Tolmans Expectancy Theory

Edward C. Tolman (1934) adhered to many of the fundamental tenets of behaviorism but also introduced several new perspectives into the study of behavior and learning— some of which were highly controversial and inconsistent with the behaviorist platform. Tolman viewed the study of behavior both as an experimental process (fact finding, hypothesizing, and falsifying) but also emphasized an interpretative component that evaluated the meaning or purposiveness of the behavior being studied. Most behaviorists before him viewed behavior as a molecular phenomenon composed of individual S-R effects and relationships. Tolman believed that behavior had to be investigated in the context of the subject's intended purpose, thus extending the study of behavior to include an evaluation of its purpose, that is, its molar implications.

Tolman's scientific thrust aimed at developing hypothetical constructs inferred from concrete experimental observations of behavior. The study of purposiveness does not imply observations based on empathy or introspection (methods that Tolman rejected) but rather the formulation of inferences derived from observed behavior. In the scientific study of behavior, three experimental variables co-interact to arrive at significance:

1. Independent variables: The various controlled aspects of the experiment, especially the stimulus conditions and motivational state of the animal.

2. Dependent variables: All measured changes occurring in the behavior of the subject under the influence of controlled experimental conditions.

3. Intervening variables: Abstract constructs necessary to explain the observed S-R relationship.

The intervening variable is not a subjective interpretation but an objectively defined presumption arrived at by holding constant all independent variables except those hypothesized significant to it. The intervening variable is inferred from experimental evidence—that is, it helps make sense of experimental results. The validity and usefulness of the intervening variable is established by making predictions based on it and then designing experiments to systematically falsify those predictions. The intervening variable is operationally defined and delimited by the results of such experimental analysis and falsification.

For example, if a dog is presented with two bowls of food, one with meat in it and the other with dry food, the dog will most likely choose the one containing meat. A reasonable conclusion that one might draw from this experiment is that the dog "prefers" meat over dry food. Although this is a possible conclusion, however, it is not the only one possible from this experiment. To demonstrate preference some quantifiable correlation needs to be elaborated, defining preference (itself unquantifiable) as the most relevant variable controlling the dog's choice of meat over dry food. A hypothetical experiment might involve making the dog expend physical energy (jumping over a barrier of increasing difficulty) or mental effort (solving a difficult puzzle or maze) to acquire the meat as a goal and then comparing the dog's effort with respect to other food items. The assumption here is that the dog's preference for the food item is positively correlated with a willingness to work harder for it. Further more, his preference can be quantified relative to other less preferred items of food.

A slightly more complicated situation occurs in a two-choice discrimination task. In this experiment, the dog is trained to choose between two cards, one patterned with a checkerboard pattern and the other left blank. Choosing the checkerboard pattern always results in the presentation of food, whereas the blank card is never reinforced. Within several trials, the dog learns to choose the patterned card when prompted to choose. In this case, many possible intervening variables may subsist between the presentation of the positive and negative cards and the pattern of subsequent choice making. One very general hypothetical construct is that the dog "thinks" about the choice options and then chooses according to cognitive rules of discrimination; another broad view might theorize that the dog "learns" to choose the correct card as the result of trial and error; another observer might claim that the dog is innately attracted to patterned objects and is more likely to attend to the checkerboard-pattern card over the blank card; another possible theory is that the positive choice is an outcome of the nonreinforcement of the blank card (extinction) rather than a result of reinforcement of the positive card; and another theorist might explain the dog's mastery as an outcome of classical condition-ing—that is, the dog is attracted through associative learning to the positive card.

As the foregoing inventory of possible intervening variables shows, there are many possible ways to explain the dog's successful discrimination. To determine how the dog manages to learn such a discrimination task requires experiments that isolate one intervening variable at a time while controlling the effects of others. The theory that the dog is innately attracted to the positive card can be easily falsified by presenting the blank card as the positive stimulus and comparing relative rates of learning with the checkered card. But what about the relative importance of trial and error versus extinction-based learning, and the role of classical conditioning? What are the most important variables influ encing discrimination learning? Answering such questions as these would require the design of several controlled experiments isolating significant from confounding variables.

