An Evaluation of the Effects of a Mild Delayed Verbal Punisher on Choice of an Immediate Reinforcer by Children With Autism

Abstract

Different combinations of immediate and delayed consequences differentially affect choice. Basic research has found that nonhuman animals are more likely to choose an alternative that produces an immediate reinforcer that is followed by a delayed punisher as the delay to punishment increases. The purpose of the current effort was to examine the choices of three individuals with autism when they were given the choice between receiving a larger amount of preferred food followed by a mild, delayed verbal punisher and a smaller amount of the preferred food. A secondary purpose was to determine whether signal presence and duration would affect the efficacy of the punisher (i.e., whether children would be more likely to select the smaller reward that was not followed by a delayed punisher). Results were idiosyncratic across children and highlight the need to evaluate choice under multiple arrangements.

Keywords

choice delayed punishment rule-following

Individuals diagnosed with autism spectrum disorder (ASD) often choose among multiple options, and each option typically is associated with both immediate and delayed consequences. For example, on one hand, complying with an instruction might be associated with an immediate aversive consequence (e.g., response effort, potential need to set aside a preferred activity) and a long-term pleasant consequence (e.g., exposure to new environments and learning experiences). On the other hand, noncompliance could be associated with an immediate pleasant consequence (e.g., no response effort, continued engagement with the preferred item or activity) and a delayed aversive consequence (e.g., physical prompt, reprimand). As this example illustrates, choices between response options can be influenced by the immediacy, magnitude, and aversiveness of the consequences of each option.

Basic researchers often have set up choice scenarios in which one response produces reinforcement (or punishment) of greater magnitude following a delay and another response produces reinforcement (or punishment) of lesser magnitude immediately (see Critchfield & Kollins, 2001, for a review). Under these arrangements, selection of the smaller, immediate reinforcer over the larger, delayed reinforcer or selection of a larger, delayed punisher over a smaller, immediate punisher has been labeled “impulsive.” The decrease in efficacy of a consequence as the time until its receipt increases is termed temporal discounting (see Green & Myerson, 2004, for a review), and it is thought to describe how individuals might choose the smaller but more immediate reinforcer (and the larger, but delayed punisher). That is to say, the discounting of the delayed reinforcer makes it less attractive and the discounting of the delayed punisher makes it less aversive.

Temporal discounting research has examined the effects of both delayed reinforcers (e.g., Green, Myerson, Lichtman, Rosen, & Fry, 1996) and delayed punishers (e.g., Abramowitz & O’Leary, 1990; Banks & Vogel-Sprott, 1965; Baron, 1965; Trenholme & Baron, 1975; Woolverton, Freeman, Myerson, & Green, 2011). Woolverton et al. (2011) found that the delivery of a delayed punisher following selection of a reinforcer could shift response allocation of five male rhesus monkeys to those responses that produced reinforcers not followed by delayed punishers. However, Woolverton et al. found that the punishers had less of an effect on choice as the delay to the aversive event increased. It is important to examine the effects of such contingencies because many responses that result in immediate reinforcement also produce delayed aversive events. For example, as noted by Woolverton et al., substance use often is associated with a variety of delayed aversive events, such as incarceration, accidents, worsened health, deterioration of social relationships, and hangovers. Likewise, someone applying for a job might quickly complete a cover letter, thereby contacting the immediate pleasant consequence of escape from work. However, this pleasant consequence could be followed by delayed, aversive events if content, spelling, or grammatical errors within the cover letter preclude that individual from obtaining a job. Despite the prevalence of these types of contingencies, explicit studies of choice between smaller reinforcers and larger reinforcers followed by delayed punishers have not been conducted with humans.

Indeed, the effects of punishment on human behavior have not been extensively studied outside of the context of severe problem behavior, despite the fact that punishment contingencies exist throughout daily life (Vollmer, 2002). For example, failure to complete academic work might result in restricted access to preferred activities, such as going out for recess later on in the day, a potential negative punishment contingency. From an ethical standpoint, failure to complete academic work might not justify the use of punishment; however, these types of contingencies are so prevalent that it seems ethically indefensible not to study their effects.

Thus, the purpose of the present study was to extend work by Woolverton et al. (2011) and examine how choice behavior of children diagnosed with ASD is affected by the magnitude of preferred consequences and the delay to the delivery of a mildly aversive consequence (in the form of delayed feedback). The effects of delayed feedback were examined with individuals diagnosed with ASD because many of these individuals have limited verbal repertoires (Kjelgaard & Tager-Flusberg, 2001), which might make their behavior less sensitive to verbal statements (e.g., “You do not get dessert tonight because you did not finish your chores this morning”), that typically are used with neurotypical children to link the target response to the delayed consequence. In addition, we examined how signal presence and duration affected choice.

