Overall, the reliability of the modern Binet is quite good. Data on the internal consistency as well as test-retest reliabilities are reported in the manual as a function of age range. Internal consistency estimates are based on the Kuder Richardson formula, KR20, which calculates the highest reliability from all posible half-splittings of the data. use this formula, the authors that all items below the basal were passed and all items above the ceiling were failed As the authors of the manual note, this procedure is not very conservative and most likely leads to slightly inflated (at best, the upper bounds of) reliability.
Table 10-4 shows the median reliability coefficients across age groups for individual tests, area scores, and the composite. Examination of the table re veals that the reliabilities of area scores and the composite are well above the 90 minimum that most experts believe a test should have if it is to be used to make decisions about individuals. Coefficients for individual tests, however, often fall below this minimum or are right on the borderline. Given that these coefficients represent the upper bounds of reliability, the need for caution in using individual test scores is evident. Test-retest reliability data were obtained from a sample of 57 5-year-olds and 55 8-year-olds. The average interval between the two testing was about