Understanding how the built environment shapes walking behavior is central to evidence-based pedestrian planning, yet the mechanisms and spatial scale at which environmental characteristics operate remain contested. Most existing research relies on perceived environmental measures and self-reported behavioral outcomes, concentrates on large cities, and treats the neighborhood as the sole unit of environmental exposure. This thesis addresses these gaps by examining the relationship between objectively measured built environment characteristics and GPS-tracked utilitarian walking in two medium-sized Swedish cities, Umeå and Linköping, across multiple datasets collected in 2019 and 2021.

Built environment exposure is operationalized at two complementary spatial scales: potential exposure, measured within a 750 m radius of each participant's home location, and realized exposure, measured within a 15 m buffer along GPS-traced actual routes and their shortest-path alternatives. This dual-scale design enables a direct empirical comparison of which environmental features predict how much people walk versus which features shape the paths they take when they do. Psychological and socio-demographic variables, derived from a Theory of Planned Behavior questionnaire integrated into the tracking application, are examined as potential mediators and moderators of built environment effects. Analytical methods include bivariate correlation analysis, structural equation modelling, linear mixed models, and discrete choice analysis.

At the neighborhood level, building density, bus stop access, commercial land use, tree cover, paved surface area, and pedestrian network length show modest but statistically significant positive associations with walking distance and walking ratio. Structural equation modelling confirms that the built environment retains a significant direct effect on walking when attitudes are simultaneously controlled, though attitudes are the stronger predictor. Contrary to the mediation hypothesis, the built environment does not significantly influence walking through attitudes in the overall sample; however, moderated mediation analysis reveals conditional pathways by age, with a significant indirect effect via attitudes among middle-aged adults and a stronger direct environmental effect among young adults.

At the route level, 95.65% of observed trips follow the shortest available path, confirming distance minimization as the dominant pedestrian strategy. For trips that deviate, commercial land use is the strongest positive predictor of route selection in the discrete choice analysis, while dedicated non-vehicular pedestrian infrastructure and low-speed streets are consistently negatively associated with route deviation across both linear mixed models and discrete choice models. This counterintuitive negative association reflects the spatial structure of Scandinavian pedestrian networks, where off-road paths form the most direct routing corridors rather than alternative scenic routes. Building density effects are city-specific, with positive associations in Umeå reversing in Linköping, pointing to the moderating role of urban spatial structure. Moderation analyses further show that income, age, and scenic motivation condition how specific built environment features influence route behavior, while no socio-demographic or attitudinal variable produces a significant unconditional main effect on route choice.

The comparison across scales reveals that the environmental variables associated with walking volume are largely distinct from those shaping route selection, a finding that challenges undifferentiated walkability frameworks and suggests that planning interventions targeting modal shift and those targeting route quality require different built environment priorities. These results advance understanding of scale-dependent built environment effects on pedestrian behavior and provide an empirical basis for more targeted assessment of walking behavior in Nordic urban contexts