Probabilistic Classification Learning Research Articles

The Visual Advantage Effect in Comparing Uni-Modal and Cross-Modal Probabilistic Category Learning.

People rely on multiple learning systems to complete weather prediction (WP) tasks with visual cues. However, how people perform in audio and audiovisual modalities remains elusive. The present research investigated how the cue modality influences performance in probabilistic category learning and conscious awareness about the category knowledge acquired. A modified weather prediction task was adopted, in which the cues included two dimensions from visual, auditory, or audiovisual modalities. The results of all three experiments revealed better performances in the visual modality relative to the audio and audiovisual modalities. Moreover, participants primarily acquired unconscious knowledge in the audio and audiovisual modalities, while conscious knowledge was acquired in the visual modality. Interestingly, factors such as the amount of training, the complexity of visual stimuli, and the number of objects to which the two cues belonged influenced the amount of conscious knowledge acquired but did not change the visual advantage effect. These findings suggest that individuals can learn probabilistic cues and category associations across different modalities, but a robust visual advantage persists. Specifically, visual associations can be learned more effectively, and are more likely to become conscious. The possible causes and implications of these effects are discussed.

With or without Feedback?-How the Presence of Feedback Affects Processing in Children with Developmental Language Disorder.

Language acquisition depends on the ability to process and learn probabilistic information, often through the integration of performance feedback. Children with developmental language disorder (DLD) have demonstrated weaknesses in both probabilistic learning and feedback processing, but the individual effects of each skill are poorly understood in this population. This study examined school-aged children with DLD (n = 29) and age- and gender-matched children with typical development (TD; n = 44) on a visual probabilistic classification learning task presented with and without feedback. In the feedback-based version of the task, children received performance feedback on a trial-by-trial basis during the training phase of the task. In the feedback-free version, children responded after seeing the correct choice marked with a green border and were not presented with feedback. Children with TD achieved higher accuracy than children with DLD following feedback-based training, while the two groups achieved similar levels of accuracy following feedback-free training. Analyses of event-related potentials (ERPs) provided insight into stimulus encoding processes. The feedback-free task was dominated by a frontal slow wave (FSW) and a late parietal component (LPC) which were not different between the two groups. The feedback-based task was dominated by a parietal slow wave (PSW) and an LPC, both of which were found to be larger in the TD than in the DLD group. In combination, results suggest that engagement with feedback boosts learning in children with TD, but not in children with DLD. When the need to process feedback is eliminated, children with DLD demonstrate behavioral and neurophysiological responses similar to their peers with TD.

Stereotypes disrupt probabilistic category learning.

Racial stereotypes exert pernicious effects on decision-making and behavior, yet little is known about how stereotypes disrupt people's ability to learn new associations. The current research interrogates a fundamental question about the boundary conditions of probabilistic learning by examining whether and how learning is influenced by preexisting associations. Across three experiments, participants learned the probabilistic outcomes of different card combinations based on feedback in either a social (e.g., forecasting crime) or nonsocial (e.g., forecasting weather) learning context. During learning, participants were presented with either task-irrelevant social (i.e., Black or White faces) or nonsocial (i.e., darker or lighter clouds) stimuli that were stereotypically congruent or incongruent with the learning context. Participants exhibited learning disruptions in the social compared to nonsocial learning context, despite repeated instructions that the stimuli were unrelated to the outcome (Studies 1 and 2). We also found no differences in learning disruptions when participants learned in the presence of negatively (Black and criminal) or positively valenced stereotypes (Black and athletic; Study 3). Finally, we tested whether learning decrements were due to "first-order" stereotype application or inhibition at the trial level, or due to "second-order" cognitive load disruptions that accumulate across trials due to fears of appearing prejudiced (aggregated analysis). We found no evidence of first-order disruptions and instead found evidence for second-order disruptions: participants who were more internally motivated to respond without prejudice, and thus more likely to self-monitor their responses, learned less accurately over time. We discuss the implications of the influence of stereotypes on learning and memory. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Supplemental Material for Stereotypes Disrupt Probabilistic Category Learning

Supplemental Material for Stereotypes Disrupt Probabilistic Category Learning

What Is the Weather Prediction Task Good for? A New Analysis of Learning Strategies Reveals How Young Adults Solve the Task.

