How "blind" are double-blind studies?

Psychopharmacological studies usually attempt to eliminate "nonspecific" influences on outcome by double-blind designs. In a randomized, double-blind comparison of alprazolam, imipramine, and placebo, the great majority of panic disorder patients (N = 59) and their physicians were able to rate accurately whether active drug or placebo had been given. Moreover, physicians could distinguish between the two types of active drugs. Inasmuch as correct rating was possible halfway through treatment, concerns about the internal validity of the double-blind strategy arise.