Measuring cortisol in saliva offers important advantages compared to measurement in plasma or serum. However, the sampling procedure and also the detection limit cause problems, especially in paediatric and neonatal care. We describe a simple and efficient sampling procedure, together with a modification of a radioimmunoassay, which enables analysis of low (down to 1 nmol/L) concentrations of salivary cortisol (10 times lower detection limit than in the original procedure). This setting was used in studying salivary cortisol concentrations before and after heelstick on healthy newborn infants. A significant rise (median 81%, p <0.01) in salivary cortisol as response to this invasive stressor was noted.