After annealed, type K (Chromel vs Alumel) thermocouples were heated above 200 °C, their temperature measurements were in error up to 1.3%, as determined by comparison calibrations to working standard 90% Pt–10% Rh/Pt thermocouples or platinum resistance thermometers. Reannealing the type K thermocouples removed the errors. The errors were due to changes in the thermal emf vs temperature relationship of the type K thermocouples, which from previous work of others can be attributed to short‐ranged ordering of the Chromel thermoelements. Reannealing the thermocouples removed the errors because the order–disorder transformation is reversible; that is, short‐ranged ordering of the Ni and Cr atoms of the Chromel alloy occurs between 200° and 600 °C, and disordering occurs above 600 °C. The traveling gradient method was used to determine the effects of heat treatment on the thermal emf of type K thermocouples, to investigate the kinetics of ordering of Chromel, and to determine the amount of order produced by heat treatments. Though the order–disorder transformation could not be stopped, our results demonstrate that there exists an optimum amount of order in the Chromel thermoelement to yield a repeatable thermal emf vs temperature relationship for a type K thermocouple in a specific application. When the experimental conditions of the application were controlled carefully, isothermal heat treatment to produce the optimum order in the Chromel thermoelements and calibration of the type K thermocouples prior to use essentially eliminated the temperature measurement errors due to order. The optimum amount of order depended on the application and was determined experimentally. Variations from the optimized heating or cooling cycles, changes in the depth of thermocouple immersion during use, or modification of the temperature gradient on the thermocouple caused temperature measurement errors of up to 1% with ordered and calibrated type K thermocouples.