The compensation hypothesis postulates that the absence of green spaces in residential areas leads to a greater use of public green spaces and nature in general. This paper tests this hypothesis by focusing on the use of public and private urban green spaces. A holistic approach is adopted that considers the complexity of residential context and the use of urban, peripheral and private green spaces. A survey was conducted in two medium-sized French cities (Dijon and Besançon) to identify how their green spaces were used and perceived. The residential context was described through multiple spatial and environmental metrics (physical and visual access to nature, noise level, private green space area) that were GIS-referenced to postal addresses. On the basis of that data, the compensation hypothesis was explored using a PLS path model. The results show the absence of compensatory behaviour when considering just the overall relationships between the use of different types of green spaces. However, conditional compensatory behaviours can be detected when allowance is made for other variables (physical and visual accessibility to nature, type of housing, dwelling floor area, noise level). These results remain dependent on the geographical context of the city, with a higher compensation effect where natural areas are less accessible.