Techniques for Obtaining Food

The omnivorous blue jay faces a final cognitive challenge: it must learn to extract food from the environment. It may need to do anything from prying up bark to capture insects underneath to opening a discarded berry container. The jay must learn those food-handling techniques that are not innate.

Instrumental Conditioning

Instrumental or operant conditioning refers to a situation in which an animal learns that its own behavior, in the presence ofcertain stimuli, is instrumental in causing a particular outcome. The study of instrumental conditioning began with the work of E. L. Thorndike (1874—1949), who conducted the first controlled studies of learning in the laboratory (Thorndike 1911). To compare the "intelligence" of species directly, he developed cages known as puzzleboxes, in which a hungry animal had to trigger a release mechanism from inside the box to reach food outside. When first placed in a puzzle box, an animal moved randomly until it accidentally triggered the escape mechanism. In subsequent trials, the animal tended to repeat the behaviors that had occurred just before its escape, whether or not those behaviors opened the apparatus. This process of repeating the behaviors that preceded success produced a gradual, negatively accelerated learning curve (as discussed under "conditioning mechanisms" in section 4.4) when Thorndike plotted time to escape against trial number. From this observation, Thorndike formulated the law of effect: in a particular context, behavior that is followed by a satisfying event strengthens the association between the context and the behavior, causing the behavior to become more likely should the context recur. This law formed the basis for instrumental learning theory.

Behavioral psychologists use two types of procedures to study instrumental conditioning: discrete-trial and free-operant procedures. In discrete-trial procedures, the subject makes the instrumental response once per trial, such as triggering the escape mechanism of a puzzle box. Likewise, an experiment may require that a rat turn left in a maze to obtain a reward. After the response, the investigator removes the subject from the apparatus. In free-operant procedures, the subject repeats its response freely. The operant chamber is the original and most typical free-operant apparatus and has proved to be a critical tool in the study of instrumental conditioning due to the ease of collecting data.

Both types of procedures rely on the pairing of a behavior with a reinforcing outcome, or reinforcer, such as food. One can deliver the reinforcer every time the subject makes the required response (continuous reinforcement) or only every so often (partialreinforcement). Behavioral psychologists use fourbasic schedules of partial reinforcement. In an interval schedule the subject earns reinforcers for responses after a given time interval. In a ratio schedule the subject earns rein-forcers after a specified number ofresponses, such as lever presses or key pecks. The time and number requirements can be fixed (staying the same from trial to trial) or variable (changing from one trial to next), giving four possibilities: variable interval, variable ratio, fixed interval, and fixed ratio schedules. The reinforcement schedule influences the behavior of a subject in predictable ways; for example, subjects in fixed interval schedules begin to respond just before the end of the fixed interval (Roberts and Church 1978; see Domjan 1998 for thorough discussion of instrumental conditioning).

Biology constrains instrumental conditioning, just as it does classical conditioning. Foragers do not have to learn all the behaviors associated with feeding; the corollary of this statement, that some behaviors cannot be unlearned, is instinctive drift.

Breland and Breland (1961) first demonstrated instinctive drift in their instrumental conditioning of animals for commercial advertising. For example, they would train a raccoon to drop a coin into a box using the method of successive approximations, in which they rewarded the animal for behaviors progressively closer to the desired one. However, the raccoon's behavior proved less malleable than predicted. It would rub the single coin, or later two coins, together, thereby delaying reinforcement. Despite the obvious cost in reinforcements, the raccoon could not suppress its innate foraging movements of rubbing small objects together. These findings have inspired a movement toward a functional perspective in learning theory that emphasizes biological relevance (Domjan 2005).


A jay may learn foraging techniques by imitating a conspecific's successful technique. However, as mentioned above, researchers must carefully identify the processes involved. In one famous example, a wild population of English blue tits learned to open milk bottles and drink the cream (reviewed in Shettle-worth 1998). Debate ensued over how this skill spread through the population. Sherry and Galef (1990) showed experimentally that the spread of this skill did not require imitation, but could have been accomplished by local enhancement and social facilitation.

Imitation can also be confused with emulation. Whereas when an individual imitates, it copies the action ofa model, when an individual emulates, it learns that the environment can be manipulated to achieve a particular goal. For instance, an emulator might see a model open a hinge by poking out a pin and learn only that the pin comes out. During replication, an imitator would poke the pin out, whereas an emulator might pull it. Emulation is arguably as cogni-tively complex as imitation, but may require different mechanisms. The mechanisms involved in both processes are still highly controversial (see reviews in Caldwell and Whiten 2002; Zentall 2004).

In the most definitive test for imitation, the two-action test, models demonstrate different solutions to the same problem to different experimental groups. If the subjects use the method they observed, this indicates imitation rather than emulation. For example, demonstrator Japanese quail depressed a treadle with the foot or the beak while one experimental group watched each technique. When tested, the quail generally used the technique they had witnessed. In a further demonstration, observers were more likely to imitate a demonstrator that received food rewards for its actions than one that did not, suggesting that the imitator may also represent the action's purpose—in this case, obtaining food (reviewed in Zentall 2004).

