JOURNAL OF APPLIED BEHAVIOR ANALYSIS 1987, 20, 361-378 NUMBER 4 (WINTER 1987)

NEGATIVE REINFORCEMENT IN APPLIED BEHAVIOR ANALYSIS: AN EMERGING TECHNOLOGY

BRIAN A. IWATA

UNIVERSITY OF FLORIDA

Although the effects of negative reinforcement on human behavior have been studied for a number of years, a comprehensive body of applied research does not exist at this time. This article describes three aspects of negative reinforcement as it relates to applied behavior analysis: behavior acquired or maintained through negative reinforcement, the treatment of negatively reinforced behavior, and negative reinforcement as therapy. A consideration of research currently being done in these areas suggests the emergence of an applied technology on negative reinforcement.

DESCRIPTORS: aversive stimulation, avoidance, escape, negative reinforcement

Research published in the Journal of Applied Behavior Analysis (JABA) has, for 20 years now, demonstrated how knowledge about environment­ behavior interactions, particularly those involving response-contingent events and correlated stimuli, may be used for the benefit of individuals and the larger society. In doing so, applied research has also made significant contributions to the general science of behavior by providing extension and external validation of experimental findings from the basic research laboratoty (Baer, 1978).

Along with the development of the applied field and its expansion into a number of areas in which the outcome of an experiment often has immediate social implications (e.g., business and industty, de­ velopmental disabilities, education, medicine, men­ tal health, public affairs), there has been growing concern of a widening gap between basic and ap­ plied behavior analysis. Critics (Deitz, 1978; Pierce & Epling, 1980) have indicated that the emphasis ofcontemporary applied behavior analysis has shift­ ed away from the study of conditions that produce change to the production of change per se, and that "relevance to basic principle," a supposed char-

Appreciation is expressed to Ed Malagodi, Jack Michael, Gary Pace, Terry Page, and Terri Rodgers, whose work or helpful suggestions assisted in the preparation of this article. Support was provided by Grant HD 16052 from the National Institute of Child Health and Human Development.

Reprints may be obtained from Brian A. Iwata, Depart­ ment of Psychology, University of Florida, Gainesville, Flor­ ida 32611.

acteristic of applied behavior analysis (cf. Baer, Wolf, & Risley, 1968), is reflected less and less in the research that journals such as JABA publish. The general accuracy of these criticisms, as well as their basis and implications, will continue to be the subject of periodic debate (Baer, 1981; Cullen, 1981; Michael, 1980). Nevertheless, it is possible at this point in the development of our field to identify specific and well-established areas of basic research for which little parallel exists in the applied literature, and vice versa.

The thesis of this article is that research on neg­ ative reinforcement provides one of the clearest and most immediately relevant examples of a case in which consideration, replication, and extension of basic research would benefit the applied area. Along with positive reinforcement and punishment, neg­ ative reinforcement has long been considered one of the elementary principles of behavior. A volu­ minous amount of research on negative reinforce­ ment exists in the basic literature (see reviews by Bolles, 1970; Herrnstein, 1969; Hineline, 1977, 1981, 1984; Hoffman, 1966; Schoenfeld, 1969; Sidman, 1966), and its inclusion as a distinct topic in texts on experimental analysis (e.g., Honig & Staddon, 1977) justifies its status as a major or­ ganizing principle. For example, acquisition, main­ tenance, extinction, and stimulus control all have been studied using negative reinforcement as the operant mechanism of interest.

