Rich-Lean Transitions Produced by Stimuli Associated with Accurate and Inaccurate Responding by Pigeons

Abstract

Discriminable transitions from relatively favorable schedules of reinforcement to unfavorable schedules (rich-lean transitions) can produce disruptions in operant behavior. A prior evaluation in our laboratory (Toegel et al., 2021) found that placing a border around a key displayed on a resistive touchscreen increased pigeons’ response accuracy relative to conditions without the border. We sought to evaluate (a) whether effects of the key border on accuracy could be replicated in a within-session comparison and (b) whether transitions from a response alternative associated with accurate responding to one associated with inaccurate responding functioned as rich-lean transitions. Pigeons’ key pecks were reinforced according to a two-component multiple schedule with identical fixed-ratio (FR) requirements and reinforcer magnitudes. The components differed based on whether the virtual key was displayed behind a border or with no border. In line with prior research, within-session comparisons yielded higher response accuracy in components with the key border than in components without the border. Furthermore, transitions from the border component to the no border component functioned as rich-lean transitions for pigeons whose obtained FRs in no border components were substantially larger than the FR programmed in that component.

1. Introduction

Disruptions in operant behavior that can occur during schedules of positive reinforcement have received attention from basic researchers studying reinforcement processes (e.g., Perone & Courtney, 1992) and applied researchers and clinicians (e.g., Jessel et al., 2016; Williams, 2015), whose primary interest is in treating problematic behavior observed in clinical settings. Initial evaluations of the basic processes responsible for the behavioral disruptions associated with reinforcement schedules began with laboratory experiments that studied the control of pausing under fixed-ratio (FR) schedules (e.g., Lowe et al., 1974; Powell, 1969). Pausing was the measure of disruption. Although pausing is also an outcome on interval-based schedules of reinforcement (e.g., FI schedules, Pitts et al. 2019), pausing on FR schedules has long held the conceptual interest of researchers as it may appear to be counterintuitive or maladaptive. Response rates on FR schedules are directly related to reinforcement rates. When an individual pauses on an FR schedule, it extends the time to the receipt of the next reinforcer at the local level and reduces the rate of reinforcement at the global level.

Perone and Courtney (1992) designed a procedure to assess the potential roles that prevailing and previous conditions of reinforcement play in the disruptions observed during FR schedules. They arranged a multiple schedule with two fixed-ratio (FR) components. In one component, the schedule of reinforcement was relatively favorable or rich. When the pigeons completed the FR, a relatively large amount of food was delivered. In the other component, the schedule of reinforcement is relatively unfavorable or lean. When the pigeons completed the FR, a relatively small amount of food was delivered. These two components alternated in a semi-random sequence that produced an equal number of four types of transitions: lean-lean, lean-rich, rich-lean, and rich-rich. Their findings showed that pausing was jointly controlled by the reinforcement magnitudes in the past and upcoming components of the multiple schedule (rich or lean). Specifically, large disruptions in behavior occurred only in the rich-lean transition, and not in the other three kinds of transitions (lean-lean, lean-rich, and rich-rich). The disruptions occurred reliably when the pigeons completed a rich schedule of reinforcement, claimed the reinforcer, and were exposed to stimuli signaling the start of a lean schedule. Thus, the behavioral disruptions, in the form of pausing, were a function of discriminable degradations in the local contingencies of reinforcement.

Perone and Courtney’s general findings that discriminable degradations in reinforcement schedules produce disruptions in behavior has been replicated using a variety of species, including rats (Baron et al., 1992; Galuska & Sawyer, 2017; Sawyer et al., 2019; Wade-Galuska et al., 2005), pigeons (Everly et al., 2014; Langford et al., 2019; Langford et al., 2021; Retzlaff et al., 2017; Toegel et al., 2021; Toegel & Perone, 2022), hens (Young et al., 2017), monkeys (Galuska et al., 2007), human adults (Williams et al., 2011), and human children (Jessel et al., 2016) and using different methods of arranging rich and lean schedules, including by manipulating amounts of food (e.g., Toegel et al., 2022), amounts of money (e.g., Williams et al. 2011), concentrations of drugs (Galuska et al., 2007), kinds of leisure activities (Jessel et al., 2016), and the amount of force required to engage in the target response (Wade-Galuska et al., 2005).

