PSYCHOLOGICAL

REVIEW
VOL. 78, No. 6 NOVEMBER 1971

SPECIFIC HUNGERS AND POISON AVOIDANCE AS

ADAPTIVE SPECIALIZATIONS OF LEARNING 1
PAUL ROZIN 2 AND JAMES W. KALAT 3
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 in
thiaraine specific hunger. Consideration of the demands the environment
makes on the rat and the contingencies it faces in the natural environment,
appreciation of the importance of the novelty-familiarity dimension for
these animals, and the realization of two new principles of learning, permit
a learning explanation of most specific hungers. The two new principles,
"belongingness" and "long-delay learning" specifically meet the peculiar
demands of learning in the feeding system. In conjunction with the importance
of the novelty dimension, they are discussed in an attempt to develop the laws
of taste-aversion learning. It is argued that the laws or mechanism of learning
are 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 feeding
system need not be the same as those in other systems; manifestation of a
learning 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 evolution
and development of learning abilities and cognitive function. Full under-
standing of learning and memory involves explanation of their diversity as
welLas the extraction of common general principles.

The application of the basic principles developed in situations and species where
of evolution and adaptation to learning other solutions to the problems at hand
and memory offers a hopeful means of arc less adaptive. Furthermore, when
conceptualizing and ordering the increas- learning or memory capacities are brought
to bear on a
ingly diverse data of these fields. Learning particular type of problem or
and memory, being the result of natura situation, it stands to reason that these
selection, 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 ., • j
of the research reported in it were supported by actenstic, subject to natural selection and
National Science Foundation Grant GB 8013. We therefore adapted to handle specific types
would like to thank John Garcia, Henry and Lila of problems. Insofar as an important
Gleitman Elisabeth Rozin, Martin SeligmanW. J. environmental problem (e.g., obtaining
Smith, 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 find
Rozin, Department of Psychology, University of appropriate modifications of learning and
Pennsylvania, Philadelphia, Pennsylvania 19104. L •!•.• 1 1 i ^ j -^i
'Supported by National Institute of Health memory abilities, closely articulated with
training 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, significant part of our understanding of

there is no reason to assume that there sensory psychology and physiology. Spe-
should exist an extensive set of generally cializations in sensory function, as shown
applicable laws of learning, independent of in the classic works of Lettvin, Maturana,
the situation in which they are manifested. McCulloch, and Pitts (1959) in the visual
This is not to say that there might not be system of the frog, Capranica (1966) on
some general laws, at a minimum resulting the remarkable tuning of the frog auditory
from basic constraints and features of the
system to mating calls, and Roeder (1963)
operation of the nervous system, and
perhaps reflecting general principles of on detection of bat sonar by the moth
causality in the physical world. However, auditory system, can be fully understood
if we look at learning within an adaptive- only in the context of adaptations to
evolutionary framework, we should seek natural problems. The incredible diversity
not only to uncover some of the greatest of the visual system (e. g., variations in the
common denominators among the be- proportion of rods and cones in the retina
haviors we study, but also to explore the as a function of species and ecology) can
plasticity of the mechanisms themselves, be ordered and comprehended in a phy-
as they are shaped through selection to logenetic adaptive framework, as shown
deal with particular types of problems. in the classical work of Walls (1942). We
For many years, the leading ethologists would like to suggest that there are similar
(e.g., Lorenz, 1965; Tinbergen, 1951) have possibilities in the area of learning and
espoused a position consistent with this. memory.
They have emphasized the importance of
considering learning within a naturalistic In the major portion of this paper, we
context; learning is viewed as being shall discuss specific hungers and poisoning
genetically programmed to occur at specific in rats. We shall argue that some basic
points in an ongoing behavior sequence. features of learning and memory as applied
The thesis presented here is in harmony to food selection in the rat are strikingly
with the ethologists' position, but em- different from features characterizing the
phasizes differences in learning mechanisms rat's learning in more traditional laboratory
themselves, as a function of the situations situations, that these differences make
in which learning occurs. sense in terms of evolutionary adaptation,
In the face of recent evidence, coming and that an understanding of the role of
most significantly from the work of Garcia learning and memory in food selection in-
and his colleagues (Garcia, Ervin, & volves an elucidation of specifically adapted
Koelling, 1966; Garcia & Koelling, 1966), learning mechanisms and an integration
challenging two important generalities in of these with genetically determined be-
learning, a framework for ordering the havior patterns.
suggested diversity of laws of learning is In the remaining portion of the paper, we
desirable. We propose such a framework
in terms of the notion of adaptive specializa- shall discuss whether the feeding system
tions. Consideration of learning from the is a unique case or representative example,
adaptive point of view offers two ad- relate our position to somewhat similar
vantages: (a) it provides a heuristic, for theoretical positions, including ethology
ordering and predicting the types of learn- and the recent papers of Garcia (Garcia &
ing one will see in different situations, and Ervin, 1968; Garcia, McGowan, & Green,
(b) it provides a type of explanation of be- 1971), Revusky (Revusky, in press; Re-
havior, in that one aspect of understanding vusky & Garcia, 1970), Seligman (1970),
behavior can be considered to be explica- Lockard (1971), and Shettleworth (1971),
tion of its adaptive value. and discuss the implications of our position
Adaptive considerations have proved in the areas of comparative psychology of
useful in many areas of biology and are a learning and human function.
 ADAPTIVE SPECIALIZATIONS OF LEARNING 461
SPECIFIC HUNGERS learning had to be involved. This served
as the impetus for our investigation of
Basic Phenomenon and Problems with a
specific hungers.
Learning Interpretation
The first demonstration of vitamin B
The phenomena of specific hungers were hunger was that of Harris et al. (1933),
beautifully demonstrated in the classical who found that given a choice of three
work of Richter (1943). The question foods, one supplemented with B vitamins,
raised by this work was, How is the animal B vitamin deficient rats would quickly
(in particular, the rat) able, when deficient come to choose the correct food. For
in a particular nutrient, to select those convenience, Rozin, Wells, and Mayer
foods containing it? In the case of sodium, (1964) decided to study a more simplified
it is quite clear that sodium deficiency re- situation. One vitamin, thiamine, was
leases an innate preference for substances selected, since a clear specific hunger for
containing sodium (Nachman 1962; Rich- thiamine had been reported (Richter,
ter, 1956; Strieker & Wilson, 1970). How- Holt, & Barelare, 1937; Scott & Quint,
ever, it is hard to believe that the rat comes 1946). The most pronounced symptoms
equipped with prewired recognition systems of thiamine deficiency in the rat are weight
for each of the many substances for which loss and anorexia. (More details about this
it can show a specific hunger. The al- and other aspects of thiamine hunger are
ternative view holds that rats can learn available in a review by Rozin, 1967a).
about what foods "make them better" Following injection or ingestion of thiamine,
when they are ill and in this manner come deficient rats show clear signs of reversal
to select the proper nutrients (Harris, of symptoms within a few hours, at most.
Clay, Hargreaves, & Ward, 1933; Scott The basic design consisted of raising wean-
& Quint, 1946). Rats would, in their ling rats on a thiamine deficient diet for
lifetime, learn preferences for the tastes of 21 days, at which point they showed clear
only those foods associated with recovery signs of deficiency. At this point they
from the deficiencies they happen to have were offered a choice between this deficient
experienced. This notion has the distinct diet and the same diet supplemented with
advantage of simplicity in that it accounts thiamine.
for all of the specific hungers (except sodi- Thiamine deficient animals strongly pre-
um) with one basic mechanism. ferred the thiamine enriched choice, while
There was, in fact, some evidence for a control animals who were not deficient, but
learning interpretation of specific hungers. otherwise treated identically, did not. A
Scott and Verney (1947) offered a dis- few rats with a choice between deficient
tinctively flavored vitamin supplemented and highly enriched diet showed no clear
food and an unflavored deficient food to preference during the first few choice days,
deficient rats. After a preference developed but in the meantime ingested great amounts
for flavored, enriched food, the flavor was of thiamine and showed marked recovery.
switched to the deficient food. By and Subsequently, these rats developed a clear
large, rats now preferred the flavored defi- preference for the thiamine rich diet. It
cientfood, suggestinga learning mechanism. seemed highly unlikely that a rat would
The problem with a learning interpreta- show a preference for what it would have
tion, so far as psychologists were concerned, needed a few days ago without showing
was the long delay of reinforcement. The that same preference when the thiamine
recovery (reward) effects produced by an presumably had reinforcing value. This
ingested nutrient would not occur until anomalous observation was confirmed in an
many minutes or hours after ingestion. experiment in which rats were made
This poses serious problems for a learning deficient in thiamine and then recovered
mechanism within the context of traditional from deficiency by injection of high
learning theory. If specific hungers were amounts of thiamine while consuming the
in fact learned, then some new type of deficient diet. After a period of recovery,
462 PAUL ROZIN AND JAMES W. KALAT

the rats were offered a choice of thiamine Reinterpretation of Specific Hungers

enriched and deficient diets. Preferences
for the thiamine enriched diets in rats that The first step in the resolution of the
had been deficient emerged at all the re- problem as formulated came with the
covery intervals studied (Rozin, 1965). appreciation of the importance of novelty
This result raises a second problem for a in the determination of food preferences
learning interpretation of thiamine hunger: in deficient rats (Rodgers & Rozin, 1966).
How could rats develop a learned prefer- Deficient rats show an immediate and
ence for something that at the time the almost complete preference for new foods,
preference appeared, had no known rein- even when the new food is thiamine
forcing effects? If the preference developed deficient and the old food has a thiamine
because of the initial reinforcing effects supplement. In this case, the new-food
of the thiamine, why did it not appear preference reverses after a few days. The
until much later? We shall call this results suggested that the "specific hunger"
problem Preference after recovery. for thiamine may be simply a reflection of
Careful consideration of the specific the thiamine deficient rat's preference for
hungers situation and contingencies gave new foods: If the new food contains
rise to additional problems that could not thiamine, a learned preference could de-
be easily solved within the traditional learn- velop (mechanism still unknown).
ing framework. Given that a rat might Rodgers (1967a) succeeded in demon-
learn with the delays of reinforcement that strating quite conclusively that there is no
seem to hold here, how would the recovery specificity to thiamine specific hunger.
be specifically associated with a particular Reasoning that the novelty response might
food? If the animal had eaten from a overwhelm an existing tendency to prefer
number of the choices available (or both the vitamin, Rodgers offered deficient rats
choices in the two-choice situation), how the choice of a deficient novel diet or the
could one of these choices, presumably the same novel diet supplemented with thi-
one containing thiamine, show a specific amine. The usual strong specific hunger
increase in preference? In other words, did not appear. Furthermore, when the
if an ingestion of two or more foods is choice was between two different novel
followed by recovery, how does one of these diets, one enriched, preferences for the
foods acquire positive properties? We enriched source did not develop rapidly.
shall call this third problem, Which food? Finally, separate groups of thiamine defi-
A logical extension of this notion raises cient rats and pyridoxine deficient rats
the question of how food or the eating of were offered their basal diet in two forms:
food becomes associated with recovery one supplemented with thiamine, the other
when many other potential candidates for with pyridoxine. If there were any spe-
association exist in the environment. After cificity, one would expect each group to
all, following ingestion of foods including show a preference for the vitamin that
the vitamin enriched food, and before would produce recovery. There were no
recovery can ensue, the rat performs many significant differences between the groups.
acts, such as chewing, grooming, exploring, There seemed to be two types of ex-
and sleeping, and receives many stimuli. planations for the novelty effect. One is
By what principle does he elect to respond that the deficiency experience triggers an
to food stimuli instead of other stimuli? innately programmed "neophilia." A more
We call this fourth problem, Why food? attractive hypothesis is that the novelty
We have, then, four problems in the effect comes about through learning. The
application of traditional learning principles paradigms described all pit the new food
to the explanation of specific hungers: against the familiar food that has been
associated with deficiency. The new-food
1. Delay. preference could be a reflection of an
2. Preference after recovery. aversion to the old food. Put more col-
3. Which food? loquially, we could ask whether preferences
4. Why food ? for new foods appear because the rat
 ADAPTIVE SPECIALIZATIONS OF LEARNING 463

