- Email: [email protected]
The effect of food deprivation on within-session patterns of wheel running
Behavioural Processes 46 (1999) 121 – 129
The effect of food deprivation on within-session patterns of wheel running Jeffrey N. Weatherly *, Ashley S. McMurry, Cam L. Melville Department of Psychology, McNeese State Uni6ersity, Lake Charles, LA 70609 -1895, USA Received 20 October 1998; received in revised form 4 February 1999; accepted 9 February 1999
Abstract In the present study, rats’ wheel running was reinforced when they responded at different food deprivations. In experiment 1, fourths of a wheel turn were reinforced on a variable interval (VI) 15- or 60-s schedule during 50-min sessions. Subjects responded at 75, 85 or 95% of their free-feeding body weights, across conditions. Within-session decreases in responding were steepest at subjects’ 75% weights for the VI 60-s schedule, but were similar at different weights for the VI 15-s schedule. In experiment 2, subjects responded on a VI 60-s schedule at 75 or 95% of their free-feeding body weights. Reinforcer size was one or four food pellets. Steeper within-session decreases in responding were observed at subjects’ 75% weight than at their 95% weight, with no effect of reinforcer amount. The present results cannot disconfirm either leading theory for within-session changes because the terms involved (i.e. habituation and satiation) are not adequately defined. However, the present results seem to pose problems for both explanations. © 1999 Elsevier Science B.V. All rights reserved. Keywords: Within-session response patterns; Food deprivation; Variable interval schedule; Wheel running; Rat
1. Introduction Since first explicitly described (McSweeney et al., 1990), within-session changes in responding have received much empirical and theoretical attention (McSweeney et al., 1996; Palya and Wal* Corresponding author. Tel.: +1-318-4755435; fax: +1318-4755467. E-mail address: [email protected] (J.N. Weatherly)
ter, 1997; Bizo et al., 1998). Two explanations for these changes remain viable, the first being that they are produced by arousal and satiation and the second that they are produced by sensitization and habituation. Unfortunately, neither explanation is defined in such a way as to allow for a critical test between them. In the absence of such a test, however, there are still reasons to study these changes in responding. The present experiments studied wheel running of
0376-6357/99/$ - see front matter © 1999 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 6 - 6 3 5 7 ( 9 9 ) 0 0 0 3 1 - 5
122
J.N. Weatherly et al. / Beha6ioural Processes 46 (1999) 121–129
rats by reinforcing fourths of a wheel turn on a simple variable-interval (VI) schedule throughout the session. In experiment 1, subjects responded at three different levels of their free-feeding body weights across conditions and wheel running was studied at two different VI schedule values. In experiment 2, subjects responded at two different levels of their free-feeding body weight and amount of reinforcement was manipulated across conditions. Studying wheel running as the operant is interesting for two reasons. The first is that it is a response yet to be explicitly studied in a vast literature on within-session changes in responding. Finding that rates of wheel running change within the session would further generalize the phenomenon. The second, and perhaps the more important, reason is that by studying wheel running, the study of within-session changes in responding makes contact with the theory of activity-based anorexia (see Pierce and Epling, 1994 for a review). Activity-based anorexia is founded on two basic, and supported, tenants. The first is that increasing the level of food deprivation increases the reinforcing value of exercise. Pierce et al. (1986) (experiment 1) found that rats would press a bar that allowed access to a running wheel more often when they were food deprived than when they were not. The second tenant is that there is an inverse relationship between amount of exercise and the reinforcing value of food. Pierce et al. (1986) (experiment 2) had rats press a bar for food pellets delivered by a progressive-ratio (PR) schedule (i.e. 5, 10, 15, 20...). They found that when rats were allowed to exercise (i.e. run in a wheel) prior to the session, they quit bar pressing earlier in the PR schedule than they did when they were not allowed to exercise prior to the session. This theory is of interest because it posits an indirect interaction between food deprivation and reinforcer value. That is, as the animal becomes more food deprived, it should become more active and, as it becomes more active, the value of food reinforcement should decrease. Likewise, as the animal becomes less food deprived, it should become less active and, thus, the value of food reinforcement should increase.
This interaction poses an interesting question about within-session changes in responding. Intuitively, one would expect that within-session decreases in responding would be less steep when subjects are more food deprived than when they are less food deprived. However, if subjects are more active when they are more, rather than less, food deprived, then the value of food reinforcement should be lower when subjects are more, rather than less, food deprived. If this were the case, then one might expect steeper within-session decreases in responding when subjects are more food deprived than when they are less food deprived.
2. Experiment 1
2.1. Methods 2.1.1. Subjects A total of four experimentally naive Long– Evans rats served. Subjects were obtained from Charles River Laboratories and were approximately 6 months old at the beginning of the experiment. Subjects were housed individually, had water freely available in the home cage, and experienced a 14/10-h light/dark cycle. 2.1.2. Apparatus The apparatus was a 30× 25× 22-cm experimental enclosure for rats. On the front panel, two 4.75-cm levers were located, 1.5 cm from the side wall and 7.5 cm above the grid floor. A 2.5-cm diameter light was located 6.75 cm above each lever. A recessed 5×5-cm aperture, that could allow access to the reinforcer, was centered on the front panel. The aperture was located 2 cm above the floor. A pellet dispenser (Med Associates ENV-203), located behind the front panel, delivered reinforcers. A 0.75-cm diameter houselight was centered on the back wall, 1 cm from the ceiling. A 7× 10.5-cm opening was centered, at floor level, on the side wall opposite to the door of the chamber. This opening allowed access to a 35-cm diameter running wheel. The wheel was 6 cm wide. The chamber was enclosed within a sound attenuating box. An IBM-compatible 486
J.N. Weatherly et al. / Beha6ioural Processes 46 (1999) 121–129
computer, connected to a Med-Associates interface and running Med-State software, controlled the experimental events and recorded the data.
