Defense mechanisms have been one of the most durable constructs in psychoanalysis and dynamic psychiatry and psychology, spanning theory, therapy, and research, since Freud’s 1894 publication “The neuro-psychoses of defense” (1). Following the development of systematic assessment methods, recent studies have demonstrated that defenses can be arranged hierarchically based on their usual level of adaptiveness (2–6). Studies of adult development have demonstrated that defensive functioning tends to improve over time in stepwise fashion, that is, individuals’ repertoires of defenses tend to trade off lower for mid-level, then higher-level defenses in the hierarchy, progressing over many years (4). Subsequent naturalistic studies have also found that defensive functioning improves over the course of brief (7, 8), medium-term, and long-term psychotherapy (9–13) and psychoanalysis (14). Furthermore, defensive functioning at the outset of treatment predicts response at 6 months in major depression (15). One intriguing psychotherapy study found that change in distress was mediated by prior improvement in defensive functioning, but not in conscious coping (13).

Dynamic therapists consistently interpret defensive functioning even more frequently than they do transference. In one case series, addressing defenses surrounding therapeutic impasses was found to be critical to improving alliance (16). Another study (17) showed that addressing defenses during short-term psychotherapy improved neurotic defenses. While confirmation by controlled trials is needed, evidence to date suggests that change in defenses is related to or may promote other changes in symptoms and functioning. Defenses therefore warrant further clinical study.

In this study, we examined improvement in defensive functioning during long-term dynamic psychotherapy in relation to longer-term improvement in symptoms and functioning. We built on our previous study demonstrating that self-report defensive functioning, assessed outside of therapy, improved significantly (9). Our team subsequently developed the following four hypotheses characterizing change in defensive functioning in psychotherapy:

• Defensive functioning tends to improve during long-term dynamic psychotherapy.

• Defenses change in stepwise fashion, following the hierarchy of adaptation. Specifically, a) in the early phase of change, defenses low on the hierarchy (e.g., acting out) decrease while those in the middle of the hierarchy (i.e., neurotic) increase, whereas b) in the later phase of change, mid-level defenses decrease and high-level defenses (e.g., suppression) increase.

• An individual’s rate of change in defensive functioning is moderated by characteristics such as level of illness.

• Improvement in defensive functioning during dynamic therapy will be associated with longer-term improvement in psychosocial functioning and symptoms.

We reported preliminary confirmation of these hypotheses in a systematic examination of four therapy cases with 1 to 14 years of follow-up (12). We extend our investigation here, using a naturalistic study of long-term dynamic psychotherapy to examine evidence for these hypotheses. Confirmation would support the hypothesis that changes in defenses in dynamic psychotherapy may in part mediate changes in symptoms and functioning.

Design and Study Participants

The study design, inclusion and exclusion criteria, and participants have been described in detail previously (9, 18–21). Briefly, the overall aim of the study was to examine the course and outcome of long-term dynamic psychotherapy for individuals for whom clinicians deemed that previous, usually short-term, treatments had been insufficient. The design was naturalistic and observational. Study participants were referred from the outpatient psychiatric department of a university teaching hospital in Canada. Selection criteria included having a depressive, anxiety, and/or personality disorder, expressing a wish for psychotherapy, and agreeing to participate in the research component. Level of defensive functioning was not an inclusion criterion. Exclusion criteria included psychosis, organic brain disorders, and significant current substance abuse that might interfere with learning. Participants gave written informed consent after the study was explained to them and their questions addressed.

Therapists

The study therapists were experienced practitioners of long-term dynamic therapy, with a mean of 13.1 years of postdoctoral experience; most were psychoanalysts. Therapists treated a median of three participants each.

Psychotherapies

Dynamic psychotherapy was offered once or twice weekly at the discretion of the participant and therapist. Treatment was provided at no cost to the participants, as is the general case in Canadian hospitals. Intending to reflect long-term dynamic therapy as locally practiced, we used neither specific therapy manuals nor supervision groups. While participants were offered a minimum of 3 years of treatment, they could terminate at will or try other therapies, such as pharmacotherapy, concurrently or sequentially, reflecting real-life practice conditions.

