Despite previous evidence suggesting a lack of correspondence between self-reported and objective assessments of the built environment [4–9] we found that perceived neighborhood characteristics for the most part were higher (more positive) in objectively-determined high walkable neighborhoods than in less walkable neighborhoods. Furthermore, similar to evidence elsewhere  perceived and objectively-determined neighborhood walkability appears to be more important for neighborhood-based transportation versus recreational walking. Specifically, self-reported and objectively-measured characteristics of the neighborhood were independently associated with both weekly participation and minutes of neighborhood-based transportation walking, adjusting for neighborhood self-selection and socio-demographic characteristics. Noteworthy was that we found dence of effect modification, with stronger associations between perceived physical barriers and safety from crime and local transportation walking in HW neighborhoods only and between pedestrian infrastructure and local transportation walking in MW neighborhoods only. In support of findings elsewhere suggesting that the perceptions of built environment mediate the relationship between the objectively-measuredbuilt environment and physical activity [24, 25] we found that perceived neighborhood characteristics attenuated the association between the objectively-determined neighborhood type and participation, but not minutes, in weekly neighborhood-based transportation walking.
For most comparisons, differences in perceptions of walkability were found between the three objectively-determined neighborhood types in the expected direction. Compared with respondents from LW neighborhoods, those from HW neighborhoods in general reported their neighborhoods to be more walkable. HW neighborhoods in our study included, among other attributes, high levels of street connectivity and count of businesses  which was congruent with respondent’s positive evaluations of street connectivity, access to services and utilitarian destination mix. HW and MW neighborhoods were similar with regard to sidewalk length, the mix of park types and recreational facilities corresponding with the more positive perceptions of perceived pedestrian infrastructure and recreation destination mix within these neighborhood types. Previous studies that have measured concordance between perceived and objectively-measured walkability have found poor to fair agreement [4–7, 9], although Arvidsson  found higher agreement when overall indices of perceived and objectively-measured walkability were compared. Gebel et al.  found that approximately a third of participants living in HW neighborhoods (based on dwelling density, street connectivity, land use mix and retail density) misperceived their neighborhoods to be low walkable. We found that perceived aesthetics and safety (from crime and traffic) were more negative in objectively-determined high walkable versus low walkable neighborhoods. This result is similar to Van Dyke et al.  who found that socioeconomically disadvantaged women residing in neighborhoods with high connectivity and destination density perceived their neighborhoods to have poor aesthetics, to have lower social cohesion, and to be less safe compared with women in neighborhoods that had lower connectivity and destination density. In both studies, self-report measures of aesthetics and safety only were captured and were not objectively-determined. Thus perceptions of aesthetics and safety could reflect real differences across objectively-measured neighborhood types or could reflect differences in the way the perceived and objective built environment measures are operationalized. Comprehensive objectively-determined walkability indices that better reflect the range of built characteristics important for supporting and encouraging neighborhood walking are needed.
Different built environmental factors were found to be associated with walking participation (some or none) and duration (minutes). For example, neighborhood aesthetics had a significant positive association with participation in walking for recreation but not with duration among respondents who reported some walking for recreation, independent of objectively-determined neighborhood characteristics. The differences in potential determinants of transportation versus recreational walking, and of initiating walking versus extending the duration of walking among those who already walk, suggest that there are different decision-making processes involved in these activities. These differences have implications for designing future interventions to promote walking for different purposes. Also notable was that self-reported aesthetics only was associated with neighborhood-based recreational walking while objectively-determined neighborhood type did not appear to be associated with recreational walking. The influence of the built environment on walking behavior may be weaker for recreational than for transportation walking . Speculatively, our cross-sectional findings suggest that the increases in recreational walking might be more likely if interventions target individual-level (motivations and perceptions) factors, while increases in transportation walking might be more likely in interventions focus on changing individual-level characteristics in addition to improving neighborhood walkability.
