The Evidence on Teaching

Improving fluid intelligence with training on working memory

Jaeggi SM, Buschkuehl M, Jonides J, Perrig WJ · 2008

grade Crctdeveloper-ledfailed
Sample
Four training groups differing in training dose, plus controls (small groups)
Population
Healthy young adults (out of the 4-18 age band)
Design
The origin of the modern brain-training boom, recorded here as the claim under test rather than as evidence for it. Graded C: small groups, a no-contact control rather than an active placebo, and a single task per construct - precisely the three design features Shipstead et al. (2012) identified and Redick et al. (2013) removed, whereupon the effect disappeared.
Key findings
Reported transfer from a demanding dual n-back working-memory task to measures of fluid intelligence, with a dose-response relationship between amount of training and size of Gf gain. This finding failed direct replication (Redick 2013) and has not survived meta-analysis with active controls.
Genetic confound
Low by design (randomised), but the reported effect is not robust to expectancy and test-specific practice, which the no-contact control cannot rule out.

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Fluid intelligence (Gf)transfer from n-back trainingreported gain, dose-dependent - failed direct replication in Redick 2013standardizedpost-trainingbusiness-as-usualend-of-treatmentfar-transfer

Cited by

Improving fluid intelligence with training on working memory · The Evidence on Teaching