There is increasing evidence that the socio-spatial context of the local area in which one lives can have an effect on health, but teasing out contextual influences is not a simple task. We examine whether the incidence of Type 2 diabetes in small areas in Tayside, Scotland is associated with deprivation in neighbouring areas, controlling for the deprivation of the area itself. As such, this is a genuinely 'contextual' variable situating each small area in the context of surrounding places. We test two opposing hypotheses. First, a 'psycho-social' hypothesis might suggest that negative social comparisons made by individuals in relation to those who surround them could lead to chronic low-level stress via psycho-social pathways, the physiological effects of which could promote diabetes. Thus, we would expect people living in deprived areas surrounded by less deprived areas to have an increased risk of diabetes, compared to those living in similarly deprived areas that are surrounded by equally or more deprived areas. Alternatively, a neo-materialist approach might suggest that the social, cultural and environmental resources in the surrounding environment will influence circumstances in a particular area of interest. Poorer areas surrounded by less deprived areas would benefit from the better resources in the wider locality, while less deprived areas surrounded by poorer areas may be hampered by the poorer resources available nearby. We refer to this as the 'pull-up/pull-down' hypothesis. Our results show that, as expected, area deprivation is positively related to diabetes incidence (p<0.001), whilst deprivation inequality between areas and their neighbours is negatively related (p=0.006). Type 2 diabetes is more common in deprived areas, but lower in deprived areas that are surrounded by relatively less deprived areas. On the other hand, less deprived areas that are surrounded by relatively more deprived areas have higher diabetes incidence than would be expected from the deprivation of the area alone. Our model results are consistent with a pull-up/pull-down model and lend no support to a 'psycho-social' interpretation at this local scale of analysis.