The Weather Prediction Task (WPT) was originally designed to assess probabilistic classification learning. Participants were believed to gradually acquire implicit knowledge about cue–outcome association probabilities and solve the task using a multicue strategy based on the combination of all cue–outcome probabilities. However, the cognitive processes engaged in the resolution of this task have not been firmly established, and despite conflicting results, the WPT is still commonly used to assess striatal or procedural learning capacities in various populations. Here, we tested young adults on a modified version of the WPT and performed novel analyses to decipher the learning strategies and cognitive processes that may support above chance performance. The majority of participants used a hierarchical strategy by assigning different weights to the different cues according to their level of predictability. They primarily based their responses on the presence or absence of highly predictive cues and considered less predictive cues secondarily. However, the influence of the less predictive cues was inconsistent with the use of a multicue strategy, since they did not affect choices when both highly predictive cues associated with opposite outcomes were present simultaneously. Our findings indicate that overall performance is inadequate to draw conclusions about the cognitive processes assessed by the WPT. Instead, detailed analyses of performance for the different patterns of cue–outcome associations are essential to determine the learning strategies used by participants to solve the task.

Feedback Processing During Probabilistic Learning in Children With Developmental Language Disorder: An Event-Related Potential Study.

The purpose of this study was to examine feedback processing within the context of probabilistic learning in children with and without developmental language disorder (DLD). The probabilistic category learning task required 28 children ages 8-13 years old to classify novel cartoon animals that differed in five binary features into one of two categories. Performance feedback guided incremental learning of the stimuli classifications. Feedback processing was compared between children with DLD and age-matched children with typical development (TD) by measuring the magnitude of feedback-related event-related potentials. Additionally, the likelihood of each group to repeat a classification of a stimulus following positive feedback ("stay" behavior) and change a classification following negative feedback ("switch" behavior) served as a measure of the consequence of feedback processing. Children with DLD achieved lower classification accuracy on all learning outcomes compared to their peers with TD. Children with DLD were less likely than those with TD to demonstrate "stay" behavior or to repeat a correct response following positive feedback. "Switch" behavior or changing an incorrect response following negative feedback was found to be at chance level in both groups. Electrophysiological data indicated that children with DLD had a smaller feedback-related negativity effect (i.e., smaller differential processing of positive and negative feedback) when compared to children with TD. Although no differences were found between the two groups in the amplitude of the P3a, strong positive correlations were found between "stay/switch" behavior and the P3a for children in the TD group only. Children with DLD do not appear to benefit from incremental corrective feedback to the same extent as their peers with TD. Processing differences are captured in the initial stages of feedback evaluation and in translating information carried by the feedback to inform future actions.

Category learning in autism: Are some situations better than others?

Autism is diagnosed according to atypical social-communication and repetitive behaviors. However, autistic individuals are also distinctive in the high variability of specific abilities such as learning. Having been characterized as experiencing great difficulty with learning, autistics have also been reported to learn spontaneously in exceptional ways. These contrasting accounts suggest that some situations may be better than others for learning in autism. We tested this possibility using a probabilistic category learning task with four learning situations differing either in feedback intensity or information presentation. Two learning situations compared high- versus low-intensity feedback, while two other learning situations without external feedback compared isolated sequentially presented information versus arrays of simultaneously presented information. We assessed the categorization and generalization performance of 54 autistic and 52 age-matched typical school-age children after they learned in different situations. We found that children in both groups were able to learn and generalize novel probabilistic categories in all four learning situations. However, across and within groups, autistic children were advantaged by simultaneously presented information while typical children were advantaged by high-intensity feedback when learning. These findings question some common aspects of autism interventions (e.g., frequent intense feedback, minimized simplified information), and underline the importance of improving our current understanding of how and when autistics learn optimally. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

The effect of cognitive flexibility on probabilistic category learning

The effect of cognitive flexibility on probabilistic category learning

Strategy Development and Feedback Processing During Complex Category Learning.

In this study, 38 young adults participated in a probabilistic A/B prototype category learning task under observational and feedback-based conditions. The study compared learning success (testing accuracy) and strategy use (multi-cue vs. single feature vs. random pattern) between training conditions. The feedback-related negativity (FRN) and P3a event related potentials were measured to explore the relationships between feedback processing and strategy use under a probabilistic paradigm. A greater number of participants were found to utilize an optimal, multi-cue strategy following feedback-based training than observational training, adding to the body of research suggesting that feedback can influence learning approach. There was a significant interaction between training phase and strategy on FRN amplitude. Specifically, participants who used a strategy in which category membership was determined by a single feature (single feature strategy) exhibited a significant decrease in FRN amplitude from early training to late training, perhaps due to reduced utilization of feedback or reduced prediction error. There were no significant main or interaction effects between valence, training phase, or strategy on P3a amplitude. Findings are consistent with prior research suggesting that learners vary in their approach to learning and that training method influences learning. Findings also suggest that measures of feedback processing during probabilistic category learning may reflect changes in feedback utilization and may further illuminate differences among individual learners.