2.1.3. Procedure Subjects’ free-feeding body weights were established prior to beginning the experiment. Subjects were given free access to food for 2 months. Subjects were then weighed and each subject’s 75, 85, and 95% free-feeding body weight was determined using the outcome from this single weighing. The first phase of the experiment was to measure unconditioned rates of wheel running. Subjects had free access to the wheel for 50 min. The houselight was illuminated throughout the session. No food pellets were delivered and only wheel running was recorded. Each subject was given three of these sessions at its 75, 85, and 95% weight, conducted in that order for all subjects. In the second phase of the experiment, subjects’ wheel running was reinforced on a simple VI schedule. The reinforcer was a 45-mg Noyes pellet (Formula A/I) that was delivered by a VI 15- or 60-s schedule, in different conditions. Reinforcers were programmed by sampling at a probability of 0.0667 (VI 15 s) or 0.0167 (VI 60 s) every 1 s. When a reinforcer became available, the light above each lever became illuminated and remained illuminated until a wheel response was recorded and the reinforcer was delivered. These lights were extinguished during the inter-reinforcer interval (IRI). Once a reinforcer was scheduled, the IRI did not advance until it had been collected. One fourth of a wheel turn was designated a response (i.e. one revolution of the wheel would be recorded as four responses). The houselight was illuminated throughout the session. Sessions lasted 50 min and began approximately 15 s after the subject was placed into the chamber. Each subject responded in a total of six conditions of 15 sessions each. There were two subjects who experienced the following order of VI schedules at the following body weights (in parentheses): VI 60 s (95%), VI 15 s (75%), VI 15 s (85%), VI 60 s (85%), VI 60 s (75%), VI 15 s (95%). The other two subjects received the reverse order of conditions. A session was conducted if the subject
123
weighed within 9 3% of its target body weight (e.g. 72–78% for the 75% conditions). Sessions were conducted daily, 7 days per week (if the subjects met the weight criterion). Maintenance feedings, when necessary, occurred after all subjects had completed their daily sessions.
2.2. Results and discussion Appendix A contains the rates of wheel running and obtained rates of reinforcement for individual subjects during the final five sessions of each condition of the entire study. It shows that unconditioned rates of wheel running increased with food deprivation. Therefore, although the subjects in the present study were older than those typically used in activity-based anorexia research (Pierce and Epling, 1994), changes in the activity rates for the present subjects were similar to those observed for younger animals. Appendix A also shows that reinforced response rates were always well above unconditioned response rates at each level of deprivation for each rate of reinforcement for each subject, indicating that wheel running was indeed reinforced by food pellets. Fig. 1 displays the observed within-session patterns of responding. Presented are the rates of wheel running (in fourths of a wheel turn per min) across successive 5-min intervals of the session. The upper graph shows the results for the conditions in which a VI 15-s schedule was in effect. The lower graph shows the results for the VI 60-s schedule conditions. Each function represents the mean for all subjects responding at 75 (closed squares), 85 (open squares), or 95% (closed triangles) of their free-feeding body weights. Results for each condition represent responding during the final five sessions that each condition was in effect. The upper graph of Fig. 1 shows that, when subjects responded on a VI 15-s schedule, food deprivation effected absolute rates of responding, but not the within-session pattern of responding. Absolute rates of wheel running varied directly with food deprivation and responding decreased within the session in a similar fashion at each level of deprivation. A two-way (deprivation by 5-min interval) repeated-measures analysis of variance
124
J.N. Weatherly et al. / Beha6ioural Processes 46 (1999) 121–129
(ANOVA), conducted on the rates of responding for individual subjects, confirmed these impres-
Fig. 1. Presented are the rates of wheel running across successive 5-min intervals in the session. The upper and lower graphs present the results for the VI 15-s and VI 60-s conditions, respectively. Each function represents the results for the mean for all subjects responding at 75% ( ), 85% ( ), or 95% () of their free-feeding body weights.
sions. The main effect of deprivation was significant [(F(2, 6)= 7.02, PB 0.05], indicating that subjects responded at different absolute rates at different food deprivations. The main effect of 5-min interval [F(9, 27) = 7.88, PB 0.001] was also significant, indicating that the decrease in responding within the session was systematic. Finally, the interaction between deprivation and 5-min interval was not significant [F(18, 54)= 0.99], indicating that the within-session pattern of wheel running did not differ for different food deprivations. Results for this analysis, and for those that follow, were considered significant at PB 0.05. The lower graph of Fig. 1 indicates that food deprivation did alter the within-session pattern of wheel running when subjects responded on a VI 60-s schedule. A steeper within-session decrease in responding was observed when subjects responded at 75% of their free-feeding body weight than at the lesser deprivations. A two-way (deprivation by 5-min interval) repeated-measures ANOVA, conducted on the rates of responding for individual subjects, produced significant main effects of deprivation [F(2, 6)= 40.77, PB 0.001] and 5-min interval [F(9, 27) = 18.66, P B 0.001]. These results indicate that absolute rates of wheel running varied directly with food deprivation and that the within-session decrease in responding was systematic. The interaction between deprivation and 5min interval was also significant [F(18, 54)= 2.41, PB 0.01], indicating that different within-session patterns of wheel running were observed at different levels of food deprivation. The significant interaction term does not necessarily lead to the conclusion that the within-session decrease in responding on the VI 60-s schedule was steeper at subjects’ 75% weight than at their 85 and 95% weights. However, the slopes of response functions shown in the lower graph of Fig. 1 indicated that this was indeed the case. Slopes of the best-fit lines, plotted using the data from the entire session, were − 2.25, − 1.64, and − 1.44 for the response patterns at subjects’ 75, 85, and 95% weight, respectively. The results for the mean of all subjects were also representative of those for individual subjects. Appendix A presents the standard error of
J.N. Weatherly et al. / Beha6ioural Processes 46 (1999) 121–129
the mean of responding within the session for individual subjects, which gives an indication of within-session variance in responding. Standard errors for individual subjects were relatively similar at different deprivations for the VI 15-s schedule. However, the standard errors were larger when subjects were more food deprived than when less food deprived for the VI 60-s schedule, indicating that individual subjects showed larger within-session changes in responding when they were more, rather than less, food deprived.
