This paper evaluates the role of residential self-selection in the spatial distribution of perceived accessibility in the Netherlands. People may self-select into residential areas that match their preferences regarding participation in out-of-home activities. Differences in the magnitude of opportunities provided by the land use and transport system may, consequently, not accurately mirror perceived accessibility differences, complicating the design of responsive policy. Perceived accessibility was found to be less variable than spatial accessibility, meaning that fewer opportunities do not fully translate into lower perceived accessibility. Expected perceived accessibility differences under random conditions were estimated by comparing residents of distinct spatial accessibility contexts with the same propensity to live in either context, yielding a quasi-experimental setting. Estimates of the expected increments in average perceived accessibility levels as the number of locally available opportunities increases indicate diminishing returns to spatial accessibility. In addition, preference-based residential self-selection was found to further mitigate spatial accessibility differences. Yet despite diminishing returns to spatial accessibility and residential self-selection, perceived accessibility remains lower in rural areas due to limited residential freedom. In addition, the non-linear benefits of spatial accessibility imply that further loss of facilities in rural areas can quickly lead to insufficiency below a certain tipping point. Already, residential self-selection in rural areas strongly relies on access to car mobility, conflicting with environmental and social inclusion accessibility planning objectives. Accessibility-based transport planning, therefore, requires not only a substantive shift away from alleviating car congestion, but also a geographical shift in favour of rural areas.