This study examines the construct validity of a new teacher interview designed to assess teachers' pedagogical content knowledge (PCK) for preschool mathematics. PCK describes the subject matter knowledge a teacher needs for effective teaching. Data from 22 teachers and 113 Head Start children in a large midwestern city in the United States were used to test predictive relationships between the PCK interview and two dependent variables: good preschool math teaching practices and improved child learning outcomes. Using a method from previous published work, frequency of math-related language is assessed as a proxy for good preschool math teaching. Changes in children's mathematical achievement were measured from fall to spring within a single school year using the Test of Early Mathematics Ability (3rd ed.). Analysis by hierarchical linear modeling found significant positive relationships between scores on the new measure and both variables, suggesting that the PCK interview adequately represents the knowledge needed for effective teaching of preschool mathematics. A theoretical framework for the interview's construction is provided. (Contains 2 tables and 1 figure.)