Sandler and Davidson (1973) reviewed some of this basic research and discussed its relevance to the

361

362 BRIAN A. IWATA

development and treatment of pathological human behavior. They concluded that " … the escape and avoidance paradigms are still plagued by a number of unresolved issues …" (p. 254) that they hoped would be clarified by additional basic research and extension to the world of humans. Since that time, a number of investigations on negative reinforce­ ment with humans have been conducted, yet a systematic and comprehensive body of applied re­ search still does not exist. Consider the two most recent texts on aversive control with humans (Ax­ elrod & Apsche, 1983; Matson & Dilorenzo, 1984). Both provide thorough discussion of topics such as positive reinforcement, extinction, time-out, response cost, and contingent aversive stimulation. Thus, one might expect that these texts would be the most likely sources of information on negative reinforcement as well, but this is not the case. One text (Axelrod & Apsche, 1983) devotes less than a half dozen of over 300 pages to the topic of avoidance, and the discussion always is limited to avoidance as a side effect of punishment. No men­ tion is made of escape or avoidance as directly produced performances. The second (Matson & DiLorenzo, 1984) describes the hypothetical fea­ tures of escape and avoidance training on two pages but does not cite any applied references.

The relative absence of integrated material on negative reinforcement with humans raises several questions concerning generality and utility. Is hu­ man behavior relatively insensitive to contingencies of negative reinforcement? Are naturalistic human situations typically characterized by the absence of stimuli that can function as negative reinforcers, or opportunities to escape from or avoid these stimuli? What types of performances are likely to be ac­ quired through negative reinforcement? Finally, do procedures based on the application of negative reinforcement, unlike those based on positive re­ inforcement and punishment, have little therapeutic or pragmatic value?

For the past few years my students, colleagues, and I have been conducting a series of investigations in two areas–self-injurious behavior and pediatric feeding disorders. Curiously, these vety different problems have brought us into direct contact with

situations involving the use of negative reinforce­ ment and have forced us to consider more generally the relevance of negative reinforcement in applied behavior analysis. Our experience and our exami­ nation of the basic and applied research suggest that the answer to each of the above questions is "No." In fact, it appears that negative reinforce­ ment plays a central role in the development of many behaviors, appropriate as well as inappro­ priate, and that its application in a number of studies has not been formally acknowledged. In what follows, I will describe three aspects of neg­ ative reinforcement as it relates to the applied sit­ uation: first, undesirable behavior acquired or main­ tained through negative reinforcement; second, the treatment of negatively reinforced behavior; and third, negative reinforcement as therapy. This or­ ganization departs somewhat from that used in reviews of the basic research literature and has been adopted here to highlight the relevance ofparticular issues to the applied researcher. Much of the re­ search included here has been done with the de­ velopmentally disabled population because there is a high prevalence of significant behavioral disorders in this group and because it provides a narrow but adequate focus for discussion.

CURRENT CONCEPTUALIZATION OF NEGATNE REINFORCEMENT

Before proceeding, it may be helpful to clarify terminology and to delineate the defining features of negative reinforcement. The purpose of this digression is to show that the task of determining whether a given contingency is an example of neg­ ative reinforcement may not always be a simple one. Although there has been little confusion re­ garding the effect of negative reinforcement, de­ scribing its operations has posed a challenge to many beyond the level of the beginning student.

The process of negative reinforcement typically involves the removal, reduction, postponement, or prevention of stimulation; these operations strengthen the response on which they are contin­ gent (Hineline, 1977). Removal and reduction of ongoing stimulation typically produce behavior that

363 NEGATIVE REINFORCEMENT

is called "escape," whereas postponement and pre­ vention of stimulus presentation produce behavior that is called "avoidance." "Typically" is used as a qualifier throughout because the terms negative reinforcement, escape, and avoidance are subject to confusion under certain conditions, as the following will illustrate.

In commenting on the distinction between pos­ itive and negative reinforcement, Michael (1975) reviewed a number of historical points related to terminological usage. More important, he noted that some stimulus changes associated with an in­ crease in behavior are difficult to classify as "pre­ sentation" (positive reinforcement) versus "remov­ al" (negative reinforcement), and that the use of either description may be nothing more than an arbitrary and incomplete abbreviation for the static "prechange" and "postchange" stimulus condi­ tions as well as for what transpires in between. For example, is a change in temperature more accurately characterized as the presentation of cold (heat) or the removal of heat (cold)? Problems such as this led Michael to suggest that "The distinction be­ tween two types of reinforcement [positive vs. neg­ ative}, based in tum upon the distinction between presentation and removal simply can be dropped'' (p. 44). An additional basis for distinguishing be­ tween positive and negative reinforcement was sug­ gested first by Catania ( 197 3) and later by Hineline (1984), who noted that "… if a stimulus or sit­ uation is to be reducible or removable by some response, that response must occur in its presence. In contrast, positively reinforced responses neces­ sarily occur in the absence of the stimuli upon which reinforcement is based" (pp. 49~97). Such a distinction is not without its own problems, as can be seen in the previous example. Is responding prior to a temperature change more accurately described as responding in the presence versus the absence of heat (cold)? Another problem with this distinction is encountered when one considers the difference between escape (responding in the presence of stim­ ulation), and avoidance (responding in the absence of stimulation), both of which are examples of negative reinforcement.