Tolman placed a stronger emphasis on stimulus or sign learning than he did on response habit formation (i.e., Thorndike's stamping-in or stamping-out process). Instead of learning a response pattern, Tolman argued that an animal learns a cognitive map of significant relations or sign-gestalts (signs, significates, and behavior routes leading from sign to significate) in the environment, leading to the satisfaction of appetitive demands and goals. In a general sense, signs correspond to the classical conception of the conditioned stimulus and significates to the unconditioned stimulus:

The sign-gestalt theory asserts that the conditioning of a reflex is the formation of a new sign-gestalt. It asserts that a conditioned reflex, when learned, is an acquired expectation-set on the part of the animal that the feature of the field corresponding to the conditioned stimulus will lead, if the animal but waits [behavior route], to the feature of the field corresponding to the unconditioned stimulus. (Tolman, 1934:393)

Hilgard and Bower (1975) described several experiments that tend to support Tol-man's cognitive interpretation of learning. One of these experiments (Tinklepaugh, 1928) involved a delayed-response test in which a hungry monkey was shown a banana that was then hidden under one of two cans. The monkey quickly mastered this discrimination and easily found the concealed banana. Later, while the monkey was out of sight, the banana was secretly removed and replaced with a leaf of lettuce. When the monkey returned and discovered the change, he rejected the lettuce (a less preferred food item) and began searching for the hidden banana. This study implies that the animal had formed a definite expectation about finding a banana. Besides forming expectations about outcomes, animals learn from signs and place cues how to reach specific goals—that is, the animal is not learning a specific series of responses but exhibits behavior that implies that he knows where the goal is located and uses various signs and routes to get there, which is a "what leads to what" theory of learning. Another study [Macfarlane (1930), reported by Hilgard and Bower (1975)] provided additional support for a connection between cognitive mapping and goal-directed behavior. Macfarlane first trained rats to wade through a flooded maze and then required that they swim the course instead. The swimmers were found to do equally well as the waders, indicating that the response sequence was not dependent on learning a set of specific motoric or kinetic actions but depended on a more general knowledge of place.

Thorndike (1946) proposed an experiment to test the role of learned expectancy versus habit formation and response reinforcement in instrumental learning. The experiment involved placing a rat on a cart and pulling it through a maze. After a number of such trials, the rat would than be tested for its ability to learn the maze route and the results compared with that of a naive rat not previously exposed. Thorndike predicted that both subjects would learn the task equally well. However, subsequent studies have contradicted Thorndike's prediction (Mazur, 1986). Experiments by Dodwell and Bessant (1960) found that such preexposure did affect learning rates in a positive direction. In their experiment, animals were pulled through a water maze in a little car. Subsequent tests demonstrated that the preexposed rats performed better than controls not exposed. Earlier studies demonstrated that rats that underwent pretraining exposure, by being permitted to explore the maze prior to training, did substantially better than controls not given such exposure. This latter evidence is somewhat confounded, however, since the benefit of such pretraining exposure may have been due to adaptation to the training environment rather than due to learning. On the whole, these results contradict Thorndike's view that instrumental learning is solely dependent on response-contingent connections.

Tolman's learning theory makes several theoretical distinctions between learning and performance. For Tolman, learning is independent of performance, but performance is not independent of learning. Motivational levels strongly impact performance by generating goal-directed tensions demanding satisfaction. In an important sense, performance is a composite of current motivational states and past learning experiences. Even though learning is not dependent on motivation, as seen in the aforementioned case of latent learning, it is not entirely independent of it either. Motivational substrates (appetite, fear, and aversion) define the specific details of the environment that an animal most alertly and selectively attends to. An animal pays greatest attention to and learns the most from items that possess significant motivational interest. Hungry dogs seek out signs of food, whereas fearful dogs search for routes of escape. The cumulative organization of all available signs and routes together with their corresponding significates forms a cognitive map of signgestalts representing the overall field of available expectancies (Tolman, 1948).

Perhaps Tolman's most significant contribution to the study of animal behavior is his emphasis on the cognitive aspects of learning. From this perspective, learning represents much more than the acquisition of a series of simple S-R outcomes or response-reward relationships. Tolman's view places learning within a much broader context or field. Learning takes place on an integrative, molar level where S-R events are interpreted and made meaningful by assimilation of the particular into the general, a mediation effected by "sign-gestalt expectations." Through learning, dogs are ever-forming predictive interpretations and expectancies about the occurrence of important stimulus events—a process that is both purposive and cognitive.

Dog Training Basics

Dog Training Basics

Are you looking for the quick and easy ways to train your dog to follow whatever you asked him/her to? or maybe you are just sick and tired of your dog behavioral problems and misbehaved without listen to what you are commanding them to follow, then this will be the most important letter you'll ever read for today! If so, are you dreaming of owning a dog that's well behaved, obedient and protective?

Get My Free Ebook


Post a comment