This paper revisits the issue of crop yield distributions using improved model specifications, estimation and testing procedures that address the methodological concerns raised in recent literature that could have invalidated previous conclusions of yield non-normality. It shows beyond reasonable doubt that some crop yield distributions are non-normal, kurtotic and right or left skewed, depending on the circumstances. A procedure to jointly estimate non-normal farm- and aggregate-level yield distributions with similar means but different variances is illustrated, and the consequences of incorrectly assuming yield normality are explored.