"loves the nev stuff or hates the old of recovery, and after a 16-hour period of
stuff." The i:ssue can be clarified by food deprivation, they were presented again
watching the behavior of deficient animals with the familiar deficient food for the
toward different foods (Rozin, 1967b). first time since the onset of recovery. These
Rats housed in relatively large individual food deprived rats, showing no signs of
cages were observed during the first 15 deficiency at this time, responded to the
minutes of an 8-hour food presentation familiar deficient food as they had before,
each day. The r responses were classified with minimal ingestion and occasional
into a number of categories, such as groom- spillage and redirected feeding. In this
ing, sleeping, eating, and chewing. The case, a "normal" hungry rat prefers staying
experimental ;ts were placed on a thi- hungry to eating its original diet which
amine deficient diet. As deficiency de- provides, in fact, perfectly adequate nutri-
veloped, less time was spent eating during tion for him at the time, and is normally
the first 15 mirutes, and two striking be- quite palatable. Preference of hunger
haviors, quite :-are in normals, appeared. (eating nothing) to ingestion in hungry
One was spillag< of the food: the rat would rats and the similarity in the rat's be-
approach the food cup, sniff at the food, havior toward deficient and highly un-
and then paw at the food in a scooping palatable (quinine adulterated) diets sug-
motion, spilling it out of the cup, so that gests strongly that we are dealing with an
it fell through l[he wire mesh floor. (This aversion to the familiar food.
spillage phenon.enon in deficient rats was The learned aversion interpretation
actually quite familiar, since we have places specific hungers in a new perspec-
always had some difficulty measuring the tive. We can consider the deficient diet
food intake of deficient rats on this ac- as a CS and the nausea or other ill effects
count.) The otaer new behavior was what produced by its ingestion as a UCS.
we call "redire :ted feeding." Following Presumably, the classically conditioned
this initial investigation of the food cup, "ill effects" lead to avoidance of the
rats would occasionally move over to the familiar food. The mechanism suggested
wooden barrier separating the nest area for the specific hunger phenomenon, as
from the rest of | the cage and begin to chew seen in the standard two-diet choice, would
it vigorously. iThey would also, occasion- then be that the rat learns an aversion to
ally, chew on the cage wires. These two the familiar deficient food. Before the
behaviors suggested an aversion to the time of choice, he has already done a most
familiar diet. The redirected feeding sug- significant part of his learning: He knows
gests conflict between desire to eat and what not to eat. The initial preference
the aversiveness of the food offered. for the new food follows. Its maintenance
Spillage is often seen in normal rats with when the new food is enriched could be
highly unpalatable foods, such as quinine accounted for as an additional learned
adulterated diet 3. preference for the new food or as a failure
When offered! the old deficient food in a to develop an aversion to it (see section
new container (metal instead of glass) in on learned preference).
a new location, (these deficient rats showed To the extent that specific aversions
little potentiati<j>n of eating and continued play a key role in specific hungers, there
to show the behavior described above. is an obvious parallel between specific
Apparently the vessel and its location were hungers and poisoning. Both involve
not controlling the aversion. However, learned aversions; vitamin deficient diet
when a completely new deficient food was is a slowly acting poison. The aversion
offered in the familiar vessel and location, experiment thus suggests that these two
uninterrupted avid eating ensued, sug- sets of phenomena are closely related.
gesting that the aversion was specific to Furthermore, the same basic problems
the food. The deficient rats were sub- raised here (Delay, Which food? Why
sequently allowed to recover on a new, food?) arise with respect to poisoning.
vitamin enriched diet. Following one week Since poisons are designed and synthesized
464 PAUL ROZIN AND JAMES W. KALAT

by man, it is unreasonable to hold that showed an increased preference for the

rats show innate aversions to them. In familiar safe food, that is, a neophobia,
comparing the literature on poisoning (see Half-wild rats showed a stronger neophobia
Barnett, 1963; Richter, 1953; Rzoska, following the aversive experience, but
1953) to that on specific hungers, there is half-wild controls also showed a higher
one apparent contrast. We have reported base-line neophobia. All experimental rats
an increased preference for new foods in almost completely avoided the familiar
white rats, while the poisoning literature aversive food (B), but ate some of the
which focuses on wild rats strongly in- completely new food. There were no
dicates the opposite: an exaggerated neo- major differences between the specific
phobia following poisoning. That is, wild hunger and poisoning groups. Therefore,
rats, who show a much greater base-line it appears that we can consider specific
tendency to avoid new objects or events hungers as a parallel to poisoning. Again,
than do white rats (e.g., see Galef, 1970), the behavior makes sense in an adaptive
show a further exaggeration of this tend- framework: following an unpleasant food-
ency, often to an extreme (Richter, 1953) related experience, the rat tends to return
following poisoning experiences. This dis- to a known, safe food.
parity can be accounted for as a pro-
cedural difference, since the new-familiar Resolution of the Inconsistencies with a
choices we offered to our domestic rats Learning Interpretation
were different from those usually offered
poisoned wild rats. In particular, in the The specific aversion explanation of
novelty experiments, which were done specific hungers, the realization of the
before we realized that specific aversions importance of the familiarity-novelty di-
were involved, the rat was offered a choice mension, and the appearance of two
between a familiar food associated with articles of major significance by Garcia
deficiency and a new food. Even if the and his colleagues enable us to resolve
white rat were somewhat neophobic, this the four basic problems with a learning
might not appear since the alternative interpretation.
choice was a strongly aversive familiar diet. Preference after recovery. This can be
In order to provide a meaningful com- accounted for as a retained aversion to the
parison between poisoning and specific familiar food. Rats made deficient in
hungers, both sets of phenomena would thiamine and then recovered by injection
have to be demonstrated under the same for 12 hours or 5 days (while continuing
sets of conditions and in the same strain to eat the same deficient diet) show a
of rats. In an experiment meeting these strong initial novel-food preference when
criteria and comparing half-wild and do- offered a novel-familiar food choice, with
mestic rats, a paradigm was employed that thiamine absent from both choices (Rozin &
allowed fuller expression of the rat's neo- Rodgers, 1967). In these experiments,
phobic or neophilic tendencies. Rats were recovery took place in the presence of the
raised on Diet A, prior to induction of familiar food. The existence of a novelty
aversive consequences. Deficiency or poi- effect suggests that the aversion had not
soning (or nothing in the case of controls) fully extinguished by the time of testing.
occurred in the presence of Diet B. In It is significant in this regard that in most
the final test, rats were offered these two of our experiments (some unpublished)
diets and a completely new one, Diet C. the recovery effect showed some diminu-
Therefore, rats were faced with a choice tion as time of recovery increased. The
among a familiar safe diet (A), a familiar recovery effect should be maximal when
aversive diet (B), and a completely new recovery is associated with a new food, so
diet (C) (Rozin, 1968). that the old diet aversion does not ex-
The single important result is that tinguish. Strong "recovery" effects are
all rats suffering poisoning or deficiency in fact seen under such circumstances in
 ADAPTIVE SPECIALIZATIONS OF LEARNING 465

the aversion experiment, with a single Feeding was restricted to eight hours a day,
choice (Rozin, 1967b) and in typical and the food intake from each cup was
two-choice experiments with long recovery measured at hourly intervals. Four of the
periods (Rozin, unpublished data). 10 deficient rats studied developed clear
Which food? In the "standard" two- preferences for the enriched choice within
choice situation, this problem is easy to a few days, and two other rats developed
deal with; the rat learns to avoid the strong preferences for two of the four
deficient food when it is the only diet choices, where one of the preferred choices
available. There is no Which food? was enriched. Analysis of the meals in-
problem here. Because of this learned dicates a characteristic pattern, both on a
aversion, rats show an immediate, vir- daily and hourly basis. Means, except for
tually complete preference for the novel an initial daily sampling of many or all of
food, thus making it possible for them to the choices, tended to be restricted to one
learn about its consequences. food. The rat's feeding pattern maximizes
The testing situations described up to the possibility of associating each diet
this point, with the partial exception of with its appropriate consequences, since
the testing environment in the aversion meals tend to be isolated in time and to
experiment, have been very limited, well consist of a single food. The emergence of
defined, and unnatural. Rats normally a strong preference for the enriched food is,
live in a much larger area than was per- in each case, preceded by a clearly defined
mitted to them in these studied, have an meal of that food. Furthermore, no rat
elaborate social life, and seldom face failed to develop a clear enriched food pref-
simple binary food choices. It is not un- erence if it ate an isolated meal of at least
reasonable to assume that one or only a 1 gram from the enriched food. The rats
few foods might be available to the rat that failed to show adaptive preferences
during the deficiency period in nature, in the initial part of the experiment failed
since with a wide variety of foods available, to sample significant amounts of the en-
a deficiency would be quite unlikely. In riched food. Similar, though less well
their classic work on B vitamin hungers, defined, sampling patterns are seen in
Harris et al. (1933) offered deficient rats normal rats. Normals mix meals more,
a choice among a large variety of foods, but this may be a direct consequence of the
with only one containing B vitamins in fact that their meals are larger. The
significant amounts. By and large, they anorexia of deficiency may, in and of itself,
found that with large numbers of choices exaggerate in an adaptive way the normal
(6-10 foods), rats were unable to select feeding pattern of the rat. It is clear that
the vitamin enriched source, and they we have here another part of the answer
found it necessary to "educate" the rats to the Which food ? problem.
by offering only the enriched food for a Finally, social factors have become im-
period of days in order to establish a main- plicated in food selection in some sig-
tained preference for this food. Examina- nificant recent experiments. Galef and
tion of the day-to-day intake of these Clark (in press) studied responses to poison-
animals before the education period sug- ing in colonies of wild rats observed in the
gests a distinctly nonrandom food selection laboratory under seminaturalistic condi-
pattern: food intake seemed to be restricted tions. They offered a group of wild rats
to only a few of the many choices on any two foods, one of which was poisoned.
particular day. This seemed to be an After a short period of time, all rats
adaptive way of unconfounding the situa- avoided this food. The poison was re-
tion and suggested an analysis of the meal moved, but the rats still completely
patterns of deficient rats faced with the avoided the base. A litter was born, and
choice of a number of new foods (Rozin, the behavior of the young toward these new
1969a). A thiamine supplement was placed foods was observed. The parents and
in one of four diet choices for each rat. young were fed for three hours a day and
466 PAUL ROZIN AND JAMES W. KALAT

were constantly under observation during closer to solving this important problem.
this period. During a two-week period It does not appear to be possible to rein-
(Days 14-28 of life in the pups), the pups terpret what appeared to be long-delay
came out to eat, but ate only the safe learning in terms of short delays. However,
food (possibly because this was the only a critical experiment by Garcia et al. (1966)
food being eaten by the parents). When demonstrates that long-delay learning can
the young were fully weaned and placed occur in this system. They induced an
in a new cage, separated from their parents, aversion to saccharin in rats by injections
they continued for a six-day period to eat of apomorphine, a drug that presumably
only the safe diet and ignore the formerly produces gastrointestinal upset. Aversions
poisoned diet. Further experiments (Galef were produced when the interval between
1971; Galef & Clark, in press) suggest a termination of drinking and injection of
straightforward explanation for this effect. the drug was one-half hour or more. Only
Rat pups tend to approach areas where a few trials were necessary to establish a
adults are located and begin feeding there. clear aversion.
In this way, they become familiar with the
foods eaten by the adults and avoid al- A New Problem—Are There Learned Pref-
ternative diets as a result of their neo- erences as well as Learned Aversions!
phobia. More recent work (Galef & Clark,
1971; Galef & Henderson, in press) sug- Before we leave the problem of specific
gests an additional mechanism: chemical hungers, there is one new issue to resolve.
cues from mother's milk seem to function Much of the specific hunger phenomenon
to familiarize the young with certain char- can be explained in terms of aversion; Is
acteristics of the mother's food and produce there a positive side also ?
initial postweaning preferences. Up to this point we have provided
Why food? Rats in the aversion experi- evidence for three categories of foods based
ment (Rozin, 1967b) did not show an on the animal's experience (Rozin, 1968).
aversion to the container or the location of These are unexperienced or novel, familiar-
the deficient food. While introduction of safe, and familiar-aversive. The question
a new food dramatically increased inges- is, Is there a fourth category, familiar-
tion, change to a new location and con- positive? That is, is the experiencing of
tainer did not. Apparently what the positive consequences any different from
animal learned was specific to the food. the experiencing of neutral consequences
In an independent and much more com- when these experiences are contingent upon
pelling experiment, Garcia and Koelling ingestion of a particular substance?
(1966) provided evidence that there was a Much of the evidence which appears to
specific tendency for taste "CSs" to be demonstrate positive preferences can be
associated with certain types of visceral reinterpreted in terms of learned aversions.
"UCSs," while exteroceptive CSs such as For example, Harris et al. (1933) found
light and sound would be preferentially that when only one of a large number of
associated with exteroceptive UCSs, such diets contained adequate thiamine, most
as shock. They paired light, sound, and deficient rats were not successful in select-
taste simultaneously with either electric ing the correct diet. However, if the rats
shock or poisoning in different groups of were "educated" by being offered only
animals. The shocked animals developed the adequate food for several days, they
an avoidance of the light and sound, but showed a preference for this food when the
not of the taste. The poisoned animals, larger number of diets were again offered.
subjected to the same light-sound-taste This experiment does not distinguish be-
pairing, avoided the taste and not the tween learned-preference and learned-aver-
sound or light. sion interpretations. Rats could be avoid-
Delay. The reinterpretation of specific ing the nonthiamine diets by a combination
hungers, in itself, does not bring us much of learned aversions and neophobia. That
 ADAPTIVE SPECIALIZATIONS OF LEARNING 467