Linguistic and neurocognitive correlates of probabilistic classification learning in schizophrenia

Individuals with schizophrenia demonstrate impaired implicit learning on cognitively complex tasks and preserved implicit motor learning. However, little is known about how implicit learning may be related to other linguistic and cognitive variables, including development of complex language including comprehension and syntax. This study explored the relationship between probabilistic classification learning, a type of implicit learning style, and linguistic and cognitive skills in schizophrenia. This was done by examining how schizophrenia patients perform on the Weather Prediction Task (WPT) relative to controls, particularly during a dual-task interference condition that assesses task automaticity. Individuals with schizophrenia (N = 34) demonstrated depressed cognitive functioning relative to the controls (N = 18) across nearly all cognitive functions. On the Weather Prediction Task, the schizophrenia group performed less accurately than the control group in later blocks and had a relatively flat learning curve. A significant Group X Block effect when controlling for age and sex suggested differential learning throughout the task. A subgroup of patients did not develop automaticity during the repeated blocks of trials. For those patients who did not develop automaticity over the course of the WPT, linguistic and cognitive skills were strongly correlated with their Block 1 performance. For patients who developed automaticity, overall neurocognitive ability was correlated with their ultimate level of performance on the WPT but not with their Block 1 performance. That language was related to differential learning emphasizes the role of explicit, verbal processes on making initial rapid improvement on the WPT.

Associations between long-term psychosis risk, probabilistic category learning, and attenuated psychotic symptoms with cortical surface morphometry.

Neuroimaging studies have consistently found structural cortical abnormalities in individuals with schizophrenia, especially in structural hubs. However, it is unclear what abnormalities predate psychosis onset and whether abnormalities are related to behavioral performance and symptoms associated with psychosis risk. Using surface-based morphometry, we examined cortical volume, gyrification, and thickness in a psychosis risk group at long-term risk for developing a psychotic disorder (n = 18; i.e., extreme positive schizotypy plus interview-rated attenuated psychotic symptoms [APS]) and control group (n = 19). Overall, the psychosis risk group exhibited cortical abnormalities in multiple structural hub regions, with abnormalities associated with poorer probabilistic category learning, a behavioral measure strongly associated with psychosis risk. For instance, the psychosis risk group had hypogyria in a right posterior midcingulate cortical hub and left superior parietal cortical hub, as well as decreased volume in a right pericalcarine hub. Morphometric measures in all of these regions were also associated with poorer probabilistic category learning. In addition to decreased right pericalcarine volume, the psychosis risk group exhibited a number of other structural abnormalities in visual network structural hub regions, consistent with previous evidence of visual perception deficits in psychosis risk. Further, severity of APS hallucinations, delusional ideation, and suspiciousness/persecutory ideas were associated with gyrification abnormalities, with all domains associated with hypogyria of the right lateral orbitofrontal cortex. Thus, current results suggest that structural abnormalities, especially in structural hubs, are present in psychosis risk and are associated both with poor learning on a psychosis risk-related task and with APS severity.

Stress-induced bias of multiple memory systems during retrieval depends on training intensity

Stressful events promote a shift from hippocampus-dependent ‘cognitive’ learning towards dorsal striatum-dependent ‘habit’ learning. Beyond modulating the recruitment of multiple memory systems during learning, recent evidence suggests that stress may also affect which of these memory systems is employed during retrieval, thereby affecting the nature of remembering. However, while some studies reported increased reliance on ‘habit’ memory retrieval after stress, other studies suggested even a bias towards ‘cognitive’ memory retrieval after stress. In the present experiment, we tested the hypothesis that the nature of the stress effect on the control of memory retrieval depends on the extent of initial training. To this end, participants completed a probabilistic classification learning (PCL) task that can be solved by both the ‘cognitive’ and the ‘habit’ memory systems, which is reflected in the engagement of specific behavioral strategies. Critically, participants received either moderate (100 trials) or intensive (200 trials) training in the PCL task. Participants then underwent a stress protocol or a non-stressful control procedure, before they completed a retrieval version of the PCL task. The effectiveness of the stress manipulation was verified by increases in salivary cortisol and autonomic arousal. Our results further revealed that participants who received moderate training showed, during retrieval, a stress-induced shift towards strategies indicative of the dorsal striatal ‘habit’ memory system. After prolonged training, however, stress did not affect which memory system guided retrieval. The present results indicate that the effect of stress on the engagement of multiple memory systems during retrieval is critically dependent on the extent of initial training and, by inference, on the strength of the multiple memory traces established during learning.

