Effects of long- and variable-duration signals for food on activity, instrumental responding, and eating

LEARNING

AND

MOTIVATION

11,

164-184

(1980)

Effects of Long- and Variable-Duration Signals for Food on Activity, Instrumental Responding, and Eating PETER F. LOVIBOND University

of New

South

Wales

Incentive-motivation theories typically assume that the conditioning of appetitive motivation involves the same parameters as Pavlovian salivary conditioning. In contrast, the Soltysik-Konorski model asserts that drive is inhibited by stimuli closely associated with food (salivary CSs) and augmented by stimuli more loosely associated with food (long and variable CS-US interval). Experiment 1 examined this latter proposition. Sixty-four rats were given extensive exposure to each of four environmental CSs, two while hungry and two while satiated. Within each deprivation condition, food was given 30-300 set after placement of the rats in one environment, and was not given in the other environment. Performance on three separate measures-activity, lever-pressing, and food consumption-was higher in the environments previously associated with food. Experiment 2 examined the effects of discrete stimuli presented in advance of eating; in accord with the results of Experiment 1, food consumption was greater after a stimulus (1- to 9-min duration) previously paired with food than after no stimulus or after a stimulus unpaired with food. The overall results indicate (a) that stimuli associated with food become capable of facilitating a variety of food-directed behaviors, possibly via the conditioning of a common appetitive system, and (b) that a close association between the stimuli and food is not essential for such conditioning to occur.

Incentive-motivation theories maintain that a CS associated with a food US acquires the capacity to motivate food-directed behavior via a mediating process of classical conditioning. These theories differ in a variety of ways, for instance as to whether mediation is achieved via overt CRs (e.g., Spence, 1956) or a central motivational state (e.g., Bindra, 1968). However, they share the common assumptions (a) that mediation inThe author wishes to thank Professor S. H. Lovibond, who suggested the line of research reported in this paper, and Dr. R. F. Westbrook, who provided helpful comments on an earlier version of the paper. This research was conducted while the author held an Australian Commonwealth postgraduate research scholarship. Requests for reprints should be sent to Peter F. Lovibond, School of Psychology, University of New South Wales, P.O. Box 1, Kensington, N.S.W. 2033, Australia. 164 0023-9690/80/020164-21$02.00/0 Copyright AU rights

@ 1980 by Academic Ress, Inc. of reproduction in any form reserved.

SIGNALS

FOR

FOOD

165

volves the transfer of unconditioned motivational properties from the US to the CS, and (b) that motivational conditioning follows the same laws as the conditioning of overt CRs such as salivation (e.g., Rescorla & Solomon, 1967, p. 172). Thus, most incentive studies have used relatively short CS-US intervals, usually in the salivary conditioning range: 10 set or less. In a recent review of incentive theory, however, Mackintosh (1974, pp. 224-231) has concluded that the majority of evidence is inconsistent with the notion that salivation, or a central state indexed by salivation, motivates instrumental responding for food. For instance, transfer studies have typically shown suppression of instrumental performance by CSs that would be expected to elicit salivary CRs (e.g., Azrin & Hake, 1969). On the other hand, there is some evidence that the expected incentivemotivational effects can be obtained at longer CS-US intervals. Those transfer studies which have demonstrated facilitation of responding have used CS-US intervals of 1 to 2 min (e.g., Estes, 1943, 1948). When CS-US interval has been explicitly manipulated within a transfer design (Meltzer & Brahlek, 1970; Meltzer & Hamm, 1978; Miczek & Grossman, 1971), it has been found that suppression of responding occurs with a short CS-US interval, whereas facilitation of responding appears to depend on the use of a long CS-US interval. Furthermore, food-conditioned increases in activity, often interpreted as reflecting elevated motivation, occur reliably when the CS precedes food by several minutes (Sheffield & Campbell, 1954). There is also some evidence that this effect does not appear at very short CS-US intervals (Longo, Klempay, & Bitterman, 1964). One theory which appears to be capable of accounting for these effects of CS-US interval is the model developed by Soltysik (1975) and Konorski (1967). These theorists identify appetitive motivation with the hunger drive, and consequently assume that food has an inhibitory, rather than an excitatory, effect on appetitive motivation. Stimuli which are closely associated with food (short, constant CS-US interval) are assumed to become conditioned primarily to the food center, and thereby to elicit consummatory responses and inhibit drive (consummatory conditioning). Stimuli which are more loosely associated with food (long, variable CS-US interval) are assumed to become conditioned to the hunger drive center and thereby to facilitate learned and innate foodseeking behaviors (preparatory or drive conditioning). The separation of consummatory behavior and appetitive motivation allows the Soltysik-Konorski model to provide an account of the divergent effects of short and long duration food signals on activity and instrumental responding. The present studies were designed to evaluate the hypothesised drive-inducing properties of stimuli preceding food by a long and variable CS-US interval. Although based on the Soltysik-Konorski

166

PETER

F. LOVIBOND

notion of preparatory conditioning, such an evaluation is also relevant to the general class of theories which argue that the effects of stimuli paired with food are mediated by a central motivational system. Bindra (1968, p. 10) has pointed out that the case for a central appetitive motivational state would be strengthened if a single CS could be shown to exert influences on several different aspects of behavior. Experiment 1 attempted to provide such a set of converging measures by examining whether stimuli loosely associated with food could augment a variety of food-related behaviors-general activity, an explicitly trained instrumental response, and eating. EXPERIMENT

