Statistics are presented to illustrate the influence of growing season and location on the accuracy of NIR testing of hard red spring wheat for protein. The influence of season was greater than that of growing location, and highly significant errors are likely to occur if protein is predicted in samples from one growing season using a calibration based on samples from another season. The same was true to a lesser extent if samples from one growing location were analysed using calibrations based on samples drawn from different areas. There was evidence that some growing areas were less susceptible than others to the seasonal influence, whereas certain seasons (e.g. 1976) exerted more influence than others. Errors in predicting protein were considerably greater than errors in predicting moisture. The development and use of universal constants incorporating samples from a wide range of location and several seasons effectively counteracted the influences of both season and location, and universal constants have been developed for use in the NIR instruments of leading manufacturers, for hard red spring and durum wheats, and barley. The concept and principle of 2·point calibration using normalized mathematical treatment of the log 1/R signal is discussed and its efficiency in providing a stable calibration for the prediction of protein in hard red spring wheat has been demonstrated.