Within-session changes in responding during concurrent variable interval variable ratio schedules

369

JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR 1996, 66, 369–390 NUMBER 3 (NOVEMBER)

WITHIN-SESSION CHANGES IN RESPONDING DURINGCONCURRENT SCHEDULES WITH DIFFERENT REINFORCERS IN

THE COMPONENTS

FRANCES K. MCSWEENEY, SAMANTHA SWINDELL, AND JEFFREY N. WEATHERLY

WASHINGTON STATE UNIVERSITY

Rats and pigeons responded on several concurrent schedules that provided different reinforcers inthe two components (food and water for rats, Experiment 1; wheat and mixed grain for pigeons,Experiment 2). The rate of responding and the time spent responding on each component usuallychanged within the session. The within-session changes in response rates and time spent respondingusually followed different patterns for the two components of a concurrent schedule. For mostsubjects, the bias and sensitivity to reinforcement parameters of the generalized matching law, aswell as the percentage of the variance accounted for, decreased within the session. Negative sensitivityparameters were sometimes found late in the session for the concurrent food-water schedules. Theseresults imply that within-session changes in responding could cause problems for assessing the validityof quantitative theories of concurrent-schedule responding when the components provide differentreinforcers. They question changes in a general motivational state, such as arousal, as a completeexplanation for within-session changes in responding. The results are compatible with satiation for,or sensitization-habituation to, the reinforcers as explanations.

Key words: concurrent schedule, within-session patterns of responding, matching law, lever press,key peck, rats, pigeons

The present experiments were conductedto examine within-session changes in re-sponse rates and time spent responding dur-ing concurrent schedules that provided dif-ferent reinforcers in the two components.The experiments determined whether thewithin-session patterns were similar or differ-ent for the different reinforcers. The answerto this question is important for two reasons.

First, the answer may help to identify thetheoretical variables that produce within-ses-sion patterns of responding. Suppose, for ex-ample, that within-session changes are pro-duced by changes in a general motivationalstate of the animal, such as arousal (e.g., Duf-fy, 1962). If changes in a single state producethe within-session changes in responding dur-ing both components of a concurrent sched-ule, then those changes should be similar forthe two components. In contrast, supposethat within-session changes are produced bysatiation for the reinforcer or by sensitization-habituation to aspects of the experimental sit-uation that are presented repeatedly (e.g., re-inforcers) or for a prolonged period (e.g.,

This material is based on work supported by the Na-tional Science Foundation under Grants IBN-9207346and IBN-9403719.

Reprints may be obtained from Frances K. McSweeney,Department of Psychology, Washington State University,Pullman, Washington 99164-4820.

the experimental enclosure, McSweeney, Hin-son, & Cannon, in press). In that case, within-session patterns might differ for the two com-ponents. Rates of satiation might differ forreinforcers that differ in caloric density, taste,stomach load, and so forth. Rates of sensiti-zation-habituation depend on the prepara-tion under study (e.g., Hinde, 1970) andmight differ for different reinforcers.

Second, the answer to whether respondingchanges in a similar or different pattern hasimplications for quantitative theories of con-current-schedule responding. Many of thesetheories attempt to predict the ratio of therates of responding during the two compo-nents averaged over the session (e.g., Herrn-stein, 1970). One example, the generalizedmatching law (Baum, 1974), appears in Equa-tion 1.

bP T R1 1 15 5 a . (1)1 2P T R2 2 2

P1, T1, and R1 are the rates of respondingemitted, the time spent responding on, andthe rates of reinforcement obtained from thefirst component, respectively. P2, T2, and R2

refer to the same variables for the other com-ponent. The a and b parameters are bias andsensitivity to reinforcement, respectively. Biasrepresents preference for a component that

370 FRANCES K. MCSWEENEY et al.

is not explained by differences in the rates ofreinforcement provided by the components.Sensitivity represents the degree to whichpreference changes with changes in the ratioof the rates of reinforcement.

Within-session changes in respondingcould cause problems for assessing the valid-ity of quantitative theories, such as Equation1, if the changes occur differently for the twocomponents. Suppose that concurrent-sched-ule responding increases to a peak and thendecreases within experimental sessions, as re-sponding often does during multiple andsimple schedules (e.g., McSweeney, 1992).Suppose also that the peak rate of respondingoccurs earlier in the session and that within-session changes are larger for componentsthat provide more highly preferred reinforc-ers than for those that provide less preferredreinforcers. In that case, the ratio of the morepreferred to the less preferred response ratewould not be constant, but would increase toa peak and then decrease within the session.If the peak rate of responding was reached ata constant time after the beginning of a ses-sion regardless of session length (McSweeney,1992; McSweeney, Roll, & Cannon, 1994),then the ratio of response rates would alsodiffer for sessions of different lengths whenthe ratio was calculated across the entire ses-sion.

Within-session changes in respondingwould cause fewer problems for evaluatingquantitative theories if the changes occurredsimilarly for the two components. Suppose,for example, that within-session changes arerelated to changes in a multiplier that mod-ulates the absolute rates at which subjects re-spond. If this multiplier changed in the sameway within the session for the two compo-nents of a concurrent schedule, then its ef-fect would cancel when the ratios of the ratesof responding were calculated.

McSweeney, Weatherly, and Roll (1995)and McSweeney, Weatherly, and Swindell(1996b) studied responding by rats and pi-geons during concurrent schedules that pro-vided a wide range of rates of reinforcement.They reported that the within-session pat-terns of responding were usually similar forthe two components of a concurrent sched-ule regardless of whether similar (McSweeneyet al., 1996b) or different (McSweeney et al.,1995) types of responses produced reinforc-

ers in the two components, and regardless ofwhether reinforcers were delivered at similaror different rates during those components.

