The rates of change in the language status of children with language impairment unaccompanied by other developmental or sensory disorders (primary language disorder) were studied in a longitudinal sample of 196 children who were followed from kindergarten through 4th grade. Previous studies have shown that children with such language impairments have moderate rates of improvement during this age range. Also, those with the most specific deficits have the greatest likelihood of improvement. Cole and colleagues have hypothesized that such results could be due to the effect of regression to the mean (K. Cole, I. Schwartz, A. Notari, P. Dale, & P. Mills, 1995). This study used a baseline measure of language that was independent of the measure used for diagnosis in order to control for factors leading to regression to the mean. Patterns of change using the kindergarten diagnostic measure were compared to those using the baseline measure. Rates of diagnostic change between kindergarten and subsequent observation intervals showed patterns of change similar to those of past research. Comparisons using the baseline measure revealed no significant change in relative language status across the 4-year time period. The results showed that when the conditions for regression to the mean were controlled, the poor language of children with language impairments was very likely to persist during the primary school years.