Basically, a double-blind study is one in which neither the subjects nor
the researchers know which treatment a particular subject is receiving;
the information is kept secret and is not revealed until the end of the
study.
Double-blind studies arose as a necessary complement to
placebo-controlled studies. Doctors have known about the
placebo effect for years--if you tell a patient that a drug
is effective, he'll often seem to get better, even if you've really give him
a sugar pill. Therefore, when you test a new drug, you have to give half
the patients the drug and half of them a placebo.
Now, it turns out that if the researchers know which treatment
a patient is receiving, they may start to treat that patient differently.
That difference may in turn evoke the placebo effect and make the drug
seem effective when it really isn't. For example, if a doctor or nurse
knows that a patient is receiving the real drug, not the placebo, they may
look harder for positive effects or tell the patient that the drug looks
like it's really working. Conversely, if they know he's receiving the placebo, they may downplay any positive effects that the patient mentions.
These responses aren't necessarily deliberate; a doctor or nurse can
reveal the truth simply through his or her tone of voice.
In medical research, new drugs are virtually never accepted unless they
are tested in a strict double-blind experiment. Most drug companies go
along with this principle, although I've heard stories in which some
companies try to circumvent this in crude ways (e.g. by giving
researchers nicotine patches with the word "PLACEBO" or
"REAL DRUG" printed on them.)