The present experiments were designed tofind out whether a similar conclusion wouldbe reached if the two components provideddifferent reinforcers. Heyman (1993) provid-ed prelimary information on this topic. Hestudied responding on concurrent schedulesthat provided ethanol in one component andsucrose in the other. Rate of respondingchanged differently within the session for thetwo components when response rate was plot-ted as a function of the number of obtainedethanol reinforcers.

The present experiments extended thisfinding. They were designed to examine with-in-session changes in responding for non-drug reinforcers (food and water for rats inExperiment 1; wheat and mixed grain for pi-geons in Experiment 2). Within-session pat-terns of time spent responding were exam-ined, as were within-session patterns ofresponse rates. The ratio of the programmedrates of reinforcement was varied so that thegeneralized matching law could be fit to thedata. Subjects were also exposed to schedulesthat delivered rates of reinforcement withinthe range of values typically studied. In Hey-man’s (1993) experiment the variable-inter-val (VI) schedules comprising the concurrentpair were unusually short (VI 5 s). Finally, twospecies of subjects and four types of reinforc-ers were examined to determine the gener-ality of the results.

EXPERIMENT 1

Experiment 1 examined responding whenrats pressed levers during concurrent sched-ules that provided food in one componentand water in the other.

Method

Subjects. The subjects were 5 experimentallynaive male rats, bred from Sprague-Dawleystock. They were maintained at approximate-ly 85% of their free-feeding weights by post-session feedings given when all subjects hadcompleted their daily sessions. Weights wereestablished immediately before the start ofthe experiment, which began when subjectswere approximately 120 days old. During dayswhen a session was conducted, subjects were

371WITHIN-SESSION CHANGES

given 30 min of access to water when all sub-jects had completed the experimental ses-sion. During days when a session was not con-ducted, subjects were given free access towater all day. Subjects were housed individu-ally and were exposed to a 12:12 hr light/dark cycle.

Apparatus. The apparatus was a two-leveroperant conditioning chamber, measuring 29cm by 23 cm by 21.5 cm. Two levers (5.5 cmby 1.5 cm) were located 11.5 cm above thefloor. The left lever was 5 cm from the leftwall; the right lever was 2.5 cm from the rightwall. The levers were made of clear Plexiglasand could be illuminated by lights centeredinside of them. An opening (5.5 cm diame-ter) allowed access to a 0.25-ml dipper. Theopening was located 3 cm above the floor and8.5 cm from the right wall. A rectangularfood cup (4.5 cm wide) extended 4 cm intothe chamber, 5 cm above the floor and 2.5cm from the left wall. The cup was 1.5 cmdeep. A houselight (3 cm diameter) was lo-cated 3 cm from the top of the chamber and11.5 cm from the right wall. The apparatuswas enclosed in a sound-attenuating chamber.An exhaust fan masked noises from outside.Experimental events were presented and datawere recorded by MED Associates softwarerun by an IBMt-compatible 486 computer, lo-cated in another room.

Procedure. Subjects were trained to press theleft and the right levers by a shaping-by-suc-cessive-approximations procedure. The rein-forcer obtained by pressing the left lever wasone 45-mg Noyes pellet. The reinforcer ob-tained by pressing the right lever was 5-s ac-cess to the 0.25-ml dipper that contained wa-ter. The rate of reinforcement obtained bypressing each lever was gradually reduced un-til subjects responded at a steady rate on a VI60-s schedule. Then the experiment began.

In the first condition, subjects respondedon a concurrent VI 60-s VI 60-s schedule. Thelights in the left and right levers were illu-minated at all times during the session, ex-cept during presentation of the water rein-forcers. Reinforcers were scheduled accordingto two 25-interval Fleshler and Hoffman(1962) series. The series used for each leverwas independent of that used for the otherlever. A 3-s changeover delay (COD), duringwhich responses were not reinforced, fol-lowed all changes from one operandum to

the other. Sessions were 60 min long, exclud-ing the time of dipper presentation, and wereconducted daily, five to six times per week.The chamber was illuminated throughout thesession by a houselight.

When subjects had responded on the con-current VI 60-s VI 60-s schedule for 30 ses-sions, they were exposed to the followingschedules in the following order: concurrentVI 15 s VI 240 s, concurrent VI 120 s VI 30 s,concurrent VI 30 s VI 120 s, concurrent VI240 s VI 15 s, and concurrent VI 60 s VI 60s. Here, and throughout this paper, the watercomponent is listed first, and the food com-ponent is listed second. Each schedule wasstudied for 30 sessions.

Results and Discussion

Figures 1 and 2 present the within-sessionpatterns of responding for each subject dur-ing each component of each concurrentschedule. Response rates were calculated bydividing the number of responses in eachcomponent by the total session time. Thetime for which the dipper was available wasexcluded from these calculations. Figures 1and 2 show that subjects often responded atdifferent average rates during the two com-ponents of a concurrent schedule. These dif-ferences could be attributed to several differ-ences between the components, includingthe type of reinforcer provided, the rate ofreinforcement obtained, the response oper-anda used, and so forth. The rate of respond-ing, averaged over the session, usually in-creased with increases in the rates ofreinforcement provided by the componentsfor both food and water. The mean rates ofresponding for food, averaged over all sub-jects, were 2.1, 1.5, 5.2, 7.5, and 5.8 responsesper minute during the VI 240-s, VI 120-s, VI60-s, VI 30-s, and VI 15-s components, respec-tively. The mean rates of responding for waterwere 1.4, 2.9, 3.7, 3.6, and 3.5 responses perminute for the same components presentedin the same order.