1

Four distinct environments were constructed to serve as CSs in a 2 x 2 factorial, repeated measures design, with food as the US. Experience with the four environments differed according to: (a) .whether the subject was hungry or not at the time, and (b) whether food was delivered in that environment or not. The CSs will be referred to as H. F, H. NF, NH. F, and NH. NF (H, hungry; NH, not hungry; F, fed; NF, not fed). When food was delivered, it followed placement of the subject in the environment (CS) by an interval which ranged from 30 to 300 sec. This range of CS-US intervals was chosen in order to provide conditions which, according to the Soltysik-Konorski model, should allow the development of preparatory (drive) conditioning but should minimize the development of consummatory conditioning. Three measures of appetitive motivation were employed: general activity, instrumental response rate, and food consumption. The predictions of the model in relation to the four environmental CSs are as follows. Environment H. F should acquire conditioned drive properties due to its loose association (long and variable CS-US interval) with food. Environment NH. F may acquire conditioned drive properties, depending on whether a high level of hunger drive is necessary for preparatory conditioning to occur (cf. Konorski, 1967, p. 275). Environment NH. NF should remain relatively neutral, since it is associated with neither food nor hunger. Environment H. NF should become capable of inhibiting drive, since it predicts a lack of food while the hunger drive system is aroused (Konorski, 1967, pp. 325-326). The hypothesized ordering of the CSs, in terms of their conditioned motivational properties, is therefore H. F, NH. F, NH. NF, and H. NF. Method

Subjects The subjects were 64 experimentally naive male Sprague-Dawley rats, approximately 35 days old at the beginning of the experiment. They were housed in individual wire cages with free access to water, and had been

SIGNALS

adapted to a 23-hr food deprivation course of the experiment.

FOR

FOOD

167

schedule. Eight rats died during the

Apparatus

The main apparatus consisted of 64 boxes which served as environmental CSs during classical conditioning. The boxes were of four distinct types, the 16 boxes of each type being housed in separate rooms. The Type 1,2,3, and 4 boxes differed in terms of their visual properties (grey, black, vertically striped, or horizontally striped walls), tactile properties (flat aluminium, sawdust, mesh, or sand floors), and olfactory properties (eucalyptus, peppermint, aniseed, or musk odor). All boxes were constructed from 12-mm particle board and measured 23 x 20 x 20 cm. The boxes were arranged in rows of eight, each unit of eight boxes having a common 195cm Perspex lid. Each box contained a cylindrical copper food cup, 5 cm high and 5 cm in diameter, mounted on the rear wall 2 cm above the floor and 5 cm from the left hand wall. Above each cup was a 6-cm-diameter, 5-cm-deep tin containing foam plastic which, when lowered, sealed and prevented access to the cup. Each tin was supported by an l&cm-long, 6-mm brass rod which passed through a rubber grommet in the Perspex lid above. The brass rods from the eight boxes in each unit were all connected to a horizontal 170-cm length of T-section ahrminium, so that the eight tins could be raised and lowered together. Within each conditioning room, the two units of eight boxes each were placed on parallel tables 60 cm apart such that both aluminium bars could be lifted together, thus allowing all 16 rats access to food at the same moment without any warning signal. In order to measure instrumental responding and activity, the 16 Type 1 boxes had provision for mounting response levers and were fitted with activity-sensitive floors. When mounted, the Gerbrands response levers were located on the rear wall, 7 cm from the right hand wall and 7 cm above the floor. The aluminium tray of each Type 1 box rested on a flat aluminium plate which was supported by four pieces of lO-mm foam plastic. A stereo ceramic phonograph cartridge, mounted in one end of a IO-cm aluminium arm, rested on the plate (as if playing a record). Output from the cartridge, in response to vibrations of the plate, was sent to a summation circuit which advanced a counter when the stored voltage reached a given threshold. A further piece of apparatus consisted of six standard Gerbrands experimental chambers, housed in a fifth room. The 23-cm side walls were constructed from Perspex, and the 20-cm front and rear walls from aluminium. On the rear wall of each box a Gerbrands response lever was mounted, 7 cm above the floor and 7 cm from the right hand wall. A food hopper was located 7 cm to the left of the lever, 1 cm above the floor. A

168

PETER F. LOVIBOND

food dispenser connected to the hopper allowed presentation Noyes reinforcement pellets.

of 45mg

Procedure Classical conditioning. A completely within-subjects design was used, with all rats being exposed equally to all four CSs during the course of training. Thus, each rat had experience with two particular boxes only when hungry, and two further boxes only when not hungry. In one of the “hungry” boxes, food was always presented (H. F), whereas in the other it was never presented (H. NF). Similarly, in one of the “not hungry” boxes, food was always presented (NH. F), whereas in the other it was not (NH. NF). The specific type of box which served as a particular CS was counterbalanced across rats, such that each type (1,2,3, or 4) served as an H. F stimulus for 16 rats, H. NF for 16 rats, NH. F for 16 rats, and NH. NF for 16 rats. Experience with the four CSs was regulated by the use of four different types of daily schedule, shown in Table 1. Each block of four conditioning days included each type of schedule once, with the order randomized. In all cases, the same type of trial was given to all rats at the same time. The combination of schedules shown in Table 1 was chosen for the following reasons: (a) it provided equal experience with each CS, (b) it ensured that, averaged across days, deprivation levels for the two hungry conditions, H. F and H. NF, were equal (20 hr, range 16-24), and that deprivation levels for the two not hungry conditions, NH. F and NH. NF, were equal (45 min, range 30-60), and (c) it ensured that, within each deprivation condition, no extraneous cues such as time of day provided any information as to the nature of the next trial. Schedules 2 and 4 involved one 60-min feeding period per day; schedules 1 and 3 involved two 60-min feeding periods. The average daily access to food was therefore 90 min. Each conditioning trial involved placing the 64 rats in the conditioning boxes for 60 to 65 min. After the 16 rats assigned to a particular type of box had been loaded, they were presented with the odor characteristic of TABLE 1 The Four Types of Daily Schedule Used During Classical Conditioning Trial 1 Schedule Schedule Schedule Schedule Note.

min.

