A hysteresis effect in multiplier phototube noise as a function of tube gain has been found for both RCa 5819 and DuMont 6292 tubes. The effect is a function of both temperature and the length of time that a tube has ``rested.'' The hysteresis is seen to be a noise buildup with time but not a change in dynode multiplication. The reasons for this effect are not clear. Several possible mechanisms are discussed. Conditions are determined for obtaining reproducible noise counts for liquid scintillation counting of low‐energy beta emitters.