The first, hidden stage of the Supreme Court’s agenda-setting process is the formation of the “discuss list,” the small set of cases actually considered in conference. Yet few have systematically considered the operation of and the influences on this critical initial phase of decision making. No systematic, empirical work makes comparisons over time of how features of cases shape the makeup of the chief justice’s discuss list. Here, we examine the composition of the discuss list through a multivariate analysis of all paid petitions for certiorari filed in October Term 1939, 1968, and 1982. We are thereby able to compare the tendencies and efficacy of three long-serving chief justices—Hughes, Warren, and Burger—in making up the discuss list. And, methodologically, we present an alternative to the “observed-value” and the “representative-case” methods of calculating effect sizes for second differences, with software to implement our proposal.