Study Procedures

Participants had an initial guided clinical interview with a psychiatrist who made DSM-IV axes I through V diagnoses, including provisional personality disorder types, and obtained a personal lifetime history (9), which was scored for 10 aspects of childhood adversity using the Traumatic Antecedents Interview scoring method (21). At baseline and every 6 to 12 months, research assistants interviewed participants using the Longitudinal Interval Follow-Up Evaluation–Adapted for the Study of Personality (22, 23). All psychotherapy sessions were audiotaped. Transcripts were made of sessions 3, 5, and 7, another three sessions at 6 months, and two more sessions at 2.5 years for those still in treatment. Audiotapes of these eight sessions were disguised as to session number and rated in random order using the Defense Mechanisms Rating Scales, quantitative method (24). A single rater rated the block of sessions for the participant, with a second, blinded rater rating the first and fifth sessions for reliability determination. After rating, both raters reviewed the session together and developed a consensus rating. Substantive analyses used the individual rating, except when a consensus rating was available. Rating procedures are detailed elsewhere (25). Rating sessions commenced after most treatments were completed.

Measures

The Defense Mechanism Rating Scales, 5th edition (24, 25) (English edition available from the first author) is an observer-rated method, nearly identical to the Provisional Defense Axis in Appendix B of DSM-IV (6, 7). Using quantitative scoring directions, raters identify each defense in order of occurrence in a session. This method differs from other observer-rated methods that yield qualitative or semiquantitative ratings based on global ratings for the whole interview (26). Three levels of scoring are used, yielding continuous, ratio scales. Individual defense scores are proportional or percentage scores, calculated by dividing the number of times each of 30 defenses was identified by the total instances of all defenses for the session. Defense level scores are proportional or percentage scores of each of seven defense levels. Each level consists of the sum of the scores of the three to eight constituent defenses with shared functions. The defenses are arranged hierarchically into seven defense levels based on their general level of adaptiveness (3). Table 1 lists the levels and their constituent defenses. Overall defensive functioning is a summary score obtained by taking the average of each defense level score, weighted by its order in the hierarchy, yielding a number between 1 (lowest) and 7 (highest). Finally, defense level scores are combined into superordinate categories: mature, neurotic, immature, and psychotic, although psychotic defenses were not included in this study. The intraclass R_(2,1) interrater reliability statistics (I_R) for the defense summary variables were as follows (N=32): number of defenses identified per session I_R=0.82; overall defensive functioning I_R=0.81; defense levels, median I_R=0.71; immature category I_R=0.83; neurotic category I_R=0.70; mature category I_R=0.67.

The following symptom and personality measures were obtained at baseline and at each follow-along interview. The Symptom Checklist–90–Revised (SCL-90-R) is a self-report measure of psychiatric symptoms, summarized by the Global Severity Index (27). The 21-item Hamilton Depression Rating Scale (HAM-D) and Hamilton Anxiety Rating Scale (HAM-A) were rated by interview (28, 29). Each axis I disorder with any positive symptoms was coded week by week on a 5-point intensity-of-symptom scale using the Longitudinal Interval Follow-Up Evaluation method (22). The Social Adjustment Scale (longitudinal version) is a semistructured interview of social functioning using a 5-point ordinal scale (30), adapted to the above longitudinal format. Overall life satisfaction was scored by similar format. Monthly ratings for each follow-along interval were summarized as a mean of each social role rating. The Global Assessment of Functioning Scale (GAF) was rated in a month-by-month format.

