Specific hungers and poison avoidance as adaptive specializations … · 2021. 4. 26. · 460 PAUL ROZIN AND JAMES W. KALAT of the organism. Biologically speaking, there is no reason

PSYCHOLOGICALREVIEW

VOL. 78, No. 6 NOVEMBER 1971

SPECIFIC HUNGERS AND POISON AVOIDANCE ASADAPTIVE SPECIALIZATIONS OF LEARNING1

PAUL ROZIN 2 AND JAMES W. KALAT3

University of Pennsylvania

Learning and memory are considered within an adaptive-evolutionary frame-work. This viewpoint is illustrated by an analysis of the role of learning inthiaraine specific hunger. Consideration of the demands the environmentmakes on the rat and the contingencies it faces in the natural environment,appreciation of the importance of the novelty-familiarity dimension forthese animals, and the realization of two new principles of learning, permita learning explanation of most specific hungers. The two new principles,"belongingness" and "long-delay learning" specifically meet the peculiardemands of learning in the feeding system. In conjunction with the importanceof the novelty dimension, they are discussed in an attempt to develop the lawsof taste-aversion learning. It is argued that the laws or mechanism of learningare adapted to deal with particular types of problems and can be fully under-stood only in a naturalistic context. The "laws" of learning in the feedingsystem need not be the same as those in other systems; manifestation of alearning capacity in one area of behavior does not imply that it will be acces-sible in other areas. This notion leads to speculations concerning the evolutionand development of learning abilities and cognitive function. Full under-standing of learning and memory involves explanation of their diversity aswelLas the extraction of common general principles.

The application of the basic principles developed in situations and species whereof evolution and adaptation to learning other solutions to the problems at handand memory offers a hopeful means of arc less adaptive. Furthermore, whenconceptualizing and ordering the increas- learning or memory capacities are broughtingly diverse data of these fields. Learning to bear on a particular type of problem orand memory, being the result of natura situation, it stands to reason that theseselection, should be expected to be best capacities should be shaped by or adapted

to the situation.1 This manuscript is an expanded version of an We propose to treat learning and

invited address by the first author to APA Division memory as any Other biological char-6 in 1969. The preparation of this article and some ^ • ,.• i • 4. 4. 4 i i ., • jof the research reported in it were supported by actenstic, subject to natural selection andNational Science Foundation Grant GB 8013. We therefore adapted to handle specific typeswould like to thank John Garcia, Henry and Lila of problems. Insofar as an importantGleitman Elisabeth Rozin, Martin SeligmanW. J. environmental problem (e.g., obtainingSmith, and Richard Solomon for their helpful com- f , -. . , ,ments and criticisms. adequate foods) presents unique demands

2 Requests for reprints should be sent to Paul or contingencies, we would expect to findRozin, Department of Psychology, University of appropriate modifications of learning andPennsylvania, Philadelphia, Pennsylvania 19104. L • ! • . • 1 1 i ^ j -î

'Supported by National Institute of Health memory abilities, closely articulated withtraining grant. Now at Duke University. one another and with the natural behavior

459

© 1971 by the American Psychological Association, Inc.

460 PAUL ROZIN AND JAMES W. KALAT

of the organism. Biologically speaking,there is no reason to assume that thereshould exist an extensive set of generallyapplicable laws of learning, independent ofthe situation in which they are manifested.This is not to say that there might not besome general laws, at a minimum resultingfrom basic constraints and features of theoperation of the nervous system, andperhaps reflecting general principles ofcausality in the physical world. However,if we look at learning within an adaptive-evolutionary framework, we should seeknot only to uncover some of the greatestcommon denominators among the be-haviors we study, but also to explore theplasticity of the mechanisms themselves,as they are shaped through selection todeal with particular types of problems.For many years, the leading ethologists(e.g., Lorenz, 1965; Tinbergen, 1951) haveespoused a position consistent with this.They have emphasized the importance ofconsidering learning within a naturalisticcontext; learning is viewed as beinggenetically programmed to occur at specificpoints in an ongoing behavior sequence.The thesis presented here is in harmonywith the ethologists' position, but em-phasizes differences in learning mechanismsthemselves, as a function of the situationsin which learning occurs.

In the face of recent evidence, comingmost significantly from the work of Garciaand his colleagues (Garcia, Ervin, &Koelling, 1966; Garcia & Koelling, 1966),challenging two important generalities inlearning, a framework for ordering thesuggested diversity of laws of learning isdesirable. We propose such a frameworkin terms of the notion of adaptive specializa-tions. Consideration of learning from theadaptive point of view offers two ad-vantages: (a) it provides a heuristic, forordering and predicting the types of learn-ing one will see in different situations, and(b) it provides a type of explanation of be-havior, in that one aspect of understandingbehavior can be considered to be explica-tion of its adaptive value.

Adaptive considerations have proveduseful in many areas of biology and are a

significant part of our understanding ofsensory psychology and physiology. Spe-cializations in sensory function, as shownin the classic works of Lettvin, Maturana,McCulloch, and Pitts (1959) in the visualsystem of the frog, Capranica (1966) onthe remarkable tuning of the frog auditorysystem to mating calls, and Roeder (1963)on detection of bat sonar by the mothauditory system, can be fully understoodonly in the context of adaptations tonatural problems. The incredible diversityof the visual system (e. g., variations in theproportion of rods and cones in the retinaas a function of species and ecology) canbe ordered and comprehended in a phy-logenetic adaptive framework, as shownin the classical work of Walls (1942). Wewould like to suggest that there are similarpossibilities in the area of learning andmemory.

In the major portion of this paper, weshall discuss specific hungers and poisoningin rats. We shall argue that some basicfeatures of learning and memory as appliedto food selection in the rat are strikinglydifferent from features characterizing therat's learning in more traditional laboratorysituations, that these differences makesense in terms of evolutionary adaptation,and that an understanding of the role oflearning and memory in food selection in-volves an elucidation of specifically adaptedlearning mechanisms and an integrationof these with genetically determined be-havior patterns.

In the remaining portion of the paper, weshall discuss whether the feeding systemis a unique case or representative example,relate our position to somewhat similartheoretical positions, including ethologyand the recent papers of Garcia (Garcia &Ervin, 1968; Garcia, McGowan, & Green,1971), Revusky (Revusky, in press; Re-vusky & Garcia, 1970), Seligman (1970),Lockard (1971), and Shettleworth (1971),and discuss the implications of our positionin the areas of comparative psychology oflearning and human function.

ADAPTIVE SPECIALIZATIONS OF LEARNING 461

SPECIFIC HUNGERS

Basic Phenomenon and Problems with aLearning Interpretation

The phenomena of specific hungers werebeautifully demonstrated in the classicalwork of Richter (1943). The questionraised by this work was, How is the animal(in particular, the rat) able, when deficientin a particular nutrient, to select thosefoods containing it? In the case of sodium,it is quite clear that sodium deficiency re-leases an innate preference for substancescontaining sodium (Nachman 1962; Rich-ter, 1956; Strieker & Wilson, 1970). How-ever, it is hard to believe that the rat comesequipped with prewired recognition systemsfor each of the many substances for whichit can show a specific hunger. The al-ternative view holds that rats can learnabout what foods "make them better"when they are ill and in this manner cometo select the proper nutrients (Harris,Clay, Hargreaves, & Ward, 1933; Scott& Quint, 1946). Rats would, in theirlifetime, learn preferences for the tastes ofonly those foods associated with recoveryfrom the deficiencies they happen to haveexperienced. This notion has the distinctadvantage of simplicity in that it accountsfor all of the specific hungers (except sodi-um) with one basic mechanism.

There was, in fact, some evidence for alearning interpretation of specific hungers.Scott and Verney (1947) offered a dis-tinctively flavored vitamin supplementedfood and an unflavored deficient food todeficient rats. After a preference developedfor flavored, enriched food, the flavor wasswitched to the deficient food. By andlarge, rats now preferred the flavored defi-cientfood, suggestinga learning mechanism.

The problem with a learning interpreta-tion, so far as psychologists were concerned,was the long delay of reinforcement. Therecovery (reward) effects produced by aningested nutrient would not occur untilmany minutes or hours after ingestion.This poses serious problems for a learningmechanism within the context of traditionallearning theory. If specific hungers werein fact learned, then some new type of

learning had to be involved. This servedas the impetus for our investigation ofspecific hungers.

The first demonstration of vitamin Bhunger was that of Harris et al. (1933),who found that given a choice of threefoods, one supplemented with B vitamins,B vitamin deficient rats would quicklycome to choose the correct food. Forconvenience, Rozin, Wells, and Mayer(1964) decided to study a more simplifiedsituation. One vitamin, thiamine, wasselected, since a clear specific hunger forthiamine had been reported (Richter,Holt, & Barelare, 1937; Scott & Quint,1946). The most pronounced symptomsof thiamine deficiency in the rat are weightloss and anorexia. (More details about thisand other aspects of thiamine hunger areavailable in a review by Rozin, 1967a).Following injection or ingestion of thiamine,deficient rats show clear signs of reversalof symptoms within a few hours, at most.The basic design consisted of raising wean-ling rats on a thiamine deficient diet for21 days, at which point they showed clearsigns of deficiency. At this point theywere offered a choice between this deficientdiet and the same diet supplemented withthiamine.

Thiamine deficient animals strongly pre-ferred the thiamine enriched choice, whilecontrol animals who were not deficient, butotherwise treated identically, did not. Afew rats with a choice between deficientand highly enriched diet showed no clearpreference during the first few choice days,but in the meantime ingested great amountsof thiamine and showed marked recovery.Subsequently, these rats developed a clearpreference for the thiamine rich diet. Itseemed highly unlikely that a rat wouldshow a preference for what it would haveneeded a few days ago without showingthat same preference when the thiaminepresumably had reinforcing value. Thisanomalous observation was confirmed in anexperiment in which rats were madedeficient in thiamine and then recoveredfrom deficiency by injection of highamounts of thiamine while consuming thedeficient diet. After a period of recovery,


the rats were offered a choice of thiamineenriched and deficient diets. Preferencesfor the thiamine enriched diets in rats thathad been deficient emerged at all the re-covery intervals studied (Rozin, 1965).

This result raises a second problem for alearning interpretation of thiamine hunger:How could rats develop a learned prefer-ence for something that at the time thepreference appeared, had no known rein-forcing effects? If the preference developedbecause of the initial reinforcing effectsof the thiamine, why did it not appearuntil much later? We shall call thisproblem Preference after recovery.

Careful consideration of the specifichungers situation and contingencies gaverise to additional problems that could notbe easily solved within the traditional learn-ing framework. Given that a rat mightlearn with the delays of reinforcement thatseem to hold here, how would the recoverybe specifically associated with a particularfood? If the animal had eaten from anumber of the choices available (or bothchoices in the two-choice situation), howcould one of these choices, presumably theone containing thiamine, show a specificincrease in preference? In other words,if an ingestion of two or more foods isfollowed by recovery, how does one of thesefoods acquire positive properties? Weshall call this third problem, Which food?

A logical extension of this notion raisesthe question of how food or the eating offood becomes associated with recoverywhen many other potential candidates forassociation exist in the environment. Afterall, following ingestion of foods includingthe vitamin enriched food, and beforerecovery can ensue, the rat performs manyacts, such as chewing, grooming, exploring,and sleeping, and receives many stimuli.By what principle does he elect to respondto food stimuli instead of other stimuli?We call this fourth problem, Why food?

We have, then, four problems in theapplication of traditional learning principlesto the explanation of specific hungers:

1. Delay.2. Preference after recovery.3. Which food?4. Why food ?

Reinterpretation of Specific Hungers

The first step in the resolution of theproblem as formulated came with theappreciation of the importance of noveltyin the determination of food preferencesin deficient rats (Rodgers & Rozin, 1966).Deficient rats show an immediate andalmost complete preference for new foods,even when the new food is thiaminedeficient and the old food has a thiaminesupplement. In this case, the new-foodpreference reverses after a few days. Theresults suggested that the "specific hunger"for thiamine may be simply a reflection ofthe thiamine deficient rat's preference fornew foods: If the new food containsthiamine, a learned preference could de-velop (mechanism still unknown).

Rodgers (1967a) succeeded in demon-strating quite conclusively that there is nospecificity to thiamine specific hunger.Reasoning that the novelty response mightoverwhelm an existing tendency to preferthe vitamin, Rodgers offered deficient ratsthe choice of a deficient novel diet or thesame novel diet supplemented with thi-amine. The usual strong specific hungerdid not appear. Furthermore, when thechoice was between two different noveldiets, one enriched, preferences for theenriched source did not develop rapidly.Finally, separate groups of thiamine defi-cient rats and pyridoxine deficient ratswere offered their basal diet in two forms:one supplemented with thiamine, the otherwith pyridoxine. If there were any spe-cificity, one would expect each group toshow a preference for the vitamin thatwould produce recovery. There were nosignificant differences between the groups.

