Using longitudinal data on the Boston Naming Test ( Kaplan, Goodglass, & Weintraub, 1983) collected over 20 years from healthy individuals aged 30 to 94, we examined change in lexical retrieval with age, gender, education, and their interactions. We compared results between random-effects longitudinal and traditional cross-sectional models. Random-effects modeling revealed significant linear and quadratic change in lexical retrieval with age; it also showed a Gender x Education interaction, indicating poorest performance for women with less education. Cross-sectional analyses produced greater estimates of change with age than did longitudinal analyses.