Double-Blind Procedure

Blind experimental procedures are designed to eliminate biases produced by expectations of study participants, experimenters, and data analysts. Investigators commonly employ any of three levels of blindness in their experimental designs. In a simple [Page 285]blind experiment, participants are unaware of their assignment to the treatment or to the control group. In a double-blind experiment, both the participant and the experimenter are unaware of the group assignment. Finally, in a triple-blind experiment, the participants, experimenter, and those responsible for data entry and analysis are all unaware of the treatment versus control condition of each study trial.

Blind procedures are necessary because participants respond not only to the experimental stimulus but also to their impression of the experiment's hypothesis. If the manipulation of the experiment is readily apparent, participants may alter their response in accordance with what they perceive to be the desired response. In this case, the measured effect of the experiment represents an unknown combination of intended treatment effects and the demand characteristics of the experimental setting.

A more subtle source of bias lurks in the interaction between the experimenter and the participant. When an experimenter expects a certain response from participants in one condition, he or she may artificially generate or enhance this response by influencing these participants with (often unintentional) nonverbal cues, extra encouragement, or heightened attention. The bias that results from these behaviors is called the experimenter expectancy effect. Because participant and experimenter biases are systematically related to expectations of the experimental manipulation, they have the potential to confound the interpretation of experimental results. The same type of expectancy effects may lead the people handling the data to consciously or unconsciously skew the results in an expected direction.

The double-blind procedure, which is also referred to as a “masked condition” or an “experimentally naïve” procedure, is often used in medical research. For example, in a drug trial, a sugar pill identical in appearance to the experimental pill is administered to the control group. The experimental and sugar pill treatments are randomly assigned by an investigator who does not interact with participants. The identity of the pills keeps both the participants and the experimenters who actually administer the drugs blind to the nature of each experimental trial. Double blind procedure has the added benefit of preventing those administering the treatment from succumbing to the temptation to give needier patients the experimental drug rather than the placebo, thereby distorting the results.

Many social science investigations involve treatments that, unlike a pill, cannot be engineered to look the same to treatment and control subjects. Consider a study comparing the effects of telephone versus e-mail messages on attitude change; here, both participants and experimenters are fully aware of whether the study involves a phone call or an e-mail. In such instances, double blindness is approximated by keeping both participants and experimenters unaware of the hypotheses, or by minimizing contact between participants and the experimenter who knows the hypothesis. For example, a third party can administer the treatment before the experimenter collects data on the dependent variable. In sum, a “blind” protocol helps the investigator to minimize systematic human bias that is introduced when participants, experimenters, and data analysts are aware of the experimental treatment. Double-blind procedures afford the investigator more confidence when drawing causal relationships between the treatment and the outcome variable.

...