Discrete fixed-resolution representations in visual working memory

Zhang, Weiwei; Luck, Steven J. · 2008 · OpenAlex-citations

DOI: 10.1038/nature06860

archive: archived pipeline: cataloged verified

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

Summary

This supplementary material supports the main findings of Zhang et al. (2008), which argue that visual working memory relies on discrete, fixed-resolution representations rather than a continuously divisible resource pool. The text provides rigorous statistical validation of the proposed "slots+averaging" model and presents control experiments to rule out alternative explanations regarding encoding time, categorical coding, and stimulus modality. The authors demonstrate that their three-parameter mixture model provides an excellent quantitative fit to the data. Adjusted $r^2$ values for group data exceeded 95% across all conditions, and Kolmogorov-Smirnov and $\chi^2$ tests confirmed that observed data did not significantly deviate from the model. In contrast, a simple resource model, which assumes all items are encoded with varying precision, yielded negative adjusted $r^2$ values and significant deviations, indicating it is quantitatively inconsistent with the results. To ensure these findings were not artifacts of insufficient encoding time or methodological insensitivity, control experiments were conducted. Varying sample duration from 100 to 500 ms had no significant effect on memory probability ($P_m$) or standard deviation (SD). Furthermore, adding sensory noise to stimuli increased SD but did not affect $P_m$, proving the methods were sensitive to precision changes without conflating them with storage capacity. Additional analyses addressed potential confounds in representation type and task methodology. Data from 50 observers showed a linear relationship between actual and reported colors, ruling out the hypothesis that observers store categorical prototypes rather than continuous values. A comparison between the color recall task used in the study and the standard change detection task revealed a strong correlation ($r^2 = .572$) in estimated storage capacity, suggesting both tasks measure similar aspects of working memory. Finally, experiments using shape-defined stimuli replicated the color results: $P_m$ dropped dramatically as set size increased from 3 to 6, while SD remained constant. Crucially, attentional cuing in shape experiments increased $P_m$ for cued items but did not reduce the precision of uncued items, further supporting the conclusion that resources are allocated in discrete chunks rather than being infinitely divisible.

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