An alternative method of identifying effective schools, based on the concept of effective schools as statistically atypical, was tested. This paper investigates the issue of consistency of the identification of what is termed "exceptional" schools. The following tests were administered to all first through fourth graders in 431 schools in South Carolina: (1) South Carolina Basic Skills Assessment Program, given to all first, second and third graders; (2) Comprehensive Tests of Basic Skills, given to all fourth graders; and (3) Cognitive Skills Assessment Battery, given to all entering first graders. Data were collected during 1984-85 and 1985-86. Using studentized residuals for reading, mathematics, and reading and mathematics combined, this study tested for both cross-grade and cross-year consistency of scores. The strongest consistencies were found with combined reading-mathematics scores, at each grade level within schools. Because this combined score appeared to be consistent within grades and across years, it was suggested that it might be a more accurate statistic for predicting which schools are effective (positively exceptional) and which are ineffective (negatively exceptional). (JGL)