst addressed the patient's conflicts. Also, when the patient was more productive, the analyst was able to address the patient's conflicts to a greater degree. There were also many differences in the relationships observed between the pairs, so that this method of analysis of the data served as well to describe individual ways of working. For example, patient BB has been described as tending to intellectualize and keep some distance from his feelings. However, when his analyst addressed conflict and resistance, this patient was much more likely to become expressive of feeling, although not about the analyst specifically. The other two patients responded to interpretations of the transference by becoming more expressive of feeling about the analyst, but not patient BB. We will see in the next section how this characteristic avoidance by patient BB was also demonstrated by graphing his segment-by-segment responses. Patient AA also showed a finding specifically characteristic for her: she showed the greatest expression of, and response to self-esteem issues, and a number of measures of her productivity and participation went up or down depending on the degree to which her analyst addressed her self esteem concerns. This was a patient who suffered from anticipatory shame whenever she would go out, fearing that she would have to run to the bathroom. She clearly displayed more self-esteem problems than the other two at the outset of treatment.
The BB patient-analyst pair, which was the most productive analytically of the three, showed an appreciable positive correlation between the nature of the analyst's interventions and the productivity of the patient's responses for five of the thirteen analyst variables. There was a strong correlation between the nature and quality of the intervention and BB's subsequent productivity: the higher the quality of the analyst's remark, the more productive the patient was in the next segment. There was also a correlation denoting a pattern of mutual enhancement: when the patient was more productive, the analyst was able to make a better intervention. These correlations take on more meaning from the discovery that there was not a strong correlation between patient productivity in a given segment and in the segment immediately preceding it. That is, we are warranted in attributing an observed increase in this variable to a specific intervention rather than to a more general rise in the patient's productivity.
The reader will recall that the CC patient-analyst pair was the one where the analytic process deteriorated and the analyst was angrily accusing the patient of foiling the treatment. In studying the relationship between analyst variables and patient productivity in this pair, we found that there was nevertheless some degree of increased productivity whenever the analyst interpreted, addressed conflict and addressed reactions to the analyst, even though these three activities were rare. We also saw clearly that the patient CC's productivity declined when the analyst became confrontational and expressed negative feelings.
An especially interesting finding emerged from comparing the correlations between patient productivity and responsiveness to analyst intervention for all three patients. The correlation was .80 for patient BB and .71 for patient AA, but only .57 for patient CC. The differences among these may well be significant. They suggest, (on the basis of this very small sample), that the stronger the correlation is between the patient's overall productivity and the patient's productive responses to the analyst, the better the outcome is for the patient. This in turn can be related to patient analyst match: it makes sense that if a treatment is proceeding well, patient productivity is strongly related to the analyst's contribution.
Can we study graphically the scores for analyst and patient work through the course of a session, to develop a better understanding of the degree to which each participant facilitates or impedes the progress of the treatment? The use of the APS to show the interactive process of a patient variable and a related analyst variable unfolding segment by segment is illustrated in figure 2, where we have charted the patient and analyst self-esteem scores for one of AA's sessions. It graphically shows that when the patient expresses concern about issues of self esteem, the analyst responds with interventions which address self esteem issues to a greater degree than before. Observe in the figure how the peaks in the analyst addressing self-esteem issues follows closely upon the patient peaks from segment to segment. A similar pattern occurred for BB (not shown). Both of these patients appear clinically to have benefitted more from their treatments than patient CC, in whose sessions this pattern did not occur. One of CC's important problems was self esteem issues, and it is likely that much of the reason this analysis did not progress was because the analyst did not respond significantly to these issues.
Another example is shown in figure 3, where we see how patient BB and his analyst interacted on transference issues. Close inspection of this figure shows that when the patient addressed his own reactions to the analyst, the analyst responded by remarks also addressing the transference, as can be seen by the way the analyst line peaks after the patient peaks. In addition, the figure reveals another aspect: each time that the analyst addresses a transferential aspect of the patient's feelings, the patient turned away from this topic. This is shown by the way the patient's expression of transference related issues fell to lower levels each time the analyst made a transference related remark: the patient shows a strongly descending line from each previous peak, following the analyst's responsive peak. In other words, the patient preferred to keep his distance, a reaction consistent with the patient's basic character problems. This tendency of the patient to avoid a response was also shown in the transcript by significant pauses before giving his response, after the analyst commented on
transference related feelings.
