Hull, Clark L.
Hull, Clark L.
Hull, Clark L.
Clark L. Hull (1884-1952) was the most influential figure in the experimental psychology of learning during the decades immediately preceding and following World War II [seeLearning; Learning Theory]. This was not because his researches were unusually definitive or his theory universally accepted. If there was a single cause, it was that Hull presented his theoretical ideas with a degree of rigor and analytical detail then unfamiliar in psychology. A substantial proportion of the research in learning in this era was formulated in terms of Hull’s theory, whether designed to sustain or to disprove it, and virtually all such research at least acknowledged a relationship to Hullian conceptions. Although Hull’s theory contained some ambiguities and inconsistencies, it set a challenging level of explicitness that contributed in an important way to the developing image of psychology as a science.
Hull’s ideas drew on many sources. Darwin had proposed the principle of natural selection that formed a rationale for much of Hull’s theory. Many of the phenomena of conditioning that Hull described and some of the theoretical constructs that he espoused had been discussed by Pavlov. Freud had asserted the central role of motivation in behavior, a position Hull increasingly adopted. Watson had established the behavioristic departure from earlier introspectionism; Hull followed in the less radical, neobehavioristic tradition. Thorndike’s law of effect, asserting the important role of rein forcement in learning, was central to Hull’s major theorizing. Tolman had described the place of the intervening variable in behavior theory and had drawn the critical distinction between learning and performance, which Hull later formalized. Not only was Hull indebted to these predecessors, but he was unusually careful also to acknowledge the impor tant contributions of his colleagues, notably Neal E. Miller and Kenneth W. Spence, in developing his approach.
To this potpourri of influences, Hull added inte gration and purpose. He made more explicit the ideas of these and other men and in so doing added a wealth of his own conceptual innovations. He believed that controlled studies of simple learning situations would reveal a core set of principles of behavior and that these could then be applied to the more complex conditions of individual and group behavior. If successful, this program, he felt, would have inestimable value for understanding, predicting, and, where appropriate, controlling behavior.
Hull was born in a log cabin on a farm in New York State in 1884. His education, which began in a rural school, was interrupted by a job as a mining engineer and an attack of poliomyelitis that made the use of a cane necessary thereafter. He was 34 when he received a ph.d. from the University of Wisconsin. Early in life, Hull began a series of personal notebooks in which he recorded a variety of facts, ideas, and reflections, hoping to preserve the thoughts of his youth for later, less creative years. These “idea books” show an early sense of destiny and a developing obsession with urgency. They also contain many detailed analyses of his theory-in-the-making which illustrate his conviction that the explicit, formal statement of an idea is most likely to reveal its weaknesses. As he said in Principles of Behavior, “It is believed that a clear formulation, even if later found incorrect, will ultimately lead more quickly and easily to a correct formulation than will a pussyfooting statement which might be more difficult to convict of falsity” (1943, p. 398). Hull was uncompromising in his expectation of speed, effort, and punctuality from his associates and intolerant of slipshod thinking by his students. He was generous both with per sonal attention and professional credit to all who worked with him. He consciously avoided vindictive arguments, to which some of the reviews of his work could easily have led, but accepted, indeed welcomed, considered criticism.
Hull’s early scientific activities are noteworthy in themselves. Continuing at the University of Wisconsin after obtaining his degree, Hull worked in the area of the measurement and prediction of achievement. His first major work, Aptitude Test ing (1928), reveals his principal concern—the problems of validation of tests. To reduce computational time and errors, Hull designed an auto matic correlation machine, a beautifully complex unit that, as he said in his autobiography, “really worked,” and that illustrates his unusual talent for gadgetry. He then became interested in hypnosis and conducted a series of experimental studies, published in his second major work, Hypnosis and Suggestibility (1933), that brought scientific rigor to this often quasi-scientific area. Hull moved to the Institute of Human Relations at Yale University in 1929 and there, at the age of 45, started to formalize a mechanistic theory of behavior. He was to make his major contribution to psychology in this area. [SeeAchievement Testing; Aptitude Testing; Hypnosis; Suggestion.]