3. Experiment 2 The results of experiment 1 suggest that withinsession patterns of wheel running sometimes, but not always, decrease more steeply when subjects are more food deprived than when they are less food deprived. This was the case when subjects responded on a VI 60-s schedule. However, when subjects received more food (i.e. responded on a VI 15-s schedule), within-session patterns of wheel running did not differ at different levels of deprivation. Experiment 2 was conducted to determine whether the results for the VI 60-s schedule could be replicated if it provided more food than in experiment 1. Subjects responded on a simple VI 60-s schedule at either 75 or 95% of their freefeeding body weight, across conditions. The size of the reinforcer was either one or four 45-mg food pellets. The reinforcer size of four food pellets was chosen because a VI 60-s schedule delivering four pellets per reinforcer programmed the equal number of reinforcers per session as did the VI 15-s schedule in experiment 1.
3.1. Method 3.1.1. Subjects and apparatus Of the four subjects used in experiment 1, three served. The fourth subject died before completing all the conditions and its data were excluded from the following analyses. Subjects were housed and maintained as in experiment 1. They responded in the same apparatus as described in experiment 1.
125
3.1.2. Procedure The procedure was identical to that of experiment 1 with the following exceptions. First, the first phase of experiment 1 was not repeated, rather subjects were immediately placed on the experimental procedure. Second, reinforcers were always scheduled by a VI 60-s schedule. Third, in some conditions, a response would be reinforced with four 45-mg Noyes pellets rather than just one. Subjects 70 and 71 responded in the following conditions in the following order (body weights appear in parentheses) four pellets (95%), one pellet (95%), four pellets (75%), and one pellet (75%). Condition order was reversed for subject 72. 3.2. Results and discussion Fig. 2 shows that patterns of responding differed with food deprivation, but not with reinforcer size. A steeper within-session decrease in responding was observed when subjects were more food deprived than when they were less food deprived. It presents the rates of wheel running (in fourths of a wheel turn per min) across successive 5-min intervals of the session. Each function represents the mean for all subjects responding at 75 (solid symbols) or 95% (open symbols) of their free-feeding body weight when the reinforcer size was one (triangles) or four (squares) pellets. Results were calculated using the final five sessions that each condition was in effect. The above impressions were confirmed by statistical analyses. A two-way (condition by 5min interval) repeated-measures ANOVA, conducted on responses rates for individual subjects, showed that the main effect of condition [F(3, 6)= 5.48, PB 0.05], main effect of 5-min interval [F(9, 18) = 18.28, PB 0.001], and interaction between condition and 5-min interval [F(27, 54)= 1.80, PB 0.05] were significant. These results indicate that absolute rates of responding differed across conditions, that responding changed systematically within the session, and that different within-session patterns of responding were observed in the different conditions. Because the interaction term was significant, follow-up two-way (reinforcer size by 5-min inter-
126
J.N. Weatherly et al. / Beha6ioural Processes 46 (1999) 121–129
Fig. 2. Presented are the rates of wheel running across successive 5-min intervals in the session. Each function represents the results for the mean for all subjects responding at 75% for four pellets ( ), 75% for one pellet (), 95% for four pellets ( ) or 95% for one pellet (open triangles).
val) repeated-measures ANOVAs were conducted on the data at each level of food deprivation. When subjects responded at 75% of their freefeeding body weight, the main effect of 5-min interval [F(9, 18) =9.99, P B0.001] was significant, but neither the main effect of reinforcer size [F(1, 2) = 1.63] nor the interaction term [F(9, 18)=1.36] reached significance. In other words, similar response patterns were observed for both reinforcer sizes. Similar results were observed when subjects responded at 95% of their free-feeding body weight. Again, the main effect of 5-min interval [F(9, 18) = 57.01, P B0.001] was significant, but neither the main effect of reinforcer size [F(1, 2) =0.04] nor the interaction term [F(9, 18)=1.55] reached significance. Simply because response patterns differed at 75 and 95% body weights does not necessarily mean
that steeper decreases in responding were observed when subjects were more, rather than less, food deprived. However, the slopes of the response functions presented in Fig. 2 support this claim. Slopes of the best-fit lines for the 75% conditions, plotted using the data from the entire session, were − 1.97 when one pellet was delivered and − 2.21 when four pellets were delivered. Slopes for the 95% conditions were − 1.30 when one pellet was delivered and − 1.59 when four pellets were delivered. The results in Appendix A again indicate that these results were representative of those for individual subjects. Standard errors of the mean for responding within the session were almost always larger at subjects’ 75% weight than at their 95% weight, supporting the idea that larger changes in responding were observed when subjects were more, rather than less, food deprived.