These examples illustrate the utility of the APS scores as sensitive indicators of the interaction between patient and analyst. To the best of our knowledge, our study represents the first demonstration of these kinds of relationships from recorded analytic data. We are able to evaluate reliably, for the first time, the degree to which a patient immediately responds to central psychoanalytic techniques, such as clarification, interpretation, and analysis of transference and resistance. In addition, we can evaluate the effect that a patient's productivity has on an analyst's work. The APS provides us with the means for systematically studying a range of patients and phases of treatments, and contrasting them with each other on central patient, analyst, and patient/analyst interaction psychoanalytic variables. In summary, we have found that the APS provides a reliable, detailed, systematic description of central aspects of the psychoanalytic process as widely understood by psychoanalysts today and discriminates between different psychoanalytic processes at different phases of treatment. It therefore promises to greatly expand our knowledge about the processes of psychoanalyses and psychoanalytic psychotherapies, and about the relationship between processes and outcomes.
QUESTIONS AND HYPOTHESES ABOUT ANALYTIC PROCESSES ANSWERABLE USING APS SCORES
How may the APS best be used to advance psychoanalytic understanding of clinical work? There are two aspects to our answers to this question at this time: first, there are several hypotheses we have already formed based upon the data from this small sample which, if confirmed, would be valuable additions to our knowledge about the conduct of psychoanalyses and psychotherapies. They include:
These hypotheses demonstrate the way ongoing study of clinical work using the APS can generate specific hypotheses worthy of further investigation. They are presented solely for illustrative purposes; they are strictly preliminary and await further study. Although many of our colleagues may consider them to be self-evident, it should be emphasized that they have never been systematically validated. Many more specific hypotheses will no doubt be developed by groups studying individual recorded psychoanalytic cases.
In addition, there are a number of general hypotheses which we have formulated which derive from a careful study of the nature of the APS instrument itself, and from our clinical experience. These represent an ambitious program requiring data from more recorded cases than we have collected so far. In general, the nature and quality of the analytic process of a given treatment can be assessed by measuring how well the analyst is working, how well the patient is working, and how well they are working together. These three aspects serve to organize the hypotheses presented below. In addition, in regard to each of these three perspectives on the work, there are are two special sets of questions which can be addressed. What are optimal ranges for certain activities, and what are desirable rhythmic features of the work. These issues have scarcely been addressed in studies of psychoanalyses because hitherto the data has scarcely been available. Some illustrative hypotheses are included below, with the understanding that rapid change in such hypotheses may be expected based upon actual study of more data.
A. There are many hypotheses bearing on how well the analyst is working, including the four above which derived from examining our preliminary data. Here follow others.
B. The assessment of how well the patient is working has been illustrated in the discussion of our findings, particularly in describing the differences between our three cases (section 2 of Preliminary Results above). However, there is an additional hypothesis we wish to mention here.
C. How well the patient and analyst are working together is assessed by two variables directly: how responsive is the patient to the analyst's intervention and how well does the analyst follow the patient's immediate emotional focus. In addition, there are a number of more complex hypotheses bearing on their work together, which can be studied using APS scores.
The APS makes a wide range of investigations possible which will allow us to better identify precisely which components of psychoanalytic psychotherapy, applied by whom, are actually beneficial to which patients under what circumstances. The ensuing long-term benefit to patients should be considerable.