There seemed to be two types of ex-planations for the novelty effect. One isthat the deficiency experience triggers aninnately programmed "neophilia." A moreattractive hypothesis is that the noveltyeffect comes about through learning. Theparadigms described all pit the new foodagainst the familiar food that has beenassociated with deficiency. The new-foodpreference could be a reflection of anaversion to the old food. Put more col-loquially, we could ask whether preferencesfor new foods appear because the rat

"loves the nevstuff." The i:

stuff or hates the oldssue can be clarified by

watching the behavior of deficient animalstoward different foods (Rozin, 1967b).Rats housed incages were observed during the first 15minutes of aneach day. Theinto a number of categories, such as groom-ing, sleeping, eating, and chewing. The

;ts were placed on a thi-experimentalamine deficient

One was spillag<

and then pawmotion, spilling

actually quite

food intake of

we call "redire

wooden barrier


relatively large individual

8-hour food presentationr responses were classified

diet. As deficiency de-veloped, less time was spent eating duringthe first 15 mirutes, and two striking be-haviors, quite :-are in normals, appeared.

of the food: the rat wouldapproach the food cup, sniff at the food,

at the food in a scoopingit out of the cup, so that

it fell through l[he wire mesh floor. (Thisspillage phenon.enon in deficient rats was

familiar, since we havealways had some difficulty measuring the

deficient rats on this ac-count.) The otaer new behavior was what

:ted feeding." Followingthis initial investigation of the food cup,rats would occasionally move over to the

separating the nest areafrom the rest of | the cage and begin to chewit vigorously. iThey would also, occasion-ally, chew on the cage wires. These twobehaviors suggested an aversion to thefamiliar diet. The redirected feeding sug-gests conflict between desire to eat andthe aversiveness of the food offered.Spillage is often seen in normal rats withhighly unpalatable foods, such as quinineadulterated diet 3.

When offered! the old deficient food in anew container (metal instead of glass) ina new location, (these deficient rats showedlittle potentiati<j>n of eating and continuedto show the behavior described above.Apparently the vessel and its location werenot controlling the aversion. However,when a completely new deficient food wasoffered in the familiar vessel and location,uninterrupted avid eating ensued, sug-gesting that the aversion was specific tothe food. The deficient rats were sub-sequently allowed to recover on a new,vitamin enriched diet. Following one week

of recovery, and after a 16-hour period offood deprivation, they were presented againwith the familiar deficient food for thefirst time since the onset of recovery. Thesefood deprived rats, showing no signs ofdeficiency at this time, responded to thefamiliar deficient food as they had before,with minimal ingestion and occasionalspillage and redirected feeding. In thiscase, a "normal" hungry rat prefers stayinghungry to eating its original diet whichprovides, in fact, perfectly adequate nutri-tion for him at the time, and is normallyquite palatable. Preference of hunger(eating nothing) to ingestion in hungryrats and the similarity in the rat's be-havior toward deficient and highly un-palatable (quinine adulterated) diets sug-gests strongly that we are dealing with anaversion to the familiar food.

The learned aversion interpretationplaces specific hungers in a new perspec-tive. We can consider the deficient dietas a CS and the nausea or other ill effectsproduced by its ingestion as a UCS.Presumably, the classically conditioned"ill effects" lead to avoidance of thefamiliar food. The mechanism suggestedfor the specific hunger phenomenon, asseen in the standard two-diet choice, wouldthen be that the rat learns an aversion tothe familiar deficient food. Before thetime of choice, he has already done a mostsignificant part of his learning: He knowswhat not to eat. The initial preferencefor the new food follows. Its maintenancewhen the new food is enriched could beaccounted for as an additional learnedpreference for the new food or as a failureto develop an aversion to it (see sectionon learned preference).

To the extent that specific aversionsplay a key role in specific hungers, thereis an obvious parallel between specifichungers and poisoning. Both involvelearned aversions; vitamin deficient dietis a slowly acting poison. The aversionexperiment thus suggests that these twosets of phenomena are closely related.Furthermore, the same basic problemsraised here (Delay, Which food? Whyfood?) arise with respect to poisoning.Since poisons are designed and synthesized


by man, it is unreasonable to hold thatrats show innate aversions to them. Incomparing the literature on poisoning (seeBarnett, 1963; Richter, 1953; Rzoska,1953) to that on specific hungers, there isone apparent contrast. We have reportedan increased preference for new foods inwhite rats, while the poisoning literaturewhich focuses on wild rats strongly in-dicates the opposite: an exaggerated neo-phobia following poisoning. That is, wildrats, who show a much greater base-linetendency to avoid new objects or eventsthan do white rats (e.g., see Galef, 1970),show a further exaggeration of this tend-ency, often to an extreme (Richter, 1953)following poisoning experiences. This dis-parity can be accounted for as a pro-cedural difference, since the new-familiarchoices we offered to our domestic ratswere different from those usually offeredpoisoned wild rats. In particular, in thenovelty experiments, which were donebefore we realized that specific aversionswere involved, the rat was offered a choicebetween a familiar food associated withdeficiency and a new food. Even if thewhite rat were somewhat neophobic, thismight not appear since the alternativechoice was a strongly aversive familiar diet.

In order to provide a meaningful com-parison between poisoning and specifichungers, both sets of phenomena wouldhave to be demonstrated under the samesets of conditions and in the same strainof rats. In an experiment meeting thesecriteria and comparing half-wild and do-mestic rats, a paradigm was employed thatallowed fuller expression of the rat's neo-phobic or neophilic tendencies. Rats wereraised on Diet A, prior to induction ofaversive consequences. Deficiency or poi-soning (or nothing in the case of controls)occurred in the presence of Diet B. Inthe final test, rats were offered these twodiets and a completely new one, Diet C.Therefore, rats were faced with a choiceamong a familiar safe diet (A), a familiaraversive diet (B), and a completely newdiet (C) (Rozin, 1968).

The single important result is thatall rats suffering poisoning or deficiency

showed an increased preference for thefamiliar safe food, that is, a neophobia,Half-wild rats showed a stronger neophobiafollowing the aversive experience, buthalf-wild controls also showed a higherbase-line neophobia. All experimental ratsalmost completely avoided the familiaraversive food (B), but ate some of thecompletely new food. There were nomajor differences between the specifichunger and poisoning groups. Therefore,it appears that we can consider specifichungers as a parallel to poisoning. Again,the behavior makes sense in an adaptiveframework: following an unpleasant food-related experience, the rat tends to returnto a known, safe food.

Resolution of the Inconsistencies with aLearning Interpretation

The specific aversion explanation ofspecific hungers, the realization of theimportance of the familiarity-novelty di-mension, and the appearance of twoarticles of major significance by Garciaand his colleagues enable us to resolvethe four basic problems with a learninginterpretation.

Preference after recovery. This can beaccounted for as a retained aversion to thefamiliar food. Rats made deficient inthiamine and then recovered by injectionfor 12 hours or 5 days (while continuingto eat the same deficient diet) show astrong initial novel-food preference whenoffered a novel-familiar food choice, withthiamine absent from both choices (Rozin &Rodgers, 1967). In these experiments,recovery took place in the presence of thefamiliar food. The existence of a noveltyeffect suggests that the aversion had notfully extinguished by the time of testing.It is significant in this regard that in mostof our experiments (some unpublished)the recovery effect showed some diminu-tion as time of recovery increased. Therecovery effect should be maximal whenrecovery is associated with a new food, sothat the old diet aversion does not ex-tinguish. Strong "recovery" effects arein fact seen under such circumstances in


the aversion experiment, with a singlechoice (Rozin, 1967b) and in typicaltwo-choice experiments with long recoveryperiods (Rozin, unpublished data).

Which food? In the "standard" two-choice situation, this problem is easy todeal with; the rat learns to avoid thedeficient food when it is the only dietavailable. There is no Which food?problem here. Because of this learnedaversion, rats show an immediate, vir-tually complete preference for the novelfood, thus making it possible for them tolearn about its consequences.

The testing situations described up tothis point, with the partial exception ofthe testing environment in the aversionexperiment, have been very limited, welldefined, and unnatural. Rats normallylive in a much larger area than was per-mitted to them in these studied, have anelaborate social life, and seldom facesimple binary food choices. It is not un-reasonable to assume that one or only afew foods might be available to the ratduring the deficiency period in nature,since with a wide variety of foods available,a deficiency would be quite unlikely. Intheir classic work on B vitamin hungers,Harris et al. (1933) offered deficient ratsa choice among a large variety of foods,with only one containing B vitamins insignificant amounts. By and large, theyfound that with large numbers of choices(6-10 foods), rats were unable to selectthe vitamin enriched source, and theyfound it necessary to "educate" the ratsby offering only the enriched food for aperiod of days in order to establish a main-tained preference for this food. Examina-tion of the day-to-day intake of theseanimals before the education period sug-gests a distinctly nonrandom food selectionpattern: food intake seemed to be restrictedto only a few of the many choices on anyparticular day. This seemed to be anadaptive way of unconfounding the situa-tion and suggested an analysis of the mealpatterns of deficient rats faced with thechoice of a number of new foods (Rozin,1969a). A thiamine supplement was placedin one of four diet choices for each rat.

Feeding was restricted to eight hours a day,and the food intake from each cup wasmeasured at hourly intervals. Four of the10 deficient rats studied developed clearpreferences for the enriched choice withina few days, and two other rats developedstrong preferences for two of the fourchoices, where one of the preferred choiceswas enriched. Analysis of the meals in-dicates a characteristic pattern, both on adaily and hourly basis. Means, except foran initial daily sampling of many or all ofthe choices, tended to be restricted to onefood. The rat's feeding pattern maximizesthe possibility of associating each dietwith its appropriate consequences, sincemeals tend to be isolated in time and toconsist of a single food. The emergence ofa strong preference for the enriched food is,in each case, preceded by a clearly definedmeal of that food. Furthermore, no ratfailed to develop a clear enriched food pref-erence if it ate an isolated meal of at least1 gram from the enriched food. The ratsthat failed to show adaptive preferencesin the initial part of the experiment failedto sample significant amounts of the en-riched food. Similar, though less welldefined, sampling patterns are seen innormal rats. Normals mix meals more,but this may be a direct consequence of thefact that their meals are larger. Theanorexia of deficiency may, in and of itself,exaggerate in an adaptive way the normalfeeding pattern of the rat. It is clear thatwe have here another part of the answerto the Which food ? problem.

Finally, social factors have become im-plicated in food selection in some sig-nificant recent experiments. Galef andClark (in press) studied responses to poison-ing in colonies of wild rats observed in thelaboratory under seminaturalistic condi-tions. They offered a group of wild ratstwo foods, one of which was poisoned.After a short period of time, all ratsavoided this food. The poison was re-moved, but the rats still completelyavoided the base. A litter was born, andthe behavior of the young toward these newfoods was observed. The parents andyoung were fed for three hours a day and


were constantly under observation duringthis period. During a two-week period(Days 14-28 of life in the pups), the pupscame out to eat, but ate only the safefood (possibly because this was the onlyfood being eaten by the parents). Whenthe young were fully weaned and placedin a new cage, separated from their parents,they continued for a six-day period to eatonly the safe diet and ignore the formerlypoisoned diet. Further experiments (Galef1971; Galef & Clark, in press) suggest astraightforward explanation for this effect.Rat pups tend to approach areas whereadults are located and begin feeding there.In this way, they become familiar with thefoods eaten by the adults and avoid al-ternative diets as a result of their neo-phobia. More recent work (Galef & Clark,1971; Galef & Henderson, in press) sug-gests an additional mechanism: chemicalcues from mother's milk seem to functionto familiarize the young with certain char-acteristics of the mother's food and produceinitial postweaning preferences.

Why food? Rats in the aversion experi-ment (Rozin, 1967b) did not show anaversion to the container or the location ofthe deficient food. While introduction ofa new food dramatically increased inges-tion, change to a new location and con-tainer did not. Apparently what theanimal learned was specific to the food.

In an independent and much more com-pelling experiment, Garcia and Koelling(1966) provided evidence that there was aspecific tendency for taste "CSs" to beassociated with certain types of visceral"UCSs," while exteroceptive CSs such aslight and sound would be preferentiallyassociated with exteroceptive UCSs, suchas shock. They paired light, sound, andtaste simultaneously with either electricshock or poisoning in different groups ofanimals. The shocked animals developedan avoidance of the light and sound, butnot of the taste. The poisoned animals,subjected to the same light-sound-tastepairing, avoided the taste and not thesound or light.

