A human-operant investigation of preceding- and following-schedule behavioral contrast.
Research suggests that the magnitude of contrast depends on whether the change in reinforcement occurs in the preceding versus in the following component (Williams, 1979, 1988; Williams & Wixted, 1986). For example, Williams (1981) used a three-component (ABC) mult schedule in which Components A and C were target components and reinforcement was delivered according to the same variable-interval (VI) 3-min schedule across conditions. The schedule of reinforcement in Component B varied (VI 1 min or VI 6 min) across conditions. Variations in Component B resulted in larger contrast in Component A than in Component C. With respect to the target components (A and C), the following schedule influenced behavior to a larger degree (B followed A) than the preceding schedule (B preceded C).
Research comparing positive and negative contrast, however, suggests that the relation between following and preceding schedules and contrast holds for positive contrast only. For example, Weatherly, Melville, Swindell, and McMurry (1998) found that negative contrast was larger in the target component preceded by the varied component (the opposite of the relation with positive contrast), and Williams (1992b) found similar magnitudes of negative contrast regardless of whether the target component was preceded or followed by the varied component. Generally, researchers have found that positive contrast is easier to produce than negative, and that the magnitude of positive contrast is larger than negative (McSweeney, Dougan, Higa, & Farmer, 1986; Pliskoff, 1963; Wilton & Gay, 1969).
In addition to molar changes in behavior, contrast has been analyzed on a molecular (i.e., within component) level. For example, Williams (1988) used a three-component mult schedule in which Components A and C were targets and reinforcement was delivered according to the same VI 90-s schedule across conditions. The schedule of reinforcement in Component B varied (VI 30 s or VI270 s) across conditions. To examine within-component effects, Williams divided Components A and C into bins and calculated the extent to which behavior changed across conditions within each bin. Contrast in Component A increased as time progressed into the component (i.e., across bins), while contrast in Component C decreased as time progressed. Several terminological conventions have been used to describe within-component effects, but, for the purpose of this paper, within-component contrast in Component A (largest magnitude of contrast at the end of the component) will be referred to as anticipatory contrast (Williams, 2002). Within-component contrast in Component C (largest magnitude of contrast at the beginning of the component) will be referred to as preceding-local contrast. Researchers have consistently demonstrated preceding-local contrast but have had difficulty reliably demonstrating anticipatory contrast (Weatherly et al., 1998; Williams, 1981, 1988).
The vast majority of research on behavioral contrast (and the research discussed thus far here) has been conducted with nonhumans (usually, pigeons). Research on behavioral contrast with humans largely has been conducted within human-operant arrangements (i.e., with humans engaging in arbitrary responses; Edwards, 1979; Fagen, 1979; Hantula & Crowell, 1994; O'Brien, 1968; Rovee-Collier & Capatides, 1979; Tarbox & Hayes, 2005; Terrace, 1974; Waite & Osborne, 1972; Weatherly, Melville, & McSweeney, 1996). Fewer applied studies have explicitly examined behavioral contrast (Johnson & Kaye, 1979; Kistner, Hammer, Wolfe, Rothblum, & Drabman, 1982; Koegel, Egel, & Williams, 1980), although researchers in other applied areas (e.g., task interspersal, stimulus generalization) have reported contrast-like effects (Charlop, Kurtz, & Milstein, 1992; Wahler, Vigilante, & Strand, 2004).
No research with humans has systematically investigated preceding- and following-schedule effects. Furthermore, few studies with humans have directly compared positive and negative contrast, and those that have were human-operant arrangements (cf. Waite & Osborne, 1972; Weatherly et al, 1996). Waite and Osborne (1972) used a two-component mult schedule with typically developing children to assess positive contrast in one group of participants and negative contrast in a second group. The positive-contrast condition began with equal reinforcement rates in baseline (mult VI 20 s VI 20 s) and the manipulation consisted of extinction in the second component (mult VI20 s EXT). The negative-contrast condition began with unequal rates of reinforcement in baseline (mult VI 20 s EXT) and the manipulation consisted of an increase in reinforcement rate in the second component (mult VI 20 S VI 20 s). Positive contrast occurred with all three subjects and negative contrast occurred with two of three, which parallels the relative inconsistency of negative contrast that has been found in the nonhuman literature (McSweeney et al., 1986; Pliskoff, 1963; Wilton & Gay, 1969). Weatherly et al. (1996) also compared positive and negative contrast in a human-operant arrangement but with typically developing adults, and similar to Waite and Osborne found that positive contrast occurred more often (five of six participants) than negative contrast (four of six). Only one study (Waite & Osborne) investigated within-component effects with humans, but such effects did not occur.
Despite the potential implications of contrast with humans (e.g., treatment of disruptive behavior in one setting may worsen behavior in another setting, contrast may be more likely to occur at different times of day), the vast majority of research on contrast has been conducted with nonhumans. No studies have been conducted with humans regarding preceding- and following-schedule effects, few have directly compared positive and negative contrast, and only one has investigated within-component contrast. Therefore, the current study was a translational investigation that assessed the extent to which nonhuman findings on behavioral contrast generalized to humans. One approach to translational research is to replicate basic-research arrangements with "clinical populations" (Lerman, 2003). We considered individuals with intellectual and developmental disabilities (IDDs) a "clinical population" insofar as a large proportion of those individuals engages in problematic behavior (Kiernan & Alborz, 1996; Matson & Rivet, 2008; Rojahn, Matson, Lott, Esbensen, & Smalls, 2001), and ultimately, findings on behavioral contrast may prove useful in providing those individuals with clinical services. We therefore chose to evaluate the extent to which results obtained with nonhumans generalized to adults with IDDs. Furthermore, because so little research on behavioral contrast has been done with humans, arbitrary responses were used in lieu of socially significant behavior to more closely approximate the control attained with nonhumans (i.e., a human-operant arrangement; Mace & Critchfield, 2010). Specifically, we used a three-component mult schedule (Weatherly et al, 1998; Williams, 1981, 1988) and manipulated the schedule of reinforcement in the second component in order to evaluate positive and negative contrast, preceding-and following-schedule effects, and within-component contrast.
Subjects and Setting
Subjects were three adults with IDDs who attended a day program in northern Utah. All subjects were able to follow simple instructions and had expressive vocal-verbal repertoires, although the complexity of their verbal repertoires varied. Jimmy and Lucy spoke in complete sentences, and Molly spoke in phrases. Lucy also engaged in repetitive, unusual speech (e.g., "You're a funny bunny, you're a funny bunny rabbit queen lover") throughout the majority of each session. Molly and Lucy wore corrective lenses. Jack was ambulatory with the aid of a walker and had fine motor difficulties. These individuals were selected because they attended the day program 5 days per week and did not have therapeutic obligations during the times when researchers were available to conduct sessions. Subjects did not have any health or medical conditions that precluded the use of edibles as reinforcers and did not engage in severe problem behavior (e.g., aggression, property destruction, self-injury). We obtained informed consent from the subjects' caregivers and assent from the subjects. Sessions took place at the day program, and only the subject, therapist and data collectors were present. Subjects sat in a chair in front of a table on which experimental materials were presented.