The institute housed a vigorous group of workers of varying ages, fields, and persuasions, all dedi cated to the concept of a science of behavior.
Nevertheless, it was 11 years before Hull and several colleagues formalized a set of learning prin ciples in Mathematico-Deductive Theory of Rote Learning (1940), a work less well known than it deserves to be, probably because of its use of the language of formal logic and mathematics. Three years later a more general and readable statement of the fundamental principles appeared in Hull’s third and most influential major work, Principles of Behavior (1943). Another nine years passed be fore Hull’s A Behavior System (1952a) described selected examples of more complex individual be havior in terms of his theoretical principles. This last work was published posthumously; the intend ed application to social and cultural problems remained to be written at the time of Hull’s death in 1952.
Reinforcement theory of learning. Hull’s theory can be characterized as a mechanistic stimulus-response conception of behavior. Learning, Hull believed, consists in the bonding together of some stimulus event (a light, a tone, a maze) with some response of the organism (salivating, running, turning). Behavior in general is describable in terms of hierarchies and constellations of such connections. Hull did not, however, deny the reality of ideas, knowledge, and purpose; indeed, he accepted these as obvious descriptive properties of molar be havior. He argued, rather, that these are not the basic units of behavior and that they are understandable in terms of more fundamental stimulus-response events. He introduced, therefore, the notion of the fractional, anticipatory goal response, which assumes that a portion of the reinforcing event at the end of a behavior chain can have an effect at the beginning of the chain. This response, or more properly, its response-produced stimulus, serves to guide overt behavior, as if purposefully, toward the goal. Whatever the complexity of be havior apparent to direct inspection, it could be reduced to simpler principles by the further dis covery of response mechanisms, often hypothetical internal responses with no other function than to provide the organism with guiding stimuli.
Principles of Behavior provides the epitome of a reinforcement theory of learning: the formation of stimulus-response connections is assumed to depend on a subsequent reinforcing state of affairs, and the degree to which the connection is strength ened depends upon the quantitative properties of the reinforcement. That is to say, learning occurs only if the response is rewarded, and the larger and more immediate the reward, the better the learning. Recognizing the possible circularity of this theory in the absence of an independent definition of reinforcement, Hull proposed the drive reduction hypothesis [seeDrives]. He assumed that a hungry rat learns to run through a maze because the turning responses at the choice-points are reinforced by the eventual reduction in hunger drive oc casioned by eating in the goal box. Furthermore, since choice-points near the goal are followed most quickly by this reinforcement, Hull correctly in ferred that errors would be eliminated in backward fashion, the appropriate responses at the beginning of the maze being learned last. These two explicit postulates—that learning occurs only if responses are reinforced and that reinforcement occurs only if a drive is reduced—generated a substantial proportion of the research of Hull and his colleagues.
Habit and drive. The stimulus-response connection, which Hull called habit strength and sym bolized .SH", is itself not the sole determinant of be havior. In order to appear in overt performance, habits must be energized by drive. This latter construct, symbolized D, is related to presumed bio logical and survival needs of the organism and species (hunger, thirst, pain, sex) and is assumed to combine multiplicatively with habit to determine excitatory potential (SER), which in turn deter mines performance. This equation—SE — SHR x D —formalizes the distinction between learning and performance and inextricably entangles learning with motivation. When elaborated and expand ed, the formulation became a powerful source of hypotheses.
The proper use of this equation requires a so phisticated understanding of both habit and drive. Consider first HR The actual stimulus to which a response is attached is not the distal event as it occurs in the environment or even the proximal event as it occurs on the receptor. Rather, it is a trace of this latter event, which, according to Hull, decays over time. Moreover, because stimuli are never exactly identical, Hull further hypothesized that habit learned to one stimulus generalizes some habit strength to other similar stimuli. The postu lates of the stimulus trace and habit generalization illustrate the way Hull derived behavioral phenomena.
