A performance baseline provides the standard against which we measure the effectiveness of our intervention. A change following an intervention relative to the baseline level of the behavior signals the intervention’s utility. No change sends us back to the drawing board. In the immediately preceding Behavior Watch commentary, we discussed the criteria by which we might judge the stability of a behavioral baseline. A stable baseline is an important element in assessing an intervention’s effectiveness, but there are other things about the baseline that we need to consider, too. Three other important questions to be answered when using baselines are: (1) How might “participant reactivity” affect the baseline measurement? (2) How long should the baseline remain in place prior to an intervention? and (3) Do we need a post-intervention return to the baseline?
Sometimes just the act of recording our own behavior or that of others changes the behavior independently of any intervention. If you know someone is monitoring your weight or your frequency of social interactions, how might this affect such behavior? When others are observing us, we tend to mind our P’s and Q’s more than we do when not being observed. At least for a while. Then, we often (but not always, of course) habituate to being observed and we get back to our same old ways of behaving.
How long should the baseline remain in effect? The baseline needs to be long enough that we know the behavior has reached a steady state so that it isn’t changing wildly from day to day. In other words, until the baseline is stable as discussed in the previous commentary on baseline performance. If the behavior has never been observed or has been observed very infrequently, then a short baseline condition may be in order. If, however, behavior fluctuates considerably from observation period to observation period, then the baseline needs to be long enough that the variability drops to a manageable level or we are satisfied that considerable fluctuation is the baseline state. Bottom line: A baseline should be long enough to convince both you and others that the behavior is not changing in any predictable or orderly way.
Removing an intervention is the best way of showing that the behavior change was due to the intervention and not something else, like the simple passage of time. That said, however, in many situations involving human participants it is ill-advised, and, indeed, unethical to remove an intervention that is helping someone. Sometimes the intervention will be removed to show that the behavior reverts to baseline levels in the absence of the intervention, but then the intervention is thereafter reinstated so the participant can continue to benefit from it. At other times, other methods are used to show that it was the intervention that was responsible for the behavior change. These include a special design called a multiple baseline design and certain statistical analyses created specifically for determining the reliability of change across the baseline-intervention, or A-B, cycle.
Establishing a proper baseline is an essential first step in evaluating any intervention in any setting. Without it, there is no way to determine whether what we are doing to make things better is really doing so.