A recent study from our laboratory evaluated the suitability of resistive touchscreen monitors as a tool for operant research with pigeon subjects (Toegel et al., 2021). In one part of the study (Test B, Conditions 6–8), effects of a rubber key border on the accuracy of pigeons’ pecking were evaluated by manipulating the presence of the key border across conditions. Pigeons’ pecking accuracy increased substantially when the key was recessed behind the key border relative to conditions without the key border and decreased back to pre-border levels when the border was removed during a replication of a previous condition. To gain a clearer understanding of the effects of the key border on pecking accuracy, we wanted to compare accuracy with and without the key border within sessions, rather than across sessions in different conditions. This could be accomplished by arranging a multiple schedule in which the response key displayed on the touchscreen monitor appeared either behind a key border or at another location without a key border.

This arrangement permitted an evaluation of the effects that transitions between keys with and without a key border had on operant behavior. If pigeons are less accurate on one response option than the other, it is possible that this arrangement functionally creates transitions between rich (more difficult) and lean (less difficult) components in the multiple schedule. Previous research has evaluated transition-related disruptions by manipulating the difficulty of emitting responses. Wade-Galuska et al. (2005) arranged varying degrees of response difficulty by altering response force requirements. In this study, rats’ lever presses were reinforced with food pellets according to a multiple FR-FR schedule with equivalent ratio sizes and reinforcer magnitudes, but with different amounts of force required to press each of two levers. In the rich component, the lever that was inserted required 0.25 Newtons of force to register a response. In the lean component, the lever that was inserted required 0.25–0.85 Newtons. This arrangement produced reliable behavioral disruptions, in the form of extended pausing, in the rich-lean transition and short pauses in the other three types of transitions, but only when the force requirement in the lean component was raised suitably (0.40–0.70 Newtons, depending on the rat) to create a large enough discrepancy between the force requirements in the rich and lean components.

Results from our previous study (Toegel et al., 2021) showed that pigeons’ response accuracy on the touchscreen monitor was reliably affected by manipulating the presence of the key border. It is plausible that using a multiple schedule to arrange transitions between response options with and without a key border will produce behavioral disruptions similar to those observed in rich-lean transitions as arranged in Wade-Galuska et al. (2005).

The present experiment served two purposes. First, we evaluated whether differences in accuracy observed across conditions in Toegel et al.’s (2021) study would be observed in a within-session comparison arranged with a multiple schedule. Second, if reliable differences in response accuracy were generated within sessions, it would allow an evaluation of whether the transitions between the response alternatives associated with accurate or inaccurate responding produced behavioral disruptions consistent with the literature on rich-lean transitions.

2. Method

2.1. Subjects

Four individually housed White Carneau pigeons served as subjects. All pigeons had previous experience with FR schedules in a previous evaluation using the same apparatus (see Toegel et al., 2021 for details). The pigeons were maintained at 80% (± 2%) of their free-feeding body weights by food deliveries during the sessions and, if necessary, supplemental feedings at least 30 min after the end of the session. Water was freely available in the pigeons’ home cages, which were kept in a temperature-controlled room with a 12:12 hr light/dark cycle. The treatment of the pigeons, in and out of the experimental sessions, complied with a protocol approved by the West Virginia University Animal Care and Use Committee.

