Early temperament predicts various aspects of development. In large-scale studies, temperament is often assessed via parental report because naturalistic and structured observations are costly and bear the risk of subject loss. However, the validity of such parental reports has been disputed repeatedly. This article compared parental reports on Negative Affectivity and Orienting/Regulation with three naturalistic observations and a Still-Face Paradigm. It thereby investigated the role of situational similarity on the congruence of assessment approaches while controlling selected covariates. Results showed that whereas parental reports merely reflected naturalistic observations, they strongly predicted child behaviour in the SFP. Negative mood was highly volatile over the observed situations whereas motor activity showed some consistency. In sum, the results underline the importance of kind and intensity of stimulus for consistency and visibility of temperament and support the validity of parental reports on early temperament by demonstrating that these reports can predict observable child behaviour.