JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR

THE INTEGRATION OF HABITS MAINTAINED BY FOODAND WATER REINFORCEMENT THROUGH

STIMULUS COMPOUNDINGSTANLEY J. WEISS, CHARLES W. SCHINDLER, AND RAMONA EASON'

THE AMERICAN UNIVERSITY AND NIDA ADDICTION RESEARCH CENTER

In Experiment 1, a light and a tone were correlated independently with water reinforcement of barpressing by rats. With different naive subjects in Experiment 2, one of these stimuli was correlatedwith food and the other with water reinforcement (counterbalanced). In both experiments the absenceof tone and light signaled extinction. Tests of stimulus-reinforcer independence in Experiment 2indicated that tone and light controlled behavior whose rate was specifically affected by deprivationstate. In the stimulus-compounding tests of both experiments, response rates were higher to tone-plus-light than to tone or light presented alone (additive summation). This is the first report of additivesummation produced through compounding stimuli paired with different reinforcers. The results arediscussed in the context of the effects of incentive motivation on operant performance.Key words: additive summation, stimulus compounding, water reinforcement, food and water re-

inforcement, incentive motivation, appetitive-aversive interactions, rats

Additive summation is observed when moreresponses are emitted to the simultaneous pre-sentation of two independently established dis-criminative stimuli (SDs) than to either pre-sented alone. This is an exceptionally robustphenomenon. It has usually been reported af-ter discrimination training in which (a) re-sponding was maintained by the same rein-forcer in each of two separately presented SDs,often a tone and a light, and (b) the simulta-neous absence of these stimuli (T + L) wasassociated with extinction and came to controlresponse cessation (Weiss, 1978). However, inspite of the numerous reports of additive sum-mation, in all of these experiments respondingwas maintained in each of the SDs by one oftwo reinforcers, food (e.g., Miller & Ackley,1970; Weiss, 1971) or shock avoidance (Emur-ian & Weiss, 1972; LoLordo & Hart, 1972).When SDS that control free-operant avoid-

ance were compounded, additive summationwas reported (Emurian & Weiss, 1972) that

I S. J. Weiss and R. Eason are at The American Uni-versity; C. W. Schindler is at the NIDA Addiction Re-search Center. This research was supported in part byGrant MH-16853 from the National Institute of MentalHealth, United States Public Health Service, awarded toS. J. Weiss. It was presented at the 1985 convention ofthe Psychonomic Society in Boston. The authors wouldlike to thank Richard Warren for his helpful commentson the manuscript. Reprints may be obtained from StanleyJ. Weiss, Department of Psychology, The American Uni-versity, Washington, D.C. 20016.

was comparable in magnitude to that producedwhen food-related SDS were compounded(Weiss, 1969, 1971, Experiment 2). Weiss andSchindler (1985) also brought rats' respondingunder the discriminative control of tone andof light SDS, with their subjects bar pressingin tone and in light and ceasing in extinction-correlated T + L. However, this experimentdiffered from the single-reinforcer experi-ments in that performance was maintained inone SD by negative reinforcement (free-operant.shock avoidance) and in the other by positivereinforcement (food). In contrast to the robustadditive summation controlled by tone-plus-light (T + L) during stimulus-compoundingtests of the single-reinforcer studies, Weiss andSchindler found similar rates in testing to tone,light, and T + L. They related this absenceof additive summation to the simultaneous ac-tivation during stimulus compounding of twodifferent, reciprocally inhibiting, motivationalsystems-the appetitive and the aversive (e.g.,Dickinson & Pearce, 1977; Konorski, 1967;Millenson & de Villiers, 1972; Mowrer, 1960).

In the current series of experiments, wesought to investigate the effects of stimuluscompounding in a situation in which respond-ing was maintained in each of the single SDSby a different positive reinforcer, food in oneand water in the other. As described earlier,there is a substantial literature on the effectsof stimulus compounding when food main-tained bar pressing in the single SD. Although

237

1988, 50, 237-247 NUMBER 2 (SEPrEMBER)

STANLEY J. WEISS et al.

Table 1Discrimination training in Experiments 1 and 2.

Experiment 1 Experiment 2

Training stimuli Tone or light Light or tone T + La Tone or light Light or tone T + La

Schedule VI VI Ext VI VI ExtReinforcer Water Water None Food Water None

a T + L = tone and light off.

there is no reason to anticipate different testresults if responding is maintained in each ofthe SDs by water reinforcement, that experi-ment has not been reported to date. Therefore,before investigating the effects of compoundingan SD in whose presence responding producesfood with an SD in whose presence respondingproduces water, we will examine the effects ofstimulus compounding when responding ismaintained in the single SDs by water rein-forcement.The characteristics of the discriminative

control generated in baseline training that de-termine the degree to which behavior is en-ergized during stimulus compounding havebeen discussed in detail elsewhere (Weiss, 1972,1978) and will not be elaborated here. Sufficeit to say that the three-component baselinetraining schedule used in Experiment 1 waschosen to maximize the likelihood of additivesummation. The training was structurallysimilar to that employed in previous stimulus-compounding studies reporting substantial be-havioral enhancement. It differed from thosebaselines only in that responding was main-tained in each of the SDs by water, rather thanby food or by shock avoidance. The trainingschedule used in Experiment 1 is outlined inthe left section of Table 1. Employing a proveneffective schedule in training is especially im-portant here in light of several unsuccessfulattempts at obtaining summation with waterreinforcement that have come to our attention(J. A. Nevin, personal communication).

