What happens to the things in working memory we are not attending to?
When attention is focused on one remembered item, nearby unattended items are not stored less precisely, but they are more likely to be swapped into the wrong location.
Source
The Graded Fate of Unattended Stimulus Representations in Visuospatial Working Memory
Study at a glance
- Design
- Human experiment — Within-subject retro-cue colour-recall experiments: continuous colour-wheel report (Exp 1, four items) and categorical five-colour report (Exp 2), 80% valid cues.
- N
- Experiment 1: 21 students recruited, 20 analysed after one did not finish the second session; Experiment 2: 20 different students.
- Population
- Ghent University students with normal colour vision
- Outcome
- Recall error; mixture-model precision, guessing and distance-specific misbinding (swap) rates; categorical error proportions by non-target distance
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
Students saw arrays of coloured discs, then a retro-cue pointed to one remembered location before a probe asked for the colour at one location. On 80% of trials the cue was valid, and on invalid trials the researchers varied how far the probe was from the cued spot. In Experiment 1 people picked the colour on a continuous colour wheel and a mixture model split their errors into precision, random guessing and swaps with non-targets at each distance; Experiment 2 used five clearly distinct colours so swaps could be counted directly.
What they found
Valid cues reduced recall error and raised the chance of reporting the correct item, confirming the cue worked. The precision of uncued items did not change with distance from the cued location. Swap errors, however, fell steadily as non-targets got further from the probed location, and this gradient was steeper when attention had to shift to a location close to the cued one. Experiment 2 replicated the distance gradient in swaps without any modelling, though there the gradient did not differ between valid and invalid cues.
The limits
What it doesn't show
Each experiment tested only about 20 mostly female students, so the precision null result could reflect limited power. The authors note the model was fitted to fewer than 50 trials per condition, which may make parameter estimates unreliable, and that the outermost positions were tested more often at the largest distance, possibly biasing how people prioritised them. The high cue validity (80%) may have encouraged people to neglect uncued items, so the conclusions may not hold with less reliable cues. The account that items are stored as abstract gist is explicitly speculative.
Key terms
- Retro-cue
- A cue shown after the items have disappeared that points to which remembered item is likely to be tested, directing attention within memory.
- Binding (swap) error
- Reporting the feature of a different remembered item than the one probed, because the item-to-location link was lost.
- Mixture model
- A statistical model that splits recall errors into precise target responses, random guesses and swaps with non-targets.
- Representational precision
- How finely a remembered feature is stored, measured here by the spread of responses around the true colour.
- Cue validity
- Whether the probed item matches the cued item; valid trials test the cued item, invalid trials test an uncued one.
Flashcards
0 of 10 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
What was the main result about unattended items near the cued location?
Common questions
Why use a colour wheel instead of a yes/no memory test?
A continuous report shows how close the answer was, so the researchers could tell apart fuzzy memories, pure guesses and answers that matched a different item.
Does this mean unattended items are forgotten?
Not exactly: their colours were stored as precisely as ever, but they were more likely to be tied to the wrong place, especially when near the attended location.
Why run a second experiment?
Because the first result depended on a model; using five clearly different colours let the team count swaps directly and check the gradient was real.
More on Working memory