3.4: Measure the Behavior with a Reliable Data Collection System

Behavioral means that the behavior itself is the subject matter, so the dependent variable must be operationalized in observable and measurable terms with the goal of having inter-observer reliability, or a high percentage of agreement between multiple observers that the same behavior occurred or did not occur during observations. In order for direct observations to occur, there must be a precise definition of the behavior and a measurement system to record the behavior. Choosing a measurement system depends on how the behavior is exhibited (Alberto & Troutman, 2003). For example, the behavior may be problematic because of how often it occurs (i.e, too many times or not enough times) or the length of the response (i.e, too short or too long). When the number of occurrences of the behavior is the problem, one would choose an event-recording system if the behavior has a specific start and end in order to teach occurrence, like the number of times a student talks out of turn during instruction time. If the response does not have a distinct beginning and end, use an interval recording system. When the length of time the behavior occurs is the problem, choose duration recording to measure the span of each occurrence. Like duration, latency measures time; however in this measurement, one measures the elapsed time after a signal and the onset of the behavior. For example, one could measure the amount of elapsed time between when the teacher gives the direction and when the student starts the first problem.

The context in which the behavior is exhibited provides a picture of when the behavior occurs or does not occur in relation to other events (Johnston & Pennypacker, 1993). Observing the changes of behavior requires knowledge of an analytic approach. To be analytic requires a believable demonstration of the events that caused the change in behavior. These changes of behavior are empirical demonstrations of functional relations between antecedent events, behavior, and consequence events. There are different degrees to this level of analytic behavior change. At the highest level, research methodology using single-subject research designs (i.e., reversal, multiple-baseline line) is used to determine if functional control is established, verified, and replicated (Kennedy, 2005). A less-rigorous level still includes demonstrations of prediction and control (i.e., baseline or phase with no intervention, intervention phase, intervention removal, and reintroduction of the intervention) but may have lower levels of inter-observer reliability of data-collection procedures. The value of the data depends on the accuracy and reliability of the data to make decision about which intervention to choose, continuation of the procedures, or when to stop the intervention.