In Japan, many scholars and policymakers as well as ordinary people, have accepted the family nuclearization theory—that is, that the Japanese family system has changed from a traditional stem family into the modern nuclear or conjugal family in the latter half of the twentieth century. Although the number of nuclear family households doubled during the period of high economic growth (1955-1974), this can be brought about under the stem family principle by a marriage boom among non-heir sons and daughters born in the 1930s and the 1940s, who have many siblings due to the demographic transition. This paper examines continuity, change, and regionality in the Japanese family over the twentieth century, using retrospective longitudinal data from nationally representative survey implemented in 2002. The proportion of couples co-residing with the husband’s parent(s) or with the wife’s parent(s) at the time of marriage decreased from about 35% for those born in the 1930s to 20% for those born in the 1960s. However, the latter cohort started living with their parent(s) soon after marriage, and then over 30% of them co-reside 10–15 years after marriage, showing a delayed co-residence tendency. The results from a discrete-time logistic regression analysis reveal that the spread of conjugal family ideology, industrialization, and urbanization are the primary causes of the decrease in intergenerational co-residence at marriage. However, these effects start to weaken immediately after marriage and then fade out sooner or later, inconsistent with the nuclearization theory. By contrast, the intergenerational transfer of family property, especially house and/or land, and intergenerational reciprocity have powerful impacts from the time of marriage onward, never weakening with the passing of time. These features are clearer among younger cohorts than among their parental cohorts, suggesting intrinsic forces that form stem families continue to work in depth. Moreover, there is the persistence of the stem family in terms of regionality over the twentieth century, concluding that the Japanese family is still based on the stem family system. The evidence presented in this paper could provide several insights into the family systems of other strong family countries.