Does Null Hypothesis Testing Create a False Dichotomy?
Olga Eizner Favreau, Dpartement de psychologie, Universit de Montral, C.P. 6128, Succursale A, Montral, Qubec, CANADA H3C 3J7 In an ongoing debate about the value of doing tests for sex differences, those in favour claim that if sex differences exist, it is important to know about them. However, the null hypothesis (NH) tests that are used for inferring group differences can detect only mean differences and provide no information about how the differences are distributed across groups. Theoretical and empirical examples show how NH rejection can occur when only a small proportion of individuals differ from all others, demonstrating that these tests are incapable of supporting inferences to general group differences. This forces a reevaluation of sex difference research which has been interpreted as distinguishing males from females in general, particularly where inferences have been to general biological factors. However, even knowing the limitations of these tests may not lead to more
