A study of hot-film anemometer measurements in water resulted in a hypothesis that: Dirt and air bubbles accumulating on the hot-film sensor change the mean voltage for a given velocity, but have only a minor, outside the frequency domain in water, effect on the frequency response of the sensor to velocity fluctuations. And for a given sensor there is a unique family of voltage/velocity relations that can be defined by calibration with different overheat ratios. The hypothesis was experimentally verified by comparing turbulence measurements made in the same flow field with well filtered and with very dirty water. The method was used to measure the characteristics of turbulence measurements made in ordinary laboratory water for flow over hydraulically-rough and hydraulically-smooth boundaries. Reynolds numbers ranged from 8,000 to 20,000 and mean velocities from 0.3 fps to 2.8 fps. Relative turbulent intensities for flow over the smooth boundary were in good agreement with Laufer's measurements in air. Power spectrum analysis of the turbulence showed very little energy in frequencies larger than 100 Hz.