Most psychoanalytic clinicians and researchers know that psychoanalysis and long-term psychoanalytic therapy work. What remains uncertain is what contributes to that beneficial result. Freud’s emphasis on the role of insight has not been sufficiently confirmed in empirical studies, nor has the role of interpretation in bringing about favorable change been easy to demonstrate.
Thirty years ago a group of analysts in New York started tackling the job of connecting the processes of psychoanalytic work with outcomes. We didn’t think we had a chance to accomplish this except by studying actual recordings. It was supremely difficult to persuade fellow analysts to record. We had the benefit of groundbreaking efforts by Merton Gill and by a group led by Lester Luborsky in Philadelphia, and I started recording my own work to give us more to study. My background contributed to my making such a culturally unacceptable decision: I had been educated as an undergraduate at the Department of Social Relations at Harvard. This gave me a view of the science of the social sciences (including statistics), and inspired me to do a small research project of my own for my thesis, so that I had the benefit of a somewhat unique perspective as I pursued the goal of becoming a psychoanalyst.
The research group I was able to form 30 years ago continues to this day, with gradually changing membership. Our first task was to develop measures of what is going on in the consulting room, viewed from a psychoanalytic point of view. We developed scales and variables to assess core psychoanalytic activities such as the analyst’s clarifying, interpreting, addressing transference, conflicts and defenses, along with other aspects of the nature and quality of the analyst’s activity and the therapeutic relationship. Also, we assessed the patient’s contribution to the process—conveying his/her experiences, reflective functioning, ability to express feelings in an informative way, response to the analyst’s contributions and the quality of the patient’s own contributions.
To aid in reliability, we developed clinical examples to illustrate the different levels of each variable, from zero to four points, and the resulting APS Coding Manual (with Robert Scharf as first author, unpublished 1993 revised 2010) is helpful as a teaching instrument to students of analysis and analytic therapy. Published results from the ensuing Analytic Process Scales (APS) showed that the strongest influence of the analyst on the patient came from those analyst comments which were of the highest quality (as judged by experienced analyst-raters), irrespective of whether the comments were an interpretation, encouragement to elaborate, addressing the patient’s self-esteem or of a similar nature. The analyst-raters used their own best judgment, guided by our APS Coding Manual. The judgments were moderately reliable among the analyst-raters.
What to make of these findings? We realized we had been concentrating most on the technical aspects of the analyst’s work, but the relational aspects needed to be more directly evaluated. So back to the drawing board. We
developed another set of scales to be applied to each recorded session. We called these the Dynamic Interaction Scales (DIS). As if the 28 scales already described were not sufficient (which they were not). The DIS assesses contributions of the patient, therapist and the interaction between them.
Then a breakthrough occurred: We began a collaboration with colleagues in Rome, Vittorio Lingiardi and Francesco Gazzillo, of “La Sapienza” University Department of Psychology. They proposed to study what had now become an entire collection of 27 completely recorded psychoanalyses by seven different psychoanalysts. Back in the middle 1980s I had set up the Psychoanalytic Research Consortium (PRC), with help from the Scientific Activities Committee and the Fund for Psychoanalytic Research of APsaA, to collect, safeguard and confidentialize as many psychoanalyses as we could persuade people to make or share with us. Our group included the late Robert Wallerstein as vice-president of the PRC and other members of the Scientific Activities Committee. By the time our Roman collaboration began we had enough cases to make a small study, and received some funding from a private foundation as well as from the research arm of the International Psychoanalytical Association, and the Department of Psychology at La Sapienza University.
We decided to study 20 sessions from each of the recorded analyses, from early, middle and late in each treatment. We were able to find out about what elements contributed to forward movement, once Francesco and his team of colleagues finished applying our 39 process measures to all 540 sessions—a huge accomplishment.
First we needed to establish whether our patients actually benefitted from their analyses. Using both the Global Assessment of Functioning (GAF) and a new measure we developed called the Personality Health Index (PHI), we found most of them did benefit. While widely used, the GAF is a rather crude measure. The PHI is based upon Jonathan Shedler and Drew Westen’s Shedler Westen Assessment Procedure (SWAP). The clinician makes detailed clinical judgments of 200 known aspects of a person’s pathology and strength. In 2011, we published the PHI, which provides a percentile score compared to a nationwide sample of other patients in psychoanalysis collected by Rosemary Cogan and John Porcerelli.
