In the late 1950s, when Aaron Beck was a young psychoanalyst at the University of Pennsylvania, he practiced classic Freudian analysis. The goal of therapy, he believed, was to give voice to the repressed urges of the id, revealing those inner conflicts – most of which involved sex - that we hid from ourselves.

But then Beck began treating a patient named Lucy. In a 1997 essay, Beck describes one of their early sessions:

“She was on the couch and we were doing classical analysis. She was presumably following the ‘fundamental rule’ that the patient must report everything that comes into her mind. During this session, she was regaling me with descriptions of her various sexual adventures. At the end of the session, I did what I usually do. I asked her, ‘Now, how have you been feeling during this session?’ She said, ‘I’ve been feeling terribly anxious, doctor.’”

When Beck asked Lucy why she felt so anxious, she gave him an answer that would reshape the future of talk therapy: “Well, actually, I thought that maybe I was boring you. I was thinking that all during the session.”

This offhand remark triggered Beck’s lifelong interest in what he called “unreported thoughts." Although these thoughts are rarely expressed, Beck believed that they shaped our experiences and influenced our emotions. After Beck taught Lucy how to evaluate her negative thoughts – and how to dismiss the incorrect ones – she began feeling better. For Beck, it was proof that talk therapy could work, but only if we talked about the right things.

In many respects, Lucy was patient zero of cognitive behavioral therapy, or CBT. Since that session, CBT has become one of the most widely practiced forms of psychotherapy. This is largely because it works: hundreds of studies have confirmed the effectiveness of CBT at treating a wide range of mental illnesses, from anxiety to schizophrenia. However, CBT is mostly closely associated with the treatment of depression. In part, this is because Beck himself focused on depression. But it’s also because CBT is a remarkably good treatment, at least when it comes to mild and moderate forms of the illness. While direct comparisons with anti-depressant medication are difficult, numerous studies have demonstrated that CBT is about as effective as the latest pills, and might even come with longer-lasting benefits.

That said, many questions about CBT remain. One unknown involves refinements to the practice of CBT, including the introduction of new concepts (schema theory, etc.) and new techniques (mindfulness based CBT and related offshoots). Are these revisions improving CBT or making it worse? Another unknown involves the rapid growth of CBT as a treatment, and the impact of this growth on the quality of CBT therapists. Given these changes, it makes sense to investigate the healing power of CBT over time.

The hope, of course, is that CBT has been getting more effective. Given all that we’ve learned about the mind and mental illness since Beck began studying automatic thoughts, it seems reasonable to expect a little progress. This, after all, is the usual arc of modern medicine: there are very few healthcare interventions that have not improved over the last 60 years.

Alas, the initial evidence does not support the hope. That, at least, is the conclusion of a new paper in Psychological Bulletin by the researchers Tom Johnsen and Oddgeir Friborg. They collected 70 studies of CBT used as a treatment for depression, published between 1977 and 2014. Then, Johnsen and Friborg tracked the fluctuation of CBT’s effectiveness – measured as its ability to reduce depressive symptoms – over the decades. The resulting chart is a picture of decline, as the effect size of the treatment (as measured by the Beck Depression Inventory) has fallen by nearly 50 percent over the last thirty years:

The same pattern basic pattern also applies to studies using a different measure of depression, the Hamilton Rating Scale of Depression:

These are distressing and humbling charts. If nothing else, the decline they document is a reminder that it’s incredibly hard to heal the mind, and that our attempts at progress often backfire. Decades after Beck pioneered CBT, we’re still struggling to make it better.

So what’s causing this decline in efficacy? Johnsen and Friborg dismiss many of the obvious suspects. Publication bias, for instance, is the tendency of scientific journals to favor positive results, at least initially. (Once a positive result is established, null or inconclusive results often become easier to publish.) While Johnsen and Friborg do find evidence of publication bias in CBT research, it doesn’t seem to be responsible for the decline in efficacy. They also find little evidence that variables such as patient health or demographics are responsible.

Instead, Johnsen and Friborg focus on two likely factors. The first factor concerns the growing popularity of the treatment, which has led many inexperienced therapists to begin using it. And since there’s a correlation between the experience of therapists and the recovery of their patients – more experience leads to a greater reduction in depressive symptoms – an influx of CBT novices might dilute its power. As Johnsen and Friborg note, CBT can seem like it’s easy to learn, since it has relatively straightforward treatment objectives. However, the effective use of CBT actually requires “proper training, considerable practice and competent supervision.” There is nothing easy about it.

The second factor is the placebo effect. In general, new medical treatments generate stronger placebo responses from patients. Everyone is excited; the breakthrough is celebrated; the intervention seems full of potential. But then, as the hype gives way to reality, the placebo effect starts to fade. This phenomenon has been used to explain the diminished potency of various pharmaceuticals, from atypical antipsychotics to anti-depressants. Because we are less likely to believe in the effectiveness of these pills, they actually become less effective. Our skepticism turns into a self-fulfilling prophecy.

Johnsen and Friborg speculate that a similar trend might also apply to CBT:

“In the initial phase of the cognitive era, CBT was frequently portrayed as the gold standard for the treatment of many disorders. In recent times, however, an increasing number of studies have not found this method to be superior to other techniques. Coupled with the increasing availability of information to the public, including the internet, it is not inconceivable that patients’ hope and faith in the efficacy of CBT has decreased somewhat…Moreover, whether widespread knowledge of the present results might worsen the situation remains an open question.”

It’s an unusually meta note for a scientific paper, as Johnsen and Friborg realize that their discovery might influence the very facts they describe. After all, if patients believe that CBT is no longer the "gold standard" then its decline will accelerate. And so we are stuck with a paradox: we cannot study the power of mental health treatments without impacting future results. Belief is part of the cure.

Johnsen, Tom J., and Oddgeir Friborg. "The Effects of Cognitive Behavioral Therapy as an Anti-Depressive Treatment is Falling: A Meta-Analysis." Psychological Bulletin (2015).