The aetiology of childhood acute myeloid leukaemia (AML) is largely unknown. Maternal marijuana use just before, or during pregnancy has been previously associated with childhood AML. This case-control investigation formally tested the hypothesis that parental marijuana use increases the risk of childhood AML in offspring. Incident cases of AML <18 years of age, diagnosed between 1989 and 1993, and registered with the Children's Cancer Group (a paediatric clinical co-operative group), were eligible for inclusion. Control children were selected via random digit dialling and individually matched 1:1 to cases on age, race and residential location, except for rare morphological subtypes that were matched 1:2. Parental telephone interviews were conducted to determine exposure and covariate information. Conditional logistic regression was used to estimate matched odds ratios (OR) and 95% confidence intervals [CI] adjusted for household income, parental education and parental age. The analysis included 517 cases and 610 matched controls. A series of sensitivity analyses examined the potential for recall bias. Ever lifetime use of marijuana by mothers was not associated with childhood AML [OR = 0.89; 95% CI = 0.66, 1.19]. Maternal marijuana use any time during the 3 months before, or during pregnancy was inversely associated with childhood AML [OR = 0.43; 95% CI = 0.23, 0.80]. Paternal use during the same time period was not associated with risk. Assuming a large degree of differential exposure misclassification was present, the corrected ORs ranged between 0.82 and 1.40. The previously reported positive association between maternal marijuana use before or during pregnancy and childhood AML was not confirmed. The decreased ORs observed in this study may be due to recall bias assuming plausibly low values of sensitivity.