Abstract: The use of academic indices consisting solely of grades and test scores to determine admission to selective undergraduate institutions has resulted in an underrepresentation of students from low socioeconomic backgrounds and from certain racial and ethnic groups at these schools, notably African American and Chicano/Latino students. Policymakers and advocates who believe that talent and potential are distributed more randomly throughout the population often argue for broadening admissions criteria to value a greater range of applicants' activities and accomplishments. This study examines whether a definition of merit in undergraduate admissions that is broader than grades and test scores but still based on the goal of prediction could result in greater representation of these groups. At the University of California Berkeley, the undergraduate admission process that weighs multiple aspects of the high school record is called comprehensive review. Using ordinary least squares and logistic regression to analyze data from the 1999 enrolled and the 2004 applicant cohort to the University of California Berkeley, this study seeks to identify pre-college student characteristics that contribute to subsequent success in college. The college outcomes examined are initial GPA, change in GPA, degree completion, academic engagement, and campus leadership. The study also examines how the use of admissions formulas designed to optimize some of these outcomes would affect the demographic composition of a cohort of admitted students. Finally, it assesses the predictive power of these alternative admissions formulas, comparing the proportion of variance in first-year college GPA that they explain to that of the unitary comprehensive review score assigned by readers. This analysis suggests that there are aspects of students' high school records, beyond grades and test scores, that predict success in college. The way that these factors are combined affects the demographic representation of the admitted cohort; while none achieves demographic representation that mirrors California's population, the comprehensive review score and the leadership optimization formula come the closest. Except for the leadership formula, optimization formulas predict first-year college GPA better than the comprehensive review score does, suggesting that there may be a trade-off between prediction and representation in undergraduate admissions.