Replicate subaqueous pebble counts show that measurement and selection bias can significantly affect sediment size distribution data. The standard deviation about replicate means increases linearly with sediment size and is independent of the sediment percentile (e.g., d50, d84, etc.). For samples collected by one person, sampling error varies from 2¿2.6 mm to 300¿35.4 mm within one standard deviation. For samples collected by different individuals, sampling bias increases the error from 2¿4.2 mm to 300¿63.8 mm. Whenever possible, only one individual should select and measure pebbles, which improves the ability of the pebble count to delineate trends between sites or over time (although the absolute size values may be in error due to bias). Bias can be reduced by rigorous training of individuals, by using templates to measure sediment size, and by using grids on the stream bottom to select particles. If two or more individuals collect samples, they should conduct replicate samples to provide a basis for later calibration of results. ¿ American Geophysical Union 1995.