Census estimates of bird populations provide an essential framework for a host of research and management questions. However, with some exceptions, the reliability of numerical estimates and the factors influencing them have received insufficient attention. Independent of the problems associated with habitat type, weather conditions, cryptic coloration, ete., estimates may vary widely due only to intrinsic differences in observers? abilities to estimate numbers. Lessons learned in the field of perceptual psychology may be usefully applied to 'real world' problems in field ornithology. Based largely on dot discrimination tests in the laboratory, it was found that numerical abundance, density of objects, spatial configuration, color, background, and other variables influence individual accuracy in estimating numbers. The primary purpose of the present experiment was to assess the effects of observer, prior experience, and numerical range on accuracy in estimating numbers of waterfowl from black-and-white photographs. By using photographs of animals rather than black dots, I felt the results could be applied more meaningfully to field situations. Further, reinforcement was provided throughout some experiments to examine the influence of training on accuracy.