JOURNAL OF APPLIED BEHAVIOR ANALYSIS
2017, 9999, n/a–n/a
NUMBER
9999 ()
USING PROGRESSIVE RATIO SCHEDULES TO EVALUATE TOKENS AS GENERALIZED CONDITIONED REINFORCERS DANIELLE RUSSELL, EINAR T. INGVARSSON AND JENNIFER L. HAGGAR UNIVERSITY OF NORTH TEXAS AND CHILD STUDY CENTER
AND
JOSHUA JESSEL UNIVERSITY OF NORTH TEXAS
The properties of operant reinforcers are dynamic and dependent on a number of variables, such as schedule and effort. There has been sparse research on the generalized conditioned properties of token reinforcement. We evaluated leisure items, edible items, and tokens using a progressive ratio schedule with three children with diagnoses of ASD and developmental delays. The highest break points occurred during the token reinforcement condition for two out of three participants, but response rates tended to be higher with edibles. We then evaluated the effects of presession access to edibles on the break points of edible items and tokens with two participants. Break points decreased only in the edible reinforcement condition, and the participants chose to work for leisure items rather than edibles when presession access to edibles was in place. These findings suggest that the tokens functioned as generalized conditioned reinforcers. Key words: generalized conditioned reinforcement, progressive ratio schedule, motivating operations, tokens
Efficient identification of reinforcers and evaluation of their ongoing effectiveness is a critical aspect of behavioral interventions. Applied researchers typically conduct preference assessments to determine a relative ranking of stimulus value, under the assumption that the ranking predicts reinforcer effectiveness (e.g., DeLeon & Iwata, 1996; Fisher et al., 1992; Pace, Ivancic, Edwards, Iwata, & Page, 1985; Roane, Vollmer, Ringdahl, & Marcus, 1998; Windsor, Piche, & Locke, 1994). After preference is determined, a reinforcer assessment with mastered responses might be conducted, during which each response produces access to a preferred stimulus (Fisher &
Mazur, 1997; Piazza, Fisher, Hagopian, Bowman, & Toole, 1996). However, reinforcer effectiveness may fluctuate dependent on various parameters, including fixed ratio (FR) requirements (DeLeon, Iwata, Goh, & Worsdell, 1997; Tustin, 1994). DeLeon et al. (1997) compared the reinforcing potency of two similar stimuli previously found to be roughly equivalent in preference ranking. When these stimuli were available under concurrent FR 1 schedules, both participants—adults diagnosed with developmental disabilities—showed little or no differences in preference for either stimulus. However, when the schedule requirements increased to FR 5, a clear preference for one of the stimuli emerged for both participants. This difference maintained when the schedule requirements increased to FR 10. Thus, reinforcer potency is not an absolute value for a given stimulus. Given that schedule requirements influence reinforcer potency and the terminal goal of most applied efforts is to identify preferred stimuli that will maintain and increase
This study was conducted in partial fulfillment of the requirements for the master’s degree of the first author at the University of North Texas. We thank Kimberly Fairman and Melinda Robison for assistance with data collection. Joshua Jessel is now affiliated with Queens College. Address correspondence to: Einar Ingvarsson, who is now at the Virginia Institute of Autism, 943 Glenwood Station LN, STE 201, Charlottesville, VA 22901. Email:
[email protected] doi: 10.1002/jaba.424
© 2017 Society for the Experimental Analysis of Behavior
1
2
DANIELLE RUSSELL et al.
responding over time, Roane, Lerman, and Vorndran (2001) suggested using progressive ratio (i.e., PR) schedules to evaluate reinforcer potency. In a PR schedule, the number of responses required to access reinforcement systematically increases within-session. With PR schedules, relative reinforcement effects can be identified by comparing the obtained break point for each stimulus (Hodos, 1961). One advantage of this method is that it allows for an evaluation of reinforcing efficacy under increasing response requirements. In two experiments, Roane et al. (2001) evaluated the effectiveness of preferred stimuli using PR schedules. Four individuals who were diagnosed with developmental disabilities participated in the first study and three of those individuals participated in the second study. Roane et al. found that whereas response patterns of higher and lower preferred stimuli failed to differentiate with FR 1 schedules, PR schedules produced differences in break points between various stimuli. They also discovered that stimuli with higher break points were overall more effective in the treatment of destructive behavior. Even following more extensive reinforcer assessments using PR schedules, reinforcer effectiveness can fluctuate as a function of motivating operations (MO; Laraway, Snycerski, Michael, & Poling, 2003). A common concern with edible and leisure reinforcers is the deleterious effects of abolishing operations on the reinforcing potency of these stimuli (McAdam et al., 2005; Murphy, McSweeney, Smith, & McComas, 2003). One major advantage of tokens is that they should not be subject to MO effects specific to each backup reinforcer, if they function as generalized conditioned reinforcers (Cooper, Heron, & Heward, 2007). Token economies are reinforcement systems that involve the contingent delivery of tokens following the occurrence of the targeted behavior. Accumulated tokens can later be exchanged for back-up reinforcers (e.g., food and toys). Through their association with back-up
