The randomized double-blind clinical trial (RDBCT) is a key source of information for evidence-based medicine. However, anomalous and unexplainable results have prompted suggestions that ‘unknown and unidentifiable biases’ may exist. This paper identifies that a possible flaw in the implementation of RDBCTs may account for these biases. The flaw relates to the breaking of the double blind through the generation of beliefs and expectations in experimenters. These, in turn, may lead to unconscious biases in assessment and cues to patients. It is then uncertain how much of an observed effect is due to such expectations or the treatment itself. Therefore, any RDBCT in which the maintenance of blinding is not monitored throughout its course is at risk of its conclusions being compromised. It is not sufficient to assert that blinding must have been maintained through arguments based on design features. The burden of proof is on the researchers to demonstrate, through data, that blinding has been maintained. The need to address social psychological issues in implementing RDBCTs is discussed and it is recommended that to avoid this confound, methods of monitoring and accounting for experimenter beliefs and expectations should be routinely included in future RDBCTs.