Procedural learning of unstructured categories

Instance theory predicts categorization decisions in the absence of categorical structure: A computational analysis of artificial grammar learning without a grammar

Theories of categorization have historically focused on the stimulus characteristics to which people are sensitive. Artificial grammar learning (AGL) provides a clear example of this phenomenon, with theorists debating between knowledge of rules, fragments, whole strings, and so on as the basis of categorization decisions (i.e., stimulus-driven explanations). We argue that this focus loses sight of the more important question of how participants make categorization decisions on a mechanistic level (i.e., process-driven explanations). To address the problem, we derived predictions from an instance-based model of human memory in a pseudo-AGL task in which all study and test strings were generated randomly, a task that stimulus-driven explanations of AGL would have difficulty accommodating. We conducted a standard AGL experiment with human participants using the same strings. The model's predictions corresponded to participants' decisions well, even in the absence of any experimenter-generated structure and regardless of whether test stimuli contained any incidental structure. We argue that theories of categorization ought to continue shifting towards the goal of modeling categorization at the level of cognitive processes rather than primarily attempting to identify the stimulus characteristics to which participants are drawn.

A role for the medial temporal lobes in category learning

Despite much research, the role of the medial temporal lobes (MTL) in category learning is unclear. Two unstructured categorization experiments explored conditions that might recruit MTL category learning and memory systems—namely, whether the stimulus display includes one or two stimuli, and whether category membership depends on configural properties of the stimulus features. The results supported three conclusions. First, in agreement with prior research, learning with single stimulus displays depended on striatal-mediated procedural learning. Second, and most important, learning with pair displays was mediated by MTL declarative memory systems. Third, the use of stimuli in which category membership depends on configural properties of the stimulus features made MTL learning slightly more likely. Overall, the results suggested that the MTL are most likely to mediate learning when the participant must decide which of two configural stimuli belongs to a selected category.

A neural interpretation of exemplar theory

Exemplar theory assumes that people categorize a novel object by comparing its similarity to the memory representations of all previous exemplars from each relevant category. Exemplar theory has been the most prominent cognitive theory of categorization for more than 30 years. Despite its considerable success in providing good quantitative fits to a wide variety of accuracy data, it has never had a detailed neurobiological interpretation. This article proposes a neural interpretation of exemplar theory in which category learning is mediated by synaptic plasticity at cortical-striatal synapses. In this model, categorization training does not create new memory representations, rather it alters connectivity between striatal neurons and neurons in sensory association cortex. The new model makes identical quantitative predictions as exemplar theory, yet it can account for many empirical phenomena that are either incompatible with or outside the scope of the cognitive version of exemplar theory. (PsycINFO Database Record

Perceptual category learning and visual processing: An exercise in computational cognitive neuroscience

The field of computational cognitive neuroscience (CCN) builds and tests neurobiologically detailed computational models that account for both behavioral and neuroscience data. This article leverages a key advantage of CCN-namely, that it should be possible to interface different CCN models in a plug-and-play fashion-to produce a new and biologically detailed model of perceptual category learning. The new model was created from two existing CCN models: the HMAX model of visual object processing and the COVIS model of category learning. Using bitmap images as inputs and by adjusting only a couple of learning-rate parameters, the new HMAX/COVIS model provides impressively good fits to human category-learning data from two qualitatively different experiments that used different types of category structures and different types of visual stimuli. Overall, the model provides a comprehensive neural and behavioral account of basal ganglia-mediated learning.

Dopamine dependence in aggregate feedback learning: A computational cognitive neuroscience approach

Procedural learning of skills depends on dopamine-mediated striatal plasticity. Most prior work investigated single stimulus-response procedural learning followed by feedback. However, many skills include several actions that must be performed before feedback is available. A new procedural-learning task is developed in which three independent and successive unsupervised categorization responses receive aggregate feedback indicating either that all three responses were correct, or at least one response was incorrect. Experiment 1 showed superior learning of stimuli in position 3, and that learning in the first two positions was initially compromised, and then recovered. An extensive theoretical analysis that used parameter space partitioning found that a large class of procedural-learning models, which predict propagation of dopamine release from feedback to stimuli, and/or an eligibility trace, fail to fully account for these data. The analysis also suggested that any dopamine released to the second or third stimulus impaired categorization learning in the first and second positions. A second experiment tested and confirmed a novel prediction of this large class of procedural-learning models that if the to-be-learned actions are introduced one-by-one in succession then learning is much better if training begins with the first action (and works forwards) than if it begins with the last action (and works backwards).

Differential contributions of the globus pallidus and ventral thalamus to stimulus-response learning in humans

The ability to learn associations between stimuli, responses and rewards is a prerequisite for survival. Models of reinforcement learning suggest that the striatum, a basal ganglia input nucleus, vitally contributes to these learning processes. Our recently presented computational model predicts, first, that not only the striatum, but also the globus pallidus contributes to the learning (i.e., exploration) of stimulus-response associations based on rewards. Secondly, it predicts that the stable execution (i.e., exploitation) of well-learned associations involves further learning in the thalamus. To test these predictions, we postoperatively recorded local field potentials (LFPs) from patients that had undergone surgery for deep brain stimulation to treat severe movement disorders. Macroelectrodes were placed either in the globus pallidus or in the ventral thalamus. During recordings, patients performed a reward-based stimulus-response learning task that comprised periods of exploration and exploitation. We analyzed correlations between patients' LFP amplitudes and model-based estimates of their reward expectations and reward prediction errors. In line with our first prediction, pallidal LFP amplitudes during the presentation of rewards and reward omissions correlated with patients' reward prediction errors, suggesting pallidal access to reward-based teaching signals. Unexpectedly, the same was true for the thalamus. In further support of this prediction, pallidal LFP amplitudes during stimulus presentation correlated with patients' reward expectations during phases of low reward certainty - suggesting pallidal participation in the learning of stimulus-response associations. In line with our second prediction, correlations between thalamic stimulus-related LFP amplitudes and patients' reward expectations were significant within phases of already high reward certainty, suggesting thalamic participation in exploitation.