Delay. The reinterpretation of specifichungers, in itself, does not bring us much

closer to solving this important problem.It does not appear to be possible to rein-terpret what appeared to be long-delaylearning in terms of short delays. However,a critical experiment by Garcia et al. (1966)demonstrates that long-delay learning canoccur in this system. They induced anaversion to saccharin in rats by injectionsof apomorphine, a drug that presumablyproduces gastrointestinal upset. Aversionswere produced when the interval betweentermination of drinking and injection ofthe drug was one-half hour or more. Onlya few trials were necessary to establish aclear aversion.

A New Problem—Are There Learned Pref-erences as well as Learned Aversions!

Before we leave the problem of specifichungers, there is one new issue to resolve.Much of the specific hunger phenomenoncan be explained in terms of aversion; Isthere a positive side also ?

Up to this point we have providedevidence for three categories of foods basedon the animal's experience (Rozin, 1968).These are unexperienced or novel, familiar-safe, and familiar-aversive. The questionis, Is there a fourth category, familiar-positive? That is, is the experiencing ofpositive consequences any different fromthe experiencing of neutral consequenceswhen these experiences are contingent uponingestion of a particular substance?

Much of the evidence which appears todemonstrate positive preferences can bereinterpreted in terms of learned aversions.For example, Harris et al. (1933) foundthat when only one of a large number ofdiets contained adequate thiamine, mostdeficient rats were not successful in select-ing the correct diet. However, if the ratswere "educated" by being offered onlythe adequate food for several days, theyshowed a preference for this food when thelarger number of diets were again offered.This experiment does not distinguish be-tween learned-preference and learned-aver-sion interpretations. Rats could be avoid-ing the nonthiamine diets by a combinationof learned aversions and neophobia. That


is, after "education" the rats have learnedthat the thiamine-containing diet is "safe,"but it is not certain whether they havelearned that it has any special "recoveryfrom deficiency" properties which dis-tinguish it from other safe diets, since weknow that deficient rats prefer "old safe"foods to either old aversive or novel foods(Rozin, 1968). Similarly, in the samplingexperiment (Rozin, 1969a), the fact thatthe rat eventually prefers the only en-riched diet in a four-choice situationcould mean that he had developed aver-sions to the other choices.

There are some recent experimentswhich present more serious challenges to apure aversion model. Garcia, Ervin,Yorke, and Koelling (1967) repeatedlymade rats thiamine deficient and producedrecovery by thiamine injection. Just priorto thiamine injection, rats drank saccharinsolution. At all other times, water wasthe only fluid available. When thiaminedeficient, the rats showed an increase ofsaccharin intake over trials, both withrespect to their own water intake and tothe saccharin intake of similarly treatedanimals where thiamine injection was notcontingent on saccharin ingestion. Camp-bell (1969) has also demonstrated thatthiamine deficient rats show an increase intheir absolute intake of a sucrose solutionwhich has been associated with recoveryfrom deficiency. Zahorik and Maier (1969)used Garcia et al.'s (1967) procedure butwith a modified test and found that ratsprefer the taste associated with recoveryto both the taste associated with deficiencyand a novel taste. Furthermore, thispreference was apparent in both deficientand recovered rats.

These experiments, which certainly pro-vide evidence for learned preferences, cannonetheless be explained in terms of thethree basic categories of food. A moredecisive experiment would have to showthat rats prefer a "recovery" solution toan old "safe" solution, that is, one theydrank without aversive consequences ata time when then were not vitamin deficient.

Revusky (1967; Revusky & Garcia,1970) performed a series of experiments to

demonstrate that food with clear positiveconsequences would be preferred to foodswith relatively neutral consequences. Inhis simple design, rats were fed one nutrientsolution while hungry and a different onewhen satiated. After five days of thistraining, a significant preference developedin a two-bottle test for the solution drunkwhile deprived. This result is interpretedin terms of the greater (delayed) rein-forcing effect of the solution drunk duringdeprivation. Since both solutions wereequally familiar and the rats drank thesolution offered in satiated conditionsvoluntarily (so that an acquired aversionwould be unlikely), this experiment meetsthe requirements for demonstrating alearned preference. The effect is clear,though not large by comparison to theaversion data (see Revusky & Garcia,1970, for additional data).

It is noteworthy that the positive prefer-ence effects reported have been rathersmall by comparison with learned aversions,and difficult to obtain (Kalat & Rozin,unpublished; Revusky & Garcia, 1970).We cannot satisfactorily explain this asym-metry. Possibly the rat is better preparedto learn aversions because rapid learningthere has particular survival value; that is,mistakes are very costly. This remains, forthe moment, an intriguing problem, withpossibly great implications for the regula-tion of food intake.

Other Specific Hungers

We have offered a theory of thiaminespecific hunger. We believe that it holdsfor other learned specific hungers as well.Novel-food preferences, which imply alearned aversion mechanism, have beendemonstrated in calcium, sodium, andmagnesium deficient rats (Rodgers, 1967a)and in pyridoxine and riboflavin deficientrats (Rozin & Rodgers, 1967). The pre-ference after recovery phenomenon appearsin identical form in thiamine, riboflavin,and pyridoxine deficiency (Rozin & Rod-gers, 1967).

It seems reasonable to assume that theanorexia characteristic of most vitamindeficiencies reflects, at least in part, a


learned aversion to the deficient food. Thisis dramatically clear in the case of thiaminedeficiency, where the anorexia symptomdisappears precipitously when a new dietis offered. On the other hand, there isrelatively little anorexia in vitamin D orvitamin A deficiency, and in both casesthere has not been a clear demonstration ofa specific hunger.

There is an impressive body of researchon amino acid specific hungers and re-sponses to amino acid imbalance (Harper,1967). These "deficiencies" are char-acterized by anorexia and seem to fit intothe scheme we have described. Recently,Rogers and Harper (1970) presented evi-dence for a positive preference for a solu-tion that corrects an amino acid imbalance.

We cannot complete this reconsiderationof specific hungers without mentioning oneparticularly serious shortcoming of all themechanisms we have discussed. Although,in principle, they can account for most ofthe individual vitamin or mineral specifichungers, it is not clear how the classic"cafeteria" self-selection of Richter (1943)can be accounted for. Rats appear toself-select an extremely well balanced diet.Unless we assume that they developincipient deficiencies of a variety of nu-trients and learn aversions and prefer-ences on the basis of these minimal symp-tons and their abatement, we have no ex-planation of this remarkable phenomenon.At this time, long-delay learning mech-anisms appear inadequate to the task,since we cannot identify obvious candidatesfor unconditioned stimuli.4

'Another remaining problem concerns the diffi-culty in demonstrating specific hungers with thenutrients in an aqueous solution (Appledorf &Tannenbaum, 1967; Rozin, Wells, & Mayer, 1964).This contrasts with Richter, Holt, and Barelare's(1937) dramatic demonstration of a specific hungerfor thiamine in solution. We are inclined to believethat the difficulty frequently encountered withaqueous solutions derives from pure water being oneof the choices available. Water's great familiaritymight protect it from becoming aversive (Garcia &Koelling, 1967; Maier, Zahorik, & Albin, 1971;Nachman, 1970a). Palatability differences in thevarious choices solutions might also be involvedRogers & Harper, 1970).

WHY Is SODIUM HUNGER INNATE?

An alternative to the mechanism we havediscussed is an almost completely genetic-ally preprogrammed specific hunger. So-dium hunger seems to be such a case.Rats sodium deficient for the first timeshow a strong preference for sodium im-mediately upon tasting it (Nachman,1962; Richter, 1936, 1956; Strieker &Wilson, 1970). They also show strongpreferences for similar-tasting salts whichwill not actually alleviate their deficiency(Nachman, 1962), and they will workat a high rate in extinction at a lever whichpreviously produced a salt solution whichthey drank while sodium replete (Krieck-haus& Wolf, 1968).

We do not know why sodium differsfrom other specific hungers, but at leastthree explanations are available. First,the great importance of sodium in bodyfluid balance, its common scarcity in theenvironment (witness salt licks and saltwars), and the large amount needed by theindividual organism might place particularselection pressure in favor of a surefiresodium hunger. The various physiologicaladaptions directed toward regulation ofbody sodium are impressive. Second, it isconceivable, as Rodgers (1967b) has sug-gested, that sodium ingestion by sodiumdeficient organisms may have initial nega-tive effects8 associated with large-scaleelectrolyte shifts. There is not muchevidence on this point. Sodium deficientanimals do develop an aversion to sodiumdeficient diet (Rodgers, 1967a). However,injections of sodium into sodium deficientanimals have not successfully producedpreferences for foods ingested just priorto injection (Rodgers, 1967b). Thesepossible initial negative events may pre-vent the operation of the learning mech-

6 Scott, Verney, and Morissey (1950) noted thatmagnesium deficient rats selected magnesium defi-cient food in preference to magnesium enriched food.Rodgers (1967a) confirmed this "inverse" specifichunger and suggested that the initial ingestion ofmagnesium by magnesium deficient rats producesaversive consequences. Here, those consequencesmight be considered as "coming down" from thehyperirritability or "high" characteristic of mag-nesium deficiency.


anisms normally implicated in specifichungers. Third, sodium as a stimulus iswell defined by the taste system. It wouldbe easy for a genetic program to tie intothe proper class of stimuli. (Conversely,one might argue that the fact that sodiumrecognition seems to be so fundamental tothe taste system suggests again the greatimportance of sodium as a directive forcein evolution.) Although a unique learningability might have evolved to meet thisparticular problem, the actual geneticallyprogrammed solution seems preferable.However, important biological functions areoften overdetermined; this may be one ex-ample. It has been shown that normalrats can learn the location of sodiumsources in their environment and go to thempromptly when a sodium need is induced(Krieckhaus & Wolf, 1968).

But now we must turn the questionaround and ask, Why are all specifichungers not innate? One basic learningmechanism can provide an adaptive solu-tion for almost all specific hungers (andalso poisoning). Instead of programmingspecific innate recognitions of a variety ofnutrients, the organism can rely on thefact that the malaise produced by mostdeficiencies (and possibly the consequentrecovery on ingestion of the needed nu-trient) is adequate to establish aversions(and preferences). There remains thefascinating question of whether long-delaylearning originally evolved in response toproper selection of nutrients or the regula-tion of food (caloric) intake, or both.

Sodium hunger raises the general issueof the factors influencing the role learningwill play in solving a particular problem.With alternative mechanisms available todeal with an environmental problem, thesolution achieved by a species should bedetermined by factors of economy (cost)dictated by the actual situation and fea-tures of the development of the organism.We can see the interplay of these factorsin the case of what Strieker (personalcommunication, 1971) has pointed out isanother specific hunger: thirst for water.Morgan (1894) studied the development ofdrinking in young chicks and found that

they apparently had no innate recognitionof water. For example, they would runthrough water trays without drinking. Atsome point in the first few days of life,the bird would ingest water by pecking at awater drop or at an irregularity on thesurface or in the bottom of the water pan.As soon as water entered the bird's mouth,it would adopt characteristic drinkingmovements, and after only one or twosuch experiences, would show a clear visualrecognition of water. Hunt and Smith(1967) repeated and extended these results.

This system is highly predetermined: allcomponents seem to be genetically pro-grammed, except the recognition of thevisual stimulus for water. Even a reason-ably precise regulation of water intake isprogrammed, ready to come into operationas soon as water is experienced. Striekerand Sterritt (1967) have shown that on itsvery first drink, a chick's water intake isproportional to its deficit. We suggestthat the visual recognition of water islearned because this is an efficient way ofaccomplishing the task at hand. The situa-tion in the environment is such that a birdlooking for food (defined as small irregu-larities) will certainly end up with water inhis beak, and since he is prepared to learnthis and has all of the other genetic equip-ment necessary to handle water ingestionand balance, he is "home free." We em-phasize that this mechanism is successfulbecause the bird's normal pecking andeating behavior (genetically programmed)invariably brings it into a position wherewater enters the beak. Hunt and Smith(1967) show that the first time a bird pecksat a dew drop (likely to be his first ex-perience with water in the real world), itresponds with a "feeding" peck, which ischaracteristically different from an ap-proach to water. Only after the waterenters the beak does drinking-type be-havior appear, and subsequently, the peckat dew drops can be seen to be qualitativelydifferent from the peck at food.

