BACKGROUND: The weather may be a driver of seasonal patterns in children's physical activity (PA). A better understanding of the relationships between weather and PA may help increase children's PA. This study aims to examine the association between PA and rainfall in 9-10 year old children, and how it may be modified by school policies. METHODS: 1794 participants in the SPEEDY study in Norfolk, UK recorded PA using ActiGraph accelerometers over up to six days in the summer term of 2007. Multilevel regression models were used to determine the day-by-day association between rainfall and minutes spent sedentary, in moderate-to-vigorous PA (MVPA), and average counts per minute (cpm) over the whole day (07:00-21:00) and the lunchtime period (12:00-14:00). School policies for break times in bad weather were fitted as interaction terms with rainfall. RESULTS: Relative to days with no rain, children spent 9.4 minutes (95%CI 7.0 to 11.9) fewer in MVPA, were sedentary for 13.6 minutes (8.8 to 18.4) more, and accumulated 85.9 cpm (66.2 to 105.5) fewer over the whole day on the wettest days. Children allowed to play outside in wet weather showed the lowest lunchtime PA levels on the wettest days, undertaking 9.8 minutes (6.2 to 13.5) fewer MVPA, 16.1 minutes (10.3 to 21.9) more sedentary, and accumulating 408.0 cpm (250.9 to 565.1) fewer than those allowed to be active indoors. CONCLUSIONS: Rainfall is negatively associated with PA in primary school children, but providing indoor physical activities in wet weather may help children maintain physical activity levels irrespective of rainfall.