Milk urea nitrogen (MUN) is one of the minor milk components (i.e. in low concentration) that can be determined by PLS models using milk mid-infrared (MIR) instruments. MUN, even though in such low concentrations in milk, is used in dairy herd improvement and payment testing laboratories to gain a more clear understanding of the efficiency of nitrogen utilization by ruminants for various purposes, such as protein feeding management and monitoring physiological consequences on the dairy cows (e.g. improved reproductive efficiency). In addition, MUN testing is used to help lower dairy farm environmental impacts due to excretion of urea in urine of dairy cows. Therefore, accurate MUN concentration information in a timely manner is of great importance for dairy herd feeding and reproduction management. Milk MIR instruments provide a method to measure MUN using PLS models. However, calibration samples and an accurate and reliable reference chemical method to obtain reference calibration values are needed for the measurement of MUN with milk MIR instruments. The objectives of the current study were to add reference chemistry for a milk urea nitrogen (MUN) parameter with a wide, well-distributed range of MUN concentrations to the modified milk calibration sample set made every four weeks at Cornell University, while maintaining orthogonality with variation in fat, protein, and lactose concentrations and to determine the within and between lab variation in an enzymatic spectrophotometric chemical reference method for MUN determination on the modified milk calibration samples and degree of uncertainty in MUN reference values. Changes in the modified milk samples’ formulation procedure were performed to include the desired range of MUN concentrations, particularly in the ultrafiltration processing step. An enzymatic spectrophotometric reference method was used to obtain MUN reference values for the set of samples formulated. The MUN enzymatic spectrophotometric method was also used to obtain MUN concentrations for a set of individual-farm bulk tank raw milk samples. The within (RSDr) and between lab variation (RSDR) in the reference values for MUN were < 1%, and the average expanded analytical uncertainty for the mean MUN value of the 14-sample calibration set was 16.15 mg/100 g milk +/- 0.09 mg/100 g milk. Another objective was to use the modified milk calibration samples to evaluate and improve the performance of PLS models for prediction of MUN. The orthogonal sample set was used to determine when a PLS model did not correctly model out the background variation in fat, protein or lactose. The calculation and application of an intercorrection factor to eliminate that effect was shown to improve the model performance (i.e., 50 % reduction in standard deviation of the differences between instrument predictions and reference chemistry values for MUN).