Clinicians can evaluate their own patients, or clinician/researchers can rate patients based on listening to eight consecutive sessions to develop sufficient familiarity with the patient to perform the SWAP. The latter was the procedure we followed. The patient who had the most positive change in the course of his analysis went from a PHI of 10, which meant that 90 percent of analytic patients were healthier than he was, to a PHI of 85 at the end of his treatment, meaning he was functioning in the top 15 percent of the reference analytic patients. The chart below (Figure 1) shows how the group as a whole fared from early to late in their treatments (with identifying pseudonyms for each patient). It is clear there was considerable benefit overall, which was also supported by changes in GAF scores that were rated by a different set of raters.
Could we tell what was different about the analyses that did better and those that did do not so well? Here we ran into our biggest problem: our sample size of only 27 cases. We classified 17 as good outcome cases and 10 with limited benefit or deterioration. In view of the importance of the unanswered questions about the efficacy of psychoanalysis under varying conditions (e.g., therapist, patient, techniques used) the number should have been 270 cases. Then we could examine patterns in a way that could give more detailed meaningful conclusions. In this study the only finding we can report about differences between good and poor outcome cases was that all the variables (patient, therapist and
interaction) that would be expected to be higher in good outcome cases did indeed average higher in the eight early sessions, compared to the poor outcome cases. This result could scarcely occur by chance. It would be like flipping a coin 39 times and always coming up heads. But because of the low N, we cannot claim to having demonstrated differences between the two groups on any given variable. Our field needs the new generation of analysts to have the courage to record their own work to answer many important remaining questions.
Meanwhile, what can we learn from our process variables, as assessed for all the 540 sessions? First we wanted to know what variables go with each other. Discovering this is a way to simplify to some extent the results from so many measurements and to reveal the structure of these analyses, viewed through the lenses of analytic clinicians. A factor analysis was applied to three domains: contributions of the patient, the analyst, and the interaction between patient and analyst. It turned out the factors we found made clinical sense and could readily be named, although to get a feel for them, one would want to review the individual variables that contributed to each factor. Both the patient and therapist variables reduced to three factors, whereas the interaction only led to one factor. (Figure 2)
THERAPIST FACTORS: The first, Therapist Relational Competence, evaluates the therapist’s warmth, amicability, sharing of his/her own subjectivity, expressing feelings, being supportive, straightforward and addressing momentary shifts in the patient’s feelings. A second small therapist factor is Therapist Confronts, that is, how confrontative the therapist was in the session. The third therapist factor, Therapist Dynamic Competence, includes encouraging elaboration, clarifying, interpreting, addressing defenses, transference and conflicts, plus the more general assessment of the quality of the analyst’s communication.
INTERACTION FACTOR: Analysis of our interaction variables yielded only one factor, Interaction Quality. The variables in this factor include: Therapist helps patient to be more aware of feelings; the patient experiences the therapist as empathic; the therapeutic couple make connections between what is happening in the consulting room and the rest of the patient’s life; the patient responds productively to the analyst’s communications; and the therapeutic couple are rated as engaged with one another.
PATIENT FACTORS: The first patient factor is Patient [Communicates] About His/Her World (meaning outside the consulting room). The second, Patient Dynamic Competence, includes shifting flexibly between experiencing and reflecting, overall productivity (defined and illustrated in the APS Coding Manual), and addressing troubling patterns in his/her life. The third is Patient [Communicates] About the Analyst or Analytic Situation. All the component variables are rated on a five-point scale.
Once the factor analysis is completed, the seven factors identified become variables in our study, so we can examine how they are related to one another.
The most important result from our assessment of these variables is that we can say what aspects affected the progress of these analyses by studying changes from session to session. There are 405 pairs of sessions adjacent to one another in our database (that is, two consecutive sessions on successive days or within a few days of each other). This gives us the opportunity to find out what activities or characteristics, among our variables, influence the psychoanalytic process favorably or unfavorably in the following session. Using this approach, we have very strong evidence of the impact of a psychoanalytic approach in enhancing those features of the experience that have been shown by many other studies in psychodynamic therapy to be beneficial to patients.
Figure 3 illustrates that therapist relational competence and therapist dynamic competence both contribute to heightened interaction quality in the next session. And higher interaction quality in the previous session contributes to enhanced patient dynamic competence in the next session. In addition, therapist dynamic competence contributes to patient dynamic competence directly.
