The importance of having the investigator blind to the experimental condition of a human subject was demonstrated using meta analysis of 43 reports of the effects of carbon monoxide (CO) on behavior. It was shown that 75% of single-blind studies found significant CO effects as opposed to only 26% of double-blind studies (difference was significant, p < 0.005). It was also not possible to show that the difference could have been due to different CO exposure levels, different statistical practices, or different study group sizes. The failure to follow double-blind procedure has been partly responsible for disagreement about the effects of CO reported in the literature. Investigator blinding is also important in laboratory animal research. (Copyright (c) 1993 Pergamon Press Ltd.).