Figures 1 and 2 show that rate of respond-ing often changed within the session. Al-though there is some variability in the formof the changes from subject to subject andfrom schedule to schedule, responding forfood primarily increased early in the session.It then increased further, remained relativelyconstant, or decreased somewhat later in the

372 FRANCES K. MCSWEENEY et al.

Fig. 1. Rates of responding (responses per minute) during successive 5-min intervals in the session for individualrats responding on each component of the first concurrent VI 60-s VI 60-s schedule, the concurrent VI 15-s VI 240-sschedule, and the concurrent VI 120-s VI 30-s schedule in Experiment 1. Each row of two graphs presents the resultsfor a concurrent schedule. Responding during the water component appears on the left; responding during the foodcomponent appears on the right. Each curve presents the results for an individual rat. These results, and all of thosethat follow, have been averaged over the last five sessions for which a schedule was available.

373WITHIN-SESSION CHANGES

Fig. 2. Rates of responding (responses per minute) during successive 5-min intervals in the session for individualrats responding on the concurrent VI 30-s VI 120-s schedule, concurrent VI 240-s VI 15-s schedule, and the secondconcurrent VI 60-s VI 60-s schedule in Experiment 1. Results are presented as in Figure 1.

session. Responding for water also increasedearly in the session and then decreased orremained relatively constant.

Figure 3 compares the within-session pat-

terns of responding during the two compo-nents of each concurrent schedule for themean of all subjects. Percentages of total-ses-sion responses are presented so that the dif-

374 FRANCES K. MCSWEENEY et al.

Fig. 3. Percentage of total-session responses during successive 5-min intervals in the session, calculated for themean of all subjects responding on the water (solid line) and food (dashed line) components of each concurrentschedule in Experiment 1. Each graph presents the results for a different schedule. Percentages were calculated bydividing the number of responses for water (or food) during a 5-min interval by the total number of responses forwater (or food) during the session and multiplying by 100%.

375WITHIN-SESSION CHANGES

ferences in the absolute rates of respondingbetween the components (Figures 1 and 2)would not obscure similarities or differencesin the within-session patterns of respondingfor the two components.

Figure 3 shows that within-session changesin responding were larger during compo-nents that provided higher rates of reinforce-ment (e.g., VI 15 s and VI 30 s) than duringcomponents that provided lower rates (e.g.,VI 120 s and VI 240 s). It also shows that thewithin-session patterns of responding dif-fered for the two components, even when thecomponents provided the same rates of re-inforcement (concurrent VI 60-s VI 60-sschedules). This conclusion was confirmed bythe results of two-way (Component 3 Inter-val) within-subject analyses of variance (AN-OVAs) applied to the rates of responding dur-ing each component of each concurrentschedule. The main effect of 5-min intervalwas significant for each schedule: F(11, 44) 54.912, first concurrent VI 60-s VI 60-s sched-ule; F(11, 44) 5 7.679, concurrent VI 15-s VI240-s schedule; F(11, 44) 5 5.918, concurrentVI 120-s VI 30-s schedule; F(11, 44) 5 6.290,concurrent VI 30-s VI 120-s schedule; andF(11, 44) 5 3.553, concurrent VI 240-s VI 15-sschedule, indicating that responding usuallychanged significantly within the session. Theexception was the second concurrent VI 60-sVI 60-s schedule, F(11, 44) 5 1.515. The in-teraction term was significant for each sched-ule: F(11, 44) 5 7.227, F(11, 44) 5 13.940,F(11, 44) 5 2.239, F(11, 44) 5 9.496, F(11,44) 5 2.446, F(11, 44) 5 3.546 (presented inthe order listed above), indicating that thewithin-session patterns of responding dif-fered for the two components of each sched-ule. The main effect of component wassignificant for the first concurrent VI 60-s VI60-s schedule, F(1, 4) 5 11.549, the concur-rent VI 30-s VI 120-s schedule, F(1, 4) 59.362, and the concurrent VI 240-s VI 15-sschedule, F(1, 4) 5 7.774), indicating that ab-solute response rates averaged over the ses-sion sometimes differed for the components.Here, and throughout this paper, results willbe considered to be significant when p , .05.

Figure 4 presents within-session changes intime spent responding on the water compo-nent. Time spent responding on the watercomponent was determined by a timer thatstarted when the subject responded on the

water component and stopped when the sub-ject responded on the food component. Thetimer for food started when the subject re-sponded for food and stopped when the sub-ject responded for water. Results have beenpresented only for water because the timespent responding for food and water summedto the 300 total seconds available for all 5-minintervals except the first. The time to the firstresponse was not included in the time spentresponding for either food or water in thefirst interval.

Figure 4 shows that the time spent respond-ing for water often changed within the ses-sion. When water was provided at a high rate,subjects responded mainly for water early inthe session but shifted to responding mainlyfor food later. When water was provided at alower rate, the time spent responding on thewater component either increased or re-mained relatively constant across the session.One-way (5-min interval) within-subject AN-OVAs applied to the time spent respondingfor water failed to reach significance only forthe first concurrent VI 60-s VI 60-s schedule,F(11, 44) 5 1.597. The ANOVAs were signif-icant for all other schedules: F(11, 44) 515.506, concurrent VI 15-s VI 240-s schedule;F(11, 44) 5 3.213, concurrent VI 120-s VI 30-sschedule; F(11, 44) 5 5.263, concurrent VI30-s VI 120-s schedule; F(11, 44) 5 2.592, con-current VI 240-s VI 15-s schedule; and F(11,44) 5 3.087, second concurrent VI 60-s VI60-s schedule.

Finding different within-session patterns ofresponding for food and water implies thatthe parameters and fit of the generalizedmatching law might not be constant acrossthe session. Figures 5 and 6 confirm this con-clusion. Figure 5 presents the parameters ofthe generalized matching law and the per-centage of the variance accounted for whenresults were calculated for the mean of allsubjects. To compute these parameters, amean rate of responding and a mean timespent responding were calculated over allsubjects responding on each component ofeach concurrent schedule. A linear leastsquares procedure was used to fit the gener-alized matching law to the logarithms of theratios of these means. The left graphs (re-sponse matching) present the results whenthe generalized matching law was fit to theratio of response rates. The right graphs

376 FRANCES K. MCSWEENEY et al.

Fig. 4. Time spent responding on the water component (seconds) during successive 5-min intervals in the sessionin Experiment 1. Each curve presents the results for an individual subject. Each graph presents the results for aconcurrent schedule. The schedule for the water component appears before the schedule for the food componentin the label for the graph.

