Accurate information on behavior of young children at home is crucial to the study of child development. The present study compared parent diaries of 5-year-old children's time spent with television to concurrent automated time-lapse video observations. In addition, a number of control groups were employed to assess the effects of observational equipment in the homes. The sample consisted of 334 mostly white middle-class families, of whom 106 had observational equipment installed. Results indicated no systematic subject selection of families who were willing to have observational equipment as compared to the controls. In addition, there were no differences in reported viewing behavior between the observed families and controls. Of 3 types of parent estimates of 5-year-old TV viewing, concurrent diaries correlated best with video observation (r = .84) and produced a very small absolute mean time error. Direct parent estimates of typical time spent viewing produced smaller correlations and large overestimates as compared with diaries.