Method

Participants and Setting

Three children participated in the study. Parental consent was obtained for all three participants, and a university-affiliated institutional review board approved the research protocol. Chas was a 9-year-old male diagnosed with ASD. He previously had been administered the Clinical Evaluation of Language Fundamentals–Fourth Edition (CELF-4), a tool designed to test language skill deficits in school-aged children, adolescents, and young adults. Chas was unable to complete many of the tasks required for this assessment measure, and a standard score was unable to be obtained. Chas also was administered the Preschool Language Scale–Fourth Edition (PLS-4). Although a standard score could not be obtained because he was above the normed age level, it was determined that his receptive language skills were discrepant from the norm. Chas’s performance on the Receptive One Word Picture Vocabulary Test (ROWPVT) was below expectancy. In the past, Chas exhibited a variety of aberrant responses, including difficulty transitioning between activities, scratching, and self-injury. However, these responses were eliminated before the start of the study. Jonah was a 9-year-old male with an educational diagnosis of ASD. Naomi was a 7-year-old female with an educational diagnosis of ASD. Neither Jonah nor Naomi had been administered tests to assess their language skills. However, both were able to emit and respond to multiple-word sentences. Neither Jonah nor Naomi had histories of displaying aberrant behavior, and neither child displayed aberrant behavior during the study.

All three children received services at an agency that teaches language, social, adaptive, and pre-academic or academic skills to children with autism. This agency is specifically prohibited from using strategies that might be considered aversive. Therefore, it is unlikely that participants contacted punishment contingencies at the agency. However, it is unknown whether participants contacted punishment contingencies in their homes.

Sessions were conducted at a table in the private room at the agency where each child received behavioral services.

Response Measurement and Interobserver Agreement (IOA)

The dependent variable was the percentage of free-choice trials in which a response option was selected. Observers scored a selection when the child independently touched one of the two options (various amounts of preferred food on a plate) presented on a trial. The percentage of free-choice trials was calculated by dividing the number of trials in which the child selected a particular option (i.e., the reinforcer not followed by a “no” statement, the reinforcer followed by a “no” statement, the larger magnitude of food, the smaller magnitude of food, or the larger magnitude of food followed by a delayed “no” statement) by the total number of free-choice trials and multiplying by 100. All the children had the motor skills required to make a selection and consume the item.

A second observer simultaneously, but independently, collected data on the children’s choices during 76%, 27%, and 24% of all sessions for Chas, Jonah, and Naomi, respectively. IOA was calculated on a trial-by-trial basis by dividing the number of trials in which observers agreed about the option selected by the total number of trials and multiplying by 100. IOA averaged 100%, 99.2% (range = 90%-100%), and 100% for Chas, Jonah, and Naomi, respectively.

Experimental Design

We used a concurrent-chains arrangement embedded within a reversal design to evaluate the relative preference for (a) food followed by a “no” statement versus food not followed by a “no” statement, (b) larger versus smaller amounts of preferred foods, (c) larger amounts of preferred food followed by a delayed “no” statement versus smaller amounts of food, (d) larger amounts of preferred food followed by a partially signaled and delayed “no” statement versus smaller amounts of food, and (e) larger amounts of preferred food followed by a completely signaled delayed “no” statement versus smaller amounts of food. Delays ranged in duration from 5 s to 60 s. A combination of pre-established stability criteria and visual analysis were used to determine preference stability. During each session, the experimenter presented two forced-choice trials and 10 free-choice trials, and the position of each option was randomly alternated across trials. To the extent possible, experimenters attempted to remain neutral following every choice and during the duration of the delay to the “no” statement or to the next trial. One to two sessions were conducted per day, 1 to 3 days per week.

Procedure

Preference assessments

The purpose of these assessments was to identify highly preferred food items. Prior to each session, children were offered a choice among parent-approved food items. The item selected first was used during the remainder of that session. This assessment is similar to the modified stimulus without replacement (MSWO) preference assessment described by DeLeon and Iwata (1996) in which an individual is presented with an array of items and asked to choose one. In the MSWO, after a selection is made and access to the item provided, the remaining items are re-presented to the child who is asked to choose another item. This process continues until all the items have been selected or 30 s elapse without a selection. This was not done in the current study because we were only interested in identifying the most preferred item.

Punishment test

The purpose of this assessment was to determine whether a “no” statement functioned as a punisher. During each trial, the experimenter presented the child with two plates. The plates were differently colored, and each plate held one small piece of food identified as preferred prior to the session (e.g., if the child selected a chip prior to the start of the session, half a chip was presented on each plate). During the session, one plate was randomly designated as the “no” plate. If the child selected the food on this plate, he or she was allowed to consume the food while the experimenter immediately removed the other plate, made eye contact with the child, and said “no” in a medium-volume voice with a neutral expression. Although “medium” volume was not quantified by a decibel reader, every attempt was made to say “no” in a natural, forceful manner, as it might be said in the child’s natural environment. If the child selected the food on the other plate, the experimenter did not say anything, and the child was allowed to consume the food. Thus, the only difference between the choice options was whether the experimenter said “no” immediately following a choice. Holding access to food constant across response options allowed us to isolate “no” as the relevant independent variable.

