Although he spent his mature career in educational psychology at Columbia University’s Teachers College in New York City, American psychologist Edward Lee Thorndike’s most important work was done in animal learning, begun at Harvard under William James (18421910). Thorndike, along with Russian physiologist Ivan Pavlov (1849-1936), provided the methodological tools of behaviorism, in which psychologists used animal models to formulate theories of learning and behavior that would include humans.
When Thorndike and Pavlov began their research, comparative psychology used the so-called anecdotal method—collecting stories about animal behavior in natural and seminatural settings—in order to understand conscious animal thinking. Thorndike challenged the anecdotal method for lack of control, overestimation of animal intelligence, and tendencies to anthropomorphize the animal mind. He substituted experiments for anecdotes, establishing one of the two major paradigms for studying learning: instrumental, or operant, conditioning. Contemporaneously, Pavlov established Pavlovian, classical or respondent, conditioning. With regard to animal thought, Thorndike set out to catch the animal mind at work but concluded that animals do not reason their way to problem solutions. Rather, they engage in mindless trial-and-error learning.
In the experiments that defined instrumental conditioning, Thorndike placed young cats inside wooden cages called puzzle boxes from which they could escape by working a manipulandum inside the box, such as a foot treadle. Thorndike observed that cats tried out a variety of instinctive responses before accidentally hitting on the correct response. Nor did they show insight. Instead of stepping on the treadle immediately on the next trial, cats repeated erroneous responses, although the correct response emerged sooner as trials progressed until it became dominant. Thorndike described the process of learning as a gradual "stamping out" of connections between the box stimuli (S) and the incorrect instinctive responses (R), and the gradual "stamping in" of connections between S and the correct R. Hence, Thorndike called his theory of learning connectionism, a behavioral version of associationism.
Thorndike proposed three laws governing learning. The law of exercise held that using a connection strengthened it and disuse weakened it, while the law of readiness stated that when a connection was available, its use would be satisfying to the organism; these laws were later abandoned. Most important was the law of effect. Initially, the law of effect held that when a response to a stimulus led to pleasure, the S-R connection was strengthened, and when a response led to painful punishment, the connection was weakened. Thorndike later revised the law of effect, having found that punishment did not weaken S-R connections, but inhibited their expression, a view held today.
Because Thorndike never proposed a comprehensive system of psychology, his ideas were subjected to detailed rather than systematic criticism. Most significant were critiques of the law of effect and the methodology of the puzzle boxes. Eager to purge references to mental states from psychology, behaviorists objected to Thorndike’s reference to "pleasure"—a subjective conscious feeling—as the cause of learning. They substituted less mentalistic causes such as contiguity of stimulus and response (Edwin R. Guthrie [1886-1959]) or biological drive reduction (Clark Hull [1884-1952]), or they defined reinforcers functionally as events that strengthen the responses that produced them (B. F. Skinner [1904-1990]). Later, as information-processing views of learning gained strength, psychologists questioned Thorndike’s law of effect in a new way (anticipated by Edward C. Tolman [1886-1959] in the 1930s). Thorndike assumed that rewards and punishments work via pleasure and pain, but they also provide information (a concept not available to Thorndike) that a response was correct or incorrect. Experiments that separate the two (e.g., making a painful stimulus indicate that a response was correct) have shown that learning depends on the information value of reinforcers more than their subjective quality.
The Gestalt psychologist Wolfgang Kohler (18871967) offered an important criticism of Thorndike’s puzzle-box method. One of the driving issues in psychology of learning is whether learning occurs gradually or can occur suddenly via insight. Thorndike found no signs of insight in his puzzle-box studies, while Kohler found evidence of insight in his studies of problem solving by chimpanzees. Kohler argued that Thorndike’s method was faulty because it made insight impossible: Trapped in the puzzle box, the cat could not see the connection between the manipulandum and the door opening, and so was forced to resort to trial and error. In Kohler’s experiments, on the other hand, all the elements needed to solve a problem were available to the subject, who was able to assemble them insightfully into a solution. The force of Kohler’s critique extends beyond issues of learning. Psychologists perform experiments in order to discover laws explaining behavior in real life, but experiments are necessarily artificial, and may lead psychologists to propose universal laws of behavior that are in fact laws induced by their experiments.
Nevertheless, Thorndike’s influence was enormous. He initiated the S-R concept of learning elaborated by Clark Hull and his followers from the 1930s to the 1960s,which overshadowed the cognitive tradition of the Gestalt psychologists and Edward Tolman, which held that learning consisted of developing internal representations of the world, the main view in cognitive science today. The law of effect provided the basis for Skinner’s principles of reinforcement, though Skinner did not view operant learning as making connections. There is today a new "connection-ist" (neural network) movement, but it is not linkable to Thorndike.