Aims: To investigate the hypothesis that changes in cannabis prevalence among Dutch secondary school students (aged 12-17 years) were paralleled by shifts in the age of first cannabis use. Design and participants: Data were derived from five waves (1988, 1992, 1996, 1999 and 2003) of the Dutch National School Survey on Substance Use, a nationally representative cross-sectional study, with a total of 32 777 respondents. Measurements: Written questionnaires on cannabis, tobacco, alcohol, other drug use and socio-demographic and behavioural variables were administered in classroom settings. Findings: Survival analysis showed a strong increase in cumulative incidences by age of first cannabis use from 1988 to 1992, a further increase in 1996 and stabilization in 1999, continuing into 2003. From 1992 to 1996, age of onset shifted towards younger ages. Onset peaked at age 15 in 1992 and age 14 in 1996. The proportion of life-time cannabis users starting at age 13 or younger increased from 26% in 1992 to 41% in 1996. The overall trend was similar for boys and girls. Conclusions: The study largely confirmed the expectation that the increase in cannabis use from 1988 to 1996 was paralleled by a decrease in the age of first cannabis use. From 1996 to 2003 age of first cannabis use and prevalence stabilized, possibly occasioned by a change in cannabis policy in the mid-1990s. © 2005 Society for the Study of Addiction.