Maintenance

Description of the Strategy

Since the 1960s, the disciplines of applied behavior analysis (formerly known as behavior modification) and behavior therapy have developed a number of highly effective behavior-change strategies. Many of these strategies, most of which are based on the principles of operant or respondent conditioning, have been effectively disseminated such that they are now readily available to clinicians. A fundamental aspect of behavioral treatment, and perhaps its most important feature, is the maintenance of behavior change. Maintenance is defined as the continued performance of an appropriate target behavior (e.g., communication) or the continued reduction of a problem behavior (e.g., aggression) after treatment has been withdrawn. Maintenance can also be conceptualized as behavior that is resistant to extinction and thereby more likely to persist in the absence of reinforcement. A better understanding of maintenance may be illustrated by contrasting it with the related process of generalization.

There are two types of generalization: stimulus and response. Stimulus generalization occurs when a behavior learned under one set of circumstances (e.g., in an outpatient clinic with a therapist) occurs at other times, in novel settings, and/or in the presence of different people. For example, stimulus generalization is said to occur when an unassertive and shy child is taught by a therapist to be more socially assertive, and subsequently demonstrates those same skills at school with peers. When stimulus generalization occurs, the trained response occurs in different situations. Alternatively, response generalization occurs when different, but functionally similar, responses occur after training. For example, response generalization might occur when the child mentioned earlier uses assertive phrases that were not directly taught but are, nonetheless, equally effective in social situations. Both stimulus and response generalization are important considerations when changing behavior because it is frequently clinically advantageous for multiple variations of a target behavior to occur across a variety of situations. Although maintenance has sometimes been referred to as “temporal” generalization because it refers to the durability of behavior change over time, it is important to consider it apart from stimulus and response generalization because maintenance requires its own clinical strategies.

Successful behavior change is a prerequisite for maintenance. Obviously, if a behavior is not effectively changed, there will not be anything to maintain over time. Although generally very effective, most behavioral treatments cannot be continued indefinitely and thus must eventually be withdrawn. Of course, despite the withdrawal of some or all aspects of a treatment, behavior change still needs to be maintained. For example, a teacher who uses preferred snacks and toys to reinforce student compliance must at some point stop delivering those rewards. However, if the teacher abruptly withdrew treatment, compliance would likely worsen. Maintenance generally will not occur unless certain variables are addressed during the behavior-change process. In other words, maintenance must be explicitly programmed. Such programming can occur when behavior has already been [Page 898]changed or, better yet, during the behavior-change process before treatment has been withdrawn.

At least four strategies can be used to program maintenance. The first strategy, the use of indiscriminable contingencies, is actually a group of methods that produces a common goal: the gradual elimination of training contingencies such that the process is undetectable. The primary method by which this is accomplished is the use and thinning of intermittent reinforcement schedules. Behaviors that are intermittently, or periodically, reinforced (i.e., consequences are provided for some, but not all, of the behaviors) are generally more resistant to extinction (i.e., nonreinforcement) than behaviors that are continuously reinforced. Although it is generally recommended to first use continuous reinforcement when teaching a behavior, after it has been acquired the schedule can be gradually thinned, thus making the behavior more resistant to extinction. For example, a continuous schedule could first be thinned by reinforcing every other behavior, every third behavior, approximately every fifth behavior, and so on. In addition, the schedule might be further thinned until reinforcers are no longer delivered. A second method involves gradually increasing the delay between the behavior and its reinforcer. This can be used instead of or in addition to reinforcement-schedule thinning. For example, after a behavior has been taught using the immediate delivery of reinforcers, the time between the behavior and reinforcer could be gradually extended over time (e.g., from 1 to 60 seconds), possibly until the reinforcers are eventually removed altogether. A third method for creating an indistinguishable contingency is to gradually over time deliver the programmed consequence independent of (e.g., every 5 minutes) and contingent on the target behavior. Thus, the likelihood of the child receiving these consequences independent of the target behavior would eventually increase. However, one should take great care not to introduce response-independent reinforcers too quickly or the target behavior will likely extinguish. The three aforementioned methods for creating indistinguishable contingencies should always be incorporated with the use of natural reinforcers (described below) such that when programmed consequences are gradually removed, natural contingencies will remain to maintain the behavior. The fourth method for creating indistinguishable contingencies applies to the gradual withdrawal or thinning of treatments designed to reduce problem behavior. This is sometimes referred to as treatment fading. For example, with a differential reinforcement of other behavior (DRO) procedure, a reinforcer is delivered when a problem behavior has not occurred for a specified period of time. This time period can be gradually extended over time (e.g., from 5 to 30 minutes) such that the DRO treatment is easier to implement but still effective at maintaining the behavioral reduction. Thinning procedures have also been developed for several other reductive treatments, including functional communication training and non-contingent reinforcement.

...