EXPERIMENT IMETHOD

SubjectsFive adult male Long-Evans rats served as

subjects. They were maintained in individualcages with free access to food (Tekland RatDiet) but were deprived of water to the extent

necessary to maintain them at 80% of theirweights under free food and water. Their free-feeding weights ranged from 356 to 513 g. Sub-jects received water during training sessionsand directly thereafter.

ApparatusThe three similar operant training cham-

bers measured 20 cm high, 21 cm long, and17.5 cm wide. The front and rear chamberwalls were constructed of aluminum. The sidewalls and ceiling were 0.6-cm white translu-cent and clear plastic, respectively. The ceilingwas vented with 0.6-cm-diameter holes. Thechamber floor was composed of 0.3-cm-di-ameter stainless steel rods spaced 1.3 cm be-tween centers. A Gerbrands microswitch leverwas located on the right side of the front wall6.9 cm above the floor. A force equivalent to0.146 to 0.195 N was required to operate thebar. A food trough (that was not used through-out Experiment 1) was mounted on the leftside of the front wall at floor level. A 1-cm-diameter water dish was located on the rightside wall towards the rear of the chamber 2cm above the floor. The amount of water de-posited into the dish was controlled by theopening of a solenoid-operated valve (SkinnerB Series).The 2000-Hz tone stimulus used in training

and testing was generated by a BRS AA-201audio oscillator, amplified by a BRS AO-201amplifier, and presented through an 8-ohm20-cm speaker mounted within an enclosurecentered 20.6 cm above the training chamber.The tone's intensity was approximately 90 dBmeasured at the lever with a Type 1565-AGeneral Radio sound-level meter, scale C, withthe microphone parallel to the speaker. Withthe exhaust fan operating, the ambient noiselevel was approximately 80 dB. The tone stim-ulus was almost inaudible outside the atten-uation chest.The light stimulus used in training and test-

238

FOOD AND WATER REINFORCEMENT

ing was generated by two 15-cm, 25-W, 120-Vtubular showcase bulbs, each mounted hori-zontally 1 cm from the translucent side walls.These two bulbs produced approximately 130.2cd/M2, measured with a Honeywell Pentax 1°/210 photometer that was positioned 12.5 cmfrom, and directed at, the chamber's side wall.A shielded 7-W, 120-V nightlight bulb op-

erating at 3 W served as a houselight that wason continuously. The illumination it producedwas too dim to activate the photometer, but itnonetheless allowed the experimenter to dis-cern the subject within the apparatus. Eachtraining chamber and its associated stimulus-presentation devices were enclosed in separatesound-attenuation chests described elsewhere(Weiss, 1970). Solid-state programmingequipment (BRS) was located in a room ad-jacent to that housing the training chambers.

ProcedureInitial training. After being deprived of water,

the rats were trained to approach the waterdish when the solenoid valve was operated.Next, bar pressing was established by rein-forcing successive approximations, and all cri-terion bar presses were reinforced (a continu-ous reinforcement schedule) until approximately50 water reinforcers had been obtained. Forthe next training session bar presses were rein-forced on a variable-interval (VI) 15-s sched-ule. That is, water could be produced by a barpress on the average of once every 15 s. Theintervals on this and all other VI schedulesranged from 2 s to three times the mean value,with intervals sequenced so as to keep the lengthof any interval independent of the precedinginterval. The VI was increased to 22 s and then30 s over approximately six sessions. The lightwas on during half of these initial trainingsessions; the tone was on during the other half.Approximately 130 reinforcers were earned ineach of these (and subsequent) training ses-sions.

Discrimination training. During discrimi-nation training a VI 30-s or VI 45-s scheduleoperated in the tone and light components;reinforcement was discontinued in the absenceof these stimuli (T + L). The tone and lightcomponents were separated by T + L andsequenced so as to prevent either stimulus fromfollowing T + L more than three times insuccession. Tone and light components aver-aged approximately 3.5 min (range, 2 to 5

Table 2Schedule, deprivation, and reinforcement values on finalbaseline of Experiment 1.

Rein-80% Daily forcement

deprived water magni-weight ration tude VI Training

Subject (g) (cc) (cc) (s) sessions

1 410 17.5a .1 45 792 332 13.5 .05 30 293 280 14.5 .08 45 504 275 14.0 .05 30 39

a This subject was trained on alternate days during whichhis total water ration (reinforcement plus supplement) was35 cc.

min). The duration of T + L ranged from 60to 140 s, with a response-correction contin-gency operating. Bar presses reset the T + Lcomponent duration timer. Table 2 gives thedeprivation weight, daily water ration, rein-forcer size, VI schedule, and number of train-ing sessions for each of the subjects successfullytrained. (To keep session length roughly com-parable over subjects, the VI schedule was in-creased to 45 s for those subjects requiring alarger reinforcer magnitude to maintain stablerates.) One subject was removed from the studybecause water reinforcement did not maintainits responding, and we were concerned that amore stringent water-deprivation regimencould adversely affect its health.

