Accessibility is essential in land-use transport interaction frameworks. For residential location choices in particular, it has always been important at the theoretical level. At the empirical level, the place of accessibility is questioned in some works, considering other more important location characteristics, like the social environment and the neighbourhood amenities. However, this result can be caused by the measurement of accessibility. In view of the wealth of approaches, this paper examines whether different accessibility measures can lead to divergent results. Using a residential location choice model for the Lyon urban area in France, we test various accessibility indicators and we compare the results. We conclude that accessibility is an indispensable variable. Without it, the model gives inconsistent results. Complex accessibility measures give better results but simple measures are also relevant for residential location choices modelling. The choice highly depends on the objectives of the application especially if the model is to be used for simulation.