I read with great interest the article by Arseneault et al (Reference Arseneault, Cannon and Witton2004). It demonstrates without any doubt that cannabis use in adolescence acts as a causal risk factor for schizophrenia in adulthood. It is, therefore, a pity that the authors had to add the caveat that, since not all adults with schizophrenia used cannabis in adolescence and since the majority of cannabis users do not develop schizophrenia in adulthood, cannabis can be neither a sufficient nor a necessary cause for psychosis. This formulation is erroneous and was used, in exactly the same words, many years ago when prospective studies in the UK established the aetiological role of tobacco in lung cancer. In an elementary textbook on statistics, Schwartz (Reference Schwartz1999) explains that this error arises from the faulty use of the term ‘cause’, which applies to the domain of certainty, whereas in the domain of uncertainty (i.e. of illness) the definition of a causal factor is that it provokes an increase in risk, as perfectly demonstrated by the authors. One wonders why they make this elementary error. It is unlikely to be due to psychological resistance, as was the case with tobacco smokers at that time. Perhaps they believe that schizophrenia (or psychosis) is a known disease entity, as defined according to international systems of classification (DSM–IV, ICD–10) which, unfortunately, continue to exclude substance use from their diagnostic criteria.