Taking working memory training from the laboratory into schools

Publication: Educational Psychology

Institution: MRC Cognition & Brain Sciences Unit (Cambridge)

Investigator(s): Joni Holmes, Ph.D., Susan E. Gathercole, Ph.D.

Program: Cogmed RM

Background & Aim: Poor working memory (WM) is commonly observed in children with educational underachievement. A large proportion of children with low WM fail to meet expected standards for either math, reading, or both and children designated as having special education needs are 6 times more likely to have WM impairments than those not needing special education. It thus seems that low WM puts children at risk for poor scholastic attainment.

Previous research has demonstrated that WM training, such as Cogmed, improves WM and may lead to improved academic outcomes. However, the extant literature points largely to highly controlled studies that do not reflect the conditions under which training typically occurs. It is thus important for researchers to demonstrate how WM training can be extended to children at high educational risk in their typical school setting, using existing support resources and classroom structure.

The aim of Trial 1 in this research was to investigate whether teacher led WM training results in the same pattern of generalization to non-trained WM tasks as observed in tightly controlled research settings. Assessment in Trial 1 was conducted using the Automated Working Memory Assessment (AWMA), which consists of 8 WM tasks that are distinct from those practiced on during training (ie., non-trained tasks).

The aim of Trial 2 was to evaluate whether children with poor academic performance could improve their performance on national achievement tests. Assessment in Trial 2 included test questions taken from previous Standard Assessment Test (SAT) paper tests for English and math. The English questions tested reading, writing, speaking, and listening skills. The math questions tested the children’s ability to use and apply math, complete tests of number of algebra, shape space and measures of handling data. Children’s performance on these tests combine with teacher observations to inform judgments about a child’s progress compared to Assessing Pupil’s Progress grids. Department for Education National Curriculum levels range from 1 to 10 for children in compulsory education in the United Kingdom. There are three sublevels within each level (a,b, and c). An ‘a’ indicates that a student is performing consistently at a level and is ready to progress to the next. A ‘b’ means that they are secure at a particular level and a ‘c’ means that they are just starting on a level. Children are expected to progress by 2 sublevels per school year and achieve a level 4c by the end of Year 6.

Trial 1

Population & Sample Size: N = 22 mixed ability children, ages 8 – 10 years (Year 4) in adaptive Cogmed RM training group

* Children were trained in a single group of 22 children at the beginning of each school day and supervised by their teacher and a teaching assistant

Design: Field trial, blind raters, test-retest

T1 = baseline, T2 = post-test


I. 90% of children completed the full Cogmed training protocol.

II. Children significantly improved at T2 on non-trained tasks of visuo-spatial and verbal STM and visuo-spatial and verbal WM (Automated Working Memory Assessment; AWMA).

III. Children with low baseline WM made greater gains than children with high baseline WM on non-trained tasks of visuo-spatial STM and verbal WM (AWMA).

Trial 2

Population & Sample Size: N = 50 children with low academic performance, ages 9 – 11 years (Year 5 and 6)

• n = 25 children ages 9 -10 years (Year 5) in adaptive Cogmed RM training group
• n = 25 children ages 10 -11 years (Year 6) in adaptive Cogmed RM training group

* Year 5 children trained in a single group of 25 students supervised by the Head teacher and a teaching assistant at the end of the school day

** Year 6 children trained in 2 smaller groups of 12 and 13 students and were supervised by the same school staff at the end of the school day

Design: Field trial, blind raters, age-gender matched passive controlled, test-retest

T1 = baseline, T2 = post-test


I. Children in Year 5 Cogmed training group significantly improved compared to the control group on a standardized measure of math (SAT). Note: Control group had a drop in performance across the school year.

II. Children in Year 6 Cogmed training group significantly improved compared to the control group on both a standardized measure of math and English (SAT).

III. 85% of Year 6 Cogmed training group reached nationally expected levels of attainment in English at the end of Year 6 compared to 72% of the control group.

IV. Impact of training on academic progress not mediated by children’s baseline academic performance.

Summary and Implications: The purpose of the current study was to investigate whether computerized WM training (Cogmed) could enhance WM performance for students when led by teachers in schools. Compliance for children across both trials was high, with 80% of students completing at least 20 sessions of training. Children’s improvement on the trained tasks were equivalent in both trials, including mixed ability and low achieving students in groups ranging in size from 12 to 25. Thus, this research demonstrated that it was feasible for students to complete WM training in large groups in schools with both high rates of compliance and good rates of progress on trained activities.

Younger children (ages 8 – 9 years) that trained as whole class before school demonstrated improved performance on non-trained standardized measures of WM, with improvements in all assessed aspects. The greatest improvements were found when students needed to recall sequences of visuo-spatial information or simultaneously hold in and mind and manipulate visuo-spatial or verbal information. This finding is important as performance on these non-trained tasks is strongly associated with the ability to control and focus attention, as well as, highly predictive of children’s learning abilities across the school years.

Trial 2 provided evidence that low achieving children (ages 9 -11 years) who participated in school-led WM training had enhanced academic performance at the end of the year. Year 5 children showed improvement in math compared to the passive age-gender matched control group though, the control group did have a drop in performance across the year. Year 6 students showed improvements in math and English compared to the passive control group and a greater proportion reached target levels of attainment in English by year’s end than the control group.

The transfer of training gains to improvements in National Curriculum assessment of English and math provides an important step forward in the consideration of cognitive training as an educational intervention. Rather than report on artificial WM tasks benefiting children, this research demonstrates transfer to educationally relevant measure of academic ability under real world conditions. Future research should investigate whether gains transfer to larger, possible whole school, interventions with randomized controlled designs so to more precisely establish the value of implementing WM training at school.