Separating cognitive capacity from knowledge: a new hypothesis

Cowan, Nelson · 2007 · OpenAlex

DOI: 10.1016/j.tics.2007.04.001

archive: archived pipeline: cataloged

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper proposes a unified hypothesis that working memory (WM) and reasoning share related capacity limits, quantified by the number of items or interrelationships that can be actively maintained. The authors argue that these limits stem from the limited ability to form and preserve bindings between elements in memory, a process that requires attention. By employing principled procedures to control for recoding and other strategies that reduce memory and processing loads, the study aims to separate pure capacity limitations from knowledge-based effects, allowing for an orderly quantification of these limits in both domains. The methodology involves reviewing diverse experimental techniques designed to prevent grouping and rehearsal, thereby isolating chunk capacity. These include spatial array comparisons, perfect serial recall, running span tasks with unpredictable endpoints, and multi-object tracking. The paper also details the Method of Analysis of Relational Complexity (MARC), which assesses reasoning complexity by defining the number of slots or arity required to represent a cognitive process. For instance, transitive inference is defined as having a relational complexity (RC) of 3, while 5-way interactions represent the upper limit of human parallel processing. The authors analyze how conceptual chunking and segmentation serve as mechanisms to reduce processing loads, noting that while chunking compresses representations, it renders internal variables inaccessible unless unpacked. Key findings indicate that immediate memory capacity is roughly constant when measured in meaningful units, converging on a range of 3 to 5 chunks for adults, with a central tendency of approximately 3.5 items. This limit applies regardless of modality; studies show a 1-to-1 trade-off between verbal and spatial memory when sensory cues are eliminated, suggesting a central capacity allocated across modalities. In reasoning, performance declines as complexity increases, reaching chance levels at 5-way interactions, which implies a limit of four variables that can be processed in parallel. The paper notes that latent variable constructs of WM capacity account for approximately 0.60 of the variance in reasoning and fluid intelligence. Furthermore, neuroimaging evidence suggests that parietal lobes, which integrate information across modalities, may house this central capacity, distinct from frontal areas involved in attention control. The significance of this work lies in its integration of WM and reasoning theories, suggesting that both are constrained by the same attentional binding mechanism. This framework allows researchers to distinguish between capacity limits and knowledge effects, providing a clearer understanding of cognitive development and expertise. The hypothesis that optimal group size in recall (2–4 items) mirrors the maximal number of chunks that can be inter-associated at once offers a coherent explanation for observed memory patterns. Ultimately, this approach opens new research opportunities by enabling the independent investigation of complexity and capacity effects, contributing to a more precise model of human information processing limits.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via openalex_abstract on 2026-05-08 (13 acquisition events logged).

Summary generated by qwen3.8-27b-gittensor on 2026-08-23; verification: pending re-verification.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• working memory
• capacity resource theory
• cognitive capacity variation