(time matching) present the results when thegeneralized matching law was fit to the ratioof the times spent responding. In all cases,results for water were divided by those for

food. Figure 6 presents the results when asimilar procedure was used to fit the gener-alized matching law to data for individual sub-jects. Because results were somewhat variable

377WITHIN-SESSION CHANGES

Fig. 5. The bias (top graphs) and sensitivity (middle graphs) parameters of the generalized matching law, as wellas the percentage of the variance accounted for (bottom graphs), during successive 5-min intervals when the gen-eralized matching law was applied to the ratios of the rates of responding (left graphs) and to the ratios of the timesspent responding (right graphs) for the mean of all rats in Experiment 1.

378 FRANCES K. MCSWEENEY et al.

Fig. 6. The bias (top graphs) and sensitivity (middle graphs) parameters of the generalized matching law, as wellas the percentage of the variance accounted for (bottom graphs), when the generalized matching law was appliedto the ratios of the rates of responding (left graphs) and to the ratios of the times spent responding (right graphs)for individual rats in Experiment 1. These parameters were calculated from the data in successive 5-min intervals.Each plotted point is a mean of three successive 5-min intervals. Each curve presents the results for an individualrat.

379WITHIN-SESSION CHANGES

when ratios were calculated over intervals asshort as 5 min, Figure 6 presents the resultswhen the parameters or fits, calculated foreach 5-min interval, were averaged over threesuccessive 5-min intervals.

Figures 5 and 6 show that the bias and sen-sitivity parameters mainly decreased from thebeginning to the end of the session for allsubjects and for the mean of all subjects, forboth response and time matching. The de-creases in the size of both parameters weresteeper for time matching than for responsematching. Bias was sometimes greater than 1early in the session and less than 1 later, in-dicating that bias shifted from one compo-nent to the other as the session progressed.The percentage of variance accounted forusually decreased across the session, with late-session increases for some subjects and forthe mean of all subjects. The sensitivity pa-rameters were sometimes negative late in thesession for time matching. Finally, the size ofthe sensitivity parameter was small (less than0.8) relative to the size usually reported whenthe components of concurrent schedules pro-vide the same type of reinforcers and resultsare averaged over the session. For example,Taylor and Davison (1983) reported that themean size of the sensitivity parameter was0.97 for response matching and 0.96 for timematching for experiments that used relativelystandard procedures and that scheduled re-inforcers according to an exponential pro-gression similar to that used here.

EXPERIMENT 2

Experiment 2 examined responding whenpigeons pecked keys during concurrentschedules that provided wheat in one com-ponent and mixed grain in the other.

Method

Subjects. The subjects were 3 experimentallyexperienced pigeons, maintained at approx-imately 85% of their free-feeding bodyweights by postsession feedings deliveredwhen all subjects had completed the daily ses-sion. Subjects were housed individually andwere exposed to a 12:12 hr light/dark cycle.

Apparatus. The apparatus was a three-keyexperimental enclosure, measuring 32.5 by30.5 by 35.5 cm. Three response keys (2.5 cmdiameter) were located 23.5 cm above the

floor and 7.5 cm apart. The left and rightkeys were located 6.5 cm from the left andright walls, respectively. A force of approxi-mately 0.25 N was required to operate eachkey. Two openings (6.5 cm by 4 cm) allowedaccess to food magazines. They were located9.5 cm below the left and right keys. A lightbehind a panel (4.5 cm diameter) served asthe houselight. It was located 0.75 cm fromthe left wall and 23 cm above the floor. Theexperimental panel was housed in a sound-attenuating chamber. A ventilating fanmasked noises from outside the chamber. Ex-perimental events were programmed anddata were recorded by MED Associates soft-ware run by an IBMt-compatible 486 com-puter, located in another room.

Procedure. Subjects had pecked keys in pre-vious experiments. Therefore, they wereplaced directly on the experimental proce-dure. Subjects responded on the followingconcurrent schedules in the following order:concurrent VI 60 s VI 60 s, concurrent VI 15s VI 240 s, concurrent VI 120 s VI 30 s, con-current VI 30 s VI 120 s, concurrent VI 240s VI 15 s, and concurrent VI 60 s VI 60 s. Hereand throughout this paper, responding onthe right key produced the reinforcers for thecomponent listed first (5-s access to wheat,obtained from the right magazine). Respond-ing on the left key produced the reinforcersfor the component listed second (5-s accessto mixed grain, obtained from the left mag-azine). The left and right keys were illumi-nated with white light except when a rein-forcer was presented. Reinforcers werescheduled according to an independent25-interval Fleshler and Hoffman (1962) se-ries for each component. A 3-s COD, duringwhich no responses were reinforced, followedall changes from one operandum to the oth-er. Sessions were 60 min long, excluding re-inforcement time, and were conducted daily,five to six times per week. Each concurrentschedule was presented for 30 sessions. Thechamber was illuminated throughout the ses-sion by the houselight.

Results and Discussion

Figures 7 and 8 present the within-sessionpatterns of responding for each subject re-sponding during each component of eachconcurrent schedule. Response rates werecalculated as in Figures 1 and 2. Again, the

380 FRANCES K. MCSWEENEY et al.

Fig. 7. Rates of responding (responses per minute) during successive 5-min intervals in the session for individualpigeons responding on each component of the first concurrent VI 60-s VI 60-s schedule, the concurrent VI 15-s VI240-s schedule, and the concurrent VI 120-s VI 30-s schedule in Experiment 2. Each row of two graphs representsresponding during a single concurrent schedule. Responding during the wheat component appears on the left;responding during the mixed grain component appears on the right. The labels present the schedule in the wheatcomponent before the schedule in the mixed grain component. Each curve presents the results for an individualpigeon.