We are suggesting that there are twotypes of advantages (and therefore evolu-tionary pressures) to learned solutions toproblems. The first, and most obvious,


has to do with the variable environment.In the face of a changing and variable en-vironment, it would be maladaptive to pro-gram specific modes of response, since thesewould often be inappropriate. Clearly, aplastic organism can do better in a plasticenvironment. (We expect that mono-phagous animals are less capable of learn-ing about food than omnivorous ones.)The second factor predisposing to learninghas to do with the "cost of preprogram-ming." This has two components. First,preprogramming of a particular behavior orstimulus configuration preempts a portionof the total genetic material, which couldotherwise be used for some other purpose.Second, there may be greater costs as-sociated with the evolution of genetic-ally programmed solutions, insofar asbuilt-in solutions involve more geneticreorganization.

Of course, in some cases, a geneticallyprogrammed recognition may be replacedby a learning mechanism. If an animalhad the capability of learning somethingthat was genetically programmed, other-wise favorable mutations in the geneticmaterial responsible for this behavior couldbe selected for, even though the geneticbasis for the behavior would be destroyed.Given a general learning capability, itcould often be cheaper to allow the en-vironment to supply the appropriate stimu-lus configurations, even if these will notvary much throughout life. We suggestthat this "balance sheet" type of approachmay be fruitfully applied to all specifichungers. In the case of the chick, we sug-gest that it is necessary to geneticallyspecify those portions of the feeding anddrinking system which are distinctivelydifferent (the ingestive responses andregulation). Since the chick's behaviorleads to water in the beak anyway, be-cause the preprogrammed food recognitionoverlaps with this aspect of drinking, acheap way out can be found. The costsare minimal, and the genetic savings,significant.

SUMMARY

The reanalysis of specific hungers restson three fundamental points: (a) Theactual contingencies in the feeding situa-

tion are not what they were thought to be.(b) The novelty-familiarity dimension isof particular importance to the rat. (c)To account fully for the phenomenon, twonew "principles" of learning were needed:belongingness and long delay. It is re-markable that each of Garcia et al.'s two"classic" experiments (Garcia & Koelling,1966; Garcia et al., 1966) dealt directlywith one of these problems at just the timethat these two issues became the two prob-lems in specific hungers.

In the context of the problems of specifichungers, it seemed clear that the basicprinciples demonstrated in these experi-ments must be essentially correct. Bothprinciples, belongingness and long-delaylearning, seem highly adapted to the pro-perties of the feeding system. Tastes arein fact causally linked to gastrointestinalevents, and there is a long inherent delaybetween the taste and its consequences.We suggest that specific learning mech-anisms have evolved in response to specificproblems. In the following section, weshall consider taste-aversion learning ingreater detail, in order to demonstrate theextensive ways in which learning mech-anisms may be modified to satisfy particulardemands.

TASTE-AVERSION LEARNING: ANEXAMPLE OF ADAPTIVE SPECIAL-

IZATION OF LEARNING

In this section we shall consider learningabout food. In considering whether it is a"new type" of learning, we shall examineboth the belongingness and delay principlesin some detail, consider the importance offamiliarity and novelty, and explore otherpossible differences between learning inthe feeding system and more traditionaltypes of learning.

Principles of Stimulus Selection

can presume, from the material onspecific hungers, that when faced with abewildering array of stimuli as candidatesfor association with a gastrointestinalevent, the rat has available principles withwhich to sort them out. One concerns hispast experience with these stimuli, the


novelty-familiarity dimension, and theother certain presumably built-in tend-encies to associate certain categories ofstimuli with certain relevant events (be-longingness). We shall consider each inturn.

Novelty-Familiarity Dimension andAs sociability

We have argued that the novelty (orfamiliarity) of a stimulus is of particularimportance for a rat (see Galef, 1970;Rozin, 1968). This distinction has par-ticular significance in determining themagnitude of a learned aversion to a giventaste. Revusky and Bedarf (1967) andWittlin and Brookshire (1968) showed thatrats learn aversions much more readily tonovel than to familiar solutions, even whenthe familiar solution is drunk after thenovel solution (and, of course, beforepoisoning). In these experiments, familiar-ization occurred over a period of days, but,in fact, a 20-minute exposure to a solutionfollowed by neutral consequences will pro-duce virtually the same effect, for such asolution is then quite resistant to associa-tion with poisoning (Kalat & Rozin6).This minimal (single) experience can haveoccurred three weeks before the poisoningwithout significant attenuation of theeffect.

Belongingness

The "belongingness" (Garcia & Koelling,1966) or "preparedness" (Seligman, 1970)or "stimulus relevance" (Capretta, 1961)phenomenon—that is, the tendency to as-sociate tastes with aversive internal con-sequences as opposed to associating eitherelement with anything else—seems emi-nently sensible from an adaptive point ofview. Gastrointestinal and related internalevents are, in fact, very likely to be in-itiated or influenced by substances eaten,and taste receptors, by virtue of their loca-tion, provide information about these samesubstances. It is essential for an omnivoreto have the ability to learn rapidly which

6J. Kalat and P. Rozin. Learned safety as anexplanation for taste-aversion delay of reinforcementgradients. In preparation.

substances to eat and which to avoid.However, an equal ability to associatelights and sounds with gastrointestinalconsequences would be far less adaptive;in fact, the common result would be"superstitious" learning. The belonging-ness phenomenon receives support not onlyfrom these adaptive arguments but alsofrom neurological considerations. Thegustatory receptors and the gut receptorsare similarly classified as visceral sensoryinputs and show close neurological re-lationships, specifically in the medulla andpossibly in the hypothalamus.

It appears that under some circum-stances, exteroceptive cues can become as-sociated with poisoning (Garcia, Kimel-dorf, & Hunt, 1961; Rozin, 1969b).Rather rapid learning of aversion to loca-tions and shape of drinking vessel can bedemonstrated if the UCS (apomorphine)is introduced during drinking from theappropriate vessel (Rozin, 1969b). On theother hand, even with a modest 30-minuteCS-UCS interval and the use of a con-catenation of nontaste cues includingvessel shape, location, solution texture, andtemperature, virtually no aversion appearsafter many trials (Rozin, 1969; but seeNachman, 1970a, for a demonstration thatunder proper circumstances oral tempera-ture—a nontaste but interoceptive cue—alone may be an effective cue). The sug-gestion here is that at least part of the"weakness" of exteroceptive cues derivesfrom a very rapid decay of their associ-ability with time.

The belongingness principle in relationto taste-aversion learning is elaborated byGarcia and Ervin (1968) and Garciatet al.(1971). Further extensions of the principleto areas outside of feeding have been dis-cussed by Seligman (1970) and Shettle-worth (1971).7

7 The principle of belongingness makes some pre-dictions about what aversive stimuli should be em-ployed in various types of aversion therapy. Forexample, it would be expected that nausea-producingdrugs such as apomorphine should be more effectivethan electric shock in treating alcoholism. At thepresent time, the data are not clear on this point(Rachman & Teasdale, 1969). The issue is com-plicated because it may be the case that humanslearn different things with electric shock or nauseaas UCSs.


Salience. There is evidence for "intra-modality" belongingness. Rats tend toassociate poisioning with some novel solu-tions more than others (Kalat & Rozin,1970). Rats drank one novel solutionbriefly, 15 minutes later drank a secondnovel solution, and another 15 minuteslater were poisoned. The following day,the rats were offered both solutions simul-taneously. Under these conditions, certainsolutions, which we describe as highlysalient, became more aversive than others.The salience of a solution proved to be amore potent predictor of amount ofacquired aversion than temporal proximityto poisoning. It was found possible torank novel solutions in a stable, transitive"salience" hierarchy, such that each solu-tion proved more salient (associable withpoisoning) than all solutions lower on thelist. Evidence for a salience effect on the"positive" side (recovery from thiaminedeficiency) has been reported (Campbell,1969).

As yet it is not known what constitutesthe denning characteristics of the saliencedimension. It is probable that the "rela-tive novelty" of these operationally novelsolutions contribute to the effect. By thiswe mean that more salient solutions maybe more different from previously ex-perienced solutions. There may be addi-tional factors associated with intrinsic prop-erties of the solutions. Certain tastes (e.g.,bitter) may have a particular tendency tobe associated with aversive consequences.

A functional definition of belongingness,At this time we know that some categoryof CSs, including tastes, preferentiallyassociates with some category of UCSs, in-cluding "gastrointestinal distress." Wewould like to define these categories moreprecisely. The unconditioned stimuli em-ployed have been described variously as"poisoning," "nausea," and "gastrointesti-nal upset." But the category of effectivestimuli may include stimuli not suggestedby these terms. Ingested foods can cer-tainly produce significant internal effectsoutside of the gastrointestinal system, andit is indeed possible that the primary actionof some of the UCSs presently employed is

outside of the gastrointestinal system. Onthe other hand, it is hard to understand,from an adaptive point of view, why painin the lungs or heart, for example, shouldbe selectively associated with taste. As amatter of fact, it is not known whethergastrointestinal pain is selectively as-sociated with taste. Much more researchis needed to better define the range ofvisceral sensations with which tastes havehigh associability. The extensive Russianliterature on interoceptive conditioning(see Bykov, 1957) demonstrates that ex-teroceptive cues can be attached to visceralUCSs. Whether there is some belonging-ness operating here remains to be seen.

Yet another dimension of belongingnessconcerns the temporal parameters of thevarious CSs and UCSs. In the taste-visceral system, stimuli tend to have slowonsets and to last for long periods. Ex-teroceptive CSs and UCSs, in contrast, arecharacteristically brief and well defined intime. The importance of these dimensionsand the visceral field for taste-aversionlearning could both be determined by usingexperimentally controlled UCSs, such aselectric shock to the stomach mucosa in-stead of the ill-defined poisoning procedures.

The category of effective CSs shoulddiffer from species to species. It wasargued above that the specific ability tolearn taste-poison associations is highlyadaptive because foods are the usual causeof any aversive gastrointestinal stimula-tion. This argument assumes that theanimal identifies its food by taste. How-ever, some species use other modalitiesas well. In particular, there is reason tobelieve that birds put main emphasis onvisual cues in the identification and selec-tion of food. A number of experiments byBrower (1969) indicate that birds canreadily learn aversions to the sight offood,8 and experiments by Wilcoxon, Dra-

8 One type of mimicry is based on the fact thatmany birds and some other species readily learn toavoid toxic or unpalatable prey. In these cases, asafe, palatable species evolves coloration similar tothat of an unsafe or unpalatable one and therebyobtains some protection due to the predator'sgeneralization of its learned aversion (Brower, 1969;Rothschild, 1967; Wickler, 1968).


goin, and Krai (1971) indicate that Jap-anese quail learn poison-based aversionsmore readily to the color (or opticaldensity) than to the taste of a solution.Rats and bobwhite quail were poisonedhalf an hour after drinking a solutionwhich was either blue, sour, or both blueand sour. The rats subsequently showedconsiderable aversion to the sour water,but none to the blue (dense) water. Thequail, however, showed an aversion toboth; furthermore, the quail poisoned onblue-sour water generalized their aversionto blue water and not to sour water. Toshow that the aversion to blue water wasnot based on the taste of the blue coloring,the investigators showed that quail couldform an aversion to plain water in a dis-tinctively colored tube.

We suggest that the critical dimensionfor poison-based aversion learning maynot be "taste versus other modalities" but"eating-related cues versus other cues."This type of functional categorization ofinput is in harmony with Gibson's (1966)view of perceptual systems. Eating-relatedcues would be whatever type of cue—gustatory, visual, or otherwise—a particu-lar species uses to identify food. Becauseof the intimate and relatively invariantrelationship between taste receptors andfood ingestion, and because of the neuro-logical association between taste and vis-ceral receptors, it is likely that taste cueswould always be classified as eating-related so that taste-poisoning belonging-ness should be practically universal. How-ever, in some species, other modalitiesmay also be eating related.

Nontaste sensory modalities could in-clude both eating-related and non-eating-related cues. For instance, it may be thatbirds can form poison-based aversions tothe sight of potential prey, but not toother sights. Also, it is known that odorsare less effective than tastes in poison-based aversions in rats. Possibly odorsare more effective when they emanate froma food source or if they are experiencedsimultaneously with a taste. It would beinteresting to investigate the effective cuesfor poison-based aversion learning in other

species in which food recognition is knownto occur partly via nongustatory cues.For instance, frogs have specific visual cellswhich respond maximally to flying insectsand similar stimuli (Lettvin et al., 1959).The above analysis suggests that stimuliwhich excite these "bug detectors' mightbe more easily associated with poison thanother types of visual stimuli.

LONG-DELAY LEARNING

Until recently it appeared that learningof both classical and operant types wasalmost impossible with delays of rein-forcement exceeding a few seconds. Theimportance of close temporal contiguityhas been demonstrated in a considerablevariety of experimental paradigims, andapparent exceptions seemed to depend onsecondary reinforcement (Kimble, 1961).There were cogent theoretical and adaptivereasons for assuming the universal im-portance of close temporal continguity.Premonitions existed, however, of a pos-sible problem with learning about foodingestion.