Following two forced-choice trials in which the experimenter prompted the child to sample the consequences of choosing each plate, the experimenter presented 10 free-choice trials in which the child was allowed to choose between the two options. Each trial began immediately following consumption (i.e., absence of mouth movement), and only data from the 10 free-choice trials were recorded. “No” was assumed to function as a punisher if the child selected the food not followed by the “no” statement during at least 60% of free-choice trials across three consecutive sessions. Although the 60% criterion was relatively lenient, it was enough to indicate that selection of the food not followed by the “no” statement was greater than chance. We chose not to use a higher criterion because the aversive stimulus (“no”) had to be relatively minor given the non-clinical nature of its inclusion. In some cases, we extended the punishment test after criterion was met so as to verify the stability of responding. This phase was replicated later in the evaluation to ensure that “no” continued to function as a punisher.

Magnitude tests

The purpose of these assessments was to determine the child’s preference for larger over smaller amounts of the preferred food. At the start of each session, experimenters presented the child with two plates and said, “Pick one.” The plates were identical in color. Initially, one plate held one piece of a preferred food and the other plate held two pieces of the preferred food. Following either selection, the other plate was removed, and the child was allowed to consume the food. Any attempt to access both plates was blocked, and the trial was repeated. Each subsequent trial began immediately following consumption (i.e., absence of mouth movement). Following two forced-choice trials, the experimenter presented 10 free-choice trials during each session. Larger amounts of preferred food were assumed to be relatively more reinforcing if the child selected the larger amount during at least 60% of free-choice trials in the last three sessions. Once again, although the 60% criterion was relatively lenient, it was deemed sufficient to indicate that selection was greater than chance. We did not aim for a higher criterion because uncontrolled pre-session variables (e.g., participants occasionally arrived at the session having just finished a snack) made it unlikely that larger amounts of food would be consistently selected. In addition, the repeated trials may have made differences in magnitude less salient. Indeed, during this evaluation, a child could select the smaller amount of food, consume it, and be offered a choice between two new plates of food just a few seconds later. Functionally, this might not be that different from selecting the larger amount of food and eating each piece one at a time, with a brief pause between pieces. Thus, we decided to use the lenient criterion of 60% selection across three consecutive sessions. In some cases, we extended the magnitude test after criterion was met to verify the stability of responding. This phase was replicated later in the evaluation to ensure that a larger magnitude of reinforcement was still preferred.

In Jonah’s second magnitude test, preference between one piece and three pieces of food was evaluated. This was done to verify that three pieces of food were more preferred than one piece of food prior to evaluating choice between these food magnitudes in subsequent delayed punishment phases. Jonah had pretty consistently selected the smaller, one-piece magnitude of food over the larger, two-piece magnitude of food in the immediately preceding punishment phases, even as the delay between selection of the larger magnitude of food and the mild, verbal punisher increased to 60 s. Thus, we hypothesized that if the magnitude of the larger amount of food was increased, this might make Jonah more likely to select the larger amount of food, even when it was followed by “no.” However, prior to this manipulation, we wanted to verify that three pieces of food were more preferred to one piece of food.

For Chas, in his second magnitude test, preference between two pieces and three pieces of food was evaluated. This was done to verify that three pieces of food were more preferred than two pieces of food prior to evaluating choice between these magnitudes in the subsequent partially signaled delayed punishment phase. In the immediately preceding punishment phases, Chas had almost exclusively chosen the larger magnitude of food, despite the fact that the larger magnitude was followed by a delayed “no,” which had been shown to function as a punisher during each of two punisher tests. We, therefore, hypothesized that if the magnitude of the smaller amount of food was made more comparable with that of the larger amount of food, this might make Chas less likely to select the larger amount of food. Thus, we decided to manipulate parameters to identify the point at which Chas would stop making “impulsive” choices and come to select the smaller magnitude of food. However, prior to this manipulation, we wanted to verify that three pieces of food were more preferred relative to two pieces of food.

Unsignaled delayed punishment

The purpose of this phase was to evaluate whether the delivery of a delayed, mild verbal punisher following selection of the larger magnitude reinforcer would shift preference from the larger magnitude reinforcer to the smaller magnitude reinforcer. Procedures were identical to those in the reinforcer magnitude phase but, if the child chose the larger magnitude of food, the experimenter made eye contact with the child following a pre-specified delay and said the word “no” in a medium-volume voice with a neutral expression. The next trial immediately followed the “no” statement. If the child chose the smaller magnitude of food, the experimenter did not say “no” but waited the duration of the delay before presenting the next trial. Following two forced-choice trials, the experimenter presented 10 free-choice trials.