Materials and Response Measurement
The therapist sat at a table across from the subject. A laptop computer faced the therapist and signaled component transitions. Response materials for Molly and Lucy included two bowls (17.8 cm in diameter) and a crayon, and materials for Jack included a box of pegs (35.6 cm x 35.6 cm) and a pegboard (varied in size depending on the condition). Simple arbitrary responses were selected for each subject on the basis of their motor skills.
The target response for Molly and Lucy consisted of picking up a crayon from one bowl and placing it at the bottom of a second bowl. The target response for Jack consisted of inserting plastic pegs into a pegboard. A box of pegs of assorted colors was placed in front of Jack in addition to a pegboard of the color corresponding to the current component. A response was scored when Jack inserted a peg into a board of the same color.
Colors were used to signal each of the three components of the mult schedule and included the therapist's shirt, the response materials (bowls and crayon for Molly and Lucy; pegboard for Jack) and a laminated sheet of paper (21.6 cm x 27.9 cm) on the table in front of the subject. The written number 1, 2, or 3 on a laminated sheet of paper (21.6 cm x 27.9 cm) was also placed on the table in front of the subject to indicate Component A, B, or C, respectively.
Interobserver Agreement and Treatment Fidelity
Trained observers collected frequency data on subjects' responses and reinforcers delivered by the therapist. A second observer independently collected data for 35 %, 48 %, and 33 % of sessions for Molly, Lucy, and Jack, respectively. Interobserver agreement (IOA) was calculated by dividing each session into 10-s bins and comparing the observers' data with respect to the number of occurrences of the event (subject responses and reinforcers delivered). A block-by-block-agreement score was calculated for each interval by dividing the smaller recorded number of responses by the larger and multiplying by 100 %. Intervals in which neither observer recorded a response were scored as 100 % agreement. The IOA score for each session was calculated by obtaining the mean of the interval IOA scores for that session. Mean IOA for the target response was 90 % (range 73 %-95 %), 92 % (range 85 %-98 %), and 85 % (range 80 %-95 %) for Molly, Lucy, and Jack, respectively. Mean IOA for reinforcer delivery was 95 % (range 80 %--100 %), 94 % (range 82 %-100 %), and 92 % (range 78 %-99 %) for Molly, Lucy, and Jack, respectively.
Observers also collected data on therapist behavior (changing stimuli that signaled each component and delivery of reinforcers) to assess treatment fidelity. Percentage of sessions with treatment fidelity data differed across therapist behavior (changing stimuli and reinforcer delivery) based on the availability of multiple data collectors (i.e., one data collector scored data on changing stimuli and one scored data on reinforcer delivery). The fidelity with which the therapist changed stimuli was assessed during 41 %, 37 %, and 32 % of sessions for Molly, Lucy, and Jack, respectively. During treatment integrity sessions, data collectors sat behind the therapist (or viewed digital recordings of sessions) and recorded whether the therapist changed stimuli within 5 s of the computer-generated prompt. If the therapist changed the stimuli within 5 s of the prompt, the response was scored as correct. If the therapist changed the stimuli after 5 s of the prompt, the response was scored as incorrect. A percentage-correct score was calculated by dividing the number of correct stimulus changes by the sum of correct and incorrect stimulus changes, and multiplying the quotient by 100 %. The mean percentage correct score was produced by calculating the mean of all integrity scores. Mean treatment fidelity for changing stimuli was 96 % (range 88 %-100 %), 96 % (range 78 %-100 %), and 92 % (range 78 %-l 00 %) for Molly, Lucy, and Jack, respectively.
The fidelity with which the therapist delivered reinforcers was assessed during 33 %, 33 %, and 31 % of sessions for Molly, Lucy, and Jack, respectively. Two criteria were used to assess reinforcer delivery fidelity. The first criterion specified which responses resulted in reinforcement. A response was considered to be eligible for reinforcement if it (a) occurred fewer than 3 s prior to the end the interval, (b) was one of several responses that occurred fewer than 3 s following the end of the interval, or (c) was the first response to occur following the end of the interval. The second criterion specified the timing of reinforcer delivery. A reinforcer was correctly delivered if it was delivered within 3 s of a response. A percentage-correct score was calculated by dividing the number of correct reinforcer deliveries by the sum of correct and incorrect reinforcer deliveries and multiplying the quotient by 100 %. The mean percentage-correct score was produced by calculating the mean of all integrity scores. Mean treatment fidelity for reinforcer delivery was 94 % (range 75 %-100 %), 97 % (range 88 %-100 %), and 95 % (range 91 %-100 %) for Molly, Lucy, and Jack, respectively.
Preference Assessment To identify potentially preferred edibles to use as reinforcers, interviews were conducted with caregivers and employees to produce a list of between five and 10 edibles. These edibles were then assessed for preference with a paired preference assessment (Fisher et al., 1992) during which the edibles were presented in pairs, and subjects were instructed to pick one. Each edible was paired once with every other edible. A hierarchy was produced by dividing the number of selections of each edible by the number of opportunities to select that edible and multiplying the quotient by 100 %. Prior to each session, the top three edibles were presented to subjects, and they were instructed again to pick one; the chosen edible was delivered during that session. Staff members were asked to limit the availability of the top three edibles outside of sessions.
Response Training and Schedule Thinning Prior to experimental conditions, subjects were taught to engage in the target response. The purpose of this phase was to ensure that subjects emitted the response consistently throughout the session. Sessions lasted 5 min and consisted of a preexposure to the contingency prior to the session in the form of a physical prompt from the therapist to engage in the response, which was followed by the delivery of the preferred edible. Edibles were delivered according to a fixed-ratio (FR) 1 schedule, and physical prompts to engage in the response were delivered according to a delay, which increased by 5 s across sessions (the delay in the first session was 5 s). When three sessions occurred with at least 80 % independent responding, with the first response in each session occurring independently, the schedule of reinforcement was thinned to variable-ratio (VR) 2, VR 4, VI 15 s, and, finally, VI 30-s schedules. All VI schedules in the experiment consisted of 15 intervals (selected without replacement) and were programmed according to the distribution described by Fleshier and Hoffmann (1962).
Experimental Arrangement To determine the extent to which changes in the reinforcement conditions in one context affected responding in preceding and following contexts, we used a single-operant arrangement within a three-component (ABC) mult schedule. Components A and C were the designated target components, and Component B was the varied component. The target components were both associated with the same color (red) and schedule of reinforcement regardless of the condition. In the varied component, the color and schedule of reinforcement differed depending on the contingencies in effect. The therapist delivered a brief praise statement (e.g., "Good job!") along with each edible delivery.
