Background. Evidence for the effects of marital status and marital status change on physical health is equivocal.

Method. Structural equation models examined whether marital status predicted physical health. Six groups were examined simultaneously: married (M); widowed (W); divorced (D); never married (NM); newly widowed (NW); and newly divorced (ND). There were four annual measures (T1–T4). Both NW and ND were married at T1, but had lost their partners by T2. Four physical health variables were examined: Problems, Limitations, Service use, and Self-rated health (SRH). Age and gender were included.

Results. Previous health predicted future for all measures and for all groups. However, the specific strengths and time-courses varied between marital status groups and between health measures. The most marked patterns were associated with marital status change. Service use was influenced most strongly by NW, whilst Limitations was influenced by ND. Problems distinguished NW and ND from stable marital status groups but also from each other. SRH was influenced by W and not by recent marital status change. The effects of age and gender were modest and restricted to specific health variables and specific marital statuses.

Conclusions. The results demonstrate that marital status and marital status change, in particular, influence health longitudinally. The impact of a change to divorced or to widowed status is not the same. No two health variables responded in the same way, suggesting that marital status has a differential effect on health.