Statistical Analysis

All analyses were conducted with SAS 9.2 for Windows (31). As many distributions were nonnormal, bivariate and partial correlations were computed using the nonparametric Spearman rank correlation (r_s). Because in long-term follow-along studies data collection is subject to irregularities of time and missing data, as previously detailed (19), we calculated individual simple linear regression models for each participant for each measure for which there were two or more follow-along interview observations. We estimated the slope (rate of change) and the predicted scores at intake and last observed follow-up, calculating raw change. The Wilcoxon rank-sum or median tests were used to examine whether the rate or amount of change was significant. Within-condition (pre-post) effect sizes are presented as an approximate measure of the magnitude of change in defenses. Because this measure presumes normal distributions, it should be considered with caution. Given the small study group size, we considered the trade-off between type I and II errors, and given that our initial and final estimates of each variable were modeled using multiple observations, we considered the precision obtained a partial guard against false positive findings.

To avoid a problem with multiple comparisons when examining the 5-year outcomes, we combined the effect sizes of our eight outcome measures into two composite variables, as follows. First, each subject’s effect size for the eight outcome measures was corrected for the subject’s initial value, using simple linear regression. This produced eight residualized effect size variables. Taking the mean of the residualized effect sizes of the three functioning measures (GAF, Social Adjustment Scale, life satisfaction), we created a composite variable, called residualized functioning effect size (Cronbach’s alpha=0.86). The mean residualized effect size of the five symptom measures (HAM-D, HAM-A, Global Severity Index, number of current axis I disorders, and proportion of all current axis I disorder symptom levels) was called residualized symptom effect size (Cronbach’s alpha=0.85). After partialling out initial defense scores, we correlated raw change in defense levels with both composite variables.

Of the 21 patients rated for defenses, 16 (76%) were female; the mean age was 34.2 years (SD=10.5, range=18–53), and the mean current Hollingshead II Factor Index was 3.7 (SD=1.3, range=1–5), indicating a mid to low socioeconomic stratum. The mean number of axis I disorders was 4.4 lifetime (SD=2.7) and 3.0 current (SD=1.7). Sixteen participants (76%) had an axis II disorder, and the most prevalent type was borderline personality disorder (N=6, 29%). The mean number of axis III conditions was 1.6 (SD=1.5), the mean axis IV stress score (DSM-III-R version) was 2.5 (SD=0.9), and the mean GAF score at intake was 54.4 (SD=4.1).

Patients received a mean of 0.74 sessions per week (SD=0.23). Treatment lasted a mean of 248 weeks (SD=103) with a mean of 176 sessions (SD=79). Follow-along lasted a mean of 5.1 years (SD=1.6). At the end of the study, two patients (10%) had dropped out and three (14%) had terminated because of a move; 16 (76%) had either completed treatment or were continuing treatment.

Table 2 lists the median and mean estimates of the proportional defense scores at the first month and 2.5 years, as calculated by the individual regression models. Two defense levels produced significant rates of change: level 1 (action) defenses decreased, while level 7 (high adaptive) defenses increased, yielding effect sizes of –0.83 and 0.80, respectively. Defense levels 2 through 6 did not show significant change, although the major image-distorting defenses, which had a very low base rate, showed a significant negative direction of change (sign test). In general, lower levels (1 through 4) showed a negative direction of change, while higher levels (5 through 7) showed a positive direction of change. There were two exceptions. Level 3 (disavowal defenses) showed a positive direction of change (increasing), while level 5a (hysterical defenses) showed a negative (decreasing) direction of change. Of the tripartite categories, the direction of change was negative for immature but positive for neurotic and mature defenses. The number of defenses used per session declined nonsignificantly. Overall defensive functioning score improved significantly (effect size=0.71), but the final median and mean values remained below 5.0, consistent with depressive or higher-level personality disorders, which, while improving, had not yet attained neurotic or healthy-neurotic levels of functioning.

Table 3 addresses the second hypothesis, namely, that improvement in defenses proceeds stepwise up the hierarchy. It displays the intercorrelations of raw change (late value minus early value) for each defense level and category, after partialling out the initial values of each pair of variables. Decreasing action level defenses correlated significantly with decreasing hysterical but increasing obsessional and high adaptive level defenses.