381WITHIN-SESSION CHANGES

Fig. 8. Rates of responding (responses per minute) during successive 5-min intervals in the session for individualpigeons responding on each component of the concurrent VI 30-s VI 120-s schedule, the concurrent VI 240-s VI 15-sschedule, and the second concurrent VI 60-s VI 60-s schedule in Experiment 2. Results are presented as in Figure 7.

time for which reinforcement was availablewas excluded from total session time. Thesefigures show that subjects responded at dif-ferent average rates during the two compo-

nents. This difference could be attributed todifferences in the reinforcers, the responseoperanda, the obtained rates of reinforce-ment, and so forth, available on the two com-

382 FRANCES K. MCSWEENEY et al.

ponents. Rates of responding, averagedacross the session, usually increased with in-creases in the rates of reinforcement provid-ed by the component. The mean rates of re-sponding for wheat were 3.8, 7.0, 19.8, 23.4,and 27.6 responses per minute for the VI240-s, VI 120-s, VI 60-s, VI 30-s, and VI 15-sschedules, respectively. The mean rates of re-sponding for mixed grain were 14.1, 18.0,28.4, 32.3, and 24.2 responses per minute forthe same schedules presented in the same or-der.

Figure 9 compares the within-session pat-terns of responding during the two compo-nents of each concurrent schedule. Again,percentages have been presented instead ofabsolute response rates so that differences inthe absolute rates of responding averagedover the session (Figures 7 and 8) would notobscure similarities or differences in the with-in-session patterns of responding. Althoughthere is some variability in the form of thewithin-session changes in responding fromsubject to subject and from schedule toschedule, Figures 7, 8, and 9 show that re-sponse rates often changed within the ses-sion. The changes were also larger when thecomponents provided higher (e.g., VI 15 sand VI 30 s) rather than lower (e.g., VI 120s and VI 240 s) rates of reinforcement.

Figure 9 shows that the within-session pat-terns of responding often differed for thecomponents. This impression was confirmedby the results two-way (Component 3 Inter-val) within-subject ANOVAs applied to therates of responding during each concurrentschedule. The main effect of 5-min intervalwas significant for each schedule: F(11, 22) 52.542, first concurrent VI 60-s VI 60-s sched-ule; F(11, 22) 5 2.696, concurrent VI 120-sVI 30-s schedule; F(11, 22) 5 4.028, concur-rent VI 30-s VI 120-s schedule; F(11, 22) 54.447, concurrent VI 240-s VI 15-s schedule.The exceptions were for the concurrent VI15-s VI 240-s schedule, F(11, 22) 5 1.605, andthe second concurrent VI 60-s VI 60-s sched-ule, F(11, 22) 5 1.182. These results indicatethat responding usually changed significantlywithin the session. The interaction termswere also significant for all schedules, F(11,22) 5 6.332, F(11, 22) 5 6.001, F(11, 22) 52.710, and F(11, 22) 5 7.840 (presented inthe same order as above) except for the con-current VI 15-s VI 240-s schedule, F(11, 22)

5 1.760, and the second concurrent VI 60-sVI 60-s schedule, F(11, 22) 5 1.334. This in-dicates that the within-session patterns of re-sponding often differed for the two compo-nents. The main effect of component wassignificant for the concurrent VI 120-s VI 30-sschedule, F(1, 2) 5 29.147, and the concur-rent VI 240-s VI 15-s schedule, F(1, 2) 528.699, indicating that the absolute rates ofresponding averaged over the session some-times differed for the two components.

Figure 10 presents within-session patternsof time spent responding on the wheat com-ponent for individual subjects. Again, resultshave been presented only for the wheat com-ponent because the time spent respondingon the two components must sum to the 300total seconds available for all 5-min intervalsexcept the first. Figure 10 shows that the timespent responding for wheat changed withinthe session, often increasing and then de-creasing. One-way (5-min interval) within-subject ANOVAs applied to the time spent re-sponding for wheat were significant for thefirst, F(11, 22) 5 5.605, and the second, F(11,22) 5 5.423, concurrent VI 60-s VI 60-s sched-ules but not for the concurrent VI 15-s VI240-s schedule, F(11, 22) 5 1.262, the con-current VI 120-s VI 30-s schedule, F(11, 22)5 1.144, the concurrent VI 30-s VI 120-sschedule, F(11, 22) 5 1.529, and the concur-rent VI 240-s VI 15-s schedule, F(11, 22) 51.424.

Examination of Figure 10 suggests that thetime spent responding may have changed sig-nificantly within sessions for individual sub-jects even when the change was not signifi-cant for the mean of all subjects. This wasconfirmed by the results of one-way (5-mininterval) ANOVAs applied to the time spentresponding on the wheat component by in-dividual subjects during the last five sessionsfor which a concurrent schedule was avail-able. ANOVAs were calculated only for sched-ules with nonsignificant effects of time for themean of all subjects. Time spent respondingfor wheat changed significantly within the ses-sion for all schedules for Pigeon 5502, F(11,44) 5 17.781, concurrent VI 15-s VI 240-sschedule; F(11, 44) 5 2.198, concurrent VI120-s VI 30-s schedule; F(11, 44) 5 5.425, con-current VI 30-s VI 120-s schedule; F(11, 44)5 4.519, concurrent VI 240-s VI 15-s sched-ule; and Pigeon 5503, F(11, 44) 5 5.796,

383WITHIN-SESSION CHANGES

Fig. 9. Percentage of total-session responses during successive 5-min intervals, calculated for the mean of allsubjects responding on the wheat (solid line) and mixed grain (dashed line) components of each concurrent sched-ule in Experiment 2. Each graph presents the results for a different schedule. The labels on the graphs present thewheat schedule before the mixed grain schedule. Percentages are calculated and presented as in Figure 3.

