rule, and 0.59 was that obtained by Lander and Irwin (1968). These results are in agree- nment with data reported by Shimp and Wheatley (1971) from a similar ...
1972, 17, 45-49
JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR
NUMBER
1
(JANUARY)
COMPONENT DURATION AND RELATIVE RESPONSE RA TES IN MULTIPLE SCHEDULES1 JOAO CLAUDIO TODOROV UNIVERSIDADE DE SAO PAULO
Pigeons were trained on a multiple variable-interval 30-sec, variable-interval 90-sec schedule with each component presented alternately for an equal (on the average) duration. This average duration of exposure to each component was varied from 5 to 300 sec. The main concern was with rate of response in the variable-interval 30-sec component relative to rate of response in the variable-interval 90-sec component. In all cases, rate of response was higher in the variable-interval 30 sec component, but the discrepancy in the rate produced by the two schedules tended to be greatest when the duration of component presentation was brief. The mean proportion of responses emitted during the variable-interval 30sec component (responses in variable-interval 30-sec component divided by total responses) varied from about 0.60 to 0.71, where 0.75 would be expected on the basis of a matching rule, and 0.59 was that obtained by Lander and Irwin (1968). These results are in agreenment with data reported by Shimp and Wheatley (1971) from a similar experiment.
Lander and Irwin (1968) suggested that the of a may depend on the immediate consequences of changeovers between the concurequation rent schedules (Shull and Pliskoff, 1967; (1) Todorov, 1969, 1971). For a given pair of un1 N2 (n2) equal VI schedules, the relative rate associated describes the relationship between the number with the most favorable schedule seems to inof responses per component and the number crease (and so increases the value of a) as the of reinforcements per component in concur- rate of changeovers decreases. When the rate rent and in multiple schedules. N and n rep- of changeovers is high, the subjects alternate resent the number of responses and reinforce- responses on both schedules, and the relative ments, respectively; the subscripts represent response rate is insensitive to reinforcement the components of the multiple schedule or distribution between the schedules (a = 0). If the concurrent operants; and a is a constant the relationships between relative response assuming the value 1.0 for concurrent sched- rate and relative reinforcement rate in conules and %/3 for multiple schedules. This equa- current VI VI schedules can be affected by the tion was based on data obtained by Lander length of exposure to each schedule between and Irwin and on data reported by Reynolds changeovers (interchangeover time), it is possible that that relationship in multiple VI VI (1963). Lander and Irwin (1968) suggested that schedules also will be a function of the frea = 1/3 in multiple variable-interval, variable- quency of alternation of the schedules. interval (VI VI) schedules, and that a = 1.0 in concurrent VI VI schedules. However, reMETHOD sults from a number of experiments suggest that in concurrent VI VI schedules the value Subjects Three adult, White Carneaux pigeons were maintained at 80% of normal body weight 'This research was supported in part by research determined during free access to grain. The grant NsG-450 from the National Aeronautics and Space Administration. The experiment was conducted subjects had previous experience with several while the author was in the faculty of the Department reinforcement schedules. of Psychology, Mary Washington College of the University of Virginia. Reprints may be obtained from the author, Departamento de Neuropsiquiatria e Psicologia M6dica, Faculdade de Medicina de Ribeirao Prkto, Av. 9 de Julho, 980, Ribeirao Pr&to, Sao Paulo, Brasil.
Apparatus A standard experimental box for pigeons, with two response keys (Ferster and Skinner,
45
46
JOAO CLAUDIO TODOROV
1957), was used. The left response key re- five durations were used (3, 4, 5, 6, and 7 sec) mained inoperative during the entire experi- due to teclhnical problems. The red and green ment; the right key was transilluminated components alternated and lhad, at the end of either by a red or a green light. A minimum a session, the same average duration. The force of 0.09807 N was required to operate average duration of the component schedules the response key. The reinforcer was a brief was, at first, 10 sec each. Table 1 shows the period (3 sec to one subject, 4 sec to another, values of average component duration investiand 6 sec to the third one) of free access to gated and their order of presentation. As the grain, presented to the pigeons through a tape arranging component duration stopped solenoid-operated hopper. when the feeder was operated, the occurrence of a reinforcement in a component did not reProced ure duce the time during which the subject could The key-pecking behavior of the pigeons respond in that component. had been established in previous experiments. Other contingencies. Contrary to the proBeginning with the first session in the present cedure used by Slhimp and Wheatley (1971), experiment, the subjects' responses were rein- no changeover delay was scheduled to prevent forced on a two-component multiple variable- reinforcement of the first response in each cominterval variable-interval (mult VI VI) sched- ponent. When a component ended during the ule of reinforcement. emission of a burst of responses, reinforcement VI schedule. A separate VI schedule was could be initiated in a component different associated with each component. Interrein- from the one in effect when the response burst forcement intervals were based on an arith- started. The sequence of