1 2 3 4

H.NF H.NF H.F H.F

Trial 2 -

a a b b

-

HCF HCF NH.F NH.NF

Trial 3 -

b b -

NH.F NH.NF

HCF = 60-min feeding in home cage; a = delay of 120-180 min; b = delay of 30-60

SIGNALS

FOR

FOOD

169

those boxes by wiping the underside of the lids with cotton soaked in the appropriate oil. A delay which constituted the CS-US interval was then interposed. If food was scheduled for that trial, it was presented by lifting the sealing tins at the end of the CS-US interval, which was varied randomly across trials between 30 and 300 sec. Each food cup contained approximately 40 g of Allied food pellets, more than a single rat could eat in 60 min. Food was placed in the cups on both feeding and nonfeeding trials, so that any food odor which escaped the sealing tins would not predict delivery of food. For the same reason, the Experimenter remained in the room during the CS-US interval whether the tins were to be lifted or not. The rats were left in the boxes (with or without food) for a further 60 min before being returned to their home cages. On feeding trials (H. F and NH. F), the procedure was therefore one of delay conditioning, with a CS-US interval of 30-300 set and full overlap of the US by the CS. General procedure. The 146 experimental days may be divided into eight phases. Phase 1 (Days l-56) consisted of 56 days of classical conditioning, as described above. Phase 2 (Days 57-71) was the instrumental training period. The rats were given a 40-min training period in the Skinner boxes each day, followed by 60-min of access to food in their home cages. Each rat was initially trained on a continuous reinforcement schedule (Days 57-62) and was then switched to a variable interval schedue. The mean interval was increased from 10 to 60 set over Days 63-66, and maintained at 60 set on Days 67-71. Phase 3 (Days 72-83) consisted of 12 more days of classical conditioning. Phase 4 (Days 84-98) was the instrumental test period, in which the levers used in the Skinner boxes for instrumental training were installed in the classical conditioning boxes. This is a variant of the transfer of control paradigm (Trapold & Overmier, 1972). Since only the Type 1 boxes had facilities for mounting levers, the design for this measure reduced to four independent groups of 16 rats each. The counterbalancing ensured that the Type 1 boxes served as an H. F stimulus for 16 rats, H. NF for 16 rats, NH. F for 16 rats, and NH. NF for 16 rats. The groups defined by the role of the Type 1 boxes will be referred to as the H. F, H. NF, NH. F, and NH. NF groups. Testing was carried out at an intermediate deprivation level of 12 hr. On each of 3 test days, the rats were placed in the Type 1 boxes, as if for a normal trial, for 5 min (the longest CS-US interval used during training). During this time, lever-press responses, which were under extinction, were recorded. The rats were then returned to their home cages without being fed. They were later given 60-min of access to food in the home cages, 12 hr prior to the next day’s test. After these initial data had been collected, the instrumental test period was extended since observation of the rats during testing indicated that the food delivery apparatus may have

170

PETER

F. LOVIBOND

been having a distracting effect on the instrumental performance, especially in groups H. F and NH. F. The rats were first retrained on the lever-press response (Days 87-91), and given four more days of classical conditioning (Days 92-95). A further instrumental test period was then conducted (Days 96-98). The conditions were identical to the first test, except that the entire food delivery mechanism (food cups, sealing tins, and brass rods) was removed. Phase 5 (Days 99-118) consisted of a further 20 days of classical conditioning. Phase 6 (Days 119-126) was the activity test period. The procedure was similar to the instrumental tests, except that deprivation level was included as a factor. Once again, since activity measurement was possible only in the Type 1 boxes, the repeated measures design reduced to an independent groups design, with four groups defined by the role of the Type 1 boxes. On each of the 8 test days, all rats were placed in the Type 1 boxes for 5 min, during which time activity rates were recorded. On the odd days, the rats were tested at the average “not hungry” deprivation level (45 min), whereas on the even days they were tested at the average “hungry” deprivation level (20 hr). Phase 7 (Days 127-134) consisted of a final 8 days of classical conditioning. During the whole experiment, the rats received a total of 100 days of classical conditioning (200 trials: 50 exposures to each CS). Phase 8 (Days 135-146) was the food consumption test period. All rats were tested three times in each of the four types of conditioning box (fully repeated design). Deprivation level for all 12 tests was 2 hr. The CS-US interval was held constant at 5 min, and food was delivered regardless of condition. After the full 60 min had elapsed, the rats were removed and the uneaten food weighed. Statistics Three orthogonal sets of planned contrasts were written to analyze the data from the three dependent variables. In all cases, the basic contrasts of interest were the same. They tested (a) the main effect for feeding (H. F + NH. F vs NH. NF + H. NF), (b) the difference between the two “fed” conditions (H. F vs NH. F), ‘and (c) the difference between the two “not fed” conditions (NH. NF vs H. NF). The contrasts were tested using the technique described by Hays (1969, pp. 464-466), but using (~/3 rather than (Y, in order to control for the inflation of the Type 1 error rate which occurs when multiple univariate analyses are carried out on multivariate data (Bird, 1975). Repeated measures contrasts were evaluated using a multivariate model (Poor, 1973). Sample sizes at the time of testing were as follows: Lever pressing, N=62; Activity, N=60; Food consumption, N=56. A rejection criterion of b G .05 was used throughout this paper.

SIGNALS

171

FOR FOOD

Results

Lever-Pressing The left-hand panel of Fig. 1 presents the number of lever-press responses emitted during the initial test period, averaged across Subjects and days. Contrary to prediction, the response rates for the rats normally fed in the Type 1 boxes (H. F and NH. F) were lower than for the “not fed” groups (NH. NF and H. NF). Observation of the rats indicated that the “fed” groups spent a great deal of time biting and investigating the food cups and sealing tins used to deliver food during classical conditioning. When retested with the food delivery apparatus removed, the response rates were as shown in the right-hand panel of Fig. 1. This time the “fed” groups responded more than the “not fed” groups. The relevant interaction contrast showed that this reversal was significant [F(l, 58) = 10.61. In both test periods, the difference between the two “fed” groups, H. F and NH. F, was in the predicted direction but was not significant [F( 1, 58) = 1.31. There was, however, a significant inhibitory effect. Group H. NF responded consistently less than group NH. NF [F( 1,58) =

6.4-j. Activity The activity levels recorded during the hungry and not hungry tests, averaged across Subjects and days, are shown in Fig. 2. Under both

I I I I-

-

I

lo-

l4.F WITH

NH.F

NH.NF FOOD

H.NF CUPS

H.F WITHOUT

NH.F

NH.NF FOOD

H.NF CUPS

FIG. 1. Mean number of lever-press responses for the four groups during the S-min instrumental tests. The left panel refers to testing with the food delivery apparatus in place; the right panel refers to subsequent testing with the apparatus removed. All tests were carried out at 12 hr of deprivation.

