Running Head: NEGATIVE REINFORCEMENT

Negative Reinforcement: A Review of the Literature Cicely Irene Nickerson Utah State University

NEGATIVE REINFORCEMENT Abstract Negative reinforcement is an operant mechanism in which aversive stimuli are removed contingent on a response. Aspects of negative reinforcement, including its comparison to positive reinforcement and positive punishment, are explored as this article reviews current and past research in both the applied and basic settings. The findings of each study are discussed, and recommendations for future research are explored. Keywords: avoidance, escape, negative reinforcement, positive punishment, positive reinforcement,

NEGATIVE REINFORCEMENT Negative Reinforcement: A Review of the Literature Much of human behavior is shaped by the consequences of those behavior. One classification of those consequences is negative reinforcement. Contacting negative reinforcement contingencies is a part of a typical day. Completing course evaluations is negatively reinforced by escape from the continued emails from the university. Opening an umbrella is negatively reinforced by avoidance of being wet. Pushing a button is negatively reinforced by escape from a clocks alarm. Wearing sunblock is negatively reinforced by avoidance of sunburns. As seen in these examples, negative reinforcement is characterized by the removal of a stimulus from the environment contingent on a response. There are two basic types of negative reinforcement, escape and avoidance. Postponing or preventing stimuli from presentation is termed avoidance. Removal of stimuli is categorized as escape. Negative reinforcement is an important tool for behavior analysts. It has been used to study acquisition, maintenance, extinction, and stimulus control (Iwata, 1987). Negative reinforcement can also serve as the function for aberrant behavior. In a meta-analysis, Iwata et al. (1994) found that social-negative reinforcement was the most common function for self-

injurious behavior. Studies that employ or focus on negative reinforcement have been conducted for many years. An early study using negative reinforcement demonstrated its effectiveness in the acquisition of behavior. This early basic research study was conducted by Solomon & Wynne (1953). The purpose was to study the acquisition and extinction of jumping as an avoidance behavior in a modified shuttle box. The shuttle box was made of two compartments which were separated by a barrier. The floor of each compartment had a steel grid which could be electrified independent of the other side. Dogs that had no known history as research subjects were used in

NEGATIVE REINFORCEMENT this study. Following baseline and an acclimatization period, each dog was exposed to the

acquisition procedure. During each session, a light in the shuttle box was lit for ten seconds. At the end of the ten seconds, the grid on the floor of the shuttle box was electrified, thus shocking the subject. Shocking continued until the dog had jumped or climbed over the barrier to the other side of the shuttle box where the floor was not electrified. Trials were counted as escape if the dog was shocked prior to jumping over the barrier, and counted as avoidance if the dog was not shocked prior to jumping over the barrier. After each of the dogs had acquired the response, they were exposed to an extinction phase where no shocks were given. Solomon et al. state, We had expected these dogs to extinguish spontaneously. . . . Instead, the experimenters found themselves running the animals day after day with no signs of extinction (p. 291). This early study shows that behavior that is maintained through negative reinforcement, especially those behavior which are linked to traumatic experiences, are highly resistant to extinction. In addition, Solomon et al. found that even after the behavior had been successfully put on extinction, it was not unusual for the behavior to spontaneously recover. Of course, one very significant limitation to this study is the severity of the stimuli used. It is expected that the acquisition of behavior in a similar study which used stimuli of a milder nature would be less resilient to extinction. Perhaps partially as a result of the adverse side-effects which have accompanied studies involving negative reinforcement in the basic setting, not much research has been done in the applied setting. More than 30 years ago, Iwata (1987) called for this to change. He stated, research on negative 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 (p. 361). Since that time, there has been a slight increase in the research

NEGATIVE REINFORCEMENT dedicated to negative reinforcement, but much more is needed. This article explores some possible directions for future research as it discusses current and past studies in both the basic and applied settings. Discussion Escape and Avoidance Negative reinforcement is divided into two categories: escape and avoidance. In an escape procedure, an aversive stimulus is terminated contingent on the subjects behavior. Comparatively, the subjects behavior in an avoidance procedure delays or deletes the imminent occurrence of an aversive stimulus. To be clear, for avoidance the behavior occurs prior to the presentation of a stimulus, resulting in postponement or cancellation. For escape, the behavior occurs during the presentation of the stimulus, resulting in its removal.