reinforcers, the tokens may acquire conditioned reinforcing properties (Cooper et al., 2007; DeFulio, Yankelevitz, Bullock, & Hackenberg, 2014; Hackenberg, 2009). Early laboratory research evaluated token systems and the extent to which tokens functioned as conditioned reinforcers (Cowles, 1937; Kelleher, 1956, 1958; Wolfe, 1936). These studies demonstrated that poker chips could be established as conditioned reinforcers for chimpanzees. There are a number of advantages to using generalized conditioned reinforcers (Kazdin & Bootzin, 1972). Conditioned reinforcers are less susceptible to the effects of satiation and can be used to bridge the delay between initial responses and access to preferred items or activities. Additionally, conditioned reinforcers maintain performance and responding over extended periods of time, involve minimal interruptions, and allow for sequences of responses to be reinforced easily. They can also be more potent than any single primary reinforcer. Ayllon and Azrin (1968) described several potential advantages of tokens over other generalized conditioned reinforcers (such as praise statements). First, the number of tokens quantitatively represents the amount of reinforcement earned. Second, tokens are portable, can remain with the client, and can be delivered easily across a variety of contexts and environments. Third, there is no limit to the potential number of tokens one can earn. Fourth, tokens are durable and allow for consistency and standardization. Fifth, tokens offer a visual representation of improvement and progress (see also Kazdin & Bootzin, 1972). Whereas a considerable number of basic research studies with nonhuman subjects have explored the reinforcing properties of tokens, little is known about the extent to which tokens function as generalized conditioned reinforcers in the context of application (Hackenberg, 2009). An exception is a study by Moher, Gould, Hegg, and Mahoney (2008). The participants were five children with autism and other developmental
TOKENS AS GENERALIZED CONDITIONED REINFORCERS disabilities. Moher et al. conducted reinforcer assessments with tokens and manipulated the MO (i.e., presession free access) for the backup edible reinforcers. In a parametric analysis, tokens were exchangeable with one, two, or three backup reinforcers. Moher et al. found that tokens exclusively exchangeable for one backup reinforcer lost the ability to maintain responding on dense FR schedules when preceded by sessions of free access to that reinforcer. However, tokens continued to maintain responding when exchangeable for multiple backup reinforcers, even when the value of one of the reinforcers was previously abolished via satiation. These findings suggest that the tokens functioned as generalized conditioned reinforcers. We conducted the current study as an extension of Moher et al. (2008) by including PR schedules to determine reinforcer effectiveness. In addition, we included qualitatively different reinforcers (leisure items and edible items) in an evaluation of the effects of different levels of presession exposure on the effectiveness of tokens (exchangeable for both edibles and leisure items) versus edibles only. METHOD Participants Three students attending a school for children with developmental disabilities served as participants. The school director identified participants based on whether they had experience with token economies, had low levels of problem behavior and noncompliance, and had acquired a sufficient number of math facts to be used as tasks during the study. All three children, Damien, Carmen, and Zane, participated in Phase 1 of the study. However, only Damien and Carmen participated in Phase 2. All three participants had received standard token training, based largely on the procedures described by Cooper et al. (2007). The participants had extensive experience (at least 14 months) with token economies at the school. Damien was a 7-year-old boy who was diagnosed with developmental delays and autism
3
spectrum disorder. According to the Woodcock-Johnson III Test of Achievement (WJ-III; Woodcock, McGrew, & Mather, 2001), he functioned on a late-first grade/earlysecond grade level. Prior to the study, Damien had experience using tokens for about 18 months. On average, Damien completed two to three responses before earning each token, and he earned about 10 tokens before accessing reinforcers. According to his teacher’s report, Damien typically chose to exchange his tokens for video games, board games, and occasionally more lessons. Carmen was an 8-year-old girl who was diagnosed with developmental delays and functioned on a kindergarten grade level, according to the WJ-III. She had experience using tokens for about 14 months prior to the study. On average, she earned about 10 tokens before accessing reinforcers in the classroom. With difficult tasks, only one response was required per token, but with mastered tasks, two to three responses were required before earning a token. According to her teacher’s report, Carmen typically chose to exchange her tokens for small toys (i.e., Polly Pockets, LEGO® bricks, etc.), board games, and candy. Zane was a 7-year-old boy who had a diagnosis of attention deficit hyperactivity disorder and autism spectrum disorder. He functioned on a first grade level according to the WJ-III. Zane had experience using tokens for about 2 years. The number of responses required to earn a single token varied from one to five, depending on task difficulty. According to his teacher’s report, Zane typically earned about 10 tokens before accessing reinforcers and chose to exchange his tokens for LEGO® bricks, Wii®, iPad®, and jelly beans. Setting All conditions and sessions were conducted in a room (approximately 2 m x 3 m) containing a bookshelf, two desks, four chairs, a tripod with video camera, and a three-tier plastic storage bin. Other children were not present in the room
4
DANIELLE RUSSELL et al.
