Meta-analysis of the relationships between space syntax measures and pedestrian movement

Ample research has been conducted investigating the built environment impacts on pedestrian movement (PM). A clear division is also evident in the literature on this topic: one group tends to use geographic measures (metric distance) of the environment to explain pedestrian behaviour; the other group uses syntactic measures (visual distance). Many review articles have been published on the former. However, relatively little is known about the effect size (ES), directions, and consistency of syntactic measures in explaining PM. This paper fills this gap through a meta-analysis of published and unpublished studies on this topic spanning over 1975-2016, and answers the following five research questions: (a) What are the different measures used in the space syntax literature to explain PM?; (b) What are the magnitudes and directions of associations between space syntax measures and PM?; (c) Which space syntax measure has a more consistent relationship with PM?; (d) To what extent do the explanatory powers of different measures vary between their derivation methods?; and (e) What are the likely causes of variations of the reported results in prior studies? This research examined four syntactic measures (integration, connectivity, choice, and control) in a random effect model with 95% confidence interval (CI). The choice and integration measures were further investigated based on their operational approaches (topological, angular and metric). Results show that integration (ES=0.206, 95% CI=0.173-0.238, p<0.001), choice (ES=0.481, 95% CI=0.391-0.561, p<0.001), and connectivity (ES=0.305, 95% CI=-0.225-0.696, p=0.257) measures positively influence PM with choice being the strongest predictor. Both connectivity and control (ES=-0.001, 95% CI=-0.117-0.115, p=0.990) were found to have a statistically insignificant impact. The choice and integration measures show stronger effects when derived using the angular approach (ES=0.493 for choice, and ES=0.502 for integration) compared to topological approach (ES=0.374 for choice, and ES=0.124 for integration). However, the reported results of all measures are highly heterogeneous, perhaps due to the differences in research design. The significance, magnitude, and consistency of integration and choice measures justify their relevance in built environment interventions to promote PM.