As disavowal level defenses decreased, obsessional defenses increased, although this relationship fell short of significance. As minor image-distorting defenses decreased, both other neurotic defenses (displacement/reaction formation) and high adaptive defenses increased significantly. As hysterical defenses decreased, obsessional level defenses increased significantly. Given a low initial prevalence, only change in major image-distorting level defenses failed to demonstrate some relationship with other changes. Among the tripartite categories, decreasing immature defenses correlated with increasing high adaptive and neurotic defense categories. There was no relationship between changes in the neurotic and high adaptive categories.

The third hypothesis examined selected intake demographic, diagnostic, and childhood adversity variables as potential predictors of the rate of change in overall defensive functioning score, after partialling out the initial overall defensive functioning estimate (all N=21). Age, sex, and socioeconomic status were nonsignificant. The total numbers of lifetime (r_s=–0.45, p=0.05) and current (r_s=–0.55, p=0.01) axis I disorders were significant predictors, while the number of axis III disorders (r_s=–0.44, p=0.06) approached significance. The presence of significant traits or a definite axis II personality disorder, axis IV level of stress in the past year, and axis V current GAF did not reach significance. Two childhood abuse variables were significant predictors—total sexual abuse (r_s=–0.55, p=0.01) and total witnessing violence (r_s=–0.45, p=0.05)—while two others were nonsignificant—total physical abuse (r_s=–0.22, p=0.34) and total verbal abuse (r_s=–0.30, p=0.20). Negative direction indicates predicting a slower rate of change in overall defensive functioning score.

The fourth hypothesis examined whether change in defenses in line with the hierarchy at 2.5 years was associated with improvement in functioning and symptom levels over a mean of 5.0 years, using the residualized composite effect size scores. In Table 4, positive correlations indicate that an increase in the defense level or category was associated with improvement, whereas a negative direction indicates that a decreasing defense score was associated with improvement (Figure 1). Decreasing action defenses correlated with improvement in both functioning and symptoms. Increasing obsessional defenses correlated with significant improvement in functioning, while increasing high adaptive defenses correlated with improvement in symptom levels as well as with improvement in functioning, although this association fell short of significance. All correlations with change in functioning below level 6 were negative in direction.

Among the tripartite categories, decreasing immature defenses correlated with improvement in both functioning and symptoms. Increasing neurotic defenses correlated with improved functioning but not with symptom levels. Increasing high adaptive defenses were positively correlated with improved functioning and symptoms, as noted above. Finally, improving overall defensive functioning score correlated equally significantly with improvement in both functioning and symptoms.

We previously found that a rating of “mostly satisfied” on life satisfaction was highly associated with improvement and recovery on other variables (19). In an exploratory analysis in this sample, we found that the six individuals who had attained this level of life satisfaction had all shown substantial positive increases in overall defensive functioning score (mean effect size=2.12, SD=1.16). Raw change in overall defensive functioning score was significantly greater than for the 14 participants who had not attained this level of life satisfaction (0.61 compared with –0.01; t=3.49, df=19, p=0.003).

All of our hypotheses were derived from the theory-based, empirically validated hierarchy of adaptation of defenses. In line with our first hypothesis, the lowest-level action defenses decreased and the high adaptive level increased significantly by 2.5 years in therapy, as did overall defensive functioning.

Generally, immature defenses decreased, while neurotic and high adaptive defenses increased, in line with the second hypothesis of stepwise improvement up the hierarchy of defenses. However, high adaptive defenses increased more than did neurotic levels, indicating that a stepwise developmental pathway from immature through neurotic to mature defenses is only partly accurate. Instead, as immature defenses decrease, both neurotic and high adaptive defenses increase, slightly favoring the high adaptive level. Action and hysterical defenses decreased in tandem, correlating with increasing obsessional defenses. This indicates important differences between hysterical and obsessional levels, although both are neurotic. Overall, this hypothesis is largely but not completely upheld, because of the indication that immature defenses can directly change to high adaptive as well as to neurotic defenses.

