Theoretical considerations suggest that the temperature distribution within a meter of a cold surface may be partially stratified as a result of radiative and turbulent processes. Even though stratification would be of crucial importance in certain boundary-layer processes, it probably would be undetectable by conventional thermometry. Observations using direct refractive thermometry have been used to determine with great accuracy the temperature distribution above a cold water surface. These observations reveal a persistent temperature anomaly about 10 centimeters above the surface, which is consistent with theory.