Calcutta after the rains, ankle-deep in filthy water, awash with sewage. People emerge dripping from the bits of bric-a-brac that pass for housing in the shanty towns, children squatting in the gutter with a large block of ice, chipping off pieces to add to a drink. This is an environment of poverty, and there is little difficulty imagining it being bad for health.

The effects of environment on health are seen most vividly in the deaths of children. In India, of every 1,000 babies born alive, more than 80 will die before the age of five. In much of sub-Saharan Africa, it is worse. Sierra Leone wins the grim bouquet: 316 deaths by age five for every 1,000 children born alive.

Can environment harm health in a rich country such as Britain? Have we not solved the major environmental threats to health? Happily, the death of a child is uncommon. Seven deaths per 1,000 live births puts us on a different scale from Africa or India. The health problems we face are mental illness, heart disease, cancer and diabetes - not illnesses normally attributed to dirty water or inadequate housing, but the environment does cause health and disease in Britain. It is the social environment that is crucial; people, their interactions, and what they do.

Where a person lives is a guide to longevity. This can be illustrated with a tour of the borough of Camden, in north London. Between the sleek elegance of Belsize Park and Hampstead, and the not-so-sleek areas of Kilburn and Holborn, there is a 10-year variation in life expectancy for men (79.6 years in Belsize Park, 69.7 in Kilburn). To put this into context, if heart disease - the chief cause of death - was abolished, it would add just under four years to life expectancy. These huge differences are accompanied by big differences in illness statistics.

This problem is not peculiar to London. Cities vary. American cities such as Washington DC and New York have greater variations in health than London. Helsinki has fewer. These variations relate to the way cities are segregated. In Helsinki, manual workers, the unemployed and single mothers are spread more evenly than in the US, where geographical separation is clearer.

In seeking to understand why Belsize Park appears more salubrious than Kilburn, the obvious should not be overlooked. In addition to predicting illness and death, where a person lives is also a guide to their standing in society. It might be that standing that predicts longevity, not location. In other words, poor places may have worse health because poor people live in poor places, and poor people have worse health than the rich.

This distinction is rather important. If it were the people and not the places that mattered for health, then moving people from a poor area to a rich area would do nothing for them. Similarly, if geographical variations in health were all down to who lives where, investing in neighbourhoods would do nothing either. One would need to invest in people, not places.

Our research shows that people and places both matter for health in Britain. With Mai Stafford at University College London, and in collaboration with Sally Macintyre's unit in Glasgow, we have shown that for people of equivalent social status, it is the area that matters. More deprivation in an area means worse health. Lower social position means worse health. Both are important. In fact, the two interact.

Looking at the whole of England, we find a much more striking variation in mortality for people at the bottom of the social tree than for those at the top. In London, for example, people classified (in the old registrar-general's scheme) as unskilled manual have about twice the mortality rate of top professionals. In the north-east of England, the difference in mortality is sixfold. It is the people at the bottom who show the biggest variation, not the people at the top.

Our own studies confirm that mental illness and self-reported health problems vary more among areas for people lower in the social hierarchy. One question therefore had a simple answer. For those of low social status, health is made even worse by living in a poor area. There is a kind of double jeopardy.

This last finding was rather challenging. In my book, Status Syndrome, I assembled evidence that social status matters for health. I argued that when it comes to the importance of money for health - for people who had adequate food and shelter - it was not so much the absolute amount that mattered but how much relative to others.

The important question is relative to whom? One might imagine that people compared themselves with their neighbours. Hence, the poor might feel even worse about their position in society if they lived in a rich neighbourhood and were confronted daily with evidence of their low social status.

This argument that local comparisons matter would suggest that health would be better if the poor were all together. At least they would be in the same boat as others. But that is not what the data shows. Poor people living in poor neighbourhoods have worse health than if they lived in richer ones. This does not mean that the theory of relative position and health is wrong. It may mean that people do not feel better about their place in society simply because their neighbours are similarly disadvantaged.

Rather than simply speculate, we tested it out. We gave respondents in a study of civil servants a drawing of a ladder representing the social hierarchy and asked them to place a cross on the rung that marked their place. As expected, the lower the grade of employment, the lower people were likely to place themselves. Where they lived mattered to self-ratings, but not in the way set out above. For people of low employment grade, the more deprived the area, the lower they ranked themselves on the social ladder. For those of low status at work, their perception of social position was enhanced by living in a more affluent area, and further diminished by living in a poor one.

My general thesis is that relative position matters for health for two connected reasons: degree of control over life circumstances and opportunities for full social participation.

Living in a poor area is likely to be bad for both of these. The question is what is it about neighbourhoods that matter for health?

One obvious candidate is physical hazard. If home or workplace has lead paint peeling off the walls, unsafe stairs and a high fire risk, there will be health risks too. Important as they are, they will not explain differences in mortality from heart disease or homicide that we see in US cities.

A second aspect is amenities. Lack of transport and poor access to services and facilities are all associated with worse health.

There is a third important aspect grouped under the heading of social cohesion. Robert Putnam, professor of international affairs at Harvard, set out the argument for the importance of social capital in his book, Bowling Alone. He detailed the connections among individuals - social networks and the norms of reciprocity and trustworthiness that arise from them.

Ichiro Kawachi, professor of social epidemiology at Harvard, put this concept into operation in his investigation of the variations in mortality among US states. He used responses to a simple social survey and showed that the higher the income inequality of a state, the lower was the level of trust, perceived fairness, perceived helpfulness or civic engagement. He concluded that low social capital provides the mechanism that links income inequality to mortality.

In our studies, we characterised neighbourhoods according to responses to a simple survey. We found that trust, tolerance and sense of attachment to the neighbourhood were strongly related to health. These links were independent of the physical infrastructure.

Interestingly, the links of social cohesion to health were stronger among women than men. The likely explanation was that for women involved in childcare, the social cohesion of the neighbourhood is of vital importance.

The connection with social position is that lower status is associated with a greater likelihood of living in a neighbourhood characterised by lack of tolerance and trust. While this research does not, in itself, provide ready answers as to how to improve social cohesion at the local level, recognition of its importance is, at least, a place to start. It suggests that segregation of poorer people geographically is a bad policy. Poorer people have, if anything, more to gain from a more cohesive neighbourhood than those with higher social position.

More generally, recognition of the importance for health of social cohesion suggests that this should be a central focus of policy development.

· Sir Michael Marmot is professor of epidemiology and public health and director of the International Centre for Health and Society at UCL. He is speaking at a conference next week, on the healing environment, at the Royal College of Physicians. Status Syndrome is published by Bloomsbury, 2004

· Details of the RCP conference on the healing environment, at www.rcplondon.ac.uk/calendar/2005/conf_2005_he.htm