In many applied situations, it is possible to iden-

tify unambiguously a stimulus change as one in­ volving presentation (e.g., of physical contact) or removal (e.g., of a token) and to determine whether or not the response of interest occurs in the presence or the absence of stimulation. However, because research outside of the laboratory is subject to great­ er variation of and less control over a multitude of potentially relevant stimuli, the motivational fea­ tures of some stimulus changes are difficult to spec­ ify. Consider, as a case in point, Osborne's (1969) "Free-time as a reinforcer in the management of classroom behavior," which examined the out-of­ seat behavior of six students. During the baseline condition, the students worked for approximately 4 hours daily without recess, and data showed that the target behavior occurred frequently. During treatment, students could earn 5 min of free time at the end of every 15-min work period by re­ maining in their seats, and the data showed a de­ crease in out-of-seat. It is interesting to note the target behavior. Defined and recorded as "out-of­ seat,'' free time was made available for its absence; this type of contingency usually is described as differential reinforcement of other behavior (ORO). However, the instructions given to students spec­ ified that they were to remain in their seats, sug­ gesting "in-seat" as the functional target. If so, free time was made available for the occurrence of in­ seat behavior; this type of contingency is not con­ sidered an example of ORO. Depending on how one characterizes "free time," (i.e., the availability of preferred activities vs. the termination of non­ preferred activities), we would label the contingency as one involving positive or negative reinforcement for in-seat behavior. Osborne suggested both of these possibilities in his discussion and perhaps this is why he did not place an adjective in front of the term "reinforcer" in the title of the article.

As a field, we have not attended carefully to the important distinction that Osborne drew. His study is regarded as a seminal piece of research in the applied literature for expanding our notion of what constitutes a reinforcing event and for demonstrat­ ing very nicely the effects of group contingencies, although the exact nature of the contingency is still unclear. A number of interesting replications and

364 BRIAN A. IWATA

extensions have appeared in]ABA (e.g., Aaron &

Bostow, 1978; Baer, Rowbury, & Baer, 1973; Barrish, Saunders, & Wolf, 1969; Harris & Sher­ man, 1973, 1974; Long & Williams, 1973; Ma­ loney & Hopkins, 1973; Medland & Stachnik, 1973), but none have included further discussion or analysis of free time contingencies as positive versus negative reinforcement. Although such anal­ yses may have little or no impact on outcome (i.e., in either case, behavior will have been increased), our general tendency to overlook a negative rein­ forcement interpretation may lead to undue em­ phasis on the numerous forms that free time may take at the expense of considering important fea­ tures of the environment that free time replaces. That is, if free time serves as negative reinforcement, its only essential component may be alteration or termination of the preceding aversive situation.

In a more general sense, the complete analysis and specification of conditions in effect prior and subsequent to responding was the primary basis underlying Michael's (1975) suggestion to elimi­ nate the distinction between positive and negative reinforcement. It appears unlikely that the terms "positive" and "negative" will be deleted from our technical vocabulary in the near future; neverthe­ less, researchers should be cognizant of the fact that the two are potentially interchangeable and that failure to consider both possibilities may have a limiting effect on experimental procedure, inter­ pretation, and subsequent application.