is, after "education" the rats have learned demonstrate that food with clear positive
that the thiamine-containing diet is "safe," consequences would be preferred to foods
but it is not certain whether they have with relatively neutral consequences. In
learned that it has any special "recovery his simple design, rats were fed one nutrient
from deficiency" properties which dis- solution while hungry and a different one
tinguish it from other safe diets, since we when satiated. After five days of this
know that deficient rats prefer "old safe" training, a significant preference developed
foods to either old aversive or novel foods in a two-bottle test for the solution drunk
(Rozin, 1968). Similarly, in the sampling while deprived. This result is interpreted
experiment (Rozin, 1969a), the fact that in terms of the greater (delayed) rein-
the rat eventually prefers the only en- forcing effect of the solution drunk during
riched diet in a four-choice situation deprivation. Since both solutions were
could mean that he had developed aver- equally familiar and the rats drank the
sions to the other choices. solution offered in satiated conditions
There are some recent experiments voluntarily (so that an acquired aversion
which present more serious challenges to a would be unlikely), this experiment meets
pure aversion model. Garcia, Ervin, the requirements for demonstrating a
Yorke, and Koelling (1967) repeatedly learned preference. The effect is clear,
made rats thiamine deficient and produced though not large by comparison to the
recovery by thiamine injection. Just prior aversion data (see Revusky & Garcia,
to thiamine injection, rats drank saccharin 1970, for additional data).
solution. At all other times, water was It is noteworthy that the positive prefer-
the only fluid available. When thiamine ence effects reported have been rather
deficient, the rats showed an increase of small by comparison with learned aversions,
saccharin intake over trials, both with and difficult to obtain (Kalat & Rozin,
respect to their own water intake and to unpublished; Revusky & Garcia, 1970).
the saccharin intake of similarly treated We cannot satisfactorily explain this asym-
animals where thiamine injection was not metry. Possibly the rat is better prepared
contingent on saccharin ingestion. Camp- to learn aversions because rapid learning
bell (1969) has also demonstrated that there has particular survival value; that is,
thiamine deficient rats show an increase in mistakes are very costly. This remains, for
their absolute intake of a sucrose solution the moment, an intriguing problem, with
which has been associated with recovery possibly great implications for the regula-
from deficiency. Zahorik and Maier (1969) tion of food intake.
used Garcia et al.'s (1967) procedure but
with a modified test and found that rats Other Specific Hungers
prefer the taste associated with recovery We have offered a theory of thiamine
to both the taste associated with deficiency specific hunger. We believe that it holds
and a novel taste. Furthermore, this for other learned specific hungers as well.
preference was apparent in both deficient Novel-food preferences, which imply a
and recovered rats. learned aversion mechanism, have been
These experiments, which certainly pro- demonstrated in calcium, sodium, and
vide evidence for learned preferences, can magnesium deficient rats (Rodgers, 1967a)
nonetheless be explained in terms of the and in pyridoxine and riboflavin deficient
three basic categories of food. A more rats (Rozin & Rodgers, 1967). The pre-
decisive experiment would have to show ference after recovery phenomenon appears
that rats prefer a "recovery" solution to in identical form in thiamine, riboflavin,
an old "safe" solution, that is, one they and pyridoxine deficiency (Rozin & Rod-
drank without aversive consequences at gers, 1967).
a time when then were not vitamin deficient. It seems reasonable to assume that the
Revusky (1967; Revusky & Garcia, anorexia characteristic of most vitamin
1970) performed a series of experiments to deficiencies reflects, at least in part, a
468 PAUL ROZIN AND JAMES W. KALAT

learned aversion to the deficient food. This WHY Is SODIUM HUNGER INNATE?
is dramatically clear in the case of thiamine
An alternative to the mechanism we have
deficiency, where the anorexia symptom
discussed is an almost completely genetic-
disappears precipitously when a new diet
ally preprogrammed specific hunger. So-
is offered. On the other hand, there is
dium hunger seems to be such a case.
relatively little anorexia in vitamin D or
Rats sodium deficient for the first time
vitamin A deficiency, and in both cases
show a strong preference for sodium im-
there has not been a clear demonstration of
mediately upon tasting it (Nachman,
a specific hunger.
1962; Richter, 1936, 1956; Strieker &
There is an impressive body of research Wilson, 1970). They also show strong
on amino acid specific hungers and re- preferences for similar-tasting salts which
sponses to amino acid imbalance (Harper, will not actually alleviate their deficiency
1967). These "deficiencies" are char- (Nachman, 1962), and they will work
acterized by anorexia and seem to fit into at a high rate in extinction at a lever which
the scheme we have described. Recently, previously produced a salt solution which
Rogers and Harper (1970) presented evi- they drank while sodium replete (Krieck-
dence for a positive preference for a solu- haus& Wolf, 1968).
tion that corrects an amino acid imbalance. We do not know why sodium differs
We cannot complete this reconsideration from other specific hungers, but at least
of specific hungers without mentioning one three explanations are available. First,
particularly serious shortcoming of all the the great importance of sodium in body
mechanisms we have discussed. Although, fluid balance, its common scarcity in the
in principle, they can account for most of environment (witness salt licks and salt
the individual vitamin or mineral specific wars), and the large amount needed by the
hungers, it is not clear how the classic individual organism might place particular
"cafeteria" self-selection of Richter (1943) selection pressure in favor of a surefire
can be accounted for. Rats appear to sodium hunger. The various physiological
adaptions directed toward regulation of
self-select an extremely well balanced diet.
body sodium are impressive. Second, it is
Unless we assume that they develop conceivable, as Rodgers (1967b) has sug-
incipient deficiencies of a variety of nu- gested, that sodium ingestion by sodium
trients and learn aversions and prefer- deficient organisms may have initial nega-
ences on the basis of these minimal symp- tive effects8 associated with large-scale
tons and their abatement, we have no ex- electrolyte shifts. There is not much
planation of this remarkable phenomenon. evidence on this point. Sodium deficient
At this time, long-delay learning mech- animals do develop an aversion to sodium
anisms appear inadequate to the task, deficient diet (Rodgers, 1967a). However,
since we cannot identify obvious candidates injections of sodium into sodium deficient
for unconditioned stimuli.4 animals have not successfully produced
preferences for foods ingested just prior
'Another remaining problem concerns the diffi-
culty in demonstrating specific hungers with the to injection (Rodgers, 1967b). These
nutrients in an aqueous solution (Appledorf & possible initial negative events may pre-
Tannenbaum, 1967; Rozin, Wells, & Mayer, 1964). vent the operation of the learning mech-
This contrasts with Richter, Holt, and Barelare's
6
(1937) dramatic demonstration of a specific hunger Scott, Verney, and Morissey (1950) noted that
for thiamine in solution. We are inclined to believe magnesium deficient rats selected magnesium defi-
that the difficulty frequently encountered with cient food in preference to magnesium enriched food.
aqueous solutions derives from pure water being one Rodgers (1967a) confirmed this "inverse" specific
of the choices available. Water's great familiarity hunger and suggested that the initial ingestion of
might protect it from becoming aversive (Garcia & magnesium by magnesium deficient rats produces
Koelling, 1967; Maier, Zahorik, & Albin, 1971; aversive consequences. Here, those consequences
Nachman, 1970a). Palatability differences in the might be considered as "coming down" from the
various choices solutions might also be involved hyperirritability or "high" characteristic of mag-
Rogers & Harper, 1970). nesium deficiency.
 ADAPTIVE SPECIALIZATIONS OF LEARNING 469

anisms normally implicated in specific they apparently had no innate recognition

hungers. Third, sodium as a stimulus is of water. For example, they would run
well defined by the taste system. It would through water trays without drinking. At
be easy for a genetic program to tie into some point in the first few days of life,
the proper class of stimuli. (Conversely, the bird would ingest water by pecking at a
one might argue that the fact that sodium water drop or at an irregularity on the
recognition seems to be so fundamental to surface or in the bottom of the water pan.
the taste system suggests again the great As soon as water entered the bird's mouth,
importance of sodium as a directive force it would adopt characteristic drinking
in evolution.) Although a unique learning movements, and after only one or two
ability might have evolved to meet this such experiences, would show a clear visual
particular problem, the actual genetically recognition of water. Hunt and Smith
programmed solution seems preferable. (1967) repeated and extended these results.
However, important biological functions are This system is highly predetermined: all
often overdetermined; this may be one ex- components seem to be genetically pro-
ample. It has been shown that normal grammed, except the recognition of the
rats can learn the location of sodium visual stimulus for water. Even a reason-
sources in their environment and go to them ably precise regulation of water intake is
promptly when a sodium need is induced programmed, ready to come into operation
(Krieckhaus & Wolf, 1968). as soon as water is experienced. Strieker
But now we must turn the question and Sterritt (1967) have shown that on its
around and ask, Why are all specific very first drink, a chick's water intake is
hungers not innate? One basic learning proportional to its deficit. We suggest
mechanism can provide an adaptive solu- that the visual recognition of water is
tion for almost all specific hungers (and learned because this is an efficient way of
also poisoning). Instead of programming accomplishing the task at hand. The situa-
specific innate recognitions of a variety of tion in the environment is such that a bird
nutrients, the organism can rely on the looking for food (defined as small irregu-
fact that the malaise produced by most larities) will certainly end up with water in
deficiencies (and possibly the consequent his beak, and since he is prepared to learn
recovery on ingestion of the needed nu- this and has all of the other genetic equip-
trient) is adequate to establish aversions ment necessary to handle water ingestion
(and preferences). There remains the and balance, he is "home free." We em-
fascinating question of whether long-delay phasize that this mechanism is successful
learning originally evolved in response to because the bird's normal pecking and
proper selection of nutrients or the regula- eating behavior (genetically programmed)
tion of food (caloric) intake, or both. invariably brings it into a position where
Sodium hunger raises the general issue water enters the beak. Hunt and Smith
of the factors influencing the role learning (1967) show that the first time a bird pecks
will play in solving a particular problem. at a dew drop (likely to be his first ex-
With alternative mechanisms available to perience with water in the real world), it
deal with an environmental problem, the responds with a "feeding" peck, which is
solution achieved by a species should be characteristically different from an ap-
determined by factors of economy (cost) proach to water. Only after the water
dictated by the actual situation and fea- enters the beak does drinking-type be-
tures of the development of the organism. havior appear, and subsequently, the peck
We can see the interplay of these factors at dew drops can be seen to be qualitatively
in the case of what Strieker (personal different from the peck at food.
communication, 1971) has pointed out is We are suggesting that there are two
another specific hunger: thirst for water. types of advantages (and therefore evolu-
Morgan (1894) studied the development of tionary pressures) to learned solutions to
drinking in young chicks and found that problems. The first, and most obvious,
470 PAUL ROZIN AND JAMES W. KALAT

