Jennifer J. McComas and Ellie C. Hartman. Angel Jimenez

Transcription

1 The Psychological Record, 28, 58, Some Effects of Magnitude of Reinforcement on Persistence of Responding Jennifer J. McComas and Ellie C. Hartman The University of Minnesota Angel Jimenez The University of Guadalajara The influence of magnitude of reinforcement was examined on both response rate and behavioral persistence. During Phase, a multiple schedule of concurrent reinforcement was implemented in which reinforcement for one response option was held constant at VI 3 s across both components, while magnitude of reinforcement for the other response option was varied across the 2 components. In Phase 2, persistence was evaluated during extinction following steady state responding. In Phase, responding of 3 of the 4 participants was greater on the option that produced the higher magnitude of reinforcement. During Phase 2, 2 of these 3 participants response persistence under extinction was greater for the signal associated with the greater magnitude of reinforcement. The influence of magnitude of reinforcement on behavioral persistence is discussed, as well as implications for studying behavioral maintenance. The concept of behavioral momentum appears to apply directly to the study of behavioral persistence because it captures not only responding during acquisition but also persistence of responding (e.g., during extinction). Although reinforcement affects both response rate and persistence, or resistance to changes in reinforcement schedules, their determiners can be separated. Response rate depends primarily on the response reinforcement contingency, whereas resistance to change depends primarily on the stimulus reinforcer contingency signaled by a discriminative stimulus (Nevin, 984). Results of research in the area of behavioral momentum have shown that behavior is most persistent in the presence of stimuli that signal relatively richer rates of reinforcement (e.g., Nevin, Tota, Torquato, & Shull, 99). The results of basic experimental preparations have demonstrated that resistance to change depends on the stimulus reinforcer contingency in effect during training (Nevin, 984; Nevin et al., 99; Mace et. al., 99). Specifically, persistent responding during extinction depends primarily on the rate of Correspondence may be addressed to Jennifer J. McComas, Ph.D., Associate Professor of Educational Psychology, University of Minnesota, 347 Education Sciences Building, 56 East River Road, Minneapolis, Minnesota jmccomas@umn.edu

2 58 MCCOMAS ET AL. reinforcement correlated with a particular environmental stimulus (Nevin, Mandel, & Atak, 983). In an experiment designed to evaluate the effect of stimulus reinforcer contingencies, Nevin and colleagues (99) arranged three multiple concurrent reinforcement schedules. Each schedule was signaled by a distinct stimulus. In Component A, a green light signaled 45 f/hr (45 foods per hour) for key pecks on the left key and 5 f/hr for key pecks on the right. In Component B, a red light signaled no reinforcement for key pecks on the left and 5 f/hr for key pecks on the right. In Component C, a white light signaled no reinforcement on the left and 6 f/hr on the right. Following stable baseline responding, extinction was imposed on all three components. Results indicated that the baseline rate of responding was highest in Component C, where the rate of reinforcement for right key pecks was highest. Furthermore, despite equal rates of reinforcement for right-key pecks, pigeons responded less on the right key during Component A, in which left key responses produced reinforcement, than during Component B, in which no alternative reinforcement was provided. In contrast, when Components A and B were compared during the extinction test, persistence was greater in Component A, which signaled a denser overall schedule of reinforcement, whereas persistence during Component A was similar to persistence during Component C, since both components signaled an equal overall schedule of reinforcement. These findings and those of similar basic studies have demonstrated that responding is persistent in schedule components correlated with the highest overall amount of reinforcement (Bell, 999; Nevin, Mandel, & Atak, 983; Nevin et al., 99). Further study of the influence of stimulus reinforcer contingencies on behavioral persistence is needed. Specifically, the influence of different dimensions of reinforcement on behavioral maintenance other than rate, such as magnitude of reinforcement, should be explored further. In an earlier study, Nevin (974) observed that magnitude had similar effects as rate on 2 pigeons response persistence. Rate of reinforcement was held constant at a Variable Interval (VI min or VI 3 min) schedule, whereas duration of food availability varied from 7.5 s to 2.5 s. When free food was available, pigeon response rates were less resistant to change when reinforcement length was relatively less (i.e., 2.5 s), although pigeon was equally persistent during VI min regardless of the duration of reinforcement. The current study was designed to further examine whether persistence during extinction is more likely in the presence of a stimulus that signals a relatively greater magnitude of reinforcement. Method Subjects A total of 4 graduate students majoring in educational psychology or related fields participated. The project was advertised in the building that houses the Educational Psychology Department, and participants were selected on a first come, first served basis. To participate in the study, participants were required to (a) have earned a high school diploma or General Equivalency Diploma (GED); (b) be at least 8 years old; (c) have vision that was either unimpaired or corrected by contact lenses, glasses, or surgery; and (d) not be colorblind. Subjects were 4 female university students who were between the ages of 8 and 22 and who reported they met the vision criteria, including not being color-blind. Each was paid for her participation.