during sessions. One session was conducted per day for each participant, 3 to 5 days per week (depending on participant availability). Tasks and Materials During sessions, participants worked on previously mastered academic skills. The first author interviewed their teachers to identify skills that had previously been mastered and selected tasks on the basis of each participant’s individual level of development and academic ability. Addition math facts were chosen for all participants. Carmen worked on addition math facts 0 and 1 + 0 through 11. Both Zane and Damien worked on addition math facts 0 through 5 + 0 through 11. In all sessions and conditions, one participant was seated at a desk and the experimenter (either the first or third author) sat to his or her right side. Each math fact was printed on a 10.8 cm x 14.0 cm laminated paper. An aluminum pan (22.9 cm x 27.9 cm) with a card attached with the printed phrase “I’m done” was located on the far left corner of the desk. The math fact sheets were stacked in the far right corner of the desk. After taking a math fact sheet from the stack and writing the answer on the sheet under the equals sign, the participant placed the completed math fact in the pan. The participant was given a dry-erase marker to write the answers and a tissue to erase mistakes. The token board consisted of a laminated file folder with the label “token board” at the top. It had three long strips of Velcro centered beneath the label. The tokens were brown paper circles, about the size of a quarter, laminated with the word “token” printed across the center and Velcro affixed to the back. In the token condition, the therapist placed the token board in between the pan and the sheets. Measurement All sessions were video recorded and the primary observers scored responses using a data
collection app on an iPod. The experimenters (first and third author) and trained observers collected data on correct and incorrect responses, total session time, and the progressive ratio (PR) break point. Observers scored a response each time the child placed the completed task (math problem) in the pan. When the child had written the correct answer on the sheet, observers scored that response as correct. Observers counted a response as incorrect when either (a) the answer was not written clearly enough for the therapist to read or (b) the child wrote the wrong answer on the sheet. The number of correct responses was then divided by the duration of the session to calculate rate of responses per min (rpm). All participants responded at near 100% accuracy throughout the study (M = 96.4%, SD = 2); therefore, we report only correct responses. Reinforcer access time was subtracted from session duration prior to calculating rate. Session time began after delivering the initial instructions and ended when the child asked to be finished or when the break point criterion was reached. Observers circled the PR break point when the child was off-task for 2 min or when the child said, “I’m done.” If the session ended after 2 min of off-task behavior, this duration was also subtracted prior to calculating the rate. In the token condition, the experimenter also noted whether the participants selected an edible or leisure item as a backup reinforcer. Interobserver Agreement (IOA) Interobserver agreement (IOA) was assessed for at least 30% of sessions by having a second observer independently collect data from videos using pencil and paper. We then counted point by point agreements versus disagreements across the two observers for incorrect and correct responses and PR break point, and calculated IOA for each session by dividing the total number of agreements by the total number of agreements plus disagreements and multiplying
TOKENS AS GENERALIZED CONDITIONED REINFORCERS by 100. Mean IOA for both Carmen and Damien was 100% for correct responses and PR break point. Mean interobserver agreement for Zane was 99% (range, 98% - 100%) for correct responses and 100% for the PR break point. Procedural Fidelity To assess procedural fidelity, we created a checklist specifying important aspects of the procedures. These included giving instructions, presenting the task, delivering consequences according to the PR schedule, implementing the break point, managing the token system, and implementing presession access procedures. An independent observer observed randomly selected video-recorded sessions (with at least one session from each phase and condition for each participant) and scored experimenter behavior using the checklist. This observer had not previously been a part of running the study or collecting data on participant behavior. We collected procedural fidelity data for 25.8% of sessions for Damien, 23.8% of sessions for Carmen, and 25% of sessions for Zane. Procedural fidelity was 100% for Damien, 99.5% (range, 89.5%-100%) for Carmen, and 100% for Zane. We also obtained IOA for the procedural fidelity measures for 12.5% of scored sessions for Damien, 20% of sessions for Carmen, and 20% of sessions for Zane. IOA for the procedural fidelity measures was 100% for all three participants. Procedures: Reinforcer Assessment (Phase 1) Preference assessment. Two paired-stimulus preference assessments were conducted with each child (Fisher et al., 1992). One assessment included edible items and the other included leisure items. Five to eight items were paired once with every other item in a counterbalanced sequence. Following a selection response, the participant had access to the edible/leisure item for 1 min. The top three edible and leisure items were then used as reinforcers
5
in the upcoming progressive ratio conditions. For Damien, the top three edibles were 3 Musketeers Bar®, M&M®, and chocolate chip cookie. iPad®, LEGO® bricks, and an Angry Birds Game were the top three leisure items. For Carmen, the top three edibles were jelly bean, Oreo®, and KitKat®. The top three leisure items were iPad®, bowling, and an Angry Birds game. For Zane, the top three edibles were Doritos®, Oreo®, and KitKat®. The top three leisure items were iPad®, bowling, and LEGO® bricks. Progressive ratio assessment. There were three conditions in the progressive ratio assessment: edible, leisure, and token reinforcement. The top three edibles from the preference assessment were used in the edible condition, the top three leisure items were used in the leisure condition, and tokens exchangeable for any of those six items were used in the token condition. For the duration of the study (both Phases 1 and 2), the participants did not have access to these edibles and leisure items at the school. Additionally, the experimenters asked the participants’ parents to ensure that they did not have access to these items at home. Prior to each session, the experimenter placed the items relevant to the current experimental condition on the table and the child selected what he or she wanted to earn during that session. Immediately after the child selected an item, the experimenter removed all other items. Session materials were then placed on the child’s desk (marker, tissue for erasing, math facts, and “finished” pan). The experimenter then issued the following instruction: “Today you will be working on math facts. You will receive _______ for completing math problems. Just grab a sheet from the stack, answer the problem and then put it in the finished bin. If at any point you want to be all done working, you can tell me, ‘I’m done’ (experimenter points to the ‘I’m done’ card). What do you say when you want to be all done?” The session began after the experimenter delivered the instructions and
6
DANIELLE RUSSELL et al.