J.N. Weatherly et al. / Beha6ioural Processes 46 (1999) 121–129
4. General discussion The present results demonstrate that withinsession changes in wheel running are observed when running serves as the operant response. The present results also indicate that within-session rates of wheel running decrease more steeply when subjects are more food deprived than when they are less food deprived, at least at moderate rates of reinforcement (i.e. VI 60-s schedule). Different patterns of responding were not observed at different levels of deprivation when subjects responded at a high rate of reinforcement (i.e. VI 15-s schedule). Why similar within-session patterns were observed at the high, but not at the moderate, rate of reinforcement is not known, but differences in the amount of food delivered by the two schedules can be ruled out as a cause. The VI 60-s schedule, four-pellet conditions of experiment 2 delivered the equivalent number of food pellets as did the VI 15-s schedule conditions of experiment 1. However, the results of experiment 2 showed that response patterns varied with level of food deprivation and not the number of food pellets delivered. It is therefore likely that the difference in results is tied to the frequency of reinforcer delivery, not to the amount of reinforcement delivered. One possible explanation for the difference between the two schedules is the presence of competing responses. That is, competing responses (e.g. interim behaviors; Staddon and Simmelhag, 1971) may have occurred at a higher rate when subjects were more food deprived than when they were less food deprived, thus producing different response patterns. However, for the VI 15-s schedule, the inter-reinforcement interval may have been too short to allow for differences in these interim behaviors. This potential explanation must remain speculative, however, because subjects were not observed while responding. Because amount of reinforcement did not alter response patterns in experiment 2, one could argue that the present results are incompatible with some existing evidence on within-session patterns of responding. For instance, both Palya
127
and Walter (1997) and Bizo et al. (1998) report different within-session patterns of responding with different reinforcer amounts. However, the discrepancy is potentially the product of differences in procedure. Whereas both Palya and Walter and Bizo et al. used pigeons and varied reinforcer amount by varying hopper duration, the present study used rats and varied reinforcer amount by delivering a different number of food pellets. Roll et al. (1995) studied rats and reported results similar to those reported here. Furthermore, although Palya and Walter and Bizo et al. took great care to correlate hopper duration to amount eaten, reinforcer amount may have varied from reinforcer to reinforcer. Such was not the case in the present study. As noted above, the leading explanations for within-session changes in responding (i.e. satiation and habituation) are not adequately defined to allow for a critical test between them. Therefore, the present results cannot favor one explanation over the other. However, the results would seem to pose some difficulty for each explanation. For instance, finding steeper decreases in responding for subjects that are more food deprived and responding for less food than for subjects that are less food deprived and responding for more food would appear to run contrary to an intuitive interpretation of satiation. On the other hand, one could argue that finding seemingly steeper decreases in responding at lower, than at higher, rates of reinforcement (Fig. 1) runs contrary to what one would predict according to habituation. Finally, the present results may suggest avenues of future research in the area of activitybased anorexia. For instance, the present results suggest that decreases in operant responding for food depends more on the frequency of food delivery than on the amount of food delivered. The present results may therefore suggest that anorexia may be overcome more easily if food is presented infrequently in large quantities rather than frequently in small quantities. Clearly, the present results cannot be directly extended to anorexia nervosa. They do, however, indicate that the study of within-session changes in responding may have far-reaching implications.
J.N. Weatherly et al. / Beha6ioural Processes 46 (1999) 121–129
128
Acknowledgements
Experiment 1: VI 60-s schedule
Partial funding for this research was provided by the Shearman Endowed Professorship in Education awarded to C.L.M. The authors thank the Department of Biological and Environmental Sciences at MSU for sharing laboratory space, Hanna C. Rue, Jason E. Stout, Tiffani Thibodeaux, and Carolyn S. Davis for their help in collecting the data, and Frances K. McSweeney for her comments on an earlier version of this manuscript.
70
71
72
73 Appendix A Presented are the rates of wheel running (in fourths of a wheel turn per min; R) and obtained reinforcers (OR) for each subject in each condition of the study. Also presented are the standard errors of the means (S.E.M.), calculated on the mean rates of responding per 5-min interval across the session. The S.E.M.s therefore provide a measure of the change in responding within the session for each subject.
Unconditioned
R R R R
85%
1.7 2.3 2.3 3.2
1.5 0.9 1.3 2.5
95% 0.8 0.6 1.5 1.2
Experiment 1: VI 15-s schedule 70
71
72
73
R S.E.M. OR R S.E.M. OR R S.E.M. OR R S.E.M. OR
16.3 2.1 38.8 13.2 1.3 40.4 21.5 3.3 51.6 15.6 2.9 47.8
10.6 0.7 48.0 10.8 2.0 45.4 18.3 3.0 50.4 10.1 1.5 49.0
6.0 1.1 40.4 6.3 1.6 44.2 10.2 2.4 47.6 7.6 1.5 46.2
Experiment 2: VI 60-s schedule, one pellet 70
71
72
R S.E.M. OR R S.E.M. OR R S.E.M. OR
19.6 2.5 42.8 8.2 0.9 41.8 27.7 3.0 44.0
6.6 0.9 41.0 8.6 1.4 44.2 8.1 1.6 38.8
Experiment 2: VI 60-s schedule, four pellets 75%
70 71 72 73
R S.E.M. OR R S.E.M. OR R S.E.M. OR R S.E.M. OR
22.8 1.1 156.0 7.8 0.5 155.8 32.5 4.3 171.4 20.5 2.6 170.4
16.6 1.7 144.6 9.4 0.8 160.4 24.1 4.0 169.6 12.0 1.7 150.0
70
71
72 7.9 1.3 136.0 6.7 0.9 143.8 12.3 2.4 152.0 10.4 1.7 157.2
R S.E.M. OR R S.E.M. OR R S.E.M. OR
21.7 2.0 44.0 12.1 1.6 46.8 26.8 3.4 42.6
6.6 1.7 37.0 7.1 1.4 46.4 9.2 2.1 37.2
References Bizo, L.A., Bogdanov, S.V., Killeen, P.R., 1998. Satiation causes within-session decreases in instrumental responding. J. Exp. Psych. Anim. Behav. Process. 24, 439 – 452. McSweeney, F.K., Hatfield, J., Allen, T.M., 1990. Withinsession responding as a function of post-session feedings. Behav. Process. 22, 177 – 186.
J.N. Weatherly et al. / Beha6ioural Processes 46 (1999) 121–129 McSweeney, F.K., Hinson, J.M., Cannon, C.B., 1996. Sensitization-habituation may occur during operant conditioning. Psychopharmacol. Bull. 120, 256–271. Palya, W.L., Walter, D.E., 1997. Rate of a maintained operant as a function of temporal position within a session. Anim. Learn. Behav. 25, 291–300. Pierce, W.D., Epling, W.F., 1994. Activity anorexia: an interplay between basic and applied behavior analysis. Behav. Anal. 17, 7 – 23.
.
129
Pierce, W.D., Epling, W.F., Boer, D.P., 1986. Deprivation and satiation: the interrelations between food and wheel running. J. Exp. Anal. Behav. 46, 199 – 210. Roll, J.M., McSweeney, F.K., Johnson, K.S., Weatherly, J.N., 1995. Satiety contributes little to within-session decreases in responding. Learn. Motiv. 26, 323 – 341. Staddon, J.E.R., Simmelhag, V.L., 1971. The ‘superstition’ experiment: a reexamination of its implications for the principles of adaptive behavior. Psych. Rev. 78, 3 – 43.
