Longitudinal decline in IQ among fragile X males was reported recently. However, there are problems in retesting IQ that may affect scores. Two such factors are intertest time interval and score obtained on the first test. To determine the generality of IQ score changes, we examined 101 fragile X males from 6 centers. To ensure high test-retest reliability, only results from Stanford-Binet and Wechsler tests were used. Thus there were retest scores from 60 subjects. Test-retest reliability between first and last scores was very good (r = 0.85) and comparable to those seen in nonfragile X mentally retarded individuals. Also computed were z-scores of differences in IQ scores. The z-score differences were distributed about a mean at 1 SD below the expected zero value. Eighteen subjects showed statistically significant decreases in IQ, 6 showed statistically significant increases, while 5 showed the same scores. Z-score differences were not correlated with type of residence or elapsed intertest interval, but were negatively correlated with first score obtained, indicating a regression-to-the-mean effect. Using a multiple regression analysis, we found first score obtained, age tested, and age retested significant predictors of score differences, accounting for 19% of the total variance. These results suggest that factors previously identified as affecting retest scores have a smaller effect than originally thought. It is suspected that decline in IQ is associated with dynamic neurological processes and needs to be investigated further.