Elemental representation and configural mappings: Combining elemental and configural theories of associative learning

In this article, we present our first attempt at combining an elemental theory designed to model representation development in an associative system (based on McLaren, Kaye, & Mackintosh, 1989) with a configural theory that models associative learning and memory (McLaren, 1993). After considering the possible advantages of such a combination (and some possible pitfalls), we offer a hybrid model that allows both components to produce the phenomena that they are capable of without introducing unwanted interactions. We then successfully apply the model to a range of phenomena, including latent inhibition, perceptual learning, the Espinet effect, and first- and second-order retrospective revaluation. In some cases, we present new data for comparison with our model's predictions. In all cases, the model replicates the pattern observed in our experimental results. We conclude that this line of development is a promising one for arriving at general theories of associative learning and memory.     

A configural theory of attention and associative learning

A formal account of the relationship between attention and associative learning is presented within the framework of a configural theory of discrimination learning. The account is based on a connectionist network in which the entire pattern of stimulation presented on a trial activates a configural unit that then enters into an association with the trial outcome. Attention is assumed to have two roles within this network. First, the salience of the stimuli at the input to the network can be increased if they are relevant to the occurrence of reinforcement and decreased if they are irrelevant. Second, the associability of configural units can increase on trials when the outcome is surprising and decrease when the outcome is not surprising.     

Preference effects on acquisition and retention of concurrent discriminations by rhesus monkeys

Ten rhesus monkeys were trained on five tasks, each of which consisted of eight concurrently presented object discrimination problems. Sequences of presentation were devised to allow one, two, or three new tasks to intervene between acquisition and retention tests or to provide a 30-day period of no testing. Equivalent and proficient performances were obtained in all retention tests, and no relationship was observed between retention and the initial preference characteristics of various objects. Object pReferences did produce significant influences upon acquisition, but these effects were not as pronounced in early tasks as in later ones. An additional retention test provided support for the contention that monkeys do not necessarily process information about specific object pair discriminations. Rather, they appeared to retain a list of previously rewarded objects even when object pairings were different from those provided during acquisition. Concurrent discriminations involving many distinct objects were resistant to interference and independent of preference characteristics over long retention periods.     

Age differences in the acquisition and retention of concurrent discriminations by monkeys

The performances of adult and aged macaque monkeys were compared on several tasks requiring acquisition and retention of concurrent object discriminations. Cross-sectional and longitudinal comparisons between 11-year-old and over-20-year-old animals were conducted on the acquisition of several 16-problem concurrent tasks, and after 2 weeks, retention by the different age groups was also evaluated. Although the retention tests indicated no age-related impairment, acquisition tests showed that aged animals made reliably more errors-to-criterion than did adult animals. The relative contributions of intersession and intrasession improvement and preference phenomena were considered and contrasted to some previously reported age-specific outcomes. Although the present results affirmed the contention that old animals were selectively impaired on intersession task requirements, some earlier interpretations of preference effects were questioned. Old animals did not show a potentiation of specific object preferences; rather, they tended to be nonsystematic in maintaining their selection of initially preferred objects. Concurrent discrimination was indicated to be a behavioral measure that is, under appropriate circumstances, selectively sensitive to aging and may offer an appropriate comparative model of information-processing characteristics.     

Evaluation and development of a connectionist theory of configural learning

A configural theory of associative learning is described that is based on the assumption that conditioning results in associations between the unconditioned stimulus and a representation of the entire pattern of stimulation that was present prior to its delivery. Configural theory was formulated originally to account for generalization and discrimination in Pavlovian conditioning. The first part of the article demonstrates how this theory can be used to explain results from studies of overshadowing, blocking, summation, and discrimination learning. The second part of the article shows how the theory can be developed to explain a broader range of phenomena, including mediated conditioning, reinforcer devaluation effects, the differential outcomes effect, acquired equivalence, sensory preconditioning, and structural discriminations.     

Nondifferential training of pigeons retards acquisition of subsequent discriminations involving other stimuli

In Experiment 1, two groups of pigeons (n = 8) were given nondifferential (ND) training with a green keylight and a white vertical line on a dark surround nonsystematically alternated. Two groups (n = 8) received single stimulus (SS) training with the green light only. In Experiment 2, two groups of pigeons (n = 8) were given ND training with vertical and horizontal lines, while two other groups (n = 8) received SS training with only the vertical line. In both experiments, all groups were transferred to a green S+ (VI reinforced) and a red S? (extinguished) transfer problem. In each experiment, one ND and one SS group was tested in the same context as initial training (houselight off) and one ND and one SS group was tested in a changed context (houselight on). In both experiments and in both contexts, the ND groups performed less well on the transfer problem than did the SS groups. There was no evidence of greater control by the context in ND than in SS groups, which suggests that the observed difference in acquisition of the transfer task is not attributable to a purported difference in control by the context under the two conditions. The overall results favor the position that nondifferential training reduces attention to stimuli involved in the original training procedure and that this reduced attention transfers to stimuli subsequently experienced.     

