No Evidence for an Item Limit in Change Detection

Keshvari, Shaiyan; van den Berg, Ronald; Ma, Wei Ji · 2013 · OpenAlex-citations

DOI: 10.1371/journal.pcbi.1002927

archive: archived pipeline: cataloged verified

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

Summary

This study addresses the debate regarding the nature of visual working memory limitations, specifically contrasting "item-limit models," which posit a fixed capacity for storing a discrete number of items, with "continuous-resource models," which suggest memory is limited by the precision of stimulus encoding that decreases as set size increases. The authors argue that previous change detection studies failed to distinguish between these theories because they used highly discriminable stimuli and large changes, thereby ignoring the effects of limited encoding precision and comparison errors. To resolve this, the researchers conducted two change detection experiments involving orientation and color, manipulating both set size and the magnitude of change, including small changes. They rigorously compared five computational models: three item-limit variants (infinite-precision, slots-plus-averaging, and slots-plus-resources) and two continuous-resource variants (equal-precision and variable-precision). All models incorporated a Bayesian decision stage to account for noisy measurements. The variable-precision model uniquely assumed that encoding precision varies across items and trials, drawn from a gamma distribution, rather than being fixed or equally divided. The results demonstrated that human performance was best explained by the variable-precision continuous-resource model. Bayesian model comparison provided strong evidence favoring this model over all item-limit models and the equal-precision model. Specifically, the variable-precision model accurately predicted hit rates, false-alarm rates, and psychometric curves across different change magnitudes and set sizes. Crucially, the study showed that apparent "guessing" rates, often cited as evidence for item limits in other paradigms, are an epiphenomenon of variable precision and comparison errors, not evidence of discarded items. The item-limit models failed to account for the dependence of performance on change magnitude. These findings sharply challenge the theoretical basis for many neural studies of working memory that rely on the item-limit hypothesis. By demonstrating that working memory resource is continuous and variable, the study supports the view that memory limitations arise from the quality of encoding rather than a fixed quantity of storage slots. This implies that previous interpretations of working memory capacity based on change detection tasks may be fundamentally flawed, necessitating a shift toward continuous-resource frameworks in cognitive neuroscience.

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.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

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

• working memory