All components lasted 2 min. There was a 15-s period between Components A and B and between Components B and C, during which response-related materials and stimuli were removed and a black laminated sheet of paper (21.6 cm x 27.9 cm) was placed on the table in front of the subject. Each three-component presentation therefore lasted 6.5 min in addition to reinforcer access time, which differed across sessions and subjects depending on the number of reinforcers that were earned and amount of time necessary to consume the reinforcer.
Each three-component presentation was a trial, and a 45-s intertrial interval (ITI) occurred between each trial, during which all response-related materials and stimuli were removed and a white laminated sheet of paper (21.6 cm x 27.9 cm) with the word Break was placed on the table in front of the subject. Four trials constituted a session. Three trials were originally implemented per session, however Molly's data indicated that variables other than those programmed by the experiment influenced responding in the first trial. We therefore implemented a fourth trial and omitted data from the first in all analyses for all subjects, as analyzing data for three versus four trials did not change our primary conclusions. One session was conducted per day and lasted about 35 min, including time between components, ITIs, and reinforcer access time. At least eight sessions were conducted in each phase.
A reversal design (ABAC) was used with conditions occurring in the following order: baseline, extinction, baseline, FR 1.
Baseline Baseline schedules of reinforcement varied across subjects depending on response patterns during training: VI 60 s for Molly, VI 30 s for Lucy, and VI 45 s for Jack. The same color (red) was associated with all components, although each component was signaled with the number 1, 2, or 3.
Contrast Two schedule manipulations were assessed: a lower rate of reinforcement (extinction) and a higher rate of reinforcement (FR 1). Each condition was identical to baseline with the exception of the schedule of reinforcement in, and color (i.e., response-related materials and the therapist's shirt) associated with, the varied component (Component B).
The purpose of the first experimental condition, extinction, was to determine whether a decrease in the rate of reinforcement in the varied component resulted in increases in responding in the target components (i.e., positive contrast). No reinforcers were delivered for responding in Component B. Edibles were removed from the subject's view to enhance discriminability of the change in contingency from baseline to extinction.
For Molly, an additional manipulation was made during Component B. At the onset of the component, the therapist switched the stimuli to those of the color associated with Component B and removed the edibles (as with the other subjects), but the therapist and data collectors then left the room and remained out of Molly's sight for the duration of the component. Molly was thus in the experimental room alone, and a video camera was used from which data on her responding were scored following the session. Molly never attempted to leave her chair or engage with materials in the room.
We made this manipulation with Molly because, after conducting several sessions of extinction, her responding persisted at levels comparable to the preceding baseline. Responding may have persisted in the researchers' presence because of the history of reinforcement for responding (delivery of edibles and brief social interactions) in their presence in the past. By removing the researchers during Component B, responding might have extinguished due to the removal of stimuli potentially associated with such a history of reinforcement.
The purpose of the second experimental condition, FR 1, was to determine whether an increase in the rate of reinforcement in the varied component resulted in decreases in responding in the target components (i.e., negative contrast). Jack was the first subject to experience this condition. The high rate of reinforcement schedule in the varied component was initially VI10 s (compared to VI45 s in his baseline), but after 15 sessions, negative contrast had not occurred. We therefore further increased the rate of reinforcement to FR 1. To increase the likelihood of detecting an effect with Molly and Lucy in this phase, we used an FR 1 schedule with them as well.
Stability and Termination Criteria A minimum of eight sessions was conducted in all phases, including both baseline and contrast. Stability criteria for advancing to subsequent phases from baseline involved requirements for both responses and reinforcers earned. Criteria were as follows:
1. In each of the last three sessions of baseline, the rate of responding in each component must have been within 20 % of the mean of responding across all components.
2. The floating average (arithmetic mean of the most recent three sessions) of reinforcers earned in each component must have been within 20 % of the mean of the floating average of reinforcers earned across all components.
3. If 20 sessions were conducted without meeting criteria (1) and (2), the next phase began, given that the data were not occurring on a trend that was expected to continue in the next phase (e.g., an upward trend in baseline would be undesirable if the next phase entailed data most likely continuing to increase). This occurred twice, once with Molly and once with Lucy.
Stability criteria for terminating contrast phases involved response-rate requirements in Components A and C only and were assessed via visual inspection:
1. Response rates in Component A must have changed in the direction consistent with contrast, relative to the preceding baseline, for three consecutive sessions. We only required that rates occurred in the direction of contrast in Component A given that previous research suggests that the following-schedule effect (Component A) develops after the preceding-schedule effect (Component C). Thus, terminating contrast phases based on an effect consistent with contrast in Component A was more conservative than terminating phases based on such an effect in Component C. In addition, it might never have been the case that effects in both components occurred simultaneously (e.g., within the same three consecutive sessions). The criterion to terminate the phase when an effect was detected in Component A ensured terminating the phases in a reasonable number of sessions.
2. For those three sessions, response rates in Components A and C must either have not displayed a trend or have been on an increasing trend in extinction or on a decreasing trend in FR 1. This criterion was implemented to ensure that the effect consistent with contrast was not temporary. Even if response rates in Component C did not meet criterion (1), we required that they did not display a trend in the opposite direction. For example, in the extinction phase, response rates must have been elevated in Component A relative to the respective response rates in baseline for three consecutive sessions, and response rates in both components must not have displayed a decreasing trend. In the FR 1 phase, response rates in Component A must have been below the respective response rates in baseline for three consecutive sessions, and response rates in both components must not have displayed an increasing trend. Lucy and Jack met criteria (1) and (2) in the extinction phase in eight and 13 sessions, respectively, however, we conducted additional sessions given the large range of Lucy's data in Component C and the small magnitude of the effect for Jack.
3. If criteria (1) and (2) were not met within 20 sessions, the phase was terminated. This occurred twice, once with Molly and once with Jack, and each in the FR 1 phase.
Because we terminated phases when behavior changed in the direction consistent with contrast, it is possible that we created "false-positive" situations, or committed Type I errors. However, we chose these criteria based on previous research that suggests that some types of contrast, specifically that of the following schedule, develop over time (Killeen, 2014; Williams, 1981, 1988). Assuming contrast develops over multiple sessions, it is possible that ending phases after a predetermined number of sessions results in Type II errors if the phase is not carried out for a sufficient number of sessions for contrast to develop. Similarly, developing a criterion based on lack of variability (e.g., three sessions with behavior within a percentage of the mean of all three sessions) may result in termination of the phase before contrast develops. Because of the paucity of research on contrast with humans, we chose to use criteria that would maximize the likelihood of detecting an effect, although it might be the case that doing so resulted in a Type I error.
Figure 1 shows response rates for all subjects during all conditions.