History

The first such reference is Pavlov (1927).Using morphine as a UCS, Pavlov notedsigns of nausea and profuse salivation inresponse to the touch of the experimenter,which preceded injection. Needless to say,the UCS (i.e., nausea, etc.) did not occuruntil many minutes after the morphine wasadministered. Pavlov evidently regardedthe effect as a species of conditioning, thusimplicitly accepting the possibility oflearning over long delays:

This experiment provides a clue to the well-knownfact that dogs will eat meat the first time it isoffered them, after removal of their parathyroids, orafter an Eck fistula and tying of the portal vein, buton all subsequent occasions will refuse it. Evidentlyin these cases the appearance and smell of meatproduce of themselves a reaction identical with thatproduced through direct pathological action in theabsence of the parathyroids or the portal circulation,by those toxic substances resulting from digestion ofthe meat [p. 36].

A similar recognition of the phenomenonof long-delay learning is present in the


biological literature on poisoning, regula-tion of food intake, and specific hungers.Thus, Harris et al. (1933), Scott andVerney (1947), Rzoska (1953), and Le-Magnen (1969) explicitly implicated learn-ing mechanisms; to biologists without anyparticular commitments to psychologicaltheory, this explanation seemed perfectlyplausible.

One way of reconciling these data totraditional S-R learning theory was toassume that animals associated a food withthe consequences of a previous meal of thesame food, thus achieving temporal con-tiguity (Rozin et al., 1964; Smith &Capretta, 1956). But rats need only asingle exposure to a toxic food to learn toavoid it (Rzoska, 1953); thus some otherexplanation is necessary.

The discovery that long CS-UCS in-tervals are possible in learning about theconsequences of foods occurred gradually.Garcia et al. (1961) in their early workfound that rats learned food aversions onthe basis of simultaneous exposure to Xrays. Although it was known that X rayshad their main effect with a considerabledelay ("radiation sickness"), it was as-sumed that some immediate effect of Xrays was involved in the learning, andslight evidence was offered to support thisposition. McLaurin (1964) was the firstto operationally manipulate CS-UCS inter-vals over a wide range in taste-aversionlearning with long delays of reinforcement.However, a methodological flaw precludedmeaningful interpretation of this result;McLaurin tested for aversion immediatelyafter exposure to X rays, and it was laterdemonstrated (Smith & Schaeffer, 1967)that the rats were learning aversions to thetest solution during the test itself. Thatis, the drinking of the solution duringtesting was simultaneous with the develop-ment of aversive consequences of theX rays, and temporal contiguity wasachieved. Garcia et al. (1966) avoidedthis problem by giving the test for sac-charin aversion three days after exposureto apomorphine. This experiment suc-cessfully demonstrated learning with longdelays of reinforcement of the order of

one hour. Smith and Roll (1967) foundsimilar results using X rays and eithersaccharin (up to 12-hour delays) or sucrose(up to 6-hour delays). Replications byRevusky (1968) using sucrose CS andX-ray UCS and Rozin (1969b) (usingsaccharin or casein hydrolysate as CSs andapomorphine as UCS) confirmed this effect.

Hypotheses to Explain the Long Delay

Peripheral—the aftertaste theory. Be-cause of the revolutionary nature of thisfinding, there was considerable interest inthe possibility that the apparent absence oftemporal contiguity was illusory. Althoughthe CS-UCS interval was ostensibly long,some peripheral trace of the CS might re-main throughout the interval in the formof an aftertaste, regurgitation, or a highconcentration in the blood. It could bethis trace that was involved in the condi-tioning. However, a fair amount of dataare now available to oppose this explana-tion. The main import of the data is thattaste-aversion learning is possible underconditions that should greatly minimizeany aftertaste.

First, taste-aversion learning to sucroseand saccharin solutions occurs on a singletrial with maximum delays of poisoningof about 7 and 12 hours, respectively(Revusky, 1968; Smith & Roll, 1967). Itis very doubtful that an aftertaste orperceptible change in blood concentrationpersists for such long periods. It is evenless plausible to suggest that enoughsucrose or saccharin remains in the stomachat this time to be tasted by regurgitation.Actually the regurgitation model neverhad much applicability to rats anyway,since rats rarely if ever vomit. A secondargument against the aftertaste and re-lated models is the fact that rats can learnaversions to a .05% HC1 solution with aone-hour delay of reinforcement (Garcia,Green, & McGowan, 1969). A litmuspaper test showed no measurable amountof HC1 on the tongue two minutes afterthe animal stopped drinking. Thus, thelikelihood of a conventional peripheralaftertaste an hour later is minimal. Fur-thermore, the solution was less concentrated


than the HC1 already in the stomach.Consequently, this experiment is peculiarlyeffective in eliminating regurgitation, stom-ach-tasting, and blood-tasting models. Athird line of evidence is Rozin's (1969b)demonstration that rats can learn anaversion to a specific concentration of asolution as opposed to other concentra-tions of the same solution. The animallearned the aversion just as easily to thelower as to the higher concentration. Pre-sumably after the 30-minute delay used inthis experiment, the aftertaste of a moreconcentrated solution should taste morelike a less concentrated solution of thesame substance. Similarly, the blood orstomach concentration of the substance atthe time of poisoning should not be uniquelyrelated to a particular concentration oforiginal solution by a function known in-dependently by the rats. A fourth lineof evidence is Nachman's (1970a) demon-stration that rats can learn an aversion toa particular temperature of water; anaftertaste of a temperature is difficult toimagine. Fifth, rats can learn an aversionto a solution even if one or more solutionsis drunk between the first solution andpoisoning (Kalat & Rozin, 1970, 1971a;Revusky & Bedarf, 1967). The interven-ing solutions would surely minimize anyaftertaste of the first. Finally, it has beenargued (Revusky & Garcia, 1970) thateven if the aftertaste model were correct,it would be difficult to reconcile the taste-aversion learning results to conventionallearning theories. Even if there is anaftertaste, it would have been present anddeclining in intensity for such a longperiod that there would be no precedentfor expecting learning to occur—let alonein one trial.

It may be argued that none of thesearguments completely demolishes an after-taste theory, and that certain other testscould be conducted, for example, attempt-ing to produce aversions to solutionsintubated intragastrically. Some recentevidence suggests that if such aversionsoccur at all, they are much smaller thantaste-mediated aversions (Smith, 1970).However, even if demonstrable, intragastric

aversions would not solve the delay prob-lem, as the HC1 experiments (Garcia et al.,1969), concentration aversions (Rozin,1969b), and solution temperature aversions(Nachman, 1970a) are no easier to explainin terms of contiguity with intragastricstimuli.

Bradley and Mistretta (1971) havedemonstrated the development of aversionsto intravenously introduced solutions (sac-charin) in rats, using X rays as the UCS.A circulating "slug" of high-concentrationsaccharin stimulates taste receptors in thetongue. This interesting experiment pro-vides another mechanism of learned aver-sions, but it cannot be the only mechanisminvolved in orally mediated aversions.The blood concentrations used in theseexperiments are much higher than anywhich would occur naturally or in thelong-delay experiments, and the same ex-periments discussed above as raising prob-lems for an intragastric-tasting mechanismwould be equally troublesome for a "blood-tasting mechanism."

At present there is no evidence in favorof an aftertaste theory and a considerablebody of evidence against it. Establish-ment of an alternative theory seems a moreappropriate strategy than accumulationof still further evidence against aftertasteinterpretations. Of course, from the pointof view of this paper, long-delay learningis exactly what should be expected in thissituation, and the central-mediation alter-native appears quite acceptable.

Central mechanisms—interference. Re-vusky (in press; see also Revusky &Garcia, 1970) proposes that the maximumCS-UCS interval in all types of learningdepends not on time per se, but on thenumber of other potential CSs that in-tervene between the^CS being tested andthe UCS. That is, the animal tends toassociate the UCS with the most recentpotential CS, or perhaps the last severalsuch stimuli. As the CS-UCS intervalexpands, the probability increases that theorganism will perceive other sights, sounds,etc., and consequently the probability de-creases that the animal will associate theUCS with the CS in question. In taste-


aversion learning, the argument holds, therange of potential CSs is much more re-stricted. Only tastes have a substantialtendency to be associated with aversivegastrointestinal stimulation, and typicallythe test solution is the most recent tastethe animal experienced prior to poisoning.Since no other potential CSs are present,there is nothing to interfere with learningover long delays. In taste aversion as inother learning, the animal tends to as-sociate a UCS with the last potential CS;the only difference is that in taste-aversionlearning, the potential CSs are fewer andless frequent. We would like to point outthat the taste modality differs from thecommonly studied exteroceptive modalities,in that virtually all stimulation comes fromcontacts initiated by behavior. Rats donot taste unless they approach somethingand introduce it into their mouth.

This theory is very attractive becauseit proposes that the differences betweentaste-aversion learning and other typesof learning may all be derived from thegeneral principle of belongingness, withoutpostulating an independent difference inthe delay of reinforcement function.

Nevertheless, the theory, if taken as thesoleôf primary explanation of delay, facestwo serious problems. It predicts that inthe absence of interfering taste cues, thereshould be little or no decrement in learningas the CS-UCS interval is increased. Thisis not the case. Garcia et al. (1966),Nachman, (1970a), Revusky (1968), andKalat and Rozin (197 la) have all foundan orderly decrease of aversion with in-creasing CS-UCS interval. Furthermore,all experiments have found a maximalinterval, varying from about 2 to 12 hours,beyond which no learned aversion can bedemonstrated. Kalat and Rozin (197la)deprived rats of both food and water duringthe CS-UCS interval, and still observedan orderly decrease in learned aversionswith increasing CS-UCS interval.

Not only is there a decrement in learn-ing in the absence of explicit interference,but the addition of explicit interfering cuesdoes not markedly reduce learning. Theconsumption of two or even three salient

novel solutions during a 30-minute delaybetween sucrose consumption and poisondoes not eliminate the sucrose aversion(Kalat & Rozin, 1971a). The maximalinterference effect we observed using threenovel interfering solutions following sucrose,and one poisoning, left sucrose equallypalatable with water, to which it is normallystrongly preferred.

These experiments, then, suggest thatretroactive interference can clearly weakena potential association, but it is highlyunlikely that the normal delay function canbe largely accounted for in these terms:What could interfere with drinking of a testsolution during six hours of no eating ordrinking that would be more effective thanthree novel solutions? Proactive interfer-ence is an unlikely explanation. Effectsfrom past taste experiences should beminimal, since most rats have experiencedonly highly familiar rat chow, water, andmother's milk.

Recent findings by Wilcoxon et al.(1971) pose further problems for Revusky'stheory. Unlike rats, bobwhite quail poi-soned 30 minutes, after drinking unflavoredwater from a blue tube learn an aversionto drinking from that tube. Since thequail presumably saw a great many visualstimuli in the 30-minute delay, theirability to learn over long delays in thissituation cannot be explained simply interms of absence of interference.

Although it seems unlikely that inter-ference represents the only basis for theCS-UCS delay gradient, it does appear tobe a contributing factor. Since a ratnormally experiences fewer tastes thanother stimuli within a given period, theRevusky theory may account for part ofthe difference in the delay gradients be-tween taste-aversion and other types oflearning.

Slow decay of associability as an inherentproperty of the taste system. An alternativeto the view that the delay-of-reinforce-ment gradient is a function of interferenceis the view that the delay gradient is aninherent property of the system itself;memory or associability decays muchmore slowly for taste than for other


modalities. This leaves unanswered thequestion, What accounts for this decay?

One possibility is that the delay func-tion represents the decay of short-termmemory. According to this hypothesis, astimulus is associable with other eventsonly while it is held in short-term memory,and tastes remain in short-term memoryfor unusually long periods. (Althoughthere is extensive data on short-termmemory in humans, virtually none hasbeen collected on taste, and it is con-ceivable that tastes do not compete withexteroceptive cues for space in short-termmemory.)