172

PETER F. LOVIBOND

-

0

H.F

-

-

NH.F

NH.NF

HUNGRY

H.NF

-

-

H.F

NH.F NOT

NH.NF

H.NF

HUNGRY

2. Mean activity levels for the four groups during the 5-min activity tests. The left panel refers to testing at 20 hr of deprivation; the right panel refers to testing at 45 min of deprivation. FIG.

deprivation conditions, the two “fed” groups were far more active than the two “not fed” groups [F(l, 56) = 57.73. The “fed” groups were considerably less active when tested at low deprivation, whereas the activity of the “not fed” groups hardly changed as a function of test deprivation level. This finding was supported statistically by the main effect for test deprivation [F( 1, 56) = 21.73, which interacted with the fed-not fed comparison [interaction F(1, 56) = 9.81. The ordering of the “not fed” groups, NH. NF and H. NF, was as predicted, but the difference was not significant [F( 1, 56) = 1.31. The relationship between the “fed” groups, H. F and NH. F, varied as a function of test deprivation. When tested hungry, the H. F group was most active, but when tested not hungry, the NH. F group was most active [interaction F(1, 56) = 14.91. Food Consumption Figure 3 presents the mean amount of food consumed by all rats under the four CS conditions. The rats ate, on average, 11.5% more in the boxes previously associated with food, H. F and NH. F, than in the boxes unassociated with food, NH. NF and H. NF. This difference, although small (13.6 vs 12.2 g), was highly significant [F( 1,55) = 26.11. There were no other reliable differences.

SIGNALS

FOR FOOD

NH.F CS

NH.NF

173

H.NF

TYPE

FIG. 3. Mean amount of food consumed by all subjects during 60-min exposures to the four types of conditioning box at 2 hr of deprivation.

Discussion Lever-Pressing

When the lever was the only salient object in the environment (food cups absent), the rats emitted more lever-press responses if that environment had been previously associated with food than if it had not. This result is in accordance with previous studies which have shown that a long duration CS for food can facilitate performance of an instrumental response undergoing extinction (Estes, 1943, 1948; Morse & Skinner, 1958). In the past, such results have been taken as support for incentive theories, which assert that the CS, by virtue of its association with food, has acquired the capacity to intensify the motivational state underlying instrumental responding (e.g., Bindra, 1968, p. 15). There have also been a number of nonmotivational explanations, in terms of disinhibition (Baum & Gleitman, 1967), or a change in the expectation of food (Mackintosh, 1974, p. 226). Such accounts do not appear to apply to the present data, since the background cues from instrumental training were absent, and the CS was continually present throughout extinction. Another explanation of the present results is that the introduction of the lever into the test environment was more novel, and hence more disruptive, for the groups which were not expecting food (H. NF and NH. NF). However, a mildly novel event in a nonaversive context may be expected to arouse approach and manipulation rather than cause disruption (Berlyne, 1960). In this circumstance, differential novelty would have worked against the obtained result of more lever-

174

PETER

F. LOVIBOND

pressing in the “fed” groups. Finally, the facilitation of instrumental responding during transfer tests may be attributed to the frustrative effects of failing to reward the CS (Amsel, 1972). In the present study, testing was not extended past the end of the CS (5 min). However, since the average CS-US interval during training was less than 5 min, frustration effects may have occurred in the later portions of the test period. When testing was carried out with the food cups present, the ordering of the “fed” and “not fed” groups was reversed. One explanation of why the “fed” groups concentrated on the food cups to the detriment of lever-pressing is that the environmental cues had become discriminative stimuli for the incompatible instrumental response of approaching the food cups. Such a response may have been reinforced by faster access to food when the sealing tins were lifted. However, this account predicts that the rats would learn to go to the food cups and wait there, whereas in fact the rats climbed all over the sealing tins and were often lifted up with them, thus delaying access to the food. A more likely explanation is that the rats were exhibiting “sign-tracking” (Hearst & Jenkins, 1974) to the food delivery apparatus, as a result of its close spatial contiguity with food. There is good evidence (LoLordo, McMillan, & Riley, 1974; Schwartz, 1976) that sign-tracking to a cue for food can interfere with performance of a spatially separated instrumental response. In the present experiment, the environments associated with food, H. F and NH. F, may be thought of as completing the compound stimulus necessary for signtracking to occur, since the food delivery apparatus only predicted food in these “fed” environments. From this point of view, the biting and investigation of the food tins may be described as “goal-tracking” in response to the environmental cues signaling food (Boakes, 1977). A final point of interest in the lever-pressing data concerns the significantly lower performance of group H. NF as compared with group NH. NF, during both test periods. If group NH. NF is taken as a neutral reference, then the environmental cues for the H. NF group had become capable of inhibiting instrumental responding. This finding corresponds to Cell 2 of Rescorla and Solomon’s table of classical-instrumental interactions, and supports their view that appetitive inhibitors suppress appetitive instrumental responding (Rescorla 8z Solomon, 1967, pp. 172-174). Although the inhibitory effect of the H. NF environment is consistent with the motivational account of the lever-pressing data, it is incompatible with the novelty and frustration accounts, since the expectation of food in the H. NF and NH. NF environments was the same. Activity

The major finding in the activity data, that the rats were far more active if the test environment signaled food delivery, is consistent with previous research showing that activity conditioning occurs reliably at long CS-US

