The aim of this study was to compare predictions of hyperbilirubinaemia by eye, performed by trained physicians and nurses, with predictions obtained using two commercial bilirubinometers. Jaundice was assessed in 92 white and 48 non-white healthy full-term neonates using three non-invasive methods and by total serum bilirubin as the reference method. Clinical assessment of cephalocaudal progression of jaundice was carried out independently by a physician and by nurses. Simultaneously, the Minolta Airshields JM-102 was applied on the sternum, the BiliCheck on both the forehead and the sternum, and finally, serum bilirubin concentrations were determined. The Minolta JM-102 showed the best performance with r 2 =0.90, an intraclass correlation coefficient (ICC) of 0.93, and a 95% confidence interval (CI) of ±4 units (approx. 56 lmol/l). The BiliCheck performed slightly better on the forehead than over the sternum with r 2 =0.90, an ICC of 0.88, and a CI of ±62 lmol/l. Assessment of jaundice by eye was least accurate with r 2 =0.74, an ICC of 0.67, and a CI of ±1.5 zones (corresponding to ±75 lmol/l). Skin pigmentation and ambient light both adversely affected noninvasive bilirubin estimation. Conclusion: All three non-invasive methods are well suited for estimation of serum bilirubin but show large confidence intervals. In healthy term newborns, hyperbilirubinaemia (>250 lmol/l) can be safely ruled out by eye if jaundice does not reach the abdomen or the extremities (Kramer zones 1 and 2), with <22 units (<230 lmol/l) for the Minolta JM-102, or with a cut-off of 190 lmol/l for the BiliCheck. If these respective thresholds are exceeded, serum bilirubin concentrations should be measured.