4.4 Specialized Learning Abilities: Filling the Blanks in Species-Typical Behavior Patterns

Thus far we have been examining learning processes and activities that are quite general in the sense that they operate in a wide variety of contexts. Animals can learn many different things through classical conditioning, operant conditioning, play, exploration, and observation. But natural selection has also endowed animals with specialized learning abilities that have quite limited domains of operation. These may be best thought of as adjuncts to particular species-typical behavior patterns. Each such learning ability helps to mesh some aspect of the animal’s species-typical behavior with particular variable characteristics of the animal’s environment.

In Chapter 3, we cited the white-crowned sparrow’s learning of its local song dialect as an example of specialized learning. That ability does not stem from a general capacity of sparrows to imitate. Instead, it results from a special learning mechanism that is specific to song learning and narrow in scope. It allows the bird to learn any given white-crowned sparrow song dialect, but does not allow the bird to learn the song of another bird species. In this section, we will look at special learning abilities related to food preferences and then, more briefly, at a few other examples of specialized learning mechanisms.

Special Abilities for Learning What to Eat

For some animals, learning what to eat is a relatively simple matter. Koalas, for instance, eat only the leaves of eucalyptus trees. Through natural selection, koalas evolved a food-identifying mechanism that tells them that eucalyptus leaves are food and everything else is not. That simplifies their food choice, but if eucalyptuses vanish, so will koalas. Other animals are more flexible in their diets. Most flexible of all are omnivorous creatures, such as rats and humans, which treat almost all organic matter as potential food and must learn what is safe to eat. Such animals have evolved special mechanisms for learning to identify healthful foods and to avoid potential poisons.

Food-Aversion Learning: How It Differs from Typical Classical Conditioning

If rats become ill after eating a novel-tasting food, they subsequently avoid that food. In experiments demonstrating this, researchers induce illness by adding a toxic substance to the food or by administering a drug or a high dose of x-rays (inducing radiation sickness) to the animals after they have eaten (Garcia et al., 1972). Similarly, people who by chance get sick after eating an unusual food often develop a long-term aversion to the food (Bernstein, 1991; Logue, 1988). For years as a child, I (Peter Gray) hated the taste and smell of a particular breakfast cereal, because once, a few hours after I ate it, I happened to develop a bad case of stomach flu. I knew, intellectually, that the cereal wasn’t the cause of my illness, but that didn’t help. The learning mechanism kicked in automatically and made me detest that cereal.

Some psychologists choose to describe such cases of food-aversion learning in terms of classical conditioning. In that description, the feeling of illness or nausea induced by the x-ray treatment or drug is the unconditioned stimulus for a reaction of aversion or revulsion, and the taste and smell of the food become conditioned stimuli for that reaction. For example, many patients receiving chemotherapy or radiation therapy for cancer report acquiring aversions to foods that became associated with the nausea accompanying these therapies (Mattes et al., 1992). This is why such patients are often counseled to avoid eating favorite foods before treatment: Whatever aversions they acquire, those favorite foods won’t be affected. But John Garcia, the researcher who pioneered the study of food-aversion learning, argues that such learning is quite different from standard cases of classical conditioning (Garcia et al., 1989).

One special characteristic of food-aversion learning has to do with the optimal delay between the conditioned and unconditioned stimuli. In typical cases of classical conditioning, such as the salivary reflex studied by Pavlov, conditioning occurs only when the unconditioned stimulus follows immediately (within a few seconds) after the conditioned stimulus. But food-aversion learning has been demonstrated even when x-rays were administered as much as 24 hours after the animals had eaten the food (Etscorn & Stephens, 1973). In fact, food-aversion learning fails to occur if the gap between tasting the food and the induction of illness is less than a few minutes (Schafe et al., 1995).

Another special characteristic has to do with the sorts of stimuli that can serve as conditioned stimuli for such learning. In typical cases of classical conditioning, almost any kind of detectable stimulus can serve, but in food-aversion learning the stimulus must be a distinctive taste or smell (and taste generally works better than smell). Rats that become ill after eating a particular food subsequently avoid any food that tastes or smells like what they had eaten, even if it looks different, but they do not avoid a food that looks like what they had eaten if it tastes and smells different (Garcia et al., 1968, 1989). Also, when the x-ray-induced radiation sickness was paired with flashing lights or sounds, it was very difficult for the rats to relate the two experiences. But not when novel food was involved.