has to do with the variable environment. tion are not what they were thought to be.
In the face of a changing and variable en- (b) The novelty-familiarity dimension is
vironment, it would be maladaptive to pro- of particular importance to the rat. (c)
gram specific modes of response, since these To account fully for the phenomenon, two
would often be inappropriate. Clearly, a new "principles" of learning were needed:
plastic organism can do better in a plastic belongingness and long delay. It is re-
environment. (We expect that mono- markable that each of Garcia et al.'s two
phagous animals are less capable of learn- "classic" experiments (Garcia & Koelling,
ing about food than omnivorous ones.) 1966; Garcia et al., 1966) dealt directly
The second factor predisposing to learning with one of these problems at just the time
has to do with the "cost of preprogram- that these two issues became the two prob-
ming." This has two components. First, lems in specific hungers.
preprogramming of a particular behavior or In the context of the problems of specific
stimulus configuration preempts a portion hungers, it seemed clear that the basic
of the total genetic material, which could principles demonstrated in these experi-
otherwise be used for some other purpose. ments must be essentially correct. Both
Second, there may be greater costs as- principles, belongingness and long-delay
sociated with the evolution of genetic- learning, seem highly adapted to the pro-
ally programmed solutions, insofar as perties of the feeding system. Tastes are
built-in solutions involve more genetic in fact causally linked to gastrointestinal
reorganization. events, and there is a long inherent delay
Of course, in some cases, a genetically between the taste and its consequences.
programmed recognition may be replaced We suggest that specific learning mech-
by a learning mechanism. If an animal anisms have evolved in response to specific
had the capability of learning something problems. In the following section, we
that was genetically programmed, other- shall consider taste-aversion learning in
wise favorable mutations in the genetic greater detail, in order to demonstrate the
material responsible for this behavior could extensive ways in which learning mech-
be selected for, even though the genetic anisms may be modified to satisfy particular
basis for the behavior would be destroyed. demands.
Given a general learning capability, it
could often be cheaper to allow the en- TASTE-AVERSION LEARNING: AN
vironment to supply the appropriate stimu- EXAMPLE OF ADAPTIVE SPECIAL-
lus configurations, even if these will not IZATION OF LEARNING
vary much throughout life. We suggest
that this "balance sheet" type of approach In this section we shall consider learning
may be fruitfully applied to all specific about food. In considering whether it is a
hungers. In the case of the chick, we sug- "new type" of learning, we shall examine
gest that it is necessary to genetically both the belongingness and delay principles
specify those portions of the feeding and in some detail, consider the importance of
drinking system which are distinctively familiarity and novelty, and explore other
different (the ingestive responses and possible differences between learning in
regulation). Since the chick's behavior the feeding system and more traditional
leads to water in the beak anyway, be- types of learning.
cause the preprogrammed food recognition
overlaps with this aspect of drinking, a Principles of Stimulus Selection
cheap way out can be found. The costs can presume, from the material on
are minimal, and the genetic savings, specific hungers, that when faced with a
significant. bewildering array of stimuli as candidates
SUMMARY for association with a gastrointestinal
The reanalysis of specific hungers rests event, the rat has available principles with
on three fundamental points: (a) The which to sort them out. One concerns his
actual contingencies in the feeding situa- past experience with these stimuli, the
 ADAPTIVE SPECIALIZATIONS OF LEARNING 471
novelty-familiarity dimension, and the substances to eat and which to avoid.
other certain presumably built-in tend- However, an equal ability to associate
encies to associate certain categories of lights and sounds with gastrointestinal
stimuli with certain relevant events (be- consequences would be far less adaptive;
longingness). We shall consider each in in fact, the common result would be
turn. "superstitious" learning. The belonging-
ness phenomenon receives support not only
Novelty-Familiarity Dimension and from these adaptive arguments but also
As sociability from neurological considerations. The
gustatory receptors and the gut receptors
We have argued that the novelty (or are similarly classified as visceral sensory
familiarity) of a stimulus is of particular inputs and show close neurological re-
importance for a rat (see Galef, 1970; lationships, specifically in the medulla and
Rozin, 1968). This distinction has par- possibly in the hypothalamus.
ticular significance in determining the It appears that under some circum-
magnitude of a learned aversion to a given stances, exteroceptive cues can become as-
taste. Revusky and Bedarf (1967) and sociated with poisoning (Garcia, Kimel-
Wittlin and Brookshire (1968) showed that dorf, & Hunt, 1961; Rozin, 1969b).
rats learn aversions much more readily to Rather rapid learning of aversion to loca-
novel than to familiar solutions, even when tions and shape of drinking vessel can be
the familiar solution is drunk after the demonstrated if the UCS (apomorphine)
novel solution (and, of course, before is introduced during drinking from the
poisoning). In these experiments, familiar- appropriate vessel (Rozin, 1969b). On the
ization occurred over a period of days, but, other hand, even with a modest 30-minute
in fact, a 20-minute exposure to a solution CS-UCS interval and the use of a con-
followed by neutral consequences will pro- catenation of nontaste cues including
duce virtually the same effect, for such a vessel shape, location, solution texture, and
solution is then quite resistant to associa- temperature, virtually no aversion appears
tion with poisoning (Kalat & Rozin6). after many trials (Rozin, 1969; but see
This minimal (single) experience can have Nachman, 1970a, for a demonstration that
occurred three weeks before the poisoning under proper circumstances oral tempera-
without significant attenuation of the ture—a nontaste but interoceptive cue—
effect. alone may be an effective cue). The sug-
gestion here is that at least part of the
Belongingness "weakness" of exteroceptive cues derives
The "belongingness" (Garcia & Koelling, from a very rapid decay of their associ-
1966) or "preparedness" (Seligman, 1970) ability with time.
or "stimulus relevance" (Capretta, 1961) The belongingness principle in relation
phenomenon—that is, the tendency to as- to taste-aversion learning is elaborated by
sociate tastes with aversive internal con- Garcia and Ervin (1968) and Garciatet al.
sequences as opposed to associating either (1971). Further extensions of the principle
element with anything else—seems emi- to areas outside of feeding have been dis-
nently sensible from an adaptive point of cussed by Seligman (1970) and Shettle-
view. Gastrointestinal and related internal worth (1971).7
events are, in fact, very likely to be in- 7
The principle of belongingness makes some pre-
itiated or influenced by substances eaten, dictions about what aversive stimuli should be em-
and taste receptors, by virtue of their loca- ployed in various types of aversion therapy. For
tion, provide information about these same example, it would be expected that nausea-producing
drugs such as apomorphine should be more effective
substances. It is essential for an omnivore than electric shock in treating alcoholism. At the
to have the ability to learn rapidly which present time, the data are not clear on this point
6
(Rachman & Teasdale, 1969). The issue is com-
J. Kalat and P. Rozin. Learned safety as an plicated because it may be the case that humans
explanation for taste-aversion delay of reinforcement learn different things with electric shock or nausea
gradients. In preparation. as UCSs.
472 PAUL ROZIN AND JAMES W. KALAT

Salience. There is evidence for "intra- outside of the gastrointestinal system. On

modality" belongingness. Rats tend to the other hand, it is hard to understand,
associate poisioning with some novel solu- from an adaptive point of view, why pain
tions more than others (Kalat & Rozin, in the lungs or heart, for example, should
1970). Rats drank one novel solution be selectively associated with taste. As a
briefly, 15 minutes later drank a second matter of fact, it is not known whether
novel solution, and another 15 minutes gastrointestinal pain is selectively as-
later were poisoned. The following day, sociated with taste. Much more research
the rats were offered both solutions simul- is needed to better define the range of
taneously. Under these conditions, certain visceral sensations with which tastes have
solutions, which we describe as highly high associability. The extensive Russian
salient, became more aversive than others. literature on interoceptive conditioning
The salience of a solution proved to be a (see Bykov, 1957) demonstrates that ex-
more potent predictor of amount of teroceptive cues can be attached to visceral
acquired aversion than temporal proximity UCSs. Whether there is some belonging-
to poisoning. It was found possible to ness operating here remains to be seen.
rank novel solutions in a stable, transitive Yet another dimension of belongingness
"salience" hierarchy, such that each solu- concerns the temporal parameters of the
tion proved more salient (associable with various CSs and UCSs. In the taste-
poisoning) than all solutions lower on the visceral system, stimuli tend to have slow
list. Evidence for a salience effect on the onsets and to last for long periods. Ex-
"positive" side (recovery from thiamine teroceptive CSs and UCSs, in contrast, are
deficiency) has been reported (Campbell, characteristically brief and well defined in
1969). time. The importance of these dimensions
As yet it is not known what constitutes and the visceral field for taste-aversion
the denning characteristics of the salience learning could both be determined by using
dimension. It is probable that the "rela- experimentally controlled UCSs, such as
tive novelty" of these operationally novel electric shock to the stomach mucosa in-
solutions contribute to the effect. By this stead of the ill-defined poisoning procedures.
we mean that more salient solutions may The category of effective CSs should
be more different from previously ex- differ from species to species. It was
perienced solutions. There may be addi- argued above that the specific ability to
tional factors associated with intrinsic prop- learn taste-poison associations is highly
erties of the solutions. Certain tastes (e.g., adaptive because foods are the usual cause
bitter) may have a particular tendency to of any aversive gastrointestinal stimula-
be associated with aversive consequences. tion. This argument assumes that the
A functional definition of belongingness, animal identifies its food by taste. How-
At this time we know that some category ever, some species use other modalities
of CSs, including tastes, preferentially as well. In particular, there is reason to
associates with some category of UCSs, in- believe that birds put main emphasis on
cluding "gastrointestinal distress." We visual cues in the identification and selec-
would like to define these categories more tion of food. A number of experiments by
precisely. The unconditioned stimuli em- Brower (1969) indicate that birds can
ployed have been described variously as readily learn aversions to the sight of
"poisoning," "nausea," and "gastrointesti- food,8 and experiments by Wilcoxon, Dra-
nal upset." But the category of effective 8
One type of mimicry is based on the fact that
stimuli may include stimuli not suggested many birds and some other species readily learn to
by these terms. Ingested foods can cer- avoid toxic or unpalatable prey. In these cases, a
tainly produce significant internal effects safe, palatable species evolves coloration similar to
outside of the gastrointestinal system, and that of an unsafe or unpalatable one and thereby
obtains some protection due to the predator's
it is indeed possible that the primary action generalization of its learned aversion (Brower, 1969;
of some of the UCSs presently employed is Rothschild, 1967; Wickler, 1968).
 ADAPTIVE SPECIALIZATIONS OF LEARNING 473

goin, and Krai (1971) indicate that Jap- species in which food recognition is known
anese quail learn poison-based aversions to occur partly via nongustatory cues.
more readily to the color (or optical For instance, frogs have specific visual cells
density) than to the taste of a solution. which respond maximally to flying insects
Rats and bobwhite quail were poisoned and similar stimuli (Lettvin et al., 1959).
half an hour after drinking a solution The above analysis suggests that stimuli
which was either blue, sour, or both blue which excite these "bug detectors' might
and sour. The rats subsequently showed be more easily associated with poison than
considerable aversion to the sour water, other types of visual stimuli.
but none to the blue (dense) water. The
quail, however, showed an aversion to LONG-DELAY LEARNING
both; furthermore, the quail poisoned on
blue-sour water generalized their aversion Until recently it appeared that learning
to blue water and not to sour water. To of both classical and operant types was
show that the aversion to blue water was almost impossible with delays of rein-
not based on the taste of the blue coloring, forcement exceeding a few seconds. The
the investigators showed that quail could importance of close temporal contiguity
form an aversion to plain water in a dis- has been demonstrated in a considerable
tinctively colored tube. variety of experimental paradigims, and
We suggest that the critical dimension apparent exceptions seemed to depend on
for poison-based aversion learning may secondary reinforcement (Kimble, 1961).
not be "taste versus other modalities" but There were cogent theoretical and adaptive
"eating-related cues versus other cues." reasons for assuming the universal im-
This type of functional categorization of portance of close temporal continguity.
input is in harmony with Gibson's (1966) Premonitions existed, however, of a pos-
view of perceptual systems. Eating-related sible problem with learning about food
cues would be whatever type of cue— ingestion.
gustatory, visual, or otherwise—a particu-
lar species uses to identify food. Because History
of the intimate and relatively invariant The first such reference is Pavlov (1927).
relationship between taste receptors and Using morphine as a UCS, Pavlov noted
food ingestion, and because of the neuro- signs of nausea and profuse salivation in
logical association between taste and vis- response to the touch of the experimenter,
ceral receptors, it is likely that taste cues which preceded injection. Needless to say,
would always be classified as eating- the UCS (i.e., nausea, etc.) did not occur
related so that taste-poisoning belonging- until many minutes after the morphine was
ness should be practically universal. How- administered. Pavlov evidently regarded
ever, in some species, other modalities the effect as a species of conditioning, thus
may also be eating related. implicitly accepting the possibility of
Nontaste sensory modalities could in- learning over long delays:
clude both eating-related and non-eating-
This experiment provides a clue to the well-known
related cues. For instance, it may be that fact that dogs will eat meat the first time it is
birds can form poison-based aversions to offered them, after removal of their parathyroids, or
the sight of potential prey, but not to after an Eck fistula and tying of the portal vein, but
other sights. Also, it is known that odors on all subsequent occasions will refuse it. Evidently
are less effective than tastes in poison- in these cases the appearance and smell of meat
produce of themselves a reaction identical with that
based aversions in rats. Possibly odors produced through direct pathological action in the
are more effective when they emanate from absence of the parathyroids or the portal circulation,
a food source or if they are experienced by those toxic substances resulting from digestion of
the meat [p. 36].
simultaneously with a taste. It would be
interesting to investigate the effective cues A similar recognition of the phenomenon
for poison-based aversion learning in other of long-delay learning is present in the
474 PAUL ROZIN AND JAMES W. KALAT