Factors influencing the acquisition of tactual random figure discriminations by rhesus monkeys

Rhesus monkeys learned to perform two-object random figure discrimination problems by touch. This performance was compared to the visual acquisition of the same problems (reported earlier). In both modalities, choice latencies were shorter to the correct than to the incorrect figure. As compared to visual, however, tactual learning required significantly more trials to criterion, was independent of stimulus figure complexity, and was characterized by significantly shorter choice latencies. Rotation of the figures 180 deg after criterion performance was achieved, yielded partial recognition of the figures visually, but no recognition tactually. These results are explained by different observational or cue-utilization strategies. The majority of the figure is inspected visually, but only the same small section of the perimeter is used tactually. Thus, restricted use of the available cues characterizes tactual performance and is responsible for generating differences in various measures of learning between vision and touch in the monkey.     

Within-event learning contributes to value transfer in simultaneous instrumental discriminations by pigeons

Following simultaneous discrimination training with pigeons, in which responding to the S−was reinforced on half of the trials and responding to the S− was never reinforced, we examined the effect on the S− of presenting the S− by itself and the effect on the S+ of presenting the S− by itself (relative to an S− or an S− for which there had been no single-stimulus presentations). For Group A−, responding to the S− presented by itself was always reinforced, whereas for Group A−, such responding was extinguished. For Group B−, responding to the S− presented by itself was always reinforced, whereas for Group B+, responding was extinguished. Although both Group A+ and Group A−tended to avoid their associated S− (relative to a control S−), Group A+ avoided its associated S− less than did Group A−. In contrast, although for Group B−, presentation of the S− alone increased the pigeons’ preference for its associated S−(relative to a control S+), for Group B−, presentation of the S−alone had little effect on its preference for its associated S+. These results suggest that presentation of one stimulus from a simultaneous discrimination has two independent and sometimes opposite effects on the other discriminative stimulus. First, it reduces the strength of within-event conditioning between the S+ and the S−, and second, if the value of the singly presented stimulus has increased, some of its newly acquired value will transfer retroactively to the stimulus with which it was originally paired.     

Extinction produces context inhibition and multiple-context extinction reduces response recovery in human predictive learning

Two experiments with human participants were used to investigate recovery of an extinguished learned response after a context change using ABC designs. In an ABC design, the context changes over the three successive stages of acquisition (context A), extinction (context B), and test (context C). In both experiments, we found reduced recovery in groups that had extinction in multiple contexts, and that the extinction contexts acquired inhibitory strength. These results confirm those of previous investigations, that multiple-context extinction can produce less response recovery than single-context extinction, and they also provide new evidence for the involvement of contextual inhibitory processes in extinction in humans. The foregoing results are broadly in line with a protection-from-extinction account of response recovery. Yet, despite the fact that we detected contextual inhibition, predictions based on protection-from-extinction were not fully reliable for the single- and multiple-context group differences that we observed in (1) rates of extinction and (2) the strength of context inhibition. Thus, although evidence was obtained for a protection-from-extinction account of response recovery, this account can not explain all of the data.     

The effect of a pretest in an interactive,multimodal pretraining system for learning science concepts

Multiple traits of language proficiency as well as test method effects were concurrently analyzed to investigate interrelations of construct validity, convergent validity, and discriminant validity using multitrait-multimethod (MTMM) matrices. A total of 585 test takers' scores were derived from the field test of the Pearson Test of English Academic. An MTMM confirmatory factor analysis model was parameterized using 4 traits and 3 assessment methods. The 4 traits included listening, reading, speaking, and integrated skills, while the 3 methods included prescribed multiple-choice responses, constructed responses, and summarized responses. The trait factor loadings were systematically greater than those of methods, providing evidence that the indicators were strongly related to their latent constructs, after adjusting for the method effects. The results showed robust convergent validity, moderate discriminant validity, and insignificant method effects. Implications are discussed.     

Changes in response to S+ and S− during acquisition and overtraining of simultaneous discriminations in rats

Two experiments with rats investigated learning about S+ and S? during various stages of discrimination training. Transfer tests, in which either S+ or S? was retained, or two new stimuli were presented, were used to assess learning early in acquisition, at criterion, or following overtraining. Both choice and latency measures were used. Results indicated that learning about S+ occurs early in acquisition when noncorrection training is given, and little improvement occurs after that time, at least up to 70 trials of overtraining under present conditions. Learning about S? begins somewhat later in acquisition and continues throughout overtraining. When correction training is used, learning about S? occurs earlier in acquisition, and further learning about both S+ and S? occurs during overtraining.     

