Context: Obesity and related non-communicable diseases represent a large and growing disease burden, and various aspects of contemporary urban environments may be important drivers of these health outcomes. Neighbourhood resources for physical activity, exposure to fast-food outlets and greenspace may play a role, but the evidence regarding these neighbourhood effects remains equivocal. A possible explanation is that neighbourhood effects may be heterogeneous; some people may be more sensitive than others to the built environment, and neighbourhoods may matter more in some places than in others. We know little about how neighbourhood characteristics interact with other factors at various scales to influence health. Methods: Using UK Biobank, a sample of >300,000 adults in mid-life, I conducted a multiscalar examination of heterogeneity in associations between neighbourhood built environments and adiposity, CVD and cancer. I examined cross-sectional associations between characteristics of the physical activity and food environments around people's home addresses, and multiple measures of adiposity, exploring potential effect modification by individual-, neighbourhood- and local authority-level variables. Using linked hospital data, I examined effect heterogeneity in longitudinal associations with CVD and cancer. Results: Population-wide associations obscured substantial effect heterogeneity. The magnitude of these associations varied with genetic risk, socioeconomic position, gender, and other environmental factors. Greater availability of formal physical activity facilities near home was associated with lower adiposity, particularly for people without public parks nearby, but less so for people also living near a fast-food outlet. Proximity to a fastfood outlet was associated with BMI, but this was modified by genetic risk of obesity: highrisk individuals appear to be more sensitive to their food environment. Associations between adiposity and neighbourhood characteristics varied across England, and potential drivers of this geographical heterogeneity were identified. Neighbourhood greenspace was not associated with any outcomes across the sample as a whole, but appeared protective against cancer in more deprived areas. Conclusions: This thesis contributes empirically and conceptually to our understanding of how, where and for whom neighbourhoods matter for health. The results highlight the role of individual and contextual factors operating at multiple scales to moderate people’s sensitivity to characteristics of their neighbourhood environment. A better understanding of effect modification relationships such as these may help to guide the development and evaluation of future interventions and policies involving the urban built environment.