2.2. Apparatus

Sessions were conducted in four sound-attenuating chambers equipped with resistive touchscreens (see Toegel et al., 2021 for specific information regarding the design and construction of the touchscreen chambers). During sessions, a ventilation fan circulated air and white noise (80 dB) was played through an 8-ohm speaker to mask extraneous sound. The interior of the chamber area was 34.3 cm long, 30 cm wide, and 37 cm high. General illumination was provided by a 28-v houselight (No.1820) located on the bottom left corner of the experimental panel. Food reinforcers (Purina Mills Nutriblend Green) were delivered through an illuminated (No. 1820 bulb) 5-cm × 6-cm rectangular aperture centered approximately 10.5 cm from the floor of the chamber, measured from the bottom of the aperture.

Response keys were created by presenting square buttons at different locations on a touchscreen monitor with a black background (rgb 0, 0, 0). The touchscreen monitor had a resolution of 480 pixels tall and 800 pixels wide. The monitor had a Mylar screen protector to prevent damage from pigeons’ pecks. The keys 2.7 cm (150 pixels) wide and 2.7 cm (150 pixels) tall. The keys were red (rgb 255, 0, 0) and had a 0.18 cm by 0.18 cm (10 pixels by 10 pixels) white (rgb 255, 255, 255) target centered in the middle of the key. There were two locations of the response key to allow one of the keys to be presented recessed behind a key border. One key, the No Border key, was always presented in the lower center portion of the touchscreen. This location corresponded to the same location used in the previous evaluation of the touchscreen apparatus. The other key, the Border key, was always presented behind a key border in the lower left portion of the touchscreen. The key border was constructed of black adhesive rubber, 5 mm wide and 3 mm thick, which was cut to the shape and size of the response key and glued to the Mylar screen protector. Visual and auditory feedback was provided for 100 ms after each peck on a key by darkening the pecked key (rgb 102, 0, 0) and operating a 28-V feedback relay. Pecks were recorded and experimental events were controlled by a computer program written in Visual Basic 2010.

2.3. Procedure

Sessions were normally conducted 6 or 7 days per week at approximately the same time each day. Before each session, the pigeon was placed in its chamber and exposed to a 5-min blackout to minimize any effects of handling in the trip from the vivarium to the laboratory. After the blackout, the houselight was turned on and a key was displayed on the touchscreen. Keypecking was reinforced with access to food according to a multiple schedule with two FR components. The components had identical ratio requirements, which were individualized for each pigeon as described below, and produced identical 4-s food reinforcers. The components differed based on which key was used. In the No Border component, the No Border key was used; in the Border component, the Border key was used. The two components alternated semi-randomly each session according to a component sequence that was determined by a computer program before each session. The program generated sequences of 41 components at random, until a sequence was generated that met the following restrictions: (a) the sequence included exactly 10 transitions of each type (Border-Border, Border-No Border, No Border-Border, and No Border-No Border) and (b) no more than three Border or No Border components were arranged in succession. All sequences meeting these criteria began and ended with the same type of component. Sessions ended after 41 reinforcers were delivered or 4 hr elapsed, whichever came first.

2.3.1. Preliminary Training

Before the start of this experiment, the pigeons were exposed to evaluations of the touchscreen apparatus described in a previous study of the touchscreen apparatus (see Toegel et al., 2021 for additional details). All pigeons were trained to eat food pellets promptly upon delivery and peck a key displayed in the center of the touchscreen without and with a key border.

2.3.2. The Multiple FR-FR Schedule

To evaluate the effects of the key border on response accuracy and transition-related pausing, pigeons were first exposed to two sessions in which the multiple FR-FR schedule was in effect and the response requirement in both components was FR 1. From there, FR was raised for each pigeon after the completion of at least two sessions with the FR in place unless signs of ratio strain were observed, according to the following progression: From FR 1 to FR 5 in steps of 2, from FR 5 to FR 50 in steps of 5, from FR 50 in steps of 10. Signs of ratio strain included (a) the termination of sessions on the basis of time (the 4-hour maximum) rather than the completion of all 41 FR components and either (b) extremely long pauses (e.g., 10+ min) or (c) multiple long pauses (e.g., 1+ min) observed after part of the ratio was completed. When signs of ratio strain were observed for more than one session, the ratio requirement was reduced to the previous step and the requirement at that step was used as the pigeon’s terminal FR. Table 1 shows the reinforcer durations and FR requirements for each pigeon.

