Perhaps the greatest problem faced by the academic social sciences is that what is measurable is often irrelevant, and what is truly relevant often cannot be measured. Sigmund Freud’s mechanisms of defense are an excellent example of the latter.

In 1894, the year that Freud branched out into psychiatry, he was still a hard-nosed scientist—he was a practicing neurologist and the secretary of the Society of German Naturalists and Physicians. Yet, that very year, he was hard at work describing defense mechanisms for the first time (1). Freud observed not only that affect could be “dislocated or transposed” from ideas (by the unconscious mechanisms that he would later call dissociation, repression, and isolation), but also that affect could be “reattached” to other ideas (by the mechanism of displacement). Freud also noted that subject and object could be cognitively reversed by the process he called projection.

The years 1894–1896, during which Freud published his two seminal papers on defenses, were the most anxiety-ridden of Freud’s life, which had a profound impact on the focus of his work at the time. By 1897, however, he had directed his focus elsewhere (very possibly as a defense mechanism), and his list of defenses went virtually ignored for a quarter of a century. In 1926, Freud’s “repression” lifted, and he wrote, “I have revived a concept . . . of which I made exclusive use thirty years ago when I first began the study of the subject [of anxiety] but which I later abandoned. I refer to the term ‘defensive process’ . . . it will be an undoubted advantage, I think, to revert to the old concept of ‘defense’” (2). In the same book, Freud pointed out the utility of distinguishing the defense of isolation from repression, a distinction he had clearly defined in 1894 but which in 1926 he asserted “we are setting out to describe . . . for the first time” (2). I suspect that the history of DSM-III thru DSM-6 will echo Freud’s struggle with his brilliant discovery.

Over a period of 40 years, Freud and his daughter outlined most of the defense mechanisms that we speak of today and identified five of their important properties:

• Defenses are a major means of managing conflict and affect.

• Defenses are relatively unconscious.

• Defenses are discrete from one another.

• Although often the hallmarks of major psychiatric syndromes, defenses are reversible.

• Defenses are adaptive as well as pathological.

Defenses, like rainbows and shooting stars, exist but are difficult to capture. Defenses are, after all, metaphors; they are very complex affective and cognitive styles that the brain uses to alter conflictual inner and outer realities. Like creativity, defense mechanisms reflect integrated mental processes that cannot be broken into component parts, precisely measured, and uniformly labeled. Thus, like creativity, defenses have not yielded easily to rating scales (3).

In the fall of 1977, to plan for DSM-III, Bob Spitzer called a group of psychoanalysts together in New York City to plan for a possible axis VI (defenses and coping) for the new DSM-III. After several hours, it was clear that we were unable to reach consensus on a list of the important defenses or to agree on their definition or their significance for psychopathology. By the time DSM-III went to press, the world of defenses resembled the Tower of Babel. Within a 30-mile radius of San Francisco, five different investigators—Horowitz, Haan, Moos, Lazarus, and Vaillant—had each developed influential nomenclatures for defenses. All were different, and no investigator had yet tried to translate his own hierarchy into those of the others.

By including an optional axis and a glossary for a hierarchy of defenses, DSM-IV achieved the consensus that had been unobtainable for DSM-III. Chris Perry (4, 5), Andy Skodol, and Allen Frances deserve enormous credit for achieving this consensus. However, rumor has it that for DSM-5, defenses will again slip into the mists. Perry and Bond’s methodologically complex but elegant paper in this issue (6) and the increasing promise of neuroimaging (briefly reviewed in reference 7) offer hope that the pendulum will swing back for DSM-6.

Over the past 40 years, I have been content to demonstrate that in adult life a hierarchy of defenses exists; that if organic insults to the brain are avoided, the hierarchy matures over the adult life span; and that maturity of defenses powerfully predicts the future (8, 9). However, I have failed to communicate how clinicians might reliably identify these evanescent defenses and how clinicians might catalyze their patients’ maturation of defenses. In contrast, over the years Perry and his team, by perfecting the research art of quantitative clinical observation, have established progressively more reliable ways of rating defenses (10). Moreover, they have shown that defenses improve with psychotherapy and that this change is significantly associated with clinical change (11).

The article by Perry and Bond in this issue (6) is the capstone to Perry’s ongoing efforts. Using a naturalistic study of 21 patients with severe depression and/or personality disorders who had a mean Global Assessment of Functioning score of 52.9 and had been treated for a median of 228 weeks, the authors demonstrate that improvement in defenses in the first 2.5 years was associated with significant improvement in external measures of life functioning and of symptoms 5 years later. They also demonstrate that change in defenses does follow a meaningful progression, as predicted by the hierarchy; furthermore, even modest improvement in defenses is associated with sizable improvement in real life. Throughout this longitudinal study, Perry’s team has consistently delivered interrater reliability to burnish the long-term validity of their hierarchy of defenses. How many studies of psychoanalysis can make such a claim?

With this paper, Perry’s group—alas, too late for DSM-5—meets the hopes entertained for the optional axis on adaptive (defense) mechanisms offered by DSM-IV and shows that dynamic psychotherapy may in part mediate changes in both symptoms and functioning.

The study does have two notable limitations. The first, which the authors acknowledge, is that in any 5-year study a control group is needed to distinguish association from causality. Might not the changes observed be explained by normal maturation and by regression to the mean in a group of patients who were severely ill at the start of long-term psychotherapy? The second limitation is that to be clinically useful and experimentally simple, one needs to apply Occam’s razor. Any consensually accepted hierarchy of defenses should probably be limited to three or four levels and to 10 to 15 widely accepted adaptive mechanisms and not the seven levels and 30 mechanisms utilized by Perry’s group and by DSM-IV. However, only time will tell.