Skip to content

Response Expectancy and the Placebo Effect

Chapter here.

Kirsch kicks it off by highlighting the beef he has with the general understanding of placebo:

  • There is not just “one” placebo effect - different placebos produce different effects and physiological responses
  • Similarly, there is not just “one” type of expectancy. Response and stimulus expectancies are separate things, and understanding the difference is important.

Ah - this puts it plainly:

  • Stimulus expectancies anticipate external events. (EG - expecting something to be less painful than before.)
  • Response expectancies predict nonvolitional responses. (EG - expecting to feel less pain, despite there being no change in the stimulus.) These are more related to placebo.

Aaaand a rapid fire list of thoughts on response expectancies:

  • They tend to be self-confirming
  • They create/modify non-volitional responses (emotion/subjective perception)
  • While stimulus expectancies can affect perception, response expectancies are “stronger, more stable, and more resistant to extinction” (!)
  • The more ambiguous the stimulus, the more the response set and expectation will change it

Bayesian models are beyond my pay-grade as a hobbyist, but Kirsch highlights that there, predictions can operate outside of awareness. One of the more interesting thoughts here that can be applied to hypnosis is that you don’t need to actively hold on to the thought that influences pain perception in order for it to have an effect.

A study compared stimulus and response expectancies:

  • Light colors were paired to a painful stimulus.
  • In the stimulus expectancy group, they said the light color was paired to the (matching) pain intensity.
  • In the response expectancy group, they said it was paired with analgesia.
  • When testing, the response expectancy group maintained analgesic effects longer.

They cite a study where both subliminal and supraliminal conditioning in a stimulus context seems to work well for modifying pain response.

Separate from measurement, expectancies have at least two dimensions:

  • Expected response magnitude
  • Expected likelihood

Expectancies of administering “a lot” of a placebo follow a quadratic curve. Or, more plainly, in a study where they made people watch them add a metric shit-ton of (unknowingly decaf) coffee to a drink…

  • Small amounts produced a small but measurable physiological response
  • Medium amounts produced a bit more
  • Absurd amounts produced less effect, likely making the dose appear implausible, similar to the law of reversed effect maxim

They assert that measuring expectations is often not complicated.

  • The simplest way is to ask how effective they think the treatment will be.
  • Asking what they think will happen (EG - how much pain do you expect to feel) is a better indicator of outcome, but less correlated with credibility.
  • Expectancies change, even during your study.

The chart from the chapter is useful for understanding, but here’s the gist.

  • Classical conditioning, verbal information (although I don’t see why it’d be just verbal), therapeutic relationship, and modeling all goes into response expectancy
  • Classical conditioning and affect goes into unconscious outcomes
  • The therapeutic relationship and response expectancy affect each-other bidirectionally (EG, if you think it won’t work, it’ll hurt your relationship)
  • Response expectancy and your therapeutic relationship change your affect
  • Affect and response expectancies affect conscious outcomes

In short…

In other words, feeling good psychologically makes you feel good physically.

Irving Kirsch, Chapter Five - Response Expectancy and the Placebo Effect - https://doi.org/10.1016/bs.irn.2018.01.003

Outside of the paper, from an emotional constructivist perspective, I can see physical (interoceptive) information and affect changing each-other. EG, if you feel better physically, you’ll likely feel better emotionally.

In summary:

  • Create high-certainty, small expectations and grow them
  • Frame setbacks as temporary