Systematic reviews of the literature have identified some

Systematic reviews of the literature have identified some consistent physical environment correlates of physical activity, including land use mix, connectivity and residential density which all have positive associations with activity (McCormack & Shiell, 2011; Saelens & Handy, 2008). Access to green space may also be important: a study including over 200,000 adults reported cross-sectional associations between green space access and increased self-reported walking and moderate-to-vigorous physical activity (MVPA) (Astell-Burt, Feng, & Kolt, 2014). The social environment has also been examined in relation to physical activity. In particular, cognitive and structural social capital constructs have been explored, encompassing aspects of perceived or objective social cohesion, trust, social support, safety, social participation and social resources (e.g. collective efficacy to enforce normative behaviours and reciprocity in sharing personal resources) (Moore & Kawachi, 2017). In a recent systematic review of 38 studies, Samuel, Commodore-Mensah, & Himmelfarb (2014) identified several characteristics of the social environment associated with overall physical activity, walking and sports participation, with higher quality social environments (i.e. increased sense of community, trustworthiness, reciprocity, social potassium phosphate monobasic and social control) indicating higher levels of activity. There is also some evidence for a negative association between physical activity and crime and a positive relationship between physical activity and perceived safety, although findings are inconsistent. Several reasons could contribute to inconsistent results: i) inadequate measurement of crime resulting in measurement error, ii) use of physical activity outcomes that are not neighbourhood-based and therefore may have weaker relationships with the neighbourhood environment and iii) lack of consideration of features of the physical and social environment that may mediate or moderate the effects under investigation (Foster & Giles-Corti, 2008). A core tenet of ecological models of physical activity is that correlates are embedded in a complex system whereby multiple environmental and individual characteristics are interrelated and exert independent and interactive effects (Sallis et al., 2006). While a growing literature examines independent effects of environmental correlates, there has been very little focus on their interactive or synergistic effects on physical activity despite empirical and theoretical evidence of interplay between social and physical environments (e.g. social interaction is related to structural elements including provision of communal space (Yancey, 1971), physical disorder is associated with collective efficacy (Sampson & Raudenbush, 1999) and bidirectional reciprocal associations existing between social and physical disorder as purported by broken windows theory (Keizer, Lindenberg & Steg, 2008)). The scientific value of examining social and physical effects simultaneously (rather than only controlling for other environmental correlates) is to explore the concurrent influences of social and physical environmental features on physical activity, as hypothesised in ecological models. Conceptualising concurrent influences could elucidate counter-intuitive relationships between the environment and physical activity. For example, although there is an established relationship between area deprivation and poorer health outcomes and behaviours, including physical activity (Ecob & Macintyre, 2000), a study in two Scottish neighbourhoods found that the deprived neighbourhood had more recreation centres, sport centres and street cleaning than the affluent neighbourhood, undermining the assumption that more deprived areas would be physically less supportive of activity (Macintyre, Maciver & Sooman, 1993). Various studies in Europe, USA and Australia also report that physical activity resources are not fewer in more deprived areas (Cradock et al., 2005; Giles-Corti & Donovan, 2002; Van Lenthe, Brug & MacKenbach, 2005). In Canada and USA, lower levels of physical activity were self-reported in areas that are objectively-classified as highly walkable (according to physical metrics like connectivity) than in less walkable areas (Jack & McCormack, 2014; King, 2008). In such instances, features of the social environment or micro-scale features of the physical environment may modify the impact of physical walkability metrics.