The third hypothesis addressed predictors of the rate of change in overall defensive functioning. Age, sex, social class, axis II disorders (borderline personality disorder in particular), stressors in the past year (axis IV), and GAF score were insignificant. Two illness variables—the total number of lifetime and of current axis I disorders—predicted a slower rate of improvement. The number of axis III medical conditions was a negative predictor, approaching significance. Childhood total sexual abuse and witnessing violence histories were significant negative predictors. These findings, that higher numbers of psychiatric and medical illnesses and certain types of childhood adversity predicted slower rate of improvement in defenses, suggest why some patients may require longer-term therapies to achieve a given level of improvement. This may inform patient selection in studies of longer-term therapies or indicate how study samples may not be comparable.

Our final hypothesis was upheld: improvement in defensive functioning was associated with improvement in longer-term functioning and symptom levels. Overall defensive functioning demonstrated large effects, as did action, obsessional, and high adaptive defense levels (32). However, the two highest defense levels showed differential effects. Increased obsessional defenses were associated with improved functioning more than symptom levels, while high adaptive defenses were associated with improved symptom levels more than functioning. Obsessional defenses allow awareness of facts and acceptance of taking personal action, thus promoting better functioning. The increase in obsessional defenses was also derived from decreases in hysterical and disavowal defenses, which detract from functioning by inhibiting awareness of conflicting ideas and by avoiding appropriating one’s own problems. High adaptive defenses promote the expression of affect and wishes in accordance with one’s interests while adapting to internal and external constraints, thereby relieving emotional contributions to symptoms, whereas obsessional defenses minimize affective engagement. Finally, it is remarkable that improvement in defenses measured moment to moment in therapy is associated with external improvements, including attaining a “mostly satisfied” level of life satisfaction.

This study had several limitations. Because it was a noncontrolled, naturalistic study, we cannot attribute the significant changes we observed to treatment. In fact, the direction of causal influence between change in defenses and change in symptoms and functioning was not determined. It may be that improvement in symptoms and functioning leads to changes in defenses, or alternatively, that some third variable leads to changes in both over time. Future studies should examine whether change in defenses mediates subsequent improvement in functioning, as others have shown for an increase in insight and improved interpersonal functioning (33). This limitation, however, has less influence on our hypotheses regarding the hierarchical order in which defenses change and the relationship to changes in other external measures. For some purposes, diagnostic heterogeneity is a limitation, potentially introducing confounds of treatment response. However, in this study, subject variability allowed the detection of relationships between illness and history variables and rate of change in overall defensive functioning, which homogeneity may have suppressed. While study group size was limited, modeling data from multiple occasions resulted in fairly precise and significant measurements of change, including recovery (19). Finally, our findings are most generalizable to sicker, nonpsychotic patients with anxiety, depressive, and/or personality disorders in long-term therapy.

While other studies have demonstrated that defenses change in therapies ranging from 1 to 12 months’ duration (7, 8, 10), the present study and our previous studies (9, 19) demonstrated that change during longer-term therapy was associated with improvement in both functioning and symptoms over long-term follow-up. This finding is consistent with developments in more than a century’s theoretical and clinical understanding of defenses: they are our first, automatic or spontaneous, adaptive response to threat, and they may play a role in the formation of symptoms (34). Future studies may include replication or controlled trials, such as comparing shorter and longer-term therapies on improvement in defensive functioning and other outcomes. Comparative treatment trials should be conducted to determine whether defenses improve similarly in other treatments, such as cognitive-behavioral therapy. Additional studies should incorporate illness and history predictors to delineate clinically meaningful moderators of rate of improvement, thereby identifying characteristics requiring longer-term therapies. Finally, elsewhere our team has enumerated hypotheses regarding specific therapist interventions that address defensive functioning in-session (35). Preliminary evidence suggests that the process of addressing defenses is related to overall outcome (17, 36). Future studies that examine those processes leading to improved defensive functioning should directly inform good clinical practice.