The aim of the study was to examine associations of social support with early retirement and reported retirement preference. Logistic regression analyses of early retirement (retired before the age of 55) were based on a cohort of 10,489 respondents (5960 female, 4529 male) aged 40-55 years. Analyses of retirement preference (planning of early retirement) were based on a sub-cohort of 7759 full-time employees (4233 female, 3526 male). The measures for social support were the social network size (number of close personal relationships) and social network heterogeneity (number of different roles as sources of social support). Among women, early retirement was associated with small social networks (OR 5.1, 95% CI = 2.8-9.2) and low social network heterogeneity (OR 9.1, 95% CI = 3.2-25.9). Among men, the corresponding associations were somewhat smaller (1.5; 0.9-2.6 and 8.0; 1.8-35.7, respectively). Adjustments for age, occupational training, and spouse at home did not considerably affect these associations. Among men, the association between social network heterogeneity and early retirement was not statistically significant after additional adjustments had been made for marital status and perceived health status. The characteristics of the social network were not associated with a retirement preference among employed men. Among women, the lack of social relations was associated with low retirement preference. Our findings suggested that social support is not independently associated with a retirement preference among men. Socially isolated women prefer retirement less often than other women do. Being retired before the age of 55 years may be predicted by limited social support or it may restrain one's possibilities to establish and maintain social relationships.