384 FRANCES K. MCSWEENEY et al.

Fig. 10. The time spent responding on the wheat component (seconds) during successive 5-min intervals in thesession in Experiment 2. Each function presents the results for an individual pigeon. Each graph presents the resultsfor a concurrent schedule. The schedule of wheat reinforcement is presented before the schedule of mixed grainreinforcement in the labels for the graphs.

385WITHIN-SESSION CHANGES

F(11, 44) 5 7.264, F(11, 44) 5 19.779, F(8,32) 5 2.915 (presented in the same order asabove). Time spent responding for wheatchanged significantly for Pigeon 5501 only onthe concurrent VI 30-s VI 120-s schedule,F(11, 44) 5 8.211. Fewer degrees of freedomappear for Pigeon 5503 on the concurrent VI240-s VI 15-s schedule because this subject didnot respond on the wheat componentthroughout the session during that schedule.

Figures 11 and 12 present within-sessionchanges in the parameters and fit of the gen-eralized matching law. Figure 11 presents theresults calculated for the mean of all subjects,and Figure 12 presents the results for individ-ual subjects. Results were analyzed and pre-sented as in Figures 5 and 6. Results for wheatwere divided by those for mixed grain. Thesize of the sensitivity parameter decreasedfrom the beginning to the end of the sessionfor all subjects, although the size of thechange was small for Pigeon 5503 when timematching was considered. The percentage ofthe variance decreased for 2 subjects but in-creased for Pigeon 5503. The bias parameteralso changed differently across the session fordifferent subjects, resulting in erratic changesin the bias parameter for the mean of all sub-jects. As in Experiment 1, the sensitivity pa-rameters were small (usually less than 0.8)relative to those observed for concurrentschedules that provide the same reinforcersin the two components when results are av-eraged over the session. Unlike the resultspresented in Experiment 1, the sensitivity pa-rameters were not negative late in the sessionand the bias parameters were usually less than1 throughout the session.

GENERAL DISCUSSIONImplications for Within-Session Changesin Responding

Rate of responding and time spent re-sponding usually changed within sessionswhen rats responded on concurrent food-wa-ter schedules (Experiment 1) and when pi-geons responded on concurrent wheat–mixed grain schedules (Experiment 2). Thisextends the generality of within-sessionchanges in responding to concurrent sched-ules that provide different reinforcers in thetwo components and to time spent respond-ing as a measure of behavior.

Different within-session patterns of re-sponding were usually reported for the twocomponents of the concurrent food-waterand the concurrent wheat–mixed grainschedules. These results are similar to thosereported by Heyman (1993) for concurrentethanol-sucrose schedules. The results differfrom the similar patterns of respondingfound for concurrent schedules that providethe same reinforcers in the two components(McSweeney et al., 1995, 1996b).

Finding different within-session patterns inthe two components questions the idea thatchanges in a general motivational variable,such as arousal, provide a complete accountfor within-session changes in responding.Arousal can refer to a state of the organism(e.g., Duffy, 1962). If within-session changesin responding for food and water (or wheatand mixed grain) were both produced solelyby changes in a single state, then the patternsof responding should have been similar dur-ing the two components.

McSweeney, Swindell, and Weatherly (inpress) also reported results that question anarousal-based interpretation of within-sessionchanges in responding. They studied within-session changes in operant responding whenadjunctive drinking or wheel running wasalso available. The rates of operant and ad-junctive responding often changed within ex-perimental sessions. The correlation betweenthe rates of these two types of behavior, cal-culated over the session, was inconsistentlypositive and negative. It was not strongly pos-itive, as would be expected if changes in asingle variable, such as arousal, produced thewithin-session changes in both types of be-havior.

Finding different within-session changes inresponding for different reinforcers is morecompatible with explanations for thesechanges in terms of sensitization-habituationto, or satiation for, the reinforcer. As arguedearlier, these processes might occur at differ-ent rates for different reinforcers. To clarifythe difference between sensitization-habitua-tion and satiation, satiation can refer to a de-crease in consumption when an ingestivestimulus (e.g., food, water) is presented re-peatedly. Sensitization-habituation refers toan increase (sensitization) followed by a de-crease (habituation) in responding to a stim-ulus when that stimulus is presented repeat-

386 FRANCES K. MCSWEENEY et al.

Fig. 11. The bias (top graphs) and sensitivity (middle graphs) parameters of the generalized matching law, aswell as the percentage of the variance accounted for (bottom graphs), during successive 5-min intervals when thegeneralized matching law was applied to the ratios of the rates of responding (left graphs) and to the ratios of thetimes spent responding (right graphs) for the mean of all pigeons in Experiment 2.

387WITHIN-SESSION CHANGES

Fig. 12. The bias (top graphs) and sensitivity (middle graphs) parameters of the generalized matching law, aswell as the percentage of the variance accounted for (bottom graphs), when the generalized matching law was appliedto the ratios of the rates of responding (left graphs) and to the ratios of the times spent responding (right graphs)for individual pigeons in Experiment 2. These parameters were calculated from the data in successive 5-min intervals.Each plotted point is a mean of three successive 5-min intervals. Each curve presents the results for an individualpigeon.

388 FRANCES K. MCSWEENEY et al.

edly or for a prolonged time. Becausereinforcers are stimuli, sensitization-habitua-tion might occur in their presence, produc-ing systematic changes in their strength orvalue when they are repeatedly presentedwithin sessions (e.g., McSweeney, Weatherly,& Swindell, 1996a). It should be noted thatsensitization-habituation to the sensory char-acteristics of an ingestive stimulus may con-tribute to satiation for that stimulus, butadditional factors such as gastric fill, nutri-tional state, and postingestive consequencesalso contribute to satiation (e.g., Swithers-Mulvey & Hall, 1993).