the child answered the question by saying, “I’m done.” During sessions, the children worked on math fact tasks by writing the solution to each problem in the designated area on the laminated sheet. The participant then placed the completed sheet in the “finished” pan, took the next sheet from the stack, and continued. When the participant completed the current ratio schedule requirement, the experimenter placed a hand on the math sheets and delivered the programmed consequence. The ratio schedule requirements increased arithmetically (step size of 1) beginning with one response to access reinforcement, then 2, 3, and so on. The experimenter delivered an edible item in the edible condition, a leisure item in the leisure condition, and a token in the token condition. The child’s selection at the beginning of session determined which edible or leisure item was used. In the edible and leisure conditions, the participant was given 1-min access to consume the edible or play with the leisure item after completing each schedule requirement. If the participant made an error, the therapist said, “That’s not right” and allowed the child a few seconds to correct the error. If the participant did not know the answer, the experimenter vocally prompted the correct answer. Any incorrect or prompted answers were counted toward the FR schedule requirement after the child wrote the correct answer on the sheet and placed it in the pan. Thus, an incorrect response followed by a prompted response was counted only once towards the schedule requirement. If the participant tried to engage in conversation during the task or reinforcer time, the experimenter responded by saying, “We can talk about that later if you want” and ignored all additional initiations within 30 s of the verbal redirection. The experimenter verbally redirected the child again if needed after the 30 s had passed. The PR break point was reached and the session was terminated when the child asked to be
finished or when no target responses occurred for 2 min. When this occurred in the edible and leisure conditions, the experimenter immediately walked the child back to his or her classroom. In the token condition, the experimenter and child counted the number of tokens and the child was given access to the item he or she selected prior to the session. Each token was worth one edible or 1-min access to the leisure item. For example, if the break point was PR 10, the child had earned 10 tokens that were exchangeable for 10 min of play time with the leisure item or 10 pieces of the edible. Design. The three reinforcement conditions were arranged sequentially according to an ABC design with the order of conditions counterbalanced across participants. Sessions were conducted until the experimenters (first and second authors) judged that the data were sufficiently stable. A minimum of five sessions was conducted in each condition. Stability was judged by first looking at the most recent three data points. The condition ended if a trend was absent. If a possible trend was present, the experimenters examined the most recent five data points and ended the condition if a trend was absent. The number of sessions in each condition for all participants ranged from 5 to 13 (M = 8.9, SD = 2.6). Procedures: Presession Access Evaluation (Phase 2) Baseline. There were two conditions in this phase: edible and token. The leisure items condition was not included because only the motivating operation (MO) for edibles was manipulated. However, the participants could still choose to earn leisure items in the token condition. The procedures were similar to the progressive ratio assessment. Unlike the progressive ratio assessment, the participants did not choose backup reinforcers at the beginning of the session, but rather at the end of session after all the tokens had been earned. This
TOKENS AS GENERALIZED CONDITIONED REINFORCERS change was made to better approximate the procedures in place in their classroom. Presession access comparison. During this phase, each session in both conditions was preceded by a presession access period. During presession access, the child was seated at the desk and given access to the top three edibles from the preference assessment. The instructions given were as follows: “You can eat as much of the snacks as you want. If at any point you are full, you can just tell me, ‘I’m full’.” The experimenter then started a timer. During the first presession access condition the maximum time was set to 5 min. When the timer beeped or when the participant stated that he or she was full or wanted to stop eating, the PR session for that day was initiated. The initial duration of presession access did not result in a change in behavior. Because the participants consumed the edibles rather slowly, we concluded that 5 min might not be enough to allow for satiation. Thus, we increased the presession access period to 10 min. At the same time, we observed that the participants had rarely picked a leisure item as a backup reinforcer in the token condition in baseline. In fact, Carmen had picked leisure items only once, and Damien had never picked the leisure items. We suspected that the leisure items might have lost their reinforcing value. Additionally, Damien refused to exchange his tokens for either leisure or edible items in the token condition during the 5-min presession access, opting instead to go back to his classroom. Therefore, we also increased the range of leisure items that the participants could choose as backup reinforcers in the token condition. From that point on, the participants could select any toy, game, or activity available in the classroom. However, the edibles available in the edible condition remained unchanged. Subsequently, we conducted a third presession access condition with Carmen only, in which we didn’t set specific time limits for her to consume the edibles prior to the session
7
(i.e., unlimited time). Again, this was because she tended to eat very slowly. Presession access lasted on average 23 min and 40 s for Carmen during this phase. Return to baseline. The second baseline was identical to the initial baseline, except the participants had access to a broader array of leisure items as backup reinforcers in the token condition, as in the preceding presession-access condition. Experimental design. During the presession access comparison, the progressive ratio conditions (token vs. edible) were conducted in a multielement design. The order of the conditions was determined in a quasirandom fashion such that no condition was conducted more than two times in a row. The experimental evaluation of the effects of presession access was conducted in a reversal design.
