Paper-pencil tests and classroom work-samples were used to evaluate the effectiveness of three learning units for teachers in early childhood education. Half of the sample had BA degrees or above, and half high school diplomas or less. Both methods assessed the same four competencies: quality of classroom arrangement, personalization of a literature experience, awareness of sense perception uses of materials, and diversity of teaching strategies which develop children's sense perception. Stability coefficients for the paper-pencil measures ranged from .29 to .83; generalizability coefficients for the work-sample measures ranged from .67 to .98. In general, validity coefficients increased from pre to post for the high school educated teachers, but decreased for the college educated teachers. The validity coefficients were higher than the average correlations between different competencies assessed by different methods. This suggested some degree of discriminant validity of the competency measures. However, the mean correlations between different competencies assessed by the same method suggested that at pre-test, the paper-pencil test might be the more appropriate method for the college educated teachers, and the work-sample method might be more appropriate for the high school educated teachers. After training, discriminant validity improved for both groups. (Author)