A recent study distinguished between action imitation and cognitive imitation (Subiaul et al. 2004). In a typical serial learning task, demonstrator rhesus monkeys were taught series of photographs. The monkeys were required to press each photograph on the screen in order, although the location of the photographs was changed in each screen. The observer monkeys were able to gain some information about ordinal position by watching the demonstrators that raised their performance significantly above baseline. This effect was not the result ofsocial facilitation or emulation based on the feedback given by the computer. Therefore, under some circumstances, animals may learn rule-like information from observing conspecifics.

In other cases, animals may learn not to imitate one another. Pigeons in a situation in which the actions of a skill demonstrator deliver food to the observer regardless of the observer's behavior do not learn the skill. In contrast, with a small change in the apparatus, the observer is not rewarded during the experience, and under these conditions, observers readily learn to copy the movements of the demonstrator (Giraldeau and Lefebvre 1987). This observation suggests that learning of a particular food-handling technique may depend on whether the subject stands to gain from learning that skill.


If animals can learn from others, it stands to reason that behaviors that promote such learning experiences could also evolve. Caro and Hauser (1992) defined teaching functionally as a change in behavior in the presence of a naive individual that is not immediately beneficial to the teacher and helps the naive individual learn. Common chimpanzees may teach their young how to use stone hammers and anvils to open coula nuts (Boesch 1991). Mother chimpanzees in Tai National Park behaved in ways that could facilitate learning, including leaving hammers near anvils when offspring were present, although they usually carried the hammers away (the hammers were used by offspring on 46.2% of 387 such occasions), or bringing nuts or hammers to a young chimpanzee at an anvil (588 occasions, leading to a 20% increase in nuts eaten per minute by offspring). On two occasions, mothers adjusted the orientation of the hammer or the nut, seemingly correcting the infant's use of the technique.

Teaching may be prevalent in species with elaborate predatory behavior, such as birds of prey and carnivores. Among these species, ospreys, domestic cats, and cheetahs demonstrably increase the foraging effort they require from their offspring, from bringing them dead prey to live but wounded prey and finally live prey that are allowed to escape for recapture (reviewed in Caro and Hauser 1992). Some spiders may behave similarly (Wilson 1971). In most of these species, it remains to be demonstrated that this behavior actually facilitates learning. However, a laboratory study with domestic cats found that kittens whose mothers were present and interactive during exposures to live prey learned hunting skills earlier than control kittens whose mothers were not present (reviewed in Caro and Hauser 1992).

As with imitation, cognition researchers want to understand the cognitive processes underlying teaching. It might seem that teachers require a theory of mind (a representation another's mental states) to be sensitive to the needs of the pupil. Caro and Hauser maintain that although such a representation would "almost certainly enhance the utility of teaching" and may be present in some species, it is not necessary. To be useful, the teacher must have a mechanism for discriminating which individuals lack skills or knowledge. Distinguishing the actual mechanisms involved will require experimental manipulations. As with other behaviors we have discussed, species differences in the cognitive basis of teaching are likely to emerge.


Can an animal use existing knowledge to produce a novel foraging technique? One way of doing so might be through insight, a novel viewpoint on a situation that can enable undetected relationships to suddenly become apparent. Animals must solve problems without overt trial-and-error learning, innate programmed responses, or observation before insight can be considered. Early experiments by Kohler (1925) are frequently cited as the seminal research on insight in human and nonhuman psychology (reviewed in Ormerod et al. 2002). Working with a group of captive chimpanzees, in one experiment Kohler (1925) hung bananas from a high place and gave the chimpanzees a box. The chimpanzees solved this problem by moving the box so that stepping on it allowed them to reach the bananas. Later they were also able to stack several boxes to solve a similar problem (fig. 4.6). Success tended to come suddenly after a period ofno progress, not gradually after many approximations, suggesting insight.

Figure 4.6. Testing for insight in chimpanzees. Captive chimpanzees trying to reach a hanging banana appearto suddenly realize a solution to the problem, suggesting insight. In the drawing atthe right, a chimpanzee has stacked three boxes to reach the bananas overhead. In the drawing atthe left, another is in the process of stacking four boxes to reach the goal. (After photographs in Kohler 1925.)

Figure 4.6. Testing for insight in chimpanzees. Captive chimpanzees trying to reach a hanging banana appearto suddenly realize a solution to the problem, suggesting insight. In the drawing atthe right, a chimpanzee has stacked three boxes to reach the bananas overhead. In the drawing atthe left, another is in the process of stacking four boxes to reach the goal. (After photographs in Kohler 1925.)