RESULTS Phase 1: Reinforcer Assessment Figure 1 displays the response rate and break point performance across the three reinforcement conditions during the progressive ratio assessment. Break points for each condition were relatively similar for Damien, with the highest mean break point in the edible condition (M = 10.2, SD = 2.1), followed by the leisure items (M = 9.4, SD = 3.6), and token reinforcement (M = 7.6, SD = 3.7). The response rates were also highest in the edible condition (M = 6.5 rpm, SD = .8), followed by the leisure items (M = 5.9 rpm, SD = .9), and then tokens (M = 5.6 rpm, SD = 1.1). For Carmen, leisure items had a lower mean break point (M = 3, SD = 2.3) than edible (M = 5.1, SD = 2.9) and token (M = 5.2, SD = 3.1) reinforcers. The highest mean response rate occurred in the edible condition (M = 5.7 rpm, SD = 1.2), followed by the leisure items (M = 5.2 rpm, SD = 1.1), and tokens (M = 4.5 rpm, SD = 1.5).
DANIELLE RUSSELL et al.
8 8
Damien
6
15
10
4 5
2
Token Edible Leisure
0
Carmen
6
15
10
4 5
2 0
Break point
Responses per minute
8
0
0
15
Zane
30
10
20
5
10
0
0 10
20
30
Sessions Figure 1. Rate of responding and break points during the progressive ratio assessment (phase 1).
The highest mean break point for Zane was observed for the token condition (M = 15.7, SD = 5.2), followed by the edibles (M = 11.7, SD = 3.3), and then leisure items (M = 7.4, SD = 2.8). The highest response rates were observed in the edible condition (M = 10.7 rpm, SD = 1.3). The token condition was next (M = 10 rpm, SD = 1.4), and the lowest response rates occurred in the leisure item condition (M = 6.5 rpm, SD = .7).
Phase 2: Presession Access Comparison During the presession access baseline for Damien (Figure 2, top panel), there was little
difference in break points across the edible reinforcers condition (M = 13.4, SD = 5.7) and the token reinforcers condition (M = 15.4, SD = 6.3). These undifferentiated results between the edibles (M = 9.3, SD = 1.2) and tokens (M = 9, SD = 1) maintained after the 5-min presession access to the edible items was introduced. Following the introduction of the 10min presession access condition, Damien contacted the ratio requirements in only one session during the edible reinforcement condition (M = 1.75, SD = 3.5). However, break points remained stable and high during the token reinforcement condition (M = 10.5, SD = 2.6). Break points during the edible
TOKENS AS GENERALIZED CONDITIONED REINFORCERS
Figure 2. Break points for Damien (top panel) and Carmen (bottom panel) across varying durations of presession access to edible reinforcers in phase 2.
reinforcement condition (M = 5.5, SD = 3.8) returned to previous levels during the second baseline. For Carmen, higher break points were achieved in the token reinforcement condition (M = 7.1, SD = 2.1) in comparison to the edible reinforcement condition (M = 6.4, SD = 2.9) during baseline (Figure 2, bottom panel). Although there was a downward trend in token condition break points during the 5-min presession access period (M = 6, SD = 2.2), this condition was largely unaffected by the changes in presession access. High break points continued to occur during the 10 min (M = 6.7, SD = .6) and unlimited conditions (M = 6.8, SD = 1.5). A slight decrease in the level of edible condition break points was observed for
9
Figure 3. Cumulative selections of backup reinforcers that were exchanged during each session in which tokens were earned in phase 2.