Stress-induced modulation of multiple memory systems during retrieval requires noradrenergic arousal

Stress has been shown to favor dorsal striatum-dependent ‘habit’ memory over hippocampus-dependent ‘cognitive’ memory during learning. Here, we investigated whether stress may modulate the engagement of these ‘cognitive’ and ‘habit’ systems also during memory retrieval and if so, whether such a stress-induced shift in the control of memory retrieval depends on noradrenergic activation. To this end, participants acquired a probabilistic classification learning (PCL) task that can be solved by both the ‘cognitive’ and the ‘habit’ system, reflected in the distinct behavioral strategies. Twenty-four hours later, participants received either the beta-adrenergic receptor antagonist propranolol or a placebo before they underwent a psychosocial stressor or a non-stressful control manipulation, followed by a retrieval version of the PCL task. Overall, participants showed a practice-dependent shift from ‘cognitive’ to ‘habit’ memory. Stressed participants that had received a placebo fell back to a ‘cognitive’ strategy during retrieval, which was linked to an impairment in retrieval performance. Propranolol blocked this stress-induced shift towards the less efficient strategy. Moreover, our results showed that salivary cortisol was related to the retrieval strategy only when paralleled by increased autonomic arousal. Together, these results indicate that stress effects on the modulation of multiple memory system during retrieval necessitate noradrenergic arousal, with relevant implications for retrieval performance under stress.

Stress-induced cortisol modulates the control of memory retrieval towards the dorsal striatum.

Stress can modulate the recruitment of multiple memory systems during learning, favouring dorsal striatal "habit" learning over hippocampal "cognitive" learning. Here, we tested whether stress may also bias the engagement of "cognitive" and "habit" systems during retrieval and thereby affect the nature of remembering. To this end, participants first performed a probabilistic classification learning task that can be solved by both the "cognitive" and the "habit" system. Twenty-four hours later, participants underwent either a stress manipulation or a non-stressful control procedure before they completed a retention test for the previously learned task in the MRI scanner. During this retention test, stress-induced cortisol levels were linked to a relative bias towards behavioural strategies indicative for the "habit" system. At the neural level, stress led to increased dorsal striatal activity during retrieval. Elevated cortisol levels were directly correlated with increased activity in the dorsal striatum and further linked to reduced functional connectivity between the hippocampus and the amygdala, which is assumed to orchestrate the stress-related shift from "cognitive" to "habitual" control. Together, our data suggest that stress may bias the contributions of multiple memory systems also at retrieval, in a manner that promotes dorsal striatal "habit" processes and most likely driven by cortisol.

Does Acute Stress Impact Declarative and Procedural Learning?

It is well established that acute stress can influence memory function, yet its influence may differ across memory systems. Whereas stress sometimes exerts a negative influence on declarative learning, it does not necessarily harm learning in general, as demonstrated in the case of procedural learning. Probabilistic category learning is mediated by the striatum, but delaying feedback by a few seconds shifts learning to become more hippocampal-dependent. Here, we examined the influence of acute stress on this type of learning, under different conditions that favor either procedural-based (immediate feedback) vs. declarative-based (delayed feedback) learning. Sixty-two participants randomly assigned to either stress or non-stress groups, performed a probabilistic category learning task, in which they were instructed to learn associations between cues and outcomes under different feedback conditions (immediate feedback, short-delayed feedback, and long-delayed feedback). Acute stress was induced by the Maastricht Acute Stress Test (MAST), and stress levels were gauged by Galvanic Skin Response (GSR) measures and a self-reported questionnaire. Results showed that although the MAST was effective in inducing stress, this did not harm learning in either of the feedback conditions. These findings suggest that not all hippocampal-based learning types are negatively influenced by stress.

Modeling the Effect of Speed Emphasis in Probabilistic Category Learning

For many decisions in life, there is uncertainty about the outcome even if the stimulus is a familiar one. Under these circumstances, people often rely upon previous experience to make a choice. In general, more accurate responding is achieved when one considers decision-relevant information for longer. There are practical limits, however, to how long one can deliberate, and so decision-makers must balance decision speed against accuracy. Traditionally, the overall speed with which people respond has been understood in terms of changes in the evidence threshold used for decision-making. However, more recent findings suggest that speed pressure may also affect the quality of information people recruit for making decisions. The current study investigated the effects of emphasizing either response speed or response accuracy on decision-making in a probabilistic category learning task. Strikingly, conclusions from two model-based analyses of performance depended on whether data were analyzed at the level of individuals or the group average. For the group-level analyses, speed pressure selectively influenced evidence threshold. For the individual-level analyses, speed pressure affected both evidence threshold as well as the quality of information driving decision-making for four out of six participants. Our findings are consistent with recent results that identify common brain areas associated with setting evidence thresholds decision-making and representing category-level information in trial-by-trial associative learning.