biological literature on poisoning, regula- one hour. Smith and Roll (1967) found
tion of food intake, and specific hungers. similar results using X rays and either
Thus, Harris et al. (1933), Scott and saccharin (up to 12-hour delays) or sucrose
Verney (1947), Rzoska (1953), and Le- (up to 6-hour delays). Replications by
Magnen (1969) explicitly implicated learn- Revusky (1968) using sucrose CS and
ing mechanisms; to biologists without any X-ray UCS and Rozin (1969b) (using
particular commitments to psychological saccharin or casein hydrolysate as CSs and
theory, this explanation seemed perfectly apomorphine as UCS) confirmed this effect.
plausible.
One way of reconciling these data to Hypotheses to Explain the Long Delay
traditional S-R learning theory was to Peripheral—the aftertaste theory. Be-
assume that animals associated a food with cause of the revolutionary nature of this
the consequences of a previous meal of the finding, there was considerable interest in
same food, thus achieving temporal con- the possibility that the apparent absence of
tiguity (Rozin et al., 1964; Smith & temporal contiguity was illusory. Although
Capretta, 1956). But rats need only a the CS-UCS interval was ostensibly long,
single exposure to a toxic food to learn to some peripheral trace of the CS might re-
avoid it (Rzoska, 1953); thus some other main throughout the interval in the form
explanation is necessary. of an aftertaste, regurgitation, or a high
The discovery that long CS-UCS in- concentration in the blood. It could be
tervals are possible in learning about the this trace that was involved in the condi-
consequences of foods occurred gradually. tioning. However, a fair amount of data
Garcia et al. (1961) in their early work are now available to oppose this explana-
found that rats learned food aversions on tion. The main import of the data is that
the basis of simultaneous exposure to X taste-aversion learning is possible under
rays. Although it was known that X rays conditions that should greatly minimize
had their main effect with a considerable any aftertaste.
delay ("radiation sickness"), it was as- First, taste-aversion learning to sucrose
sumed that some immediate effect of X and saccharin solutions occurs on a single
rays was involved in the learning, and trial with maximum delays of poisoning
slight evidence was offered to support this of about 7 and 12 hours, respectively
position. McLaurin (1964) was the first (Revusky, 1968; Smith & Roll, 1967). It
to operationally manipulate CS-UCS inter- is very doubtful that an aftertaste or
vals over a wide range in taste-aversion perceptible change in blood concentration
learning with long delays of reinforcement. persists for such long periods. It is even
However, a methodological flaw precluded less plausible to suggest that enough
meaningful interpretation of this result; sucrose or saccharin remains in the stomach
McLaurin tested for aversion immediately at this time to be tasted by regurgitation.
after exposure to X rays, and it was later Actually the regurgitation model never
demonstrated (Smith & Schaeffer, 1967) had much applicability to rats anyway,
that the rats were learning aversions to the since rats rarely if ever vomit. A second
test solution during the test itself. That argument against the aftertaste and re-
is, the drinking of the solution during lated models is the fact that rats can learn
testing was simultaneous with the develop- aversions to a .05% HC1 solution with a
ment of aversive consequences of the one-hour delay of reinforcement (Garcia,
X rays, and temporal contiguity was Green, & McGowan, 1969). A litmus
achieved. Garcia et al. (1966) avoided paper test showed no measurable amount
this problem by giving the test for sac- of HC1 on the tongue two minutes after
charin aversion three days after exposure the animal stopped drinking. Thus, the
to apomorphine. This experiment suc- likelihood of a conventional peripheral
cessfully demonstrated learning with long aftertaste an hour later is minimal. Fur-
delays of reinforcement of the order of thermore, the solution was less concentrated
 ADAPTIVE SPECIALIZATIONS OF LEARNING 475

than the HC1 already in the stomach. aversions would not solve the delay prob-
Consequently, this experiment is peculiarly lem, as the HC1 experiments (Garcia et al.,
effective in eliminating regurgitation, stom- 1969), concentration aversions (Rozin,
ach-tasting, and blood-tasting models. A 1969b), and solution temperature aversions
third line of evidence is Rozin's (1969b) (Nachman, 1970a) are no easier to explain
demonstration that rats can learn an in terms of contiguity with intragastric
aversion to a specific concentration of a stimuli.
solution as opposed to other concentra- Bradley and Mistretta (1971) have
tions of the same solution. The animal demonstrated the development of aversions
learned the aversion just as easily to the to intravenously introduced solutions (sac-
lower as to the higher concentration. Pre- charin) in rats, using X rays as the UCS.
sumably after the 30-minute delay used in A circulating "slug" of high-concentration
this experiment, the aftertaste of a more saccharin stimulates taste receptors in the
concentrated solution should taste more tongue. This interesting experiment pro-
like a less concentrated solution of the vides another mechanism of learned aver-
same substance. Similarly, the blood or sions, but it cannot be the only mechanism
stomach concentration of the substance at involved in orally mediated aversions.
the time of poisoning should not be uniquely The blood concentrations used in these
related to a particular concentration of experiments are much higher than any
original solution by a function known in- which would occur naturally or in the
dependently by the rats. A fourth line long-delay experiments, and the same ex-
of evidence is Nachman's (1970a) demon- periments discussed above as raising prob-
stration that rats can learn an aversion to lems for an intragastric-tasting mechanism
a particular temperature of water; an would be equally troublesome for a "blood-
aftertaste of a temperature is difficult to tasting mechanism."
imagine. Fifth, rats can learn an aversion At present there is no evidence in favor
to a solution even if one or more solutions of an aftertaste theory and a considerable
is drunk between the first solution and body of evidence against it. Establish-
poisoning (Kalat & Rozin, 1970, 1971a; ment of an alternative theory seems a more
Revusky & Bedarf, 1967). The interven- appropriate strategy than accumulation
ing solutions would surely minimize any of still further evidence against aftertaste
aftertaste of the first. Finally, it has been interpretations. Of course, from the point
argued (Revusky & Garcia, 1970) that of view of this paper, long-delay learning
even if the aftertaste model were correct, is exactly what should be expected in this
it would be difficult to reconcile the taste- situation, and the central-mediation alter-
aversion learning results to conventional native appears quite acceptable.
learning theories. Even if there is an Central mechanisms—interference. Re-
aftertaste, it would have been present and vusky (in press; see also Revusky &
declining in intensity for such a long Garcia, 1970) proposes that the maximum
period that there would be no precedent CS-UCS interval in all types of learning
for expecting learning to occur—let alone depends not on time per se, but on the
in one trial. number of other potential CSs that in-
It may be argued that none of these tervene between the^CS being tested and
arguments completely demolishes an after- the UCS. That is, the animal tends to
taste theory, and that certain other tests associate the UCS with the most recent
could be conducted, for example, attempt- potential CS, or perhaps the last several
ing to produce aversions to solutions such stimuli. As the CS-UCS interval
intubated intragastrically. Some recent expands, the probability increases that the
evidence suggests that if such aversions organism will perceive other sights, sounds,
occur at all, they are much smaller than etc., and consequently the probability de-
taste-mediated aversions (Smith, 1970). creases that the animal will associate the
However, even if demonstrable, intragastric UCS with the CS in question. In taste-
476 PAUL ROZIN AND JAMES W. KALAT
aversion learning, the argument holds, the novel solutions during a 30-minute delay
range of potential CSs is much more re- between sucrose consumption and poison
stricted. Only tastes have a substantial does not eliminate the sucrose aversion
tendency to be associated with aversive (Kalat & Rozin, 1971a). The maximal
gastrointestinal stimulation, and typically interference effect we observed using three
the test solution is the most recent taste novel interfering solutions following sucrose,
the animal experienced prior to poisoning. and one poisoning, left sucrose equally
Since no other potential CSs are present, palatable with water, to which it is normally
there is nothing to interfere with learning strongly preferred.
over long delays. In taste aversion as in These experiments, then, suggest that
other learning, the animal tends to as- retroactive interference can clearly weaken
sociate a UCS with the last potential CS; a potential association, but it is highly
the only difference is that in taste-aversion unlikely that the normal delay function can
learning, the potential CSs are fewer and be largely accounted for in these terms:
less frequent. We would like to point out What could interfere with drinking of a test
that the taste modality differs from the solution during six hours of no eating or
commonly studied exteroceptive modalities, drinking that would be more effective than
in that virtually all stimulation comes from three novel solutions? Proactive interfer-
contacts initiated by behavior. Rats do ence is an unlikely explanation. Effects
not taste unless they approach something from past taste experiences should be
and introduce it into their mouth. minimal, since most rats have experienced
This theory is very attractive because only highly familiar rat chow, water, and
it proposes that the differences between mother's milk.
taste-aversion learning and other types Recent findings by Wilcoxon et al.
of learning may all be derived from the (1971) pose further problems for Revusky's
general principle of belongingness, without theory. Unlike rats, bobwhite quail poi-
postulating an independent difference in soned 30 minutes, after drinking unflavored
the delay of reinforcement function. water from a blue tube learn an aversion
Nevertheless, the theory, if taken as the to drinking from that tube. Since the
sole^of primary explanation of delay, faces quail presumably saw a great many visual
two serious problems. It predicts that in stimuli in the 30-minute delay, their
the absence of interfering taste cues, there ability to learn over long delays in this
should be little or no decrement in learning situation cannot be explained simply in
as the CS-UCS interval is increased. This terms of absence of interference.
is not the case. Garcia et al. (1966), Although it seems unlikely that inter-
Nachman, (1970a), Revusky (1968), and ference represents the only basis for the
Kalat and Rozin (197 la) have all found CS-UCS delay gradient, it does appear to
an orderly decrease of aversion with in- be a contributing factor. Since a rat
creasing CS-UCS interval. Furthermore, normally experiences fewer tastes than
all experiments have found a maximal other stimuli within a given period, the
interval, varying from about 2 to 12 hours, Revusky theory may account for part of
beyond which no learned aversion can be the difference in the delay gradients be-
demonstrated. Kalat and Rozin (197la) tween taste-aversion and other types of
deprived rats of both food and water during learning.
the CS-UCS interval, and still observed Slow decay of associability as an inherent
an orderly decrease in learned aversions property of the taste system. An alternative
with increasing CS-UCS interval. to the view that the delay-of-reinforce-
Not only is there a decrement in learn- ment gradient is a function of interference
ing in the absence of explicit interference, is the view that the delay gradient is an
but the addition of explicit interfering cues inherent property of the system itself;
does not markedly reduce learning. The memory or associability decays much
consumption of two or even three salient more slowly for taste than for other
 ADAPTIVE SPECIALIZATIONS OF LEARNING 477

