Many studies have shown that CBT is effective for treating depression and anxiety disorders. The studies are convincing, and the effect sizes have been large, usually comparable to medication treatments.
CBT studies are also usually well-designed. The therapy itself is very clear. While some complain that a "manualized" therapy is too mechanical or detached, it is true that a very standardized therapy approach allows a much more reliable scientific study. A less structured therapeutic style would be expected to show much more variability between one therapist and the next, or between one patient and the next. This fact does not mean that standardized, manualized therapies are superior to less structured types, but it does mean that the standardized varieties give more meaningful research results showing without any doubt whether a psychological therapy works or not.
I believe that CBT is a type of "fitness training" focusing on psychological symptoms and goals. The CBT therapist is an educator and coach. Actual CBT sessions are analogous to having a workout with a personal trainer.
Just as with the literal situation of seeing a personal trainer, perhaps two or three times a week for 6 weeks, one could have a lot of fitness gains from the sessions alone.
But most fitness gains--especially for skill-related activities such as learning tennis, skiing, skating, dance, or bowling--happen as a result of the hours of dedicated, earnest daily practice. These practice hours would take place between training sessions with the coach!
Similarly, there is some improvement in symptoms due to CBT sessions alone. But most of the gains, in my opinion, will occur as a result of focused daily practice and homework between the sessions.
Most CBT research does not clearly indicate the number of hours of practice the patients have done, and do not have any measure of the quality of the practice done. Just as with children doing homework activities, it matters how much time is spent, but it matters even more how good the quality is. Was the work done in a sloppy, bored, rushed, haphazard manner, or was there evidence that the work was done with care, attention, organization, and devotion?
Similarly, very little behavioural therapy research has looked specifically at exactly how long an exposure task needs to be in order to produce an optimal effect.
In my own look at these topics, I have reached the following conclusions:
1) Daily homework of high quality is necessary for CBT to work best. This is no different from getting good results in a university class, or following music lessons.
2) Exposure tasks (which I believe are an essential part of all CBT) need to last 20 minutes in order to be most effective. Many clinics advise 45-90 minutes at a time. The difficulty of the exposure task has to be adjusted so that it is moderately challenging (not too easy, and not overwhelming), with some feeling of mastery when it is over. Just as with physical workouts, at least 3-5 exposure tasks per week should be a goal.