Pavlov had shown that if the sound of a bell regularly occurs, say, ten seconds before food is delivered to a hungry dog, the salivation that had occurred only after the food was in the dog’s mouth becomes anticipatory, i.e., begins to occur when the bell sounds and before the food arrives. Hull’s theoretical interpretation is that the bell initiates a stimulus trace in the nervous system of the dog which immediately begins to decay, but which is present in some unique intensity ten seconds later when the food is delivered. Since the food elicits salivation as an unlearned reflex, and since it also provides reinforcement by reducing hunger, the appropriate conditions are present for the formation of a habit connection between the ten-second trace of the bell and salivation. Now, because earlier traces of a bell are similar to the ten-second trace of that same bell, the habit being formed to the latter will generalize to the former. Hence, the dog salivates anticipatorily as a result of generalization. By making explicit quantitative assumptions about the shape of the stimulus trace function and the degree of generalization between different traces, one can make detailed derivations about the expected latency of salivation at different stages of training and the importance of the time interval separating the ringing of the bell and the presentation of the food.
Hull also expanded the construct of drive. Perhaps his most important elaboration is the concept of the association of any particular intensity of any particular drive (e.g., a rat deprived of food for 24 hours) and a characteristic drive stimulus which differs to some extent from that produced by other drives (e.g., deprivation of water) or other inten sities of the same drive (e.g., 12 hours of food deprivation). Hence, although Hull’s conception of drive is that of a general energizer—hunger can as well motivate a response learned under thirst— drives do have a directing function, because each drive stimulus will tend to become selectively at tached to those responses most likely to reduce it. Thus, a rat in a maze can learn to turn right when hungry and left when thirsty even though, as mo tivators, these drives act indiscriminately on all habits.
Given the fact that there is experimental extinction—the weakening and disappearance of responses when reinforcement is terminated—Hull was faced with two theoretical alternatives. One was the assumption that SHR itself is weakened by nonreinforcement. However, there were a variety of reasons for believing that learning is a relatively permanent process, not the least of which is the dramatic evidence of memory Hull had obtained using hypnosis. Hence, Hull followed Pavlov in as suming that experimental extinction does not rep resent the loss of habit, but rather the accumulation of inhibition. Excitatory potential is reduced by inhibition resulting from nonreinforcement; the dog does not forget how to salivate to the bell; instead, learning is inhibited if food no longer follows the bell.
The system was completed by relating excitatory potential to performance. One common response measure is the probability that a response will occur on any presentation of a stimulus. To adapt this measure to his theory, Hull borrowed the notion of a threshold from physiology and assumed that excitatory potential is subject to oscillation from moment to moment according to an essentially ran dom distribution. A response will occur only if SER is above threshold at the moment and, if this is so, the speed and amplitude of response will depend directly upon the momentary value of excitatory potential.
One of the persistent problems that Hull faced in developing his theory was that the intervening variables such as habit and drive are inherently unobservable by direct inspection and hence cannot be measured. Since the theory is cast in quanti tative terms, Hull’s derivations typically consist of illustrative examples, with arbitrary values for the important variables. He foresaw the importance of quantifying these variables in units that might be applied to the wide range of situations to which the theory was intended to apply. Toward this end, Hull undertook a meticulous series of studies employing the Thurstone scaling technique to quantify habit and drive. Hull valued this work greatly, in keeping with his belief that mathematics is a fundamental tool of behavioral as well as physical science.
Stimulus dynamism and incentive motivation. Almost as soon as Principles of Behavior was published, major modifications of its theory began. One unsolved problem was the role of stimulus inten sity; Hull became convinced that not only do drive stimuli have energizing and directing functions, but that all stimuli have dynamogenic as well as cue properties. Furthermore, the rapid changes in an animal’s response after a change in the amount of reward suggested that reinforcement is a mo tivational variable as well as a learning variable. A new equation for excitatory potential therefore appeared. It included the concepts of stimulus dynamism based on the intensity of the eliciting stimulus, and of incentive motivation based on the amount and time of delay of reward as additional motivational factors multiplying habit. According to this new conception, a dog salivates more to a strong than a weak bell not because he has learned to salivate better, but because the stronger stimulus energizes a more vigorous response. A rat runs a maze faster for a larger reward than for a smaller not because he has learned the maze better, but because the larger reward provides more incentive to run. Motivational factors thus began to dominate Hull’s learning theory, although he clung to the assumption that some reinforcement is required for the accretion of habit strength.