Different delays were evaluated in separate phases. Delays ranged in duration from 5 s to 60 s and were held constant within sessions and identical phases. The initial delay was set at 10 s for Jonah and Naomi and at 30 s for Chas. The delay then was increased if the child consistently chose the smaller magnitude reinforcer and decreased if the child consistently chose the larger magnitude reinforcer. In some cases, phases with identical delays were replicated later in the evaluation to verify stability of responding.

Partially signaled delayed punishment

The purpose of this phase was to evaluate whether adding a brief signal to indicate that the delayed, mild verbal punisher was forthcoming would shift preference from the larger magnitude reinforcer to the smaller magnitude reinforcer. Procedures were identical to those in the unsignaled delayed punishment evaluation but, if the child chose the larger magnitude of food, the experimenter pressed a clicker once immediately following the child’s selection so that the clicker would become associated with the delayed “no” statement. If the child chose the smaller magnitude of food, the experimenter did not press the clicker or say “no,” but waited the duration of the delay before presenting the next trial. Following two forced-choice trials, the experimenter presented 10 free-choice trials.

Completely signaled delayed punishment

This phase was conducted with Naomi and Jonah, and the purpose was to evaluate whether adding a signal throughout the delay to the delivery of the mild verbal punisher would shift preference from the larger magnitude reinforcer to the smaller magnitude reinforcer. The method was based on previous laboratory research (Schaal & Branch, 1990) that found that the effectiveness of a delayed outcome on responding was increased as the duration of the signal during the delay to the outcome increased. Procedures were identical to those in the unsignaled delayed punishment evaluation phase but, if the child chose the larger magnitude of food, the experimenter presented a red 7.62 × 12.7 cm index card approximately 12.7 cm in front of the child and kept the index card out for the duration of the delay. If the child chose the smaller magnitude of food, the experimenter did not present the index card or say “no,” but waited the duration of the delay before presenting the next trial. Following two forced-choice trials, the experimenter presented 10 free-choice trials.

Results

Figure 1 displays the results for Jonah, Naomi, and Chas, in the upper, middle, and lower panels, respectively. In phase labels, numbers in parentheses indicate the length of the delay. In data path labels, numbers in parentheses indicate the number of pieces of food.

Figure 1.

Percentage of free-choice trials (out of 10) each option was selected during each session across the various phases.

Punishment Test

Data from this evaluation are depicted in Figure 1, under phases labeled Pun. Test. Initially, “no” functioned as a punisher for all three children. The mean percentage of choices for the food not associated with “no” was 96.7%, 93.3%, and 97.5% for Jonah, Naomi, and Chas, respectively. Thus, the children were more likely to choose the food not associated with “no.” When this phase was reintroduced with Naomi during Sessions 18 through 49, she selected the food not paired with “no” during 55% of free-choice trials (range = 0%-100%). When this phase was reintroduced with Chas during Sessions 18 through 20, he selected the food not paired with “no” during 90% of free-choice trials.

Magnitude Tests

Data from these evaluations are depicted in Figure 1, under phases labeled Mag. Test. All three children initially preferred the larger amount (two pieces) of their preferred food to the smaller amount (one piece). Jonah selected the larger amount during 100% of free-choice trials, Naomi selected the larger amount during 82.5% of free-choice trials (range = 70%-100%), and Chas selected the larger amount during 75% of free-choice trials (range = 30%-100%). However, these effects were not fully replicated with Jonah when we later compared three pieces versus one piece of his preferred food. He selected the larger amount during 46.43% of free-choice trials (range = 10%-70%). His preference eventually stabilized, however, and he chose the larger magnitude of food during at least 60% of free-choice trials during the last three sessions. With Chas, we also examined preference for three pieces of food versus two pieces of food and found that he selected three pieces of food during 95% of free-choice trials (range = 90%-100%).

Unsignaled Delayed Punishment

Data from this evaluation are depicted in Figure 1, under phases labeled Delay. Jonah selected the larger magnitude of food (i.e., two pieces) during 3.33% (range = 0%-10%) of free-choice trials when selection of the larger magnitude of food was followed by the word “no” after a 10-s delay. Jonah chose the larger magnitude of food during 12.5% (range = 0%-20%), 10% (range = 0%-30%), and 0% of free-choice trials when the delay to “no” was increased to 20 s, 30 s, and 60 s, respectively. We later evaluated whether reinstituting the 60-s delayed “no” would still shift preference away from the larger magnitude of food when the larger magnitude of food was increased to three pieces. When this was done, the 60-s delayed “no” now had less of an effect on preference: Jonah selected the larger magnitude of food (i.e., three pieces) during 54.29% of free-choice trials (range = 50%-60%), which was similar to his selection of the larger magnitude of food when it was not followed by a delayed “no” in the last half of the prior phase (i.e., 57.14%, range = 40%-70%).

