The objectives of this longitudinal study were to determine the prevalence of smoking among primary school children in Liverpool, and to identify the predictors of experimentation with cigarettes during pre-adolescence. A cohort of children (n = 270) completed questionnaires that elicited patterns of child smoking behaviour and children's experiences of smoking in their families and communities each year between the ages of 9 and 11 years. Parents also completed questionnaires. Children's first trials with cigarettes and repeated smoking were reported. The independent variables measured were socio-economic status, familial and peer smoking, and intentions to smoke. By age 11, 27% of children had tried smoking, 12% had smoked repeatedly and 3% were smoking regularly. Variables measured at age 9 predicting experimentation with cigarettes by age 11 were male gender 9P = 0.041) paternal smoking (P = 0.001) fraternal smoking (P = 0.017) a best friend who smoked (P = 0.026) and knowing someone with a smoking-related disease (P = 0.006) Intentions to smoke at age 9 did not predict smoking at age 11 (P < 0.001). In univariate analyses, child smoking was also associated with maternal smoking (P = 0.002 at age 11), living in a low-income household (P < 0.001 at age 10) and living in a deprived area ( P = 0.025 at age 11). Early smoking presents a considerable challenge to health promoters, not least because it is socially patterned. The interventions required must tackle the structural and social pressures that shape smoking behaviour during childhood.