The relationship between television program variables and children's attention was examined using sample shows from two nationally distributed educational programs. Shows were coded at 30-second intervals on four measures of program complexity and two measures of the relationship between verbal and visual content. Children's visual-attention levels were measured individually and averaged over the sample. Single time series analysis indicated the predictability of children's attention levels to be no more than 30 seconds. Younger children's attention in adjacent program segments was found to be more highly interdependent than was older children's attention. Multiple time series anlaysis indicated the attention of older children could be explained very well, based on the program variables measured. Older children also appeared to be affected more by variations in program complexity and, in particular, by visual complexity. Younger children were more affected by verbal variables, and attention levels were found to be more due to children's consistently attending than to program variables. Forty-four percent to 58% of the variance in attention was explained by these models. (Author)