Albert Bandura\'s Social Learning Theory

¶ … dominant models of human behavior by the late 1950s and early 1960s were based on Neo-Freudian models and B.F. Skinner's brand of operant behaviorism. However, there were theorists that rejected the mechanistic views of behaviorism and Freudian instinct-drive-based models. Perhaps the most influential of these theorists was Albert Bandura. Bandura had received his Ph.D. from the University of Iowa and had been exposed to the work of Robert Spears who was studying familial influences of social behavior and identification in children. Bandura was also heavily influenced by other theorists at Iowa such as James Dollard and Neal Miller who had merged Freudian and Hullian learning principles. Bandura believed that learning principles were sufficient to explain and predict behavior, but he also believed that humans thought and regulated their behavior and were not at the mercy of environmental stimuli as in Skinnerian models of behavior. Furthermore, he believed that many functions of personality and learning involve interactions with other people and therefore any theory of personality or of learning should take interpersonal factors into account.

Bandura's social learning theory is based heavily on the ground breaking notions of Kurt Lewin (1943) and his Field Theory. Lewin's Field Theory was radical for its time given Freudian notions of behavior that dominated, but by today's standards his ideas seem obvious. Lewin developed the famous B = f (P, E) formula that states behavior is a function of the person and the environment. Bandura's notion of reciprocal determinism (Bandura, 1978) improves on Lewin's concept by maintaining the three factors of B, P, and E. By adding reciprocal lines of influence. Thus, the person has direct influences on the environment and his/her behavior and both can directly influence the person. Likewise the environment and the person's behavior also directly and reciprocally influence one another (see Figure 1).

Bandura believed that this reciprocal determinism, the idea that people influence their fate by controlling environmental forces but are also influenced by them, was the basic principle in his social learning theory (Bandura, 1997b). Moreover, Bandura departed from the traditional views of behaviorists in asserting that learning could occur without a change in one's behavior and without direct reinforcement.

In spite of the reciprocal influences of behavior, environment, and personal factors Bandura also understood that must be a center to the process. Bandura termed this center the self-system (Bandura, 1978). This self-system refers to a set of cognitive structures that effects perception and evaluation of the person, environment, self and others' actions and also and the regulation of one's behavior. In a sense there are reciprocating factors within the self-system. The self-system engages in constant monitoring of the self and environment. People evaluate themselves, their actions, and their goals based on their experience. These evaluations lead to important expectations. Bandura defined outcome expectations as estimates that one's actions will lead to specific outcomes, which are often learned. These outcome expectations are heavily influenced by internal notions of self-efficacy, one's perception of how well or poorly one can function in a given situation (Bandura, 1977a). The concepts of self-efficacy and outcome expectations are key components to behavioral change and to learning. The other important self-efficacy expectation in Bandura's social learning theory is the belief that a person can effectively perform the behavior required to produce the desired outcome (efficacy expectations).

Bandura viewed efficacy expectations as an important part of adaptation and coping. For example, Bandura (1982) compared group and person performances at on tasks at different levels of perceived self-efficacy and found that there was a direct relationship between self-efficacy and performance such that higher levels of perceived self-efficacy were associated with higher levels of performance. However, this is a relationship that could be explained either way: self-efficacy could lead to higher performance or better performers have greater levels of perceived self-efficacy. When self-efficacy of the individual or the group was increased performance subsequently increased at both the group and individual levels, indicating that perceived self-efficacy affects performance. Bandura, Adams, Hardy, and Howells, (1980) provided clinical support for increasing self-efficacy in the treatment of anxiety. People with agoraphobia were subjected to cognitive-behavioral treatment that increased their perceptions of being able to leave their homes safely and this was associated with a significant increase in their leaving their homes alone and performing daily activities such as shopping.

Bandura saw reinforcement as a causal mechanism of learning; however, he also maintained that we can learn by being reinforced for something, observing someone else perform, by vicarious reinforcement (by observing a model being reinforced), or without reinforcement at all (Bandura, Adams, & Beyer, 1977). Bandura believed that much learning occurs through the expectation of reinforcement by noting how others are reinforced or not reinforced for their actions. These observations lead to cognitive expectations as to the results of one's behavior. Reinforcement guides learning and behavior based on the anticipatory outcomes one has and learning is not always accomplished by direct reinforcement in the Skinnerian sense. These expectations can lead to developing other expectations about the consequences or outcomes of future actions based on past actions. Through interactions with socializing agents such as parents, peers, teachers, etc. people develop internalized behavioral standards and learn to reward themselves by self-approval or to degrade themselves by self-criticism. These anticipatory and internal self-reinforcing mechanisms guide behavior, foster new learning, and maintain consistency with established behaviors.

