It has frequently been shown that the never married, divorced and widowed have higher rates of self-reported morbidity than married people. The purpose of this study was to assess to what extent morbidity differences by marital status can be explained by differences in living arrangement (i.e. living with a partner or not). If living arrangement plays a major role, one expects to find that: 1. people who live with a partner have lower morbidity rates than those who live alone; 2. morbidity differences by marital status decrease substantially after controlling for living arrangement. Data from the GLOBE study, a large prospective cohort in the Netherlands, on 18,973 people from Eindhoven and its surroundings, aged 15-74, were used to test the hypotheses. The measures for morbidity were perceived general health, subjective health complaints, chronic conditions and work disability. Multiple logistic regression models were used to calculate odds ratios for morbidity by living arrangement (question 1) and for morbidity by marital status, firstly without control and subsequently with control for living arrangement (question 2). In all models we controlled for other sociodemographic variables (age, sex, educational level, degree of urbanization, religion and country of birth). Our analyses showed that people who live with a partner have lower morbidity rates than those who live alone. They further showed that the excess risks of the never married, widowed and divorced decreased by 40-70% for all health measures (except subjective health complaints) after controlling for living arrangement.(ABSTRACT TRUNCATED AT 250 WORDS)