Correction to: Nature Neuroscience https://www.nature.com/articles/s41593-018-0206-1, published online 27 August 2018.

In the version of this article initially published, several occurrences of the word ‘schizophrenia’ should have more accurately been worded as ‘liability to schizophrenia’ or ‘schizophrenia risk’, including in the title. Specifically, in the dataset used for the genome-wide association meta-analysis of cannabis use, only a small proportion of the participants will actually have been diagnosed with schizophrenia. Therefore, it is more precise to say that the analyses reported in the paper show that the ‘liability to schizophrenia’ (rather than schizophrenia itself) leads to cannabis use. Additionally, there was a typo in the original report. The genetic correlation is very marginally higher and more significant. Congruent with Supplementary Table 6, the observed correlation is rg = 0.25 (s.e. = 0.03, P = 5.81 × 10–15).