Sir: Meinhofer et al. 1 analyzes administrative data from state medical marijuana laws and substance use treatment admissions in an ecological study design to show that medical marijuana laws are associated with state‐level increases in marijuana, alcohol and cocaine treatment admissions among pregnant women. While these findings are provocative, the utility of ecological analyses in assessing the effects of medical marijuana use is severely limited and this study's results could reflect bias, rather than an individual‐level effect. Ecological studies are susceptible to a bias commonly referred to as the ‘ecological fallacy’—the (often implied) assumption that correlations occurring at the population‐level parallel correlations at the individual level 2. This assumption is sometimes true, but not always; indeed, correlations that exist when analysing population‐level data may not exist or may even be reversed when analysing individual‐level data. Researchers must use individual‐level data to reliably discern the individual‐level effect of medical marijuana use, an individual‐level exposure.