Bandura's social leaning theory then expanded the boundaries of what constitutes or defines reinforcement. Observers can be reinforced by a model such as when we adopt attitudes and the behaviors of peers. Observers can be reinforced by a third party as when we model the behavior of another such as an older sibling and are given praise by a parent. The modeled behavior can result in reinforcing consequences. For example, a person could observe how exercise such as bike riding is fun and starts riding a bike to receive enjoyment.

Bandura's social learning theory also places a great deal of emphasis on vicarious learning and on observational learning, learning by watching a model perform a task. In possibly his most famous study Bandura, Ross, and Ross (1961) demonstrated that children can learn merely by watching others and without having seen the model being reinforced for their behavior. A group of nursery school children, one at a time, observed an adult model perform a series of aggressive acts on a large rubber inflatable doll (the acts were both physically and verbally aggressive), whereas a second group of children (again one at a time) watched an adult model sit in the room with the doll and perform no action. A third group of children had no model and just viewed a room full of toys that also contained the rubber doll (this is the famous Bobo doll experiment). Afterwards each child was subjected to mild frustration by placing them alone in a room full of toys and first allowing them to play with the toys but after a few minutes prohibiting them to play with them. Then when the children were again allowed to play with the toys their behavior was observed and coded. The behaviors of the children tended to mimic the model to which they were previously exposed. Children observing the aggressive model acted significantly more aggressively towards the dolls than children observing the passive model. Furthermore, children observing the passive model acted even less aggressively than children who observed no model at all. There was also a gender effect such that same gender models and observers demonstrated stronger associations with aggressive actions and male children behaved overall more aggressively than female children. Bandura's conclusion was that the performance or learning of novel responses by a person as a result of solely watching someone else perform them is possible only because of the cognitive abilities people possess (even children). Later, this concept of learning via modeling would be expanded. Bandura found that people can combine and transform what they observe to develop novel or innovative behaviors based on observations. For example, a child may observe that when his older brother gets angry he does not express anger overtly but speaks in a slow-paced and low-toned voice, but when his father gets angry he becomes sarcastic and speaks loud tone. The child may combine these behaviors to produce his/her own form of self-expression.

Bandura maintained that modeling can result in learning through four different routes: 1) modeling teaches the observer new behaviors; 2) modeling can influence the frequency that previously learned behaviors will occur; 3) unfortunately, modeling may encourage previously forbidden or discouraged behaviors such as adolescent smoking or aggression; and 4) modeling can lead to an increase in the frequency of similar behaviors. For instance someone may view a person or friend excel in academics in school and attempt to excel in sports because he does not believe that he can perform as well academically (Bussey & Bandura, 1999). Not only novel behaviors are modeled. Observing a model may result in the extraction of behaviors already in the observer's repertoire, especially for socially acceptable behaviors. The classic examples include learning to stand in the presence of being introduced to a new person or learning to shake hands at an introduction.

Bandura also realized that the effects of modeled behavior can occur when models are not live humans, but instead are depicted symbolically. Bandura, Ross, and Ross (1963b) found that live models, actors in film, and even cartoon characters were equivalent in eliciting modeled behaviors from children. However, just the mere act of observation was not enough to allow the observer to learn and imitate a response. There are several important factors of both the model and the observer that facilitate observational learning.

An important factor in social learning theory is attention; the observer must attend to the model in order to learn. The characteristics of the model are important in affecting the observer's attention. Bandura, Grusec, and Menlove (1967) found that children who observed peers were more likely to model the behavior they observed than when observing adult models. Demographic factors are important such as age, sex, and social status can affect the probability of a behavior being modeled. The more similar the model is to the observer or the more the observer values the characteristics of the model will increase the probability that the observer will try to model behaviors. Also, the behavior stands a better chance of being modeled if it is perceived as being within reach of the observer. Behaviors that are perceived as being too difficult or beyond the standards of the observer have a lower probability of being modeled (Bandura and Whalen, 1966).

