Spence's Theory Research Articles

Evolution of an elemental theory of Pavlovian conditioning

Kenneth Spence (1936, 1937) formalized a quantitative, elemental approach to association theory that has had a broad and dominating influence on learning theory for many years. A set of challenges to the basic approach has spurred the subsequent evolution of elemental theory in various ways. Four of the challenges and some resulting theoretical accommodations are described in the context of Pavlovian conditioning. The evolution involves departures from important specifics of Spence's theory, but is viewed as demonstrating the utility of the basic, elemental approach that is one of his legacies.

Reinforcers following greater effort are preferred: A within-trial contrast effect.

Psychologists who study human behavior often posit underlying mechanisms that are difficult to measure directly. And because humans are known to be complex organisms with complex motivational systems, researchers studying human behavior may avoid asking if there are simpler accounts of the behavior being studied. One such example is the phenomenon known as transposition. This phenomenon is familiar to musicians who after playing a tune in one key can seamlessly transpose the tune to a different key with no apparent change in the melody (the relations among the notes). A similar effect can be shown in rats, when after training them on a simultaneous light-gray (S+) versus dark-gray (S-) discrimination with responding to the light-gray reinforced, rats are tested with the light-gray stimulus and a still lighter-gray stimulus (Sn). In spite of the fact that responding to the S+ had a history of reinforcement and responding to the Sn had not, rats typically choose the Sn over the S+. That is, it appears that they learned to choose the lighter one during training and they continued to choose the lighter one on test trials. That is, it appears that they had learned the relation between the two training stimuli rather than their absolute properties. But Spence (1937) proposed that when organisms learn to respond to a stimulus for reinforcement, the tendency to respond generalizes to other similar stimuli (defining a gradient of stimulus generalization). Similarly, responding to the S- results in a gradient of inhibition around the value of the S- stimulus. According to Spence, it is the algebraic summation of those gradients at each point along the dimension that determines the strength of responding (the net response strength) at other points on the dimension (brightness in the example described). If one assumes that the gradients are convex (i.e., flatter near the training values) then the net gradient typically has its peak value not at the S+ but at a point away from the S+ in the direction away from the S-. This occurs because, given the presumed shape of the gradients, there would be a smaller decline in response strength as one begins to move away from the S+ than there would be a decline in inhibition at that point in the inhibitory gradient (see Spence, 1937). In this way, paradoxically, Spence's theory of learning about the absolute properties of a stimulus can account for an animal's choice of a novel stimulus, never seen before, over the S+ training value, and hence it can account for transposition without appealing to relational learning. Cognitive Dissonance An even more striking example of a presumably complex human behavior that is likely to be influenced by simpler behavioral processes is the supposed human motivation to reduce cognitive dissonance. Cognitive dissonance occurs when there is a discrepancy between one's beliefs and one's behavior. This can occur when one acts in a way that is inconsistent with the way one believes one should act. According to cognitive dissonance theory (Festinger, 1957), one should work to reduce the dissonance and if it is too late to avoid the behavior, one would modify one's beliefs to account for or justify one's behavior. Festinger and Carlsmith (1957) tested this theory by giving subjects a task considered to be boring and then asking them to tell a prospective subject that the task was interesting. In payment for this service, the experimenter would compensate them with a payment (the experimental manipulation) of either $1 or $20. But before carried out their assignment they were asked to fill out a questionnaire that included judgments about the task. Surprisingly, those subjects who were paid $1 judged the task to be more interesting than those who were paid $20. Festinger and Carlsmith explained the difference in terms of an attempt to reduce cognitive dissonance for the subjects paid only $1. For those subjects, the dissonance produced by having engaged in a boring task but having agreed to tell a potential subject that the task was interesting could not be resolved by appealing to the payment of $1. …

The Effects of Anxiety and Intelligence on Learning from Programmed Instruction

An attempt is made to analyse and relate the findings concerned with individual differences in learning achievement from programmed learning materials, to those from learning studies generally. In particular intelligence, anxiety and task difficulty are considered in terms of an extension of Spence's theory of anxiety and performance. The subsequent study carried out revealed that significant effects due to anxiety and intelligence were only found on a moderate as opposed to difficult programme. In addition the use of a state anxiety measure revealed in interaction between anxiety resulting from the learning milieu and intelligence on the moderately difficult programme: among high‐intelligence Ss those with high state anxiety gained more; whereas, among low‐intelligence Ss those with high state anxiety gained less.

