Not a great article. It omits several relevant known mechanisms. The placebo effect is NOT "all in the mind"; some of it is (and some of the data in the article argues to that), but some of the placebo effect is "in the math" and some is "in the pill" and some of it is "in the measurement".<p>IN THE MATH: For depression in particular, regression toward the mean is huge - people tend to seek treatment when they are <i>unusually depressed</i> and thus tend to feel better at some random later time regardless of whether they get any treatment at all. You can verify this by having two control groups, one of which gets NO treatment rather than a fake treatment. Or assign different groups to start treatment at different times after a variable delay.<p>IN THE PILL: There's really no such thing as an "inactive" pill. The pill is made of <i>something</i>, and that something - even if it's milk sugar - might by chance actually do something useful to the condition you're testing. To rule this out you also need a no-treatment control - simply calling something an "inactive control" doesn't make it one.<p>IN THE MEASUREMENT: If you apply "a treatment" and ask people "is this better now?" the polite ones will say "Yes, I feel a little better" just to be polite, even if they don't. To tease this out you need objective rather than subjective measurements of success.