The effect of food deprivation on within-session patterns of wheel running Jeffrey N. Weatherly *, Ashley S. McMurry, Cam L. Melville Department of Psychology, McNeese State Uni6ersity, Lake Charles, LA 70609 -1895, USA Received 20 October 1998; received in revised form 4 February 1999; accepted 9 February 1999
Abstract In the present study, rats’ wheel running was reinforced when they responded at different food deprivations. In experiment 1, fourths of a wheel turn were reinforced on a variable interval (VI) 15- or 60-s schedule during 50-min sessions. Subjects responded at 75, 85 or 95% of their free-feeding body weights, across conditions. Within-session decreases in responding were steepest at subjects’ 75% weights for the VI 60-s schedule, but were similar at different weights for the VI 15-s schedule. In experiment 2, subjects responded on a VI 60-s schedule at 75 or 95% of their free-feeding body weights. Reinforcer size was one or four food pellets. Steeper within-session decreases in responding were observed at subjects’ 75% weight than at their 95% weight, with no effect of reinforcer amount. The present results cannot disconfirm either leading theory for within-session changes because the terms involved (i.e. habituation and satiation) are not adequately defined. However, the present results seem to pose problems for both explanations. © 1999 Elsevier Science B.V. All rights reserved. Keywords: Within-session response patterns; Food deprivation; Variable interval schedule; Wheel running; Rat
1. Introduction Since first explicitly described (McSweeney et al., 1990), within-session changes in responding have received much empirical and theoretical attention (McSweeney et al., 1996; Palya and Wal* Corresponding author. Tel.: +1-318-4755435; fax: +1318-4755467. E-mail address: [email protected] (J.N. Weatherly)
ter, 1997; Bizo et al., 1998). Two explanations for these changes remain viable, the first being that they are produced by arousal and satiation and the second that they are produced by sensitization and habituation. Unfortunately, neither explanation is defined in such a way as to allow for a critical test between them. In the absence of such a test, however, there are still reasons to study these changes in responding. The present experiments studied wheel running of
0376-6357/99/$ - see front matter © 1999 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 6 - 6 3 5 7 ( 9 9 ) 0 0 0 3 1 - 5
122
J.N. Weatherly et al. / Beha6ioural Processes 46 (1999) 121–129
rats by reinforcing fourths of a wheel turn on a simple variable-interval (VI) schedule throughout the session. In experiment 1, subjects responded at three different levels of their free-feeding body weights across conditions and wheel running was studied at two different VI schedule values. In experiment 2, subjects responded at two different levels of their free-feeding body weight and amount of reinforcement was manipulated across conditions. Studying wheel running as the operant is interesting for two reasons. The first is that it is a response yet to be explicitly studied in a vast literature on within-session changes in responding. Finding that rates of wheel running change within the session would further generalize the phenomenon. The second, and perhaps the more important, reason is that by studying wheel running, the study of within-session changes in responding makes contact with the theory of activity-based anorexia (see Pierce and Epling, 1994 for a review). Activity-based anorexia is founded on two basic, and supported, tenants. The first is that increasing the level of food deprivation increases the reinforcing value of exercise. Pierce et al. (1986) (experiment 1) found that rats would press a bar that allowed access to a running wheel more often when they were food deprived than when they were not. The second tenant is that there is an inverse relationship between amount of exercise and the reinforcing value of food. Pierce et al. (1986) (experiment 2) had rats press a bar for food pellets delivered by a progressive-ratio (PR) schedule (i.e. 5, 10, 15, 20...). They found that when rats were allowed to exercise (i.e. run in a wheel) prior to the session, they quit bar pressing earlier in the PR schedule than they did when they were not allowed to exercise prior to the session. This theory is of interest because it posits an indirect interaction between food deprivation and reinforcer value. That is, as the animal becomes more food deprived, it should become more active and, as it becomes more active, the value of food reinforcement should decrease. Likewise, as the animal becomes less food deprived, it should become less active and, thus, the value of food reinforcement should increase.
This interaction poses an interesting question about within-session changes in responding. Intuitively, one would expect that within-session decreases in responding would be less steep when subjects are more food deprived than when they are less food deprived. However, if subjects are more active when they are more, rather than less, food deprived, then the value of food reinforcement should be lower when subjects are more, rather than less, food deprived. If this were the case, then one might expect steeper within-session decreases in responding when subjects are more food deprived than when they are less food deprived.
2. Experiment 1
2.1. Methods 2.1.1. Subjects A total of four experimentally naive Long– Evans rats served. Subjects were obtained from Charles River Laboratories and were approximately 6 months old at the beginning of the experiment. Subjects were housed individually, had water freely available in the home cage, and experienced a 14/10-h light/dark cycle. 2.1.2. Apparatus The apparatus was a 30× 25× 22-cm experimental enclosure for rats. On the front panel, two 4.75-cm levers were located, 1.5 cm from the side wall and 7.5 cm above the grid floor. A 2.5-cm diameter light was located 6.75 cm above each lever. A recessed 5×5-cm aperture, that could allow access to the reinforcer, was centered on the front panel. The aperture was located 2 cm above the floor. A pellet dispenser (Med Associates ENV-203), located behind the front panel, delivered reinforcers. A 0.75-cm diameter houselight was centered on the back wall, 1 cm from the ceiling. A 7× 10.5-cm opening was centered, at floor level, on the side wall opposite to the door of the chamber. This opening allowed access to a 35-cm diameter running wheel. The wheel was 6 cm wide. The chamber was enclosed within a sound attenuating box. An IBM-compatible 486
J.N. Weatherly et al. / Beha6ioural Processes 46 (1999) 121–129
computer, connected to a Med-Associates interface and running Med-State software, controlled the experimental events and recorded the data.