Discrimination learning and stimulus generalization by impulsive and reflective children

The proposition that impulsive children inhibit expressive behavior less than reflective children was investigated by providing third grade impulsive and reflective boys with either differential or nondifferential forms or successive discrimination training prior to tests for stimulus generalization. Consistent with specific predictions derived from Spence's theory on the roles of excitation and inhibition, it was found that: (a) Reflectives acquired the discrimination more rapidly, (b) reflectives achieved a higher level of discriminative performance, (c) impulsives exhibited more variable response reates during the nonreinforced stimulus, and (d) the type of training interacted with the child's conceptual tempo to determine the elevation and shape of the post-discrimination generalization gradients. The results provide support for the construct validity of the Matching Familiar Figures Test of reflectivity-impulsivity.

Performance as a function of resultant achievement motivation (perceived ability) and perceived difficulty

Subjects worked on a task which was described as either easy or difficult. When the task was thought to be difficult, Ss high in resultant achievement motivation performed better than those low in resultant achievement motivation. However, when the task was perceived as easy, the high motive group performed worse than the low group. These results confirm a prediction from Kukla's attributional theory of performance, in which resultant achievement motivation is conceived as a measure of perceived ability. They are not, however, deriveable from Spence's theory of the effects on performance of objective task difficulty, nor from Weiner's hypothesis concerning the motivational effects of success and failure. On the other hand, Kukla's theory provides an explanation for both the data usually cited in support of Spence's position and those taken to confirm Weiner's hypothesis. The relationship between the present results and Atkinson's theory of achievement motivation, which also hypothesizes an effect of perceived difficulty on performance, is discussed.

Levels of energizing drive and learning of tasks of different difficulty levels

Twenty Ss were divided into hi-SRT and lo-SRT groups on basis of their scores on the Serial Reaction-time Test (SRT). The hi-SRT group performed significantly better than the lo-SRT group in all the items of a nonsense syllable learning task irrespective of the difficulty values of the items, except in the case of extremely easy items. Implica­ tions of this finding on the relationship between energizing drive and learning of difficult tasks have been discussed. Spence and the followers of his theory believe that high drive leads to better performance in easy but worse performance in difficult tasks (Spence, 1956). This prediction has been amply verified in verbal learning experiments (e.g., Standish & Champion,1960; Spence, Farber, & McFann, 1956, etc.). Most of these experiments use Manifest Anxiety Scale (MAS) (Taylor, 1953) to discriminate drive levels since, in terms of Spence's theory, anxiety can be regarded as an index of Hullian general drive. However, recently Eysenck and his associates have reported that persons under induced drive did in fact perform better than persons without induced drive in both easy and difficult verbal learning tasks (Willet & Eysenck, 1962; Willet, 1964). One possible explanation of such contradiction of findings on the relationship between drive and perfor­ mance on difficult tasks lies in the dual nature of drive. It has often been indicated that drive is energizing as well as disrupting in nature (e.g., Jones, 1960). How­ ever, it is also possible, as indicated by some, that there are two different kinds of drive-energizing and dis­ rupting-and that disrupting drive is quite close to anxiety (Sarason, 1957). When Hull formulated his theory on the facilitating effect of drive, most probably he had energizing drive, and certainly not disrupting drive, in mind. The results of anxiety studies, therefore, have no serious bearing on this aspect of Hull's theory. This exploratory experiment is designed to study the rela­ tionship between individual differences in energizing drive and learning of verbal materials of different difficulty levels. Serial Reaction-time Test (SRT) has been used in this study as measure of individual difference in energizing

Transposition with minimal errors during training trials

Twenty-six Ss were trained to respond to the larger of two stimulus squares under a procedure whereby the smaller, unrewarded stimulus was faded in, starting with a 2 10 - inch line and increasing in 30 equal increments until it was completed on the last training trial. One S verbalized the principle of “bigger than” in explaining her choice of the larger stimulus. Of the remaining 25 Ss, 18 completed the training series while making only one or less responses to the unrewarded stimulus. These Ss were divided into three groups and treated under the following schedule: Group I—immediate transposition test; Group II—15 additional training trials with both stimuli complete, followed by the transposition test; and Group III—30 additional training trials with both stimuli complete, followed by the transposition test. The seven Ss who made two or more errors on the original 30 fading-in trials made up a fourth group that was tested immediately after the training series. The Ss in all groups showed a significant tendency to transpose. The present results, like other findings cited herein, demonstrated certain conditions under which Spence's theory of transposition does not appear to hold for human Ss.

