在古典線性模式下 (X= T +ε)，本文就信度及分數分佈互之統計層面，討論77學年度大學聯招測驗各科分數的統計量。Spearman-Brown及α公式估計各科測驗信度，而Kolmorov-Smirnov (Lilliefors Test) 單樣本檢驗及Box-Cox轉換公式分別查證測驗分數及其轉換後是否接近常態。本文發現只有生物信度指標符合0.90之標準，而以自然組及社會組教學之信度指標最低，分別為0.52及0.43。各科測驗分數皆以不同程度偏離常態分佈，如二組數學為正偏態，顯示測驗對二組考生皆偏離。

This study examines score data from the 77th Annual College Entrance Examinations with respect to test reliability and test scores distribution under the classical linear model (X = T +ε) The Spearman­Brown formula and the Cronbachαare employed to estimate the reliability coefficients of the ten subject test, while the Lilliefors’, version of the Kolmogorov-Smirnov test is used to test the normality of the test scores and the corresponding Box-Cox transformed scores. The results of this study show that (i) except for the biology test, the reliability coefficients of the other nine tests are unable to meet the criterion of 0.90; (ii) the score distributions of ten tests are all deviated from the normal distribution. The implication of the results suggests that the improvement of test construction for ten tests be needed and that fur­ther research on the fit of the classical linear model be investigated.