Background: Few studies have focused on the social inequalities associated with environmental noise despite its significant potential health effects. This study analysed the associations between area socioeconomic status (SES) and potential residential exposure to road traffic noise at a small-area level in Marseilles, second largest city in France. Methods: We calculated two potential road noise exposure indicators (PNEI) at the census block level (for 24-h and night periods), with the noise propagation prediction model CadnaA. We built a deprivation index from census data to estimate SES at the census block level. Locally estimated scatterplot smoothing diagrams described the associations between this index and PNEIs. Since the extent to which coefficient values vary between standard regression models and spatial methods are sensitive to the specific spatial model, we analysed these associations further with various regression models controlling for spatial autocorrelation and conducted sensitivity analyses with different spatial weight matrices. Results: We observed a non-linear relation between the PNEIs and the deprivation index: exposure levels were highest in the intermediate categories. All the spatial models led to a better fit and more or less pronounced reductions of the regression coefficients; the shape of the relations nonetheless remained the same. Conclusion: Finding the highest noise exposure in midlevel deprivation areas was unexpected, given the general literature on environmental inequalities. It highlights the need to study the diversity of the patterns of environmental inequalities across various economic, social and cultural contexts. Comparative studies of environmental inequalities are needed, between regions and countries, for noise and other pollutants.