The project has 2 phases. Each teacher participant will be assigned one of their students, between the ages of 3 and 6, as the “child of focus” for the 10-week training period. Parents of selected child will be the observers in the first phase of training period. In the second phase, the roles of teachers and parents will be swapped.

In Phase 1 (July 2011–January 2012), 28 teachers will be randomly selected to receive the 10-week filial play training in 7 kindergartens. In each kindergarten, 4 selected teachers will be required to give filial play training to 10 children with either internalizing or externalizing problems. The selected teachers will be trained to be the filial play trainer in their kindergartens. Parents with selected child will act as the observer in this phase. They will be asked to observe their children’s performance in the play sessions through the videos taken in each teacher-training session.

In Phase 2 (February 2012–June 2012), 70 parent participants in 10 kindergartens will then be trained to be the filial play parent trainer receiving the 10-week filial play training. Twenty-eight teachers will become the observers in Phase 2. All trainings will also be offered in ten kindergartens after school hours.

Due to the small sample size, usually less than 100 cases, data screening is considered to examine the normality of the data, linearity among variables, and homogeneity of variance (Tabachnick & Fidell, 2007).

An analysis of covariance (ANCOVA) is conducted to test the significance of the differences between the experimental and the control groups on the adjusted post-test means for each hypothesis. In each case the post-test specified in each of the hypotheses is used as the dependent variable, and the pre-test as used as the covariant. ANCOVA is used to adjust the group means on the post-test on the basis of the pre-test, thereby statistically equating the control and experimental groups. Significance of difference between means was tested at the .05 level. On the basis of the ANCOVA, the hypotheses are either retained or rejected. Effect size is measured by eta-squared. The appropriateness of the use of covariance is determined by ensuring that no significant correlation exists among the dependent measures (Stevens, 1992; Dancy & Reidy, 2004). The assumption of homogeneity of regression for the ANCOVAs is also checked (Dancy & Reidy, 2004).

The following strategies are built on the successful experience and good practices of the experimental study done in June 2010. The concept and operation of the 10-week filial play training model is based on Landreth and Bratton’s Child–parent Relationship Therapy (CPRT) in 2006. The training model has been successfully implemented in enhancing teacher–child relationship and reducing children’s externalization or internalization of problems. The proposed project will extend the training to parents. The proposed project also leads to parent–school cooperation in enhancing children’s social competency.