Table 1.

Programmed and Obtained Fixed Ratio (FR) Information in No Border and Border Components by Pigeon

Pigeon	Programmed	Reinforcer	Mean Obtained FR No		Additional % of Programmed FRs
	FR	Duration (s)	No Border	Border	Obtained in No Border Component
88	25	4	50.3	32.3	101
363	50	4	116.7	66.7	133
1108	200	4	223.4	200.9	12
4089	150	4	196.8	151.4	31

Note. The rightmost measure, Additional % of programmed FRs obtained in No Border Component, was calculated by subtracting the programmed FR for each pigeon from the mean obtained FR in No Border components during stable sessions, dividing the remainder by the programmed FR, and multiplying the quotient by 100.

2.3.3. Stability Criteria

Behavior was judged to be stable after a minimum of 10 sessions and when the response accuracy in Border and No Border components did not form a monotonic increasing or decreasing trend over the most recent 5 sessions.

3. Results

3.1. Response Accuracy

Figure 1 shows an illustration of the layout of keys on the touchscreen monitor during No Border (left) and Border (right) components and response locations (x and y) relative to the center of the No Border Key during No Border and Border components in the final session of the evaluation. All four pigeons were more accurate in components with the key border, as evidenced by the higher ratio of circles:x’s shown in graphs in the left column compared to the right column. The pigeons showed idiosyncratic patterns of response locations, which appeared to depend on whether the key border was in place. Generally, the peck locations were concentrated for Pigeons 88, 363, 1108, and 4089 in the lower left, upper right, lower right, and upper right corners of the key, respectively, with the border in place and in the left, upper, lower left, and upper portions of the key, respectively, without the border.

Figure 1.

Response Locations by Component Type in the Final Stable Session

Note. Coordinates (X and Y) relative to the center of the key, in pixels, for all responses during the final stable session of the multiple-schedule evaluation. The dashed squares in each panel represents the key displayed under either the No Border (left) or Border (right) component of the multiple schedule. Gray circles represent On-key pecks; black X’s represent Off-key pecks.

Figure 2 shows response accuracy (left column) and the obtained FRs (right column) in No Border and Border components during the 10 stable sessions of the evaluation. Response accuracy was measured as the percentage of pecks on the key in each component relative to the total number of pecks (On-key plus Off-key) in that component. The figure shows the percentage of On-key pecks in No Border and Border components in each session (dots) and across sessions (bars) of the evaluation. For all pigeons, accuracy was higher in the Border component than in the No Border component. Pigeons 88 and 363 had fairly high accuracy (means of 75–78%) during Border components and poor accuracy (means of 43–50%) during No Border components. Pigeons 1108 and 4089 had near perfect accuracy (means of 99–100%) during Border components and fairly high accuracy (means of between 77–90%) during No Border components.

Figure 2.

Response Accuracy and Pecks per Reinforcer in No Border and Border Components

Note. Accuracy (percent of pecks on key; left) and obtained FRs (pecks per reinforcer; right) across components of each type in each of the 10 stable sessions (dots) and aggregated across sessions (bars). Dashed horizontal lines represent the programmed FR for each pigeon.

3.2. Obtained FRs

The right column of Figure 2 shows the obtained FRs in No Border and Border components in each session (dots) and across sessions (bars), as well as the programmed FR (indicated by the dashed line) for each pigeon. Obtained FRs were calculated as the total number of pecks per reinforcer (On- and Off-key pecks) recorded in components of each type. The combination of On- and Off-key pecks for this measure provides functionally obtained FRs – that is, the actual number of pecks that contacted the touchscreen monitor before the programmed FR requirement was met. The Border component produced obtained FRs that were slightly higher than the programmed FRs for Pigeons 88 and 363 (32 and 67, respectively), and were nearly identical for Pigeons 1108 and 4089 (201 and 151, respectively).