These distinguishing characteristics of food-aversion learning make excellent sense when considered in the light of the function that such learning serves in the natural environment. In general, poisons and spoiled foods do not make an individual ill immediately, but only after many minutes or several hours. Moreover, it is not the visual quality of the food that produces illness, but rather its chemical quality, detectable in its taste and smell. For example, a food that has begun to rot and makes an animal sick may look identical to one that has not begun to rot, but its taste and smell are quite different. Thus, to be effective, a learning mechanism for food aversion must tolerate long delays and be tuned especially to those sensory qualities that correspond with the food’s chemistry.

Counter to the conventional wisdom of his day, Garcia argued that the rats were prepared to make an association between nausea and food consumption (especially novel food), something that would be adaptive in the wild. A few years after Garcia published his work, Martin Seligman (1970) extended this idea, proposing that all associations between events and behavior are not equally learnable. Rather, there is a continuum of preparedness, such that animals (including people) are prepared by natural selection to make some associations and unprepared, or even contraprepared, for others. Prepared behaviors include the association between food ingestion and nausea, as shown by Garcia in rats, as well as learned behaviors that are vital to an organism’s survival, such as imprinting in ducks and geese (an infant bird forming an attachment to a moving and/or vocalizing stimulus, usually its mother), which is most easily acquired hours after hatching (see discussion of imprinting later in this chapter). Unprepared behaviors are those acquired through the normal processes of operant conditioning and usually take repeated trials to acquire. Confraprepared behaviors, in contrast, are those that are impossible or difficult to learn despite extensive training, such as the association between nausea and patterns of light and sounds in rats.

Seligman’s three-part classification shows that the rules of operant conditioning are not as uniform, or general, as Skinner and other behavioral theorists proposed; rather, there are some biological constrainfs on learning, shaped over the course of evolution, that make some associations more easily acquired than others.

Food-Preference Learning

The other side of the coin of learning to avoid harmful foods is learning to choose foods that satisfy a specific nutritional requirement. Just as rats can learn to associate the taste of a food with subsequent illness and thereafter avoid that food, they can also associate a taste with a subsequent improvement in health and thereafter prefer that food.

A number of experiments have shown that when rats are deprived of a mineral (such as calcium) or a vitamin that is essential for health, they will learn to prefer the flavor of a new food that contains that mineral or vitamin (Rozin & Schull, 1988; Tordoff, 2002). In one series of experiments, researchers deprived rats of thiamine (one of the B vitamins, essential for health) for a period of time and then offered them a choice of foods, only one of which contained thiamine (Overmann, 1976; Rozin & Kalat, 1971). Each food had a distinct flavor, and thiamine—which itself has no flavor—was added to a different food for different rats. The result was that, within a few days of experience with the foods, most rats strongly preferred the thiamine-containing food.

How did the rats “figure out” which food contained the thiamine? Close inspection of their eating patterns suggests a possible answer (Rozin & Kalat, 1971). When first presented with the choices, a rat usually ate just one or two of the foods. Then, typically after several hours, the rat would switch to a different food or two. Such behavior—eating just one or two foods at a time—seems ideally suited for isolating particular foods that lead to an increase or a decrease in health. If the rat had sampled all the foods at once, it would have had no basis for knowing which one had affected its health.

We don’t know if humans have a similar ability to learn which foods have a vitamin or mineral that we need, as no controlled experiments have been conducted to find out. It would not be ethical to deprive people of necessary nutrients for the sake of such research. However, there is evidence that people, as well as rats, learn to prefer a food that is high in calories (Brunstrom, 2005). This learning mechanism, which was no doubt valuable both to our evolutionary ancestors and to some people today, may have an unfortunate effect on those of us who are overweight and surrounded by wide choices of foods.

In the typical human flavor-preference learning experiment, college students are presented each day with one of two differently flavored foods, which is either laced with a high-calorie substance or not so laced. Initially the two foods are rated as equally pleasant (or unpleasant) in taste, but, over the course of days, the students’ average rating of the high-calorie food goes up, while their rating of the low-calorie food stays the same or declines (Brunstrum, 2005; Brunstrom & Mitchell, 2007). Apparently some delayed satisfying effect of the calories causes the students to develop a preference for the high-calorie version. This may be no news to people trying to lose weight, who are already convinced that all of human nature is stacked against them.