It can be difficult to determine whether a behavior is reinforced by escape from stimuli or avoidance of stimuli. Thompson, Bruzek, and Cotnoir-Bichelman (2011) found that both escape-type negative reinforcement and avoidance-type negative reinforcement maintained infant care-giving behavior. The purpose of their study was to demonstrate that findings in simulations of care-giving in a laboratory setting could be applied to clinical settings and treatment programs. Eleven typically developing university students were observed in a small room through a one-way mirror. Each student was told they were part of a simulated care-giving experiment and given charge of a baby doll with the instruction to do what comes naturally. Participants also had access to a crib, blanket, and, in some cases, toys and a bottle. Also present in the room was a tape player with a recording of a crying baby. This recording served as the independent variable and was turned on and off by the experimenters in the adjoining room. A specific topography of care-giving behavior (i.e. feeding, playing, or

NEGATIVE REINFORCEMENT rocking) was targeted as the dependent variable in each phase of the reversal design. During each session of the study, the recording of a babys cry would play until and unless the target

behavior was exhibited. All of the participants engaged in the target behavior which stopped the crying. When new sessions were started, the participants continued to engage in that behavior to avoid the crying. The initial engagement in the target behavior is an example of escape maintained behavior. The continuation of that behavior is an example of avoidance maintained behavior. In addition to what was previously mentioned, nine of the eleven participants discontinued the target behavior when it was no longer paired with reinforcement. The remaining two participants did not meet the criterion for moving past the extinction phase because they continued to engage in the care-giving behavior that was targeted in the initial phase of the experiment even though it was no longer being reinforced. However, when an alternate care-giving behavior resulted in negative reinforcement, both participants engaged in the new target behavior and discontinued the previous target behavior. The authors conclude that, caregiver behavior is, at least in part, under the control of negative reinforcement (p. 303). They continue by adding that other contingencies are most likely in place, but those contingencies were not the focus of this particular study. This study also evidences that negative reinforcement influences the behavior of individuals in day-to-day situations, and as such, additional research should be done to add to our limited knowledge. Negative and Positive Reinforcement Moving to a broader picture of behavior change, there is also a difference between negative reinforcement and positive reinforcement. Negative reinforcement and positive reinforcement are similar in that, in either case, the response being reinforced has an increased probability of occurring in the future. However, for positive reinforcement it is the presentation

NEGATIVE REINFORCEMENT of a stimulus that acts as a reinforcer, while for negative reinforcement it is the removal of a stimulus that acts as a reinforcer. Not only are they procedurally different, but positive and negative reinforcement have different effects. Bouxsein, Roane, and Harper (2011) compared positive and negative reinforcement as treatments for the increase of compliance to instructional tasks. The participant was a fourteenyear-old boy with Down syndrome who displayed non-compliant behavior. The sessions were run in a room that contained a garbage can and pieces of crumpled paper. The participant was instructed to pick up pieces of paper and put them in the garbage can. Completion of this instruction was measured as the dependent variable. The study followed a reversal design in

which different phases were characterized by different types of reinforcement. Those types were (a) negative reinforcement in the form of a 60-second break, (b) positive reinforcement, in this case access to music for 60-seconds with the continuation of demands, and (c) a combination of positive and negative reinforcement, where both access to music and a break from demands were given for 60-seconds. In phases where compliance resulted in access to music for 60-seconds with the continuation of demands, compliance percentages increased slightly. Interestingly, compliance percentages decreased in the phases where compliance resulted in a 60-second break. This means that a 60-second break did not function as negative reinforcement for this participant. For this study, the combination of positive and negative reinforcement for compliance to instructions was the most effective at increasing compliance. One possible explanation for these results is that the value of the reinforcer was greater during phases where both access to music and a break from demands were given. Also, the authors do not inform the reader whether the non-compliant behavior for which the participant was referred was maintained by positive or negative reinforcement. Certainly, if the behaviors function was to