UNDESIRABLE BEHAVIOR ACQUIRED AND MAINTAINED BY NEGATIVE

REINFORCEMENT

Hineline (1977) noted that a typical negative reinforcement paradigm includes three features: the presence of aversive stimulation, the availability of a response, and a suitable contingency between the response and the stimulation. Any behavior thus developed or maintained, including a variety of disruptive, destructive, aggressive, self-injurious, and otherwise problematic acts, could be considered "normal" or "adaptive" in that it is the orderly outcome of specific conditioning operations. The

term "undesirable" is used here only as a means of classifying behaviors that are considered inap­ propriate given the usual social context.

An initial question of particular interest to those working in applied areas relates to factors that de­ termine the form of the response. Acquisition of negatively reinforced behavior has been a subject of interest to basic researchers as well because it has been found that some topographies are more readily produced than others. For example, the treadle­ press and shuttle responses of pigeons are more easily controlled by negative reinforcement than is the key peck, which is highly responsive to positive reinforcement (Ferrari, Todorov, & Graeff, 1973; Foree & Lolordo, 1970; MacPhail, 1968; Rachlin & Hineline, 1967; Smith & Keller, 1970). Similar data based on the study of different species have provided some support for the hypothesis that neg­ ative reinforcement involves selective control over preexisting "species-specific defense reactions" to aversive stimulation (Bolles, 1970, 1971). This account, however, does not provide adequate ex­ planation for the wide range of human behaviors that apparently is succeptable to negative reinforce­ ment. A more likely explanation is that aversive stimulation initially produces one or more of a variety of responses characteristic of both human and nonhuman subjects, including flinching, freez­ ing, jumping, visual scanning, and related and dif­ fuse motor activity (see reviews by Davis, 1979; Hutchinson, 1977; Myer, 1971), and that the eventual and more elaborate form of the behavior is determined by the individual's previous history and the prevailing contingency.

Thus, many of the serious behavioral disorders that are seen in, for example, mentally retarded individuals may be a function of negative rein­ forcement applied to a particular behavioral rep­ ertoire and shaped over time. It is possible that certain instructional sequences (e.g., requests or even the appearance of specific training materials or the instructor) become discriminative for aversive stim­ ulation in the form of physical contact, which is a common element in many teaching routines. At first, the stimulation and its associated cues may produce behaviors similar to those noted above. If,

365 NEGATIVE REINFORCEMENT

however, other behaviors have been successful in eliminating similar types of stimulation in the past, their eventual occurrence should not be surprising. Tantrums, attempts to flee, or destruction of ma­ terials are examples of such behavior, particularly if the individual is unskilled at more subtle or socially acceptable forms of escape. Although dis­ ruptive, these behaviors usually are not considered insurmountable barriers to instruction. A number of informal and formal interventions (e.g., pro­ ceeding in spite of the tantrum, "scooting" the individual's chair under a table and backing both against a wall, bolting the materials to the table, etc.) are successful in managing disruptive behavior in some cases. In other cases, the interventions may provide a means for shaping more serious forms of escape. The immediate result of aggression for the individual toward whom it is directed suggests that physically harmful acts could serve as very effective escape behaviors, and their ability to terminate aver­ sive instruction is most likely a function of the relative size and strength of client and trainer. Fi­ nally, self-injurious behavior, if severe enough, will quickly terminate any situation.

Data relevant to a negative reinforcement hy­ pothesis for the development of behavior disorders in the mentally retarded exist in retrospective form only because it would be unethical to produce pathological behavior in humans when it does not already exist. Nevertheless, support for such a hy­ pothesis can be found in several studies. Carr and Durand (1985) and Weeks and Gaylord-Ross