modalities. This leaves unanswered the tastes remain in short-term memory for a
question, What accounts for this decay? long time.
One possibility is that the delay func- Another possible mechanism for the delay
tion represents the decay of short-term function is that some central long-term
memory. According to this hypothesis, a "trace" of the taste is decaying over time.
stimulus is associable with other events Unfortunately, this is a difficult hypothesis
only while it is held in short-term memory, to test. One experiment (Rozin & Ree9)
and tastes remain in short-term memory at least puts certain constraints on the
for unusually long periods. (Although type of decay which is possible. Rats
there is extensive data on short-term were anesthetized for 6-8 hours during the
memory in humans, virtually none has interval between consumption of a sucrose
been collected on taste, and it is con- solution and LiCl poisoning. These rats
ceivable that tastes do not compete with showed strong learned aversions at delays
exteroceptive cues for space in short-term of poisoning considerably longer than those
memory.) which are effective in the absence of
One way of testing this hypothesis is by anesthesia. Thus anesthesia seems to
means of electroconvulsive shock, which retard whatever process mediates the
has been assumed to eradicate short-term delay of reinforcement gradient. If this
memories or to block their conversion to process is to be described as "decay," it is
long-term memories. If taste associations evidently an active rather than a passive
must be made from short-term memory type.
extended in time, it should be possible to Central mechanisms—learned safety. The
demonstrate disruptive effects of electro- suggestion of an "active decay" process
convulsive shock, presented within the raises still another possibility, which is not
CS-UCS interval but with longer delays a decay mechanism at all. Perhaps the
following the CS than are ordinarily delay gradient should be viewed not as a
effective in disrupting other types of forgetting curve but as a learning curve.
memory. In a very careful study, Nach- That is, in the absence of unfavorable
man (1970b) found amnesic effects of gastrointestinal events, as time passes
electroconvulsive shock in some rats when .following consumption of a novel solution,
electroconvulsive shock was presented im- the animal learns that the solution is safe
mediately after 10 seconds of drinking but (Kalat & Rozin, see Footnote 6).
not after 30 seconds of drinking. He also As evidence for this, it has been demon-
found some amnesic effects from electro- strated that an animal does learn some-
convulsive shock presented 25 seconds thing about a solution when consumption
after a 5-second drinking period. Krai of that solution is followed by no negative
(1970) also found small amnesic effects consequences. We have already discussed
with electroconvulsive shock, using delays the evidence indicating that rats learn less
of 2 or 25 minutes following a 10-minute aversion to familiar solutions than to novel
drinking period. As is usual with elec- solutions (Revusky & Bedarf, 1967) even
troconvulsive shock experiments, the tem- when the familiar solution was experienced
poral parameters seem to vary widely from once for only 20 minutes, three weeks
one experiment to another. However, in
both cases, the effect of electroconvulsive before poisoning (Kalat & Rozin, see
shock is small; it impeded but did not Footnote 6). This interpretation goes on
prevent learning. And in both cases, the to make a stronger claim: The learning of
effective electroconvulsive shock times were "safety" takes place within the period
within the range of times reported for measured by the maximum delay of
electroconvulsive-shock-amnesic effects in reinforcement. At the end of that period,
other systems. Thus there is no evidence 9
P. Rozin and P. Ree. Long extension of effective
that the transfer of taste stimuli into long- CS-UCS interval by anesthesia between CS and US.
term memory is unusually slow, or that In preparation.
478 PAUL ROZIN AND JAMES W. KALAT

the "trace" has not decayed; it has merely What Kind of Learning Is This ?
been associated with the absence of aversive
consequences. At intermediate delays, the We have already described a number of
intermediate amount of learned aversion differences between taste-aversion learn-
reflects the fact that the animal has ing and traditional learning. We shall now
learned an intermediate amount of safety. consider some additional evidence in order
Although this interpretation is grossly to determine how fundamental this dis-
different (and not inconsistent with) the tinction is. Recent experiments have
interference or trace-decay interpretations, suggested two additional differences. First,
it is not easy to separate the alternatives it appears that rats can learn taste aver-
sions when poisoned under anesthesia
experimentally. The results of Rozin and
(Berger10; Roll & Smith, in press). Anes-
Ree (see Footnote 9), which suggest that
thesia was continued for a long period fol-
indefinitely long CS-UCS intervals will
lowing UCS administration, so that it was
support aversion learning if the rat is
highly likely that the learning occurred
anesthetized during the interval, are con-
under anesthesia. Second, one of the more
sistent with learned safety, since it is
"complex" characteristics of classical con-
reasonable that the rat would be unable to
ditioning, the Kamin blocking effect (Ka-
learn safety while anesthetized.
min, 1969), is either nonexistent or rela-
tively weak in taste-aversion learning
WHAT ALL THIS MEANS TO THE RAT (Kalat & Rozin, 197Ib; Revusky, in press).
We are now prepared to describe how a One interpretation of why taste-aversion
rat can handle some of the complex prob- learning does not clearly show the Kamin-
lems in food selection. The first thing to type effects is that the taste does not be-
realize is that the situation is probably less come a signal for poison in the sense that
complex than it might appear when the a tone or light becomes a signal for shock.
rat's natural behavior is considered. For In taste-aversion learning, the animal's
example, the rat who gets sick in the gar- perception of the taste itself or of its
bage dump probably did not recently affective value may change (suggested by
sample all the choice delicacies available H. Gleitman, personal communication,
(Rozin, 1969a). His choice behavior 1971). The taste itself may become
itself will help to unconfound the situation. aversive or unacceptable, as if it were un-
He may have eaten a few different foods. palatable (Rozin, 1967b). By contrast,
He "knows" it was & food that made him stimuli associated with shock do not them-
sick (the belongingness principle) and selves become aversive; they evoke little
can discount any familiar safe foods (the avoidance outside the training situation.
novelty principle). With the capability This difference seems related to the re-
of forming associations over long delays, ported "irrationality" of learned taste
he is now likely to associate his illness aversions; humans commonly maintain
with the last relevant (as defined above) an aversion to foods they ate prior to
thing or few things he ate over the last becoming nauseous, even when they are
few hours. Although some of these foods sure that some other agent was responsible
may become more aversive than others be- for the nausea.
cause of their intrinsic properties (salience Seligman (1970) has suggested that
effect), the rat will acquire a significant taste-aversion learning may be a particu-
aversion to each of them (small inter- larly primitive type of learning. The
ference effect), with those closer in time evidence mentioned above is quite con-
to the aversive event picking up somewhat sistent with this view. Furthermore, in a
more aversion (temporal contiguity). Simi- biological context, it makes sense that this
lar mechanisms would be employed to ability should be primitive. The problems
account for important aspects of the regula- 10
B. Berger. Learning in the anesthetized rat.
1ion of food intake. In preparation.
 ADAPTIVE SPECIALIZATIONS OF LEARNING 479

of food selection and the regulation of food perform just that function. How specifically these
intake are pervasive ones, and both are mechanisms are differentiated for one particular
function is borne out by the fact that they are very
likely to involve long-delay learning. The often quite unable to modify any but one strictly
essential problems are relatively invariant determined system of behavior mechanisms fJLorenz,
across species: any new food has some 1965, p. 47].
probability of being toxic, and for almost We are aware of only two well-studied
all species, the caloric density of foods will systems showing adaptive specializations of
vary. Therefore, it makes sense that taste learning. One is feeding, which we have
aversion and related types of learning already discussed, and the other involves
should be more or less the same throughout imprinting, as a mechanism of species
most of the vertebrate subphylum.
recognition.
Two of the prominent features of im-
ADAPTIVE SPECIALIZATIONS OF LEARNING: printing—the special sensitivity during a
GENERALITY AND RELATION TO critical period early in life and the great
OTHER POSITIONS resistance to extinction—can be seen as
The work on specific hungers and poi- adaptations to limit the likelihood of
soning suggests that there are two aspects errors in species recognition. The learning
of adaptive specializations of learning. should take place soon after hatching, since
First, some mechanisms of learning may the probability of exposure to a conspecific
differ markedly (at least in terms of large (i.e., the mother) is highest at this time,
quantitative differences in basic parameters) and the bird should be less sensitive to
in the feeding system, as compared to other later experiences, since the frequency of
systems. Second, understanding of an contact with members of other species is
adaptive specialization includes delimita- likely to increase greatly after the nestling
tion of the situations in which it applies period.
and elaboration of its relationship and Feeding and imprinting can be con-
interaction with the animals' natural be- sidered as two exceptions to an otherwise
havior (e.g., sampling, neophobia) in the correct "general process" view of learning,
relevant situation. The variation in inter- or they can be considered as examples
play among naturalistic context, genetic of a basic adaptational principle pervading
programming, and learning is clearly il- much or all of learning. We prefer the
lustrated in the contrast between water latter alternative and believe that the
or sodium hunger, and other specific absence of additional known instances of
hungers. adaptive specializations may reflect learn-
Our emphasis has been on the first aspect ing psychologists' reluctance to study
of adaptation, while the ethologists have potentially learned behaviors which do not
focused more on the context of learning fit into the general process paradigms.
and interplay of prestructured and ac- (Significantly, the case for adaptive spe-
quired components: cializations in both imprinting and feeding
came from outside the psychology of
The student of innate behavior, accustomed to learning.)
studying a number of different species and the entire At this point one can only speculate
behavior pattern, is repeatedly confronted with the
fact that an animal may learn some things much about what other systems will show special
more readily than others. . . . In other words, adaptations of learning. Bees may possess
there seem to be more or less strictly localized a wide variety of adaptive specializations
"dispositions to learn." Different species are pre- (von Frisch, 1953, 1967). A particularly
disposed to learn different parts of the pattern. So
far as we know, these differences between species promising example is their navigational
have adaptive significance [Tinbergen, 1951, p. 145]. ability. First, there is evidence for
"belongingness."
. . . innumerable observations and experiments
tend to show that modifiability occurs, if at all, only Honey bees can learn to^use irregular forms, like
in those performed places where built-in learning those of trees or rocks, as landmarks by which to
mechanisms are phylogenetically programmed to steer a course to and from the hive; but they cannot,
480 PAUL ROZIN AND JAMES W. KALAT
even by the most subtle conditioning technique, be not been thoroughly studied as possible
taught to use the same forms as positive or negative
signals indicating the presence or absence of food in
specialized learning mechanisms.
a tray, as von Frisch (1914) has shown. As signals The "heuristic" value of an adaptive
for food, bees can distinguish different forms only if evolutionary point of view can be suggested
they are geometrically regular, preferably radially by considering the types of predictions
symmetrical (Hertz, 1937) [Lorenz, 1965, p. 47].
that might be made about some basic
Second, a limited amount of experience in learning and memory relationships. For
observing a piece of the sun's arc enables example, an organism's memory of some
the bee to "project" the rest of the arc. aspect of the environment is useful only
That is, bees raised without the oppor- if that aspect is predictable or controllable;
tunity to see the sun have great difficulty otherwise, rapid forgetting might be ad-
compensating for sun movement during vantageous. It is probably of no use for
navigation. But bees that have observed bees, birds, or other organisms with com-
sun movement only during a limited period plex navigational abilities to remember
in the morning are fully capable of com- whether it was cloudy yesterday or which
pensating for sun movement in the after- way the wind was blowing, though both
noon (von Frisch, 1967). In this situa- may have been important at the time.
tion, the environmental input produces a Under these circumstances one might
long-term change, and does not act "as- expect to find a specialization in short-
sociatively," but rather provides a refer- term memory such that information could
ence point. Other examples of such be stored for longer periods and in greater
"calibrational learning" (Lorenz, 1965) quantities than usual without entering
might include adaptation to visual dis- long-term memory.
placement produced by prisms and caloric Another example concerns extinction.
regulation.11 Extinction, from an adaptive point of
The naturalistic literature is replete with view, allows an organism (a) to correct for
other examples of surprising abilities of mistakes (fortuitous conditioning) and (b)
animals—such as digger wasps' memory of to constantly reshape itself to adapt to a
the location and state of their nests variable environment. One might expect
(Baerends, 1941), gobies' latent learning the rapidity of extinction to depend on the
of the location of tidepools (Aronson, probability that either of these events
1951), salmon's recognition of home-stream would occur. In the case of imprinting, we
odor (Hasler, 1966), doves' individual mate have great resistance to extinction in a
recognition (Morris & Erickson, 1971), case where clearly the environment will
sparrows' acquisition and storage of song not vary (i.e., the species will not change),
dialects (Marler, 1970), etc. These have and the proper imprinting object is almost
11 certain to be present at the time of im-
We can suggest a role for "calibrational learn-
ing" in the regulation of food intake. It has been printing. The great resistance to extinc-
known for some time that rats and other mammals tion of avoidance conditioning (Solomon &
respond to changes in caloric density of food by Wynne, 1954) may be an adaptively
appropriate modulation of volume intake. In the selected feature of this learning: the costs
rat, this compensation occurs largely as an increase of errors of omission here are high. One
in meal size, rather than in number of meals
(Snowdon, 1969; Teitelbaum & Campbell, 1958). might expect rapid extinction of the
When a standard diet is diluted, the rat ends up learned location and stimulus properties
eating larger meals: he has "recalibrated" his meal of food sources, where these sources are
size on the basis of the metabolic aftereffects of his subject to marked seasonal or other tem-
meals. It is possible to imagine a mechanism which
compares some measure of the amount ingested with poral variations. An example might be
its delayed metabolic effects, and adjusts future extinction of responses to particular types
intake downward or upward so that the metabolic and locations of flowers in foraging bees.
consequences will approach some preferred or ideal One might observe slow learning, but high
value. The demonstration of a long-delay learning
mechanism makes this type of explanation feasible resistance to extinction in a situation where
and subject to investigation. the environment is stable but a critical
 ADAPTIVE SPECIALIZATIONS OF LEARNING 481

difference is relatively hard to detect More significantly, the long-delay learning

