An elemental model of associative learning: I. Latent inhibition and perceptual learning

This paper presents a brief, informal outline followed by a formal statement of an elemental associative learning model first described by McLaren, Kaye, and Mackintosh (1989). The model assumes representation of stimuli by sets of elements (i.e., microfeatures) and a set of associative algorithms that incorporate the following: real-time simulation of learning; an error-correcting learning rule; weight decay that distinguishes between transient and permanent associations; and modulation of associative learning that gives high salience to and, hence, promotes rapid learning with novel, unpredicted stimuli and reduces the salience for a stimulus as its error term declines. The model is applied in outline fashion to some of the basic phenomena of simple conditioning and, in greater detail, to the phenomena of latent inhibition and perceptual learning. A detailed account of generalization and discrimination will be provided in a later paper.     

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.     

Inhibitory associations between neutral stimuli: A comparative approach

In Experiments 1A, 1B, and 1C, nonhuman subjects, rats, received long alternated exposures to two compound flavors, AX and BX, that shared one flavor in common, X. Following this, conditioning of an aversion to A was sufficient to establish B as a conditioned inhibitor of the aversive unconditioned stimulus, passing both summation and retardation tests. Two additional experiments (Experiments 2 and 3) expanded the generality of these results to humans, using similar designs but an auditory discrimination learning task. A set of notes sequentially presented served as cues and fictitious composers served as outcomes. Both summation and retardation effects were found (Experiments 2 and 3, respectively). Experiment 4 then sought to clarify the mechanism underlying these effects. The results are discussed within several theoretical frameworks, most centrally the McLaren, Kaye, and Mackintosh (1989) theory of perceptual learning.     

An easy-to-hard effect after nonreinforced preexposure in a sweetness discrimination

Experiments 1A and 1B used a taste-aversion procedure with rats to demonstrate that exposure to easily discriminated flavors along a dimension (1 % and 10 % sucrose) can facilitate learning a subsequent hard discrimination (4 % and 7 % sucrose) when one of those flavors is paired with illness. Experiment 1A compared the effects of preexposure to the easily discriminated flavors against exposure to the same stimuli used in the discrimination training or no exposure at all. Experiment 1B replicated the conditions in Experiment 1A, with 2 additional days of training and unrestricted access to the flavors on CS+/CS– trials in discrimination training. Contrary to findings with multidimensional stimuli (Scahill & Mackintosh, Journal of Experimental Psychology: Animal Behavior Processes, 30, 96–103, 2004; Suret & McLaren, The Quarterly Journal of Experimental Psychology, 56B, 30–42, 2003), we found that preexposure to the easily discriminable stimuli varying along a single dimension of sweetness facilitated subsequent discrimination training over the other conditions in each experiment. We discuss the results in terms of the ideas presented by Gibson (1969) and Mackintosh (Psychological Review, 82, 276–298, 1975) and in terms of hedonic variables not considered by theories of perceptual learning.     

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.     

Human and animal perceptual learning: Some common and some unique features

A selective summary of the four contributions to this special issue of Learning & Behavior on perceptual learning is presented. Mackintosh and Hall propose an associative analysis of perceptual learning. It is argued that Tsushima and Watanabe’s psychophysical evidence and Fiser’s Bayesian-modeling work represent (in different ways) challenges to the associative approach. Some tentative suggestions are explored with regard to how animal learning theorists might meet these challenges. Finally, the role of awareness in perceptual learning is briefly examined.     

Stimuli that signal the absence of reinforcement are paid more attention than are irrelevant stimuli

According to established theories of attention (e.g., Mackintosh, 1975; Sutherland & Mackintosh, 1971), simple discriminations of the form AX+ BX- result in an increase in attention to stimuli A and B, which are relevant to the outcome that follows them, at the expense of X, which is irrelevant. Experiments that have apparently shown such changes in attention have failed to determine whether attention is enhanced to both A and B, which signal reinforcement and nonreinforcement, respectively, or just to A. In Experiments 1 and 2, pigeons were trained with a number of discriminations of the kind AX+ BX-, before compounds that had been consistently nonreinforced were involved in a subsequent discrimination. Both experiments provided support for theories that propose that more attention is paid to stimuli that consistently signal nonreinforcement than to irrelevant stimuli in simple discriminations.     

