The purpose of this study was to compare the accuracy of capillary and venous reagent strip tests (RST) against a reference plasma glucose method, and to assess the impact of haematocrit. One hundred and eighty infants admitted to the Neonatal Unit had blood glucose measured by reagent strip tests using capillary and venous blood samples. Venous plasma glucose was assayed by the Hexokinase method. Each infant had a venous haematocrit performed in the Neonatal Unit. Comparable inaccuracies were noted with both capillary and venous reagent strip tests at all levels. The mean difference between capillary RST and plasma glucose was 0.058 mmol/l (S.D. 1.39). The corresponding mean venous RST plasma glucose difference was 0.138 mmol/l (S.D. 0.96). The two means were statistically different from each other (P = 0.024), but this difference disappeared if the comparison was made only in babies with a PCV of 35-55%. At higher haematocrits (PCV >55%, N= 96) the mean difference between venous RST and plasma glucose was significantly more than the mean difference between capillary RST and plasma glucose (0.018 versus 0.295. P = 0.002). Hence the higher the haematocrit the more inaccurate the venous RST. This study confirms the limited value of reagent strip tests in the assessment of blood glucose in the neonatal period. It suggests that venous RST may be more inaccurate in comparison to capillary and that high haematocrits have a greater effect on venous RST than capillary RST.