Although Kohler's chimpanzees had no previous experience with the exact problem presented to them, an experiment by Epstein et al. (1984) cast doubt on Kohler's results. Pigeons trained separately to push a box toward a randomly placed target and to stand on a box to peck a fake banana put these behaviors together to solve the equivalent problem, reportedly through stimulus-response chaining rather than insight. Pigeons trained to perform only one of the subtasks (e.g., climbing but not pushing) failed to reach the banana. However, why the pigeons pushed the box specifically toward the banana was unclear.

A study of hand-reared ravens controlled more precisely for previous experience (Heinrich 1995). The ravens faced the following problem: how to retrieve food attached to a branch by a long string. A raven had to land on the branch and use its beak and foot to pull up the string in stages. Once the raven obtained the food, it had to suppress its natural tendency to fly away because the food was still connected by the string. Despite the complexity of the motor sequence involved, several ravens performed this task correctly without apparent trial-and-error learning. Although pulling and stepping may be an innate motor pattern in birds (see review in Thorpe 1963), several ravens never completed the task, and the ones that did showed a prolonged delay. Heinrich argued that assembly ofthe steps into a coherent, novel action, not the origin ofthe individual steps, is crucial for demonstrating insight. These studies suggest that under appropriate circumstances, animals may create novel foraging techniques without trial and error.

Tool Use

Techniques for obtaining food may include the use of tools. Many animals have been observed using tools, including insects, crabs, rodents, elephants, and many primates (reviewed in Griffin 2001). A tool is a material object that an animal manipulates as an extension of its body to achieve an immediate goal. Sea otters, for example, use a rock to crack a prey item's shell; Egyptian vultures and chimpanzees use rocks in a similar way. Many other taxa use a thin stick to extract insects or other food items from crevices; examples include the Darwin's woodpecker finch, common chimpanzee, and New Caledonian crow. Tool use may be acquired by the processes described previously or may be innate. Cognition researchers are particularly interested in whether the tool-using animal understands the relationship between the tool and its use (the means-ends or cause-effect relationship).

Hauser (1997) demonstrated that cotton-top tamarins can discriminate the functional properties ofa tool. Hauser gave tamarins a choice between a functionally intact tool and one that he had modified to make it nonfunctional. For example, the tool might be a cane placed with a piece ofcandy inside its hook so that the monkey could use it to pull the candy in. A nonfunctional option might be the cane with the candy outside of its hook. In a series of experiments, tamarins chose the functionally intact tool more frequently.

However, capuchin monkeys can successfully use tools without understanding the means-ends relationship. Visalberghi and colleagues (Visalberghi and Limongelli 1996) tested capuchins and chimpanzees using a clear plastic tube with a cuplike depression in the middle, known as the "trap tube" (fig. 4.7), and a reward placed outside the trap at one end of the tube. To extract the food, the animal had to push a stick through the tube, pushing the food out of the tube while avoiding the trap. Previously, three of four monkeys had used sticks to obtain rewards from tubes without traps (Visalberghi and Trinca 1989). With the trap, however, the monkeys needed to push from the correct end. When tested with the trap, three of four monkeys could not extract food more than half the time, even after 140 trials. The fourth monkey learned the task after 90 trials, but apparently learned by rote. She continued to push from the side farthest from the food (as the trap requires), even when the investigators rotated the trap upward (and it no longer acted as a trap). Chimpanzees

Figure 4.7. Testing for means-ends understandingwith the trap tube. In this experiment, the subject must use a stick to push a reward out of the tube. If the subject pushes from the wrong direction, the reward will fall into the trap. Here, a capuchin monkey is about to push the reward into the trap. (After a drawing in Shettleworth 1998 of a photograph in Visalberghi and Limongelli 1994.)

Figure 4.7. Testing for means-ends understandingwith the trap tube. In this experiment, the subject must use a stick to push a reward out of the tube. If the subject pushes from the wrong direction, the reward will fall into the trap. Here, a capuchin monkey is about to push the reward into the trap. (After a drawing in Shettleworth 1998 of a photograph in Visalberghi and Limongelli 1994.)

showed more signs of means-ends understanding in performing this task. Of five, two solved the original trap tube and transferred this skill to a variant in a way that suggested they understood the intermediate goal of avoiding the trap.

Modification of tools for a particular task also suggests understanding of the means-ends relationship. New Caledonian crows modify their tools into two different shapes (a hook or a jagged tool) as appropriate for removal ofinsects from different holes, and they shorten the length of a tool when necessary (Hunt 1996). Recent studies have shown that these crows can choose the right length of stick without trial and error (Chappell and Kacelnik 2002), and one individual bent a piece of wire into an appropriate tool (Weir et al. 2002).


Animals can use different cognitive skills to acquire foraging techniques. A forager may learn techniques by trial and error through instrumental conditioning, but within the constraints of innate biases. Imitation may be an efficient way to learn a successful technique from a conspecific. Teaching may also play a role in transferring foraging techniques. Sometimes animals may use insight to produce a correct technique the first time they encounter a problem. Many animals use tools to forage, though they may not always understand why the tool works. The cognitive mechanisms underlying many of these behaviors are still being investigated.

Was this article helpful?

0 0

Post a comment