(e.g., prey density), and "averaging" must found in feeding is probably not character-
be done before a clear choice or preference istic of other "prepared" associations, and
is established. One might observe rapid in our view, it should not be, since in most
learning and rapid extinction in a rapidly cases close temporal contiguity is the best
changing environment (see Shettleworth, predictor. Furthermore, we see no reason
1971, for additional comments on this to expect a consistent relationship between
point). rapid learning and high resistance to ex-
Recently, a number of authors have ex- tinction (see discussion of extinction above).
pressed positions related to ours (Garcia & In short, we disagree with Seligman that
Ervin, 1968; Garcia et al., 1971; Revusky, diversity can be ordered along any single,
in press; Revusky & Garcia, 1970; Selig- operationally meaningful dimension, that
man, 1970). We differ from all of these is, preparedness. If preparedness meant
authors in the sense that they see be- adaptedness to situational demands, it
longingness, in one form or another, as would be an acceptable but not clearly
the unique phenomenon to be explained, meaningful dimension. Seligman has given
whereas we see it as an example of the it operational meaning, but narrowed the
general adaptational principle; animals scope of the phenomena he can account for
may not only learn some things more in the process. In a more recent statement
easily than others, but they may also of the preparedness position, Seligman and
learn some things in a different way than Hager (in press) have acknowledged some
others. The contrast is clearest in the of these possible limitations.
case of Revusky, who attempts to pre- Shettleworth (1971), in a paper written
serve traditional learning theory intact, concurrently with this one, has presented
with the introduction of a new "belonging- a position very similar to ours. It focuses
ness" assumption. on belongingness, broadly conceived, but
Seligman (1970) proposes a new di- also describes learning in general in terms
mension, "preparedness," based on the of adaptive specializations, and provides
belongingness relationship, to incorporate examples from the naturalistic literature.
recent findings into a more viable learning A recent paper by Lockard (1971) also
theory. Preparedness represents the tend- explicitly discusses the diversity of learn-
ency of certain inputs (and/or outputs) to ing mechanisms as an important feature
be associated with one another, this ignored by most psychologists and implies
tendency resulting from natural selection. that the search for common elements in
Seligman proposes that highly prepared learning across species and situations is
associations are established with a minimal almost hopeless.
input (e.g., number of pairings). In We differ slightly from Shettleworth and
addition to very rapid learning, prepared markedly from Lockard with respect to
associations would tend to show learning our optimism about the possibility of find-
with long delays of reinforcement and per-
ing order within the diversity of learning
haps high resistance to extinction. We
are certainly in accord with the general mechanisms. Given the constraints on
flavor of Seligman's position, but we feel adaptations produced by basic properties
that in his desire to reorder the phenomena of the nervous system, the cost of evolving
of learning, he has not fully appreciated specializations, and the fact that most
the diverse natural forces that can shape species face a common set of problems, we
behavior and learning mechanisms. In doubt that a separate learning mechanism
imprinting, for example, wh£re Seligman's would exist for every situation, or that
view of rapid learning (and high resistance there would be separate laws for each
to extinction) fits very wellj the presence species. It may yet be possible to formu-
of the critical period canntot be accom- late laws of some degree of generality,
modated without additional assumptions. taking ecological factors into account (see
482 PAUL ROZIN AND JAMES W. KALAT