Testing. When response rate in T + L wasless than 10% of the rate in tone and lightcomponents (or less than one per minute) andthe discrimination was stable, with no appar-ent trend over 4 successive days, a stimulus-compounding test was administered. This testcontained 15 blocks, each consisting of 60-spresentations of tone, light, and T + L. Orderof presentation was varied over blocks with therestriction that no stimulus follow itself. Eachof these stimuli within a block were separatedby 60s of T + L, making a representativeblock: tone, T + L, T + L, T + L, light, T+ L. Testing started after a subject receivedapproximately 20 reinforcers on its final train-ing schedule. Reinforcement was discontinuedduring the test.

RESULTS AND DISCUSSIONTable 3 presents response rates in tone,

light, and their simultaneous absence (T + L)for each subject over the final 4 days of train-

239

STANLEYJ. WEISS et al.

Table 3Criterion performance in Experiment 1 (responses perminute).

Schedule component

Extinc-Variable interval tion

Subject Tone Light T + L

1 Mean 11.4 18.2 0.8SD 3.3 5.0 0.5

2 Mean 10.1 10.9 0.6SD 1.0 .9 0.2

3 Mean 5.9 6.4 0.4SD 1.0 1.4 0.2

4 Mean 4.4 5.4 0.5SD 0.5 0.2 0.2

ing. The degree of stimulus control is clear,with all subjects responding at least 10 timesas rapidly in tone or light as in T + L. Subjects2, 3, and 4 responded at similar rates in toneand in light, whereas S-1 responded more rap-

idly in light than in tone.Additive summation was produced in T +

L by all subjects on the stimulus-compoundingtest (Figure 1). T + L controlled between 2.3(S-1) and 5.6 (S-3) times the response outputof the higher rate single stimulus (light). Asignificantly higher percentage of test re-

sponses was emitted to T + L than to tone,t(3) = 10.28, p < .01, or to light, t(3) = 7.35,p < .01, whereas light controlled a signifi-cantly higher percentage than tone, t(3) = 5.89,p < .01.

With 72.2% of the total test responses emit-ted to T + L, this is one of the strongest relativesummative effects reported to date; but overall,many fewer responses were emitted on thesetests than on comparable single-incentive (food)stimulus-compounding tests performed in thislaboratory (Weiss 1969, 1971, Experiment 2).On average, the subjects in the current exper-iment emitted only 154 total test responses (seeFigure 1). Correcting for total session time,this is only 31.6% of the total responses emittedby the subjects in Weiss (1969) and 21.7% ofthe mean total outputs reported by Weiss (1971,Experiment 2). The reasons for this three- tofive-fold increase in total test responses whenfood rather than water reinforcement main-tained responding in training are unclear. Alltests were performed in extinction.As anticipated, when bar pressing was

(5 T+L

50

i25 L

1 2 3 4SUBJECTS

Fig. 1. The distribution of stimulus-compounding testresponses among tone (T), light (L), and tone-plus-light(T + L) for individual subjects in Experiment 1. Thetotal test responses on which these percentages are basedwere 125, 112, 339, and 40 for Subjects 1, 2, 3, and 4,respectively.

maintained independently in tone and in lightby water reinforcement, and T + L was cor-related with extinction, T + L controlled pow-erful additive summation during the stimulus-compounding test. This phenomenon has beenreported numerous times when responding wasmaintained by food (Weiss, 1978), but thepresent experiment is the first report of ad-ditive summation when responding to the sin-gle stimulus was maintained by water rein-forcement.

EXPERIMENT 2The objective of Experiment 2 was to es-

tablish a training baseline, using qualitativelydifferent appetitive reinforcers, that was struc-turally similar to those used in stimulus-com-pounding experiments reporting additive sum-mation when responding to the single SDS wasmaintained by food, water, or shock avoidance.To achieve this, subjects were trained to barpress on variable-interval schedules to producefood in one SD and to produce water in another.Bar pressing was in extinction when both SDSwere absent. (The SDS were tone and lightcounterbalanced over reinforcers.) In additionto generating stable discriminative perfor-mance, an objective of this discriminationtraining was to induce the rats to emit barpresses at similar rates in the food-correlated

240

FOOD AND WATER REINFORCEMENT

Table 4Schedule, deprivation and reinforcement values on final baseline of Experiment 2.