Figure 2 shows results from extinction (left-hand panels) and FR 1 (right-hand panels) phases for all subjects. Proportion-of-baseline rates were obtained by calculating the mean of the last three sessions of baseline for each component. Each data point reflects the quotient of the response rate in a given component and the mean of the last three sessions of baseline for that component. The dotted horizontal line represents the baseline level of responding (i.e., proportion of 1.0).
Extinction Results for Molly during extinction are shown in the upper-left panel of Fig. 2. Rates in Component B, in which responding was on extinction, were at or near zero during the entire phase. Responding in Component A occurred at or above the baseline level (with the exception of Session 7), while responding in Component C varied above and below baseline.
Lucy's responding during extinction (middle-left panel of Fig. 2) generally covaried across all components, with responding in Component A elevated relative to Component C (with the exception of Sessions 3, 7, 8, and 10), which were both elevated relative to Component B. Responding in Component B, in which responding was on extinction, was consistently at or below baseline. Rates in Components A and C were elevated relative to baseline in all sessions. Rates in Component C were more variable than in A and were often just above baseline (Sessions 1, 6, and 9).
Jack's results during extinction are shown in the bottom-left panel of Fig. 2. Following Session 4, no responding occurred during Component B, in which responding was on extinction. Responding in Components A and C began to increase about midway through the phase (Session 10 and 8, respectively), and by the end of the phase, proportion of baseline response rates in Components A and C were comparable.
FR 1 The upper-right panel of Fig. 2 shows results during FR 1 for Molly. The proportion of baseline rates in Component B was stable throughout the phase, occurring at about 0.5. Decreases in responding in a variable component associated with an increased rate of reinforcement have been reported elsewhere in contrast research (McSweeney & Melville, 1991). McSweeney and Melville argued that negative contrast in target components can still be assessed despite decreases in responding in the varied component, provided that more reinforcers are earned in the varied component in the negative contrast condition relative to the preceding baseline (see Table 1). Response rates in Component A were variable but were generally higher than baseline (with the exception of Sessions 1 and 12). Rates in Component C varied above and below baseline throughout the phase, although the proportion of baseline rates in each of the last three sessions was less than 1.0.
[FIGURE 1 OMITTED]
The middle-right panel of Fig. 2 shows results during FR 1 for Lucy. Responding in Component B, in which responding was reinforced on an FR 1 schedule, was consistently above baseline. Proportion of baseline rates in Components A and C covaried, with rates in Component A below baseline in all sessions and those in Component C at or near baseline for the first four sessions but then decreasing to a level comparable to Component A.
The lower-right panel of Fig. 2 shows results during FR 1 for Jack. Rates in Component B were variable but most often occurred above baseline. Responding in Components A and C was inconsistent across the phase as well, varying between below (e.g., Sessions 9-11) and above baseline (Sessions 5, 18), but most often occurred at the baseline level (Sessions 2-4, 6, 7, 12, 14, 17, 20).
Effect Sizes As it is possible that different types of contrast differ in their development over time (e.g., Williams, 1981), data from the first three sessions of contrast conditions were analyzed separately from data from the last three sessions. Because we analyzed the last three data points of each phase only, we may not have had adequate statistical power to detect effects with an inferential statistic (e.g., t test). We therefore used Cohen's d as our primary metric of contrast (see Sullivan & Feinn, 2012 for discussion of effect size and statistical significance). Effect sizes were calculated by obtaining the mean difference score between response rates in the last three sessions of the contrast condition and in the last three sessions of its preceding baseline. The mean difference score was then divided by the pooled standard deviation of the last three sessions of the contrast condition and its preceding baseline. For the purpose of this paper, effect sizes with absolute values of greater than 2.0 reflect an effect consistent with contrast. The designation of an effect size of a given magnitude as reflecting a meaningful (e.g., clinically significant) change in behavior is somewhat arbitrary, but others have described effect sizes of 1.0 as "large" (Cohen, 1988). Increasing the criterion to 2.0 was our attempt to be conservative in concluding an effect was consistent with contrast.
Figure 3 shows effect sizes during both contrast conditions for all subjects and for both target components. Left-hand panels show effect sizes from the first three sessions of each condition (extinction on the top; FR 1 on the bottom), and right-hand panels show effect sizes from the last three sessions of each condition. Effect sizes greater than zero reflect an increase in behavior relative to the preceding baseline, and those less than zero reflect a decrease.
[FIGURE 2 OMITTED]
In the first three sessions of extinction (top left-hand panel), Lucy was the only subject to demonstrate an effect consistent with positive contrast, and in Component A only (the following-schedule effect). By the end of extinction (upper right-hand panel), all subjects demonstrated effects consistent with positive contrast, with the exception of Molly in Component C.
In the first three sessions of FR 1 (bottom left-hand panel), none of the subjects demonstrated an effect consistent with negative contrast. By the end of FR 1, (bottom right-hand panel), Lucy was the only subject to demonstrate an effect consistent with negative contrast, and she did so in both components.
It is important to note that across-session variability in response rates had different effects on proportion of baseline rates (see Fig. 2) and Cohen's d (see Fig. 3). For example, Lucy's proportion of baseline rates in Components A and C are comparable in extinction in Fig. 2; however, the effect size for Component A is larger than Component C in Fig. 3. This is because response rates in Component C were more variable than those in Component A, which reduced the effect size in Component C relative to Component A. It was also possible to have relatively small changes in behavior relative to baseline but to have relatively large effect sizes. For example, Jack's proportion of baseline rates in Components A and C are comparable in extinction in Fig. 2 and are just above 1.0; however, the effect sizes for both components are large (5.0 and 4.0, respectively). This is because Jack's response rates were relatively stable (see Fig. 1), which resulted in a small standard deviation (i.e., the denominator in the effect-size calculation) and, ultimately, a large effect size. Each session's proportion-of-baseline-rate calculation was not influenced by across-session variability (e.g., standard deviation).
[FIGURE 3 OMITTED]
The frequency of responding per 12-s bin was calculated for each of the first three and last three sessions of extinction and FR 1 phases. The mean frequency of responding in each bin was then calculated for the first three sessions and separately for last three sessions. The same calculations were made for the last three sessions of baseline phases, against which to compare results from contrast phases.
To determine whether anticipatory or preceding-local contrast occurred, data were first fit to regression lines. If the slope of the line was consistent with that of anticipatory effects (positive in extinction and negative in FR 1) or preceding-local effects (negative in extinction and positive in FR 1), statistical significance was assessed at p < .05 to determine whether the slope was significantly different from zero. Within-session contrast was not observed with any subject in any condition, therefore data are not reported here.
Table 1 shows obtained reinforcers per hr for all components for Molly, Lucy, and Jack, respectively.
