Infrared (IR) analysis is widely used for routine analysis of cow milk in dairies. The aim of this study was to evaluate the precision and accuracy of an IR analyzer (Milko-scan 104) for measuring protein, fat, carbohydrate, and, indirectly, the energy content of human milk. The results of the IR analysis were compared with those of the following reference methods: Protein∼ Kjel-dahl (nitrogen minus nonprotein nitrogen); fat∼Roese Gottlieb; carbohydrate∼lactose enzymatic assay; energy∼bomb calorimetry. The precision (repeatability coefficient of variation) of the IR results was high for all four components: protein 0.4%, fat 1.0%, carbohydrate 0.2%, and energy 0.1%. There was a close linear covariation between IR results and reference results. [Protein content was determined with an error (SD) of 0.01 g/100 ml and fat with an error of 0.03 g/100 ml.] The covariation between IR carbohydrate results and the results of the lactose assay was poor, probably because the oligosac-charides in the milk were included in the results from the IR analysis and not in the results from the lactose assay. IR analysis is a valuable method in research, especially in epidemiological surveys, in which large numbers of samples are analyzed, and for continuous monitoring of the nutritional value of human milk in milk banking programs.