(1981) showed that several different topographies of inappropriate behavior occurred more frequently during a "difficult task" condition when compared to an "easy task" condition, suggesting that the former condition contained aversive properties and that the resulting behavior was escape- or avoid­ ance-motivated. Carr, Newsom, and Binkoff ( 1976) examined variables that apparently exerted stim­ ulus control over the self-injurious behavior of a psychotic boy. In one of their experiments, they presented the boy with three alternating situations: a free-play period, a condition in which the exper­ imenter spoke descriptive sentences to the child (e.g., "The sky is blue")-this was called the "tact"

condition, and a third condition in which the ex­ perimenter spoke instructions to the child-this was called the ''mand'' condition. Higher levels of self­ injurious behavior were associated with the mand condition. Carr, Newsom, and Binkoff (1980) con­ ducted a similar analysis of aggressive behavior in two boys, showing that aggression was more likely to occur when demands were present than when they were absent. Finally, Iwata, Dorsey, Slifer, Bauman, and Richman ( 1982) described a general methodology that allowed one to differentiate self­ injury associated with positive versus negative re­ inforcement. In one of the conditions, self-injury was followed by adult attention; in another, self­ injury produced brief escape from adult demands. Some subjects consistently exhibited self-injury dur­ ing the latter condition, suggesting that their be­ havior was more sensitive to and maintained by negative reinforcement.

It is important for us to identify how environ­ ments that we create may provide negative rein­ forcement for undesirable behaviors. When faced with situations in which our students and clients are disruptive, we should immediately examine the antecedent as well as the consequent conditions to determine if the difference between the two provides reduction of aversive stimulation, keeping in mind that negative reinforcers may be just as idiosyncratic as positive ones. If we conclude that our clients and students exhibit bizarre and potentially dangerous behaviors to terminate instruction, we might ques­ tion whether or not our well-intentioned efforts to teach are in our clients' best interest; at the very least, we must question one or more aspects of our teaching technique. Perhaps most important from the standpoint of contingencies, our ability to iden­ tify negative reinforcement as a maintenance vari­ able for undesirable behavior may directly influence treatment selection and outcome. This is particu­ larly true with respect to extinction and time-out. Their use typically calls for one or more therapist responses (e.g., turning away from the client, re­ moving stimuli from immediate access, removing the client from the setting, etc.) that terminate the ongoing situation. Studies conducted with non­ human (Appel, 1963; Azrin, 1961; Thompson,

366 BRIAN A. IWATA

1964) and human (Plummer, Baer, & LeBlanc, 1977; Solnick, Rincover, & Peterson, 1977) sub­ jeas, however, indicate that the effeas of rime-out are highly dependent on features of the "rime-in" environment. Thus, although time-out might be an effective means of extinguishing most positively reinforced behavior, it might directly strengthen negatively reinforced behavior.

TREATMENT OF NEGATNELY REINFORCED BEHAVIOR

A number of procedures based on the application of extinction, differential reinforcement, and pun­ ishment have been evaluated as treatments for problematic behavior of unspecified origin. Their use with behavior maintained by negative rein­ forcement will be discussed in this section, along with an additional procedure involving stimulus fading.

Extinction

Traditional rime-out will not provide for the extinction of behavior that has been maintained by negative reinforcement, but other procedures might. One rather obvious possibility is elimination of the supposed aversive stimulation and its related cues, which should produce a reliable decrease in escape or avoidance behavior (Boren & Sidman, 1957; Shnidman, 1968). However, as Hineline (1977) has noted, this procedure may not be a true ex­ tinction operation. The complete removal of aver­ sive stimulation during extinction of negatively reinforced behavior can be considered analogous to the continuous presence of, for example, food dur­ ing extinction ofpositively reinforced behavior. Both procedures amount to noncontingent reinforce­ ment, which removes the basis for responding and indirectly reduces the frequency of behavior. That is, if food is always present during extinction of food-maintained behavior, there is no basis for re­ sponding; a similar situation exists ifshock is always absent during extinction of shock-avoidance be­ havior. Following these procedures, food removal or, alternatively, reappearance of the shock should immediately produce the target response (see Mi-

chael, 1982, for an extended discussion of this topic).