It is worth noting that Fucy earned a higher rate of reinforcement in Components A and C in extinction than in baseline, and she continued to earn a higher rate in both components in the replication of baseline. This may explain why Fucy's responding did not return to the initial baseline rates, as the increased response rates in extinction in the form of contrast contacted a higher rate of reinforcement (see Fig. 1). Similarly, Jack earned a higher rate of reinforcement in Component C in extinction than during baseline, which may also explain the failure of responding in Component C to return to baseline levels in the replication of baseline (see Fig. 1). However, Jack earned a lower rate of reinforcement in Component A during extinction compared to baseline; the failure of responding to return to initial baseline levels in this component cannot be explained by contacting an increased rate of reinforcement.
All three subjects' behavior demonstrated an effect consistent with positive behavioral contrast (i.e., Cohen's d greater than or equal to 2.0) in at least one component. Conversely, only one subject's behavior demonstrated an effect consistent with negative behavioral contrast (i.e., Cohen's d less than or equal to -2.0; Lucy). In the positive contrast condition, the following-schedule effect was larger than the preceding-schedule effect for Lucy and Jack, and Molly only demonstrated the following-schedule effect. In the negative contrast condition, the preceding-schedule effect was larger than the following-schedule effect for Lucy. Finally, neither anticipatory nor local contrast was demonstrated with any subject.
The larger and more consistent results with respect to the following-schedule effect with positive contrast replicate findings from studies with non-humans (e.g., Williams, 1979, 1988; Williams & Wixted, 1986). Further, the finding that the preceding-schedule effect was larger than the following-schedule effect with negative contrast (the opposite of the relation with positive contrast) is consistent with what others have found regarding the interaction between order of components and direction of contrast (Weatherly et al., 1998; Williams, 1992b). The results with respect to negative contrast should be considered tenuous, however, as only one subject demonstrated negative contrast at all.
The finding that negative contrast occurred with only one of three subjects is consistent with findings from studies with non-humans regarding the relative difficulty of producing negative contrast (McSweeney et al, 1986; Pliskoff, 1963; Wilton & Gay, 1969). Conceptually, a possible reason for the difficulty of demonstrating negative contrast may be related to the notion that aversive events influence behavior to a larger extent than appetitive events. For example, Rasmussen and Newland (2008) reported asymmetry in the magnitude of behavior change produced with reinforcement versus punishment, in that the magnitude of behavior change was larger when one unit of a stimulus was removed (punishment in the form of response cost) than when one unit of the same stimulus was presented (reinforcement). A decrease in the rate of reinforcement may be conceptualized as an aversive stimulus change in the same way that an increase in the rate of reinforcement may be conceptualized as an appetitive stimulus change. Assuming that the former influences behavior to a larger degree than the latter, it could be that even indirect effects of that change (i.e., behavioral contrast) are larger as well. Accordingly, positive contrast resulting from an aversive stimulus change (in terms of the reduction in reinforcement rate from baseline to extinction conditions) may be expected to be larger than negative contrast resulting from an appetitive stimulus change (the increase in rate of reinforcement).
Unlike behavioral contrast, behavior in a target component may change in the same direction as the change in reinforcement rate in a varied component--an effect known as induction (e.g., McSweeney & Melville, 1991; Reynolds, 1963; Shanab & Kong, 1977; Weatherly et al., 1998; Williams, 1981). For example, in a parametric evaluation of the influence of component duration on behavioral contrast, McSweeney and Melville (1991) found that a decreased rate of reinforcement in the varied component produced positive contrast in target components of longer durations (30 s, 60 s, 3 min, and 16 min) but that the same decrease in rate of reinforcement produced negative induction at the shortest component duration (5 s).
Across both contrast conditions in the current experiment, Lucy's behavior never demonstrated induction (see Fig. 3). In other words, her behavior in the target components always changed in the opposite direction of the change in reinforcement rate in the varied component. Molly and Jack, however, demonstrated behavior change consistent with induction in at least one component (see Fig. 3), although the magnitudes of the effects were small, not meeting our 2.0 or -2.0 criteria for concluding behavioral positive or negative behavioral contrast, respectively. Molly's and Jack's behavior both changed in a direction consistent with positive induction in the FR 1 condition in Component A, and Jack's behavior changed similarly in Component C, although the effects were small. Jack's behavior also changed in a direction consistent with negative induction in extinction in Component C in the first three sessions.
One reason for which Molly's and Jack's behavior changed in a manner consistent with positive induction in Component A may relate to the competition between contrast effects and Pavlovian contingencies that are inherent to the mult-schedule arrangement (Williams, 1992a; Williams & McDevitt, 2001). Williams (1992a) pointed out that following-schedule contrast (contrast in Component A) is an effect that is inconsistent with what would be expected if Pavlovian contingencies governed behavior. Stimulus contexts should, in accordance with Pavlovian processes, acquire value in the same direction as the change in reinforcement rate in the context that follows. In other words, the value of Component A should increase when the rate of reinforcement in Component B increases. If stimulus contexts are similar (i.e., discrimination between contexts is difficult), behavior should be more susceptible to Pavlovian conditioning in that target components acquire the value of the varied components that follow, which would be reflected by induction (a change in behavior in the same direction as the change in reinforcement rate in the following context). This is the opposite of the effect that occurs when stimulus contexts are more discriminable in that the determinant of behavior is not the Pavlovian relation but the relative difference in reinforcement rate, which would be reflected by contrast (a change in behavior in the opposite direction as the change in reinforcement rate in the following context).
Indeed, researchers have found that the degree of following-schedule contrast depends in part on the degree of similarity between discriminative stimuli associated with each component, with larger contrast in arrangements in which components are highly discriminable (e.g., Williams, 1988). It is possible that discriminations between components were easier for some subjects (Lucy) than for others (Jack). This may have resulted in the Pavlovian contingency (Component A followed by Component B) overriding the difference in relative rate of reinforcement across conditions. We utilized multiple discriminative cues (colors of materials and shirts to signal contingencies, numbers that signaled each component) for each component to attempt to enhance discriminability between components; however, it could be that different colors were not enough to differentiate the contingencies for at least some subjects in some conditions (e.g., Jack in Component A in FR 1). Future investigations may use different areas of a room or different therapists for each component, or investigate further the notion that varying degrees of stimulus (dis)similarity may be more or less likely to produce behavioral contrast.
Similar to the results obtained by McSweeney and Melville (1991) regarding induction, it is possible that negative contrast did not occur with Jack or Molly due to component duration, reinforcement rates during baseline, or the interaction between component duration and baseline reinforcement rates. For example, Weatherly et al. (1998) found negative contrast with baseline reinforcement schedules of VI 30 s when components lasted 60 s but not when components lasted 30 s. Parametric evaluations (e.g., Ettinger & Staddon, 1982; McSweeney, 1982) have yet to be conducted with humans in a three-component arrangement and may be useful in determining whether there are between-subject differences regarding the relation between component duration and behavioral contrast.
