Differential Reinforcement of Alternative Behavior
Monet013
OnthedefinitionofDRA.docx
Journal of Applied Behavior Analysis
Volume 53, Issue 3 p. 1299-1303
Comment
Full Access
On the definition of differential reinforcement of alternative behavior
Timothy R. Vollmer , Kerri P. Peters , Faris R. Kronfli , Lindsay A. Lloveras , Vivian F. Ibañez
First published: 20 March 2020
https://doi-org.sagecolleges.idm.oclc.org/10.1002/jaba.701
Citations: 18
Abstract
Differential reinforcement of alternative behavior (DRA) has a long history as a behavioral treatment. The term has usually been defined in a manner that suggests one form of behavior (usually some appropriate alternative) is reinforced, while another form of behavior (usually problem behavior) is placed on extinction. In this paper, we will suggest that DRA should not require extinction, either procedurally or by definition. Ideally, problem behavior would be placed on extinction when possible; however, when problem behavior is not or cannot be placed on extinction, the procedure used is still DRA. Thus, we propose the following definition: Providing greater reinforcement, along at least one dimension, contingent on the occurrence of one form or type of behavior, while minimizing reinforcement for another form or type of behavior.
Differential reinforcement of alternative behavior (DRA) has been used as a behavioral treatment essentially since the inception of applied behavior analysis. For example, Zimmerman and Zimmerman ( 1962 ) increased appropriate classroom behavior and decreased problem behavior in two “emotionally disturbed” students. They provided praise and smiles when appropriate behavior occurred, and “ignored” problem behavior. Other early studies used DRA procedures to treat self-injurious behavior (SIB) and classroom disruptive behavior (Allen & Harris, 1966 ; Hall et al., 1968 ). Although differential reinforcement is applicable not only for treatment of behavior disorders, the focus of the current discussion is on its usage as a behavioral treatment. Nonetheless, our proposed definition will provide an opportunity for broader usage in the behavioral literature.
Currently, DRA is commonly used as a treatment following a functional analysis. As Vollmer and Iwata ( 1992 ) described DRA with attention as the putative reinforcer:
Arranging a differential reinforcement procedure…essentially involves a contingency reversal. That is, after demonstrating the reinforcing effect of attention in baseline, treatment would involve no longer providing attention contingent on inappropriate behavior (extinction) but rather on its absence (DRO) or after some alternative response (DRA) (p. 404).
A common application of DRA is functional communication training, where some appropriate communicative response is reinforced using the reinforcer that previously maintained problem behavior (Carr & Durand, 1985 ; Tiger et al., 2008 ).
Many definitions of DRA either directly refer to, or at least allude to, an extinction component. For example, Vollmer and Iwata ( 1992 ) also stated, “An extinction procedure (or in some cases a variation of timeout) is a necessary component of differential reinforcement when using functional/relevant reinforcing consequences. This is so by definition because behavior targeted for reduction cannot produce these reinforcing consequences” (p. 403). Even popular contemporary textbooks contain definitions that include an extinction component. For example, Miltenberger ( 2018 ) states that DRA is “a procedure in which a specific desirable behavior is followed by a reinforcer and other behaviors are not. The result is an increase in the desirable behavior and extinction of other behaviors” (p. 529). Similarly, Cooper et al. ( 2019 ) refer to DRA as “a procedure for decreasing problem behavior in which reinforcement is delivered for a behavior that serves as a desirable alternative to the behavior targeted for reduction and withheld following instances of problem behavior” (p. 790).
Following nearly 40 years of research on behavioral treatments based on a functional analysis, it appears that the common definitions of DRA that include extinction are overly restrictive. More recently, in their review of studies using DRA without extinction, Trump et al. ( 2019 ) identified that effective treatment packages using DRA without extinction included a more favorable schedule of reinforcement associated with appropriate behavior when compared with a less advantageous schedule for inappropriate behavior. Although an extinction component is ideal, extinction is not always possible or feasible (Briggs et al., 2019 ; Kunnavatana et al., 2018 ; MacNaul & Neely, 2017 ; Manente et al., 2010 ). As it relates broadly to circumstances involving differential reinforcement, we propose the following definition of DRA: Providing greater reinforcement, along at least one dimension, contingent on the occurrence of one form or type of behavior, while minimizing reinforcement for another form or type of behavior. As it relates more specifically to problem behavior, the greater reinforcement would occur contingent on some specific alternative response (e.g., a mand for attention), or some general type of response (e.g., appropriate vocal-verbal behavior in general). However, the proposed definition is worded broadly enough to include differential reinforcement during skill development (such as when reinforcing independent responses more than prompted responses), shaping (such as when reinforcing a closer approximation more than a previously established approximation), and in fact during unfavorable baseline circumstances (such as when problem behavior is reinforced more than appropriate alternative behavior). It should be noted our emphasis may apply to differential reinforcement in general, not only DRA. For example, particular arrangements of differential reinforcement of other behavior (DRO) and differential reinforcement of low-rate behavior (DRL) can be conceptualized similarly (e.g., Jessel & Borrero, 2014 ).