IMMEDIATE RESEARCH PLANS: DEVELOPMENT OF THE APS, THE RESEARCH NETWORK, REPLICATION AND EXPANSION OF THE PRELIMINARY FINDINGS
Over the next two years we are planning to study the relationships between process and outcome in three recorded analyses. There have only been a few researchers who have tape-recorded analyses, but those who have been recorded have served to stimulate systematic research in a way which could be compared to the study of the prehistoric "Lucy" from the Rift Valley in Kenya. The case of "Mrs. C." has been the most productive among the various cases studied, but it is clear that scientific progress in psychoanalysis requires broadening of the sample, and we propose to add the study of two new fully recorded psychoanalyses to the case of Mrs. C. in the next two years. The three cases are and between six hundred and twelve hundred hours in length, and we have already determined clinically that they have a broad range of scores on analytic process and substantial variations in outcome. While still a very small sample, there is a great potential for enrichment of the field by careful study of a few cases. We will examine eight consecutive hours from the early, middle and late periods of each treatment, producing a total of seventy-two rated hours and allowing us to study the ebb and flow of analytic work in unprecedented depth. Each hour will be rated by three analysts--we have already found that the mean scores from three raters are very reliable. Because this project will be extremely time-consuming, we will need to divide the work between nine groups of three analysts each.
Our goals for the next two years are:
1. to more thoroughly explore the four areas we have already considered. In brief, these are:
2. to explore the concurrent validity of our variables via a comparison (correlation) of scores with the scores from the Psychotherapy Process Q-set (see appendix 2 for listing of the relevant items from the Q-set). The degree to which these two general instruments for assessing psychoanalytic and psychotherapy process demonstrate an overlap and continuity will provide support for the validity of both instruments.
3. to make a pilot investigation of hypotheses relating analytic work by both patient and analyst to actual progress made (see previous section). We will relate our findings using the APS to outcome scores generated in collaboration with the outcome subgroup we have been working with. This will prepare us for an investigation involving many more patients later on (see "Long-Range Plans" below).
4. to determine whether we can use the APS to reliably rate whole sessions as well as segments, and whether such ratings can be carried out by junior clinicians with comparable reliability as that obtained by senior clinicians. Segmental rating with the APS is very time-consuming. If whole-session ratings prove to be reliable, we will be able to study a much broader range of sessions, and reserve segmental ratings to obtain "high magnification" views of the process at points of special interest.. This part of the research will be carried out by senior analysts as they rate the sessions segmentally, and by junior clinicians with strong psychoanalytic interests.(15)
5. to develop a network of clinicians interested in tape-recording and/or rating sessions. Not only will our long-range plans require the assistance of many more analyst recorders, but we also want to encourage the formation of a number of independent research groups around the country to study recorded materials using the APS. Gathering a much larger collection of recordings under the auspices of the Psychoanalytic Research Consortium will support this development, and serve the research needs of psychoanalysis generally. Two psychoanalytic centers have already expressed interest in forming such groups -- The Baltimore-Washington Psychoanalytic Society and the Western New England Psychoanalytic Society. Ten analysts have already agreed to record portions of their own work, and another ten analysts have agreed to serve as raters using the APS.
LONGER RANGE RESEARCH PLANS
Upon completing our study of three analytic patients, we plan to study recordings of seven additional analytic patients, using the APS to rate the early, middle and late periods of their treatments.(16) To provide a contrasting group with regard to our measures of process and outcome, we will also study recordings from ten psychotherapy patients.(17) We anticipate that our process measures of the quality of the exchange between patient and therapist will be more strongly correlated with a favorable outcome than whether the patient is in psychotherapy or psychoanalysis; however, this remains to be demonstrated.
We will concomitantly evaluate the level of psychological health of each patient at the beginning and end of treatment, using various well established measures, and new measures currently being developed by our collaborating group applied to recorded psychoanalytic sessions. This will allow us to study how processes relate to outcomes in a number of psychoanalyses and psychoanalytic psychotherapies, and what characteristics of patient and analyst work distinguish good outcomes. We will be able to compare the processes of our psychoanalyses with those of the psychotherapies. Further studies, described below, will follow this work.
In order to determine the validity and stability of outcome assessments, we will need in-person follow-up assessments with a subgroup of the recorded patients. Waldron has already conducted an NIMH supported follow-up study of neurotic children into adult life,(18) and has accumulated some follow-ups on analytic cases of his own.
One of the great obstacles to the development of psychoanalysis as a science has been the lack of knowledge about how psychoanalytic theory, method and training is related to the psychotherapeutic benefit that a specific patient derives from working with a specific analyst. Unable to compare their results, the various "schools" of psychoanalytic thought have tended to espouse different approaches to the patient. Our instrument will be able to provide a means of systematically studying the clinical consequences of different psychoanalytic theories. Do experienced analysts agree on the quality of a treatment process, regardless of their theoretical orientations? This question can be answered by having groups of analysts from centers favoring different theoretical approaches rate the same series of cases. Similar ratings would indicate that experienced clinicians understand clinical material in a similar way, despite their differing theories, while ratings which varied according to theoretical differences would demonstrate that different theories produce different clinical effect.
