The aim of this study was to test the reliability and validity of the COMFORT scale as a postoperative pain instrument for children aged 0-3 years. Subjects were 158 neonates and toddlers after major abdominal or thoracic surgery. Trained nurses rated the children's pain at 3, 6 and 9 h postoperative on the Pediatric Surgical Intensive Care Unit using the COMFORT and a VAS for pain. Interrater reliability of the COMFORT items proved to be good (Kappa 0.63-0.93) for all items with the exception of the item 'Respiratory response', which was moderate (Kappa 0.54). LISREL analyses showed that the structure of the COMFORT data was best represented by three latent variables: COMFORT 'behaviour' with loadings from the behavioural items (Alertness, Calmness, Respiratory response/Crying, Physical movement, Muscle tone and Facial tension) and separate latent variables for 'Heart rate baseline' (HR) and 'Mean arterial blood pressure baseline' (MAP). Factor loadings of the items were invariant across time, indicating stability of the structure. The latent variables COMFORT 'behaviour' and VAS pain were highly interrelated indicating congruent validity. Stability of COMFORT 'behaviour' and VAS pain was moderate which might be due to varying painful episodes in this sample. HR and MAP, although stable across time, were weakly related to VAS pain and COMFORT 'behaviour'. These findings support the use of the COMFORT 'behaviour' scale to assess postoperative pain in neonates and infants.