CONCEPT I: The Nature of Learning

Learning is involved in almost every phenomenon psychologists study, although it is neither observable nor measurable. Psychologists measure performance and infer that changes in performance reflect learning, although factors such as fatigue, emotion, motivation, health, and physical maturation can affect performance without affecting learning. Learning, thus, gives us the potential for a change in performance. Learning is defined as a relatively permanent change in performance potential that arises from experience. Different kinds of learning include associative learning (learning that one event is associated with another), of which classical conditioning and operant conditioning are examples, and cognitive learning (learning that involves concept formation, schemas, theories, and other mental abstractions).

CONCEPT II: Classical Conditioning

Classical conditioning involves reflex behavior in which a stimulus naturally produces (or elicits) an involuntary response. In classical conditioning, an originally neutral stimulus (one that elicits no response) is repeatedly presented just before the reflex stimulus (or unconditioned stimulus, UCS). The neutral stimulus eventually becomes associated with the unconditioned stimulus-unconditioned response (the reflex response) pattern and soon comes to elicit that unconditioned response (UCR), even in the absence of the unconditioned stimulus. This previously neutral stimulus becomes a conditioned stimulus (CS), and the reflex response it produces is called the conditioned response (CR). Ivan Pavlov, a Russian physiologist, was one of the first scientists to study the phenomenon of classical conditioning systematically.

One of the necessary conditions for classical conditioning is a contingency, or relationship, between CS and UCS. Generally, the CS must precede the UCS by only a short time for maximal conditioning to occur. This contingency situation is called forward conditioning. Backward conditioning involves presenting the UCS before the CS and is much less effective in establishing a conditioned response. Random presentation of CS and UCS also is ineffective in producing a CR. The CS and UCS contingency seems to depend upon the subject's use of the CS to predict the UCS. When this relationship is unclear, conditioning is more difficult to establish, as Rescorla's work demonstrates.

In classical conditioning, extinction is the slow weakening and eventual disappearance of the conditioned response. It is accomplished by no longer pairing the CS with the UCS. If an animal is removed from an experimental chamber for a while after its response has been extinguished and is then put back in, the extinguished response will recur. This is called spontaneous recovery.

Generalization is the expression of a learned response in a new but similar situation. It can be limited through the technique of discrimination training, in which only certain stimuli (the discriminative stimuli) are associated with reinforcement. Discrimination is learning to make a particular response only to a particular stimulus. A generalization gradient depicts response strength to both the original stimulus and similar stimuli. Typically, the more similar a new stimulus is to the original stimulus, the greater the generalization is. Together, generalization and discrimination work to help us respond to new situations in appropriate ways. Emotional responses are especially susceptible to classical conditioning, and difficult to extinguish.

John B. Watson found that classically conditioned fear could be diminished by counterconditioning, the pairing of a pleasant stimulus with a feared one. A similar technique, called systematic desensitization, is now being used on phobias. A person is taught to relax and then gradually introduced to anxiety-producing situations, starting with mildly fear-arousing ones and working up to situations that, before desensitization, would have produced excessive anxiety. Classical conditioning can also be used to instill desired behaviors, as in the case of treating children who wet their beds.

CONCEPT III: Operant Conditioning

Using cats and puzzle boxes, Edward L. Thorndike examined another form of association learning which we now call operant conditioning. His law of effect states that responses which lead to unsatisfying consequences are weakened.

B. F. Skinner proposed that the basic mechanism for controlling human behavior is the application of the principles of operant conditioning. According to this mechanism, a consequence that increases the probability of the behavior that caused it is called a reinforcement (or reward). A consequence that results in suppression of the causal behavior is called a punishment. Operant behavior is any behavior the organism emits spontaneously—that is, without a stimulus.

Positive reinforcement increases the frequency of response because that response is followed by a pleasant stimulus. Negative reinforcement also increases the frequency of a response, but it does so by removing a painful or unpleasant stimulus. (Punishment decreases the frequency of response by following it with an unpleasant stimulus.) Negative reinforcement can be used to establish escape learning, or learning a response that allows the organism to escape from an unpleasant situation, and avoidance learning, in which the organism learns a response which prevents a negative event from occurring. When an organism mistakenly assumes a contingency between a response and some consequences when in fact none exists, the result is called superstitious behavior.

Establishing a conditioned operant response is difficult because the response must first occur naturally before it can be reinforced and strengthened. The Skinner box (a cage with a lever that delivers a food reward when pressed) is one of many devices that control the presentation of rewards to experimental animals. It is useful in the study of operant conditioning because it allows easy measurement of an animal's responses.

Sometimes researchers who study operant responses choose to work with responses which often occur naturally because they are easier to train. When a subject does not typically make the response a psychologist wants to condition, a technique called shaping is used. Shaping involves reinforcement for closer and closer approximations of the desired behavior.

