In literature on neighbourhood effects and on resources accessibility, number of neighbourhood resources to which residents may have access are often estimated from spatial units whose constant size fails to account for unique ways residents experience their neighbourhoods. To investigate this "constant size neighbourhood trap", we compared numbers of health-care resources included in Constant Size Buffers (CSBs) and in Perceived Neighbourhood Polygons (PNPs) from cognitive neighbourhood data collected among 653 residents of the Paris metropolitan area,. We observed that residents of deprived and peripheral areas had smaller PNPs than their counterparts. Studying residents assessments of the quantity of neighbourhood practitioners, we then assessed the validity of using PNPs rather than CSBs to estimate number of neighbourhood resources. Lastly, resource inequalities across the Paris metropolitan area were found to be far wider when considering PNPs rather than CSBs. Using constant neighbourhood delineation can lead to inaccurately measure individual accessibility to neighbourhood resources and to downplay the extent of inequalities in urban resources.