Attention and salience in associative blocking

The associative learning effect called blocking has previously been found in many cue-competition paradigms where all cues are of equal salience. Previous research by Hall, Mackintosh, Goodall, and dal Martello (1977) found that, in animals, salient cues were less likely to be blocked. Crucially, they also found that when the to-be-blocked cue was highly salient, the blocking cue would lose some control over responding. The present article extends these findings to humans and suggests that shifts in attention can explain the apparent loss of control by the previously learned cue. A connectionist model that implements attentional learning is shown to fit the main trends in the data. Model comparisons suggest that mere forgetting, implemented as weight decay, cannot explain the results.     

Normalization between stimulus elements in a model of Pavlovian conditioning: Showjumping on an elemental horse

Harris and Livesey. Learning & Behavior, 38, 1-26, (2010) described an elemental model of associative learning that implements a simple learning rule that produces results equivalent to those proposed by Rescorla and Wagner (1972), and additionally modifies in "real time" the strength of the associative connections between elements. The novel feature of this model is that stimulus elements interact by suppressively normalizing one another's activation. Because of the normalization process, element activity is a nonlinear function of sensory input strength, and the shape of the function changes depending on the number and saliences of all stimuli that are present. The model can solve a range of complex discriminations and account for related empirical findings that have been taken as evidence for configural learning processes. Here we evaluate the model's performance against the host of conditioning phenomena that are outlined in the companion article, and we present a freely available computer program for use by other researchers to simulate the model's behavior in a variety of conditioning paradigms.     

An instance theory of associative learning

We present and test an instance model of associative learning. The model, Minerva-AL, treats associative learning as cued recall. Memory preserves the events of individual trials in separate traces. A probe presented to memory contacts all traces in parallel and retrieves a weighted sum of the traces, a structure called the echo. Learning of a cue–outcome relationship is measured by the cue’s ability to retrieve a target outcome. The theory predicts a number of associative learning phenomena, including acquisition, extinction, reacquisition, conditioned inhibition, external inhibition, latent inhibition, discrimination, generalization, blocking, overshadowing, overexpectation, superconditioning, recovery from blocking, recovery from overshadowing, recovery from overexpectation, backward blocking, backward conditioned inhibition, and second-order retrospective revaluation. We argue that associative learning is consistent with an instance-based approach to learning and memory.     

An attention-modulated associative network

We present an elemental model of associative learning that describes interactions between stimulus elements as a process of competitive normalization. Building on the assumptions laid out in Harris (2006), stimuli are represented as an array of elements that compete for attention according to the strength of their input. Elements form associations among each other according to temporal correlations in their activation but restricted by their connectivity. The model moves beyond its predecessor by specifying excitatory, inhibitory, and attention processes for each element in real time and describing their interaction as a form of suppressive gain control. Attention is formalized in this model as a network of mutually inhibitory units that moderate the activation of stimulus elements by controlling the level to which the elements are suppressed by their own inhibitory processes. The model is applied to a range of complex discriminations and related phenomena that have been taken as evidence for configural-learning processes.     

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.     

Symmetrical generalization decrements: Configural stimulus processing in human contingency learning

Models of associative learning differ in their predictions concerning the symmetry of generalization decrements. Whereas Pearce’s (1994) configural model predicts the same response decrement after adding elements to and after removing elements from a previously trained stimulus, elemental models, such as the replaced elements model and Harris’s (2006) model, anticipate more of a decrement for removing than for adding elements. In three contingency learning experiments, we manipulated the motion and the spatial arrangement of colored dots in order to induce configural or elemental processing by perceptual grouping. The results reliably showed symmetrical decrements for the added and removed groups. The manipulations of the stimuli had no effect on stimulus processing. This is in line with Pearce’s configural model, but it is at variance with the elemental models and previous studies.     

Varieties of perceptual learning

Although most studies of perceptual learning in human participants have concentrated on the changes in perception assumed to be occurring, studies of nonhuman animals necessarily measure discrimination learning and generalization and remain agnostic on the question of whether changes in behavior reflect changes in perception. On the other hand, animal studies do make it easier to draw a distinction between supervised and unsupervised learning. Differential reinforcement will surely teach animals to attend to some features of a stimulus array rather than to others. But it is an open question as to whether such changes in attention underlie the enhanced discrimination seen after unreinforced exposure to such an array. I argue that most instances of unsupervised perceptual learning observed in animals (and at least some in human animals) are better explained by appeal to well-established principles and phenomena of associative learning theory: excitatory and inhibitory associations between stimulus elements, latent inhibition, and habituation.     

