A placebo is an inert substance used as a control in an experiment or given to a patient for its psychological effects.
The placebo effect is the tendency of some people to feel better when given any new treatment, even if the treatment is inert.
Imagine a simple controlled experiment where one group of migraine headache sufferers is given a new treatment, X, while another group of migraine headache sufferers is given a placebo. The experimental group (the one's treated with X) might claim a 30% lessening of pain. This might seem significant and dramatic but if the placebo (i.e., a control) group claims a 20% lessening of pain, the significance of treatment X is diminished substantially. And if a third group of migraine sufferers is given no treatment (another control group) but claims a 10% lessening of pain over the same time period, the effectiveness of X will be even better understood. A testimonial from any individual in any of the groups would be of no value in determining whether or not to administer X to migraine sufferers. In our example, the effectiveness of X is exaggerated if the control groups are not brought in for comparison. The placebo effect is demonstrated by the difference between those who were given nothing and knew it and those who were given nothing but thought they might have been given something effective (the placebo).
The placebo effect should make it evident that testimonials from people as to the effectiveness of a psychotherapy or an exercise program should be taken with a grain of salt. Unless controlled studies are done, any improvement might as well be attributed to the placebo effect or to the passage of time.
further reading
How to Think Straight About Psychology, 3rd ed., Keith E. Stanovich (New York: Harper Collins, 1992).