Carmen during the 5-min (M = 3.5, SD = .6) and 10-min (M = 3.7, SD = .6) presession access condition. However, level of break points reduced dramatically and responding was eliminated altogether in many sessions during the unlimited access condition (M = 1.4, SD = 1.8). Higher break points were again obtained in the edible (M = 3.7, SD = .6) and token (M = 6.7, SD = 2.9) conditions during the return to baseline. Figure 3 shows whether the participants selected edible or leisure backup reinforcers during token reinforcement sessions. Both Damien and Carmen exclusively exchanged their tokens for edible items during baseline. During the 5-min presession access, Damien refused to exchange his tokens for either edibles
DANIELLE RUSSELL et al.
10
or tokens, choosing instead to return to his classroom. Carmen continued exchanging her tokens for edibles (with the exception of the last session in this phase). Their selections shifted to leisure items when the 10-min presession access to edibles was introduced (which coincided with a broader range of leisure items being made available), and Damien continued to select leisure items even during the second baseline. Carmen chose leisure items in the first token session in this phase, edibles in the second session, but declined both in the third and final session, opting to go to her classroom instead. After the range of available leisure items was increased, both participants selected leisure items that had not been available before. Damien chose iPod Touch® six times. However, it should be noted Damien played the same game on his personal iPod Touch as he had on the iPad in the leisure condition; the difference was that a higher game level was available on his personal iPod. Interestingly, there were three occasions on which he chose not to engage in leisure activities, but requested to go back to working on academic math tasks. Carmen chose LEGO® bricks (an item that was available during the initial preference assessment, but had not been highly preferred at that time) seven times, but also picked one of the original items (iPad®) three times. Carmen declined leisure items and chose to rejoin scheduled activities in the classroom on one occasion. DISCUSSION In the first phase of the current study, we evaluated leisure items, tokens, and edibles as reinforcers with three children with autism using progressive ratio schedules. Results indicated that all influenced high rates of behavior, with slightly higher break points occurring during the token reinforcement condition for two participants. In the second phase, we evaluated the extent to which tokens functioned as generalized
conditioned reinforcers with two of the participants. The results showed that presession access to food detrimentally affected progressive ratio break points only when edibles were used as reinforcers, but not when tokens (exchangeable for leisure items and edibles) were used as reinforcers. Further, when presession access to food was implemented, the participants chose to exchange their tokens for leisure items rather than edibles. These results suggest that tokens functioned as effective generalized conditioned reinforcers in the current study. The results of Phase 1 suggested that tokens maintained responding for a longer period and at leaner schedules of reinforcement (for two participants) but at a somewhat slower pace relative to edible reinforcement. However, because the token, edible, and leisure reinforcement conditions were not replicated within participants, conclusions regarding the differential effectiveness of tokens cannot be made. Further research should incorporate a sound experimental design to examine if tokens are indeed differentially effective. We found that presession access to edibles abolished the value of the edible reinforcers. The tokens, on the other hand, were unaffected by the presession access and continued to maintain high break points. It is likely that the combination of multiple categories of backup reinforcers, both leisure and edible items, resulted in the tokens’ resilience to singular abolishing operations. Additionally, the shift in the exchange for backup reinforcers from edibles to leisure items during the token condition supports this interpretation. The current results differ somewhat from those found by Vargo and Ringdahl (2015; Experiment 4). In their study with four typically developing preschoolers, these authors found that presession access to edible items had a more deleterious effect on responding maintained by token reinforcement than responding maintained by edible reinforcement. However, the tokens in the Vargo and Ringdahl study
TOKENS AS GENERALIZED CONDITIONED REINFORCERS were exchangeable only for the same edible items that were delivered in the presession period. Therefore, it seems unlikely that the tokens functioned as generalized conditioned reinforcers. Instead, the delivery of the tokens may have signaled the availability of delayed edibles. Further, the participants received only a set amount of edibles presession, which may not have been sufficient to produce satiation. Thus, the presession period in the Vargo and Ringdahl study may have served to reduce the value of delayed edible reinforcement more than the value of immediate reinforcement, resulting in greater reduction in response rates in the token condition. The results of the current study, however, suggest that token reinforcement is likely to be resistant to disruption by presession access to one type of reinforcement, if the tokens are exchangeable for a variety of backup reinforcers. Because access to a greater variety of leisure items was instituted at the same time as presession access to edibles, we cannot be certain whether the generalized function of the tokens was established by the enrichment of the leisure items, substitution of multiple dissimilar reinforcers, or both. Nevertheless, the fact remains that breakpoints occurring in the edible condition were disrupted as a result of presession access to edibles, whereas breakpoints occurring in the token condition were unaffected, and this effect was demonstrated in a reversal design. The current findings add to those of Moher et al. (2008), in that the properties of token reinforcers may extend across individual stimuli creating categories that are affected by a more encompassing MO. Moher et al. included only edible items and manipulated the value of each item, whereas the current study included a three-item array of the most highly preferred edibles and leisure items. However, the current study is limited in that the value of only one category of backup reinforcers was manipulated. Future research could focus on determining the boundaries of generalized conditioned
11
reinforcers by including more categories of reinforcers and measuring the effects on tokens when abolishing operations are manipulated for multiple categories. In addition, we have seen only deleterious effects of satiation, but the inverse may be possible as well. In other words, presession consumption of one category of reinforcers may increase the break points of tokens exchangeable for different, independent backup reinforcers. However, this remains an empirical question. Although the results suggest that tokens functioned as generalized conditioned reinforcers, there is a slight possibility that the tokens themselves were reinforcers. Due to practical constraints, there was no baseline token condition included. In such a condition, the tokens would be earned, but no backup reinforcers would be available for exchange. If high break points were maintained in such contexts, we could say that the value of the tokens might be somewhat independent of their relation to the backup reinforcers. During the 5-min presession access period, Damien continued to achieve high break points without exchanging the tokens for edibles or leisure items. This might indicate that the tokens functioned as reinforcers, independent of the backup reinforcers. However, this is difficult to determine because both participants had an extensive history with token economies and backup reinforcers. Along similar lines, a noreinforcement baseline condition was not conducted, leaving the possibility of the behavior being maintained in the absence of programmed reinforcement. Although inclusion of a noreinforcement baseline would reduce speculation about the tasks being automatically reinforcing, we can affirm the consequent. During presession access periods, responding was effectively eliminated even though the contingency was still in place. In other words, if completing the tasks was maintained by automatic reinforcement, task engagement should have continued irrespective of the MO manipulations.