Learning from Others What to Eat

In addition to learning from their own experiences with foods, rats learn what to eat from one another. Newly weaned wild rats generally limit their diets to foods that older rats in the colony regularly eat. Through this means, they can avoid even tasting a food that older animals have learned is poisonous (Galef & Clark, 1971) and can choose, from the beginning, a nutritious food that older animals have learned to prefer (Beck & Galef, 1989). Similar results have been found with kittens (Wyrwicka, 1996). Even in adulthood, rats are strongly influenced by one another’s food choices. Bennett Galef (1990, 2002) has found that rats in a colony sniff near the mouth of a rat that has recently eaten and then show a strong preference for the food they had smelled on the demonstrator rat’s breath. Through this and other means, adult rats introduced into a new colony acquire the colony’s food preferences. The tendency to eat what others of one’s kind have been eating has been demonstrated in many other species of animals as well (Galef & Giraldeau, 2001).

Observational learning has its limits Children acquire the food preferences of their culture by observing their elders, but sometimes it takes a while.

Mitch York/Tony Stone/Getty Images

We humans don’t learn food preferences by smelling one another’s breath (at least not consciously), but we are certainly influenced by our observations of what those around us eat. In one experiment, children between 1 and 4 years old were more willing to taste a new food if they saw an adult eat it first than if they had never seen anyone eat it (Harper & Sanders, 1975). Other research suggests that children are most open to new foods from about 1 to 2 years of age, which is when they are most likely to be closely watched and fed by adults, and are least willing to try new foods between about 4 and 8 years of age, a time when they have greater freedom of movement and are not so closely watched but have not yet learned to distinguish foods from poisons (Cashdan, 1994). From this point of view, the finicky eating of 4- to 8-year-olds is an evolutionary adaptation that reduces the chance of eating something poisonous. However, even finicky eaters in this age range can be rewarded with stickers to try new foods, including vegetables, and can develop a liking for the new foods that will last at least 3 months after rewards have been stopped (Cook et al., 2011).

Food preferences can even begin while still in the womb. For example, in one experiment pregnant women ate anise-flavored food while others did not. (Anise tastes like licorice.) At birth and 4 days later, infants born to anise-consuming mothers showed a preference for anise odor, whereas those born to non-anise-consuming mothers displayed aversion or neutral responses to anise (Schaal et al., 2000).

Summary of Rules for Learning What to Eat

Suppose that you were a wise teacher of young omnivorous animals and wanted to equip your charges with a few rules for food selection that could be applied no matter what food was available. Two that you would probably come up with are these: (1) When possible, eat what your elders eat. Such food is probably safe, as evidenced by the fact that your elders have most likely been eating it for some time and are still alive. (2) When you eat a new food, remember its taste and smell. If the food is followed within a few hours by feelings of improved health, continue choosing foods of that taste and smell, but if you feel sick, avoid such foods.

Notice that these rules do not specify what to eat, but specify how to learn what to eat. The first rule describes a specific variety of observational learning, and the second describes a specific, efficient variety of associative learning. As you have just seen, rats do in fact behave in accordance with these rules, and humans may also. Of course, we assume that these rules have been imparted not by a wise teacher of young omnivores but by natural selection, which has shaped the brain to operate automatically in accordance with the rules.

Other Examples of Special Learning Abilities

Food selection is by no means the only domain in which special learning abilities have apparently come about through evolution. Here are some other well-studied examples.

Prepared Fear-Related Learning

Do you remember the demonstration by Watson and Rayner (1920), in which a young child was conditioned to fear a white rat by pairing it with a loud noise? Several years later, a graduate student working with Thorndike named Elsie Bregman (1934) tried to repeat that demonstration with one important modification. Instead of using a rat as the conditioned stimulus, she used various inanimate objects, including wooden blocks and pieces of cloth. Despite numerous attempts, with 15 different infants as subjects, she found no evidence of conditioning. What are we to make of this apparent discrepancy? One possibility, suggested by Martin Seligman (1971), is that people are biologically predisposed to acquire fears of situations and objects, such as rats and snakes, that posed a threat to our evolutionary ancestors and are less disposed to acquire fears of other situations and objects.