Reinforcement can be applied in many patterns or schedules. Continuous reinforcement, in which every response is followed by a reward, is usually most effective for establishing a new behavior. Partial reinforcement schedules, in which only some instances of the desired behaviors are rewarded, are usually the best way to maintain a behavior. There are four types of partial schedules. A fixed-ratio schedule provides a reward each time the subject makes a specified number of responses. A fixed-interval schedule rewards the first response made after a specified period of time has elapsed. Variable-ratio and variable-interval schedules are like their fixed counterparts, except the number of responses or amount of elapsed time varies randomly from trial to trial. Variable schedules are, thus, unpredictable, and therefore tend to produce high response rates and persistent behaviors.

Stimulus control refers to the situation in which the reinforcement schedule operates only in the presence of a particular stimulus—for example, when a red light is lit. Consequently, the subject learns to respond only when the stimulus is present. This stimulus is a discriminative stimulus because it allows you to distinguish, or discriminate, this stimulus from others.

Primary reinforcers, such as food or water, satisfy a basic biological need. Secondary, or conditioned reinforcers, such as praise or money, signal that a primary reinforcer is forthcoming. Learning a sequence of behaviors that eventually ends with primary reinforcement is called chaining.

Extinction occurs in operant conditioning when the reinforcement or punishment no longer follows the response. Although the probability of the response occurring eventually decreases during extinction, at first it may increase. Spontaneous recovery may occur in operant, as well as classical, conditioning situations.

Operant conditioning techniques have been applied to humans in the form of programmed instruction and behavior modification. Programmed instruction involves active repetition, frequent testing, and immediate feedback for every response. The most sophisticated method of programmed instruction is computer-assisted instruction, which increases the amount of branching that can be included in the system. Behavior modification is the selective use of operant conditioning principles to change human behavior. Today, behavior modification is also known as contingency management. Reward programs often consist of token economies, in which tokens (such as poker chips that can later be exchanged for other reinforcers) are earned by demonstrating appropriate behavior.

Punishment is generally successful in helping to suppress maladaptive behaviors, but it has some limitations. It is fairly temporary; it can cause emotional disturbances; and it, and the circumstances that surround it, can come to be aversive stimuli to be escaped or avoided. Furthermore, it does not indicate what the desirable behavior would be. Consequently, it is usually best used in conjunction with reinforcement of appropriate behavior.

CONCEPT IV: Biological Constraints on Associative Learning

If an organism is not biologically disposed to learn a behavior, such learning is difficult to accomplish. The biological or genetic predisposition to learn is called prepared learning. When a trained animal reverts to genetically based behaviors, like a pig's rooting behavior, the reversion is called instinctive drift.

A series of experiments conducted by Garcia demonstrates that some stimulus-response associations are learned much more easily than others. Rats, for instance, associated the taste of a substance with physical illness and the visual cues in a situation with electric shock, but they did not make the reverse associations. Garcia explained this tendency as a function of the evolutionary history of the species: natural selection has favored a nervous system that allows easy learning of contingencies crucial to survival.

Bernstein has found that cancer patients associate foods eaten prior to chemotherapy with the nausea caused by the therapy; so novel foods introduced prior to therapy can serve as scapegoats for the nausea association. Pretherapy fasting also prevents the food associations. The goal of these applications is to prevent unnecessary weight loss.

The cognitive approach to learning, first proposed by Edward Tolman, objects to the exclusion of the mental activity that goes on inside the organism during the learning process. Cognitive psychologists stress that association learning involves learning an expectation that two events will be associated. Blocking occurs when prior conditioning to one stimulus prevents the conditioning of a second, thus supporting the idea that we tend to select for learning only those associations that will be useful.

Support for the cognitive view comes from experiments demonstrating that learning can occur without reinforcement. Latent learning experiments, for example, show that an organism can learn a new behavior but not demonstrate it until an incentive to do so arises. Social learning experiments demonstrate that organisms can learn through observation and without direct reinforcement. Such learning is called observational learning. It can include the use of vicarious reinforcement and punishment, which means the subject's behavior is modified by watching someone else being rewarded or punished. Cognitive psychologists, like Bandura, also stress intrinsic reinforcement and punishment, in which we assess our behavior against our own internal standards.

According to social learning theorists, we do not take camera-like snapshots of other people's behavior, but, instead, we form mental abstractions, called schemas, about what they are doing. These schemas, once formed, are difficult to change, since they guide our expectations about our experiences. Some forms of psychotherapy have been developed to help people disconfirm unrealistic schemas they hold. Cognitive theories contribute to our understanding of many areas of psychology.