2.1.3. Procedure Subjects’ free-feeding body weights were established prior to beginning the experiment. Subjects were given free access to food for 2 months. Subjects were then weighed and each subject’s 75, 85, and 95% free-feeding body weight was determined using the outcome from this single weighing. The first phase of the experiment was to measure unconditioned rates of wheel running. Subjects had free access to the wheel for 50 min. The houselight was illuminated throughout the session. No food pellets were delivered and only wheel running was recorded. Each subject was given three of these sessions at its 75, 85, and 95% weight, conducted in that order for all subjects. In the second phase of the experiment, subjects’ wheel running was reinforced on a simple VI schedule. The reinforcer was a 45-mg Noyes pellet (Formula A/I) that was delivered by a VI 15- or 60-s schedule, in different conditions. Reinforcers were programmed by sampling at a probability of 0.0667 (VI 15 s) or 0.0167 (VI 60 s) every 1 s. When a reinforcer became available, the light above each lever became illuminated and remained illuminated until a wheel response was recorded and the reinforcer was delivered. These lights were extinguished during the inter-reinforcer interval (IRI). Once a reinforcer was scheduled, the IRI did not advance until it had been collected. One fourth of a wheel turn was designated a response (i.e. one revolution of the wheel would be recorded as four responses). The houselight was illuminated throughout the session. Sessions lasted 50 min and began approximately 15 s after the subject was placed into the chamber. Each subject responded in a total of six conditions of 15 sessions each. There were two subjects who experienced the following order of VI schedules at the following body weights (in parentheses): VI 60 s (95%), VI 15 s (75%), VI 15 s (85%), VI 60 s (85%), VI 60 s (75%), VI 15 s (95%). The other two subjects received the reverse order of conditions. A session was conducted if the subject
123
weighed within 9 3% of its target body weight (e.g. 72–78% for the 75% conditions). Sessions were conducted daily, 7 days per week (if the subjects met the weight criterion). Maintenance feedings, when necessary, occurred after all subjects had completed their daily sessions.
2.2. Results and discussion Appendix A contains the rates of wheel running and obtained rates of reinforcement for individual subjects during the final five sessions of each condition of the entire study. It shows that unconditioned rates of wheel running increased with food deprivation. Therefore, although the subjects in the present study were older than those typically used in activity-based anorexia research (Pierce and Epling, 1994), changes in the activity rates for the present subjects were similar to those observed for younger animals. Appendix A also shows that reinforced response rates were always well above unconditioned response rates at each level of deprivation for each rate of reinforcement for each subject, indicating that wheel running was indeed reinforced by food pellets. Fig. 1 displays the observed within-session patterns of responding. Presented are the rates of wheel running (in fourths of a wheel turn per min) across successive 5-min intervals of the session. The upper graph shows the results for the conditions in which a VI 15-s schedule was in effect. The lower graph shows the results for the VI 60-s schedule conditions. Each function represents the mean for all subjects responding at 75 (closed squares), 85 (open squares), or 95% (closed triangles) of their free-feeding body weights. Results for each condition represent responding during the final five sessions that each condition was in effect. The upper graph of Fig. 1 shows that, when subjects responded on a VI 15-s schedule, food deprivation effected absolute rates of responding, but not the within-session pattern of responding. Absolute rates of wheel running varied directly with food deprivation and responding decreased within the session in a similar fashion at each level of deprivation. A two-way (deprivation by 5-min interval) repeated-measures analysis of variance
124
J.N. Weatherly et al. / Beha6ioural Processes 46 (1999) 121–129
(ANOVA), conducted on the rates of responding for individual subjects, confirmed these impres-
Fig. 1. Presented are the rates of wheel running across successive 5-min intervals in the session. The upper and lower graphs present the results for the VI 15-s and VI 60-s conditions, respectively. Each function represents the results for the mean for all subjects responding at 75% ( ), 85% ( ), or 95% () of their free-feeding body weights.
sions. The main effect of deprivation was significant [(F(2, 6)= 7.02, PB 0.05], indicating that subjects responded at different absolute rates at different food deprivations. The main effect of 5-min interval [F(9, 27) = 7.88, PB 0.001] was also significant, indicating that the decrease in responding within the session was systematic. Finally, the interaction between deprivation and 5-min interval was not significant [F(18, 54)= 0.99], indicating that the within-session pattern of wheel running did not differ for different food deprivations. Results for this analysis, and for those that follow, were considered significant at PB 0.05. The lower graph of Fig. 1 indicates that food deprivation did alter the within-session pattern of wheel running when subjects responded on a VI 60-s schedule. A steeper within-session decrease in responding was observed when subjects responded at 75% of their free-feeding body weight than at the lesser deprivations. A two-way (deprivation by 5-min interval) repeated-measures ANOVA, conducted on the rates of responding for individual subjects, produced significant main effects of deprivation [F(2, 6)= 40.77, PB 0.001] and 5-min interval [F(9, 27) = 18.66, P B 0.001]. These results indicate that absolute rates of wheel running varied directly with food deprivation and that the within-session decrease in responding was systematic. The interaction between deprivation and 5min interval was also significant [F(18, 54)= 2.41, PB 0.01], indicating that different within-session patterns of wheel running were observed at different levels of food deprivation. The significant interaction term does not necessarily lead to the conclusion that the within-session decrease in responding on the VI 60-s schedule was steeper at subjects’ 75% weight than at their 85 and 95% weights. However, the slopes of response functions shown in the lower graph of Fig. 1 indicated that this was indeed the case. Slopes of the best-fit lines, plotted using the data from the entire session, were − 2.25, − 1.64, and − 1.44 for the response patterns at subjects’ 75, 85, and 95% weight, respectively. The results for the mean of all subjects were also representative of those for individual subjects. Appendix A presents the standard error of
J.N. Weatherly et al. / Beha6ioural Processes 46 (1999) 121–129
the mean of responding within the session for individual subjects, which gives an indication of within-session variance in responding. Standard errors for individual subjects were relatively similar at different deprivations for the VI 15-s schedule. However, the standard errors were larger when subjects were more food deprived than when less food deprived for the VI 60-s schedule, indicating that individual subjects showed larger within-session changes in responding when they were more, rather than less, food deprived.