NEGATIVE REINFORCEMENT

access music, and not to escape demands, the data would show greater compliance during phases where access to music was given. Nevertheless, it is of value that even when escape was not negatively reinforcing for this participant, the addition of escape from demands to a positive reinforcer increased compliance. DeLeon, Neidert, Anders & Rodriguez-Catter (2001) conducted another study comparing the effects of positive and negative reinforcement schedules on compliance. This study targeted the value of positive versus negative reinforcement as schedules of reinforcement were thinned. Participant preference was also evaluated. The participant in this study was a ten-yearold girl with autism whose aberrant behavior was determined to be escape maintained. The independent variable used in this study was type of reinforcement (positive vs. negative) earned for compliance to demands. The dependent variables were rate of aberrant behavior and compliance. Following a functional analysis, a combined multi-element and reversal design was used for treatment. In the baseline condition academic demands were given in a sequential prompting order (verbal, gestural, physical), and aberrant behavior resulted in a 30-second break from demands. After baseline, sessions in which positive reinforcement (potato chip) was available and sessions where negative reinforcement (30-second break) was available were alternated and compared in terms of their influence on compliance. Both positive and negative reinforcement were delivered on an FR1 schedule during this phase, and aberrant behavior continued to result in a 30-second break from demands. After a short return to baseline, the two elements were again compared. Both remained at an FR1 schedule. In the final phase, aberrant behavior no longer resulted in a break from demands. Additionally, the participant was allowed to choose between the positive reinforcer and the negative reinforcer after she had complied with instructions. Thinning of the reinforcement schedule also began during this phase. The results

NEGATIVE REINFORCEMENT of the comparison at an FR1 schedule were similar to those of Bouxsein, Roane, and Harper

(2011). Sessions where compliance was positively reinforced were marked by near-zero rates of problem behavior and high percentages of compliance. Sessions where compliance was negatively reinforced showed minimal reduction in problem behavior and only some change in percentages of compliance. However, in the third phase, as the schedule of reinforcement was thinned, the participant began to choose the negative reinforcer more frequently. This implies that while positive reinforcement may be preferred during rich schedules of reinforcement, as schedules are thinned, the value of negative reinforcement in the form of escape from demands increases. This study was replicated by Kodak, Lerman, Volkert & Trosclair (2007) for five children with autism. Functional analyses for these participants indicated that problem behavior was maintained, at least in part, by escape from demands. For four of the five participants, positive reinforcement (edible) was preferred over negative reinforcement (break from demands) even when schedules were thinned. It should be noted, however, that when negative reinforcement was enriched to include positive reinforcement in the form of toys and attention the participants still preferred the edible reinforcer. This could mean that positive reinforcement was chosen instead of negative reinforcement, not because positive reinforcement as a whole is more effective, but because the particular edible used was viewed as very valuable by the participants. This conjecture is supported by data from three of the participants that show when the edible was switched for a less preferred edible, the participant preferred the break. Additional studies which compare positive reinforcement and negative reinforcement have found varying results. For the most part these studies show that positive reinforcement is more effective at increasing compliance and reducing problem behavior (e.g. Carter, 2010; Lalli,

NEGATIVE REINFORCEMENT et al., 1999). However, there are studies, and participants within studies that stand as evidence that negative reinforcement can be more effective in some cases (Kodak, Lerman, Volkert, & Trosclair, 2007; Zarcone, Fisher, & Piazza, 1996). Aversive Control Just as particular aspects of negative reinforcement match aspects of positive reinforcement, negative reinforcement shares common features of punishment. Consequently, negative reinforcement falls into a category termed aversive control. The term aversive

