Admissions practices at the North Carolina School of Science and Mathematics, a residential high school for gifted students, were examined. Data from students applying in 1983 and 1984 (over 800 per year) were available, as were grades of students admitted in 1983. A series of multiple regression studies showed that selection and first year grades could be accurately predicted. This finding suggested that statistical methods of selection could logically replace at least a portion of a more labor intensive clinical method. Additional analyses showed that students admitted in each year significantly outperformed other applicants on all objective measures. Recommended is a two-stage approach to selection with 260 semi-finalists being selected strictly on a statistical basis and the remaining 140 semifinalists selected by school officials. Semifinalists would then be interviewed by admissions committees and final selection be based partly on committee recommendations and partly on the discretion of school officials. Other recommendations involved more detailed definitions of admissions standards and expanded studies of school outcomes. (DB)