A benefit of such a study is the way the ensuing familiarity with the use and value of the APS may encourage participating analysts to record their work. In particular, when we have involved analysts from different theoretical backgrounds as raters, we can expect a growing interest, on the part of each group, in studying their own cases with adaptations of the same methodology. Then we, or they, can develop additional scales to measure precisely those patient and analyst features which are pertinent to their theories and methods. This will provide the basis for evaluating the benefits derived from analysts of different theoretical backgrounds, and, most importantly, for ascertaining which of various approaches work better for particular patients. It is possible, for instance, that a patient who has great problems with self esteem, and perceives the analyst's interpretations as threats to the way he or she experiences him or herself, will work more productively with a self psychologist, whose approach to interpretation may differ from that of a more typical ego-psychological analyst. The APS, together with additional scales, will allow us to evaluate what approach is in fact being used in a particular treatment, with what degree of skill, with what participation of the patient, and with what results. Thus this study will generate a much broader set of studies focusing on the varieties of psychoanalytic approaches. Meanwhile, new recordings of analyses will become available through the Psychoanalytic Research Consortium for the benefit of the entire field. This summary of our variables and their preliminary application demonstrates the importance of obtaining a sufficient sample of cases for systematic studies.
Coordinating our measures with other investigations will help us to refine the use of the APS and make it more efficient. For example, Spence's computer-generated measures of patterns of association of meaningful words in psychoanalytic discourse, Mergenthaler's computer-generated measures of emotional and abstract language, and Bucci's measure of referent activity, may allow us to enlist the computer in identifying the most likely places to find important turning points, or locate the sessions most likely to reflect moments of important work between patient and analyst.
Until now, the APS research group and the Psychoanalytic Research Consortium have been largely self-funded. The American Psychoanalytic Association and the New York Psychoanalytic Society have both contributed funds to the PRC, the former primarily to secure legal assistance in establishing safeguards and to obtain 501(c)(3) status as a not-for-profit corporation, the latter for purchasing the audio equipment which we currently offer to psychoanalysts willing to record their own work. The members of the APS research group have paid most of the expenses for transcription, as well as for the services of a psychologist research assistant. For the next period of the study we are seeking substantial outside funding to carry out the extensive studies of three patients, and to develop and support the network of clinicians whose participation is crucial to its success. For those who are interested, we have a grant application showing our proposed budget for the next two years.
Version March 28, 1998 C:\RG\introco3.htm
1. The data from the study, as reported in Bachrach, H., Galatzer-Levy, R., Skolnikoff, A., and Waldron, S.(1991) On the Efficacy of Psychoanalysis. Journal of the American Psychoanalytic Association. 39: 878-885, have been re-worked by Waldron to demonstrate impact on quality-of-life based on percent changes in the Health-Sickness Rating Scale scores from initial to follow-up two years after treatment. Mean improvement in HSRS scores was 37%, for the 80% of patients who benefitted.
2. . T. D. Borkovec and J. Miranda (1996). Between-Group Psychotherapy Outcome Research and Basic Science. Psychotherapy & Rehabilitation Research Bulletin, #5: 14-20, NIMH, Rockville, MD.
3. Although we focus our discussion on analysts and the analytic method of therapy, the APS can be used to evaluate all psychotherapeutic work derived from a psychoanalytic perspective. Our plans include the study of psychodynamic psychotherapies later.
4. The reader will note that the variable assessing how much the analyst encourages elaboration is on a separate petal of the flower diagram. We have found that encouraging elaboration as an activity tends to come at different points in the session than all other analyst activities, hence is negatively correlated with the other variables.
5. Notice that the patient and analyst cluster variable scores in this sample co-vary; that is, the column heights of each pair appear to go up and down together. This may reflect a general phenomenon of mutual influence; however, we will need to study material from several patient-analyst pairs of the same analyst to see if this is consistent finding.
