The more the merrier? A systematic review and meta-analysis of working memory capacity and category learning

The role of working memory capacity (WMC) in category learning (CL) remains theoretically contested. While the COVIS model posits that information-integration (II) learning is WMC-independent, empirical findings are mixed. Through a systematic review and subsequent three-level meta-analysis of 21 articles (164 effect sizes), we synthesized the WMC–CL relationship. Results revealed a significant overall positive correlation (r = 0.258; trim-and-fill adjusted r = 0.118). Notably, category structure (rule-based vs. information-integration) did not significantly moderate this relationship, challenging the COVIS prediction that implicit learning is independent of WMC. However, WMC task-specific moderation revealed significantly weaker correlations for reading span compared to operation span, memory updating, and spatial storage. No moderating effects were found for age, gender, or stimulus modality. These findings suggest that performance across diverse category structures shares a common reliance on WMC.