3. Experiment 2 The results of experiment 1 suggest that withinsession patterns of wheel running sometimes, but not always, decrease more steeply when subjects are more food deprived than when they are less food deprived. This was the case when subjects responded on a VI 60-s schedule. However, when subjects received more food (i.e. responded on a VI 15-s schedule), within-session patterns of wheel running did not differ at different levels of deprivation. Experiment 2 was conducted to determine whether the results for the VI 60-s schedule could be replicated if it provided more food than in experiment 1. Subjects responded on a simple VI 60-s schedule at either 75 or 95% of their freefeeding body weight, across conditions. The size of the reinforcer was either one or four 45-mg food pellets. The reinforcer size of four food pellets was chosen because a VI 60-s schedule delivering four pellets per reinforcer programmed the equal number of reinforcers per session as did the VI 15-s schedule in experiment 1.
3.1. Method 3.1.1. Subjects and apparatus Of the four subjects used in experiment 1, three served. The fourth subject died before completing all the conditions and its data were excluded from the following analyses. Subjects were housed and maintained as in experiment 1. They responded in the same apparatus as described in experiment 1.
125
3.1.2. Procedure The procedure was identical to that of experiment 1 with the following exceptions. First, the first phase of experiment 1 was not repeated, rather subjects were immediately placed on the experimental procedure. Second, reinforcers were always scheduled by a VI 60-s schedule. Third, in some conditions, a response would be reinforced with four 45-mg Noyes pellets rather than just one. Subjects 70 and 71 responded in the following conditions in the following order (body weights appear in parentheses) four pellets (95%), one pellet (95%), four pellets (75%), and one pellet (75%). Condition order was reversed for subject 72. 3.2. Results and discussion Fig. 2 shows that patterns of responding differed with food deprivation, but not with reinforcer size. A steeper within-session decrease in responding was observed when subjects were more food deprived than when they were less food deprived. It presents the rates of wheel running (in fourths of a wheel turn per min) across successive 5-min intervals of the session. Each function represents the mean for all subjects responding at 75 (solid symbols) or 95% (open symbols) of their free-feeding body weight when the reinforcer size was one (triangles) or four (squares) pellets. Results were calculated using the final five sessions that each condition was in effect. The above impressions were confirmed by statistical analyses. A two-way (condition by 5min interval) repeated-measures ANOVA, conducted on responses rates for individual subjects, showed that the main effect of condition [F(3, 6)= 5.48, PB 0.05], main effect of 5-min interval [F(9, 18) = 18.28, PB 0.001], and interaction between condition and 5-min interval [F(27, 54)= 1.80, PB 0.05] were significant. These results indicate that absolute rates of responding differed across conditions, that responding changed systematically within the session, and that different within-session patterns of responding were observed in the different conditions. Because the interaction term was significant, follow-up two-way (reinforcer size by 5-min inter-
126
J.N. Weatherly et al. / Beha6ioural Processes 46 (1999) 121–129
Fig. 2. Presented are the rates of wheel running across successive 5-min intervals in the session. Each function represents the results for the mean for all subjects responding at 75% for four pellets ( ), 75% for one pellet (), 95% for four pellets ( ) or 95% for one pellet (open triangles).
val) repeated-measures ANOVAs were conducted on the data at each level of food deprivation. When subjects responded at 75% of their freefeeding body weight, the main effect of 5-min interval [F(9, 18) =9.99, P B0.001] was significant, but neither the main effect of reinforcer size [F(1, 2) = 1.63] nor the interaction term [F(9, 18)=1.36] reached significance. In other words, similar response patterns were observed for both reinforcer sizes. Similar results were observed when subjects responded at 95% of their free-feeding body weight. Again, the main effect of 5-min interval [F(9, 18) = 57.01, P B0.001] was significant, but neither the main effect of reinforcer size [F(1, 2) =0.04] nor the interaction term [F(9, 18)=1.55] reached significance. Simply because response patterns differed at 75 and 95% body weights does not necessarily mean
that steeper decreases in responding were observed when subjects were more, rather than less, food deprived. However, the slopes of the response functions presented in Fig. 2 support this claim. Slopes of the best-fit lines for the 75% conditions, plotted using the data from the entire session, were − 1.97 when one pellet was delivered and − 2.21 when four pellets were delivered. Slopes for the 95% conditions were − 1.30 when one pellet was delivered and − 1.59 when four pellets were delivered. The results in Appendix A again indicate that these results were representative of those for individual subjects. Standard errors of the mean for responding within the session were almost always larger at subjects’ 75% weight than at their 95% weight, supporting the idea that larger changes in responding were observed when subjects were more, rather than less, food deprived.