A more appropriate extinction procedure would entail continued presentation of the aversive stim­ ulus or its cue and elimination of the consequence that was provided formerly (i.e., avoidance or es­ cape). In this manner, the basis for responding (aversive stimulation) remains, but reinforcement does not (Bankart & Elliott, 1974; Coulson, Coul­ son, & Gardner, 1970; Davenport, Coger, & Spec­ tor, 1970; Schiff, Smith, & Prochaska, 1972). Techniques derived from this type of extinction actually have been used for a number of years in the treatment of clinical phobias and provide the major theoretical basis for interventions collectively known as "implosion therapies" (Levis, 1979).

An example of extinction for negatively rein­ forced behavior was reported recently by Heidorn and Jensen (1984). After noting that demand­ related situations were associated with an increase in their subject's self-injurious behavior, a treatment was developed that included the following: (a) con­ tinued presentation of demands, (b) physical guid­ ance to complete the requested performance con­ tingent on the occurrence of self-injury, (c) termination of the session contingent on compli­ ance, and (d) gradual increase in performance cri­ teria across sessions. Positive reinforcement in the form of praise, food, and physical contaa also was provided, but its role as an active component of treatment may have been minimal. A similar pro­ cedure was used in one of the experiments reported by Carr et al. ( 1980) on the treatment of aggression. Extinction consisted of belting the subject in a chair to prevent escape while a therapist wearing pro­ tective gear sat across a table from him. The in­ tervention differed from that used by Heidorn and Jensen in that no attempt was made to deliver instructions during extinction sessions; instead, de­ mands were introduced after aggressive behavior was eliminated almost completely.

As with extinction of positively reinforced be­ havior, it is possible to foresee situations in which extinction of negatively reinforced behavior might not be in the immediate best interest of either the client (as in the case of severe self-injury) or the

367 NEGATIVE REINFORCEMENT

therapist (as in the case of aggression). Extinction procedures may be compromised further by the potential effects of what procedurally may resemble noncontingent aversive stimulation (see earlier dis­ cussion on acquisition of avoidance responding). To the extent that these "elicited" responses occur during the extinction of negatively reinforced be­ havior in applied situations, attempts to increase alternative behaviors, as well as to reduce the target behavior, may be disrupted. Finally, research show­ ing that time-based delivery of aversive stimulation can maintain (Powell & Peck, 1969) and even increase (Kelleher, Riddle, & Cook, 1963; Sidman, Hermstein, & Conrad, 195 7) the rates of avoidance behavior suggests that schedule-related variables and the subject's previous history may be important considerations in the use of extinction.

Differential Reinforcement

Applications of reinforcement to decrease a target behavior (differential reinforcement of other be­ havior [DRO}, differential reinforcement of incom­ patible behavior [ORI}, etc.) are well documented in the applied literature, although the maintaining variable for the target behavior rarely is noted. The reinforcement contingency itself typically involves the use of positive reinforcement, and discussion here will be similarly confined. Applications of neg­ ative reinforcement will be addressed separately.

An experiment designed to examine the sup­ pressive effects of differential reinforcement on neg­ atively reinforced behavior may take several forms. First, access to an appetitive reinforcer (e.g., food) could be made contingent on the absence of the target behavior (DRO) while the escape/avoidance contingency is still operative. Although this ap­ proach might be considered unusual, it may resem­ ble very closely situations in the natural environ­ ment in which ORO is implemented without attempting co identify the behavior-maintaining contingency. To my knowledge, this study has not been reported in the basic literature, probably due to difficulties associated with equating reinforce­ ment. It is possible that this rype of study has been reported in the applied literature but that it was not explicitly identified.

A second approach might involve appetitive re­ inforcement for a competing behavior (DRI) with the escape/avoidance contingency again operative. Ruddle, Bradshaw, Szabadi, and Foster ( 1982) studied human operant performance (button press­ ing) using exactly this procedure. They presented subjects with concurrent avoidance/positive rein­ forcement schedules, and obtained matched re­ sponding when the schedules were equated (this was made possible by using points exchangeable for money). Performance shifts were correlated with schedule shifts roughly in a manner predicted by Hermstein's (1961) matching law. Our assessment research on self-injury (Iwata et al., 1982) provides an approximation to the Ruddle et al. methodol­ ogy. During one condition, we p