Naomi selected the larger magnitude of food during 82.5% of free-choice trials. This preference shifted minimally when a “no” statement followed selection of the larger amount of preferred food by 10 s. In this phase, Naomi selected the larger amount of food during 80% of free-choice trials (range = 60%-90%). In the 5-s delayed “no” phase, Naomi selected the larger amount of food during 93.33% (range = 90%-100%), 61.82% (range = 0%-100%), and 49.86% (range = 0%-100%) of free-choice trials across three replications. Naomi’s responding was quite variable within this phase and across replications of this phase, making it difficult to draw conclusions about her sensitivity to delayed punishment. When we replicated the punisher assessment to ensure that “no” still functioned as a punisher, results were rather variable, and she selected food not paired with “no” during an average of 55% of the free-choice trials (range = 0%-100%). However, a punishment effect is evident over the last five sessions in which Naomi selected food not paired with “no” during 98% of the free-choice trials (range = 90%-100%), suggesting that “no” functioned as a punisher, and that the results of Naomi’s unsignaled delayed punishment evaluation did indeed indicate that the delay reduced the punishing efficacy of the “no” statement.

The bottom panel of Figure 1 indicates that Chas’s preference for the larger magnitude of food did not shift when the selection of the larger magnitude of food was followed by the word “no” after a 30-s delay: He chose the larger magnitude of food during 96.67% of free-choice trials (range = 90%-100%). He continued to choose the larger magnitude of food during 100% of the free-choice trials when the delay was decreased to 10 s, and then to 5 s.

Partially Signaled Delayed Punishment

Results from this evaluation are depicted Figure 1, under phases labeled Delay + Partial Signal. For Jonah, adding a brief signal following the selection of the larger magnitude of preferred food slightly decreased his preference for the larger magnitude of food. On average, Jonah selected the larger magnitude of preferred food during 54.29% (range = 50%-60%) of free-choice trials when the delayed punisher was unsignaled and during 45% (range = 40%-50%) of free-choice trials when the delayed punisher was partially signaled.

For Naomi, adding the brief signal following the selection of the larger magnitude of preferred food had small, and inconsistent, effects on preference. Naomi selected the larger magnitude of preferred food during 57.1% (range = 0%-100%) of free-choice trials when the 5-s delayed punisher was unsignaled. On average, Naomi selected the larger magnitude of preferred food during 54.4% (range = 50%-60%) of free-choice trials the first time the 5-s delayed punisher was partially signaled and during 34.7% (range = 0%-50%) of free-choice trials when the partially signaled 5-s delayed “no” phase was reinstated.

For Chas, adding a brief signal following selection of the larger magnitude of preferred food did not alter preference for the larger magnitude of food. Chas selected the larger magnitude of preferred food during 100% of free-choice trials when the 5-s delayed punisher was unsignaled and when the 5-s delayed punisher was partially signaled. When the partially signaled delay was reduced to 0 s, Chas selected the larger magnitude of preferred food during 96.67% of free-choice trials (range = 90%-100%). When the amount of both food options was increased so that Chas could choose between two or three food items, Chas still selected the larger magnitude of food during 100% of the free-choice trials. However, only one session was conducted before Chas’s mother was no longer able to bring him to sessions due to time constraints.

Completely Signaled Delayed Punishment

This phase was implemented only with Jonah and Naomi. Results from this evaluation are depicted in Figure 1, under phases labeled Delay + Complete Signal. For Jonah, adding a signal during the entire duration of the delay following the selection of the larger magnitude of preferred food slightly decreased preference for the larger magnitude. On average, Jonah selected the larger magnitude of preferred food during 38.6% (range = 0%-70%) of free-choice trials when the delayed punisher was completely signaled. However, the association between selection of the larger magnitude of food and the length of the signal was somewhat muddled by Jonah’s variable responding and our inability to replicate the partially and completely signaled 60-s delay phases due to time constraints.

For Naomi, adding a signal during the entire duration of the delay following selection of the larger magnitude of preferred food initially had a drastic effect on preference. On average, Naomi selected the larger magnitude of preferred food during 4% (range = 0%-20%) of free-choice trials when the delayed punisher was completely signaled. However, when we replicated the completely signaled phase, Naomi selected the larger magnitude of food during 60% (range = 50%-70%) of free-choice trials.

Discussion

Previous research on choice typically examined choice between two options, each of which produced a single outcome (e.g., Lerman, Addison, & Kodak, 2006; Vollmer, Borrero, Lalli, & Daniel, 1999), and the present investigation expanded on this work by examining choice behavior when one option produced both an immediate and delayed outcome. Although Woolverton et al. (2011) examined choices of monkeys between varying magnitudes of reinforcement, one of which was followed by a punisher, similar evaluations had yet to be conducted with individuals with ASD.

As previously noted, many choices that individuals with ASD make produce both immediate and delayed reinforcing and punishing outcomes. For example, noncompliance might be more immediately reinforcing to the individual than compliance with academic demands. However, teachers might implement prompting procedures following noncompliance, and prompts have been found to function as punishers for some individuals (Heal & Hanley, 2011). These prompts may be delayed if a teacher must first attend to other students. Results from the current study suggest that it may be helpful to provide a signal to bridge the delay between noncompliance and prompts. For example, a teacher might say, “I’ll be there shortly to help you” if he notices that a student is off-task.