Verbal Factors in the Discrimination Learning of Children

Most so-called stimulus-response theories in psychology are based on principles established in research on conditioning. The last 25 years has witnessed an increasing influence of such theories on research programs dealing with discrimination learning. That is, conditioning principles have been applied to discrimination learning data in an attempt to explain or account for the phenomena discovered in the discrimination learning situation. The Spence theory, published in 1936, was the first systematic attempt to apply conditioning principles to discrimination learning. For the first Io or 15 years of this quarter century, the application of such theories was largely restricted to the discrimination performance of infrahuman subjects. Recent years have seen a marked increase in the number of applications of stimulusresponse theories to the discrimination learning of human subjects, and, in particular, to that of children. Since even behavioristic psychologists do not ignore introspection as a potential source of information, it was inevitable that behavior theorists would make increasing use of the notion of response-produced stimulation in their attempts to apply conditioning principles to human learning. The logical or philosophical justification for the use of such concepts was laid many years ago in the writings of such people as Tolman and Hull. These writers, and others, rejected the Watsonian doctrine that mentalistic, or, more precisely, mentalistic-sounding concepts could not appear in the vocabulary of psychology. They insisted, however, that such concepts be introduced only after having been defined by means of terms which referred to public phenomena. The empirical utility of such a view was demonstrated in several experiments conducted in the 1930's. Experiments by Leeper (1935) and Hull (1933) demonstrated that rats could discriminate on the basis of being hungry or being thirsty, which led to the notion of drive stimuli (SD). Shipley (1933, 1935), in his research on mediated or secondary generalization, demonstrated that stimuli resulting from a subject's response can serve as a cue to which other responses can be conditioned. This led to the concept

UNIPROCESS VS. DUOPROCESS IN MONKEY'S DISCRIMINATION LEARNING

The purpose of this experiment was to test the SPENGE theory of discrimination learning. Eight monkeys were given 112 problems each, 13 trials in length. The 8 th to 13 th trials were reversals of the 1 st to 6 th trials after a single cue on the 7 th trial in one half of the problems. The performance of learning set under a shift from the positive to the negative cue was superior to the shift from the negative to the positive. These trends of the learning set curves in terms of Trial 8 differed from each other. These differences may be interpreted in terms of neither SPENCE theory, nor Uniprocess theory. It was suggested that BIRCH'S interpretation might be appropriate to learning set situation.

ANXIETY IN SECONDARY MODERN AND GRAMMAR SCHOOL CHILDREN

Summary. Theory regarding anxiety now yields a variety of predictions concerning its incidence in secondary modern and grammar school children. An investigation was designed to test some of these predictions.Anxiety was measured by means of eight self‐rating questionnaires. These were used in twelve single‐sex secondary modern and grammar schools to obtain a sample of approximately nine hundred children, divided equally into eight groups according to age, sex and school type. Differences between mean scores of secondary modern and grammar school groups were tested for significance by a series of t‐tests. The secondary modern school groups had significantly higher mean scores on three measures of manifest and separation anxiety, and higher mean scores on a measure of emotional introversion. The secondary grammar school groups had a significantly higher mean score on a measure of social introversion, and higher mean scores on measures of self‐blame and thinking introversion.The results are discussed in relation to Davis's theory of “adaptive, socialised anxiety.” This theory is not supported unless it is assumed that the kind of anxiety referred to by Davis bears little relation to what is generally understood by the word. Predictions following from generalisations of Lynn's work on anxiety and attainment also fail to be verified. It is concluded that if any assertion may be made regarding the general relationship of anxiety and attainment among secondary school children, it is that they are negatively correlated. This is supported by Vernon's finding of a positive correlation of attainment and emotional stability among children.Predictions made from Spence's theory are verified. Predictions made from a theory which identifies neuroticism with autonomic drive and introversion with the slow accumulation and fast dissipation of reactive inhibition are not supported.