One way of testing this hypothesis is bymeans of electroconvulsive shock, whichhas been assumed to eradicate short-termmemories or to block their conversion tolong-term memories. If taste associationsmust be made from short-term memoryextended in time, it should be possible todemonstrate disruptive effects of electro-convulsive shock, presented within theCS-UCS interval but with longer delaysfollowing the CS than are ordinarilyeffective in disrupting other types ofmemory. In a very careful study, Nach-man (1970b) found amnesic effects ofelectroconvulsive shock in some rats whenelectroconvulsive shock was presented im-mediately after 10 seconds of drinking butnot after 30 seconds of drinking. He alsofound some amnesic effects from electro-convulsive shock presented 25 secondsafter a 5-second drinking period. Krai(1970) also found small amnesic effectswith electroconvulsive shock, using delaysof 2 or 25 minutes following a 10-minutedrinking period. As is usual with elec-troconvulsive shock experiments, the tem-poral parameters seem to vary widely fromone experiment to another. However, inboth cases, the effect of electroconvulsiveshock is small; it impeded but did notprevent learning. And in both cases, theeffective electroconvulsive shock times werewithin the range of times reported forelectroconvulsive-shock-amnesic effects inother systems. Thus there is no evidencethat the transfer of taste stimuli into long-term memory is unusually slow, or that

tastes remain in short-term memory for along time.

Another possible mechanism for the delayfunction is that some central long-term"trace" of the taste is decaying over time.Unfortunately, this is a difficult hypothesisto test. One experiment (Rozin & Ree9)at least puts certain constraints on thetype of decay which is possible. Ratswere anesthetized for 6-8 hours during theinterval between consumption of a sucrosesolution and LiCl poisoning. These ratsshowed strong learned aversions at delaysof poisoning considerably longer than thosewhich are effective in the absence ofanesthesia. Thus anesthesia seems toretard whatever process mediates thedelay of reinforcement gradient. If thisprocess is to be described as "decay," it isevidently an active rather than a passivetype.

Central mechanisms—learned safety. Thesuggestion of an "active decay" processraises still another possibility, which is nota decay mechanism at all. Perhaps thedelay gradient should be viewed not as aforgetting curve but as a learning curve.That is, in the absence of unfavorablegastrointestinal events, as time passes.following consumption of a novel solution,the animal learns that the solution is safe(Kalat & Rozin, see Footnote 6).

As evidence for this, it has been demon-strated that an animal does learn some-thing about a solution when consumptionof that solution is followed by no negativeconsequences. We have already discussedthe evidence indicating that rats learn lessaversion to familiar solutions than to novelsolutions (Revusky & Bedarf, 1967) evenwhen the familiar solution was experiencedonce for only 20 minutes, three weeksbefore poisoning (Kalat & Rozin, seeFootnote 6). This interpretation goes onto make a stronger claim: The learning of"safety" takes place within the periodmeasured by the maximum delay ofreinforcement. At the end of that period,

9 P. Rozin and P. Ree. Long extension of effectiveCS-UCS interval by anesthesia between CS and US.In preparation.


the "trace" has not decayed; it has merelybeen associated with the absence of aversiveconsequences. At intermediate delays, theintermediate amount of learned aversionreflects the fact that the animal haslearned an intermediate amount of safety.Although this interpretation is grosslydifferent (and not inconsistent with) theinterference or trace-decay interpretations,it is not easy to separate the alternativesexperimentally. The results of Rozin andRee (see Footnote 9), which suggest thatindefinitely long CS-UCS intervals willsupport aversion learning if the rat isanesthetized during the interval, are con-sistent with learned safety, since it isreasonable that the rat would be unable tolearn safety while anesthetized.

WHAT ALL THIS MEANS TO THE RAT

We are now prepared to describe how arat can handle some of the complex prob-lems in food selection. The first thing torealize is that the situation is probably lesscomplex than it might appear when therat's natural behavior is considered. Forexample, the rat who gets sick in the gar-bage dump probably did not recentlysample all the choice delicacies available(Rozin, 1969a). His choice behavioritself will help to unconfound the situation.He may have eaten a few different foods.He "knows" it was & food that made himsick (the belongingness principle) andcan discount any familiar safe foods (thenovelty principle). With the capabilityof forming associations over long delays,he is now likely to associate his illnesswith the last relevant (as defined above)thing or few things he ate over the lastfew hours. Although some of these foodsmay become more aversive than others be-cause of their intrinsic properties (salienceeffect), the rat will acquire a significantaversion to each of them (small inter-ference effect), with those closer in timeto the aversive event picking up somewhatmore aversion (temporal contiguity). Simi-lar mechanisms would be employed toaccount for important aspects of the regula-1ion of food intake.

What Kind of Learning Is This ?

We have already described a number ofdifferences between taste-aversion learn-ing and traditional learning. We shall nowconsider some additional evidence in orderto determine how fundamental this dis-tinction is. Recent experiments havesuggested two additional differences. First,it appears that rats can learn taste aver-sions when poisoned under anesthesia(Berger10; Roll & Smith, in press). Anes-thesia was continued for a long period fol-lowing UCS administration, so that it washighly likely that the learning occurredunder anesthesia. Second, one of the more"complex" characteristics of classical con-ditioning, the Kamin blocking effect (Ka-min, 1969), is either nonexistent or rela-tively weak in taste-aversion learning(Kalat & Rozin, 197Ib; Revusky, in press).

One interpretation of why taste-aversionlearning does not clearly show the Kamin-type effects is that the taste does not be-come a signal for poison in the sense thata tone or light becomes a signal for shock.In taste-aversion learning, the animal'sperception of the taste itself or of itsaffective value may change (suggested byH. Gleitman, personal communication,1971). The taste itself may becomeaversive or unacceptable, as if it were un-palatable (Rozin, 1967b). By contrast,stimuli associated with shock do not them-selves become aversive; they evoke littleavoidance outside the training situation.This difference seems related to the re-ported "irrationality" of learned tasteaversions; humans commonly maintainan aversion to foods they ate prior tobecoming nauseous, even when they aresure that some other agent was responsiblefor the nausea.

Seligman (1970) has suggested thattaste-aversion learning may be a particu-larly primitive type of learning. Theevidence mentioned above is quite con-sistent with this view. Furthermore, in abiological context, it makes sense that thisability should be primitive. The problems

10 B. Berger. Learning in the anesthetized rat.In preparation.


of food selection and the regulation of foodintake are pervasive ones, and both arelikely to involve long-delay learning. Theessential problems are relatively invariantacross species: any new food has someprobability of being toxic, and for almostall species, the caloric density of foods willvary. Therefore, it makes sense that tasteaversion and related types of learningshould be more or less the same throughoutmost of the vertebrate subphylum.

ADAPTIVE SPECIALIZATIONS OF LEARNING:GENERALITY AND RELATION TO

OTHER POSITIONS

The work on specific hungers and poi-soning suggests that there are two aspectsof adaptive specializations of learning.First, some mechanisms of learning maydiffer markedly (at least in terms of largequantitative differences in basic parameters)in the feeding system, as compared to othersystems. Second, understanding of anadaptive specialization includes delimita-tion of the situations in which it appliesand elaboration of its relationship andinteraction with the animals' natural be-havior (e.g., sampling, neophobia) in therelevant situation. The variation in inter-play among naturalistic context, geneticprogramming, and learning is clearly il-lustrated in the contrast between wateror sodium hunger, and other specifichungers.

Our emphasis has been on the first aspectof adaptation, while the ethologists havefocused more on the context of learningand interplay of prestructured and ac-quired components:

The student of innate behavior, accustomed tostudying a number of different species and the entirebehavior pattern, is repeatedly confronted with thefact that an animal may learn some things muchmore readily than others. . . . In other words,there seem to be more or less strictly localized"dispositions to learn." Different species are pre-disposed to learn different parts of the pattern. Sofar as we know, these differences between specieshave adaptive significance [Tinbergen, 1951, p. 145].

. . . innumerable observations and experimentstend to show that modifiability occurs, if at all, onlyin those performed places where built-in learningmechanisms are phylogenetically programmed to

perform just that function. How specifically thesemechanisms are differentiated for one particularfunction is borne out by the fact that they are veryoften quite unable to modify any but one strictlydetermined system of behavior mechanisms fJLorenz,1965, p. 47].

We are aware of only two well-studiedsystems showing adaptive specializations oflearning. One is feeding, which we havealready discussed, and the other involvesimprinting, as a mechanism of speciesrecognition.

Two of the prominent features of im-printing—the special sensitivity during acritical period early in life and the greatresistance to extinction—can be seen asadaptations to limit the likelihood oferrors in species recognition. The learningshould take place soon after hatching, sincethe probability of exposure to a conspecific(i.e., the mother) is highest at this time,and the bird should be less sensitive tolater experiences, since the frequency ofcontact with members of other species islikely to increase greatly after the nestlingperiod.

Feeding and imprinting can be con-sidered as two exceptions to an otherwisecorrect "general process" view of learning,or they can be considered as examplesof a basic adaptational principle pervadingmuch or all of learning. We prefer thelatter alternative and believe that theabsence of additional known instances ofadaptive specializations may reflect learn-ing psychologists' reluctance to studypotentially learned behaviors which do notfit into the general process paradigms.(Significantly, the case for adaptive spe-cializations in both imprinting and feedingcame from outside the psychology oflearning.)

At this point one can only speculateabout what other systems will show specialadaptations of learning. Bees may possessa wide variety of adaptive specializations(von Frisch, 1953, 1967). A particularlypromising example is their navigationalability. First, there is evidence for"belongingness."

Honey bees can learn toûse irregular forms, likethose of trees or rocks, as landmarks by which tosteer a course to and from the hive; but they cannot,


even by the most subtle conditioning technique, betaught to use the same forms as positive or negativesignals indicating the presence or absence of food ina tray, as von Frisch (1914) has shown. As signalsfor food, bees can distinguish different forms only ifthey are geometrically regular, preferably radiallysymmetrical (Hertz, 1937) [Lorenz, 1965, p. 47].

Second, a limited amount of experience inobserving a piece of the sun's arc enablesthe bee to "project" the rest of the arc.That is, bees raised without the oppor-tunity to see the sun have great difficultycompensating for sun movement duringnavigation. But bees that have observedsun movement only during a limited periodin the morning are fully capable of com-pensating for sun movement in the after-noon (von Frisch, 1967). In this situa-tion, the environmental input produces along-term change, and does not act "as-sociatively," but rather provides a refer-ence point. Other examples of such"calibrational learning" (Lorenz, 1965)might include adaptation to visual dis-placement produced by prisms and caloricregulation.11

The naturalistic literature is replete withother examples of surprising abilities ofanimals—such as digger wasps' memory ofthe location and state of their nests(Baerends, 1941), gobies' latent learningof the location of tidepools (Aronson,1951), salmon's recognition of home-streamodor (Hasler, 1966), doves' individual materecognition (Morris & Erickson, 1971),sparrows' acquisition and storage of songdialects (Marler, 1970), etc. These have

11 We can suggest a role for "calibrational learn-ing" in the regulation of food intake. It has beenknown for some time that rats and other mammalsrespond to changes in caloric density of food byappropriate modulation of volume intake. In therat, this compensation occurs largely as an increasein meal size, rather than in number of meals(Snowdon, 1969; Teitelbaum & Campbell, 1958).When a standard diet is diluted, the rat ends upeating larger meals: he has "recalibrated" his mealsize on the basis of the metabolic aftereffects of hismeals. It is possible to imagine a mechanism whichcompares some measure of the amount ingested withits delayed metabolic effects, and adjusts futureintake downward or upward so that the metabolicconsequences will approach some preferred or idealvalue. The demonstration of a long-delay learningmechanism makes this type of explanation feasibleand subject to investigation.

not been thoroughly studied as possiblespecialized learning mechanisms.

The "heuristic" value of an adaptiveevolutionary point of view can be suggestedby considering the types of predictionsthat might be made about some basiclearning and memory relationships. Forexample, an organism's memory of someaspect of the environment is useful onlyif that aspect is predictable or controllable;otherwise, rapid forgetting might be ad-vantageous. It is probably of no use forbees, birds, or other organisms with com-plex navigational abilities to rememberwhether it was cloudy yesterday or whichway the wind was blowing, though bothmay have been important at the time.Under these circumstances one mightexpect to find a specialization in short-term memory such that information couldbe stored for longer periods and in greaterquantities than usual without enteringlong-term memory.

Another example concerns extinction.Extinction, from an adaptive point ofview, allows an organism (a) to correct formistakes (fortuitous conditioning) and (b)to constantly reshape itself to adapt to avariable environment. One might expectthe rapidity of extinction to depend on theprobability that either of these eventswould occur. In the case of imprinting, wehave great resistance to extinction in acase where clearly the environment willnot vary (i.e., the species will not change),and the proper imprinting object is almostcertain to be present at the time of im-printing. The great resistance to extinc-tion of avoidance conditioning (Solomon &Wynne, 1954) may be an adaptivelyselected feature of this learning: the costsof errors of omission here are high. Onemight expect rapid extinction of thelearned location and stimulus propertiesof food sources, where these sources aresubject to marked seasonal or other tem-poral variations. An example might beextinction of responses to particular typesand locations of flowers in foraging bees.One might observe slow learning, but highresistance to extinction in a situation wherethe environment is stable but a critical


difference is relatively hard to detect(e.g., prey density), and "averaging" mustbe done before a clear choice or preferenceis established. One might observe rapidlearning and rapid extinction in a rapidlychanging environment (see Shettleworth,1971, for additional comments on thispoint).