More recently, Susan Mineka and her colleagues (1984) showed that rhesus monkeys are not afraid of snakes when first exposed to them but easily learn to fear them. In one experiment, monkeys raised in the laboratory did not react fearfully to snakes until they saw a monkey that had been raised in the wild do so. After that, they showed strong fear reactions themselves when a snake was present (see Figure 4.18). In subsequent experiments, Michael Cook and Mineka (1989, 1990) used splicing to produce films in which a monkey was shown reacting fearfully in the presence of various objects, including toy snakes, flowers, and a toy rabbit. Through observing the films, monkeys that previously feared none of these objects developed a fear of toy snakes (and real snakes) but not of flowers or toy rabbits.

Figure 4-18: A biologically prepared learned reaction Monkeys that have never been harmed by snakes nevertheless learn quickly to fear them through watching the fearful reactions of other monkeys.

Susan Mineka

From an evolutionary perspective, this learning bias makes a good deal of sense. In some regions where rhesus monkeys live there are dangerous snakes, but in other regions all of the snakes are harmless. In places where snakes are harmless, an inflexible instinctive fear of them would be maladaptive. Thus, the learning mechanism may have evolved because it allows monkeys living in areas where snakes are dangerous to learn quickly to fear and avoid them, while it allows monkeys living elsewhere to go about their business relatively oblivious to snakes. We humans also vary greatly in the degree to which we fear snakes. Research suggests that we learn to fear snakes and other objects that posed threats to our evolutionary ancestors—such as spiders, rats, and angry faces—more readily than we learn to fear equally dangerous objects that were not present in our early evolutionary history, such as electrical outlets, guns, and automobiles (Mineka & Öhman, 2002; Seligman, 1971).

How early do such fears develop? In a clever study, Judy DeLoache and Vanessa LoBue (2009) showed 7- to 9-month-old infants and 14- to 16-month-old toddlers videos of snakes and other animals (giraffes, rhinoceroses). The infants initially showed no greater fear to the snakes than to the other animals, suggesting that a fear of snakes is not inborn. The infants and toddlers then saw brief video clips of snakes and other animals associated with either a happy or fearful voice. Both the infants and toddlers looked longer at the snakes when they heard the fearful voice than when they heard the happy voice. There was no difference in looking time to the two voices when they saw videos of other animals. DeLoache and LoBue suggested that, much like Mineka’s monkeys, human children are prepared to acquire a fear of snakes (see also LoBue & DeLoache, 2010; Rakison, 2005).

Imprinting in Precocial Birds: Learning to Identify One’s Mother

Some of the earliest evidence for specialized learning abilities came from studies of young precocial birds. Precocial birds are those species—such as chickens, geese, and ducks—in which the young can walk almost as soon as they hatch. Because they can walk, they can get separated from their mother. To avoid that, they have acquired, through natural selection, an efficient means to determine who their mother is and a drive to remain near her. The means by which they learn to recognize their mother was discovered by Douglas Spalding near the end of the nineteenth century.

Spalding (1873/1954) observed that newly hatched chicks that were deprived of their mother, and that happened to see him (Spalding) walk by shortly after they were hatched, would follow him as if he were their mother. They continued to follow him for weeks thereafter, and once attached in this way they would not switch to following a real mother hen. Some 60 years later, Konrad Lorenz (1935/1970) made the same discovery with newly hatched goslings. Lorenz labeled the phenomenon imprinting, a term that emphasizes the very sudden and apparently irreversible nature of the learning process involved. It’s as if the learning is immediately and indelibly stamped in.

One interesting feature of imprinting is the rather restricted critical period during which it can occur. Spalding (1873/1954) found that if chicks were prevented from seeing any moving object during the first 5 days after hatching and then he walked past them, they did not follow. Instead, they showed “great terror” and ran away. In more detailed studies, Eckhard Hess (1958, 1972) found that the optimal time for imprinting mallard ducklings is within the first 18 hours after hatching.

Konrad Lorenz and followers Lorenz conducted research on imprinting and many other aspects of behavior in ducks and geese. These geese, which were hatched by Lorenz in an incubator, followed him everywhere, as if he were their mother.