Learning functions in short-term cocaine users

ObjectiveThis study examined learning functions in short-term cocaine users and control participants. MethodSeventeen active cocaine users (reporting 3.5 mean years of cocaine use) and seventeen non-cocaine-using controls (with similar reported levels of alcohol and marijuana use) were compared on tasks measuring different aspects of learning. ResultsThe cocaine users performed more poorly on the Weather Prediction and List-Learning tasks, as well as supplementary executive and psychomotor function tasks, than controls. ConclusionsIndividuals with a relatively short duration of cocaine use exhibited moderate weaknesses in probabilistic category learning, verbal learning and psychomotor functions, relative to controls. These weaknesses may underpin difficulty in learning from the probabilistic consequences of behavior and hinder the ability to respond to cognitive-behavioral treatments.

Combining error-driven models of associative learning with evidence accumulation models of decision-making.

As people learn a new skill, performance changes along two fundamental dimensions: Responses become progressively faster and more accurate. In cognitive psychology, these facets of improvement have typically been addressed by separate classes of theories. Reductions in response time (RT) have usually been addressed by theories of skill acquisition, whereas increases in accuracy have been explained by associative learning theories. To date, relatively little work has examined how changes in RT relate to changes in response accuracy, and whether these changes can be accounted for quantitatively within a single theoretical framework. The current work examines joint changes in accuracy and RT in a probabilistic category learning task. We report a model-based analysis of changes in the shapes of RT distributions for different category responses at the level of individual stimuli over the course of learning. We show that changes in performance are determined solely by changes in the quality of information entering the decision process. We then develop a new model that combines an associative learning front end with a sequential sampling model of the decision process, showing that the model provides a good account of all aspects of the learning data. We conclude by discussing potential extensions of the model and future directions for theoretical development that are opened up by our findings.

Glucocorticoids, Noradrenergic Arousal, and the Control of Memory Retrieval

Glucocorticoids and noradrenaline can enhance memory consolidation but impair memory retrieval. Beyond their effects on quantitative memory performance, these major stress mediators bias the engagement of multiple memory systems toward "habitual" control during learning. However, if and how glucocorticoids and noradrenaline may also affect which memory system is recruited during recall, thereby affecting the control of retrieval, remain largely unknown. To address these questions, we trained healthy participants in a probabilistic classification learning task, which can be supported both by cognitive and habitual strategies. Approximately 24 hr later, participants received a placebo, hydrocortisone, yohimbine (an α2-adrenoceptor antagonist increasing noradrenergic stimulation), or both drugs before they completed a recall test for the probabilistic classification learning task. During training, all groups showed a practice-dependent shift toward more habitual strategies, reflecting an "automatization" of behavior. In the recall test, after a night of sleep, this automatization was even more pronounced in the placebo group, most likely due to offline consolidation processes and with beneficial effects on recall performance. Hydrocortisone or yohimbine intake abolished this further automatization, preventing the shift to a more efficient memory system and leading, in particular in the hydrocortisone group, to impaired recall performance. Our results suggest that glucocorticoids and noradrenergic stimulation may modulate the engagement of different strategies at recall and link the well-known stress hormone-induced retrieval deficit to a change in the system controlling memory retrieval.

Impaired sequential and partially compensated probabilistic skill learning in Parkinson's disease.

The striatal dopaminergic dysfunction in Parkinson's disease (PD) has been associated with deficits in skill learning in numerous studies, but some of the findings remain controversial. Our aim was to explore the generality of the learning deficit using two widely reported skill learning tasks in the same group of Parkinson's patients. Thirty‐four patients with PD (mean age: 62.83 years, SD: 7.67) were compared to age‐matched healthy adults. Two tasks were employed: the Serial Reaction Time Task (SRT), testing the learning of motor sequences, and the Weather Prediction (WP) task, testing non‐sequential probabilistic category learning. On the SRT task, patients with PD showed no significant evidence for sequence learning. These results support and also extend previous findings, suggesting that motor skill learning is vulnerable in PD. On the WP task, the PD group showed the same amount of learning as controls, but they exploited qualitatively different strategies in predicting the target categories. While controls typically combined probabilities from multiple predicting cues, patients with PD instead focused on individual cues. We also found moderate to high correlations between the different measures of skill learning. These findings support our hypothesis that skill learning is generally impaired in PD, and can in some cases be compensated by relying on alternative learning strategies.