WaterbActual/80% Daily ration/stimulusa reinforcement

deprived weight magnitude TrainingSubject (g) Food (g) Water (mL) (mL) VI (s) sessions

6 325/315 14.5/L 16/T 0.05 60 638 344/347 15/T 15/L 0.05 30 74

13 320/322 15/T 16.5/L 0.05 45 2715 310/312 16.5/T 18/L 0.06 45 1616 321/314 15/L 16.5/T 0.07 45 1819 323/308 13/T 15.5/L 0.07 30 25

a T, tone; L, light.b All food reinforcers were 45-mg Noyes rat pellets.

and water-correlated SDs. From this similar-ity, some measure of reinforcement equiva-lence over incentives might be inferred. Evenif one has difficulty with this assumption, thisrate similarity creates a clear behavioral re-ferent for use in defining operationally the re-inforcement conditions of the current experi-ment for future research and analysis.

There is a substantial body of evidence thatfood and water deprivation states are inter-related (Bolles, 1961; Collier & Knarr, 1966;Verplanck & Hayes, 1953). When animals arebeing deprived of food and water simulta-neously, and have to make the same responsein the presence of different stimuli to gain eachcommodity, the interrelatedness of these sub-stances may prevent differential stimulus-reinforcer associations form forming. That is,both tone and light could be associated withfood and water. For an informed interpretationof Experiment 2, it is necessary to knowwhether reinforcer-specific stimulus controlwas generated in training. Therefore, Exper-iment 2 included an assay to determine whetherspecific, and independent, associations wereestablished between tone-and-food and light-and-water (counterbalanced) by each subject.The independence of the control established

to the food- and water-correlated stimuli wasmeasured by switching, after discriminativecontrol was established to the food and waterSDS, from food and water deprivation to fooddeprivation alone. If bar pressing decreasesabruptly during the water-correlated SD aftera subject is no longer deprived of water, butthe subject continues to bar press in the food-correlated SD, we would have evidence of in-centive-specific stimulus control. This assay ofstimulus-incentive specificity was performed

after the stimulus-compounding test of Ex-periment 2.

METHODSubjects and Apparatus

Fourteen naive adult male Long-Evanshooded rats were given discrimination train-ing. They were maintained in individual cagesat approximately 80% of their free-feedingweights by the simultaneous deprivation of foodand water. When food and water were receivedduring a training session, both substances weresupplemented directly thereafter in an amountnecessary to maintain an animal's target weightand ensure that all supplemental food was con-sumed within about an hour. During thosetraining sessions in which only food periodswere programmed, the rat received approxi-mately half its water ration prior to the session.The apparatus was the same as that used

in Experiment 1. Noyes 45-mg rat pellets weredeposited in the food trough when a Gerbrandsfeeder was activated.Procedure

Initial training. Initially, eight animals weretrained with food (Subjects 6, 7, 10, 11, 12,15, 16, aind 19) and six with water (Subjects8, 9, 13, 14, 17, and 20). The tone or lightwas on continuously during these sessions. Thestimulus-incentive combinations for each sub-ject satisfying the training criteria are given inTable 4. The water-trained animals weretaught to approach the water cup at the soundof the solenoid valve operation, and the food-trained animals were taught to approach thefood trough when the feeder was operated.Subsequently, they were trained by the methodof successive approximations to press the bar

241

STANLEY J. WEISS et al.

to produce their respective reinforcers, a 45-mg Noyes pellet or 0.05 mL of water. After asession in which they were permitted to earnabout 50 reinforcers on a continuous schedule,they were placed on a VI 15-s schedule inwhich a response produced a reinforcer, on theaverage, 15 s after the previous reinforcer hadbeen earned. (Subject 17 had to be eliminatedat this stage because water would not maintainconsistent bar pressing.) Over six to 12 sessionsthe animals progressed to VI 30-s, VI 45-s, orVI 60-s schedules.

Discrimination Training 1. After VI training,each subject was studied on a two-componentmultiple schedule. Food or water, as appro-priate, was programmed on a VI schedule intone or light as before, but a tone-off (T) orlight-off (L) component was added duringwhich responses had no consequences (extinc-tion). Variable-interval and extinction periodsalternated, with component durations similarto those of Experiment 1. After the first severalsessions on this multiple schedule, a response-reset contingency like that described in Ex-periment 1 was introduced during the extinc-tion components. Animals required 12 to 24sessions to form a discrimination in which theirVI rates were relatively stable and about 10times their rates in T or L.

Discnrmination Training 2. The next phase oftraining was like the two just described, exceptthat food-trained animals were taught to barpress for water and water-trained animals weretaught to bar press for food. This training wasgiven in the presence of the stimulus not en-countered before. Thus, a rat that had receivedfood training in tone would receive watertraining in light. Training progressed until theVI-extinction discriminations under thisschedule were similar to those achieved in Dis-crimination Training 1. This required ap-proximately 12 sessions. Next, the rats weregiven approximately six sessions on the mul-tiple VI extinction schedule of DiscriminationTraining 1 to reestablish the stimulus controlof that phase.

