BackgroundPrior longitudinal and correlational research with adults and adolescents indicates that spatial ability is a predictor of science learning and achievement. However, there is little research to date with primary‐school aged children that addresses this relationship. Understanding this association has the potential to inform curriculum design and support the development of early interventions.AimsThis study examined the relationship between primary‐school children's spatial skills and their science achievement.MethodChildren aged 7–11 years (N = 123) completed a battery of five spatial tasks, based on a model of spatial ability in which skills fall along two dimensions: intrinsic–extrinsic; static–dynamic. Participants also completed a curriculum‐based science assessment.ResultsControlling for verbal ability and age, mental folding (intrinsic–dynamic spatial ability), and spatial scaling (extrinsic–static spatial ability) each emerged as unique predictors of overall science scores, with mental folding a stronger predictor than spatial scaling. These spatial skills combined accounted for 8% of the variance in science scores. When considered by scientific discipline, mental folding uniquely predicted both physics and biology scores, and spatial scaling accounted for additional variance in biology and variance in chemistry scores. The children's embedded figures task (intrinsic–static spatial ability) only accounted for variance in chemistry scores. The patterns of association were consistent across the age range.ConclusionSpatial skills, particularly mental folding, spatial scaling, and disembedding, are predictive of 7‐ to 11‐year‐olds’ science achievement. These skills make a similar contribution to performance for each age group.