Declarative knowledge acquisition in immersive virtual learning environments

The author investigated the interaction effect of immersive virtual reality (VR) in the classroom. The objective of the project was to develop and provide a low-cost, scalable, and portable VR system containing purposely designed and developed immersive virtual learning environments for the US Army. The purpose of the mixed design experiment was to compare lecture-based and immersive VR-based multimedia instruction, in terms of declarative knowledge acquisition (i.e. learning) of basic corrosion prevention and control with military personnel. Participants were randomly assigned to the control group (N?=?115) or investigational group (N?=?25) and tested immediately before and after training. The author accessed learning outcomes from the pre-exam and post-exam scores and VR system usability from exit questionnaires. Results indicate that both forms of instruction will increase learning. VR-based did produce higher gain scores and there was a statistically significant interaction between instruction type and time.     

Effects of pretraining on conditioning-enhanced neophobia: Evidence for separable mechanisms of neophobia and aversion conditioning

In two experiments, rats (n = 228) received pretraining access to a distinctive novel flavor (saline) followed by aversion conditioning to a different novel conditioned stimulus (CS) (saccharin). Then the rats were tested for aversion to the CS (saccharin) or for conditioning-enhanced neophobia to a third novel flavor (casein hydrolysate). Pretraining access to a distinctive novel flavor that differed from the CS reliably reduced the magnitude of conditioning-enhanced neophobia to casein, but did not reliably affect conditioned aversion effects to the CS. Pretraining access to the CS reduced aversion effects to the CS and reduced postconditioning neophobia to casein to the performance level shown by ingestion-toxin controls. Results were consistent with the view (Braveman & Jarvis, 1978) that conditioned aversion and neophobia may be independent phenomena with separable underlying mechanisms.     

The effect of feature-target intervals in conditional discriminations on acquisition and expression of conditioned nictitating membrane and heart rate responses in the rabbit

Nictitating membrane (NM) and heart rate (HR) responses were investigated in a conditional discrimination (A→X+ vs. B→X−), using feature-target intervals of 0, 5, 15, and 45 sec. Conditional control of NM responses, but not of HR responses, was acquired to the 400-msec X stimulus in all the groups tested. However, differential conditioning of both the NM and the HR responses to A versus B feature cues appeared for the three shorter intervals. Following acquisition, all the rabbits were tested with the four different feature-target intervals. All the groups showed a gradient of NM responding to X, in which the highest level of responding occurred at or near the interval used in training. The results are discussed with respect to the relationship of simple conditioning of the feature cues to their control over responding during presentation of the target stimulus, the putative role of HR as an index of preparatory processes during presentation of feature cues, and mechanisms of temporal specificity in conditional discriminations.     

Motivation in a computer-supported collaborative learning scenario and its impact on learning activities and knowledge acquisition

Addressing a drawback in current research on computer-supported collaborative learning (CSCL), this study investigated the influence of motivation on learning activities and knowledge acquisition during CSCL. Participants’ (N = 200 university students) task was to develop a handout for which they had first an individual preparing phase followed by a computer-supported collaborative learning phase immediately afterwards. It was hypothesized that in both phases current motivation (in terms of expectancy and value components) influences both learning activities and knowledge acquisition in a positive way. According to main results, only goal orientations (before learning) were associated with knowledge acquisition respectively observed learning activities during the collaborative phase. Expectancy and value components of current motivation related neither to observed learning activities nor to knowledge acquisition during collaborative learning but were in part associated with learning activities and knowledge acquisition during individual learning. The discussion addresses several possible explanations for these unexpected results.     

Effects of morphological instruction on vocabulary acquisition

The effects of a 20-session intervention targeting morphological word structure on vocabulary knowledge were investigated in four Grade 4 and 5 classes, assigned randomly to treatment and control conditions. Hierarchical regression analyses controlling for initial vocabulary showed significant instructional effects on morphological analysis and vocabulary with words that were taught directly and novel words built on bases that were taught in the context of other derivations, but not for words with untaught bases. Results indicated that the treatment group made better use of pre-test vocabulary knowledge in learning new vocabulary. Results are discussed in light of the growing debate regarding whether to teach many words in a shallow way or to provide deep, rich instruction about fewer words.     

Networking vs. rote learning strategies in concept acquisition

Following presentation of a slide-tape instructional program, the performance of subjects in two learning strategy groups-networking and rote-was compared to a control group on a concept learning task and a spatial learning task. Networking proved more effective than rote learning and the control group on both tasks; networking also allowed for improved retention over time (in this case, one week). This article reviews some of the research on learning strategies and suggests how the network strategy can enhance learning in academic situations.