The purpose of this investigation was to identify the differences in classroom structure, children’s interaction type, teacher role, and child behavior in 4 different early education program models and to determine the effectiveness of these models on children’s development over time. For the present study 4 urban kindergarten classes representing four different models were selected. They were: Montessori program (Model Ⅰ), Open Education program (Model Ⅱ), Unit Based-Child Centered Traditional program (Model Ⅲ), and Academic Drills-Oriented program (Model Ⅳ). The subjects for the study were 121 children enrolled in those classes. The data for the present study were collected through time-sampled observation of individual children over 17-week period, and pre- and post-test of children on IQ, socio-emotional development rating, school readiness, and ceativity. ANOVA was used to analyze group differences in pre- and post-test and observational data, and the stepwise multiple regression technique was applied to determine the relative effectiveness of independent variables contributing to children’s growth. The major findings of the study were as follows: 1. There were significant differences among the 4 models in classroom structure (p＜01). Model Ⅳ indica ted the highest structure among the four, while Model I was the lowest in structure. 2. There were significant differences among the four models in interaction type (p＜01). The childchild interaction was most frequently observed in Model Ⅱ, child-teacher interaction in Model Ⅲ, and child-material interaction in Model Ⅲ and Model In term of teacher role, teachers in Model Ⅰ and Ⅱ most frequently used direct teaching behavior with open-ended questions, while teachers in Model Ⅳ used controlling and punishment behavior most often. Children’s initating behavior was most frequently observed in children in Model Ⅱ, active participation in Model Ⅲ, and passivity in Model Ⅳ. 3. There were no significant differences among the four models in cognitive gains measured by the intelligence test. The regression analysis indicated that the pre-test IQ contributed most (33%) among the independent variable in predicting post-test IQ. 4. There were significant differences among the four mode is in socio-emotional development rating gains (p＜.01); Model Ⅲ made the most significant gains. It was also found from the regression analysis that pre-test ratings on socio-emotional development explained about 25% of the variance of the post-test ratings, and child-child interaction added about 12% of the variance. 5. There were significant differences among the four models in post-test scores on school readiness (P＜.01). Models I and Ⅳ indicated high averages, while Model Ⅱ was the lowest. It was found that the pre-test IQ, selective activity, and pre-test socio-emotional ratings explained approximately 25% of the variance of the school readiness scores. 6. There were significant differences among the four models in post-test scores on creativity (p＜.01). Model Ⅱ indicated the highest average, and Models Ⅰ and Ⅳ indicated low averages. It was found that selective activity and interaction type explained about 22% of the variance of children’s creativity.