There is a report titled "Educing Information: Interrogation: Science and Art
Foundations for the Future" prepared by the National Defense Intelligence College under the auspices of the Defense Department in December 2006. This report sought to evaluate alternative approaches for interrogation. A copy may be found at
http://www.fas.org/irp/dni/educing.pdf. It notes that most professionals believe that "harsh" interrogation is not a reliable method for gathering intelligence because of the difficulty of assessing the accuracy of information in a context where the subject will say anything to halt the pain. The same study also notes that there is very little scientific evidence of any kind concerning the efficacy of torture -- either showing it works or proving it doesn't.
Note that the problem is not with getting people to confess; that is easy. They will confess to anything. Maybe this is what has historically made torture popular with members of the inquisition and with some police interrogators. When all you care about is getting a confession, and "truth" is a lower priority, torture works just fine.