WME: Visual Working Memory: Capacity, Binding, and Updating

Visual Working Memory on the WME tests active maintenance and manipulation of visual information over seconds. It is not long-term visual memory and not passive short-term storage: tasks such as spatial span, change detection, n-back, feature binding, and updating require you to hold information while attention is occupied or while the content is transformed.

Core rules and concepts

• Capacity is limited to roughly three or four objects, but capacity is better described in chunks or precision than as a fixed number of pixels. Slot models propose a small number of discrete representations; resource models propose a shared pool whose precision varies with load.
• Change detection separates storage from guessing. Raw percent correct confounds sensitivity with response bias, so scoring uses hits minus false alarms or d-prime. A candidate who says 'changed' often can look accurate on change trials while being poor on same trials.
• Binding is fragile. Color-shape, shape-location, and object-location bindings can fail even when the individual features are remembered. A task that swaps features between objects is testing binding, not feature detection.
• Manipulation is distinct from maintenance. Forward spatial span mainly taxes storage; backward span, n-back, and updating tax the central executive. A normal forward span with a poor n-back points to executive or updating difficulty.
• Verbal recoding is a threat. Naming colors, shapes, or positions moves the task into the phonological loop. Complex or name-resistant stimuli, articulatory suppression, or concurrent tasks reduce this.

Worked example

A candidate scores 95% correct at set size 2 and 70% at set size 6 on a color change-detection task; false alarms are 10% at set size 6. Their 2-back accuracy is 60%, while forward spatial span is normal. Reasoning: near-ceiling performance at two objects shows adequate encoding. The drop at six objects shows a capacity limit. With 70% hits and 10% false alarms, d-prime is approximately 1.9, consistent with about three objects held-not a floor effect. The 2-back deficit, with intact forward span, suggests an updating or executive problem rather than a storage problem. Before concluding, check for verbal labeling: if the colors were easily named, the capacity estimate may be inflated by verbal strategies.

Traps that cost marks

• Interpreting raw accuracy without correcting for guessing or false alarms.
• Confusing perceptual load with memory load; a cluttered display can impair performance before storage is taxed.
• Ignoring ceiling and floor effects when comparing set sizes.
• Treating working memory as identical to short-term memory.
• Failing to control verbal recoding.
• Overlooking speed-accuracy tradeoffs and practice effects.