3 REINFORCER MAGNITUDE AND PERSISTENCE 59 Setting and Apparatus Experimental sessions were conducted in a room containing two computers and other research equipment. During each session the participant was alone in the room and seated at a computer, where she responded to a Visual Basic program (Dixon & MacLin, 23) by moving and clicking the mouse. At the beginning of each session, the screen was gray and contained two colored squares, one on the right and one on the left. During the Y component, both of these squares were yellow; during the G component, both of these squares were green. Clicking a square brought up a new screen with the background the color associated with the component (e.g., during the Y component the screen color was yellow). A black arrow (changeover requirement) was positioned in one of the upper corners of the screen. If the participant originally clicked on the left square, the arrow was in the upper, right corner pointing to the right. To respond on the right, the participant had to first click on the right arrow. Once the screen changed to the right side, the arrow was in the upper, left corner pointing to the left. Again, to respond on the left, the participant had to first click on the left arrow. In the other upper corner (whichever corner had no arrow), the word Points was written in black text followed by a white box. When the session began, the white box contained the number ; as the participant earned points, this box provided the participant with a cumulative point total. To receive points, participants had to click on the gray square that appeared on the screen below the arrow and point total. The gray box always started in the middle of the screen but randomly moved around the screen throughout the course of the session. General Procedures Experimental sessions began with general instructions for the computer and the operanda printed on a sheet of paper. The general instructions were as follows: The program is designed to pay you points for clicking the mouse on the moving square. At the end of the session, you will trade in your points for money: The more points you earn, the more money you will earn. The number of points you can earn will vary systematically depending on the condition (e.g., yellow or green), the square you choose (left or right), and a number of other factors. The number of points you have earned will appear in a small window on the screen. To begin, put the mouse cursor either on the left or right square on the screen and press the left mouse button. After that, a new screen will appear where you can click on the moving square to earn points. If you want to change screens, click on the arrow. Again, the number of points you can earn will vary systematically. In all experiments, reinforcement consisted of points exchangeable for money; participant earned $. per point, and the other 3 earned $.5 for every point. Participants were not told how much the other participants earned during the experiments. Further, the participants were not told who the other participants were, nor did they communicate with other participants about the study. After completing two 2-min sessions, participants were paid half