A Test of Spence's Theory of Incentive-Motivation

A Test of Spence's Theory of Incentive-Motivation

関係把握におけるSpence理論の検討

Some experiments of transposition of size discrimination learning in human infants and rats were made to test whether the results were in accord with Spence's theory.In Exp. I, subjects were divided into two groups. Group I was, trained to the larger one of the two stimuli and Grgup II was trained to the smaller. Transposition tests were made in the positive direction, in the negative direction, and in the stimulus pairs which were made by combining the two stimuli which were located on both sides of the training stimuli.The results were compared with the conclusion deduced from Spence's theoretical generalization curves.1) In the group trained to the larger, both infants and rats showed high percentage of relative responses in every transposition test. From Spence's theory it was deduced that some absolute responses should be made in the tests of the positive direction. So the results did not strictly accord with this inference.2) In the group trained to the smaller both infants and rats showed very low percentage of relative responses in the tests of the positive direction. This accorded with the results of Spence's experiments in chimpanzees And with the inference from his theory. But, in the tests of stimulus pairs made by combining the two stimuli which were located on both sides of the training stimuli, infants showed chance score. Spence's theory is unable to account for this result.3) Since in infants, the low percentage of the relative responses is due to the increase of absolute responses, the consideration of these two types of responses will be sufficient. In animals, on the other hand, the position habit, which is a different type of response from the above two, should be taken into consideration. Spence's theory only predicts the low percentage of relative responses in general, hence it can not explain the difference of response-types depending upon the individual subject or species. In Exp. II, the proportion of areas of stimuli was increased and whether, in rats, the absolute response was made or not was examined.4) when subjects were trained to the larger of the stimuli they skowed very high percentage of relative responses. But, when they were trained to the smaller, most of all the animals showed the position habit. That is, in the transposition of size discrimination learning at least, rats do not make the absolute response, while the fact that infants make absolute response in such a situation is well known. Spence's theory can not explain the difference between these two species.It should be concluded that Spence's theory can not explain strictly the transposition behavior of infants and rats as dealt with in the present study. However, whether this is due to the intrinsic defect in his theory or can be settled by partial correction, is at issue in the present learning theory so that the writer would not attempt to draw any definite conclusion. As there are so many factors which affect the characteristics of his theoretical curves and which he did not consider, it is no wonder that Spence's theory is not applicable to any particular situation

An Analysis of the Formation of Visual Discriminative Habits of Rhesus Monkeys

An early variant of S-R theory states that the difficulty of a discriminative problem involving recurrent stimuli is a function of the differential excitatory strengths associated with the positive and negative stimuli; the excitatory strength of a stimulus being dependent on the relative number of times it has been reinforced or non-reinforced in previous discriminations.1 Thus, if the positive member of a new pair has been rewarded more frequently than the negative in previous pairings with other stimuli, this new problem will be solved quickly, and the number of errors made by individual Ss will be correlated (negatively) with the magnitude of the difference in frequency of reinforcement in favor of the positive stimulus. Spence obtained significant rank order correlations between R (a measure of the difference in the excitatory strengths of the positive and negative stimuli) and performance on a series of tests involving recombinations of a small number of stimuli. He concluded that "discrimination-learning is a cumulative process resulting from differential reinforcement and non-reinforcement of responses to relevant stimulus-components. "2 Lashley later tested Spence's theory with respect to the relevance of stimuluscomponents and the cumulative effects of reinforcement, and demonstrated that differential reinforcement affects only those components of a compound stimulus toward which S is set to respond.8 Since this demonstration of perceptual dominance in 1942, S-R theorists have largely restricted themselves to consideration of stimuli varying in a single dimension. Lashley also repeated Spence's experiment in an attempt to evaluate the adequacy of cumulative reinforcement and non-reinforcement as a sufficient explanation of discriminative learning. He found, however, that correlational analyses, based on small samples, were not sufficiently precise to distinguish the alternative predictions of continuity and non-continuity theory. More recently, studies of the formation of learning sets in primates suggest that Spence's experimental design may yet provide a critical test of continuity theory.4 Especially relevant is Riopelle's observation that both positive and negative transfer are suppressed as the formation of learning set progresses: "During the first phase

Relational Learning in the Rat

The phenomenon of transposition revealed in experiments on discrimination has long provided one of the most crucial challenges to the neoPavlovian conception of cortical functioning.' The fact of transposition suggested the existence of a process of relational learning which did not seem understandable in terms of the independent acquisition of excitatory and inhibitory properties by isolated stimuli. Some years ago, however, Spence advanced an interpretation which met this challenge in a most ingenious manner.2 By making certain assumptions concerning the generalization of excitatory and inhibitory effects along relevant stimulus-dimensions, Spence finds it possible to explain, in essentially Pavlovian terms, not only the occurrence of transposition, but also its failure to occur under given experimental conditions. Even the fact that it is often more difficult for animals to learn to respond to stimuli of given characteristics, independently of context, rather than to a relation between pairs, can, paradoxically enough, be deduced from this theory which postulates no perception of relations by the animal.3 Recently, however, Lashley and Wade have vigorously criticized the conception of stimulus-generalization which is central to Spence's theory.4 Gradients of similarity, they maintain, do not come ready-made in the organism, but are established in the course of the animal's experience. Stimulus-dimensions are defined for the animal not by conditioning to a single stimulus but by training which provides opportunity to compare two or more stimuli-"either a direct comparison of sensory impressions