This study identified some factors associated with teachers' knowledge and beliefs that are related to scoring mathematics constructed-response (CR) assessment tasks. Five groups of teachers (n = 274) who either had different teaching experiences or had different cultural beliefs about teaching and learning mathematics were selected to score 28 students' responses to seven CR math tasks. Among the 274 teachers, the first four groups (n=222) were selected from China. Group 1 was composed of pre-service elementary school teachers; group 2, pre-service secondary teachers; group 3, elementary in-service teachers; and group 4, secondary in-service teachers. The fifth group (n = 52) was composed of in-service middle school teachers from the United States. A number of analyses of variance (ANOVA) on teachers' scores of the 28 responses and subsets of the responses were conducted to examine the mean differences across different groups of teachers. Four factors were found to have significant relationships with the rating differences. They were teaching experience, experience with students at particular grade levels, the nature of students' responses, and beliefs in teaching and learning mathematics. The identification of the factors has implications both for promoting validity of test scores and for examining teachers' understanding of student learning targets.