Results of the present study also indicate that delayed mild punishment following the selection of a larger magnitude of preferred food can have idiosyncratic effects across children. For one child, the reinforcing efficacy of the large-magnitude reinforcer competed with the punishing efficacy of a “no” statement. For two children, punishment was most effective when the signal lasted the duration of the delay. However, varying the delay durations and whether the delay was unsignaled, partially signaled, or completely signaled did not produce reliable changes in responding. Interestingly, results were least consistent for the two participants with the strongest verbal repertoires (i.e., the two children who could both produce and respond to multiple-word sentences).

Jonah’s results were the most variable. During the second 60-s unsignaled delayed “no” phase, Jonah’s choice of the large magnitude of edibles increased relative to the first 60-s unsignaled delayed “no” phase (from a M of 0% to a M of 54.3%). One possible reason for why findings differed across replications of this phase is that when punishment delays were initially implemented (Sessions 7-19), delays were gradually increased to 60 s. This procedure could be characterized as delay fading (e.g., Dixon, Rehfeldt, & Randich, 2003) and thus may have increased the likelihood that delayed reprimands affected responding. In the second replication of this phase, however, 60-s delays were not preceded by gradually increasing shorter delays. This may explain the failure to replicate the initial pattern of findings. It also should be noted that Jonah often stated rules during the second 60-s unsignaled delayed “no” phase. For example, during a few sessions, Jonah stated that he was on a diet and therefore needed to eat less. Therefore, his results might not be surprising, given research on the effects of verbal rules on behavior (e.g., Bentall, Lowe, & Beasty, 1985; Faloon & Rehfeldt, 2008; Taylor & O’Reilly, 1997). It would be of interest to replicate procedures used in the current study with individuals with less-developed verbal repertoires to clearly isolate the effects of delayed punishment on choice behavior.

Interestingly, it appeared as though the punishing efficacy of “no” decreased over time, and this is clearly seen in the extent to which Naomi selected the option associated with “no” during the first punishment test (M = 6.67%) relative to the second punishment test (M = 44.69%). A decrease in the punishing efficacy of “no” also may account for the fact that Naomi selected the larger magnitude of preferred food followed by a delayed “no” during a mean of 4% of free-choice trials during the initial completely signaled delayed punishment phase, but during a mean of 60% of free-choice trials when the same phase was replicated. A decrease in the punishing efficacy of “no” is not surprising because “no” may only function as a conditioned punisher when paired with aversive events. In the present study, “no” was never paired with a programmed aversive event. In addition, “no” was said in a neutral tone at a medium volume, which likely differs from the way “no” is said when used as a reprimand in the natural environment.

The finding that the punishing efficacy of “no” decreased with time highlights the need to use individualized assessments to identify stimuli that might be more likely to retain their punishing efficacy over time (see also Fisher, Piazza, Bowman, Hagopian, & Langdon, 1994). In the present investigation, we used the same stimulus for all children. In addition, we opted to use a relatively minor punisher. It may be argued that the effects being reported are quite reasonable given the use of such a mild punisher, and one that very well might not be sustained long-term. However, future researchers may wish to evaluate whether the periodic pairing of mild verbal punishers with other more potent punishers delivered contingent on problem behavior outside of the session would maintain the efficacy of the mild punisher.

Naomi’s results tentatively support the findings from the basic literature on the effects of signal length (e.g., Schaal & Branch, 1990). Schaal and Branch (1990) evaluated the relation between signal duration and rates of key pecking by pigeons. Briefly signaled 27-s delays to reinforcement typically maintained low rates of responding, whereas when the delay was completely signaled, rates of responding increased. However, similar to the present investigation, results were variable across subjects. Therefore, individual evaluations should be conducted to determine whether signals increase the efficacy of mild, delayed verbal punishers. In addition, future researchers may wish to equate the types of signals used. In the present investigation, the topographies of the partial (clicker) and full (red index card) signal differed and may have influenced responding independent of signal length. For example, if Naomi was not attending or was engaging in other behavior when the experimenter presented the click during the partial signal phase, this phase would be functionally similar to the unsignaled delay phase. However, the index card used in the completely signaled phase may have been more salient. In the current study, we hypothesized that the partial signal would be sufficient to increase the efficacy of the mild, delayed verbal punisher and therefore did not initially program a signal that then could easily be continued throughout the entire delay.