SIGNALS

FOR

FOOD

175

intervals (Sheffield & Campbell, 1954). Observation of the rats in the “fed” groups indicated that their activity was directed primarily toward the food cups and sealing tins, as it was during lever-press testing with the food cups present. Although this tendency has not been reported in previous studies which have measured activity in the presence of the food-delivery apparatus (Amsel & Work, 1961; Sheffield & Campbell, 1954; Zamble, 1967), it opens the possibility that a mechanism such as goal-tracking may be operating in such situations. According to an associative account of goal-tracking, activity directed toward the food cup resulted from approach behavior to the localizable component of the food-paired compound stimulus consisting of the “fed” environment and the food cup. On the other hand, the various motivational accounts of previous activity conditioning studies may also be applied to the more directed activity observed in the present experiment. For instance, excitement arising from the anticipation of food may have acted to energize approach behavior elicited by the food cup. Bindra (1978) has argued that appetitive motivation leads to undifferentiated exploratory activity only in the absence of associative cues for more specific appetitive behaviors such as sign-tracking. In comparison to the fed-not fed difference in activity, the influence of deprivation level during training was rather small. Averaged across the two test deprivation levels, there was almost no difference between the two “fed” groups. The observed superiority of group H. F during “hungry” tests and of group NH. F during “not hungry” tests may be attributed to generalization decrement affecting each group when it was tested at a different deprivation level than that used during training. In the case of the “not fed” groups, the difference favored the NH. NF group, but, unlike the lever-pressing difference, was not significant. The fact that a high level of deprivation during testing acted to selectively increase the activity of the two “fed” groups provides support for the proposition that deprivation level per se does not affect appetitive activity. Rather, as Campbell and Sheffield (1953) have argued, it appears to influence the degree to which external stimuli can provoke activity. In the present case, however, it was the conditioned, not the unconditioned, capacity of stimuli which was altered by deprivation level. This finding supports the extension of the Campbell and Sheffield position suggested by Bindra and Palfai (1967). According to their view, deprivation acts to increase “motor readiness,” which can then “facilitate the motor outflow of whatever response tendencies are evoked by the prevailing central state and environmental stimuli” (Bindra, 1968, p. 8). Food Consumption The rats in this experiment ate more in boxes previously associated with food than in boxes which were equally familiar but had not been associated with food. A similar finding was reported by Zamble (1973),

176

PETER

F. LOVIBOND

who found that a long duration (15 min) signal for food increased subsequent food consumption. If amount of food eaten is taken to be a reliable measure of hunger, a straightforward account of the present results is that the cues of the “fed” environments acted to increase hunger. The results are also compatible with theories which argue that food-related stimuli arouse a central incentive-motivational state, since enhanced appetitive motivation of this type may also be assumed to facilitate eating. An alternative account of the food consumption results is that the unexpected delivery of food in the “not fed” environments, where food had never been presented before, disrupted the animals’ behavior and delayed the commencement of eating. It is not possible to discriminate between this account and the motivational account in terms of the latency to commence eating, because one of the mechanisms by which heightened motivation may facilitate eating is by reducing the latency to eat. That is, arousal of appetitive motivation by the cues of the “fed” environments during the 5 min CS-US interval may have “prepared” the rats to engage in consummatory behavior (Konorski, 1967). Some support for this possibility comes from Zamble’s (1973) study, in which food consumption was increased by a stimulus which preceded food delivery but which did not persist throughout the feeding period. Experiment 2 employed a variant of Zamble’s design in order to further examine the effects of food-paired stimuli presented before eating, with more control over the novelty or surprise generated by food presentation during testing. EXPERIMENT

2

In the two experiments reported by Zamble (1973), one group of rats received a 15- to 16-min stimulus change before their daily 30-min period of access to food; a control group received the stimulus change at another time. In both experiments, the group receiving the warning signal before feeding ate more and gained more weight than the control group. The present study employed a repeated measures design in which all rats were trained and tested with and without a warning signal (CS,). An additional unpaired stimulus, CSu, was included as a control for the unconditioned influence of stimulus change prior to feeding. A further change from Zamble’s procedure was the use of a 22.5-hr, rather than a 23.5-hr, deprivation schedule. It was hoped that the longer period of access to food would reduce the rate of feeding so that it could be more easily influenced by the conditioned stimuli. Method Subjects

The Subjects were 32 experimentally naive male Sprague-Dawley rats, approximately 50 days old at the beginning of the experiment. Three rats died during the course of the experiment.

SIGNALS

FOR FOOD

177

Apparatus

The 64 conditioning boxes described in Experiment 1 were used as the housing and testing apparatus for Experiment 2. The various floor materials of these boxes were replaced by sawdust, and no olfactory stimuli were used. The boxes were located in two rooms which were acoustically isolated from each other; within each room 16 boxes served as home boxes and the remaining 16 at temporary holding boxes. The two rows of eight home boxes were placed on parallel tables 60 cm apart so that the 16 sealing tins could be lifted together to present food to all rats simultaneously. The holding boxes were placed on the same tables, immediately behind the home boxes. In each conditioning room, a 240-V, 25-W light bulb located at bench height provided a low level of indirect illumination throughout the experiment. A 10-W loudspeaker on the floor was used to present white noise. A further 25-W light and 10-W speaker were mounted on the ceiling immediately above the boxes to provide visual and auditory stimuli. Procedure Adaptation (Days. l-13). The 32 rats were weighed and divided into two matched subgroups of 16 rats each, which were housed in the two conditioning rooms. After a l-week period of adaptation to the boxes on ad lib. food, a 22.5hr deprivation schedule was gradually introduced. Daily access to food was reduced from 4 hr on Day 8 to 90 min on Day 13, and was maintained at 90 min for the duration of the experiment. Onset of the 90-min feeding period was varied randomly between 9 AM and 4:30 PM, to prevent temporal cues or regular laboratory noises from providing information about the presentation of food. Deprivation level at feeding time thus varied between 15 and 30 hr. Food was delivered via the same mechanism as used in Experiment 1. Approximately 40 g of Allied food pellets were placed in each food cup and were covered throughout the nonfeeding period by the sealing tins. The aluminium lifting bars were heavily weighted to prevent the rats from lifting the sealing tins. At the onset of the scheduled feeding period, the Experimenter entered the room and lifted the sealing tins to present the 16 rats in that room with food with as little warning as possible. Fifteen minutes before the end of the feeding period, the rats were transferred to the holding boxes while the uneaten food from the home boxes was removed and the food cups were refilled. Throughout the experiment, food and water were always available in the holding boxes; water was always available in the home boxes. White noise was presented continuously to provide a noise level of approximately 70 db(A) in the boxes. Training (Days 24-133). During training the 90-min feeding schedule was continued and in addition the rats were exposed to two stimuli: CSP, a stimulus paired with food, and CSU, a stimulus unpaired with food. Within