Notwithstanding that effect sizes and proportion-of-baseline rates demonstrated results consistent with behavioral contrast, response patterns between and within subjects during contrast conditions were highly variable. For example, Jack's and Molly's behavior in Component B in extinction (see Figs. 1 and 2) completely extinguished while Lucy's maintained at 50 % of her previous baseline level. Lucy experienced the most baseline sessions of the three (20 for Lucy vs. 8 for Jack and 11 for Molly), although Jack earned reinforcers in Component B at a higher rate than Lucy (see Table 1); thus, an appeal cannot be made to mechanisms responsible for behavioral momentum (e.g., Nevin, Mandell, & Atak, 1983). Furthermore, Molly's behavior in Component B in FR 1 decreased to about 50 % of responding in the preceding baseline whereas Lucy's and Jack's increased--Lucy's to a greater and more consistent extent than Jack's (see Figs. 1 and 2). Across-subject differences in the direction of behavior change with increases in reinforcement rate (i.e., some subjects' behavior increases while others' decreases) have been reported in other studies on contrast (e.g., McSweeney, Swindell, Murphy, & Kowal, 2004). And, given that the determinant of contrast is change in obtained reinforcers and not response rate (e.g., Halliday & Boakes, 1971, 1974), an evaluation of negative contrast with Molly was still possible, given she did earn a higher rate of reinforcement.
There was also considerable variability in responding within subjects. We applied the 20-session-maximum criterion during the initial baseline for Lucy and the replication of baseline for Molly, as their responding had not met the stability criteria. This was a problem mainly for Molly in Component A in FR 1, in that the amount of variability in the second baseline may have obscured a change in responding consistent with negative contrast. The level of responding in FR 1 in Components A would have had to be extremely low to produce a convincing effect, given the large range of responding in the preceding baseline.
Relatedly, responding did not return to initial baseline levels for Jack and Lucy following extinction. As a result of the increase in responding in Components A and C during extinction, Lucy's responding contacted a higher rate of reinforcement than in the initial baseline (see Table 1). It is possible that the maintenance and increased rate of responding in the replication of baseline was due to having contacted an increased rate of reinforcement in the previous phase. Although this is a plausible explanation, it is not clear that the implementation of extinction in Component B caused the increased response rates in Components A and C, as responding in Components A and C did not return to then-initial levels when the reinforcement contingency for Component B was reinstated. Jack's responding also did not return to baseline levels, but only responding in Component C (not Component A) contacted an increased rate of reinforcement when responding in Component B was placed on extinction, thus the appeal to the direct reinforcement of higher rates of responding is not applicable to Component A with Jack. In addition, the increased response rates developed slowly with Jack. It is possible that the increase in responding in Components A and C was due to a practice effect and not to extinction in Component B; responding may have increased in those components without that manipulation. This seems unlikely, though, as the increasing trends in Components A and C ceased with the reintroduction of baseline.
We chose adults with IDDs as our subjects in an attempt to evaluate the extent to which findings on behavioral contrast with nonhuman subjects generalize to a clinical population. However, research with clinical populations, and perhaps humans in general, may raise unique challenges with attempts to replicate nonhuman results. For example, it is possible that Molly and Jack did not demonstrate negative contrast as a result of a history of punishment for noncompliance (i.e., engaging in behavior other than compliance). For example, staff might have reprimanded clients in the past for engaging in behavior other than work tasks. Clients would then be more likely to work when staff are present, in order to avoid reprimands. The researchers might have functioned similarly to staff and evoked responding as a result of a history of punishment for noncompliance. Another issue with Molly and Jack may have been that responding was a result of an adventitious contingency of postponing the return to the activities of the day program. Jack, for example, often requested we work with him first or volunteered to work with us instead of attending a planned activity through the day program. It is possible that his continued high rates of responding during FR 1 were partially under the control of a motivating operation to avoid returning to the day-program activities. Finally, we sometimes conducted sessions with some subjects at slightly different times during the day due to logistical reasons, and it is possible that doing so changed the value of the programmed reinforcers such that responding was influenced accordingly. Similarly, a different room was sometimes used in which to conduct sessions, also due to logistical reasons. It is possible that subjects had different histories of reinforcement in each room, or that some other feature of the rooms exerted control over behavior (e.g., one room was slightly darker than the other, one room was upstairs). It would have been ideal to conduct sessions at exactly the same time and in the same room throughout the study, but it was not feasible given the nature of the setting.
Such challenges in establishing stimulus control to the programmed contingencies of the experiment do not necessarily preclude humans as subjects or reflect across-species differences in the laws of behavior. Rather, researchers should be mindful of subjects' sensitivities to other stimuli (e.g., events or activities that occur following experimental sessions), and perhaps enhance or make more salient the variables known to influence nonhumans' behavior (e.g., discriminability between components). It may also be useful in future investigations to utilize larger numbers of subjects to identify subject variables that may correlate with contrast (e.g., performance on tasks that require visual or temporal discriminations, nonverbal and verbal measures of intelligence, "emotional"--e.g., Binkley, Webber, Powers, & Cromwell, 2014--or verbal behavior during the experiment). An additional and related direction may be to use arrangements that more closely resemble applied or clinical situations. For example, using socially significant behavior in the context of reinforcement-based interventions to increase or decrease behavior (e.g., differential reinforcement of alternative behavior in the treatment of undesired behavior), programming more naturally discriminative stimuli that signal components (e.g., different caregivers or locations instead of color of materials), and determining more naturalistic ICIs and ITIs between components (e.g., transitions from locations instead of the removal of materials) may be future directions in assessing the generalizability of the current and other findings on behavioral contrast.
This study was the first to examine the influence of preceding and following schedules in humans, and the first study to replicate findings with nonhumans regarding the greater influence of the following schedule. It was also the first study with humans to demonstrate the larger effect of the preceding schedule with humans. Results from this preliminary investigation suggest that methods with nonhumans may be used with humans to examine behavioral contrast. We were not, however, able to replicate results with respect to within-component contrast (preceding-local and anticipatory effects). These results in some sense parallel the inconsistencies across the research on behavioral contrast in general. For example, although Williams (1981, 1983, 1988) has consistently found that the following schedule of reinforcement is more influential than the preceding, others have found that the effects are comparable (Weatherly et al., 1998). Similarly, within-component effects have been inconsistent across studies (Waite & Osborne, 1972; Weatherly et al., 1998). Researchers investigating contrast have sometimes found behavior change in the direction opposite that of contrast (i.e., induction; McSweeney et al., 1986; McSweeney & Melville, 1991; Terrace, 1968; Williams, 1981). These inconsistencies highlight the need for more research in understanding behavioral contrast.
Acknowledgments The authors would like to thank Cynthia Pietras for her comments on an earlier version of this manuscript. The authors would also like to thank Kerry Jordan and Gregory Madden for their helpful suggestions during the course of this project. Finally, the authors wish to thank Casey Clay, Chase Callard, and Rickie Ivory for assistance in data collection.