The No Border component produced obtained FRs that were higher than the programmed FRs for all pigeons. The rightmost columns of Table 1 show comparisons between the programmed FRs and obtained FRs in No Border components for each pigeon, expressed as the additional percentage of the programmed FRs added by inaccurate pecks in No Border components. For Pigeons 88 and 363, obtained FRs were much higher than programmed FRs – at least double the FR requirement (50 and 117, corresponding to an additional 101% and 133%, respectively). For Pigeons 1108 and 4089, the obtained FRs were slightly higher than the programmed FRs (223 and 197, corresponding to an additional 12% and 31%, respectively).

Together, the columns of Figure 2 show that pigeons were more accurate in the Border component than the No Border component and that this produced differences in obtained FRs that were relatively large for Pigeons 88 and 363. Accuracy and obtained FRs were similar when arranged by transition type rather than collapsed across components (Supplementary Material A).

3.3. Behavioral Disruptions

Figure 3 shows pausing (medians and interquartile ranges) across the four kinds of transitions arranged in each session for each pigeon. Pausing was measured as the time from the start of a component until the first response on the response key occurred in the next component (see Supplementary Material B for a display of an alternative measure). Because 10 transitions of each type (i.e., Border-Border, Border-No Border, No Border-Border, and No Border-No Border) occurred per session and there were 10 stable sessions, this yielded 100 measurements of pausing for each transition type. Running rates were calculated for completeness by dividing the number of On-key pecks, minus one, by the time to complete each component, minus the pause (see Supplementary Material C). As with pausing, the procedure yielded 100 measurements in transitions of each type. For pigeons 88 and 363, the pigeons with obtained FRs that were at least double the FR programmed in the No Border component, pausing in the transitions from the Border component to the No Border component were extended beyond those observed in the other three types of transitions. For Pigeons 1108 and 4089, the pigeons with obtained FRs that were not substantially higher than the programmed FRs in the No Border component, pausing in the Border-No Border transition was not appreciably different from the other three types of transitions.

Figure 3.

Pausing as a Function of the Past and Upcoming No Border (NB) or Border (B) Components

Note. Dots represent medians and error bars represent interquartile ranges during the 10 stable sessions.

4. Discussion

The multiple FR-FR schedule was effective in establishing differences in accuracy within sessions. For all pigeons, accuracy was higher in Border components than in No Border components. The differences in response accuracy in Border and No Border components yielded differences in obtained FRs, particularly for two pigeons (Pigeons 88 and 363) whose obtained FRs in No Border components were more than double the FR programmed in those components. This finding shows that pigeons’ response accuracy is sensitive to within-session manipulations of the key border, supporting the findings from Test B of Toegel et al.’s (2021) study and providing a method by which interested researchers can manipulate the environment to produce accurate and inaccurate responding by pigeons pecking on a resistive touchscreen.

Furthermore, our findings suggest that rich-lean transitions can be arranged using a multiple schedule with equivalent FR requirements and reinforcer magnitudes, but with differing levels of response difficulty – as defined by the response accuracy generated in Border and No Broder components. It is important to note that the effects of response difficulty appear to be driven by the increase in obtained FRs in the component with the more difficult response. Transitions between response alternatives that generated accurate and inaccurate responding functioned as rich-lean transitions, but only for pigeons with greater discrepancies in accuracy between the components. For Pigeons 88 and 363, pigeons with lower response accuracy, and thus relatively higher obtained FRs in No Border components, extended pausing was observed in the Border-No Border transitions. This finding is consistent with previous research that has established rich-lean transitions using different FR requirements (e.g., Sawyer & Galuska, 2017). For Pigeons 1108 and 4089, pigeons with high accuracy in No Border components, obtained FRs were not much higher than programmed FRs and extended pausing was not observed in Border-No Border transitions. For these pigeons, Border-No Border transitions were not functionally rich-lean transitions.