Anticipation to Social and Nonsocial Dynamic Cues in Preschool-Age Children

The ability to learn from expectations is foundational to social and nonsocial learning in children. However, we know little about the brain basis of reward expectation in development. Here, 3- to 4-year-olds (N = 26) were shown a passive associative learning paradigm with dynamic stimuli. Anticipation for reward-related stimuli was measured via the stimulus preceding negativity (SPN). To our knowledge, this is the first study to measure an SPN in children younger than age 6. Our findings reveal distinct anticipatory neural signatures for social versus nonsocial stimuli, consistent with previous research in older children. This study suggests an SPN can be elicited in preschoolers and is larger for social than nonsocial stimuli.     

A common error term regulates acquisition but not extinction of causal judgments in people

In three experiments, we used the allergist task to examine the role of error correction mechanisms in the acquisition and extinction of causal judgments in people. Consistent with existing human and animal studies, acquisition of causal judgments was influenced by the discrepancy between the allergenic outcome and that predicted by all of the cues present on a trial (the "common error" term). However, in the present experiments, we failed to detect any evidence for the use of a common error term in extinction learning: Judgments of the allergenic properties of a cue were unaffected by the predictive value of the other cues present on a trial. This asymmetry in the use of a common error term in acquisition and extinction learning is inconsistent with previous animal studies and also with most models of associative learning. However, approaches that allow learning to be specific to a particular arrangement of elemental cues (context specific and state based) offer some explanation of the observed asymmetry.     

Learning about associatively activated stimulus representations: Implications for acquired equivalence and perceptual learning

Pavlovian conditioning is taken to reflect the formation of links between the central representations of stimuli. A link will be formed when presentation of the relevant stimuli is scheduled in a way that ensures that two representations are activated concurrently. Once this has occurred, a representation can be activated not only by the occurrence of the appropriate stimulus but also by way of the link. Evidence is reviewed to suggest that activation produced by this second route is, in some ways, functionally equivalent to direct activation; in particular, an associatively activated representation (animage) appears capable of forming further associative links with other event representations. Learning about associatively activated stimulus representations may play a role in a range of phenomena. Its contribution to the following is discussed: sensory preconditioning, second-order conditioning, acquired equivalence and distinctiveness, equivalence class formation, and the perceptual learning effect. Finally, consideration is given to the way in which existing theories of associative learning might be modified in order to accommodate this process.     

Behavioral and associative effects of differential outcomes in discrimination learning

The role of the reinforcer in instrumental discriminations has often been viewed as that of facilitating associative learning between a reinforced response and the discriminative stimulus that occasions it. The differential-outcome paradigm introduced by Trapold (1970), however, has provided compelling evidence that reinforcers are also part of what is learned in discrimination tasks. Specifically, when the availability of different reinforcing outcomes is signaled by different discriminative stimuli, the conditioned anticipation of those outcomes can provide another source of stimulus control over responding. This article reviews how such control develops and how it can be revealed, its impact on behavior, and different possible mechanisms that could mediate the behavioral effects. The main conclusion is that differential-outcome effects are almost entirely explicable in terms of the cue properties of outcome expectancies—namely, that conditioned expectancies acquire discriminative control just like any other discriminative or conditional stimulus in instrumental learning.     

Fluid intelligence and discriminative operant learning of reinforcement contingencies in a fixed ratio 3 schedule

Although intelligence has traditionally been identified as the ability to learn (Peterson, 1925), this relationship has been questioned in simple operant learning tasks (Spielberger, 1962). Nevertheless, recent pieces of research have demonstrated a strong and significant correlation between associative learning measures and intelligence ( [Kaufman et al., 2009] and [Williams et al., 2008] ). The present paper aims to examine the relationship between intelligence and learning, complying with the experimental conditions of discriminative conditioning in a relatively complex free operant task. To that end, a sample of 1600 subjects applying for an ab initio Air Traffic Control training course was tested. The task is expected to allow identifying the specific influence of intelligence on the discrimination between stimulus dimensions. The results show how the relation among learning and intelligence strengthen on those trials which are critical in terms of discrimination.     

Effects of pretraining on acquisition of novel configural discriminations in human predictive learning

In two experiments, participants acquired one of two target configural discriminations (a biconditional or negative patterning discrimination) in a predictive learning task. In Experiment 1, participants were pretrained with either a configural or an elemental discrimination; in Experiment 2, they were pretrained with a configural discrimination, an elemental discrimination, or a control discrimination that was not expected to bias them toward elemental or configural processing. In both experiments, acquisition of the target configural discriminations was faster after configural pretraining than after elemental pretraining. In addition, the negative patterning discrimination was acquired faster than the biconditional discrimination. Finally, the results of Experiment 2 were more consistent with the notion that elemental pretraining hindered acquisition of the target discriminations than with the notion that configural pretraining enhanced their acquisition. Implications of these findings are discussed.