Final training baseline. On the final baselineschedule, bar pressing was reinforced on a VIschedule with food during one stimulus (toneor light) and with water during the other (lightor tone). The absence of tone and light (T +L) was correlated with extinction for all rats.This schedule was similar to the multipleschedules of Discrimination Training 1 and 2

with the exception that T + L could be fol-lowed by either a VI food period or a VI waterperiod, within the limitation that no more thanthree like reinforcement periods occur insuccession. A T + L extinction period followedall VI components.An animal was trained on the final baseline

schedule until, for four consecutive sessions,its response rates in food and water compo-nents were similar and relatively stable, and atleast 10 times those in T + L. Reaching thefirst criterion of response similarity and sta-bility proved extremely difficult in spite of sys-tematic efforts to adjust independently foodand water rations, VI schedules, the size ofwater reinforcers, and component durations.For some subjects, rates in the water-corre-lated stimulus were three to five times higherthan those in the food-correlated stimulus(Subjects 9 and 14) whereas for others thereverse was the case (Subjects 7 and 10). Inaddition, although usually in the directions in-dicated, these rates varied widely over sessions.Subjects 1 1, 12, and 20 had generally unstablerates, with food rates higher in some sessionsand water rates higher in others. Because theresults of a stimulus-compounding test wouldbe hard to interpret if animals entered it withunstable behavior and large rate differencesbetween tone and light stimuli, the animalsdescribed above were eliminated from the study.

Experiment 2 was conducted in two repli-cations. Two of the 5 subjects in the first rep-lication satisfied the discrimination criteria, es-sentially the same proportion as in the secondreplication, in which 4 of the 9 subjects sat-isfied the criteria within a reasonable time pe-riod. As in Replication 1, the problems causingsubjects to be eliminated were different ratesin the presence of food and water SDs andgeneral instability. Table 4 presents, for eachof the 6 subjects satisfying the discriminationcriteria, its weight, daily food and water ra-tions, stimulus-reinforcer combinations, water-reinforcement volume, VI values, and numberof training sessions to criterion. After a subjectsatisfied the discrimination criteria, it receiveda stimulus-compounding test like that de-scribed in Experiment 1.

Incentive-independence assay. After com-pleting its stimulus-compounding test, a ratwas given free access to water in its home cagefor 2 weeks while maintained on the limitedfood ration specified in Table 4. Then it was

242

FOOD AND WATER REINFORCEMENT

S-6

T+A6_ - .-

S-8

T(FOOD)

.T _-~

S-13

AmAA.A

\R T(FOOD)

_ L(WATER)

S-IS

TOOD)

WATER)A_A

S-16

TWATER)

-4 -3 -2 -I 1 2 3CBS FW

SESSIONS

S-19

A A

-4 -3 -2 -I

CBS

* T(FOOD)

L(WATER)

1 2 3FW

Fig. 2. Response rates of subjects meeting baseline performance criteria of Experiment 2 during the four trainingsessions preceding the stimulus-compounding test (left portion of each frame) and the three sessions following 2 weeksof free access to water in home cages (right of each frame).

trained for three additional sessions on the fi-nal baseline training schedule. Relative to thetraining weights indicated in Table 4, after 2weeks of free access to water the weights ofSubjects 6, 8, 13, 15, 16, and 19 changed by0, + 10, +2, +21, -10, and +8 g, respectively.

RESULTS AND DISCUSSIONCriterion baseline performance is shown for

all subjects in the left section of their respectiveframes in Figure 2. Rates in food- and water-

correlated stimuli were basically stable andsimilar for all subjects, with no trends appar-ent. The discriminations formed between thesestimuli and T + L were clear and consistent,with T + L rates usually less than 4% of therates to tone or light. Baseline behavior ap-

peared to be unaffected by which stimulus wascorrelated with the different reinforcers.The results of the stimulus-compounding

test are presented in Figure 3. Five of the 6subjects responded substantially more to T +

30-

20-

10-

L(FOOD)

T(WATER)UWATER)

50

z 40-a

conz 30-0a-CAw

20-

10

20-

10-

9 w v I v 9A

L

243

A A A A

(",- L(FOOD)

STANLEY J. WEISS et al.

13 15SUBJECTS

Fig. 3. The distribution of stimulus-compounding testresponses among tone (T), light (L), and tone-plus-light(T + L) for subjects satisfying the discrimination criteriaof Experiment 2. The total test responses on which thesepercentages are based were 974, 403, 642, 1,028, 368, and403 for Subjects 6, 8, 13, 15, 16, and 19, respectively.