J.N. Weatherly et al. / Beha6ioural Processes 46 (1999) 121–129
4. General discussion The present results demonstrate that withinsession changes in wheel running are observed when running serves as the operant response. The present results also indicate that within-session rates of wheel running decrease more steeply when subjects are more food deprived than when they are less food deprived, at least at moderate rates of reinforcement (i.e. VI 60-s schedule). Different patterns of responding were not observed at different levels of deprivation when subjects responded at a high rate of reinforcement (i.e. VI 15-s schedule). Why similar within-session patterns were observed at the high, but not at the moderate, rate of reinforcement is not known, but differences in the amount of food delivered by the two schedules can be ruled out as a cause. The VI 60-s schedule, four-pellet conditions of experiment 2 delivered the equivalent number of food pellets as did the VI 15-s schedule conditions of experiment 1. However, the results of experiment 2 showed that response patterns varied with level of food deprivation and not the number of food pellets delivered. It is therefore likely that the difference in results is tied to the frequency of reinforcer delivery, not to the amount of reinforcement delivered. One possible explanation for the difference between the two schedules is the presence of competing responses. That is, competing responses (e.g. interim behaviors; Staddon and Simmelhag, 1971) may have occurred at a higher rate when subjects were more food deprived than when they were less food deprived, thus producing different response patterns. However, for the VI 15-s schedule, the inter-reinforcement interval may have been too short to allow for differences in these interim behaviors. This potential explanation must remain speculative, however, because subjects were not observed while responding. Because amount of reinforcement did not alter response patterns in experiment 2, one could argue that the present results are incompatible with some existing evidence on within-session patterns of responding. For instance, both Palya
127
and Walter (1997) and Bizo et al. (1998) report different within-session patterns of responding with different reinforcer amounts. However, the discrepancy is potentially the product of differences in procedure. Whereas both Palya and Walter and Bizo et al. used pigeons and varied reinforcer amount by varying hopper duration, the present study used rats and varied reinforcer amount by delivering a different number of food pellets. Roll et al. (1995) studied rats and reported results similar to those reported here. Furthermore, although Palya and Walter and Bizo et al. took great care to correlate hopper duration to amount eaten, reinforcer amount may have varied from reinforcer to reinforcer. Such was not the case in the present study. As noted above, the leading explanations for within-session changes in responding (i.e. satiation and habituation) are not adequately defined to allow for a critical test between them. Therefore, the present results cannot favor one explanation over the other. However, the results would seem to pose some difficulty for each explanation. For instance, finding steeper decreases in responding for subjects that are more food deprived and responding for less food than for subjects that are less food deprived and responding for more food would appear to run contrary to an intuitive interpretation of satiation. On the other hand, one could argue that finding seemingly steeper decreases in responding at lower, than at higher, rates of reinforcement (Fig. 1) runs contrary to what one would predict according to habituation. Finally, the present results may suggest avenues of future research in the area of activitybased anorexia. For instance, the present results suggest that decreases in operant responding for food depends more on the frequency of food delivery than on the amount of food delivered. The present results may therefore suggest that anorexia may be overcome more easily if food is presented infrequently in large quantities rather than frequently in small quantities. Clearly, the present results cannot be directly extended to anorexia nervosa. They do, however, indicate that the study of within-session changes in responding may have far-reaching implications.
J.N. Weatherly et al. / Beha6ioural Processes 46 (1999) 121–129
128
Acknowledgements
Experiment 1: VI 60-s schedule
Partial funding for this research was provided by the Shearman Endowed Professorship in Education awarded to C.L.M. The authors thank the Department of Biological and Environmental Sciences at MSU for sharing laboratory space, Hanna C. Rue, Jason E. Stout, Tiffani Thibodeaux, and Carolyn S. Davis for their help in collecting the data, and Frances K. McSweeney for her comments on an earlier version of this manuscript.
70
71
72
73 Appendix A Presented are the rates of wheel running (in fourths of a wheel turn per min; R) and obtained reinforcers (OR) for each subject in each condition of the study. Also presented are the standard errors of the means (S.E.M.), calculated on the mean rates of responding per 5-min interval across the session. The S.E.M.s therefore provide a measure of the change in responding within the session for each subject.
Unconditioned
R R R R
85%
1.7 2.3 2.3 3.2
1.5 0.9 1.3 2.5
95% 0.8 0.6 1.5 1.2
Experiment 1: VI 15-s schedule 70
71
72
73
R S.E.M. OR R S.E.M. OR R S.E.M. OR R S.E.M. OR
16.3 2.1 38.8 13.2 1.3 40.4 21.5 3.3 51.6 15.6 2.9 47.8
10.6 0.7 48.0 10.8 2.0 45.4 18.3 3.0 50.4 10.1 1.5 49.0
6.0 1.1 40.4 6.3 1.6 44.2 10.2 2.4 47.6 7.6 1.5 46.2
Experiment 2: VI 60-s schedule, one pellet 70
71
72
R S.E.M. OR R S.E.M. OR R S.E.M. OR
19.6 2.5 42.8 8.2 0.9 41.8 27.7 3.0 44.0
6.6 0.9 41.0 8.6 1.4 44.2 8.1 1.6 38.8
Experiment 2: VI 60-s schedule, four pellets 75%
70 71 72 73
R S.E.M. OR R S.E.M. OR R S.E.M. OR R S.E.M. OR
22.8 1.1 156.0 7.8 0.5 155.8 32.5 4.3 171.4 20.5 2.6 170.4
16.6 1.7 144.6 9.4 0.8 160.4 24.1 4.0 169.6 12.0 1.7 150.0
70
71
72 7.9 1.3 136.0 6.7 0.9 143.8 12.3 2.4 152.0 10.4 1.7 157.2
R S.E.M. OR R S.E.M. OR R S.E.M. OR
21.7 2.0 44.0 12.1 1.6 46.8 26.8 3.4 42.6
6.6 1.7 37.0 7.1 1.4 46.4 9.2 2.1 37.2
References Bizo, L.A., Bogdanov, S.V., Killeen, P.R., 1998. Satiation causes within-session decreases in instrumental responding. J. Exp. Psych. Anim. Behav. Process. 24, 439 – 452. McSweeney, F.K., Hatfield, J., Allen, T.M., 1990. Withinsession responding as a function of post-session feedings. Behav. Process. 22, 177 – 186.
J.N. Weatherly et al. / Beha6ioural Processes 46 (1999) 121–129 McSweeney, F.K., Hinson, J.M., Cannon, C.B., 1996. Sensitization-habituation may occur during operant conditioning. Psychopharmacol. Bull. 120, 256–271. Palya, W.L., Walter, D.E., 1997. Rate of a maintained operant as a function of temporal position within a session. Anim. Learn. Behav. 25, 291–300. Pierce, W.D., Epling, W.F., 1994. Activity anorexia: an interplay between basic and applied behavior analysis. Behav. Anal. 17, 7 – 23.
.
129
Pierce, W.D., Epling, W.F., Boer, D.P., 1986. Deprivation and satiation: the interrelations between food and wheel running. J. Exp. Anal. Behav. 46, 199 – 210. Roll, J.M., McSweeney, F.K., Johnson, K.S., Weatherly, J.N., 1995. Satiety contributes little to within-session decreases in responding. Learn. Motiv. 26, 323 – 341. Staddon, J.E.R., Simmelhag, V.L., 1971. The ‘superstition’ experiment: a reexamination of its implications for the principles of adaptive behavior. Psych. Rev. 78, 3 – 43.