Compliance with Ethical Standards
Conflict of Interest All authors declare no conflicts of interest with this submission.
Research with Human Subjects All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
Informed Consent Informed consent was obtained from all individual participants included in the study.
Binkley, K. A., Webber, E. S., Powers, D. D., & Cromwell, H. C. (2014). Emotion and relative reward processing: an investigation on instrumental successive negative contrast and ultrasonic vocalizations in the rat. Behavioural Processes, 107, 167-174. doi: 10.1016/j.beproc.2014.07.011.
Charlop, M. H., Kurtz, P. F., & Milstein, J. P. (1992). Too much reinforcement, too little behavior: assessing task interspersal procedures in conjunction with different reinforcement schedules with autistic children. Journal of Applied Behavior Analysis, 25(4), 795-808. doi: 10.1901/jaba. 1992.25-795.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.
Edwards, R. P. (1979). Behavioral contrast in humans with response-independent reinforcement. The Journal of General Psychology, 700(1), 159-160. doi: 10.1080/00221309.1979.9710535.
Ettinger, R. H., & Staddon, J. E. R. (1982). Behavioral competition, component duration, and multiple-schedule contrast. Behavioural Analysis Letters, 2, 31-38.
Fagen, J. W. (1979). Behavioral contrast in infants. Infant Behavior & Development, 2(2), 101-108. doi: 10.1016/S0163-6383(79)80015-6.
Ferster, C. B., & Skinner, B. F. (1957). Schedules of reinforcement. East Norwalk, CT: Appleton-Century-Crofts.
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(2), 491-498. doi: 10.1901/jaba. 1992.25-491.
Fleshler, M., & Hoffmann, H. S. (1962). A progression for generating variable-interval schedules. Journal of the Experimental Analysis of Behavior, 5(4), 529-530. doi:10.1901/jeab.1962.5-529.
Halliday, M. S., & Boakes, R. A. (1971). Behavioral contrast and response independent reinforcement Journal of the Experimental Analysis of Behavior, 76(3), 429-434. doi:10.1901/jeab.1971.16429.
Halliday, M. S., & Boakes, R. A. (1974). Behavioral contrast without response-rate reduction. Journal of the Experimental Analysis of Behavior, 22(3), 453-^462. doi: 10.1901/jeab. 1974.22-453.
Hantula, D. A., & Crowell, C. R. (1994). Behavioral contrast in a two-option analogue task of financial decision making. Journal of Applied Behavior Analysis, 27(4), 607-617. doi:10.1901/jaba. 1994.27-607.
Johnson, C. M., & Kaye, J. H. (1979). Behavioral contrast during the acquisition of speechreading. Perceptual and Motor Skills, 49, 171-180. doi: 10.2466/pms. 19184.108.40.206.
Kiernan, C., & Alborz, A. (1996). Persistence and change in challenging and problem behaviours of young adults with intellectual disability. Journal of Applied Research in Intellectual Disabilities, 9(3), 181-193. doi: 10.1111/j. 1468-3148.1996.tb00108.x.
Killeen, P. R. (2014). A theory of behavioral contrast. Journal of the Experimental Analysis of Behavior, 702(3), 363-390. doi: 10.1002/jeab. 107.
Kistner, J., Hammer, D., Wolfe, D., Rothblum, E., & Drabman, R. S. (1982). Teacher popularity and contrast effects in a classroom token economy. Journal of Applied Behavior Analysis, 75(1), 85-96. doi: 10.1901/jaba. 1982.15-85.
Koegel, R. L., Egel, A. L., & Williams, J. A. (1980). Behavioral contrast and generalization across settings in the treatment of autistic children. Journal of Experimental Child Psychology, 30(3), 422-437. doi:10.1016/0022-0965(80)90048-X.
Lerman, D. C. (2003). From the laboratory to community application: translational research in behavior analysis. Journal of Applied Behavior Analysis, 36(4), 415-419. doi: 10.1901/jaba.2003.36-415.
Mace, F. C., & Critchfield, T. S. (2010). Translational research in behavior analysis: historical traditions and imperatives for the future. Journal of the Experimental Analysis of Behavior, 93(3), 293-312. doi: 10.1901/jeab.2010.93-293.
Matson, J. L., & Rivet, T. T. (2008). Characteristics of challenging behaviours in adults with autistic disorder, PDD-NOS, and intellectual disability. Journal of Intellectual a Developmental Disability, 33(4), 323-329. doi: 10.1080/1366825080249260.
McSweeney, F. K. (1982). Positive and negative contrast as a function of component duration for key pecking and treadle pressing. Journal of the Experimental Analysis of Behavior, 37(2), 281-293. doi: 10.1901/jeab. 1982.37-281.
McSweeney, F. K., Dougan, J. D., Higa, J., & Farmer, V. A. (1986). Behavioral contrast as a function of component duration and baseline rate of reinforcement. Animal Learning & Behavior, 14(2), 173-193.
McSweeney, F. K., & Melville, C. L. (1991). Behavioral contrast as a function of component duration for leverpressing using a within-session procedure. Animal Learning & Behavior, 79(1), 71-80. doi: 10.3758/BF03197862.
McSweeney, F. K., Swindell, S., Murphy, E. S., & Kowal, B. P. (2004). The relation of multiple-schedule behavioral contrast to deprivation, time in session, and within-session changes in responding. Learning & Behavior, 32(2), 190-201. doi: 10.3758/BF03196020.
Nevin, J. A., Mandell, C., & Atak, J. R. (1983). The analysis of behavioral momentum. Journal of the Experimental Analysis of Behavior, 39(1), 49-59. doi: 10.1901/jeab. 1983.39-49.
O'Brien, F. (1968). Sequential contrast effects with human subjects. Journal of the Experimental Analysis of Behavior, 77(5), 537-542. doi: 10.1901/jeab. 1968.11-537.
Pliskoff, S. (1963). Rate-change effects with equal potential reinforcements during the "warning" stimulus. Journal of the Experimental Analysis of Behavior, 6(4), 557-562. doi: 10.1901/jeab. 1963.6-557.
Rachlin, H. (1973). Contrast and matching. Psychological Review, 80(3), 217-234. doi: 10.1037/h0034325.
Rasmussen, E. B., & Newland, M. C. (2008). Asymmetry of reinforcement and punishment in human choice. Journal of Experimental Analysis of Behavior, 89(2), 157-167. doi: 10.1901/jeab.2008.89-157.
Reynolds, G. S. (1961). Behavioral contrast. Journal of the Experimental Analysis of Behavior, 4(1), 57-71. doi: 10.1901/jeab. 1961.4-57.