Statistically significant interaction effects in MW or HW neighborhoods but not for LW neighborhoods were found. For those who actually reside in LW neighborhoods, their perceptions of walkability appear to have limited influence on walking behavior compared with resident of MW and HW neighborhoods. Lacky and Kacynski  found that while neither perceived or objective proximity to parks alone was associated with participation in park-based physical activity, their interaction was positively associated with physical activity. Giles-Corti et al.  found that among those relocating to new neighborhoods, increases in perceptions of neighborhood walkability were positively associated with time spent walking for transportation and recreation, even after adjusting for objectively-determined neighborhood built characteristics. Among other significant interaction effects, we found that higher positive evaluations of physical barriers (higher scale scores represented fewer barriers) was positively associated with transportation walking in high walkable neighborhoods. Improving perceptions about the physical barriers in high walkable neighborhoods, through micro-scale modifications (removal of major pedestrian barriers, reducing steep hills and gradients along walking routes, etc.) might encourage additional amounts of walking among local residents. More research is needed to better understand how specific types of perceived and objectively-determined neighborhood built characteristics might combine to influence neighborhood-based walking as well as other physical activities. Longitudinal studies that attempt to improve perceptions of the built environment among residents living in neighborhoods of differing levels of walkability could provide insight into the relative contributions of perceived versus built neighborhood characteristics on walking.
Counter-intuitively, an increase in self-reported safety from crime was significantly associated with decreased minutes of walking for transportation in HW neighborhoods among those residents who reported some walking. This interaction appears to be inconsistent with evidence suggesting a positive or no relationship between perceived safety and neighborhood-based physical activity , but consistent with Michael and Carlon  who found that among older adults increases in walking was accompanied by a reduction in positive perceptions of neighborhood problems (incivilities, violent crime, abandoned buildings etc.). One explanation for this finding is that people who are more active in their neighborhood also tend to be more aware of its characteristics due to greater exposure and familiarity. Alternatively, HW neighborhoods tend to have higher population density, and may afford more potential opportunities for undesirable interactions between people. Speculatively, it is also possible that the nature of crimes differ according to neighborhood type, and that the severity of crimes within high walkable neighborhoods might result in perceived personal safety having a stronger influence on resident’s walking decisions. Improving perceptions of safety in high walkable neighborhoods – i.e., through crime prevention strategies or urban design (i.e., increased surveillance and pedestrian lighting, reduction in incivilities) – might increase local walking.
Several limitations should be considered when interpreting these study findings. Estimated associations cannot be considered causal given the cross-sectional study design. While our study advances previous cross-sectional studies by statistically adjusting for participant’s reasons for moving to their neighborhood [37, 38], unmeasured factors associated with neighborhood self-selection may still exist, thus inflating estimated associations between the built environment and physical activity. Moreover, given that participants in our sample reported living in their neighborhoods for approximately 13 years (on average), memory bias and errors associated with people’s reasons for moving to their neighborhoods likely exist. The modest response rate, study selection bias, and the age of the data may limit the generalizability of our findings. Despite using reliable context-specific (neighborhood) measures of physical activity, biases associated with capturing self-reported physical activity is a limitation . Despite selecting the best fitting principal component analysis solutions, the low internal consistency of some perceived walkability and neighborhood preference scales could have led to fewer associations between these scales and the walking outcomes being found in our study. The use of a simple random sample to recruit households across the Calgary metropolitan area meant that fewer high walkability neighborhoods were included in our sample, thus, resulting in fewer survey participants from this neighborhood type. The lower sample size in the high walkable neighborhoods likely reduced statistical power to find statistically significant differences or associations. For example, in some cases we found point estimates (group differences and associations) of similar magnitude to be statistically significant for low and medium walkable neighborhoods but not statistically significant for high walkable neighborhoods.
Both the perceived and objectively-determined neighborhood built characteristics appear to contribute to neighborhood-based walking – in particular walking for transportation. However, objective measures and perceptions of the built environment should not be considered interchangeable. Our findings suggest that interventions designed to change perceptions of neighborhood walkability might have a stronger influence on walking for adults in higher walkable neighborhoods. Creating walkable neighborhoods might increase participation and time spent walking for transportation locally; however, increasing local recreational walking might require other intervention approaches.