Research sought to develop a quantitative model for the prediction of learning effects of a bilingual children's educational television program. The stimulus program series consisted of 30 sessions of Carrascolendas, a Spanish-English series which was based on specified behavioral objectives and aimed at students through grade 2. Learning effects were measured with criterion referenced tests administered to 408 Mexican-American children. Data were collected on the independent variables of communication stimuli, individual characteristics, and contextual traits; the predictive models were based upon linear multiple regression analysis. Results showed that learning effects can be predicted with relatively high degrees of reliability and accuracy. From this it was concluded that policy decisions regarding educational television must account for contextual characteristics, as well as program content and objectives, and that a communications stimulus, such as a television series, functions as one of many stimuli in producing learning effects. More detailed measures of these independent variables are needed and contracts for the evaluation of program series should require that measurement involve linear models. (LB)