Background: A number of psychopathological and neurobiological studies on affective temperament have been conducted based on the assumption that temperament is a stable trait. However, few studies have actually assessed the long-term stability of affective temperament. The objective of this study is to evaluate the 6-year stability of affective temperaments as measured by the Temperament Evaluation of Memphis, Pisa, Paris and San Diego – Autoquestionnaire version (TEMPS-A) in a non-clinical adult population. Sampling and Methods: Study participants consisted of 178 Japanese white-collar workers (103 males and 75 females; mean age = 38.5 years, SD = 7.8) who completed the Japanese version of TEMPS-A twice over a 6-year interval, and who did not have either past or current DSM-IV affective, anxiety or psychotic disorders, as diagnosed with the Mini-International Neuropsychiatric Interview. The long-term stability of affective temperaments as measured by TEMPS-A was assessed by analyzing Pearson correlation coefficients for temperament scores over a 6-year period. Results: Temperament scores were moderately to highly correlated over the 6-year period (depressive temperament, r = 0.59; cyclothymic temperament, r = 0.68; hyperthymic temperament, r = 0.82; irritable temperament, r = 0.66; anxious temperament, r = 0.74; p < 0.01 for all values). Pearson coefficients were in the range of 0.61–0.83 for males and 0.51–0.79 for females, while they were 0.56–0.85 for younger and 0.63–0.77 for older participants. All correlations were significant at p < 0.01, irrespective of temperament type, gender and age. Conclusions: Affective temperaments as measured by TEMPS-A exhibited good long-term stability and were robust, irrespective of temperament type, gender and age. Affective temperaments as measured by TEMPS-A may be considered to be stable traits, providing a sound basis for psychopathological and neurobiological studies. Limitations of this study include the fact that our sample was not drawn from the general community, it was entirely composed of Japanese participants and the size was not large.