Background:Recent studies have suggested an association between prenatal paracetamol exposure and adverse neurodevelopmental outcomes in children. However, these findings may be confounded by unmeasured factors related to maternal use of paracetamol and child outcomes.Objective:To examine the association between duration and timing of prenatal paracetamol exposure on parent-reported communication skills, behaviour, and temperament in preschool-aged children, with focus on the role of unmeasured confounding.Methods:We used data from the Norwegian Mother and Child Cohort Study. Linear and generalised linear models with inverse probability weights and robust standard errors were used to quantify the association between prenatal paracetamol exposure and continuous and categorical outcomes.Results:Of the 32 934 children included in our study, 8374 (25.4%), 4961 (15.1%), and 1791 (5.4%) were prenatally exposed to paracetamol in one, two, and three trimesters, respectively. Children exposed to paracetamol in two trimesters scored lower on shyness compared with unexposed children (β −0.62, 95% confidence interval [CI] −1.05, −0.19). Children exposed to paracetamol in three trimesters had a moderate increased risk of internalising behaviour problems (relative risk (RR) 1.36, 95% CI 1.02, 1.80) and borderline externalising behaviour problems (RR 1.22, 95% CI 0.93, 1.60) compared with unexposed children. Children exposed to paracetamol in 2nd/3rd trimester scored lower on shyness (β −0.32, 95% CI −0.66, 0.02) compared with unexposed children. Sensitivity analyses indicated that unmeasured confounders play an important role and may potentially bias the effect estimates away from the null.Conclusions:Timing of exposure and short-term use of paracetamol during pregnancy do not seem to pose any substantial risk of the outcomes examined. Although we found an association between paracetamol use in multiple trimesters and lower shyness and greater internalising behaviour in preschool-aged children, we cannot rule out chance or unmeasured confounding as possible explanations for these findings.