Which kind of working memory matters for maths as kids grow?
Visual-spatial working memory matters for arithmetic in younger primary pupils, but by the final grades verbal working memory becomes the stronger predictor across all four operations.
Source
Verbal and visual-spatial working memory and mathematical ability in different domains throughout primary school
Study at a glance
- Design
- Cohort — One school year with three arithmetic-fluency measurements, in grades 2 to 6 (partly cross-sectional across grades), analysed with multigroup latent growth curve models; working memory measured once each.
- N
- N=4285 · 4,285 children with arithmetic data from an original 4,337; 3,830 had visual-spatial and 3,499 had verbal working-memory data.
- Population
- Dutch primary school children in grades 2-6 from 32 schools
- Outcome
- Level and within-year growth of timed addition, subtraction, multiplication and division fluency
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
Children in grades 2 to 6 at Dutch schools took a timed arithmetic test three times during one school year, covering addition, subtraction, multiplication and division. Each child also did a computerised visual-spatial working-memory game (remembering where coloured lions appeared) and a verbal one (recalling spoken words backwards). Growth curve models tested whether each working-memory type predicted children's arithmetic level and their rate of improvement, and how this changed by grade.
What they found
Both kinds of working memory predicted how well children performed in every operation, but neither predicted how fast they improved over the year. In grades 2 to 4 the two were equally predictive; in grades 5 and 6 verbal working memory was significantly stronger, while visual-spatial working memory stopped predicting most operations. The shift looked similar across all four operations, contrary to the idea that addition and subtraction rely more on visual-spatial resources.
The limits
What it doesn't show
Each working-memory type was measured by a single task, and the verbal task might partly reflect reading ability. The maths measure was a speeded fluency test, so processing speed could inflate its link with working memory. Grade comparisons are partly cross-sectional, so differences between grades could reflect cohort or teaching differences rather than development, and one school year was too short to capture meaningful individual differences in growth. The design is correlational, so it cannot show that working memory causes maths performance.
Key terms
- Visual-spatial sketchpad
- The part of Baddeley's working-memory model that briefly holds and manipulates visual and spatial information.
- Phonological loop
- The working-memory component that stores and rehearses speech-based or verbal information.
- Complex span task
- A memory task that requires both storing and processing or updating information, used to measure working memory rather than simple short-term storage.
- Latent growth curve model
- A statistical model that estimates each person's starting level and rate of change across repeated measurements.
- Math fluency
- How many simple arithmetic problems a child can solve correctly under time pressure.
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 did working memory predict in this study?
Common questions
Why would visual-spatial working memory matter less as children get older?
Younger children often solve sums with visual or counting-based procedures, while older children increasingly retrieve memorised facts and use verbal strategies.
Did working memory predict who improved fastest during the year?
No. It predicted children's level of performance but not their rate of growth, possibly because one school year gave too little variation in growth.
Did the pattern differ between addition and multiplication?
Not much. The shift toward verbal working memory looked similar in all four operations, though multiplication was the least predicted overall.
More on Working memory