We use multilevel modeling of ECLS-K data (a nationally representative sample of American kindergartners) to describe the process and content of kindergarten mathematics instruction, as well as the associations of such instruction with achievement gaps by social class and race/ethnicity. Where instructional effectiveness is concerned, time spent on two of the process characteristics--traditional math and group/interactive activities--was significantly and positively associated with achievement gains. Time spent on three of the content variables--advanced counting, practical math, and single-digit operations--was associated with increased achievement. Time spent on basic numbers/shapes significantly decreased achievement. Classes with a high percentage of African American students were particularly likely to receive full-day kindergarten, which increased total instructional time, and this may have modestly decreased the achievement growth gap for these classes. Overall, kindergarten instructional practices were found to modestly reduce the mathematics achievement growth gap of African American students, but have no significant effects on the achievement growth gaps of lower social class or Hispanic students. (Contains 5 tables, 2 figures, and 1 note.)