L than to either single stimulus alone. Onaverage, T + L controlled almost half of thetotal test responses (47.9%), with 23.5% emit-ted to the water-correlated SD and 28.6% tothe food-correlated SD. A treatments (food SD,water SD, and T + L) by subjects ANOVA(Lindquist, 1953) yielded an F(2, 10) = 10.56,p < .005. Significantly more bar presses wereemitted to T + L than to the food SD, t(10)= 3.55, p < .01, or to the water SD, t(10) =

4.30, p < .01, whereas food and water SDScontrolled similar response rates, t(10) = 0.76,p > .4. The treatments-by-subjects interactionterm supplied the degrees of freedom for theset tests (Lindquist, 1953).The relative summative effect was not as

pronounced in Experiment 2, where 47.9% ofthe total test responses were emitted to T +L, as in Experiment 1, where 72.2% of thetotal test responses were emitted to T + L.However, this difference was inversely relatedto the number of total test responses emittedby each group. The subjects in Experiment 2emitted significantly more total responses ontheir compounding tests (M = 636 per subject;see Figure 3 for individual data) than the sub-jects in Experiment 1 (M = 154 per subject;see Figure 1 for individual data) with no over-

lap over groups, t(8) = 2.99, p < .02. Themean total test responses emitted in Experi-ment 2 (636) was in the range (adjusted fortest length) of the totals emitted in comparablesingle incentive (food) compounding studiesperformed in this laboratory-488 (Weiss,1969) and 709 (Weiss, 1971, Experiment 2).This is the first report of additive summa-

tion when stimuli correlated with qualitativelydifferent appetitive reinforcers were com-pounded. Lawson, Mattis, and Pear (1968,Experiment 2) trained rats on a two-compo-nent multiple schedule in which bar pressingproduced food during light and water duringtone, but on their stimulus-compounding testT + L controlled rates intermediate to thoseof tone and light presented alone. Weiss (1971)suggested that the results of Lawson et al. weredue to their two-component training procedurerather than the different appetitive reinforcersthey used. The Lawson et al. baseline trainingschedule did not have a T + L period thatsignaled extinction, as studies reporting ad-ditive summation traditionally have had. Thistraining causes tone and light each to becomeexcitatory relative to T + L. However, withouta T + L component signaling extinction, re-inforcement rates in tone and in light can onlybe compared to each other, making one rela-tively excitatory and the other relatively in-hibitory. Weiss (1971, Experiment 1) showedthat when bar pressing produced food in bothtone and in light on a two-component schedulelike that used by Lawson et al., no additivesummation was observed-T + L controlledrates intermediate to those of tone and lightpresented alone. Because food reinforcementclearly can support summation, this confirmedWeiss' contention that Lawson et al.'s failureto obtain summation was likely caused by theabsence of a T + L extinction component intheir training schedule. The results of the cur-rent experiment further substantiated this con-tention. As Weiss (1971) predicted, whenqualitatively different appetitive reinforcers,such as food and water, are used in a three-component training paradigm that includes anextinction-correlated T + L component, ad-ditive summation is produced to T + L intesting.

After an animal completed its stimulus-compounding test, it was supplied with freeaccess to water for 2 weeks while food depri-vation was continued. Then, its performance

244

FOOD AND WATER REINFORCEMENT

was monitored for several sessions with thefinal baseline schedule in operation. As before,bar presses during one SD produced food andduring the other, water. The outcomes of thesetests for stimulus-incentive independence arepresented in the right section of each frame inFigure 2.

Free access to water drastically reduced theresponse rates of all rats in the presence of thewater-correlated stimulus. Rates to the water-correlated stimulus during these sessions wereindistinguishable from those to T + L, theextinction-correlated stimulus, for Subjects 6and 8, and were only several responses perminute above T + L rates for the remainderof the subjects. From the baseline sessions tothe sessions conducted after water had beenfreely available in the home cage for 2 weeks,rates to the water-correlated stimulus droppedby 84.8% for Subject 6, 98% for Subject 8,88.5% for Subject 13, 82.3% for Subject 15,62.2% for Subject 16, and 71.7% for Subject19. This decrease was shown from the firstsession conducted without water deprivation,further supporting the reinforcer specificity ofthe baseline stimulus control produced by thetone-water and light-food correlations for Sub-jects 6 and 16 and tone-food and light-watercorrelations for Subjects 8, 13, 15, and 19.

All subjects continued to bar press and eatduring the food components of these sessions.This means that thirst and adjunctive (Falk,1966) behavior generated by this food mayhave served to increase the probability of barpressing during the water-correlated stimulus.If so, the assay for water-deprivation controlof responding would have been conservative.Nevertheless, rates during the water-corre-lated stimulus were negligible during thesesessions.

In addition to immediately reducing bar-pressing rate during the water-correlated stim-ulus, allowing the rat free access to water inits home cage also produced a transient in-crease in response rate to the food-correlatedstimulus in all 6 subjects. However, this in-crease was clearly sustained beyond the firstor second post-free-water session only by Sub-ject 8. Rates during the food-correlated SD inthe first post-free-water session (31.4 re-sponses per minute) were significantly higherthan rates to this SD (20.9 responses per min-ute) on the criterion sessions prior to the com-pounding test, t(5) = 3.42, p < .02. This dif-

ference was not sustained in the second andthird post-free-water sessions.

If we assume, as Bolles (1961) suggested,that a thirsty animal avoids eating dry foodbecause doing so intensifies its thirst, foodshould be much more palatable, and reinforc-ing, when an animal is not water deprived.Thus, the increase in rate here could be anal-ogous to that observed when reinforcementmagnitude is increased (Meltzer & Brahlek,1968). A similar interpretation may be madeof Grice and Davis' (1957) findings that thirstyanimals respond more to food when they areprewatered than when they are tested thirsty.Therefore, we hesitate to suggest that the ratedifference between baseline criterion sessionsand the first post-free-water session to the food-correlated SD was a transient contrast effect(Bernheim & Williams, 1967; Nevin & Shet-tleworth, 1966).