McSweeney, Hinson, and Cannon (inpress) argued that sensitization-habituationprovides a better explanation than satiationfor within-session changes in responding un-der the moderate conditions (e.g., interme-diate rates of reinforcement) employed inmost studies. First, the empirical characteris-tics of the within-session patterns of operantresponding are similar to the empirical char-acteristics of behavior reported in the litera-ture on sensitization-habituation. Second,within-session changes have been reportedeven when no reinforcers are delivered (e.g.,Schoenfeld, Antonitis, & Bersh, 1950). Sen-sitization-habituation, but not satiation, canaccount for these changes because it can oc-cur in the presence of noningestive stimulisuch as the experimental enclosure. Finally,many factors that alter the rate of satiationfor food (e.g., the caloric density of the food,the subjects’ level of food deprivation, thesize of a prefeeding) do not alter within-ses-sion patterns of operant responding (Roll,McSweeney, Johnson, & Weatherly, 1995).

It should be noted that McSweeney, Hin-son, and Cannon (in press) did not arguethat factors other than sensitization-habitua-tion never contribute to within-sessionchanges in responding. Most studies of with-in-session changes provide easy-to-manipulateoperanda and intermediate amounts of rein-forcement (e.g., one Noyes pellet perreinforcer and approximately one reinforcerper minute). Under these conditions, sensi-tization-habituation to the stimulus proper-ties of the reinforcer (and possibly to otherstimuli, such as the experimental enclosure)may be largely responsible for the within-ses-sion changes in responding. However, undermore extreme conditions (e.g., very large re-

inforcers, difficult responses), other variables(e.g., other factors related to satiation, fa-tigue) might also play a role.

Implications for Quantitative Theories

The parameters and fit of the generalizedmatching law reported here were generallyconsistent with those reported in past studiesthat provided different reinforcers in thecomponents of concurrent schedules. For ex-ample, negative sensitivity parameters weresometimes reported in Experiment 1. Nega-tive sensitivity parameters have been reportedfor monkeys responding in a closed economyon three-component concurrent food-food-water schedules (Hursh, 1978). The presentsensitivity parameters were also somewhatsmaller than those reported in studies thatprovided the same reinforcers in the twocomponents. Matthews and Temple (1979)reported relatively small sensitivity parame-ters when cows responded on concurrentschedules that provided different types offood in the two components.

Figures 5, 6, 11, and 12 show that the pa-rameters and fit of at least one quantitativetheory, the generalized matching law,changed within experimental sessions. Thesize of bias decreased from the beginning tothe end of the session for all individual ratsand for 2 of 3 pigeons. The sensitivity param-eter decreased from the beginning to the endof the session for all subjects, although someof these changes were small. The percentageof the variance accounted for by the gener-alized matching law also decreased from thebeginning to the end of the session for allsubjects except Rat 41 and Pigeon 5503. Insome cases, the percentage was lowest to-wards the middle of the session and increasedagain later.

As argued earlier, these results imply thatthe fit and parameters of the generalizedmatching law may vary with session length.Shorter sessions may predominantly sampleearly-session time when the parameters arelarge and the fit is good. Longer sessions mayinclude more of the late-session time. The re-sults also indicate that applying the general-ized matching law to data calculated over thewhole session may neglect molecular and dy-namic changes in the processes that governresponding for different reinforcers. A fullaccount of the present data will undoubtedly

389WITHIN-SESSION CHANGES

depend on understanding these processes.For example, dynamic food-water interac-tions may have contributed to the results ofExperiment 1. In addition, the similarity ofthe results in Experiments 1 and 2 suggeststhat a factor common to food-water andwheat–mixed grain interactions also contrib-uted.

An understanding of these processes is notavailable at this time. Instead, the present dis-cussion will center on the implications of thepresent results for the generalized matchinglaw. Even this discussion must be interpretedwith care, however. There are many possibleexplanations for within-session changes in theparameters and fit of the generalized match-ing law. For example, late-session decreases inthe percentage of variance accounted formight have contributed to the late-session de-creases in the size of sensitivity parameters(but see the results for Rat 41 and Pigeon5503). As a result, all of the present conclu-sions should be regarded as tentative withoutfurther study.

The present within-session changes in biasare generally consistent with Baum’s (1974)interpretation of that parameter. Baum de-fined bias as preference for a component thatis not explained by the rates of reinforcementprovided by the components. He argued thatthe use of qualitatively different reinforcersin the components could result in a bias pa-rameter that differed from 1. His interpreta-tion could accommodate the present resultsif the relative values of the two reinforcerschanged within the session.

Baum (1974) suggested that changes in de-privation for the reinforcer alter the sensitiv-ity parameter. The present results, combinedwith those of past studies, show that the re-lation between deprivation and sensitivitymay be complex. Charman and Davison(1983) and Herrnstein and Loveland (1974,as interpreted by Baum, 1974) found thatsensitivity increased to approach 1 as depri-vation for the programmed reinforcer de-creased. In contrast, the present resultsshowed that sensitivity decreased with de-creases in deprivation if it is assumed thatsubjects were less deprived late in the session,after consuming many reinforcers, than theywere earlier in the session. However, depri-vation probably changed at different rates forthe different types of reinforcers presented in

the two components of this study. Such dif-ferential changes might substantially compli-cate the relation between deprivation and thesize of the sensitivity parameter.

Finding negative sensitivity parameters inExperiment 1, but not in Experiment 2, isconsistent with Hursh’s (1980, 1984) argu-ment that negative parameters occur whenreinforcers are complements rather than sub-stitutes. Reinforcers are complements if in-creasing the availability of one increases thedemand for the other. They are substitutes ifincreasing the availability of one decreasesthe demand for the other. Food and watershould be complements, but similar types offood should be substitutes (Hursh, 1980, p.235). Therefore, as reported, negative sensi-tivity parameters should have been found inExperiment 1, which provided complements,but not in Experiment 2, which provided sub-stitutes.

