BACKGROUND: England's national cycle training scheme, 'Bikeability', aims to give children in England the confidence to cycle more. There is, however, little evidence on the effectiveness of cycle training in achieving this. We therefore examined whether delivering Bikeability was associated with cycling frequency or with independent cycling. METHODS: We conducted a natural experimental study using information on children aged 10-11 years participating in the nationally-representative Millennium Cohort Study. We identified Cohort participants whose schools had offered Bikeability in 2011-2012 using operational Bikeability delivery data (children in London excluded, as delivery data not available). Our natural experimental design capitalised on the fact that Cohort participants were surveyed at different times during 2012 and were also offered Bikeability at different times during 2012. This allowed us to compare cycling levels between children whose schools delivered Bikeability before their survey interview ('intervention group', N = 2563) and an otherwise comparable group of children whose schools delivered Bikeability later in the year ('control group', N = 773). Parents reported whether their child had completed formal cycle training; their child's cycling frequency; whether their child ever made local cycling trips without an adult; and other child and family factors. We used Poisson regression with robust standard errors to examine whether cycling behaviour differed between the intervention and control groups. RESULTS: Children whose school had offered Bikeability were much more likely to have completed cycle training than the control group (68% vs. 28%, p < 0.001). There was, however, no evidence that delivering Bikeability in school was associated with cycling more often (49.0% cycling at least once per week in the intervention group vs. 49.6% in the control group; adjusted risk ratio 0.99, 95% CI 0.89, 1.10). There was likewise no evidence of an association with cycling independently (51.5% in the intervention group vs. 50.1% in the control group; adjusted risk ratio 0.97, 95% CI 0.89, 1.06). CONCLUSIONS: Offering high-quality cycle training free at the point of delivery in English schools encourages children to do cycle training, but we found no evidence of short-term effects on cycling frequency or independent cycling. Future evaluation should investigate longer-term effects on these and other stated Bikeability objectives such as increasing cycling safety.