Findings of existing primary studies and narrative reviews studying the associations between the perceived environment and PA are ambiguous. For example, pathways in close proximity to the home have been found to have a positive association with PA in some instances [12] while showing no association with PA in others [13]. Contrary to intuitive belief, positive associations have been found between the presence of unattended dogs and PA using both subjective [14] and objective [15] measures. Perceptions of neighborhood crime have been found to have both a negative association [15,16] and a lack of association [14] with PA. Similarly, the self-reported presence of street lighting has produced both positive [17] and no association [12,14] with activity using self-report measures and objectively defined geographic information system (GIS) measures [15]. Recent narrative reviews [11,18,19], while providing a useful summary of existing research, have been unable to resolve the aforementioned ambiguities.

Reporting the presence of proximal PA facilities, sidewalks, shops and services and that traffic was not a problem were all positively associated with PA. Although previous narrative reviews [11,19] were unable to provide consensus concerning how perceptions of traffic influence PA, the current review found positive associations across studies between the absence of traffic and PA using adjusted ORs. Since it has been noted that many road systems are designed without the needs of pedestrians in mind [56], it is time for planners to recognize the health relevance of their work. For instance, providing sidewalks separated from roads, pedestrian refuge islands, increased street lighting and roundabouts are all highly effective methods of reducing pedestrian crashes [56]. Engineering modifications can reduce traffic volumes and speeds if implemented effectively [57]. Although the potential for road modifications to reduce traffic volumes and increase individual PA may be small in magnitude, such modifications may contribute to sustainable positive changes in PA levels across the community.

Reviews from the transportation domain suggest the environment provides both cues and opportunities for people to engage in PA within their neighborhoods [5,6,58]. The current review supports such claims and provides evidence that these relationships exist across different populations. A plausible mechanism for observed associations is that neighborhood footpaths increase recreational and utilitarian walking/cycling by reducing the risk of falls (by providing people with opportunities to walk on a smooth, flat surfaces) and increasing perceptions of safety from traffic (due to barriers between roads and sidewalks) [59]. Sidewalk provision and town planning are under the control of various levels of government – local, state, federal. The various levels of government are capable of providing legislation that will promote or restrict community capacity to provide destinations (both utilitarian and recreational) and infrastructure (sidewalks and safe travel routes for non-motorized transportation) in neighborhood areas, influencing community behavior accordingly. The effects of such policy-level changes for changing PA in entire communities are potentially large and effective. Policy modifications are likely to influence PA by mandating the provision of safe environments close to the home (from both traffic and crime), useable green space and other recreational locations in which to engage in PA close to the home, active transport routes separate from vehicle traffic, and increases in school PE. Despite the difficulty of effecting change through policy or 'distal leverage points' [9], these and similar changes will likely contribute to longer lasting changes in behavior by making PA an easier "choice". Research examining how policy influences behavior [45,60,61] is promising, however research examining wider policy influences is needed to identify the most effective in creating environments producing sustainable increases in PA.

Because few studies have used objective measures of PA, the current review was limited to studies using measures of perceived environment. Although many of these studies used reliable measures of PA, such as the BRFSS [42,45], the lack of precise measurement of PA and environmental constructs may be obscuring true associations between PA and environmental characteristics [54]. Future studies are therefore encouraged to use both self-report and objective measures of PA [62], and when possible, combine self-report and objective measures of the environment to improve the predictive ability of studies. In addition, relatively few of included studies were conducted outside of the US, therefore research examining how environments influence activity in other countries is strongly encouraged.

Environmental changes may have differential effects on various sub-groups of the population (i.e. men and women may respond differently to similar aspects of the environment or environmental changes) [63]. The current review, limited to non-experimental studies, cannot answer these questions. As such, it is recommended that future research adopt a quasi-experimental approach to examine if similar environmental changes influence behaviors separately in regions where key socio-demographic measures are substantially different.

Some characteristics identified in this study, while likely difficult to alter, may be more readily manipulated than existing land densities and land-use mixes, which have received much attention within transportation literature [58]. Comparison studies demonstrate that rates of walking and cycling were higher in neighborhoods classified as transit orientated (higher connectivity) compared to neighborhoods classified as automobile orientated (lower connectivity) [64]. Additionally, modeling activities demonstrate that providing innovative block-cutting passages can increase connectivity of neighborhoods previously considered unfavorable to neighborhood walking [65]. Therefore, the provision of footpaths connecting previously unconnected neighborhoods and increasing connectivity may be a viable way to facilitate increased neighborhood PA. However, caution should be used in their planning and design to properly address safety and security concerns by providing good lighting and allowing residents and other street users' lines of sight into block-cutting passages. These modifications are effective strategies in reducing neighborhood crime [66,67].

The current research had several limitations that should be considered when examining the results. First, the sample was limited to studies that were published or accepted for publication, included adult populations, and used logistic regression analysis to determine associations between the perceived environment and activity. The limit to adult populations was imposed because very few studies examined features of the built or perceived environment in relation to the PA of children and youth. Thus, more research should be conducted with children and youth.

Second, two variables (PA facilities, presence of shops and services) displayed significant heterogeneity of variance suggesting that those groups of effect sizes do not represent a common population and that potential sources of heterogeneity should be sought. Unfortunately, the limited number of studies included in the analysis prohibited any search for moderators in the sample and thus the results for these two variables should be interpreted with caution. Likely sources of heterogeneity are the measurement instruments used to assess PA and the types of PA that were assessed in the original studies (e.g., walking, sufficient level of total PA). The possibility of PA measures and outcome measures as sources of heterogeneity should be examined in future analyses. Third, while several self-report measures with acceptable reliability [68,69] are available, studies used a variety of different measures to examine the perceived environment. Since alterations to measures may limit the possibility of making direct comparisons to previous studies, future research should use the most reliable measures in their entirety. Fourth, due to imprecise measurement of environmental constructs, variables were only included if a minimum of five ES were present. The number of ESs (n = 5) used in the analysis of outcome measures in previous meta-analyses [34,70] are similar to that used in the current study. However these studies used outcome measures (stroke, CVD) able to be measured with greater accuracy, and included larger within study sample sizes in pooled studies [34,70]. This minimum criterion may have reduced the number of constructs examined, however this delimitation was adopted to increase the confidence in ES estimation. Finally, all included studies were cross-sectional in design. Such designs may be subject to limitations including participants' self-selecting neighborhoods displaying design characteristics attractive to their own travel behaviors, attitudes [71], and PA preferences. For instance, using a pre-test/post-test study design, Krizek (2000) found that one-third of participants moved relatively short distances (<4 km) to neighborhoods possessing similar design characteristics, suggesting that neighborhood choice is based primarily on individual preference for a particular type of neighborhood [71]. Such limitations likely influence the results of included studies, and should be considered when interpreting the current results.