Since the mid 1970s, the impact of nearby schools on property values has been a major theme in the North-American literature, with most studies focusing on the influence of school quality. To a lesser extent, the impact proximity to school exerts on house prices has also been investigated together with the way the urban environment and distances to school affect student modal choices. Using the hedonic approach, this paper aims at assessing the price effect ensuing from proximity of, and accessibility to, the nearest primary school based on a sample of 8,285 single-family houses sold over the 1993-1996 period in Quebec City, Canada. While home-to-school walking distances are computed for properties located within one kilometre ñ a distance beyond which a school bus service is provided, car travel times measured on the 1994 GIS-operated topological street network are used for houses located further away from school. Main property features as well as neighbourhood and household descriptors are used as control variables in the model, with four urbanization patterns being distinguished on the grounds of residential densities. A major objective of this study is to test the spatial homogeneity of hedonic prices with respect to primary schools. In order to do so, the OLS method, Cassettiís expansion method and the geographically weighted regression (GWR) method are compared. Each method is assessed in regard to its ability to account for, and deal with, the presence of spatial autocorrelation in the residuals while bringing out spatial shifts in household preferences.