4 52 MCCOMAS ET AL. of the money they earned from the points they obtained, plus $. for each block of two sessions. The participants were paid the second half of the money they earned from the points after they finished all experimental sessions, including extinction sessions. Based on the arranged schedules of reinforcement and participant performance, participants earned on average $2. to $3. per experimental session (two sessions were conducted per day). Each participant required 4 to 3 experimental sessions to complete the study. Thus, the participants earned a total of between $3.75 and $33. for up to 6 hr of experimental time. The experiment was conducted in two phases: () Concurrent Schedule of Reinforcement and (2) Persistence Test. During Phase, two multiple concurrent reinforcement schedules were arranged. Each component lasted 2 min and alternated randomly throughout the session until each had been completed three times. All experimental sessions lasted approximately 2 min. In the yellow component, clicks on the left produced reinforcement on a 3-s Variable Interval (VI 3-s) schedule with a magnitude of 8 points. Clicks on the right produced reinforcement on a VI 3-s schedule, with a magnitude of point. In the green component, left clicks produced reinforcement on a VI 3-s schedule with a magnitude of 2 points. Right clicks produced reinforcement on a VI 3-s schedule with a magnitude of point. Persistence under extinction was tested in Phase 2 after responding during Phase met the criteria for stability. Responding was considered stable when it met both of the following criteria for three consecutive experimental sessions: (a) response rates varied less than % of the previous three consecutive experimental sessions within each component, and (b) consistent ordinal relationships were observed for both response options across the two components. Persistence was evaluated in terms of the proportion of responding under conditions of reinforcement. Based on the findings of previous momentum research (e.g., Nevin et al., 99), one would expect right responding to be greater in the colored component that was associated with greater magnitude of reinforcement (yellow). Further, based on the relative effects of reinforcement on response allocation, one would expect the response associated with the greater magnitude of reinforcement (left) within a component would persist longer relative to the response associated with the smaller magnitude of reinforcement (right). Data Recording, Experimental Design, and Analysis All responses were recorded by the computer software program written in Visual Basic (Dixon & MacLin, 23). The number of responses on each alternative was recorded separately. To calculate the proportion of baseline responding, the number of extinction responses on one alternative was divided by the mean frequency of clicks on that response alternative in the last 9 components in Phase. Single-subject multi-element experimental analyses were conducted via the alternation between components. Consistent with the convention for analyzing momentum data, component-by-component extinction responding was visually analyzed to evaluate the results for each participant.

5 REINFORCER MAGNITUDE AND PERSISTENCE 52 Phase : Concurrent Schedule of Reinforcement Two multiple concurrent reinforcement schedules were arranged, and the two components (yellow and green) alternated three times within each experimental session so that each experimental session consisted of 3 green components and 3 yellow components. Each component was signaled by a different color (yellow or green) background on the computer screen and involved two concurrent response options: clicking on the left box and clicking on the right box. In Component Y (signaled by a yellow screen), clicking on the left box produced eight times the magnitude of reinforcement (points) produced by clicking on the right box. In Component G (signaled by a green screen), clicking on the left box produced twice the magnitude of reinforcement (points) produced by clicking on the right box. This proportion of reinforcement within and across components was selected because it has successfully demonstrated effects when used in similar experimental arrangements with pigeons (see Grace, Bedell, & Nevin, 22). Phase 2: Persistence Test Persistence was assessed during an extinction test. Procedures were identical to those used in the preceding concurrent reinforcement phase, with one exception: No points were delivered for any responding. Here, 2 components, each signaled by a different color (yellow or green) background on the computer screen, alternated during the extinction test until responding during one of the components dropped below 75% of the average response rate for the last three sessions (9 Y components and 9 G components) in the preceding reinforcement phase, or until the response rate met the criterion for stability. Component Y (signaled by a yellow screen) signaled the relatively greater magnitude of reinforcement. Results Of the initial total number of 4 participants, participant was dropped from the study when her response pattern was undifferentiated across the conditions, suggesting that she failed to discriminate between conditions in the concurrent schedules of reinforcement, and her responding did not meet stability criteria. Figure displays results for the remaining 3 participants. Phase : Concurrent Schedule of Reinforcement Kelly completed 24 sessions (72 Y components and 72 G components), Amber completed 22 sessions (66 Y components and 66 G components), and Mary completed 24 sessions (72 Y components and 72 G components) in Phase. Figure shows only responses during the last Y and G reinforcement components. Kelly and Mary s responses during each component ranged from 5 to 4 clicks; Amber s responses during each component ranged from to 4 clicks.

6 522 MCCOMAS ET AL. Frequency of Right Responses Phase Green (Left = 2 Points Right = Point) Kelly Yellow (Left = 8 Points Right = Point) Log Proportion of Baseline on Right. Phase 2 Kelly Frequency of Right Responses Amber Log Proportion of Baseline on Right Amber Frequency of Right Responses Mary Components Log Proportion of Baseline on Right. Mary Components Figure. (For caption, please turn page.)

7 REINFORCER MAGNITUDE AND PERSISTENCE 523 Frequency of Yellow Responses Phase Left (8 Points) Right ( Point) Kelly Log Proportion of Baseline on Yellow. Phase Kelly Frequency of Yellow Responses Amber Log Proportion of Baseline on Yellow Amber Frequency of Yellow Responses Mary Components Log Proportion of Baseline on Yellow. Mary Components Figure. (For caption, please turn page.)

