USGS Home
Contact USGS
Search USGS

The term control, confusingly, has at least 3 different meanings in experimental design. The first meaning, which is more general and not specifically addressed here, involves the investigator's role. In a controlled study, the treatment (cause) is assigned by the investigator; the study is an experiment. In an uncontrolled study, the treatment is determined to some extent by factors beyond the investigator's control; the study is observational (Holland 1986). The second meaning, design control, implies that, while some experimental units receive a treatment, others (the "controls") do not. The third meaning, statistical control, means that other variables that may influence the response are measured so that we may estimate their effects and attempt to eliminate them statistically.

The major benefit of design control is to provide a basis for comparison between treated and untreated units. It reduces the error; our measured response is likely to reflect only the treatment rather than a variety of other things. Statistical control usually is less effective in reducing error and is applied after treatments are applied.

Sometimes strict design controls are not possible. Intervention analysis and BACI designs can demonstrate that some variables may have changed subsequent to the intervention or impact, but one will be less confident from that analysis that the intervention caused that change. Confidence will increase if potential confounding variables are measured and their effects are accounted for during analysis—that is, through statistical control.

Controls should be distinguished from reference units. The latter are units that represent some ideal that management actions are intended to approach. Reference sites are especially useful in restoration ecology, when evaluating the effectiveness of alternative management activities for restoring degraded areas to conditions embodied in the reference sites (Provencher et al. 2002).