Discrimination learning . All these hypotheses were generated on the basis of data obtained in relatively simple learning situations. The logic of Hull’s program, it will be recalled, is that these principles will be applicable, when properly interpreted and possibly expanded, to more complex situations as well.
In A Behavior System, Hull described the following case, among others: If an organism is confronted with two stimuli differing, let us say, in brightness, and is consistently rewarded if he chooses one of these, he will learn to behave adap-tively; he will learn to discriminate between the stimuli and consistently to choose the rewarded alternative. In Hullian language, choice between responses is somehow based on excitatory potential. The simplest hypothesis is that if several responses are in competition the one that has the largest excitatory potential will occur. With the choice measure specified in this way, discrimination learning follows readily from the theory.
It is first noted that the excitatory potential of the positive stimulus will be increased by the reinforcement following responses to it. Although generalization will produce some excitatory potential for the nonreinforced stimulus, this will be reduced by the inhibition developed whenever response is made to it. Thus, continued training will lead to a gradual separation of the excitatory poten tials of the two stimuli and hence a developing preference for the reinforced one. A number of additional and more detailed implications also follow. For example, the discrimination should be easier the more different the two stimuli are, since there will be less generalization between them and more rapid separation of their excitatory potentials. As is true of most useful theories, this analysis has more implications than appear at first glance, many of which have been derived by Spence (1960) from his related theory and tested with tolerably good results.
Assessment of Hull’s theory. Hull correctly fore saw that the details of his theory were tentative and likely to require revision as relevant data were collected. His principal lasting contribution lies more in his extensive illustration of an approach to behavior theory than in the specific theory itself. This approach leans heavily upon the availability of empirical data, since it involves the postulation of intervening variables or empirical constructs which are intended to summarize the existing data and to generalize upon them for predicting new data. For example, it is known that rats run a maze to water faster the longer they have been previ ously deprived of water. The intervening variable approach assumes that deprivation of water affects the organism by producing thirst that potentiates the response. It is important to note the two essen tial criteria of an adequately formulated intervening variable: it must be anchored to some inde pendent variable such as deprivation, and it must also be anchored to some dependent variable such as response speed.
Hull believed that intervening variables actually exist in the neurophysiological structure of the organism, and many of his postulates are couched in physiological-sounding terms. Some critics may therefore incorrectly assume that the usefulness of the theory is dependent upon the adequacy of such physiological speculations. But Hull intended only to stress the ultimate reducibility of molar behavioral events to more predictable molecular physiological events and to suggest possible rather than specific avenues toward such reduction. For example, he felt that his concept of drive might be related to endocrine and chemical mechanisms, and if this were true, an important link would be established between psychology and physiology. But such a result would be a bonus; the value of the approach is assessed by its ability to predict behav ioral facts and not by its physiological validity.
The construct “thirst” integrates a wider set of findings than the single one that longer water deprivation leads to faster running. The same in dependent variable (deprivation of water) also affects behavior in a number of other ways—the amount the rat will drink, the vigor with which he will operate a manipulandum, the intensity of elec tric shock he will tolerate in order to obtain water, etc. Furthermore, a number of operations other than the withholding of water can be found to produce the same pattern of behavioral consequences (responses)—feeding dry or salty food, injecting drugs into the rat’s brain, etc. All of these empirical relationships are tied together by the single notion of thirst. A quite similar constellation of findings appears with deprivation of food, and the construct “drive” integrates at a more abstract level the energizing and directing functions common to hunger, thirst, and other needs.
The postulation of the combination rule that drive multiplies habit permits novel predictions. For example, since the choice between two alter natives depends upon the difference between their excitatory potentials, rats will learn a maze faster under high than under low drive. This is because the developing differential in habit, resulting from the reinforcement of only one alternative, will be amplified in proportion to the prevailing level of drive. When such implications are tested and the results fail to conform to prediction, the theory necessarily is modified.