12
DANIELLE RUSSELL et al.
Additional inferences can be drawn from previous studies that examined the use of PR schedules in reinforcer assessments. Roane et al. (2001) found that stimuli with higher break points were overall more effective in decreasing problem behavior and increasing appropriate behavior. These findings, combined with the outcomes of the current study, suggest that tokens might be the most effective reinforcers to increase appropriate behavior under certain conditions. In addition, Penrod, Wallace, and Dyer (2008) found that more preferred stimuli resulted in higher break points. All the items used in the current study were ranked as the top three preferred items according to the initial preference assessments. Thus, break points on PR schedules may prove to be a more sensitive measure of reinforcer efficacy than assessments including a singular dense schedule of reinforcement. This is not to say that PR schedules should replace all other reinforcer assessments. Which assessment to conduct depends on time restrictions, the participant’s response when exposed to thin reinforcement schedules, and relevance to the task. For example, if brevity is required or if the participant is unlikely to contact thin reinforcement schedules in their everyday environment, a simple comparison using dense reinforcement schedules may be sufficient. However, if the participant is likely to contact a much broader array of reinforcement schedules, an assessment using a PR schedule may be more appropriate. It is also important to note that the rapidly thinning reinforcement schedules characteristic of PR arrangements can result in potentially aversive conditions. In the current study, we ensured that the participants could emit a low-effort escape response at any time. The fact that Damien and Carmen sometimes opted to end the sessions very quickly (sometimes without emitting any target responses) suggests that high schedule values became aversive or less preferred. Due to the fact that PR arrangements are relatively time-
consuming and may lead to lean ratios that are aversive, they should not be used unless there is a clear benefit of doing so and participants are free to opt out of the arrangement at any time (cf. Poling, 2010). Future research could help guide the appropriate application of different reinforcer assessments. REFERENCES Ayllon, T., & Azrin, N. H. (1968). The token economy: A motivational system for therapy and rehabilitation. New York, NY: Appleton-Century-Crofts. Cooper, J. O., Heron, T. E., & Heward, W. L. (2007). Applied behavior analysis (2nd ed.). Upper Saddle River, NJ: Pearson. Cowles, J. T. (1937). Food-tokens as incentives for learning by chimpanzees. Comparative Psychological Monographs, 12, 1-96. https://doi.org/10.1037/ 14268-000 DeFulio, A., Yankelevitz, R., Bullock, C., & Hackenberg, T. D. (2014). Generalized conditioned reinforcement with pigeons in a token economy. Journal of the Experimental Analysis of Behavior, 102, 26-46. https://doi.org/10.1002/jeab.94 DeLeon, I. G., Iwata, B. A., Goh, H., & Worsdell, A. S. (1997). Emergence of reinforcer preference as a function of schedule requirements and stimulus similarity. Journal of Applied Behavior Analysis, 30, 439-449. https://doi.org/10.1901/jaba.1997.30-439 Fisher, W. W., & Mazur, J. E. (1997). Basic and applied research on choice responding. Journal of Applied Behavior Analysis, 30, 387-410. https://doi.org/10. 1901/jaba.1997.30-387 Fisher, W., Piazza, C. C., Bowman, L. G., Hagopian, L. P., Owens, J. C., & Slevin, I. (1992). A comparison of two approaches for identifying reinforcers for persons with severe and profound disabilities. Journal of Applied Behavior Analysis, 25, 491-498. https://doi.org/10.1901/jaba.1992.25-491 Hackenberg, T. D. (2009). Token reinforcement: A review and analysis. Journal of the Experimental Analysis of Behavior, 91, 257-286. https://doi.org/10.1901/ jeab.2009.91-257 Hodos, W. (1961). Progressive ratio as a measure of reward strength. Science, 134, 943-944. https://doi. org/10.1126/science.134.3483.943 Kazdin, A. E., & Bootzin, R. R. (1972). The token economy: An evaluative review. Journal of Applied Behavior Analysis, 5, 343-372. https://doi.org/10.1901/ jaba.1972.5-343 Kelleher, R. T. (1956). Intermittent conditioned reinforcement in chimpanzees. Science, 124, 679-680. https://doi.org/10.1126/science.124.3224.679