178

PETER

F. LOVIBOND

each block of 6 days, there were three presentations of each stimulus. On three randomly selected days, CSr was presented immediately in advance of the scheduled feeding period, terminating 5 set after food delivery. On an independent set of three randomly selected days, CSU was presented at a random time between 9 AM and 4: 30 PM, with the restriction that it could not occur within 30 min of the feeding session. The durations of CSr and CSU were varied randomly between 1 and 9 min, with a mean of 5 min. For the subgroup in Room 1, CSr consisted of a 300-Hz square wave tone at a level of approximately 4 db(A) above the white noise, while CSu consisted of a l-Hz flashing of the stimulus light (0.5 set on, 0.5 set off). For the subgroup in Room 2, these stimuli were reversed. The schedule of three CSp and three CSU presentations per 6-day block ensured that: (a) half of the feeding periods were signaled by CSr and half were unsignaled, and (b) exposure to CSr and CSU was equal. Over the 120 days of training, the rats received 60 presentations of C& and 60 of CS”. Testing (Days 134-145). The effect of CSr and CSu on food consumption was assessed by presenting these stimuli in advance of the normal daily 90-min feeding period. Since any effects were expected to be shortlasting, food consumption was measured for only the first 15 min of the feeding period. After 15 min had elapsed, the rats were transferred to the holding boxes for the remaining 75 min access to food. During this time, the uneaten food from the home boxes was weighed and a preweighed quantity of food (lo-15 g) was placed in the food cups for the next day’s trial. The test period consisted of four blocks of 3 days each. Within each 3-day block, food delivery was preceded on 1 day by CSp, on 1 day by CSn, and on the remaining day by no CS (i.e., unsignaled). On signaled trials, CSp and CSn, the signal preceded food delivery by 5 min, and terminated 5 set after food delivery. On the day prior to testing, the feeding period occurred between 1 PM and 2:30 PM. On the following 12 test days, the feeding period was scheduled according to the following sequence of deprivation levels: 20, 25, 20, 25, 25, 20, 25, 20, 25, 20, 20, and 25 hr. This schedule allowed food delivery to vary unpredictably between IO:30 AM and 3:30 PM, while restricting the range of test deprivation levels to two values, 20 and 25 hr. Over the 12 days, each stimulus condition was tested four times, twice at each deprivation level. The order of test trials was varied between 3-day blocks and between the two subgroups. A set of planned, nonorthogonal contrasts was written to test all factors in the design, including all possible comparisons between CSr, CSU, and No CS. Repeated measures contrasts were evaluated using a multivariate model (Poor, 1973). The contrasts were tested with control over the family-wise error rate, using the Bonferroni inequality (Miller, 1966).

SIGNALS

20

FOR

CSP

NoCS

HRS

DEPRIVATION

179

FOOD

CSP

CSu 25

HRS

NoCS

CSu

DEPRIVATION

Mean amount of food consumed in 15 min after a stimulus paired with food (CSp), after no CS, and after a stimulus unpaired with food (CS,). FIG.

4.

Results and Discussion The mean food consumption following CSP, CSU, and No CS at each test deprivation level is shown in Fig. 4. The counterbalancing factor was nonsignificant [ F( 1, 27)~ l] and entered into no significant interactions, so it has been omitted for clarity. At both deprivation levels, more food was eaten after CSP than under the other two conditions. Overall, 13% more food was eaten after CSP than after No CS [F(l, 27) = 9.73; 11% more food was eaten after CSP than after CSU [F( 1, 27) = 6.91. The rats ate almost the same amount after CSU and after No CS [F( 1,27) c 11. The test deprivation factor was significant [F( 1, 27) = 86.01, indicating that more food was eaten at the higher deprivation level, but this factor did not enter into any significant interactions. The finding that more food was eaten after CSr than after No CS replicates Zamble’s (1973) report of increased food consumption after a long duration food CS. Since the rats were equally familiar with being fed with or without CSP, the lower food consumption in the No CS condition cannot be attributed simply to the surprise generated by food delivery. The fact that more food was eaten after CSP than after CS,,, a stimulus which was equally familiar but not food-paired, indicates that the higher food consumption after CSP was due to its previous association with food. It is thus possible to interpret the results in terms of the conditioned motivational properties of CSP. On the other hand, the results may still be explained in terms of surprise by assuming that the CS,-No CS difference was due to the unconditioned

