Dawes on Therapy

I finally had a chance to read Robyn Dawes’ book House of Cards, recommended to us here by Michael Crichton.  Here Dawes summarizes evidence on the effectiveness of psychotherapy: 

Randomized experiments evaluating the efficacy of psychotherapy began appearing occasionally in the scientific journals during the 1960s. … In 1977, Mary L. Smith and Gene V. Glass published a famous article in American Psychologist that concluded that psychotherapy is very effective.  They summarized the results of 375 studies of psychotherapy effectiveness that had purported to use random assignment to experimental, and control groups. …

Smith and Glass found that someone chosen at random from the experimental group after therapy had a two-to-one chance of being better off on the measure examined than someone chosen at random from the control group.  That is a very strong finding … Smith and Glass’s meta-analysis … concluded that three factors that most psychologists believed influenced this efficacy actually did not influence it. 

First, they discovered that the therapists’ credentials – Ph.D., M.D., or no advanced degree – and experience were unrelated to the efficacy of therapy.  Second, they discovered that the type of therapy given was unrelated to its effectiveness, with the possible exception of behavioral techniques, which seemed superior for well-circumscribed behavioral problems.  They also discovered that length of therapy was unrelated to its success.   …

The professional psychology community hailed Smith and Glass’s overall finding but not the three subsidiary findings.  A series of studies was thereafter conducted to indicate that at least the first finding was inaccurate.  But these studies failed to refute Smith and Glass.  I became involved in the field of meta-analysis after reading Smith and Glass’s paper.  At the time I was skeptical of it. … in the academic year 1978-79, Janet Landman … and I collaborated … to check out … Smith and Glass.  … Much to our surprise, our results were virtually identical to those of Smith and Glass.

Apparently an intelligent sympathetic ear is mainly what you need if you feel a bit crazy.  The book has much more worth pondering.  But I fear many take the wrong lesson here, thinking they knew all along something as "unscientific" as psychotherapy couldn’t work.  In fact, few experts have faced such careful scrutiny of their effectiveness.

GD Star Rating
Tagged as:
Trackback URL:
  • Caledonian

    But I fear many take the wrong lesson here, thinking they knew all along something as “unscientific” as psychotherapy couldn’t work.

    Oh, psychotherapy works, all right. But the claims about why it worked were wrong, and as a result practice was forced to conform to a variety of crazy systems that had no relationship to what was actually helpful. (Hint: friendly shoulder that’s willing to listen)

    Even a blind squirrel will find a nut now and then, but I wouldn’t bet on it.

  • Seth Roberts had an interesting post on the subject of therapy’s efficacy, which ties in nicely with this post:


    Seth seems to be going in the direction that telling your story in therapy is the important factor. It seems telling your story in a journal (as evidenced by a study on journal writing), about your experience, is therepeutic, and so story-telling might be the common factor between therapy and journal-writing of a type. He also ties blogging in, which is neat!

  • This suggests to me that psychotherapy should be more like tech support. You call a hotline and some guy in India listens to you complain. It would be a lot cheaper than going to a psychiatrist, and perhaps nearly as effective. Of course, unlike tech support, they would have to not keep you on hold, because that would just drive you crazier.

  • Constant

    Ha! I knew it all along!


  • Constant

    Have they tested the effectiveness of a simple computer program that creates the illusion of a sympathetic ear? Eliza, for example?

  • If I recall correctly – I think it’s from Dawes? – studies also showed that talking to a math professor works just as well as talking to a psychotherapist. There may even have been a similar result for talking to random teenage girls.

    • Chris Amidon

      Eliezer, The math professor study was Strupp and Hadley, (1979) “Specific vs Nonspecific Factors in Psychotherapy.”

  • Constant

    What about talking to an invisible, intangible, and inaudible person who allegedly has infinite understanding and love? Has the healing power of prayer been tested?

  • manuelg

    Talking to random teenage girls will increase your life problems, not decrease. Especially if the wife finds out.

  • cwurden


    That people do better when they think others care about them is old news to social scientists .

    The famous ‘Hawthorne Studies’ of the 1920’s-1930’s documented such a common, placebo-like effect.

    These studies of Chicago factory workers were scientifically attempting to improve worker productivity with various physical changes (better lighting & temperature/humidity in work areas, wage incentives, etc.) — but researchers discovered individual worker productivity generally increased no matter what they changed or did not change, even in the unaffected control groups. The workers markedly benefited just by knowing others were directly interested in their work and were trying to help.

    Harvard Professor Elton Mayo led this research and his “Hawthorne Effect” is a classic in professional psychology.

    The significant human variable was not physiological but psychological– and surprisingly easy to improve by simple interaction with other humans who care.

  • Robert Lindsay discusses his own mental illness and others who share it that state they benefited from discussing it with him here. There are some funny stories there.

  • Jor

    just an fyi, there now are therapies that are more effective (by trial data) than just a shoulder to cry on for particular mental disorders. Cognitive-behavioral therapy for depression, exposure therapy for some anxiety disorders, and dialectic behavioral therapy for border-line personality disorder.

  • The evidence for the Hawthorne effect is pretty weak (see Wikipedia), and the belief that the Hawthorne Studies demonstrated such an effect involves a good deal of confirmation bias.

  • Pingback: Overcoming Bias : What Purpose Self Insight?()

  • Ted Horoschak

    >had a two-to-one chance of being better off on the measure examined than >someone chosen at random from the control group. That is a very strong >finding …

    As stated here, this doesn’t seem like a strong finding to me. We’re told nothing about how strong anything is. The 2:1 says zero about the intensity (other than it was detectable by whatever measure used) of better-offedness. Were they suicidal and it saved their lives (pretty strong), or did they feel a little better for an hour (trivial)? What was the “measure”, and why would anyone care about it? A 2:1 chance of having a trivial effect would say nothing about the effectiveness of psychotherapy.

    Is this a problem with the summary, or with the study itself?

  • Pingback: Psychotherapy and Evidence – Links | Soto van Cavalera()