Recently, a number of authors have ex-pressed positions related to ours (Garcia &Ervin, 1968; Garcia et al., 1971; Revusky,in press; Revusky & Garcia, 1970; Selig-man, 1970). We differ from all of theseauthors in the sense that they see be-longingness, in one form or another, asthe unique phenomenon to be explained,whereas we see it as an example of thegeneral adaptational principle; animalsmay not only learn some things moreeasily than others, but they may alsolearn some things in a different way thanothers. The contrast is clearest in thecase of Revusky, who attempts to pre-serve traditional learning theory intact,with the introduction of a new "belonging-ness" assumption.

Seligman (1970) proposes a new di-mension, "preparedness," based on thebelongingness relationship, to incorporaterecent findings into a more viable learningtheory. Preparedness represents the tend-ency of certain inputs (and/or outputs) tobe associated with one another, thistendency resulting from natural selection.Seligman proposes that highly preparedassociations are established with a minimalinput (e.g., number of pairings). Inaddition to very rapid learning, preparedassociations would tend to show learningwith long delays of reinforcement and per-haps high resistance to extinction. Weare certainly in accord with the generalflavor of Seligman's position, but we feelthat in his desire to reorder the phenomenaof learning, he has not fully appreciatedthe diverse natural forces that can shapebehavior and learning mechanisms. Inimprinting, for example, wh£re Seligman'sview of rapid learning (and high resistanceto extinction) fits very wellj the presenceof the critical period canntot be accom-modated without additional assumptions.

More significantly, the long-delay learningfound in feeding is probably not character-istic of other "prepared" associations, andin our view, it should not be, since in mostcases close temporal contiguity is the bestpredictor. Furthermore, we see no reasonto expect a consistent relationship betweenrapid learning and high resistance to ex-tinction (see discussion of extinction above).

In short, we disagree with Seligman thatdiversity can be ordered along any single,operationally meaningful dimension, thatis, preparedness. If preparedness meantadaptedness to situational demands, itwould be an acceptable but not clearlymeaningful dimension. Seligman has givenit operational meaning, but narrowed thescope of the phenomena he can account forin the process. In a more recent statementof the preparedness position, Seligman andHager (in press) have acknowledged someof these possible limitations.

Shettleworth (1971), in a paper writtenconcurrently with this one, has presenteda position very similar to ours. It focuseson belongingness, broadly conceived, butalso describes learning in general in termsof adaptive specializations, and providesexamples from the naturalistic literature.A recent paper by Lockard (1971) alsoexplicitly discusses the diversity of learn-ing mechanisms as an important featureignored by most psychologists and impliesthat the search for common elements inlearning across species and situations isalmost hopeless.

We differ slightly from Shettleworth andmarkedly from Lockard with respect toour optimism about the possibility of find-ing order within the diversity of learningmechanisms. Given the constraints onadaptations produced by basic propertiesof the nervous system, the cost of evolvingspecializations, and the fact that mostspecies face a common set of problems, wedoubt that a separate learning mechanismwould exist for every situation, or thatthere would be separate laws for eachspecies. It may yet be possible to formu-late laws of some degree of generality,taking ecological factors into account (see


Walls, 1942, for an example in the area ofvision).

SOME SPECULATIONS

It follows from the point of view pre-sented here that an organism may have anability that manifests itself in only a smallnumber of the total possible situations inwhich an experimenter might test for it.The ability in question might be inac-cessible to or "unconnected" with themachinery for modulating and controllingmost behavior. Using a computer analogy,we might suppose that a particular routineis designed (evolved) to handle a specificproblem. At this point in time, it isphysically connected only into the relevantinputs, outputs, or systems, and is in-accessible to the rest. Under these cir-cumstances, it would be difficult to describea species' learning capacity.

The interesting demonstrations by Bit-terman (1968), Gonzalez and Bitterman(1969), and Mackintosh (1965) of dif-ferences in learning abilities in certainsituations between a few species of fishand mammals do not in themselves in-dicate complete absence of these abilitiesin any species of fish (see Gleitman &Rozin, in press, for a general review of thisissue with respect to learning and memoryin fish). A species should be tested forany ability in those situations where itsexistence would have the greatest survivalvalue. The failure of a few fish species todemonstrate abilities such as progressiveimprovement in habit reversal in a fewlaboratory situations does not meet thisdemand.

However, the data gathered by Bitter-man, by Gonzalez and Bitterman, and byMackintosh do suggest some interestingphylogenetic generalizations about learn-ing capacities. Since the laboratory ap-paratus used to study rats and othermammals is often not ideally suited totheir natural behavior, just as the fishapparatus is not, it is quite interesting thatmammals reliably show the greater plas-ticity in these "unnatural" situations. Onepossible explanation for this, offered byBitterman (1964), is that the rat possesses

certain higher learning abilities that "fish"do not possess. Another not mutually ex-clusive possibility is that the rat andprobably most other mammals are "gen-eralists," compared to the majority of othervertebrates. That is, abilities initiallyevolved to handle a specific situation andlimited in application to the appropriatesystem may turn out to be useful in othersand may through evolution be "connected"into new systems. To what extent do thehigher abilities of the mammals representan increase in accessibility of specializedcapacities, an "emancipation," to borrowa word from the ethologists, of a capacityfrom its original tight motivational system ?

The proposed increase in accessibility ofcapacities in phylogeny may have a par-allel in ontogeny. Within the Piagetianframework, it is apparent that particularcognitive structures may have only limitedapplicability at any point in development.Piaget applies the name "decalage" to thisfeature of development. (Flavell, 1963,pp. 21-24). For example, in the Piagetianscheme, the same cognitive structure isnecessary for the achievement of mass andweight conservation, yet the latter occursabout two years after the former. Cogni-tive development may consist, in part, ofthe extension of existent capacities to newsituations, in parallel to our scheme forphylogeny.

We have argued for the existence ofadaptive specializations in learning andmemory. Since, by their nature, suchspecializations are limited to a relativelynarrow range of situations, we have pointedout that they must be inaccessible to mostfunctioning systems. We believe that thisgeneral formulation has a wide application,and we are presently applying it to anunderstanding of the difficulties in initialacquisition of reading (Rozin & Kalat,in press; Rozin, Poritsky, & Sotsky, 1971).

It is the basic thesis of this paper that ina biological-evolutionary framework, spe-cifically adapted abilities are extremelyimportant and should not suffer fromneglect as a consequence of the search forgreat generalities. To understand a setof phenomena, within humans or across


the animal kingdom, is to be able to de-scribe and explain diversity, as well as toextract common elements.

REFERENCES

APPLEDORF, H., & TANNENBAUM, S. R. Specifichunger for thiamine in the rat: Selection of lowconcentrations of thiamine in solution. Journalof Nutrition, 1967,92, 267-273.

ARONSON, L. Orientation and jumping in the gobiidfish Bathygobius soporator. American MuseumNovitates, 1951, 1486, 1-22.

BAERENDS, G. P. Fortpflanzungsverhalten undorientierung der grabwespe Ammophila campcstrisJur. Tijdschr Etntomol, 1941, 84, 68-275.

BARNETT, S. A. The rat. A study in behaviour.London: Methuen, 1963.

BITTERMAN, M. E. The evolution of intelligence.Scientific American, 1964, 212, 92-100.

BITTERMAN, M. E. Comparative studies of learningin the fish. In D. J. Ingle (Ed.), The centralnervous system and fish behavior. Chicago: Uni-versity of Chicago Press, 1968.

BRADLEY, R. M., & MISTRETTA, C. M. Intra-vascular taste in rats as demonstrated by con-ditioned aversion to sodium saccharin. Journal ofComparative and Physiological Psychology, 1971,75,186-189.

BROWER, L. P. Ecological chemistry. ScientificAmerican, 1969, 220, 22-29.

BYKOV, K. M. The cerebral cortex and the internalorgans. (Trans. & ed. by W. H. Gantt) NewYork: Chemical Publishing, 1957.

CAMPBELL, C. Development of specific preferencesin thiamine-deficient rats: Evidence against me-diation by aftertastes. Unpublished master'sthesis, Library of the University of Illinois atChicago Circle, 1969.

CAPRANICA, R. R. Vocal response of the bullfrogto natural and synthetic mating calls. Journal ofthe Acoustical Society of America, 1966, 40,1131-1139.

CAPRETTA, P. J. An experimental modification offood preference in chickens. Journal of Compara-tive and Physiological Psychology, 1961, 54, 238-242.

FLAVELL, J. H. The developmental psychology of JeanPiaget. Princeton: Van Nostrand, 1963.

GALEF, B. G. Aggression and timidity: Responsesto novelty in feral Norway rats. Journal ofComparative and Physiological Psychology, 1970,70, 370-381.

GALEF, B. G. Social effects in the weaning ofdomestic rat pups. Journal of Comparative andPhysiological Psychology, 1971, 75, 358-362.

GALEF, B. G., JR., & CLARK, M. M. Social factorsin the poison avoidance and feeding behavior ofwild and domestic rat pups. Journal of Compara-tive and Physiological Psychology, 1971, 75, 341-357. (a)

GALEF, B. G., JR., & HENDERSON, P. W. Mother'smilk: A determinant of the feeding preferences of

weaning rat pups. Journal of Comparative andPhysiological Psychology, in press.

GALEF, B. G., JR., & CLARK, M. M. Mother's milkand adult presence: Two factors determininginitial dietary selection by weanling rats. Journalof Comparative and Physiological Psychology, inpress.

GARCIA, J., & ERVIN, F. R. Gustatory-visceral andtelereceptor-cutaneous conditioning—Adaptationin internal and external milieus. Communicationsin Behavioral Biology, 1968, 1 (Pt. A), 389-415.

GARCIA, J., ERVIN, F. R., & KOELLING, R. A.Learning with prolonged delay of reinforcement.Psychonomic Science, 1966, 5, 121-122.

GARCIA, J., ERVIN, F. R., YORKE, C. H., & KOEL-LING, R. A. Conditioning with delayed vitamininjection. Science, 1967,155, 716-718.

GARCIA, J., GREEN, K. F., & McGowAN, B. K.X-ray as an olfactory stimulus. In C. Pfaffman(Ed.), Taste and olfaction. New York: RockefellerUniversity Press, 1969.

GARCIA, J., KIMELDORF, D. J., & HUNT, E. L. Theuse of ionizing radiation as a motivating stimulus.Psychological Review, 1961, 68, 383-385.

GARCIA, J., & KOELLING, R. A. Relation of cue toconsequence in avoidance learning. PsychonomicScience, 1966, 4, 123-124.

GARCIA, J., & KOELLING, R. A. A comparison ofaversion induced by x-rays, toxins, and drugs inthe rat. Radiation Research, 1967, Suppl. 7,439-450.

GARCIA, J., MCGOWAN, B. K., & GREEN, K. F.Biological constraints on conditioning. In A.Black & W. F. Prokasy (Eds.), Classical condi-tioning. New York: Appleton-Century-Crofts,in press.

GIBSON, J. J. The senses considered as perceptualsystems. Boston: Houghton Miffln, 1966.

GLEITMAN, H., & ROZIN, P. Learning and memory.In W. F. Hoar & D. J. R. Randall (Eds.), Fishphysiology. New York: Academic Press, in press.

GONZALEZ, R. C., & BITTERMAN, M. E. The spaced-trials PRE as a function of contrast. Journal ofComparative and Physiological Psychology, 1969,67, 94-103.

HARPER, A. E. Effects of dietary protein contentand amino acid pattern on food intake and pre-ference. In C. F. Code & W. Heidel (Eds.),Handbood of physiology. Vol. 1, Section 6.Washington, D. C.: American PhysiologicalSociety, 1967.

HARRIS, L. J., CLAY, J., HARGREAVES, F., & WARD,A. Appetite and choice of diet. The ability ofthe vitamin B deficient rat to discriminate be-tween diets containing and lacking the vitamin.Proceedings of the Royal Society, London (Series B),1933, 113, 161-190.

HASLER, A. D. Underwater guideposts. Homing ofsalmon. Madison: University of Wisconsin Press,1966.

HERTZ, M. Beitrag zum Farbensinn und Formen-sinn der Biene. Zeilschrift fur vergleichendePhysiologie, 1937, 24, 413-421.


