The Boston Naming Test is commonly viewed as a measure of language ability, particularly, confrontational naming. Its utility in detecting word-retrieval problems in clinical populations is well documented. However, studies which would explore information-processing mechanisms involved in BNT performance are not available. Results of our repeated testing of 122 subjects between the ages of 57 and 85 with the 60-item version of BNT and other measures over three annual probes revealed high stability of the BNT scores over time, which suggests a lack of the practice effect, whereas cross-sectional analysis demonstrated some decline in the BNT scores in subjects over 70 years of age. In spite of the high stability in BNT scores on repeated testing, the pattern of correlations between the BNT and measures tapping different cognitive domains shifted over time. Results suggest predominantly the verbal mode of information processing in BNT performance on the first probe, as opposed to visuo-spatial mode on the third probe in our sample of elderly individuals.