If confirmed by future experiments, find-ing negative sensitivity parameters for timematching for some subjects in Experiment 1would question whether negative sensitivityparameters occur in closed, but not in open,economies (Hursh, 1978). In an open econ-omy, subjects are given extrasession food andwater. In a closed economy, subjects obtaintheir entire daily ration of food and waterduring the experimental session. Experiment1 employed an open economy and reportednegative sensitivity parameters.

The bias parameter and the percentage ofthe variance accounted for increased acrossthe session for Pigeon 5503, the opposite ofthe pattern reported for the other subjects.The reason for this difference is not known.One possibility is that different processes pro-duce the within-session patterns for differentsubjects. A more parsimonious explanation isthat the variables that produce the within-ses-sion patterns change at different rates for dif-ferent subjects. The results of several studies,including the present one, show that within-session patterns may take several forms. Re-sponding may increase, decrease, or increaseand then decrease within the session. Thesefindings suggest that two independent pro-cesses, one governing the increase and onegoverning the decrease in responding, pro-duce most of the within-session changes (e.g.,McSweeney, Hinson, & Cannon, in press). Ifthese processes occur at somewhat different

390 FRANCES K. MCSWEENEY et al.

rates for different subjects, as they might ifthe changes were produced by sensitization-habituation (e.g., Hinde, 1970), then thewithin-session changes in response rateswould take a different form for different sub-jects. As a result, different within-sessionchanges in the ratios of the rates of respond-ing, and therefore, the parameters and fit ofthe generalized matching law, might be ex-pected for different subjects.

REFERENCES

Baum, W. M. (1974). On two types of deviation from thematching law: Bias and undermatching. Journal of theExperimental Analysis of Behavior, 22, 231–242.

Charman, L., & Davison, M. (1983). On the effects offood deprivation and component reinforcer rates onmultiple-schedule performance. Journal of the Experi-mental Analysis of Behavior, 40, 239–251.

Duffy, E. (1962). Activation and behavior. New York: Wiley.Fleshler, M., & Hoffman, H. S. (1962). A progression for

generating variable-interval schedules. Journal of theExperimental Analysis of Behavior, 5, 529–530.

Herrnstein, R. J. (1970). On the law of effect. Journal ofthe Experimental Analysis of Behavior, 13, 243–266.

Herrnstein, R. J., & Loveland, D. H. (1974). Hunger andcontrast in a multiple schedule. Journal of the Experi-mental Analysis of Behavior, 21, 511–517.

Heyman, G. M. (1993). Ethanol regulated preference inrats. Psychopharmacology, 112, 259–269.

Hinde, R. A. (1970). Behavioral habituation. In G. Horn& R. A. Hinde (Eds.), Short-term changes in neural ac-tivity and behavior (pp. 3–40). Cambridge, England:Cambridge University Press.

Hursh, S. R. (1978). The economics of daily consump-tion controlling food- and water-reinforced respond-ing. Journal of the Experimental Analysis of Behavior, 29,475–491.

Hursh, S. R. (1980). Economic concepts for the analysisof behavior. Journal of the Experimental Analysis of Be-havior, 34, 219–238.

Hursh, S. R. (1984). Behavioral economics. Journal of theExperimental Analysis of Behavior, 42, 435–452.

Matthews, L. R., & Temple, W. (1979). Concurrentschedule assessment of food preference in cows. Jour-nal of the Experimental Analysis of Behavior, 32, 245–254.

McSweeney, F. K. (1992). Rate of reinforcement and ses-sion duration as determinants of within-session pat-terns of responding. Animal Learning & Behavior, 20,160–169.

McSweeney, F. K., Hinson, J. M., & Cannon, C. B. (inpress). Sensitization-habituation may occur duringoperant conditioning. Psychological Bulletin.

McSweeney, F. K., Roll, J. M., & Cannon, C. B. (1994).The generality of within-session patterns of respond-ing: Rate of reinforcement and session length. AnimalLearning & Behavior, 22, 252–266.

McSweeney, F. K., Swindell, S., & Weatherly, J. N. (inpress). Within-session changes in adjunctive and in-strumental responding. Learning and Motivation.

McSweeney, F. K., Weatherly, J. N., & Roll, J. M. (1995).Within-session changes in responding during concur-rent schedules that employ two different operanda.Animal Learning & Behavior, 23, 237–244.

McSweeney, F. K., Weatherly, J. N., & Swindell, S.(1996a). Reinforcer value may change within exper-imental sessions. Psychonomic Bulletin & Review, 3, 372–375.

McSweeney, F. K., Weatherly, J. N., & Swindell, S.(1996b). Within-session changes in responding dur-ing concurrent variable-interval schedules. Journal ofthe Experimental Analysis of Behavior, 66, 75–95.

Roll, J. M., McSweeney, F. K., Johnson, K. S., & Weatherly,J. N. (1995). Satiety contributes little to within-ses-sion decreases in responding. Learning and Motivation,26, 323–341.

Schoenfeld, W. N., Antonitis, J. J., & Bersh, P. J. (1950).Unconditioned response rate of the white rat in a bar-pressing apparatus. Journal of Comparative and Physio-logical Psychology, 43, 41–48.

Swithers-Mulvey, S. E., & Hall, W. G. (1993). Integrationof oral habituation and gastric signals in decerebraterat pups. American Journal of Physiology, 265, R216–R219.

Taylor, R., & Davison, M. (1983). Sensitivity to reinforce-ment in concurrent arithmetic and exponentialschedules. Journal of the Experimental Analysis of Behav-ior, 39, 191–198.

Received March 20, 1996Final acceptance July 5, 1996