8 524 MCCOMAS ET AL. Phase Phase 2 Frequency of Right Responses Frequency of Right Responses Yellow Trendline Figure. The across-condition comparisons appear in Columns and 2. The frequency of right responding graphs display participants frequency of responding during both yellow and green components during the last ten yellow and green components of Phase (baseline). The log proportion of baseline on right graphs display the logged proportion to baseline responding during both yellow and green components during Phase 2 (extinction). The within-component comparisons appear in Columns 3 and 4. The frequency of yellow responses graphs display participants frequency of responding on the left and right during the last ten yellow components of Phase (baseline). The log proportion of baseline on yellow graphs display the log proportion of baseline responding during the yellow component during Phase 2 (extinction). Right responses during reinforcement are depicted in the first column of Figure. As would be expected, based on the magnitude of alternative reinforcement available in each component, both Kelly and Amber s right responses occurred more frequently during the green component (left clicks produce 2 points) than during the yellow component (left clicks produced 8 points), whereas Mary s right responses occurred approximately equally in both the yellow and green components. Left and right responses within the yellow component are depicted in the third column of Figure. All 3 participants responded more on the left response ( click = 8 points) than on the right ( click = point) within the yellow component (overall S R+ magnitude = 9). Data on left/ right responding during the green component are available upon request from the first author. Phase 2: Persistence Test The participants responses during extinction are graphed as a log proportion of each participant s responses during the last three sessions (9 components) in Phase. Kelly completed two sessions of extinction (6 Y components and 6 G components), and Amber and Mary each completed six sessions (8 Y components and 8 G components) of extinction. The log proportion of right responses across both components is depicted in the second column of Figure. Across components, both Kelly and Amber s right responses were more persistent in the yellow component (S D = S R+ magnitude 9) than during the green component. Mary s right responding was equally persistent in the yellow and

9 REINFORCER MAGNITUDE AND PERSISTENCE 525 green components. The fourth column in Figure depicts the log proportion of right and left responding in the yellow component. Within the yellow component, both Kelly and Amber s right responding was more persistent than left responding. Mary s responding was approximately equally persistent on the right and left. Discussion These results of the across-component analyses are consistent with those of previous research in the area of momentum. Specifically, these results demonstrate that for 2 of 3 subjects, despite equal rate and magnitude of reinforcement for right responding across components, right responding was more persistent in the component that signaled the relatively greater magnitude of reinforcement (i.e., the yellow component). Somewhat sur prisingly, the within-component analyses indicated that the response that produced the smaller magnitude of reinforcement (right responses) was more persistent than the one that produced the greater magnitude of reinforcement (left responses). Although the degree to which the participants responding matched the programmed or obtained reinforcement may explain the differences across participants, a matching analysis was not conducted in this experiment because participants were exposed to only two components (only one value of magnitude was altered), and at least four different variations in components are necessary to provide a good estimate of matching regression lines. A further concern is that different magnitudes of reinforcement alter matching differently (see Dallery, Soto, & McDowell, 25). Future research should examine the conditions under which matching predicts persistence. These findings potentially represent a bridge between basic and applied research (Wacker, 996, 2). The experimental control that we can achieve in basic experiments allows us to better understand the environmental influences on behavior. Basic laboratory findings can inform our practice in applied settings. In applied settings where the goal is to address problem behavior, a common approach is to consider the problem and appropriate behavior as concurrent operants. Attempts are made to arrange the schedule of reinforcement such that responding is biased toward the appropriate rather than the problem behavior alternative. However, persistence is also an issue that must be considered. Therefore, it is important to know how well a response will persist when it contacts a change in the environment, such as a reduction in the schedule of reinforcement. Namely, it is important to strengthen the occurrence and persistence of appropriate behavior while simultaneously weakening the persistence of problem behavior. The experimenters in the present study hypothesize that Kelly and Mary s relatively greater persistence of right responding during the yellow component, as compared with right responding during the green component, was due to the stimulus reinforcer relationship observed in previous research (e.g., Nevin et al., 99). Generally, responding is expected to be more persistent in the presence of a stimulus associated with a relatively greater amount of reinforcement compared with responding in the presence of a stimulus associated with a relatively smaller amount of reinforcement. The yellow component was associated with a greater amount of reinforcement (overall magnitude of 9), whereas the green component was associated with a smaller amount of reinforcement (overall magnitude of 3). For this reason,

