A controlled experiment was conducted in 1980 with 117 first graders in two Southwestern schools near the Mexican border to determine the effects of direct instruction in teaching bilingual Hispanic children to read. Direct instruction has been defined as using modeling, reinforcement, prompting, discrimination learning, and correction/feedback for positive self concept development. Two rural schools with high percentages of Chicano students and with established bilingual education programs were selected for the study; however, School A served a poor neighborhood and was judged to be disadvantaged while School B served a more affluent area. In each school 58 first grade Hispanic bilingual children were randomly selected and randomly assigned to two experimental groups, conditions for which were identical except that one group used direct instruction and the other group used the regular bilingual method. Significant effects resulted between schools, between groups, and within a group/school interaction. Tentatively, direct instruction can significantly improve beginning bilingual children's achievement more than regular bilingual instruction; bilingual education may be enhanced by incorporating direct instruction into its teaching method. "School characteristics" may interact with the effects of any specific teaching method. (AN)