The findings are based on national census and death certification data, involving almost 67,000 adults in the USA between 1989 and 1997.
In 1989, almost one in two of the sample were married, and almost one in 10 were widowed. Around 12% were divorced and 3% were separated. Of the remainder, 5% were cohabiting, and one in five had never been married.
Unsurprisingly, older age and poor health were the strongest predictors of death by 1997, but a surviving marriage was also strongly associated with a longer life.
After taking into account age, state of health, and several other factors likely to influence the findings, those who had been widowed were almost 40% more likely to die between 1989 and 1997. Those who had been divorced or separated were 27% more likely to have done so.
But those who had never been married were 58% more likely to have died during this period than their peers who were married and living with their spouse in 1989.
The never married "penalty" was larger for those in very good or excellent health, and smallest for those in poor health, and it was greater among men than women.
For the younger age group, the primary causes of death among those who had never married were infectious disease and "external" factors. Among the middle aged and elderly, the main causes were cardiovascular and chronic diseases.
Never married men were more vulnerable than their never married female counterparts, and never married men between the ages of 19 and 44 were more than twice as likely to die as their married male peers of the same age.
"Risky" behaviours could not explain the differences, say the authors, because the unmarried group were only slightly more likely to smoke than their married counterparts, and they were less likely to drink alcohol regularly. They also exercised slightly more and were less overweight.
The authors say that marriage is a rough proxy for social connectedness, and suggest that never having married may be associated with more severe isolation.