AIMS: To investigate the association between early cannabis use and subclinical psychotic experiences, distinguishing between five levels of use: never used, discontinued use (life-time users who did not use in the preceding year), experimental use, regular use and heavy use. DESIGN: Cross-sectional observational study. SETTING: Dutch Health Behaviour in School-aged Children (HBSC) study, 2005 wave. PARTICIPANTS: A total of 4552 secondary school children aged 12-16 years. MEASUREMENTS: Cannabis use, Community Assessment of Psychic Experiences (CAPE) positive scale, confounding factors: age, gender, family affluence, household composition, social support, alcohol use, cigarette smoking, ethnicity and urbanicity. FINDINGS: The association between cannabis use and subclinical positive symptoms was confirmed, and remained significant after extensive adjustment for potential confounders. Associations were found for all user groups, with strongest associations for the discontinued use group (β = 0.061, P = 0.000) and for the heavy use group (β = 0.065, P = 0.000). CONCLUSIONS: There is an enduring association between cannabis use at an early age and subclinical positive psychotic experiences, even after abstaining from cannabis for at least 1 year.