There are other interesting interactions. Figure 3 summarizes the findings by looking at how one factor influences the other in the next session. Arrows are shown only where the results are statistically significant. The origin of each arrow is the score in the previous session, and the point of the arrow is the result in the next session (in 405 pairs of sessions).
It is of interest that there are two adjacent arrows going in opposite directions, between Therapist Dynamic Competence and Interaction Quality. This means that therapist dynamic competence in the previous session increases interaction quality in the next session, and interaction quality in the previous session increases therapist dynamic competence in the following session. This is the kind of core finding to indicate that indeed, what we communicate to our patients and how well we communicate it (tactfully, with appropriate timing and consideration for self-esteem) demonstrably makes a difference in the psychoanalytic process.
Almost all of the factors are composed of multiple individual variables. We have checked the correlations between the variables in each factor and the variables in the other factor in the next session, and the majority of the underlying variables are correlated significantly.
Figure 4 is a detailed chart showing the variables underlying one arrow. As is apparent in this chart, the therapist addressing transference and conflicts leads to increases in the three patient dynamic competence variables (The other four therapist dynamic competence variable correlations are not high enough to reach statistical significance.). The first of these two findings validates that, when used with sufficient understanding of the patient, and of his or her state of mind at the particular moment, addressing the patient’s reactions to the analyst or analytic situation can be most helpful in psychoanalysis, as other studies in psychodynamic therapy have found, for instance the groundbreaking study by Per Høglend and his colleagues from Norway . The second finding, the role of addressing the patient’s conflicts in leading to forward movement, validates this classical psychoanalytic tenet perhaps most elegantly expressed by Charles Brenner in The Mind in Conflict .
Psychoanalysis is a complex undertaking, and no two patient-analyst pairs are going to turn out alike. Figures 3 and 4 bring us in the direction of capturing that complexity. At the same time, they stimulate ideas that might not have emerged without such analysis of the data. For instance, as shown in Figure 3, the sheer fact that the patient was talking about or expressing feelings about his/her analyst did not prove to have a significant direct impact, by itself, on other dimensions of the work in the next session, nor was it impacted by previous therapist relational competence, therapist dynamic competence or interaction quality. In listening to the material, we sometimes had the impression that some patients could talk a good game in this respect, but it did not necessarily correlate with positive results.
There is one additional finding from looking at Figure 3 that is noteworthy. The third analyst factor was Analyst Confronts. It happened that higher scores on almost all the therapist relational competence variables in the previous session were predictive of increased confrontation in the next session, although this increased confrontation was not predictive of any of the other factors in the session following. We do not imply that confrontation is not useful; rather we believe the clinical material needs to be studied by clinicians to ascertain what impact such confrontation might have, not just in the next session but going forward. This is a new finding worth pursuing by further detailed study of the recorded sessions in which confrontation occurred, and the consequences. There is much room for more work, and we welcome help from interested colleagues. It is clear that, to advance psychoanalytic understanding, one needs to study the qualitative as well as quantitative.
As previously described, it may be seen from Figure 3 that therapist dynamic competence and interaction quality directly impacted patient dynamic competence, and that therapist relational competence also played an important enhancing role by enhancing interaction quality. This is consistent with the development in our field of the “relational turn.” At the same time, the “classical” analytic values appear to be well supported in our study. It is not “either-or.”
An important goal for the future is to compare the results of psychoanalysis and long-term psychoanalytic psychotherapy (LTPP). Studies such as the one by Falk Leichsenring and Sven Rabung have shown convincing evidence for the superiority of LTPP over other forms of therapy for what they call “complex mental disorders,” but we need to clarify further what difference in value there may be between LTPP and psychoanalysis for longstanding limitations in character or ways of relating to others. Here the new generation of analysts could make a great contribution by recording both cases of psychoanalysis and of LTPP. We owe it to our patients, and to the health of our practices, to investigate when and to what degree a substantially more intensive treatment effort works to the patient’s benefit, even though as psychoanalysts most of us are convinced that for some patients the more intensive form of therapy is essential. A systematic comparison of such recorded cases would be beneficial to our field. There are descriptive materials about making recordings on our website——and I will be glad to help and advise anyone who would consider joining in this work. You may also contact me at for other detailed graphs not included in this article.
My thanks to our U.S. research team, which includes currently Robert Scharf, Seymour Moscovitz, Fonya Helm and Karl Stukenberg, and to the Italian team led by Francesco Gazzillo.