The extension of the simple principles to more complex situations follows a similar progression. For example, when it is observed that a previously neutral stimulus paired temporally with a noxious stimulus takes on some of the properties ascribed to drive, the richness of the theory can be drawn upon to develop the concept of learned drive. The principles of learning should describe the way drives are acquired, and the principles of perform ance should describe the way such acquired drives affect behavior. Dollard and Miller (1950) have carefully extended the application of the basic prin ciples along these lines, and they have applied the resulting theory to psychotherapy.
Hull assumed further that these same principles could be applied to still more molar social and economic situations. This might require some additional principles not apparent in the simple situ ations from which the theory sprang. A boy and a girl on a date are reacting in accordance with their drives and habits just as much as a rat in a maze; the drives include a variety of social pressures toward conformity, approval, and morality, and the habits involve fine discriminations of the cues provided by the other person. Hull believed that the reduction of such situations to his theoretical conception was in principle possible, and certainly could not be proved impossible a priori.
Hull’s program, even as far as he carried it, was certainly not flawless, and it led psychologists to overestimate the imminent attainment of a general behavior theory. Hull tended to oversell the program, to offer more than could be delivered, to be willing to guess at systematic behavioral facts in order to construct the framework of a theory. Partly as a reaction to this, subsequent workers have tended to have more modest goals, to build mini ature models, or to delve in detail into circum scribed aspects of the larger problem. When and where behavior theory on the grand scale will again be promoted cannot be predicted. Hull, for one, was confident that it could be achieved, that it would be achieved, and that his work would provide a useful stepping stone to the goal. His last book concludes with this hope:
And finally, the crowning achievement of all will be the creation of a really quantitative system of social behavior. … It seems incredible that nature would create one set of primary sensory-motor laws for the mediation of individual behavior and another set for the mediation of group behavior. Presumably, then, the laws which are derived for social behavior will be based for the most part on the same postulates as those which form the basis of individual behavior. If this turns out to be true, we are even now an appreciable distance on our way toward the ultimate goal of inte grating the individual-social sciences with the group-social sciences. (1952a, pp. 355-356)
Frank A. Logan
1928 Aptitude Testing. New York: World.
1933 Hypnosis and Suggestibility: An Experimental Approach. New York: Appleton.
1940 Mathematico-Deductive Theory of Rote Learning: A Study in Scientific Methodology, by Clark L. Hull et al. New Haven: Yale Univ. Press; Oxford Univ. Press.
1943 Principles of Behavior: An Introduction to Behavior Theory. New York: Appleton.
1952a A Behavior System: An Introduction to Behavior Theory Concerning the Individual Organism. New Haven: Yale Univ. Press.
1952b Autobiography. Volume 4, pages 143-162 in A History of Psychology in Autobiography. Worcester, Mass.: Clark Univ. Press.
1962 Psychology of the Scientist: 4. Passages From the “Idea Books” of Clark L. Hull. Perceptual and Motor Skills 15:807-882. → Published posthumously.
Dollard, John and Miller, neal E. 1950 Personality and Psychotherapy: An Analysis in Terms of Learning, Thinking and Culture. New York: McGraw-Hill.
Hilgard, Ernest R. (1948)1956 Theories of Learning. 2d ed. New York: Appleton.
Hovland, Carl I. 1952 Clark Leonard Hull: 1884-1952. Psychological Review 59:347-350.
Koch, Sigmund 1954 Clark L. Hull. Pages 1-176 in Modern Learning Theory: A Critical Analysis of Five Examples, by William K. Estes et al. New York: Ap pleton.
Logan, Frank 1959 The Hull-Spence Approach. Vol ume 2, pages 293-358 in Sigmund Koch (editor), Psychology: A Study of a Science. New York: McGraw-Hill.
Spence, Kenneth W. 1951 Theoretical Interpretations of Learning. Pages 690-729 in Stanley S. Stevens (editor), Handbook of Experimental Psychology. New York: Wiley.
Spence, Kenneth W. 1960 Behavior Theory and Learn ing: Selected Papers. Englewood Cliffs, N.J.: Prentice-Hall.