Reynolds, G. S. (1963). Some limitations on behavioral contrast and induction during successive discrimination. Journal of the Experimental Analysis of Behavior, 6(1), 131-139. doi:10.1901/jeab. 1963.6-131.
Rojahn, J., Matson, J. L., Lott, D., Esbensen, A. J., & Smalls, Y. (2001). The Behavior Problems Inventory: an instrument for the assessment of self-injury, stereotyped behavior, and aggression/destruction in individuals with developmental disabilities. Journal of Autism and Developmental Disorders, 31(6), 577-588.
Rovee-Collier, C. K., & Capatides, J. B. (1979). Positive behavioral contrast in 3-month-old infants on multiple conjugate reinforcement schedules. Journal of the Experimental Analysis of Behavior, 32(1), 15-27. doi: 10.1901/jeab. 1979.32-15.
Shanab, M. E., & Kong, J. (1977). Failure to obtain positive behavioral contrast in the rat following extended acquisition. The Journal of General Psychology, 96(1), 95-102. doi: 10.1080/00221309.1977.9920802.
Sullivan, G. M., & Feinn, R. (2012). Using effect size--or why the p value is not enough. Journal of Graduate Medical Education, 4(3), 279-282. doi: 10.4300/JGME-D-12-00156.1.
Tarbox, J., & Hayes, L. P. (2005). Verbal behavior and behavioral contrast in human subjects. The Psychological Record, 55, 419-437. Retrieved from http://opensiuc.lib.siu.edU/tpr/vol55/iss3/5.
Terrace, H. S. (1968). Discrimination learning, the peak shift, and behavioral contrast. Journal of the Experimental Analysis of Behavior, 77(6), 727-741. doi: 10.1901/jeab. 1968.11-727.
Terrace, H. S. (1974). On the nature of non-responding in discrimination learning with and without errors. Journal of the Experimental Analysis of Behavior, 22(1), 151-159. doi: 10.1901/jeab. 1974.22-151.
Wahler, R. G., Vigilante, V. A., & Strand, P. S. (2004). Generalization in a child's oppositional behavior across home and school settings. Journal of Applied Behavior Analysis, 37(1), 43-51. doi: 10.1901/jaba.2004.37-43.
Waite, W., & Osborne, J. G. (1972). Sustained behavioral contrast in children. Journal of the Experimental Analysis of Behavior, 76(1), 113-117. doi: 10.1901/jeab. 1972.18-113.
Weatherly, J. N., Melville, C. L., Swindell, S., & McMurry, A. S. (1998). Previous- and following-component contrast effects using a three-component multiple schedule. Behavioural Processes, 42, 47-59. doi: 10.1016/S0376-6357(97)00063-6.
Weatherly, J. N., Melville, C. L., & McSweeney, F. K. (1996). Picking, pecking, and pressing: A cross-species demonstration of behavioral contrast. The Psychological Record, 46, 351-372. Retrieved from http://thepsychologicalrecord.siu.edu.
Williams, B. A. (1979). Contrast, component duration, and the following schedule of reinforcement. Journal of Experimental Psychology, 3(4), 379-396. doi:10.1037/0097-7403.5.4.379.
Williams, B. A. (1981). The following schedule of reinforcement as a fundamental determinant of steady state contrast in multiple schedules. Journal of the Experimental Analysis of Behavior, 33(3), 293-310. doi: 10.1901/jeab. 1981.35-293.
Williams, B. A. (1983). Another look at contrast in multiple schedules. Journal of the Experimental Analysis of Behavior, 39(2), 345-384. doi: 10.1901/jeab. 1983.39-345.
Williams, B. A. (1988). The effects of stimulus similarity on different types of behavioral contrast. Animal Learning & Behavior, 76(2), 206-216. doi: 10.3758/BF03209067.
Williams, B. A. (1992a). Competition between stimulus-reinforcer contingencies and anticipatory contrast. Journal of the Experimental Analysis of Behavior, 58(2), 287-302. doi: 10.1901/jeab. 1992.58-287.
Williams, B. A. (1992b). Inverse relations between preference and contrast, Journal of the Experimental Analysis of Behavior, 58(2), 303-312. doi:10.1901/jeab. 1992.58-303.
Williams, B. A. (2002). Behavioral contrast redux. Animal Learning & Behavior, 50(1), 1-20. doi: 10.3758/BF03192905.
Williams, B. A., & McDevitt, M. A. (2001). Competing sources of stimulus value in anticipatory contrast. Animal Learning & Behavior, 29(4), 302-310. doi: 10.3758/BF03192897.
Williams, B. A., & Wixted, J. T. (1986). An equation for behavioral contrast. Journal of the Experimental Analysis of Behavior, 45(1), 47-62. doi: 10.1901/jeab. 1986.45-47.
Wilton, R. N., & Gay, R. A. (1969). Behavioral contrast in one component of a multiple schedule as a function of the reinforcement conditions operating in the following component. Journal of the Experimental Analysis of Behavior, 12(2), 239-246. doi:10.1901/jeab.1969.12-239.
Megan A. Boyle  * Andrew L. Samaha  * Timothy A. Slocum  * Audrey N. Hoffmann  * Sarah E. Bloom 
Published online: 31 March 2016
[mail] Megan A. Boyle
Andrew L. Samaha
Timothy A. Slocum
Audrey N. Hoffmann
Sarah E. Bloom
 Department of Counseling, Leadership, and Special Education, Missouri State University, 901 S. National Avenue, Springfield, MO 65897, USA
 Department of Child and Family Studies, University of South Florida, 4202 E. Fowler Avenue, Tampa, FL 33620, USA
 Department of Special Education and Rehabilitation, Utah State University, 2800 Old Main Hill, Logan, UT 84322-2800, USA
Table 1 Obtained reinforcers per hour in each component in all conditions Condition Obtained Reinforcers per Hour A B C Molly Baseline 57.27 53.64 55.45 Extinction 54.17 0 58.33 Baseline 53.75 50.42 59.58 FR 1 51.00 719.50 55.00 Lucy Baseline 66.50 74.00 69.50 Extinction 95.00 0 93.00 Baseline 133.64 113.64 103.64 FR 1 96.25 1,311.25 103.75 Jack Baseline 83.75 77.50 66.25 Extinction 71.33 0 78.00 Baseline 76.25 81.25 80.00 FR 1 75.50 678.00 65.00
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||ORIGINAL ARTICLE|
|Author:||Boyle, Megan A.; Samaha, Andrew L.; Slocum, Timothy A.; Hoffmann, Audrey N.; Bloom, Sarah E.|
|Publication:||The Psychological Record|
|Date:||Sep 1, 2016|
|Previous Article:||Body mass index and caloric value of rewards as parameters that modulate delay-discounting rates.|
|Next Article:||Exploring the behavioral dynamics of the implicit relational assessment procedure: the role of relational contextual cues versus relational coherence...|