Interestingly, an unexpected finding was the way in which the magnitude of the reinforcer directly competed with the efficacy of the punisher for Chas. Indeed, during the unsignaled delayed punishment evaluation, Chas chose the larger magnitude of preferred food on 98.3% of the free-choice trials, on average (range = 90%-100%) across all delays, regardless of whether the delays were signaled or unsignaled, even though selection of the larger magnitude of food was followed by a statement that had been shown to function as a punisher during punishment tests administered both at the beginning of and halfway through the evaluation. Thus, despite “no” being a punisher when the food magnitudes were equal, it was ineffective in shifting preference from the larger to the smaller amount of preferred food. Future researchers may need to use punishers of greater magnitude that are able to compete with the reinforcing efficacy of large-magnitude reinforcers.

In any extended evaluation involving multiple phases, the potential influence of order certainly must be considered. For example, Jonah was less likely to select the larger amount of food in the second magnitude test, relative to the first magnitude test. One possible explanation might be that this was due to differences in the magnitudes compared in each test: Jonah was offered choices between one piece and two pieces in the first magnitude test and one piece and three pieces in the second magnitude test. However, this explanation seems unlikely, given that the magnitudes evaluated in the second magnitude test should have increased preference for the larger amount. A more likely explanation is that Jonah’s recent history with prior phases in which larger amounts of food were paired with mild, delayed verbal punishers decreased the reinforcing efficacy of larger amounts of food.

Phase order must also be considered along with the possibility that the punishing efficacy of “no” is likely to vary as a function of the extent to which the statement has recently been paired with aversive events. Indeed, it is possible that the results were confounded by a gradual decrease in the punishing efficacy of “no” across the entire evaluation. That is to say, as the children learned that the experimenter never implemented additional aversive contingencies after saying “no,” the statement may have become less aversive over time. This makes it difficult to evaluate controlling variables responsible for responding during phases that were only implemented later on in the evaluation (e.g., the completely signaled punishment phase). Future researchers may wish to rapidly alternate unsignaled, partially signaled, and completely signaled delayed punishment phases, and use a multielement design to examine experimental control. Future research also might add punishment test probes throughout the entire experimental evaluation so as to obtain multiple assessments of the punishing efficacy of “no” throughout the evaluation. This would allow researchers to better assess the extent to which decreases in the punishing efficacy of mild, verbal punishers might affect results of the evaluation. It also may be useful to conduct both ascending and descending sequences of delays to evaluate any hysteresis effects; however, every attempt should be made to conduct such evaluations as quickly as possible, before the efficacy of mild, verbal punishers might decrease.

In addition to controlling for phase order and conducting evaluations as quickly as possible so as to minimize the likelihood of “no” losing its punishing effectiveness during the evaluation, future researchers should consider how best to control for variations in potential establishing operations. We did not ask parents to restrict access to the food items used as reinforcers in the current study because all three children consumed only a limited range of foods. Therefore, it might be the case that variations in access to the reinforcers outside the experimental situation may have altered the reinforcing efficacy of food. However, it is to be noted that Jonah’s results were rather stable up until Session 19, and Chas’s results were stable throughout the duration of the evaluation, suggesting that, at least in these two children, variations in establishing operations likely had little effect.

Taken together, the results from the current study could be used to inform clinical practice in cases where one wishes to shift response allocation to responses that do not result in delayed punishment. For example, consider a case in which a student consistently chooses to be off-task during a work period, despite the fact that doing so means they will have to miss recess to complete the work. Teachers could state rules (e.g., “Finish your work before taking a break”) at the start of the work period and signal the loss of recess when they notice off-task behavior by stating the contingency and writing the student’s name on the board. Delay fading also could be implemented by initially scheduling difficult work periods right before recess (i.e., close in time to potential negative punishment in the form of the loss of recess) and then gradually scheduling work periods farther from recess. Finally, if the loss of recess is not a potent enough punisher, the teacher may increase the efficacy of the negative punishment contingency by increasing the reinforcing value of recess by adding new activities or increasing recess duration.

In summary, the results of the current study suggest that mild, delayed verbal punishment can be temporally effective; however, there are multiple dimensions to be taken into account when considering the effect that a mild verbal punisher will have on choice. For example, larger magnitude reinforcers may strongly compete with punishment. In addition, self-rules can potentially influence responding. Therefore, the effect of reinforcer and punisher magnitude on responding should be explored on an individual basis. The current investigation is the first attempt to replicate procedures similar to those used by Woolverton et al. (2011) with children with ASD. Although this effort represents one type of translational research outlined by Mace and Critchfield (2010), it highlights some of the difficulties in studying punishment under human operant arrangements with arbitrary responses. In particular, it is ethically indefensible to use potent punishers with arbitrary responses. Future research will need to evaluate the effects of delayed punishment under more naturalistic situations where potent punishers can be programmed following severe problem behavior. Nevertheless, results from the current study suggest that even mild verbal punishers can be affected by the delay to their delivery and the type of signal used to bridge the delay.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Author Biographies

Jolene R. Sy is an assistant professor in the Department of Psychology at the University of Maryland, Baltimore County. She studies the ways in which delayed outcomes affect skill acquisition by individuals with intellectual and developmental disabilities.