10

control is not always clear. In most definitions it includes negative reinforcement and positive punishment (Cooper, Heron, & Heward, 2007). As you will recall, negative reinforcement is when an aversive stimulus is removed contingent on a response. Positive punishment occurs when an aversive stimulus is presented contingent on a response. By some definitions aversive control also includes negative punishment (Catania, 1998). In negative punishment, a preferred stimulus is removed contingent on a response. For purposes of this article, aversive control will include positive punishment and negative reinforcement. An early basic research study conducted by Lockard (1969) compared effects of negative reinforcement and positive punishment on the acquisition of a new skill (pulling a lever) via shaping. Twenty-four rhesus monkeys participated in the first procedure where negative reinforcement was used. In this procedure, shocks were scheduled to be delivered every five seconds. Shocks were delayed contingent on the participant emitting the next behavior in the shaping chain. All of the monkeys finished this procedure. An additional procedure using positive punishment was run with six rhesus monkeys. In this procedure shock was delivered each time the participant moved away from the lever. Although no data were reported, the authors state that these subjects struggled to learn the response of pulling the lever, and some subjects stopped moving all together. For those

NEGATIVE REINFORCEMENT

11

subjects who stopped moving, the method was altered where all behavior except moving toward the lever were punished, but this did not improve training. The authors concluded that negative reinforcement is more effective when teaching a new behavior than positive punishment. This could occur because during negative reinforcement only one behavior evokes consequences, in comparison to positive punishment where all but one behavior result in the consequence. In this way, negative reinforcement teaches the correct response and positive punishment does not. Effective Negative Reinforcement The effectiveness of negative reinforcement, as with positive reinforcement and punishment, can be influenced by other factors including the quality of the reinforcement, the immediacy of the removal of the aversive stimulus after emission of the behavior, the consistency with which the behavior evokes reinforcement, and the unavailability of reinforcement through other behaviors. An example of one of these methods for effective use of negative reinforcement is shown in a study of seat belt usage systems. Geller, Casali, & Johnson (1980) observed drivers as they entered and exited university parking lots, and collected data on escape and avoidance maintained seatbelt-wearing behavior. As each driver approached the exit to the parking lot, he was stopped by a data collector and asked a few survey questions. The data collector noted whether the driver was wearing a seatbelt, or if the cars seat belt usage system had been bypassed in any way. Seat belt usage systems included any combination of lights, buzzers, or ignition interlocks that were designed to discontinue when the drivers seatbelt was in use. Data collectors also asked the driver questions about his seat belt usage and the cars seatbelt system. Additionally, drivers were asked to complete a paper survey and mail it to campus. Nearly 42% of drivers whose cars had a seat belt usage system had bypassed that

NEGATIVE REINFORCEMENT system. Of those drivers who had defeated the system, approximately 16% were wearing a seatbelt at the time of the survey. Drivers of cars where the seat belt usage systems were still intact were nearly 3 times as likely to be wearing a seatbelt. In this sample, participants whose cars had a seat belt usage system could receive

12

negative reinforcement for seat belt wearing behavior. However, only 33% of those participants were actually wearing seatbelts. It could be argued that a possible reason for low percentages of seat belt wearing behavior may be that drivers have not yet left the parking lot, and may typically buckle up upon leaving those areas. However, this low percentage is more likely due to the availability of reinforcement through alternate behavior, namely sabotage of the cars seatbelt usage system. Unlike the example of seatbelt systems, effective negative reinforcement procedures do not allow for reinforcement to be available through alternate behavior. Conclusion Despite the widespread prevalence of escape and avoidance maintained behavior, research and knowledge in applied settings are limited. An extremely small amount is known about negative reinforcement, especially when compared to its sister operant mechanism, positive reinforcement. Although it can be difficult to separate, we are aware that negativelyreinforced behavior falls under either escape-maintained behavior or avoidance-maintained behavior. However, partially because these categories are difficult to parse, we do not know much about their individual characteristics. Because much of early research focused on avoidance behavior, we do know more about it than escape, but it is not clear whether the things basic research has demonstrated can be used in applied and clinical settings. More research comparing the two types of negative reinforcement should be conducted in the applied field in order to solidify the techniques and knowledge that are a part of basic research.