Walls, 1942, for an example in the area of certain higher learning abilities that "fish"
vision). do not possess. Another not mutually ex-
clusive possibility is that the rat and
SOME SPECULATIONS probably most other mammals are "gen-
It follows from the point of view pre- eralists," compared to the majority of other
sented here that an organism may have an vertebrates. That is, abilities initially
ability that manifests itself in only a small evolved to handle a specific situation and
number of the total possible situations in limited in application to the appropriate
which an experimenter might test for it. system may turn out to be useful in others
The ability in question might be inac- and may through evolution be "connected"
cessible to or "unconnected" with the into new systems. To what extent do the
machinery for modulating and controlling higher abilities of the mammals represent
most behavior. Using a computer analogy, an increase in accessibility of specialized
we might suppose that a particular routine capacities, an "emancipation," to borrow
is designed (evolved) to handle a specific a word from the ethologists, of a capacity
problem. At this point in time, it is from its original tight motivational system ?
physically connected only into the relevant The proposed increase in accessibility of
inputs, outputs, or systems, and is in- capacities in phylogeny may have a par-
accessible to the rest. Under these cir- allel in ontogeny. Within the Piagetian
cumstances, it would be difficult to describe framework, it is apparent that particular
a species' learning capacity. cognitive structures may have only limited
The interesting demonstrations by Bit- applicability at any point in development.
terman (1968), Gonzalez and Bitterman Piaget applies the name "decalage" to this
(1969), and Mackintosh (1965) of dif- feature of development. (Flavell, 1963,
ferences in learning abilities in certain pp. 21-24). For example, in the Piagetian
situations between a few species of fish scheme, the same cognitive structure is
and mammals do not in themselves in- necessary for the achievement of mass and
dicate complete absence of these abilities weight conservation, yet the latter occurs
in any species of fish (see Gleitman & about two years after the former. Cogni-
Rozin, in press, for a general review of this tive development may consist, in part, of
issue with respect to learning and memory the extension of existent capacities to new
in fish). A species should be tested for situations, in parallel to our scheme for
any ability in those situations where its phylogeny.
existence would have the greatest survival We have argued for the existence of
value. The failure of a few fish species to adaptive specializations in learning and
demonstrate abilities such as progressive memory. Since, by their nature, such
improvement in habit reversal in a few specializations are limited to a relatively
laboratory situations does not meet this narrow range of situations, we have pointed
demand. out that they must be inaccessible to most
However, the data gathered by Bitter- functioning systems. We believe that this
man, by Gonzalez and Bitterman, and by general formulation has a wide application,
Mackintosh do suggest some interesting and we are presently applying it to an
phylogenetic generalizations about learn- understanding of the difficulties in initial
ing capacities. Since the laboratory ap- acquisition of reading (Rozin & Kalat,
paratus used to study rats and other in press; Rozin, Poritsky, & Sotsky, 1971).
mammals is often not ideally suited to It is the basic thesis of this paper that in
their natural behavior, just as the fish a biological-evolutionary framework, spe-
apparatus is not, it is quite interesting that cifically adapted abilities are extremely
mammals reliably show the greater plas- important and should not suffer from
ticity in these "unnatural" situations. One neglect as a consequence of the search for
possible explanation for this, offered by great generalities. To understand a set
Bitterman (1964), is that the rat possesses of phenomena, within humans or across
 ADAPTIVE SPECIALIZATIONS OF LEARNING 483
the animal kingdom, is to be able to de- weaning rat pups. Journal of Comparative and
scribe and explain diversity, as well as to Physiological Psychology, in press.
GALEF, B. G., JR., & CLARK, M. M. Mother's milk
extract common elements. and adult presence: Two factors determining
initial dietary selection by weanling rats. Journal
REFERENCES of Comparative and Physiological Psychology, in
press.
APPLEDORF, H., & TANNENBAUM, S. R. Specific GARCIA, J., & ERVIN, F. R. Gustatory-visceral and
hunger for thiamine in the rat: Selection of low telereceptor-cutaneous conditioning—Adaptation
concentrations of thiamine in solution. Journal in internal and external milieus. Communications
of Nutrition, 1967,92, 267-273. in Behavioral Biology, 1968, 1 (Pt. A), 389-415.
ARONSON, L. Orientation and jumping in the gobiid GARCIA, J., ERVIN, F. R., & KOELLING, R. A.
fish Bathygobius soporator. American Museum
Novitates, 1951, 1486, 1-22. Learning with prolonged delay of reinforcement.
Psychonomic Science, 1966, 5, 121-122.
BAERENDS, G. P. Fortpflanzungsverhalten und GARCIA, J., ERVIN, F. R., YORKE, C. H., & KOEL-
orientierung der grabwespe Ammophila campcstris
LING, R. A. Conditioning with delayed vitamin
Jur. Tijdschr Etntomol, 1941, 84, 68-275.
BARNETT, S. A. The rat. A study in behaviour. injection. Science, 1967,155, 716-718.
London: Methuen, 1963. GARCIA, J., GREEN, K. F., & McGowAN, B. K.
BITTERMAN, M. E. The evolution of intelligence. X-ray as an olfactory stimulus. In C. Pfaffman
Scientific American, 1964, 212, 92-100. (Ed.), Taste and olfaction. New York: Rockefeller
BITTERMAN, M. E. Comparative studies of learning University Press, 1969.
in the fish. In D. J. Ingle (Ed.), The central GARCIA, J., KIMELDORF, D. J., & HUNT, E. L. The
nervous system and fish behavior. Chicago: Uni- use of ionizing radiation as a motivating stimulus.
versity of Chicago Press, 1968. Psychological Review, 1961, 68, 383-385.
BRADLEY, R. M., & MISTRETTA, C. M. Intra- GARCIA, J., & KOELLING, R. A. Relation of cue to
vascular taste in rats as demonstrated by con- consequence in avoidance learning. Psychonomic
ditioned aversion to sodium saccharin. Journal of Science, 1966, 4, 123-124.
Comparative and Physiological Psychology, 1971, GARCIA, J., & KOELLING, R. A. A comparison of
75,186-189. aversion induced by x-rays, toxins, and drugs in
BROWER, L. P. Ecological chemistry. Scientific the rat. Radiation Research, 1967, Suppl. 7,
American, 1969, 220, 22-29. 439-450.
BYKOV, K. M. The cerebral cortex and the internal GARCIA, J., MCGOWAN, B. K., & GREEN, K. F.
organs. (Trans. & ed. by W. H. Gantt) New Biological constraints on conditioning. In A.
York: Chemical Publishing, 1957. Black & W. F. Prokasy (Eds.), Classical condi-
CAMPBELL, C. Development of specific preferences tioning. New York: Appleton-Century-Crofts,
in thiamine-deficient rats: Evidence against me- in press.
diation by aftertastes. Unpublished master's GIBSON, J. J. The senses considered as perceptual
thesis, Library of the University of Illinois at systems. Boston: Houghton Miffln, 1966.
Chicago Circle, 1969. GLEITMAN, H., & ROZIN, P. Learning and memory.
CAPRANICA, R. R. Vocal response of the bullfrog In W. F. Hoar & D. J. R. Randall (Eds.), Fish
to natural and synthetic mating calls. Journal of physiology. New York: Academic Press, in press.
the Acoustical Society of America, 1966, 40, GONZALEZ, R. C., & BITTERMAN, M. E. The spaced-
1131-1139. trials PRE as a function of contrast. Journal of
CAPRETTA, P. J. An experimental modification of Comparative and Physiological Psychology, 1969,
food preference in chickens. Journal of Compara- 67, 94-103.
tive and Physiological Psychology, 1961, 54, 238- HARPER, A. E. Effects of dietary protein content
242. and amino acid pattern on food intake and pre-
FLAVELL, J. H. The developmental psychology of Jean ference. In C. F. Code & W. Heidel (Eds.),
Piaget. Princeton: Van Nostrand, 1963. Handbood of physiology. Vol. 1, Section 6.
GALEF, B. G. Aggression and timidity: Responses Washington, D. C.: American Physiological
to novelty in feral Norway rats. Journal of Society, 1967.
Comparative and Physiological Psychology, 1970, HARRIS, L. J., CLAY, J., HARGREAVES, F., & WARD,
70, 370-381. A. Appetite and choice of diet. The ability of
GALEF, B. G. Social effects in the weaning of the vitamin B deficient rat to discriminate be-
domestic rat pups. Journal of Comparative and tween diets containing and lacking the vitamin.
Physiological Psychology, 1971, 75, 358-362. Proceedings of the Royal Society, London (Series B),
GALEF, B. G., JR., & CLARK, M. M. Social factors 1933, 113, 161-190.
in the poison avoidance and feeding behavior of HASLER, A. D. Underwater guideposts. Homing of
wild and domestic rat pups. Journal of Compara- salmon. Madison: University of Wisconsin Press,
tive and Physiological Psychology, 1971, 75, 341- 1966.
357. (a) HERTZ, M. Beitrag zum Farbensinn und Formen-
GALEF, B. G., JR., & HENDERSON, P. W. Mother's sinn der Biene. Zeilschrift fur vergleichende
milk: A determinant of the feeding preferences of Physiologie, 1937, 24, 413-421.
484 PAUL ROZIN AND JAMES W. KALAT
HUNT, G. L., & SMITH, W. J. Pecking and initial MORRIS, R. L., & ERICKSON, C. J. Pair-bond main-
^drinking responses in young domestic fowl. tenance in the ring dove (Streptopelia risoria).
IH* Journal of Comparative and Physiological Psy- Journal of Animal Behavior, in press.
chology, 1967, 64, 230-236. NACHMAN, M. Taste preferences for sodium salts by
KALAT, J. W., & ROZIN, P. "Salience": A factor adrenalectomized rats. Journal of Comparative
which can override temporal contiguity in taste- and Physiological Psychology, 1962, 55, 1124-1129.
aversion learning. Journal of Comparative and NACHMAN, M. Learned taste and temperature
Physiological Psychology, 1970, 71, 192-197. aversions due to lithium chloride sickness after
KALAT, J., & ROZIN, P. Role of interference in temporal delays. Journal of Comparative and
taste-aversion learning. Journal of Comparative Physiological Psychology, 1970, 73, 22-30. (a)
and Physiological Psychology, 1971, 77, 53-58. (a) NACHMAN, M. Limited effects of electroconvulsive
KALAT, J., & ROZIN, P. You can lead a rat to shock on memory of taste stimulation. Journal of
poison—but you can't make him think. In M. Comparative and Physiological Psychology, 1970,
E. P. Seligman & J. Hager (Eds.), Biological 73, 31-37. (b)
boundaries of learning. New York: Appleton- PAVLOV, I. Conditioned reflexes. Oxford: Oxford
Century-Crofts, 1971, in press, (b) University Press, 1927.
RACHMAN, S., & TEASDALE, J. Aversion therapy and
KAMIN, L. J. Predictability, surprise, attention, and
behavior disorders: An analysis. Coral Gables:
conditioning. In B. Campbell & R. Church University of Miami Press, 1969.
(Eds), Punishment and aversive behavior. New
York: Appleton-Century-Crofts, 1969. REVUSKY, S. H. Hunger level during food con-
sumption : Effects on subsequent preference. Psy-
KIMBLE, G. A. Hilgard and Marquis' conditioning chonomic Science, 1967, 7, 109-110.
and learning. New York: Appleton-Century- REVUSKY, S. H. Aversion to sucrose produced by
Crofts, 1961. contigent x-irradiation: Temporal and dosage
KRAL, P. Interpolation of clectroconvulsive shock parameters. Journal of Comparative and Physio-
during CS-US interval as an impediment to the logical Psychology, 1968, 65, 17-22.
conditioning of taste aversion. Psychonomic REVUSKY, S. H. The role of interference in associa-
Science, 1970, 19, 36-38. tion over a delay. In W. Honig & H. James
KRIECKHAUS, E. E., & WOLF, G. Acquisition of (Eds.), Animal memory. New York: Academic
sodium by rats: Interaction of innate mechanisms Press, in press.
and latent learning. Journal of Comparative and REVUSKY, S. H., & BEDARF, E. W. Association of
Physiological Psychology, 1968, 65, 197-201. illness with prior ingestion of novel foods. Science,
LEMAGNEN, J. Peripheral and systemic actions of 1967, 155, 219-220.
food in the caloric regulation of intake. Annals of REVUSKY, S. H., & GARCIA, J. Learned associations
the New York Academy of Science, 1969, 157, over long delays. In C. H. Bower & J. T. Spence
1127-1157. (Eds.), The psychology of learning and motivation:
LETTVIN, J. Y., MATURANA, H. R., McCuLLOcu, W. Advances in research and theory, IV. New York:
S., & PITTS, W. H. What the frog's eye tells the Academic Press, 1970.
frog's brain. Proceedings of the Institute of Radio RICHTER, C. P. Increased salt appetite in adre-
Engineers, 1959,47, 1940-1951. nalectomized rats. American Journal of Physio-
LOCKARD, R. B. Reflections on the fall of compara- logy, 1936, 115, 155-161.
tive psychology: Is there a message for us all? RICHTER, C. P. Total self-regulatory functions in
American Psychologist, 1971, 26, 168-179. animals and human beings, Harvey Lectures
LORENZ, K. Evolution and modification of behavior. Series, 1943, 38, 63-103.
Chicago: University of Chicago Press, 1965. RICHTER, C. P. Experimentally produced behavior
reactions to food poisoning in wild and domesti-
MACKINTOSH, N. J. Selective attention in animal cated rats. Annals of the New York Academy of
discrimination learning. Psychological Bulletin, Science, 1953, 56, 225-239.
1965, 64, 124-150. RICHTER, C. P. Salt appetite of mammals: Its
MAIER, S. F., ZAHORIK, D. M., & ALBIN, R. W. dependence on instinct and metabolism. In,
Relative novelty of solid and liquid diet during L'Instinct dans le comportement des animaux et de
thiamine deficiency determines development of I'homme. Paris: Masson, 1956.
thiamine-specific hunger. Journal of Comparative RICHTER, C. P., HOLT, L. E., JR., & BARKLARE, B.,
and Physiological Psychology, 1971, 74, 254-262. JR. Vitamin Bi craving in rats. Science, 1937,
MARLER, P. A comparative approach to vocal 86,354-355.
learning: Song development in white-crowned RODGERS, W. L. Specificity of specific hungers.
sparrows. Journal of Comparative and Physio- Journal of Comparative and Physiological Psy-
logical Psychology, 1970, 71(2, Pt. 2). chology, 1967, 64, 49-58. (a)
McLAURiN, W. A. Postirradiation sacharin avoid- RODGERS, W. L. Thiamine specific hunger. Un-
ance in rats as a function of the interval between published doctoral dissertation, University of
ingestion and exposure. Journal of Comparative Pennsylvania, 1967. (b)
and Physiological Psychology, 1964, 57, 316-317. RODGERS, W., & ROZIN, P. Novel food preferences
MORGAN, C. L. An introduction to comparative in thiamine-deficient rats. Journal of Compara-
psychology. London: Walter Scott, 1894. tive and Physiological Psychology, 1966, 61, 1-4.
 ADAPTIVE SPECIALIZATIONS OF LEARNING 485
ROEDER, K. Nerve cells and insect behavior. Cam- SCOTT, E. M., VERNEY, E. L., & MORISSEY, P. Self
bridge: Harvard University Press, 1963. selection diet. XI. Appetites'for calcium, mag-
ROGERS, Q. R., & HARPER, A. E. Selection of a nesium and potassium. Journal of Nutrition, 1950,
solution containing histidine by rats fed a histi- 41, 187-202.
dine-imbalanced diet. Journal of Comparative and SELIGMAN, M. E. P. On generality of the laws of
Physiological Psychology, 1970, 72, 66-71. learning. Psychological Review, 1970, 77, 406-
ROLL, D. L., & SMITH, J. C. Conditioned taste 418.
aversion in anesthetized rats. In M. E. P. Selig- SELIGMAN, M. E. P., & HAGER, J. The biological
raan & J. Hager (Eds.), The biological boundaries boundaries of learning, New York: Appleton-
of learning. New York: Appleton-Century- Century-Crofts, in press.
Crofts, in press, SHETTLEWORTH, S. Constraints on learning. In D.
ROTHSCHILD, M. Mimicry. The deceptive way of S. Lehrman, R. A. Hinde, & E. Shaw (Eds.),
life. Natural History, 1967, 76, 44-51. Advances in the study of behavior. Vol. 4. New
ROZIN, P. Specific hunger for thiamine: Recovery York: Academic Press, 1971.
from deficiency and thiamine preference. Journal SMITH, J. C., & ROLL, D. L. Trace conditioning
of Comparative and Physiological Psychology, 1965, with X-rays as an aversive stimulus. Psy-
59,98-101. chonomic Science, 1967, 9, 11-12.
ROZIN, P. Thiamine specific hunger. In, Handbook SMITH, J. C., & SCHAEFFER, R. W. Development
of physiology. Vol. 1, Section 6. Washington, of water and saccharin preferences after simul-
D. C.: American Physiological Society, 1967. (a) taneous exposures to saccharin solution and
ROZIN, P. Specific aversions as a component of gamma rays. Journal of Comparative and Phy-
specific hungers. Journal of Comparative and siological Psychology, 1967, 63, 434-438.
Physiological Psychology, 1967, 64, 237-242. (b) SMITH, M. P., & CAPRETTA, P. J. Effects of drive
ROZIN, P. Specific aversions and neophobia as a level and experience on the reward value of
consequence of vitamin deficiency and/or poison- saccharine solutions. Journal of Comparative and
ing in half-wild and domestic rats. Journal of Physiological Psychology, 1956, 49, 553-557.
Comparative and Physiological Psychology, 1968, SMITH, R. G. The role of alimentary chemore-
66, 82-88. ceptors in the development of taste aversions..
ROZIN, P. Adaptive food sampling patterns in Communications in Behavioral Biology, 1970, 5,
vitamin deficient rats. Journal of Comparative 199-204.
and Physiological Psychology, 1969, 69, 126-132. SNOWDON, C. T. Motivation, regulation, and the
(a) control of meal parameters with oral and intra-
ROZIN, P. Central or peripheral mediation of learning gastric feeding. Journal of Comparative and
with long CS-US intervals in the feeding system. Physiological Psychology, 1969, 69, 91-100.
Journal of Comparative and Physiological Psy- SOLOMON, R. L., & WYNNE, L. C. Traumatic
chology, 1969, 67, 421-429. (b) avoidance learning: The principles of anxiety con-
ROZIN, P., & KALAT, J. W. Learning as a situation- servation and partial irreversibility. Psychological
specific adaptation. In M. E. P. Seligman & J. Review, 1954, 61, 353-385.
Hager (Eds.), The biological boundaries of learning. STRICKER, E. M., & STERRITT, G. M. Osmoregula-
New York: Appleton-Century-Crofts, in press. tion in the newly hatched domestic chick. Physi-
ROZIN, P., PORITSKY, S., & SOTSKY, R. American ology and Behavior, 1967,2, 117-119.
children with reading problems can easily learn STRICKER, E. M., & WILSON N. E. Salt-seeking
to read English represented by Chinese characters. behavior in rats following acute sodium deficiency.
Science, 1971, 171, 1264-1267. Journal of Comparative and Physiological Psy-
ROZIN, P., & RODGERS, W. Novel diet preferences chology, 1970, 72, 416-420.
in vitamin deficient rats and rats recovered from TEITELBAUM, P., & CAMPBELL, B. A. Ingestion
vitamin deficiencies. Journal of Comparative and patterns in hyperphagic and normal rats. Journal
Physiological Psychology, 1967, 63, 421-428. of Comparative and Physiological Psychology, 1958,
ROZIN, P., WELLS, C., & MAYER, J. Thiamine 51, 135-141.
specific hunger: Vitamin in water versus vitamins TINBERGEN, N. The study of instinct. Oxford:
in food. Journal of Comparative and Physiological Clarendon Press, 1951.
Psychology, 1964, 57, 78-84. VON FRISCH, K. Der Farbensinn und Formensinn
RzOSKA, J. Bait shyness, a study in rat behaviour. der Biene. Zoologisches Jahrbuch, 1914,35, 1-188.
British Journal of Animal Behaviour, 1953, 1, VON FRISCH, K. The dancing bees. New York:
128-135. Harcourt, Brace, 1953.
SCOTT, E. M., & QUINT, E. Self selection of diet. VON FRISCH, K. The dance language and orientation
III. Appetites for B vitamins. Journal of Nutri- of bees. Cambridge, Mass.: Belknap Press of
tion, 1946, 32, 285-291. Harvard University Press, 1967.
SCOTT, E. M., £ VERNEY, E. L. Self selection of WALLS, G. L. The vertebrate eye and its adaptive
diet. VI. The nature of appetites for B vitamins. radiation. Bloomfield Hills, Mich.: Cranbrook
Journal of Nutrition, 1947, 34, 471-480. Institute of Science, 1942.
486 PAUL ROZIN AND JAMES W. KALAT
WICKLER, W. Mimicry in plants and animals. New phine-induced conditioned aversion to a novel
York: 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; revision
WITTLIN, W. A., & BROOKSHIRE, K. H. Apomor- received July 12, 1971)