variable component metic scale. The schedules used were VI 30-sec durations was intended to diminish the proband VI 90-sec. A reinforcement would be as- ability of the development of different resigned to a component only when its associated sponse rates in different parts of each period keylight was on; otherwise, that tape puller of exposure of the schledules. When fixed would stop until its associated keylight came component durations are used, there is the on again. When the tape puller stopped be- possibility of a high response rate in the last cause a reinforcement was scheduled before seconds of the least-favorable component the end of a given interval, it remained still maintained by the probability that a reinforceuntil the next response associated with that ment will be assigned as soon as the other comschedule was reinforced. ponent is in effect. Stimuli. The key was red when the VI 30-sec There was no cancelling of reinforcements was in effect and green when VI 90-sec was in arranged but not produced before a stimulus effect. Additional illumination in the chamber change. That reinforcement would occur after was provided by two houselights. During rein- the first response emitted on the return of the forcements, houselights and the keylight were component in which the reinforcement was turned off. Each response produced auditory arranged. feedback by operating a relay attached to the A session was terminated when the sixtieth other side of the intelligence panel. reinforcement occurred. Because of this proComponent alternation. Component dura- cedure, the sums of seconds of exposure for tion was arranged on a variable-interval basis each component could be slightly different in according to an arithmetic progression. Min- any given session. The totals of time of eximum component duration in all average posure of each component (fourth and fifth durations used was 3 sec. The maximum com- columns in Table 1) and the totals of reinponent duration would vary according to the forcements (sixth and seventh columns) show average duration used at a given experimental that the procedure did not change the schedcondition. For any average duration it would uled frequency of reinforcements per unit of be equal to the double of the average duration time in each schedule. minus three. The other eight durations, schedTraining under all component durations uled according to an arithmetic progression, was continued for at least 14 daily sessions and were d1 = 3 sec, d6 =t sec, and d1l = (2t - 3) until the rate of responding in both composec. However, there was an exception. When nents was judged to be stable over the last average component duration was 5 sec, only five sessions. The stability criterion was the
47
COMPONENT DURATION IN MULTIPLE SCHEDULES Table 1
Original data totalled across the last five sessions in Mult VI 30-sec (red) VI 90-sec (green). Average Comp. Duration (Sec)
Sessions
Red
Green
Red
Green
Red
Green
10 40 10 5 150
15 16 19 17 16
2925 3428 2999 2635
6689 6877
6609
300
17
P-14 6967 5896 6424 6017 5941 5181
5
16
6571
225 223 222 225 228 223 222
75 77 78 75 72 77 78
10 40 10 5 150 300 5
14 21 18 14 16
P-16 4151 3173 2810 4193 3916
6640 6705 7059 6794 7193 7609 6976
231 228 225 216
69 72 75 73 77 84
228
72
10
15
40 10 5 150 300
16
6621 6819 6846
227 223 224
73 77 76
6840
225 226 227
75 74 73
Responses
17 16
19 17 16
17
3298 3862 P-17 8087 9123 10158 9752 8482 7755
2971 3346 3067 1353 1572 1654 1960 2789 2716 1695
2701 4122 4323 4549 4891 4916
Time (Sec)
6785 6924 7044 6816 6825 6676 7146
7232 7006
6972 6811
7030 6648 6780 6764 6823 6910 6925
6837 6808 6912 6288 6758 6787
6576 6592
Reinf
227 223
Figure 1 also shows part of the data reported by Shimp and Wheatley (1971). The dashed line connects the means of three pigeons. The dotted horizontal lines represent the points at which relative response rates would match RESULTS relative reinforcement rates on the two sets of The original data from each subject, to- data (0.80 in the experiment by Shimp and talled across the last five sessions, are presented Wheatley and 0.75 in the present experiment). in Table 1. From these data the rate of re- Relative response rate in the Shimp and sponding under each VI schedule was com- Wheatley experiment also seems to be a deputed. Response rate is here defined as the creasing function of component duration. As the component duration decreased, in number of responses emitted in the presence of a given schedule divided by the time spent both experiments, the relative rate of respondin the presence of that schedule (reinforce- ing in a component deviated away from the ment time excluded). Relative response rate value predicted by equation 1 (0.61 for the is here defined as the rate under a given sched- Shimp and Wheatley experiment and 0.59 for ule divided by the sum of rates under both the present experiment), given Lander and Irwin's (1968) suggestion that a = % in multischedules. Figure 1 shows how the relative response ple schedules. The deviations increase toward rates under VI 30-sec changed with manipula- the values of relative rate of reinforcement of tions in average component duration. The each schedule as component duration is resolid line connects the mean of the group of duced to 10 sec or less in both experiments. points from each component duration. Rel- When the average component duration was 5 ative response rates on the most favorable sec in the, present experiment, the relative reschedule seems to be a generally decreasing sponse rate was not quite as close to the relative reinforcement rate as when the averagefunction of average component duration.
absence of upward or downward trends on relative response rates in a component, plotted against sessions, over the last five sessions.