180

PETER F. LOVIBOND

arousing properties of CSp, and that the CSp-CSU difference was due to the surprise at being fed after CS,,. However, if this step is taken, it becomes very difficult to discriminate between the surprise account (or more generally, an expectancy account) and the motivational account. It is not logically possible to provide a control condition which (a) is comparable in its unconditioned effects to CSp, and (b) can be followed by food without causing surprise, unless it is itself a food-paired stimulus. If the concept of expectancy of food is extended to explain the ability of food signals to arouse activity and facilitate instrumental responding, it would appear that there is no useful distinction to be made between the expectancy and motivational approaches in this context (cf. Mackintosh, 1974, pp. 227-231). The present results, and the food consumption results of Experiment 1, may be described then in terms of the ability of foodrelated stimuli to increase appetitive motivation (or create an expectancy for food) in such a way as to encourage eating as soon as food is presented. GENERAL DISCUSSION The present experiments were designed to evaluate the prediction generated by theories of appetitive motivation that stimuli associated with food may facilitate activity, food-rewarded instrumental responding, and eating. This prediction was confirmed. According to the motivational account, the food-related stimuli became capable of influencing a central appetitive motivational system which is responsible for regulating appetitive behavior. The case for regarding the several measures employed as indices of appetitive motivation has been argued elsewhere (e.g., Bindra, 1968; Konorski, 1967; Rescorla & Solomon, 1967). Taken separately, the results of individual variables in the present studies are open to a number of alternative interpretations; in particular the activity data may be explained in terms of an associative model of goal-tracking. However, the concordance of the three measures in Experiment 1 would seem to support the view that they are measuring a common property of food-related stimuli. That is, the motivational view provides a relatively straightforward and consistent account of the overall results. In particular, it provides a unitary account of the excitatory and inhibitory effects observed in Experiment 1. The current results have a number of implications for motivational theories. First, the food consumption data from Experiments 1 and 2 are relevant to previous research on hunger conditioning. A number of attempts have been made to condition hunger to environmental stimuli by associating a distinctive environment with a high level of food deprivation, in the absence of food. Although Calvin, Bicknell, and Sperling (1953) reported increased food consumption in a hunger-paired environment, many experiments of this type have produced negative results (cf.

SIGNALS

FOR

FOOD

181

Cravens & Rennet-, 1970; Mineka, 1975). The present results suggest that facilitation of eating may be more easily obtained by associating stimuli with food rather than with a state of food deprivation. Indeed, the activity and lever-pressing data from Experiment 1 indicate a quite different role of food deprivation during exposure to a distinctive environment. On both of these measures, H.NF performance was consistently lower than NH.NF performance. This result confirms the view that stimuli do not acquire motivational properties by simple contiguity with a high level of hunger, and suggests that (X-hunger contiguity may in fact lead to inhibitory motivational properties if the CS also predicts a period free from food. Although the results of Experiments 1 and 2 are in general agreement with motivational theories that emphasise the CS-food association, they do not lend equal support to all such theories. In particular, the results do not accord with theories which assume that motivational effects are mediated by clasically conditioned consummatory responses (Hull, 1952; Spence, 1956), since the CS-US intervals employed were well outside the range of efficient conditioning of consummatory reponses such as salivation. For the same reason, the results do not favor theories which assert that the conditioning of a central motivational state, although not reflected accurately by overt CRs, is subject to the same laws as salivary conditioning (e.g., Rescorla & Solomon, 1967, p. 172). The finding that stimuli loosely associated with food can augment appetitive behaviors is most easily explained by theories which assume that motivational conditioning can proceed at much longer CS-US intervals than consummatory conditioning (Dickinson & Dearing, 1978; Konorski, 1967; Soltysik, 1975). The relative ease of conditioning at long CS-US intervals in the present experiments, in conjunction with the evidence that stimuli conditioned with a short CS-US interval tend to suppress instrumental responding (Azrin & Hake, 1969; Meltzer & Brahlek, 1970; Soltysik, Konorski, Holownia, & Rentoul, 1976) lends some support to the Soltysik-Konorski distinction between preparatory and consummatory conditioning. However, the results do not support Konorski’s (1967) account of the mechanism underlying preparatory conditioning. According to Konorski (1967, pp. 276-279), a CS-hunger association is formed as a function of the contiguity between the CS and the “rebound” arousal of hunger after temporary inhibition by the food US. Two aspects of the current data argue against this view. First, in Experiments 1 and 2, the unrestricted access to food given after the CS would have delayed any hunger rebound by the duration of the eating period, thus preventing CS-rebound contiguity. Second, in Experiment 1, deprivation level during training did not strongly influence conditioning to the “fed” environments. The Konorski view would predict superior conditioning to the H.F environment, since rebound is assumed to occur more easily when the hunger center is under

182

PETER

F. LOVIBOND

the influence of bodily deprivation cues (Konorski, 1967, p. 275). Soltysik’s (1975, p. 466) suggestion that preparatory reflexes possess a longer “time base” than consummatory reflexes can account for preparatory conditioning at long CS-hunger intervals but still predicts that deprivation level should affect learning. An alternative approach which is consistent with the current results is to retain the view that consummatory and preparatory CSs have opposite motivational effects, but to attribute this difference to the differing information such CSs provide about food. According to this view, stimuli loosely correlated with food, such as the “fed” environments in Experiment 1, carry the information that “food is available,” and hence elicit food-seeking behaviors (cf. Soltysik, 1975, p. 470). Stimuli closely correlated with food may carry the information that “food is here” and hence inhibit food-seeking behaviors and elicit consummatory responses. Stimuli negatively correlated with food, such as the H.NF environment in Experiment 1, may carry the information that “food is not available” and hence inhibit food-seeking behaviors. The assumption that a CS-food contingency underlies preparatory conditioning implies the operation of some mechanism other than stimulus substitution, since the preparatory CS appears to become linked to hunger-based, rather than food-based, behaviors. Some evidence for such a mechanism comes from autoshaping studies in which pigeons display pecking to a keylight paired with a US that elicits swallowing (Woodruff & Williams, 1976; cf. Williams, 1974). REFERENCES Amsel, A. Inhibition and mediation in classical Pavlovian and instrumental conditioning. In R. A. Boakes L M. S. Halliday (Eds.), Inhibition and Learning. London: Academic Press, 1972. Amsel, A., & Work, M. S. The role of learned factors in “spontaneous” activity. Journal of Comparative and Physiological Psychology, 1961, 54, 527-532. Am-in, N. H., & Hake, D. F. Positive conditioned suppression: Conditioned suppression using positive reinforcers as the unconditioned stimuli. Journal of the Experimental Analysis of Behavior, 1969, 12, 167-173. Baum, M., & Gleitman, H. “Conditioned anticipation” with an extinction baseline: The need for a disinhibition control group. Psychonomic Science, 1%7, 8(3), 95%. Berlyne, D. E. Conflict, Arousal and Curiosity. New York: McGraw-Hill, 1960. Bindra, D. Neuropsychological interpretation of the effects of drive and incentivemotivation on general activity and instrumental behavior. Psychological Review, 1968, 75(l), l-22. Bindra, D. How adaptive behavior is produced: A perceptual-motivational alternative to response-reinforcement. The Behavioral and Brain Sciences, 1978, l(l), 41-52. Bindra, D., & Palfai, T. Nature of positive and negative incentive-motivational effects on general activity. Journal of Comparative and Physiological Psychology, 1967, 63(2), 288-297. Bird, K. D. Simultaneous contrast testing procedures for multivariate experiments. Multivariate Behavioral Research, 1975, 10, 343-352.

