This project was designed to study the process of parents teaching preschool age children using a direct observational method. Six mothers and their own three-year-old sons from seven ethnic groups participated: middle-class Anglo, lower-class Anglo, English-speaking Mexican-American, bilingual Mexican-American; Spanish-Speaking Mexican-American, English-Speaking Chinese-American, Chinese-speaking Chinese-American. In addition, four groups of the mothers taught a second child from their ethnic and social class background, and a third child from an ethnic and social class different from their own. Each mother taught a cognitive sorting game and a motor skill game to each of the children assigned to her. Data were organized in terms of "own child" data and analyzed by means of a one-way analysis of variance, and predicted trends; the "social distance" data were analyzed in a 3 x 2 x 2 mixed factorial design. The data were coded usin g a Parent Interaction Code which allowed analysis of the teaching interaction in terms of programmatic variables: total time, input and pacing; and teaching loop variables: alert, format, child response, and feedback. The single best predictor of maternal teaching, or child response, was ethnicity with some replication of earlier social-class differences. The results tend to support stable constellations of behaviors within ethnic groups of a subtle nature. Further research is recommended. (Author/CK)