Many professionals, from auditors, venture capitalists, and lawyers, to clinical psychologists and journal editors, divide continuous flows of judgments into subsets. College admissions interviewers, for instance, evaluate but a handful of applicants a day. We conjectured that in such situations, individuals engage in narrow bracketing, assessing each subset in isolation and then--for any given subset--avoiding much deviation from the expected overall distribution of judgments. For instance, an interviewer who has already highly recommended three applicants on a given day may be reluctant to do the same for a fourth applicant. Data from more than 9000 M.B.A. interviews supported this prediction. Auxiliary analyses suggest that contrast effects and nonrandom scheduling of interviews are unlikely alternative explanations of the observed pattern of results.