Bandura, Ross and Ross (1963b) defined four conditions that are necessary for an individual to successfully model behavior: 1) the person must pay attention to the model (attention will be increased by the aforementioned factors); 2) the observer must be able to remember the behavior he or she has observed; 3) observers must have the ability to replicate the behavior although they need not be perfect at it; and finally 4) motivation is important, learners must want to demonstrate what they have learned. Motivation can be increased by the observer's expectations and as mentioned above, the consequences of the model's behavior will also affect the probability that the behavior will be modeled. Viewing a model reinforced for a behavior will lead to a greater probability that the behavior will be modeled than viewing a model punished for a behavior. However, modeling can occur after viewing behaviors that are neither reinforced or punished (Bandura, Ross, and Ross, 1963b). The characteristics of model and observer can interact to determine whether the observed behavior will be modeled. In the case of children modeling parental behaviors it has been determined that children are more likely to model their parents' behaviors if parents are warm and nurturing or parents who are dominant.

Furthermore, Bandura believed even though children who observe a model being punished are less likely to perform the modeled behavior, they still learned from watching the model. In instances where observers view punishment the act of seeing a model being punished decreased their motivation to model or to perform the observed behavior by decreasing incentives to perform the viewed behavior, but viewing punishment does not affect acquisition or learning. This subtlety has profound implications that suggest that children could learn antisocial behaviors even though they were not given motivation to perform them. If contingencies change at a later time, the learned behavior could be put into action. To empirically test this notion Bandura (1965) had groups of children observe a movie of a model being reinforced, punished, or left without being reinforced or punished after engaging in aggressive actions. Afterwards children in the model-punished condition performed fewer instances of aggression than children in the no consequences condition, who performed fewer instances of aggression than the model-reward condition. This was consistent with previous findings. Then children in all three conditions were told to show or tell what the model did and offered rewards if they modeled the behavior. The previous group differences in the performance of modeled aggression disappeared indicating that all three groups had acquired the behavior but that the presence of viewing the model being reinforced or punished for their actions led to differential probabilities of replication.

Bandura was able to investigate a large number of different behaviors that are modeled. For instance, the Skinnerian view or Pavlovian view of emotional responses is that they are directly learned through personally being reinforced or punished. In the case of classical conditioning the learned responses are reflexive in that subjects are not considered to have much volition in displaying them. But Bandura and Rosenthal (1966) found that classically conditioned emotional responses could be modeled. In a clever study subjects watched models react painfully to a series of buzzers. The models were presented as being involved in an experiment; the subjects believed that they were witnessing an experiment, when in fact the models were confederates. The subjects were informed that the models they viewed were receiving a painful series of shocks when the buzzers sounded (however there were no shocks administered the models were feigning being in pain). Afterwards, subjects displayed a conditioned emotional response measured by their galvanic skin responses (GSR) to the buzzer even in the absence of the models and even though the subjects never experienced a shock.

A practical application of the above study would presuppose that if a response could be learned vicariously then a learned response could be extinguished vicariously. For instance, Bandura, Blanchard, and Blank (1969) incorporated desensitization of fears using models. They assigned a group of subjects with severe snake phobias into several different conditions. In one group subjects were assisted in achieving a deep state of relaxation and then imagined a series of scenes in which they became increasingly involved with a snake such as starting off by looking at a picture, then progressing to holding a toy snake, then progressing to being in a room with a live snake all the way to holding a live snake (this would be the desensitization group). A second group, the symbolic modeling group, involved relaxed subjects viewing films of models performing acts leading increasingly closer contact with a live snake. In the third group, participant modeling, the subjects observed live models interact with snakes at increasingly closer levels of contact. After each level of interaction subjects were requested to perform the action they had viewed, first with the help of the model and then alone. The fourth group was a control group that received no intervention. All groups had completed a pre-test to determine their ability to have physical contact with snakes and measures of their pre-experimental fear levels. Results indicated that the control group exhibited no changes, whereas the desensitization and symbolic modeling groups demonstrated significant changes in their ability to tolerate snakes without phobic responses. The participant modeling group demonstrated the most significant improvement. Following a post-experiment questionnaire in which subjects reported how they would respond to snakes in the future the participant modeling group reported stronger beliefs of future success. Bandura related this finding to a sense of increased self-efficacy of the participant modeling group. He also noted that while fear reduction and vicarious experiences offer some success in extinguishing unwanted behaviors, actual experience involving the feared object or situation is important to developing a sense of personal efficacy. Individuals are more likely to perform new or previously feared actions as a result of learning when they have experienced some actual prior success at performing them. The participant modeling condition allowed subjects to observe behavior, model it directly, and then perform it without assistance allowing them a greater sense of self-efficacy. Bandura's book, Principles of Behavioral Modification (1969) describes many of these techniques.