The longitudinal effects of age on confrontation naming using the 60-item Boston Naming Test (BNT) were studied in 541 "normal" elderly (ages 50-99). For participants with at least 4 annual assessments (n = 238), 150 were followed for > or =6 years, 81 for > or =8 years, and 43 for > or =10 years. A small practice effect (0.21 words, p = 0.06) and moderately high test-retest reliability were found when comparing the first 2 assessments, which were 9-15 months apart (r = 0.76, n = 353). Reliable change index scores indicated that an annual decline of > or =4 points on the BNT is needed for a statistically reliable decline in an individual. A gradient in the mean annual rate of change on the BNT was found with improvement in the 50s age group, no change in the 60s age group, and decline in the 70s and 80s age groups. When projected over 10 years, the magnitudes of the mean changes were relatively small, that is, a 1-word improvement for participants in their 50s and a 1.3-word decline for participants in their 70s. These findings demonstrate that lexical retrieval as measured by a visual object confrontation naming task is generally well preserved in aging with only subtle decline in the 7th and 8th decades of age.