GENERAL DISCUSSIONExperiment 1 provided the first demonstra-

tion of additive summation when free-operantresponding was maintained by water rein-forcement in each of the compounded stimuli.To date, this kind of additive summation hadbeen reported only when both single stimulihad been correlated with food reinforcementor with free-operant shock avoidance (Weiss,1978). With T + L controlling over 70% ofthe total test responses in Experiment 1, themagnitude of the relative summation here ap-pears at least as large as, if not larger than,that customarily reported heretofore in similarexperimental paradigms in which food was thereinforcer (Weiss, 1969, 1971, Experiment 2,1978); but the total number of compoundingtest responses emitted in Experiment 1 wasapproximately 25% of the number emitted inthe food studies cited or in Experiment 2.

Results of Experiment 2 demonstrated ad-ditive summation for the first time when free-operant responding was maintained in one ofthe compounded stimuli with water and in theother with food. This finding conflicts withthose of Lawson et al. (1968) who concludedthat summation is not forthcoming when stim-uli associated with qualitatively different rein-forcers are compounded. However, a proce-dural comparison of that study and Experiment2 may have identified the training variablesrelevant to response enhancement during stim-

245

246 STANLEY J WEISS et al.

ulus compounding and thus the basis of thedifferent outcomes of Experiment 2 and theexperiments of Lawson et al.A discriminative stimulus is correlated with

the differential reinforcement of a particularresponse. But an SD can also be correlated witha change in the rate of the reinforcing event.By this latter correlation (a Pavlovian corre-lation) the stimulus gains what in some ac-counts is called conditioned incentive proper-ties that can modulate operant performance(Rescorla, 1972; Rescorla & Solomon, 1967;Weiss, 1978). The contribution of the incen-tive variable to the results of stimulus com-pounding has been amply documented (Weiss,1978; Weiss & Schindler, 1987). A brief re-view of the various manipulations revealingthe influence of the incentive variable withinthis experimental paradigm might permit usto place the findings of Experiment 2 in alarger context.

Clear additive summation will be producedby stimulus compounding when the singlestimuli each control an increase in operant ratebut signal no change in reinforcement fre-quency. Thus, conditioned incentive is not nec-essary for summation. However, significantlygreater summation will be produced when theSDS signal an increase in reinforcement fre-quency, making them conditioned appetitiveexcitors, in addition to controlling an increasein operant rate (Weiss, 1971, Experiment 2).

In comparison to the additivity describedabove, when stimuli are compounded that sin-gly control an increase in operant rate butsignal a decrease in reinforcement (makingthem conditioned appetitive inhibitors), thesingle and compounded stimuli control the samerate in testing (Weiss & Van Ost, 1974). Thatis the same outcome (no summation) producedwhen an SD maintaining shock avoidance wascompounded with an SD that controlled com-parable food-maintained responding (Weiss &Schindler, 1985). This similarity of com-pounding-test outcomes led Weiss and Schin-dler to speculate that the reciprocal inhibitionproduced when food- and shock-correlated SDSare compounded is similar to that produced bytwo conditioned inhibitors. How might this allbe related to the outcome of Experiment 2 inthe present study?

In Experiment 2, combining food-correlatedand water-correlated stimuli that controlledsimilar operant rates produced additive sum-

mation. This suggests that water- and food-generated incentive states are not reciprocallyinhibiting, like shock- and food-generated in-centive states appear to be. However, becauseconditioned incentive enhances, but is not nec-essary for, additive summation, the outcome ofExperiment 2 in itself does not indicate howfood- and water-generated incentive statescombine-it cannot be determined whetherthey enhance or neutralize each other.The results of Experiment 2 can be consid-

ered strictly inconsistent with Konorski's(1967) contention that ". . . all drives, whetherpreservative or protective, are antagonistic toeach other" (p. 56). That food and water arecomplementary reinforcers is probably rele-vant here, for Konorski was writing about in-dependent "protective" drives (e.g., food andsex) when he hypothesized this antagonism.Complementary reinforcers are related in sucha fashion that ingestion of one increases thereinforcement value of the other (Hursh, 1978).The complementary nature of food and

water as reinforcers could have contributed tothe summative effects produced in Experiment2. A stimulus-compounding experiment inwhich responding on the same manipulandumwas maintained to each of two SDs by inde-pendent, noncomplementary positive reinfor-cers would be informative in that regard, al-though difficult to perform. However, thatadditive summation was produced in Exper-iment 2, where tests of stimulus-reinforcer in-dependence indicated that tone and light cameto control behavior whose rate was specificallyaffected by deprivation state, leads us to an-ticipate summation in the experiment pro-posed. Such a finding would suggest that pro-tective drives are not reciprocally inhibitingwhen peripheral response competition is elim-inated.