JOAO CLAUDIO TODOROV
48
ships reported by those authors for mult VI VI schedules. When the average component duration was closer to interchangeover times com'. 0.71 monly found in concurrent VI VI schedules, 'IL relative rates of responding tended to match %,V.% relative rates of reinforcement. 1*16 0.70 Similar conclusions were presented by Shimp and Wheatley (1971) from an experiment that, in spite of several procedural de0.65 tails different from the present experiment, provided similar data. The differences in the general shape of the functions in Figure 1 0.60 may be related to: different relative rates of p-1. reinforcement used (0.80 and 0.75); different overall frequency of reinforcements per hour 0.55 AVERAGE OF 3 BIRDS (+40 and 80), the use of a changeover delay FROM SHIMP AND WHEATLEY (1971) of 1 sec by Shimp and Wheatley to prevent re0.50 inforcement in one component from occurring 5 2 10 30 40 15016'0 300 after a response burst initiated on the other component; the use of fixed component duraCOMPONENT DURATION (sec) tion by Shimp and Wheatley, and of variable Fig. 1. Relative rate of responding as a function of component duration in the present experiaverage component duration in mult VI VI. The solid ment; the way in which reinforcements were curve was drawn connecting the mean point at each component duration, and represents the data from all arranged-constant reinforcement per opporthree subjects. The dashed curve represent data from tunity VI schedules used by Shimp and Wheata similar experiment conducted by Shimp and Wheatand arithmetic VI schedules on the presley (1971). The dotted lines indicate the relative rate ley of reinforcement on each experiment. It was 0.75 for ent experiment. These procedural differences the present experiment and 0.80 for the Shimp and notwithstanding, it can be concluded from Wheatley experiment. both experiments that reductions in component duration in mult VI VI move the reladuration was 10 sec, but it was still closer to tive response rate toward a value equal to the matching the relative reinforcement rate than relative reinforcement rate. On the other hand, the data from short comto the value predicted by equation (1). A similar finding is reported by Shimp and Wheatley ponent durations indicate the possibility of a with respect to the lowest component dura- maximum for the function between compotions (2 sec and 5 sec) used during their nent duration and relative response rate. Shimp and Wheatley reported that when the experiment. component duration was 2 sec, the relative response rate in a component sometimes was DISCUSSION further from the relative reinforcement rate The results clearly show the importance of in that component than when component ducomponent duration in the determination of ration was 5 sec. In the present experiment, a response rates in mult VI VI schedules. In all similar relationship was found between durations of 5 sec and 10 sec. If one assumes that cases, response rates were higher in VI 30-sec than in VI 90-sec, but the discrepancy in the as component duration tends toward zero, the rate produced by the two schedules tended to differential responding in the presence of the be the greatest when the duration of the com- discriminative stimuli tends to disappear, it can be expected that relative response rate ponent was short. When the average component duration was will tend to 0.50 as component duration tends 150 sec or 300 sec, close to the 180-sec fixed to zero. However, it should be noted that for component duration used by Lander and average component duration of 5 sec, the Irwin (1968) and by Reynolds (1963), the func- number of durations used (five) was lower than tion relating relative response rate to relative the number used on all other average durareinforcement rate was more like the relation- tions (11). This difference determines an overn An_
1%----
""I'll
qk.
11% %
W*4
0
0
*
P-14
o
o
A
p- I?
COMPONENT DURATION IN MULTIPLE SCHEDULES
49
run at short durations, and might explain the difference in inclination on the left side of the functions shown in Figure 1.
tion. Journal of the Experimental Analysis of Behavior, 1963, 6, 131-139. Shimp, C. P. and Wheatley, K. L. Matching to relative reinforcement frequency in multiple schedules with a short component duration. Journal of the Experimental Analysis of Behavior, 1971, 15, 205-
REFERENCES
210. Shull, R. L. and Pliskoff, S. S. Changeover delay and concurrent schedules: some effects on relative performance measures. Journal of the Experimental Analysis of Behavior, 1967, 10, 517-527. Todorov, J. C. Concurrent performances: effect of punishment contingent on the switching response. Journal of the Experimental Analysis of Behavior, 1971, 16, 61-62. Todorov, J. C. Timeout from positive reinforcement as punishment for choice behavior. Paper presented at the meeting of the Virginia Academy of Sciences, Fredericksburg, May, 1969.
Catania, A. C. Concurrent operants. In W. K. Honing (Ed.), Operant behavior: areas of research and application. New York: Appleton-Century-Crofts, 1966. Pp. 213-270. Ferster, C. B. and Skinner, B. F. Schedules of reinforcement. New York: Appleton-Century-Crofts,
1957. Lander, D. G. and Irwin, R. J. Multiple schedules: effects of the distribution of reinforcements between components on the distribution of responses between components. Journal of the Experimental Analysis of Behavior, 1968, 11, 517-524. Reynolds, G. S. Some limitations on behavioral contrast and induction during successive discrimina-
Received: 12 September 1970. (Final Acceptance: 19 August 1971.)