NEGATIVE REINFORCEMENT Although the majority of research suggests that it negative reinforcement is not as

13

effective as positive reinforcement for behavior change, we do know that the addition of negative reinforcement to positive reinforcement contingencies can enhance the value of reinforcement. We also know that, in some cases, negative reinforcement alone can be more effective than positive reinforcement. The most recent research does not support theories that problem behavior maintained by negative reinforcement should be paired with treated by teaching a replacement behavior that is also negatively reinforced. In fact, current research is beginning to suggest that the continued reinforcement of negatively reinforced problem behavior paired with positive reinforcement for compliant behavior is as successful, if not more successful than any other combination of contingencies (Lalli et al., 1999). To date, research has no explanation for why negative reinforcement is more successful in some cases and not in others. This may be a question for future research. That we dont know much about aversive control can make decisions concerning its use exceptionally difficult. Again, much of what we do know is from basic research. As a general rule, aversive control, and consequently negative reinforcement, has been avoided in both applied research and clinical settings. Yet, there are some circumstances, although rare, when aversive control seems to be the most effective and ethical option. In order to ensure that decisions being made at this level are based on best practice instead of chance or feeling, research in the applied field must take the information that has been found in basic research and extend it. A considerable amount of what is known concerning negative reinforcement is based on a handful of studies. While the research and publication of articles involving negative reinforcement has greatly increased since Iwatas call for change thirty years ago, the pool of

NEGATIVE REINFORCEMENT knowledge from which practice draw from is meager. Building on current and past research is essential to the growth of applied behavior analysis.

14

NEGATIVE REINFORCEMENT References Bouxsein, K. J., Roane, H. S., and Harper, T. (2011). Evaluating the separate and combined effects of positive and negative reinforcement on task compliance. Journal of Applied Behavior Analysis, 44, 175-179. Carter, S. L. (2010). A comparison of various forms of reinforcement with and without extinction as treatment for escape-maintained problem behavior. Journal of Applied Behavior Analysis, 43, 545-546. Catania, A. C. (1998). Learning (4th ed.). Upper Saddle River, NJ: Prentice Hall.

15

Cooper, J. O., Heron, T. E., & Heward, W. L. (2007). Applied behavior analysis (2nd ed.). Upper Saddle River, NJ: Pearson. DeLeon, I. G., Neidert, P. L., Anders, B. M., and Rodriguez-Catter, V. (2001). Choices between positive and negative reinforcement during treatment for escape-maintained behavior. Journal of Applied Behavior Analysis, 34, 521-525. Geller, S. E., Casali, J. G., and Johnson, R. P. (1980). Seat belt usage: A potential target for applied behavior analysis. Journal of Applied Behavior Analysis, 13, 669-675. Iwata, B. A. (1987). Negative reinforcement in applied behavior analysis: An emerging technology. Journal of Applied Behavior Analysis, 20, 361-378. Iwata, B. A., Pace, G. M., Dorsey, M. F., Zarcone, J. R., Vollmer, T. R., Smith, R. G., . . . Willis, K. D. (1994). The functions of self-injurious behavior. Journal of Applied Behavior Analysis, 27, 215-240. Kodak, T., Lerman, D. C., Volkert, V. M., and Trosclair, N. (2007). Further examination of factors that influence preference for positive versus negative reinforcement. Journal of Applied Behavior Analysis, 40, 25-44.

NEGATIVE REINFORCEMENT

16

Lalli, J. S., Vollmer, T. R., Progar, P. R., Wright, C., Borrero, J., Daniel, D., . . . May, W. (1999). Competition between positive and negative reinforcement in the treatment of escape behavior. Journal of Applied Behavior Analysis, 32, 285-296. Lockard, J. S. (1969). Shaping avoidance behavior in restrained monkeys. Journal of the Experimental Analysis of Behavior, 12, 649-652. Solomon, R.L., Kamin, L.J., and Wynne, L.C. (1953). Traumatic avoidance learning: The outcomes of several extinction procedures with dogs. The Journal of Abnormal and Social Psychology, 48, 291-302. Thompson, R. H., Bruzek, J. L., and Cotnoir-Bichelman, N. M. (2011). The role of negative reinforcement in infant caregiving: An experimental simulation. Journal of Applied Behavior Analysis, 44, 295-304. Zarconne, J. R., Fisher, W. W., and Piazza, C. C. (1996). Analysis of free-time contingencies as positive versus negative reinforcement. Journal of Applied Behavior Analysis, 29, 247250.