Placebo groups in trials of depression and anxiety treatments improve substantially, often approaching the improvement in the treated group. Several distinct mechanisms contribute, and only one of them concerns expectation.
The outcome is a rating, not a measurement
Blood pressure is read from an instrument. Mood is assessed by asking, either through a self-report questionnaire or a clinician-administered scale.
Ratings are influenced by how the question is framed, by the relationship with the assessor, and by what the person believes is expected of them.
An outcome measured this way has more room to move without any change in the underlying condition than a physical measurement does.
Regression to the mean is built into recruitment
People enter trials when their symptoms are bad, because that is when they seek help and when they meet severity criteria for entry.
Symptoms fluctuate around a personal average, so a measurement taken at a peak is likely to be followed by a lower one regardless of intervention.
This statistical effect operates in both arms of a trial and requires no biology at all, and it is one reason a control group is necessary.
Many episodes improve without treatment
Depressive and anxiety episodes are frequently self-limiting over months, following their own course.
Trials run over weeks, so natural recovery in some participants overlaps with the observation period and appears as improvement in both groups.
Separating natural course from treatment effect is exactly what the comparison is designed to do, and it means the treatment must beat a moving baseline.
Trial participation is itself an intervention
Being in a trial means regular appointments, structured attention from clinicians, and being asked systematically about one's state.
Those are non-trivial exposures for someone who was previously isolated or unmonitored, and they resemble components of some therapeutic approaches.
The placebo arm receives all of this, so it is not an untreated group in any meaningful sense.
What follows for interpreting results
A large placebo response makes it harder for a trial to demonstrate a difference, so a modest measured effect does not necessarily mean a modest real one.
It also means individual improvement after starting anything is weak evidence about that thing, since several mechanisms would have produced improvement anyway.
Decisions about treatment for these conditions depend on individual history and require a clinician, and stopping a prescribed treatment because of general reading about trial results is a decision to discuss rather than to make alone.