Objectives: To examine the association of years spent in Japan during childhood with cognitive test performance in late life among Japanese American men, and to assess the influence of the language used for testing on this association.

Design: A cross-sectional study.

Setting/participants: A total of 3734 Japanese American men, aged 71-93 years, who were first- or second-generation migrants and living on Oahu Island, Hawaii.

Measurements: The outcome variable was cognitive test performance assessed using the Cognitive Abilities Screening Instrument (CASI), which was developed for cross-cultural studies of cognitive impairment. The explanatory variable of main interest was the number of years spent in Japan during school-age childhood years (ages 6-17). The associations of CASI scores with childhood years in Japan was evaluated using a stepwise multiple linear regression model in which a total of 40 potential confounders were included as covariates.

Results: In the total sample, there was an inverse association between CASI scores and middle childhood years in Japan. This association remained significant after controlling for age, education, socioeconomic status, traditional Japanese food consumption, pulmonary function, apolipoprotein E4, proficiency in speaking Japanese, and other possible confounders. When data were analyzed separately for subgroups according to the language preferred at testing (English or Japanese), associations between childhood years in Japan and CASI scores were in opposite directions negative for the group tested in English and positive for the group tested in Japanese. The interaction between the testing language and childhood years in Japan was statistically significant.

Conclusions: There was an inverse association between years spent in Japan during school-age years of childhood and cognitive test performance in late life. This association could not be accounted for by age, education, or other confounding factors. However, this finding was not observed in participants who preferred being tested in Japanese. To assess cognitive test performance in older people, it is of prime importance to use the most optimal language for testing, usually the subject's native language.