HUNT, G. L., & SMITH, W. J. Pecking and initial^drinking responses in young domestic fowl.IH* Journal of Comparative and Physiological Psy-

chology, 1967, 64, 230-236.KALAT, J. W., & ROZIN, P. "Salience": A factor

which can override temporal contiguity in taste-aversion learning. Journal of Comparative andPhysiological Psychology, 1970, 71, 192-197.

KALAT, J., & ROZIN, P. Role of interference intaste-aversion learning. Journal of Comparativeand Physiological Psychology, 1971, 77, 53-58. (a)

KALAT, J., & ROZIN, P. You can lead a rat topoison—but you can't make him think. In M.E. P. Seligman & J. Hager (Eds.), Biologicalboundaries of learning. New York: Appleton-Century-Crofts, 1971, in press, (b)

KAMIN, L. J. Predictability, surprise, attention, andconditioning. In B. Campbell & R. Church(Eds), Punishment and aversive behavior. NewYork: Appleton-Century-Crofts, 1969.

KIMBLE, G. A. Hilgard and Marquis' conditioningand learning. New York: Appleton-Century-Crofts, 1961.

KRAL, P. Interpolation of clectroconvulsive shockduring CS-US interval as an impediment to theconditioning of taste aversion. PsychonomicScience, 1970, 19, 36-38.

KRIECKHAUS, E. E., & WOLF, G. Acquisition ofsodium by rats: Interaction of innate mechanismsand latent learning. Journal of Comparative andPhysiological Psychology, 1968, 65, 197-201.

LEMAGNEN, J. Peripheral and systemic actions offood in the caloric regulation of intake. Annals ofthe New York Academy of Science, 1969, 157,1127-1157.

LETTVIN, J. Y., MATURANA, H. R., McCuLLOcu, W.S., & PITTS, W. H. What the frog's eye tells thefrog's brain. Proceedings of the Institute of RadioEngineers, 1959,47, 1940-1951.

LOCKARD, R. B. Reflections on the fall of compara-tive psychology: Is there a message for us all?American Psychologist, 1971, 26, 168-179.

LORENZ, K. Evolution and modification of behavior.Chicago: University of Chicago Press, 1965.

MACKINTOSH, N. J. Selective attention in animaldiscrimination learning. Psychological Bulletin,1965, 64, 124-150.

MAIER, S. F., ZAHORIK, D. M., & ALBIN, R. W.Relative novelty of solid and liquid diet duringthiamine deficiency determines development ofthiamine-specific hunger. Journal of Comparativeand Physiological Psychology, 1971, 74, 254-262.

MARLER, P. A comparative approach to vocallearning: Song development in white-crownedsparrows. Journal of Comparative and Physio-logical Psychology, 1970, 71(2, Pt. 2).

McLAURiN, W. A. Postirradiation sacharin avoid-ance in rats as a function of the interval betweeningestion and exposure. Journal of Comparativeand Physiological Psychology, 1964, 57, 316-317.

MORGAN, C. L. An introduction to comparativepsychology. London: Walter Scott, 1894.

MORRIS, R. L., & ERICKSON, C. J. Pair-bond main-tenance in the ring dove (Streptopelia risoria).Journal of Animal Behavior, in press.

NACHMAN, M. Taste preferences for sodium salts byadrenalectomized rats. Journal of Comparativeand Physiological Psychology, 1962, 55, 1124-1129.

NACHMAN, M. Learned taste and temperatureaversions due to lithium chloride sickness aftertemporal delays. Journal of Comparative andPhysiological Psychology, 1970, 73, 22-30. (a)

NACHMAN, M. Limited effects of electroconvulsiveshock on memory of taste stimulation. Journal ofComparative and Physiological Psychology, 1970,73, 31-37. (b)

PAVLOV, I. Conditioned reflexes. Oxford: OxfordUniversity Press, 1927.

RACHMAN, S., & TEASDALE, J. Aversion therapy andbehavior disorders: An analysis. Coral Gables:University of Miami Press, 1969.

REVUSKY, S. H. Hunger level during food con-sumption : Effects on subsequent preference. Psy-chonomic Science, 1967, 7, 109-110.

REVUSKY, S. H. Aversion to sucrose produced bycontigent x-irradiation: Temporal and dosageparameters. Journal of Comparative and Physio-logical Psychology, 1968, 65, 17-22.

REVUSKY, S. H. The role of interference in associa-tion over a delay. In W. Honig & H. James(Eds.), Animal memory. New York: AcademicPress, in press.

REVUSKY, S. H., & BEDARF, E. W. Association ofillness with prior ingestion of novel foods. Science,1967, 155, 219-220.

REVUSKY, S. H., & GARCIA, J. Learned associationsover long delays. In C. H. Bower & J. T. Spence(Eds.), The psychology of learning and motivation:Advances in research and theory, IV. New York:Academic Press, 1970.

RICHTER, C. P. Increased salt appetite in adre-nalectomized rats. American Journal of Physio-logy, 1936, 115, 155-161.

RICHTER, C. P. Total self-regulatory functions inanimals and human beings, Harvey LecturesSeries, 1943, 38, 63-103.

RICHTER, C. P. Experimentally produced behaviorreactions to food poisoning in wild and domesti-cated rats. Annals of the New York Academy ofScience, 1953, 56, 225-239.

RICHTER, C. P. Salt appetite of mammals: Itsdependence on instinct and metabolism. In,L'Instinct dans le comportement des animaux et deI'homme. Paris: Masson, 1956.

RICHTER, C. P., HOLT, L. E., JR., & BARKLARE, B.,JR. Vitamin Bi craving in rats. Science, 1937,86,354-355.

RODGERS, W. L. Specificity of specific hungers.Journal of Comparative and Physiological Psy-chology, 1967, 64, 49-58. (a)

RODGERS, W. L. Thiamine specific hunger. Un-published doctoral dissertation, University ofPennsylvania, 1967. (b)

RODGERS, W., & ROZIN, P. Novel food preferencesin thiamine-deficient rats. Journal of Compara-tive and Physiological Psychology, 1966, 61, 1-4.


ROEDER, K. Nerve cells and insect behavior. Cam-bridge: Harvard University Press, 1963.

ROGERS, Q. R., & HARPER, A. E. Selection of asolution containing histidine by rats fed a histi-dine-imbalanced diet. Journal of Comparative andPhysiological Psychology, 1970, 72, 66-71.

ROLL, D. L., & SMITH, J. C. Conditioned tasteaversion in anesthetized rats. In M. E. P. Selig-raan & J. Hager (Eds.), The biological boundariesof learning. New York: Appleton-Century-Crofts, in press,

ROTHSCHILD, M. Mimicry. The deceptive way oflife. Natural History, 1967, 76, 44-51.

ROZIN, P. Specific hunger for thiamine: Recoveryfrom deficiency and thiamine preference. Journalof Comparative and Physiological Psychology, 1965,59,98-101.

ROZIN, P. Thiamine specific hunger. In, Handbookof physiology. Vol. 1, Section 6. Washington,D. C.: American Physiological Society, 1967. (a)

ROZIN, P. Specific aversions as a component ofspecific hungers. Journal of Comparative andPhysiological Psychology, 1967, 64, 237-242. (b)

ROZIN, P. Specific aversions and neophobia as aconsequence of vitamin deficiency and/or poison-ing in half-wild and domestic rats. Journal ofComparative and Physiological Psychology, 1968,66, 82-88.

ROZIN, P. Adaptive food sampling patterns invitamin deficient rats. Journal of Comparativeand Physiological Psychology, 1969, 69, 126-132.(a)

ROZIN, P. Central or peripheral mediation of learningwith long CS-US intervals in the feeding system.Journal of Comparative and Physiological Psy-chology, 1969, 67, 421-429. (b)

ROZIN, P., & KALAT, J. W. Learning as a situation-specific adaptation. In M. E. P. Seligman & J.Hager (Eds.), The biological boundaries of learning.New York: Appleton-Century-Crofts, in press.

ROZIN, P., PORITSKY, S., & SOTSKY, R. Americanchildren with reading problems can easily learnto read English represented by Chinese characters.Science, 1971, 171, 1264-1267.

ROZIN, P., & RODGERS, W. Novel diet preferencesin vitamin deficient rats and rats recovered fromvitamin deficiencies. Journal of Comparative andPhysiological Psychology, 1967, 63, 421-428.

ROZIN, P., WELLS, C., & MAYER, J. Thiaminespecific hunger: Vitamin in water versus vitaminsin food. Journal of Comparative and Physiological

Psychology, 1964, 57, 78-84.RzOSKA, J. Bait shyness, a study in rat behaviour.

British Journal of Animal Behaviour, 1953, 1,128-135.

SCOTT, E. M., & QUINT, E. Self selection of diet.III. Appetites for B vitamins. Journal of Nutri-

tion, 1946, 32, 285-291.SCOTT, E. M., £ VERNEY, E. L. Self selection of

diet. VI. The nature of appetites for B vitamins.Journal of Nutrition, 1947, 34, 471-480.

SCOTT, E. M., VERNEY, E. L., & MORISSEY, P. Selfselection diet. XI. Appetites'for calcium, mag-nesium and potassium. Journal of Nutrition, 1950,41, 187-202.

SELIGMAN, M. E. P. On generality of the laws oflearning. Psychological Review, 1970, 77, 406-418.

SELIGMAN, M. E. P., & HAGER, J. The biologicalboundaries of learning, New York: Appleton-Century-Crofts, in press.

SHETTLEWORTH, S. Constraints on learning. In D.S. Lehrman, R. A. Hinde, & E. Shaw (Eds.),Advances in the study of behavior. Vol. 4. NewYork: Academic Press, 1971.

SMITH, J. C., & ROLL, D. L. Trace conditioningwith X-rays as an aversive stimulus. Psy-chonomic Science, 1967, 9, 11-12.

SMITH, J. C., & SCHAEFFER, R. W. Developmentof water and saccharin preferences after simul-taneous exposures to saccharin solution andgamma rays. Journal of Comparative and Phy-siological Psychology, 1967, 63, 434-438.

SMITH, M. P., & CAPRETTA, P. J. Effects of drivelevel and experience on the reward value ofsaccharine solutions. Journal of Comparative andPhysiological Psychology, 1956, 49, 553-557.

SMITH, R. G. The role of alimentary chemore-ceptors in the development of taste aversions..Communications in Behavioral Biology, 1970, 5,199-204.

SNOWDON, C. T. Motivation, regulation, and thecontrol of meal parameters with oral and intra-gastric feeding. Journal of Comparative andPhysiological Psychology, 1969, 69, 91-100.

SOLOMON, R. L., & WYNNE, L. C. Traumaticavoidance learning: The principles of anxiety con-servation and partial irreversibility. PsychologicalReview, 1954, 61, 353-385.

STRICKER, E. M., & STERRITT, G. M. Osmoregula-tion in the newly hatched domestic chick. Physi-ology and Behavior, 1967,2, 117-119.

STRICKER, E. M., & WILSON N. E. Salt-seekingbehavior in rats following acute sodium deficiency.Journal of Comparative and Physiological Psy-chology, 1970, 72, 416-420.

TEITELBAUM, P., & CAMPBELL, B. A. Ingestionpatterns in hyperphagic and normal rats. Journalof Comparative and Physiological Psychology, 1958,51, 135-141.

TINBERGEN, N. The study of instinct. Oxford:Clarendon Press, 1951.

VON FRISCH, K. Der Farbensinn und Formensinnder Biene. Zoologisches Jahrbuch, 1914,35, 1-188.

VON FRISCH, K. The dancing bees. New York:Harcourt, Brace, 1953.

VON FRISCH, K. The dance language and orientationof bees. Cambridge, Mass.: Belknap Press ofHarvard University Press, 1967.

WALLS, G. L. The vertebrate eye and its adaptiveradiation. Bloomfield Hills, Mich.: CranbrookInstitute of Science, 1942.


WICKLER, W. Mimicry in plants and animals. New phine-induced conditioned aversion to a novelYork: McGraw-Hill 1968. food. Psychonomic Science, 1968, 12, 217-218.

,,, TT _ _ ' _ . ZAHORIK, D. M., & MAIER, S. F. Appetitive con-WILCOXON, H. C., DRAGOIN, W. B., & KRAL, P. A. ^^ with'recovery frOm thiamirTe deficiency

Illness-induced aversions in rat and quail: Rela- as the unconditioned stimulus. Psychonomic

tive salience of visual and gustatory cues. Science, Science, 1969, 17,309-310.

1971, 171, 826-828. (Received September 15, 1970; revisionWITTLIN, W. A., & BROOKSHIRE, K. H. Apomor- received July 12, 1971)

Specific hungers and poison avoidance as adaptive specializations … · 2021. 4. 26. · 460 PAUL ROZIN AND JAMES W. KALAT of the organism. Biologically speaking, there is no reason

Documents