In our view, Matthew Hill's arguments against a causal link between cannabis use and schizophrenia fail to clinch this debate (Nature 525, S14; 2015).

His contention that the increased societal use of cannabis over time is not reflected in increased rates of schizophrenia has been tested only once to our knowledge — and that study came to the opposite conclusion (J. Boydell et al. Psychol. Med. 36, 1441–1446; 2006). In multifactorial conditions such as schizophrenia, an increase in one risk factor is not necessarily balanced by a decrease in another. Deaths from cardiac disease are declining in many countries despite increased obesity, but that does not mean that obesity is unrelated to cardiac disease.

Hill misinterprets our review of cannabis use by people with psychosis (A. Kolliakou et al. Intl J. Dev. Neurosci. 29, 335–346; 2011). Contrary to his inference that this group self-medicates to mitigate negative symptoms, we found that the most commonly reported use by these individuals was purely recreational.

He suggests that cannabis is a risk factor only for those with a genetic predisposition to schizophrenia (but see M. Di Forti et al. Biol. Psychiatry 72, 811–816; 2012). Another explanation could be that some of the genes associated with a proclivity for cannabis smoking also show up among those who are predisposed to schizophrenia, because a genetic tendency for the habit could in turn increase the risk of schizophrenia.