Leonard Green is a professor in the Department of Psychological and Brain Sciences at Washington University in St. Louis. He studies choice and decision making in rats, pigeons, and people. His research on choice extends to the areas of self-control, behavioral economics, and the discounting of delayed and probabilistic outcomes.

Olivia Gratz is a graduate student in the Department of Psychology at Western Michigan University with research interests in clinical behavior analysis, including but not limited to acceptance and commitment therapy and behavioral activation. Her master’s thesis focused on evaluating the resurgence of problem behavior with three functionally equivalent responses.

Thea Ervin is a graduate of the School of Social Work at Saint Louis University. Her master’s thesis focused on the use of behavior skills training combined with observational learning to train appropriate responses to peer aggression by students with emotional and behavior disorders.

References

Abramowitz

A. J.

O’Leary

S. G.

(1990). Effectiveness of delayed punishment in an applied setting. Behavior Therapy, 2, 231-239.

Banks

R. K.

Vogel-Sprott

(1965). Effect of delayed punishment on an immediately rewarded response in humans. Journal of Experimental Psychology, 70, 357-359.

Baron

(1965). Delayed punishment of a runway response. Journal of Comparative and Physiological Psychology, 60, 131-134.

Bentall

R. P.

Lowe

C. F.

Beasty

(1985). The role of verbal behavior in human learning: II. Developmental differences. Journal of the Experimental Analysis of Behavior, 43, 165-181.

Critchfield

T. S.

Kollins

S. H.

(2001). Temporal discounting: Basic research and the analysis of socially important behavior. Journal of Applied Behavior Analysis, 34, 101-122.

DeLeon

I. G.

Iwata

B. A.

(1996). Evaluation of a multiple-stimulus presentation format for assessing reinforcer preferences. Journal of Applied Behavior Analysis, 29, 519-533.

Dixon

M. R.

Rehfeldt

R. A.

Randich

(2003). Enhancing tolerance to delayed reinforcers: The role of intervening activities. Journal of Applied Behavior Analysis, 36, 263-266.

Faloon

B. J.

Rehfeldt

R. A.

(2008). The role of overt and covert self-rules in establishing a daily living skill in adults with mild developmental disabilities. Journal of Applied Behavior Analysis, 41, 393-404.

Fisher

Piazza

C. C.

Bowman

L. G.

Hagopian

L. P.

Langdon

N. A.

(1994). Empirically derived consequences: A data-based method for prescribing treatments for destructive behavior. Research in Developmental Disabilities, 15, 133-149.

10.

Green

Myerson

(2004). A discounting framework for choice with delayed and probabilistic rewards. Psychological Bulletin, 130, 769-792.

11.

Green

Myerson

Lichtman

Rosen

Fry

(1996). Temporal discounting in choice between delayed rewards: The role of age and income. Psychology and Aging, 11, 79-84.

12.

Heal

N. A.

Hanley

G. P.

(2011). Embedded prompting may function as embedded punishment: Detection of unexpected behavioral processes within a typical preschool teaching strategy. Journal of Applied Behavior Analysis, 44, 127-131.

13.

Kjelgaard

M. M.

Tager-Flusberg

(2001). An investigation of language impairment in autism: Implications for genetic subgroups. Language and Cognitive Processes, 16, 287-308.

14.

Lerman

D. C.

Addison

L. R.

Kodak

(2006). A preliminary analysis of self-control with aversive events: The effects of task magnitude and delay on the choices of children with autism. Journal of Applied Behavior Analysis, 39, 227-232.

15.

Mace

F. C.

Critchfield

T. S.

(2010). Translational research in behavior analysis: Historical traditions and imperative for the future. Journal of the Experimental Analysis of Behavior, 93, 293-312.

16.

Schaal

D. W.

Branch

M. N.

(1990). Responding of pigeons under variable-interval schedules of signaled-delayed reinforcement: Effects of delay-signal duration. Journal of the Experimental Analysis of Behavior, 53, 103-121.

17.

Taylor

O’Reilly

(1997). Toward a functional analysis of private verbal self-regulation. Journal of Applied Behavior Analysis, 30, 43-58.

18.

Trenholme

I. A.

Baron

(1975). Immediate and delayed punishment of human behavior by loss of reinforcement. Learning and Motivation, 6, 62-79.

19.

Vollmer

T. R.

(2002). Punishment happens: Some comments on Lerman and Vorndran’s review. Journal of Applied Behavior Analysis, 35, 469-473.

20.

Vollmer

T. R.

Borrero

J. C.

Lalli

J. S.

Daniel

(1999). Evaluating self-control and impulsivity in children with severe behavior disorders. Journal of Applied Behavior Analysis, 32, 451-466.

21.

Woolverton

Freeman

Myerson

Green

(2011). Suppression of cocaine self-administration in monkeys: Effects of delayed punishment. Psychopharmacology, 186, 99-106.