REFERENCESBernheim, J. W., & Williams, D. R. (1967). Time-

dependent contrast effects in a multiple schedule offood reinforcement. Journal ofthe Experimental Analysisof Behavior, 10, 243-249.

Bolles, R. C. (1961). The interaction of hunger and thirstin the rat. Journal of Comparative and Physiological Psy-chology, 54, 580-584.

Collier, G., & Knarr, F. (1966). Defense of water bal-ance in the rat. Journal of Comparative and PhysiologicalPsychology, 61, 5-10.

Dickinson, A., & Pearce, J. M. (1977). Inhibitory in-

FOOD AND WATER REINFORCEMENT 247

teractions between appetitive and aversive stimuli. Psy-chological Bulletin, 84, 690-711.

Emurian, H. H., & Weiss, S. J. (1972). Compoundingdiscriminative stimuli controlling free-operant avoid-ance. Journal of the Experimental Analysis of Behavior,17, 249-256.

Falk, J. L. (1966). The motivational properties of sched-ule induced polydipsia. Journal of the ExperimentalAnalysis of Behavior, 9, 19-25.

Grice, G. R., & Davis, J. D. (1957). Effect of irrelevantthirst motivation on a response learned with food re-ward. Journal ofExperimental Psychology, 53, 347-352.

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

Konorski, J. (1967). Integrative activity of the brain. Chi-cago: University of Chicago Press.

Lawson, R., Mattis, P. R., & Pear, J. J. (1968). Sum-mation of response rates to discriminative stimuli as-sociated with qualitatively different reinforcers. Journalof the Experimental Analysis of Behavior, 11, 561-568.

Lindquist, E. F. (1953). Design and analysis of experi-ments in psychology and education. Boston: HoughtonMifflin.

LoLordo, V. M., & Hart, C. L. (1972). The effects ofcompounding discriminative stimuli that control vari-able-interval limited-hold avoidance. Psychonomic Sci-ence, 29, 147-148.

Meltzer, D., & Brahlek, J. A. (1968). Quantity of re-inforcement and fixed-interval performance. Psycho-nomic Science, 12, 207-208.

Millenson, J. R., & de Villiers, P. A. (1972). Motiva-tional properties of conditioned anxiety. In R. M. Gil-bert & J. R. Millenson (Eds.), Reinforcement: Behav-ioral analyses (pp. 98-128). New York: Academic Press.

Miller, L., & Ackley, R. (1970). Summation of respond-ing maintained by fixed-interval schedules. Journal ofthe Experimental Analysis of Behavior, 13, 199-203.

Mowrer, 0. H. (1960). Learning theory and behavior.New York: Wiley.

Nevin, J. A., & Shettleworth, S. J. (1966). An analysisof contrast effects in multiple schedules. Journal of theExperimental Analysis of Behavior, 9, 305-315.

Rescorla, R. A. (1972). Informational variables in Pav-

lovian conditioning. In G. H. Bower (Ed.), The psy-chology of learning and motivation (Vol. 6, pp. 1-46).New York: Academic Press.

Rescorla, R. A., & Solomon, R. L. (1967). Two-processlearning theory: Relationships between Pavlovian con-ditioning and instrumental learning. Psychological Re-view, 74, 151-182.

Verplanck, W. S., & Hayes, J. R. (1953). Eating anddrinking as a function of maintenance schedule. Journalof Comparative and Physiological Psychology, 46, 327-333.

Weiss, S. J. (1969). Attentional processes along a com-posite stimulus continuum during free-operant sum-mation. Journal of Experimental Psychology, 82, 22-27.

Weiss, S. J. (1970). An effective and economical sound-attenuation chamber. Journal of the Experimental Anal-ysis of Behavior, 13, 37-39.

Weiss, S. J. (1971). Discrimination training and stim-ulus compounding: Consideration of non-reinforce-ment and response differentiation consequences of S".Journal of the Experimental Analysis of Behavior, 15,387-402.

Weiss, S. J. (1972). Stimulus compounding in free-op-erant and classical conditioning: A review and analysis.Psychological Bulletin, 78, 189-208.

Weiss, S. J. (1978). Discriminated response and incen-tive processes in operant conditioning: A two-factormodel of stimulus control. Journal of the ExperimentalAnalysis of Behavior, 30, 361-381.

Weiss, S. J., & Schindler, C. W. (1985). Integratingcontrol generated by positive and negative reinforce-ment: Appetitive-aversive interactions. Bulletin of thePsychonomic Society, 23, 279-280. (Abstract)

Weiss, S. J., & Schindler, C. W. (1987). The composite-stimulus analysis and the quantal nature of stimuluscontrol: Response and incentive factors. PsychologicalRecord, 37, 177-191.

Weiss, S. J., & Van Ost, S. L. (1974). Response dis-criminative and reinforcement factors in stimulus con-trol of performance on multiple and chained schedulesof reinforcement. Learning and Motivation, 5, 459-472.

Received October 22, 1987Final acceptance April 30, 1988