There are numerous reasons that, at times, target behavior either cannot or should not be placed on extinction. For example, if a student engages in attention-maintained aggression, a teacher still must intervene (provide some level of attention) to ensure the safety of staff or other students (e.g., Newcomb et al., 2019 ). Similarly, if a client engages in dangerous self-injury and has potential to cause great harm to herself (e.g., detach a retina or produce a concussion), one must block the behavior for ethical and safety reasons, even if the physical contact is reinforcing the behavior. Even escape behavior is sometimes nearly impossible to place on extinction. For example, a student who is fast and wiry might slip beneath a desk or table despite the therapists' efforts to block. A client might be stronger than the therapist rendering escape extinction in the form of guided compliance difficult if not impossible (e.g., Piazza et al., 1996 ), and possibly undesirable or illegal to implement.
Another consideration is that even highly trained behavior analysts make mistakes during implementation. A search of the behavioral treatment literature shows that treatment integrity reports of values less than 100% are somewhat common. To the extent that those integrity errors are errors of commission (Vollmer et al., 2008 ), problem behavior is presumably sometimes reinforced. If even highly trained behavior analysts make errors, it is to be expected that parents, grandparents, teachers, classroom aides, and others will make errors as well, perhaps more frequently. If problem behavior is sometimes reinforced, the behavior is likely contacting reinforcement according to a schedule that may blend ratio and interval features (Nevin, 1998 ).
Basic and applied research on choice has shown that variables such as rate of reinforcement (Borrero et al., 2010 ; Herrnstein, 1961 ), delay to reinforcement (Chung & Herrnstein, 1967 ; Davison, 1988 ; Dixon & Cummings, 2001 ), duration or magnitude of reinforcement (Lerman et al., 1999 ; Volkert et al., 2005 ), and quality of reinforcement (Gardner et al., 2009 ; Hoch et al., 2002 ) control allocation of responding in concurrent choice arrangements (Fisher & Mazur, 1997 ). Contemporary theories of treatment relapse are firmly rooted in choice models that do not assume (in fact, explicitly do not assume) that extinction is arranged consistently for problem behavior (e.g., Shahan & Craig, 2017 ). Thus, in situations involving behavioral treatments, DRA can be considered a concurrent choice arrangement (Greer & Shahan, 2019 ). The topographies defined as “problem behavior” produce one outcome and the topographies defined as “alternative” (or appropriate behavior) produce some other differential outcome, and histories of such outcome yield distinct predictions about treatment outcomes (Kimball et al., 2018 ).
Usually in baseline, before a behavioral treatment is implemented, one or more (if not all) dimensions of reinforcement favor problem behavior. Functional analysis research has shown that differential reinforcement of problem behavior is the reason most problem behavior occurs (Beavers et al., 2013 ; Hanley et al., 2003 ). Usually in treatment, the task of the behavior analyst is to shift allocation of responding and this is done by making reinforcement more likely, more immediate, longer-lasting, of higher quality, or some combination, in relation to the reinforcement produced by problem behavior (e.g., Athens & Vollmer, 2010 ). Thus, even if reinforcement still occurs some of the time contingent on problem behavior, it is possible to implement differential reinforcement. Recent research has confirmed such an effect (e.g., Dowdy et al., 2018 ; Kunnavatana et al., 2018 ; Slocum & Vollmer, 2015 ).
Developing a new definition of DRA has practical importance. It may be the case that practitioners or care providers discard DRA as a treatment option when they determine that problem behavior cannot or will not be placed on extinction. If the definition instead stresses minimizing reinforcement for problem behavior, the practitioner or care provider can do what is necessary to protect themselves, the client, or others in the environment, while maximizing reinforcement when appropriate alternative behavior occurs.
Additionally, developing a new definition of DRA has conceptual importance. Functional analysis methodology has allowed behavior analytic researchers and practitioners to isolate specific sources of reinforcement for problem behavior. Prior to the development of functional analysis methods, the reinforcer(s) maintaining problem behavior were often unknown. By identifying reinforcer(s), it is easier to translate research on choice, such as the matching law, to conceptualize and favorably shift response allocation across two or more alternatives when problem behavior continues to contact reinforcement.
In summary, we propose that the definition of DRA should not be restricted in a way that includes extinction as one component. Certainly, research has shown that extinction as a part of differential reinforcement is a preferred strategy. However, recognizing that extinction either cannot or will not always be implemented during DRA, and considering that DRA can still be effective in those circumstances, the proposed definition emphasizes differentially reinforcing alternative behavior in relation to problem behavior. In research, behavior analysts could specify precisely which consequences were applied to both problem and appropriate behavior. In practice, behavior analysts should ensure that reinforcement for appropriate behavior “outweighs” reinforcement for problem behavior, even when extinction is not in place. Extinction should still be a component when possible, but when it is not very specific recommendations for the quality, duration, immediacy, and magnitude of reinforcement for alternative behavior should be routinely reported. The over-arching suggestion is to maximize reinforcement for alternative behavior and minimize or eliminate reinforcement for problem behavior.