10 526 MCCOMAS ET AL. right responding (analogous to problem behavior) was more persistent during the yellow condition than during the green condition. Future applied research should examine the conditions under which the persistence of problem behavior is weakened and appropriate behavior is simultaneously strengthened. Even though the response persistence across yellow and green components in this study can be explained by the stimulus reinforcer relationship and thus has some applied implications, the participants response persistence within the yellow component can not be explained by the stimulus reinforcer relationship. Because the same amount of reinforcement was associated with both left and right responses within a component, left and right responding should be equally persistent when compared within a component. In past research, Nevin et al. (99) did find that 2 out of 3 pigeons were more persistent on one key than another within a component; but in the Nevin et al. (99) study, these pigeons were more persistent on the key associated with more reinforcement (food). This result allowed Nevin et al. (99) to argue that the key s position could have been a stimulus that the pigeons associated with more reinforcement; therefore, the stimulus reinforcer relationship could explain this greater persistence. In the current study, within the yellow component, 2 out of the 3 participants were more persistent with right responding than left responding during extinction. Yet, unlike the Nevin et al. (99) study, right responding was associated with fewer points than left responding. Neither the stimulus reinforcer relationship nor the response reinforcer relationship can explain these results. Future basic research should explore this result. With better understanding of this result, clinicians might arrange contingencies in ways that promote desired response persistence. Therefore, future research is needed to explain this finding and to identify effective ways to strengthen the persistence of appropriate behavior while weakening the persistence of concurrent problem behavior. References BELL, M. C. (999). Pavlovian contingencies and resistance to change in a multiple schedule. Journal of the Experimental Analysis of Behavior, 72, DALLERY, J., SOTO, P. L., & MCDOWELL, J. J. (25). A test of the formal and modern theories of matching. Journal of the Experimental Analysis of Behavior, 84, DIXON, M. R., & MACLIN, O. H. (23). Visual Basic for Behavioral Psychologists. Reno, NV: Context Press. GRACE, R. C., BEDELL, M. A., & NEVIN, J. A. (22). Preference and resistance to change with constant- and variable-duration terminal links: Independence of reinforcement rate and magnitude. Journal of the Experimental Analysis of Behavior, 77, HORNE, P.J. & LOWE, C.F. (993). Determinants of human performance on concurrent schedules. Journal of the Experimental Analysis of Behavior, 59, MACE, F. C., LALLI, J. S., SHEA, M. C., PINTER LALLI, E., WEST, B. J., ROBERTS, M., & NEVIN, J. A. (99). The momentum of human behavior in a natural setting. Journal of the Experimental Analysis of Behavior, 54,

11 REINFORCER MAGNITUDE AND PERSISTENCE 527 NEVIN, J. A. (974). Response strength in multiple schedules. Journal of the Experimental Analysis of Behavior, 2, NEVIN, J. A. (984). Pavlovian determiners of behavioral momentum. Animal Learning and Behavior, 2, NEVIN, J. A. (992). An integrative model for the study of behavioral momentum. Journal of the Experimental Analysis of Behavior, 57, NEVIN, J. A., MANDELL, C., & ATAK, J. R. (983). The analysis of behavioral momentum. Journal of the Experimental Analysis of Behavior, 39, NEVIN, J. A., TOTA, M. E., TORQUATO, R. D., & SHULL, R. L. (99). Alternative reinforcement increases resistance to change: Pavlovian or operant contingencies? Journal of the Experimental Analysis of Behavior, 53, WACKER, D. P. (996). Behavior analysis research in JABA: A need for studies that bridge basic and applied research. Experimental Analysis of Human Behavior Bulletin, 4, 4. WACKER, D. P. (2). Building a bridge between research in experimental and applied behavior analysis. In J. C. Leslie & D. Blackman (Eds). Experimental and Applied Analysis of Human Behavior (pp ). Reno, NV, US: Context Press.