I think http://skeptoid.com/episodes/4147 is a wonderful discussion on the subject - particularly talking about the power of suggestion and so on. But pleasently including some facts...
A 2008 study at Hofstra University played two different binaural beats
and a control sound (a babbling brook) to patients with high blood
pressure. There was no difference between the groups. In one small
study from Japan that was published in the Journal of Neurophysiology
in 2006, they played various binaural beats to nine subjects, and
observed the resulting EEGs. They found great variability in the
results. Their conclusion was that listening to binaural beats can
produce activity on the human cerebral cortex, however the cause was
more likely a conscious auditory reaction and was not correlated to
the frequency of the binaural beat. However, a 2005 study published in
Clinical Neurophysiology found that they were able to induce a desired
frequency in the EEG matching the phantom beat frequency encoded in a
binaural beat, however this was with a single subject and was neither
blinded nor controlled.
(The studies are linked in the original)
However I think the really important part is this:
in summary, binaural beats certainly do not work the way the sellers
claim, but there's no reason to think they're any less effective than
any other music track you might listen to that effects you in a way
you like. If they make you sleepy (like they all do for me), use them
to go to sleep. If they relax you or get you amped, use them for that.
But don't expect them to be any more effective than regular music.