Time & Life Pictures/Getty Images

Although early studies suggested that young birds could be imprinted on humans or other moving objects as easily as on their mothers, later studies proved otherwise. Given a choice between a female of their species and some other object, newly hatched birds invariably choose to follow the former. Experiments with chicks indicate that this initial preference centers on visual features of the head. Newly hatched chicks will follow a box with a chicken head attached to it as readily as they will a complete stuffed chicken and more readily than any object without a chicken head (Johnson & Horn, 1988). The experience of following the object brings the imprinting mechanism into play, and this mechanism causes the chicks to be attracted thereafter to all the features of the moving object (Bateson, 2000). Under normal conditions, of course, the moving object is their mother, so imprinting leads them to distinguish their mother from any other hen.

It’s not just sight that is involved in imprinting, but also sound (Grier et al., 1967). If you put precocial birds such as ducks in circular tub and play the maternal call of their species from a speaker on one side and the maternal call of another species from a speaker on the opposite side, they will invariable approach the speaker playing the call from their own species. It’s easy to look at these findings and infer that imprinting is a classic example of an instinct—something that is under strong genetic control and requires no experience for its expression. But are these young birds really devoid of all experience? They’ve actually heard their mother’s call while still in the egg. However, when ducklings are removed from their mothers and hatched in an incubator, they still approach their maternal call, even though they had never heard it before. But they have heard the peeping of the other ducklings in the brood of eggs, and their own peeps, for that matter. (Ducklings start peeping several days before hatching.) When this auditory experience is removed (the duck’s own peeping by doing a little surgery while still in the egg that prevents it from making any sound—it wears off several days after hatching), the ducklings then approach the speakers randomly. These experiments, done by Gilbert Gottlieb (1991), show that even something that looks like a clear-cut instinct such as auditory imprinting still involves some experience. Natural selection has worked so that the brain, sensory organs (in this case, those associated with hearing), and experience are coordinated to produce a valuable adaptive behavior. For the most part, it’s only the poor duckling hatched in Gottlieb’s lab that will fail to get the appropriate experience for imprinting. But as we stressed in Chapter 3, behavior is always the product of genes and experience, and it sometimes takes a lot of effort to discover what those experiences are.

In sum, we have here a learning process for which the timing (the critical period), the stimulus features (characteristics typical of a mother bird of the species), and the behavioral response (following) are all genetically prepared—in interaction with the environment—in ways that promote its specific adaptive function—staying near the mother.

Specialized Place-Learning Abilities

Many animals have specialized abilities for learning and remembering specific locations that have biological significance to them. As one example, Clark’s nutcrackers (a species of bird inhabiting the southwestern United States) bury food in literally thousands of different sites, to which they return during the winter when the food is needed (Gould-Beierle & Kamil, 1999). Experiments have shown that the birds’ abilities to find each location depend on their memories of visual landmarks, such as stones, near the site (Kamil & Balda, 1985; Shettleworth, 1983). In other experiments, various bird species that hide seeds have been found to remember spatial locations better than do species that don’t hide seeds and to have an enlargement of an area of the brain, called the hippocampus, that is crucial for spatial memory (Papini, 2002; Shettleworth & Westwood, 2002).

A seed-hiding bird Clark’s nutcrackers, like many bird species, hide seeds in many different sites for the winter. Their ability to remember each hiding place is an example of a specialized learning ability.

© Joe McDonald/Corbis

A quite different example of specialized place learning is the ability of Pacific salmon to return to their hatching grounds. Salmon that hatch in small streams in the northwestern United States migrate into the Pacific Ocean, where they swim around for 5 years or more, carrying a precise memory of the unique smell of the water in which they hatched. Then, when they are ready to spawn, they use their sense of smell to find their way back to the same stream from which they had come (Hasler & Larsen, 1955; Navitt et al., 1994).

So, in certain very specific ways, species of birds and fish appear to be “smarter” than chimpanzees or people. The more we understand about animal behavior, the more it becomes apparent that intelligence is a relative concept. To make the concept meaningful, we have to ask, “Intelligent at what?” Animals appear extraordinarily intelligent when we observe them in nature dealing with the kinds of learning tasks and problems for which they have been biologically equipped by natural selection. The same animals appear relatively stupid when we observe them on early trials in artificial test apparatuses, such as Thorndike’s puzzle boxes or Skinner’s operant conditioning chambers. The intelligence of animals comes not from a general ability to reason but from specialized learning abilities that have evolved over thousands of generations in the wild.