A major objective in designing a randomized clinical trial is to minimize bias in the assessment of the study's primary outcome. Although we often think of bias as referring to the situation where a difference between treatment groups is found that is not real, bias also can prevent a study from finding a true treatment group difference. The former situation is referred to as differential bias and the latter as nondifferential bias.
Nondifferential bias can be thought of as “noise” that affects both the intervention group and control group equally. With respect to visual acuity testing, nondifferential bias is minimized by making the visual acuity measurement as precise as possible. Differential bias refers to the situation where there is a systematic difference in how visual acuity is measured in one treatment group compared with the other, related to either how the test is administered or how the study subject responds, that influences the results. Systematic differences of this type can only exist when either the study subject, the visual acuity tester, or both have knowledge of the subject's treatment group assignment. Thus, differential bias is avoided by having the study subject and the individual administering the visual acuity test unaware of the treatment group assignment of the subject (or of the eye, in studies that include both eyes of a subject). In most disciplines, this is referred to as “blinding,” but in eye disease studies, for obvious reasons, it commonly is referred to as “masking.”
Beck RW. To Mask or Not to Mask. Arch Ophthalmol. 2009;127(6):801-802. doi:10.1001/archophthalmol.2009.109