In a recent study, Pinkston and Moore (2020) evaluated effects of force requirements on operant behavior using a device capable of recording the force of responses that met and did not meet the criterion selected for reinforcement. In their experiment, eight rats’ presses on a 1-cm disk on a force transducer that met a predefined force criterion produced food pellets on average either once every two minutes or twice per minute (i.e., a variable-interval [VI] 120 s or 30 s schedule of reinforcement). Across four conditions, manipulations were made to the criterion selected from reinforcement (either 5.6 g or 32 g) and the average interval used in the VI. They found that the rate of presses that met the force criterion in the conditions with the 32 g force requirement was lower – by approximately 50% – than the rate in conditions with a criterion of 5.6 grams, independent of the VI schedule. Even though the rates of responses meeting the criteria in the 32-g conditions were lower than in the 5.6-g conditions, when the total number of responses – including subcriterion presses that did not meet 32-g requirement – were included in the response rate calculation, response rates were equated. Overall responding did not decrease in the 32-g force requirement conditions, just responses that met the higher reinforcement criterion.

Results from Pinkston & Moore’s (2020) study suggest that the occurrence of subcriterion responding might also increase in other situations in which force requirements are manipulated. As an example, it is possible that the force manipulations used to produce rich and lean schedules in Wade-Galuska et al.’s (2005) study may have had the unmeasured effect of increasing the number of subcriterion responses, thereby increasing the obtained FRs, on the response option with the higher force requirement. If this were the case, the results would be in-line with the findings from the present study – showing that increased obtained FRs may be critically important in establishing rich and lean schedules of reinforcement in procedures that employ manipulations to response difficulty. This possibility could be explored in future studies by arranging rich and lean schedules with differing response force requirements using force transducers – or other devices that are capable of measuring subcriterion responding.

There are several avenues that might be logical extensions for future research. Future research might consider arranging parametric analyses of response options with and without a key border on keys that are made more difficult to peck by manipulating the size of the key (larger or smaller) or by evaluating manipulations of difficulty based on the speed of moving response alternatives. Additionally, the possibility that the behavioral disruptions observed in transitions in the present experiment might be due to discrepancies between the obtained FRs and programmed FRs could be tested using an alternative multiple schedule arrangement (e.g., a multiple FI-FI schedule).

The present study is limited in a couple of ways. First, stability criteria for the multiple-schedule evaluation was based on response accuracy and not on the demonstration of disruptions in rich-lean transitions. This decision likely contributed to the lack of behavioral disruptions by two pigeons; however, it does not detract from the finding that disruptions were established in Border-No Border transitions for the two pigeons in which obtained FRs were substantially larger than the programmed FR requirements. Second, because stability criteria were based on response accuracy and not pausing, it loosened the criterion in relation to pausing in transitions relative to most of the previous research in this area and left open the possibility that trends in pausing in the four types of transitions varied systematically across stable sessions. Upon inspection of the data, this does not appear to have been the case (see Supplementary Materials C for a visual analysis of pausing across the four types of transitions during the 10 stable sessions).

4.1. Conclusions

Overall, results from this study demonstrate that pigeons’ response accuracy is sensitive to within-subject manipulations of a key border and provides a new method of arranging rich-lean transitions: using response options associated with accurate and inaccurate responding. It is also important to note that behavioral disruptions produced in transitions from a response alternative associated with inaccurate responding to an alternative associated with accurate responding appears to depend on inaccurate responding adding to pecking requirements obtained in components with the more difficult response.

Highlights.

• Pigeons’ accuracy is sensitive to within-session manipulations of key borders

• Differences in accuracy can create rich and lean schedules of reinforcement

• Pausing depends on differences in programmed and obtained FRs

Data Availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.