The association is among the most replicated in social epidemiology: across pooled prospective studies, socially isolated people have roughly a quarter to a third higher mortality than the well connected, and loneliness carries a similar excess. That much is established. What is contested is how much of it is causal. Isolation and poor health cause each other, the studies are observational, and the effect shrinks when baseline health is properly measured. Intervention trials have shown that loneliness can be reduced; none has been large or long enough to show an effect on mortality. Public health bodies have nonetheless moved: a US Surgeon General's advisory in 2023 and a WHO commission in 2025 both treated social connection as a determinant of health on the scale of smoking or obesity. That is a policy judgement running ahead of the causal evidence, which is not unusual and not necessarily wrong. The measurement is also inconsistent: isolation, loneliness and living alone are different things, and studies mix them.