6. Table of scores for all the variables may be obtained from the first author.
7. There were a large number of substantial correlations between quality scores and most of the other variables. This shows that the raters, in assessing quality, took into account many different features of the analytic process as observed in these three cases. This adds to our confidence in the validity of the differences in quality scores between patient-analyst pairs in this sample: the differences may be presumed to reflect general psychoanalytic features of the patient work and the analyst interventions.
8. Correlations measure the degree of association between two sets of scores. If they were perfectly associated, the correlation would be +1.00; if a high score in one was associated with an equally low score in the other, then the correlation would be -1.00; if there were no association, the correlation would be .00.
9. Whether the patient refers to developmental time periods and makes genetic connections may reflect a positive element, but in our view this is not suitable outcome measure at out present stage of knowledge. Neither are references to matters affecting self esteem.
10. The overall significance of the relationships found is established by the following: 21 of 117 correlations between analyst interventions and the analytic quality of the subsequent patient segment were significant at the .05 level or better (12 at .01 level); 14 of 117 correlations between the analytic quality of patient segment scores and subsequent analyst response were also significant at the .05 level (2 at the .01 level). This pattern of significant findings would not be expected to occur by chance. Because of the small sample, we must nevertheless be reserved in the degree of confidence with which we rely on particular individual correlations. The strength of our findings will be greatly enhanced if the relationships are replicated in another sample.
11. In assessing the hypotheses given, we expect to monitor the quality of the contributions of each segment simultaneously with its nature. In this instance, we will study analytic activity of good quality, not simply any analytic activity, by taking into account statistically the quality level of interventions as we examine the relationships between given interventions and subsequent response, in relation to outcome.
12. We have already found in our preliminary sample that encouraging elaboration correlates negatively with virtually all the other analyst variables. This supports this hypothesis, in that there appears to be a rhythm of work by these analysts. It is also the reason why, on our analyst flower diagram, the leaf for encouraging elaboration is separated from all the other leaves, that is, it is not part of any of our cluster variables.
13. This conceptualization owes a debt to the work of Erhard Mergenthaler, Director of the Ulm Textbank, whose description of fluctuation in levels of Emotion Tone in relation to Abstract Words, a computer generated measure he has developed in association with Wilma Bucci's measure of Computer Referential Activity (CRA), has inspired our developing this parallel hypothesis. We expect to apply his measures to our data as we explore this hypothesis later.
14. Examination of such sequences from our recorded materials has already been used to teach technique at the Psychoanalytic Institute at NYU, and at the Western New England Psychoanalytic Institute. Specific examination of such sequences might also be adapted to serve as a method of study of recorded hours as a consultative tool in cases which are not progressing as well as is hoped for.
15. One aspect of our study, not further described here, is an exploration of the degree to which clinicians agree as to the relative quality of sessions. At the same time that whole-session ratings are performed, raters will rank order sessions as to overall quality in comparison with other sessions in the same group. This will serve as another source of information about shared judgments of quality, and is inspired by similar findings in the course of the development of the Health-Sickness Rating Scales of the Menninger study.
16. The tape recordings and transcripts of these patients are available through the PRC. They were conducted by five different analysts: three analysts provided one case; one, three cases; and one, four cases. The treatments vary in length from about 200 to 1200 sessions. A larger sample would be even more statistically valuable, but this will have to await the accumulation of additional recorded cases during the course of this study.
17. The ten psychotherapy cases were primarily conducted by psychoanalysts, two of whom also provided psychoanalytic cases for this study. This will allow us to directly compare psychoanalytic and psychotherapeutic work in a pilot fashion for a small number of cases. Frequency in the psychotherapy sample varies from once to twice weekly, and length varies from about 100 to 700 sessions.
18. Waldron, S. (1976). The Significance of Childhood Neurosis for Adult Mental Health: A Follow-Up Study. American Journal of Psychiatry l33: 532-538.
SEE APS PROJECT APPENDIX FOR FURTHER DETAIL ABOUT THE VARIABLES, RELIABILITIES, AND CORRESPONDENCE WITH THE PSYCHOTHERAPY PROCESS Q-SET VARIABLES OF JONES ET AL.