TOKENS AS GENERALIZED CONDITIONED REINFORCERS Kelleher, R. T. (1958). Fixed-ratio schedules of conditioned reinforcement with chimpanzees. Journal of the Experimental Analysis of Behavior, 1, 281-289. https://doi.org/10.1901/jeab.1958.1-281 Laraway, S., Snycerski, S., Michael, J., & Poling, A. (2003). Motivating operations and terms to describe them: Some further refinements. Journal of Applied Behavior Analysis, 36, 407-414. https://doi.org/10. 1901/jaba.2003.36-407 McAdam, D. B., Klatt, K. P., Koffarnus, M., Dicesare, A., Solberg, K., Welch, C., & Murphy, S. (2005). The effects of establishing operations on preferences for tangible items. Journal of Applied Behavior Analysis, 38, 107-110. https://doi.org/10.1901/jaba. 2005.112-03 Moher, C. A., Gould, D. D., Hegg, E., & Mahoney, A. M. (2008). Non-generalized and generalized conditioned reinforcers: Establishment and validation. Behavioral Interventions, 23, 13-38. https:// doi.org/10.1002/bin.253 Murphy, E. S., McSweeney, F. K., Smith, R. G., & McComas, J. J. (2003). Dynamic changes in reinforcer effectiveness: Theoretical, methodological, and practical implications for applied research. Journal of Applied Behavior Analysis, 36, 421-438. https://doi. org/10.1901/jaba.2003.36-421 Pace, G. M., Ivancic, M. T., Edwards, G. L., Iwata, B. A., & Page, T. J. (1985). Assessment of stimulus preference and reinforcer value with profoundly retarded individuals. Journal of Applied Behavior Analysis, 18, 249-255. https://doi.org/10. 1901/jaba.1985.18-249 Penrod, B., Wallace, M. D., & Dyer, E. J. (2008). Assessing reinforcer potency of high preference and lowpreference reinforcers with respect to response rate and response patterns. Journal of Applied Behavior Analysis, 41, 177-188. https://doi.org/10.1901/jaba. 2008.41-177 Piazza, C. C., Fisher, W. W., Hagopian, L. P., Bowman, L. G., & Toole, L. (1996). Using a choice assessment to predict reinforcer effectiveness. Journal
13
of Applied Behavior Analysis, 29, 1-9. https://doi.org/ 10.1901/jaba.1996.29-1 Poling, A. (2010). Progressive-ratio schedules and applied behavior analysis. Journal of Applied Behavior Analysis, 43, 347-349. https://doi.org/10.1901/jaba.2010. 43-347 Roane, H. S., Lerman, D. C., & Vorndran, C. M. (2001). Assessing reinforcers under progressive schedule requirements. Journal of Applied Behavior Analysis, 34, 145-167. https://doi.org/10.1901/jaba.2001. 34-145 Roane, H. S., Vollmer, T. R., Ringdahl, J. E., & Marcus, B. A. (1998). Evaluation of a brief stimulus preference assessment. Journal of Applied Behavior Analysis, 31, 605-620. https://doi.org/10.1901/jaba. 1998.31-605 Tustin, R. D. (1994). Preference for reinforcers under varying schedule arrangements: A behavioral economic analysis. Journal of Applied Behavior Analysis, 27, 597-606. https://doi.org/10.1901/jaba.1994. 27-597 Vargo, K. K., and Ringdahl, J. E. (2015), An evaluation of resistance to change with unconditioned and conditioned reinforcers. Journal of Applied Behavior Analysis, 48, 643-662. https://doi.org/10.1002/ jaba.226 Windsor, J., Piche, L. M., & Locke, P. A. (1994). Preference testing: A comparison of two presentation methods. Research in Developmental Disabilities, 15, 439-455. https://doi.org/10.1016/0891-4222(94) 90028-0 Wolfe, J. B. (1936). Effectiveness of token rewards for chimpanzees. Comparative Psychological Monographs, 12, 1-72. https://doi.org/10.1037/h0093425 Woodcock, R. W., McGrew, K. S., & Mather, N. (2001). Woodcock-Johnson III. Itasca, IL: Riverside. Received April 9, 2016 Final acceptance October 9, 2017 Action Editor, Anthony DeFulio