SIGNALS

FOR FOOD

183

Boakes, R. A. Performance on learning to associate a stimulus with positive reinforcement. In H. Davis t H. M. B. Hurwitz (Eds.), Operant-Pavlovian interacn’ons. Hillsdale, NJ.: Erlbaum, 1977. Calvin, J. S., Bicknell, E. A., & Sperling, D. S. Establishment of a conditioned drive based on the hunger drive. Journal of Comparative and Physiological Psychology, 1953, 46, 173-175. Campbell, B. A., & Sheffield, F. D. Relation of random activity to food deprivation. Journal of Comparative and Physiological Psychology, 1953, 46, 320-322. Cravens, R. W., & Renner, K. E. Conditioned appetitive drive states: Empirical evidence and theoretical status. Psychological Bulletin, 1970, 73(3), 212-220. Dickinson, A., & Dearing, M. F. Appetitive-aversive interactions and inhibitory processes. In A. Dickinson & R. A. Boakes (Eds.), Mechanisms of learning and motivation: A memorial volume to Jerzy Konorski. Hillsdale, NJ.: Erlbaum, 1978. Estes, W. K. Discriminative conditioning. I. A discriminative property of conditioned anticipation. Journal of Experimental Psychology, 1943, 32, 150-155. Estes, W. K. Discriminative conditioning. II. Effects of a Pavlovian conditioned stimulus upon a subsequently established operant response. Journal of Experimental Psychology, 1948, 38, 173-177. Hays, W. L. Statistics. London: Holt, Rinehart & Winston, 1969. Hearst, E., & Jenkins, H. M. Sign-tracking: The stimulus-reinforcer relation and directed action. Monograph of the Psychonomic Society, Austin, Tex. 1974. Hull, C. L. A behavior system. New Haven: Yale Univ. Press, 1952. Konorski, J. Integrative activity of the brain. Chicago: Univ. of Chicago Press, 1967. LoLordo, V. M., McMillan, J. C., & Riley, A. L. The effects upon food-reinforced pecking and treadle-pressing of auditory and visual signals for response-independent food. Learning and Motivation, 1974, 5, 24-41. Longo, N., Klempay, S., & Bitterman, M. E. Classical appetitive conditioning in the pigeon. Psychonomic Science, 1964, 1, 19-20. Mackintosh, N. J. The psychology of animal learning. London/New York: Academic Press, 1974. Meltzer, D., & Brahlek, J. A. Conditioned suppression and conditioned enhancement with the same positive UCS: An effect of CS duration. Journal of the Experimental Analysis of Behavior, 1970, 13(l), 67-73. Meltzer, D., & Hamm, R. J. Differential conditioning of conditioned enhancement and positive conditioned suppression. Bulletin of the Psychonomic Society, 1978, 11(l), 29-32. Miczek, K. A. & Grossman, S. P. Positive conditioned suppression: Effects of CS duration. Journal of the Experimental Analysis of Behavior, 1971, 15, 243-247. Miller, R. G. Simultaneous statistical inference. New York: McGraw-Hill, 1966. Mineka, S. Some new perspectives on conditioned hunger. Journal of Experimental Psychology: Animal Behavior Processes, 1975, 104(2), 134-148. Morse, W. H., & Skinner, B. F. Some factors involved in the stimulus control of operant behavior. Journal of the Experimental Analysis of Behavior, 1958, l(l), 103-107. Poor, D. D. S. Analysis of variance for repeated measures designs: Two approaches. Psychological Bulletin, 1973, 80, 204-209. Rescorla, R. A., 8~ Solomon, R. L. Two-process learning theory: Relationships between Pavlovian conditioning and instrumental learning. Psychological Review, 1967, 74(3), 151-182. Schwartz, B. Positive and negative conditioned suppression in the pigeon: Effects of the locus and modality of the CS. Learning and Motivation, 1976, 7, 86-100. Sheffield, F. D., & Campbell, B. A. The role of experience in the “spontaneous” activity of hungry rats. Journal of Comparative and Physiological Psychology, 1954,47,97-100.

184

PETER F. LOVIBOND

Spence, K. W. Behavior theory and conditioning. New Haven: Yale Univ. Press, 1956. Soltysik, S. S. Post-consummatory arousal of drive as a mechanism of incentive motivation. Acta Neurobiologiae Experimentalis, 1975, 35, 447-474. Soltysik, S., Konorski, J., Holownia, A. & Rentoul, T. The effect of conditioned stimuli signalling food upon the autochthonous instrumental responses in dogs. Acta Neurobiologiae Experimentalis, 1976, 36, 277-310. Trapold, M. A., & Overmier, J. B. The second learning process in instrumental learning. In A. H. Black & W. F. Prokasy (Eds.), Classical conditioning II: Current research and theory. New York: Appleton-Century-Crofts, 1972. Williams, D. R. &conditional behavior: Conditioning without constraint. Unpublished manuscript, University of Pennsylvania, 1974. Woodruff, G., & Williams, D. R. The associative relation underlying autoshaping in the pigeon. Journal of the Experimental Analysis of Behavior, 1976, 26(l), l-13. Zamble, E. Classical conditioning of excitement anticipatory to food reward. Journal of Comparative and Physiological Psychology, 1967, 63(3), 526-529. Zamble, E. Augmentation of eating following a signal for feeding in rats. Learning and Motivation, 1973, 4, 138-147. Received Sept 25, 1979 Revised January 3, 1980