Declarative strategies persist under increased cognitive load

When humans simultaneously execute multiple tasks, performance on individual tasks suffers. Complementing existing theories, this article poses a novel question to investigate interactions between memory systems supporting multi-tasking performance: When a primary and dual task both recruit declarative learning and memory systems, does simultaneous performance of both tasks impair primary task performance because learning in the declarative system is reduced, or because control of the primary task is passed to slower procedural systems? To address this question, participants were trained on either a perceptual categorization task believed to rely on procedural learning or one of three different categorization tasks believed to rely on declarative learning. Task performance was examined with and without a simultaneous dual task thought to recruit working memory and executive attention. To test whether the categories were learned procedurally or declaratively, the response keys were switched after a learning criterion had been reached. Large impairments in performance after switching the response keys are taken to indicate procedural learning, and small impairments are taken to indicate declarative learning. Our results suggest that the declarative memory categorization tasks (regardless of task difficulty) were learned by declarative systems, regardless of whether they were learned under dual-task conditions.

Frontoparietal networks involved in categorization and item working memory

Categorization and memory for specific items are fundamental processes that allow us to apply knowledge to novel stimuli. This study directly compares categorization and memory using delay match to category (DMC) and delay match to sample (DMS) tasks. In DMC participants view and categorize a stimulus, maintain the category across a delay, and at the probe phase view another stimulus and indicate whether it is in the same category or not. In DMS, a standard item working memory task, participants encode and maintain a specific individual item, and at probe decide if the stimulus is an exact match or not. Constrained Principal Components Analysis was used to identify and compare activity within neural networks associated with these tasks, and we relate these networks to those that have been identified with resting state-fMRI. We found that two frontoparietal networks of particular interest. The first network included regions associated with the dorsal attention network and frontoparietal salience network; this network showed patterns of activity consistent with a role in rapid orienting to and processing of complex stimuli. The second uniquely involved regions of the frontoparietal central-executive network; this network responded more slowly following each stimulus and showed a pattern of activity consistent with a general role in role in decision-making across tasks. Additional components were identified that were associated with visual, somatomotor and default mode networks.

How may the basal ganglia contribute to auditory categorization and speech perception?

Listeners must accomplish two complementary perceptual feats in extracting a message from speech. They must discriminate linguistically-relevant acoustic variability and generalize across irrelevant variability. Said another way, they must categorize speech. Since the mapping of acoustic variability is language-specific, these categories must be learned from experience. Thus, understanding how, in general, the auditory system acquires and represents categories can inform us about the toolbox of mechanisms available to speech perception. This perspective invites consideration of findings from cognitive neuroscience literatures outside of the speech domain as a means of constraining models of speech perception. Although neurobiological models of speech perception have mainly focused on cerebral cortex, research outside the speech domain is consistent with the possibility of significant subcortical contributions in category learning. Here, we review the functional role of one such structure, the basal ganglia. We examine research from animal electrophysiology, human neuroimaging, and behavior to consider characteristics of basal ganglia processing that may be advantageous for speech category learning. We also present emerging evidence for a direct role for basal ganglia in learning auditory categories in a complex, naturalistic task intended to model the incidental manner in which speech categories are acquired. To conclude, we highlight new research questions that arise in incorporating the broader neuroscience research literature in modeling speech perception, and suggest how understanding contributions of the basal ganglia can inform attempts to optimize training protocols for learning non-native speech categories in adulthood.

Rule-based and information-integration perceptual category learning in children with attention-deficit/hyperactivity disorder

OBJECTIVE

Suboptimal functioning of the basal ganglia is implicated in attention-deficit/hyperactivity disorder (ADHD). These structures are important to the acquisition of associative knowledge, leading some to theorize that associative learning deficits might be expected, despite the fact that most extant research in ADHD has focused on effortful control. We present 2 studies that examined the acquisition of explicit rule-based (RB) and associative information integration (II) category learning among school-age children with ADHD.

METHOD AND RESULTS

In Study 1, we found deficits in both RB and II category learning tasks among children with ADHD (n = 81) versus controls (n = 42). Children with ADHD tended to sort by the more salient but irrelevant dimension (in the RB paradigm) and were unable to acquire a consistent sorting strategy (in the II paradigm). To disentangle whether the deficit was localized to II category learning versus a generalized inability to consider more than 1 stimulus dimension, in Study 2 children completed a conjunctive RB paradigm that required consideration of 2 stimulus dimensions. Children with ADHD (n = 50) continued to underperform controls (n = 33).

CONCLUSIONS

Results provide partial support for neurocognitive developmental theories of ADHD that suggest that associative learning deficits should be found, and highlight the importance of using analytic approaches that go beyond asking whether an ADHD-related deficit exists to why such deficits exist.