The aim of this paper is to update the over 20-year-old normative data for the Benton Controlled Word Association (COWA) Test. In a sample of 360 normal volunteers, the age ranged between 16–70 years, and the educational level ranged from 7–22 years. Care was taken to ensure that the population was heterogenous, yet the two stratifications of gender, four age, and three educational groups led to 24 cells with 15 individuals in each. Test-retest reliability was established by testing 30% of the sample after a 6-month delay, which represents a typical follow up duration between testings in a clinical setting. The two forms of the COWA revealed significant test-retest reliability. Generally, our updated values fall above the original normative values, which were derived from a less well-educated and rural sample. No major gender or age trends were noted, but the COWA test performances were influenced by education, i.e., as the level of education increased, the performance on the COWA increased. The only gender differences that were found were for the women in the highest educational group (> 16 years), who performed significantly better that men in the highest educational group. An error analysis of repetitions or perseverations is provided, with cut-off scores according to age levels. Finally, the updated COWA norms are compared to the original norms as well as to other measures of word fluency.