Working Memory Capacity Correlates Positively with Category Learning Across Rule Structures

A systematic review and meta-analysis of cognitive learning studies has found that higher working memory capacity correlates positively with category learning across different organizational structures. Investigating the contentious role of working memory capacity in how people learn categories, researchers led by Lingying Jin, Delin Yu, Ziyu Jin, and Hong Li synthesized empirical evidence to address conflicting claims in cognitive psychology. Their quantitative synthesis demonstrated an overall positive correlation between working memory capacity and categorization performance, providing new benchmark estimates across diverse experimental paradigms.

The investigation directly evaluated predictions stemming from the competition of verbal and implicit systems model, widely known as COVIS. In theoretical frameworks such as COVIS, category learning is thought to recruit distinct cognitive architectures depending on the underlying category structure. Specifically, the model posits that rule-based category learning relies on explicit, hypothesis-testing mechanisms governed by working memory, whereas information-integration learning is hypothesized to be implicit and independent of working memory capacity. Because prior empirical investigations had yielded mixed outcomes, the systematic synthesis was designed to test whether working memory resources truly differentiate these two categorization formats.

To rigorously examine these theoretical tensions, the authors conducted a comprehensive systematic review that evaluated 27 articles exploring the direct relationship between working memory capacity and category learning. Additionally, they reviewed seven separate articles that implemented experimental manipulations of working memory load during learning tasks. From this body of literature, the authors identified 21 empirical studies containing sufficient quantitative data to incorporate into a formal meta-analysis. In total, the statistical synthesis encompassed 164 effect sizes derived from 2,510 individual participants, offering a broad empirical foundation across multiple testing laboratories.

Across the 21 meta-analyzed articles, the primary statistical synthesis revealed a significant overall positive correlation between working memory capacity and category learning performance, showing a pooled correlation coefficient of r = 0.258. To account for potential publication bias within the literature, the investigators applied a trim-and-fill adjustment method. Following this adjustment, the relationship remained statistically positive, yielding an adjusted correlation coefficient of r = 0.118. These pooled observations confirm that higher working memory capacity is consistently linked to superior categorization outcomes overall.

Crucially, category structure did not emerge as a statistically significant moderator of the correlation between working memory capacity and learning success. When contrasting rule-based learning against information-integration learning, the analysis demonstrated that working memory capacity facilitates category acquisition regardless of whether the tasks involve explicit verbalizable rules or integrated dimensions. Furthermore, 61 percent of the reviewed studies examining information-integration tasks reported a positive correlation with working memory capacity, directly challenging the assumption that information-integration learning proceeds independently of working memory resources.

The systematic analysis of experimental working memory load manipulations uncovered phase-dependent effects during category acquisition. When concurrent working memory load was applied specifically during the feedback processing phase, it selectively disrupted rule-based learning. In contrast, when working memory load was interpolated during the stimulus-response interval or embedded directly within the categorization tasks themselves, it impaired performance across both rule-based and information-integration structures. These findings indicate that working memory load selectively or broadly interferes with learning depending on the exact temporal phase in which the cognitive demand is introduced.

The meta-analysis also determined that different working memory assessment paradigms differentially predicted category learning outcomes. Measures of reading span demonstrated relatively weaker associations with categorization performance compared to operation span tasks, memory updating tasks, and spatial storage assessments. This divergence indicates that specific cognitive processing demands and task modalities within working memory assessments capture differing degrees of the variance associated with category learning efficiency.

Finally, moderator analyses revealed that several contextual and demographic variables did not significantly alter the association between working memory and category learning. The modality of the stimuli presented to participants, specifically visual versus auditory formats, showed no moderating influence on the reported effect sizes. Similarly, demographic factors across the sampled literature, including participant age, gender distribution, and cultural individualism, did not moderate the observed relationships, pointing to a general cognitive association across tested adult demographics.

Subject of Research: Working memory capacity and category learning meta-analysis

Article Title: The more the merrier? systematic review and meta-analysis of working memory and category learning

Article References: Jin, L., Yu, D., Jin, Z., & Li, H. (2026). The more the merrier? systematic review and meta-analysis of working memory and category learning. BMC Psychology. https://doi.org/10.1186/s40359-026-05901-8

Image Credits: AI Generated

DOI: 10.1186/s40359-026-05901-8

Keywords: working memory capacity, category learning, meta-analysis, systematic review, rule-based learning, information-integration learning, COVIS model, operation span, reading span, memory updating, spatial storage, working memory load, feedback processing, cognitive psychology, stimulus-response interval, effect size