Background: Despite national policies to promote user choice for health services in many European countries, current trends in maternity unit closures create a context in which user choice may be reduced, not expanded. Little attention has been paid to the potential impact of closures on pregnant women's choice of maternity unit. We study here how pregnant women's choices interact with the distance they must travel to give birth, individual socioeconomic characteristics and the supply of maternity units in France in 2003. Results: Overall, about one-third of women chose their maternity units based on proximity. This proportion increased steeply as supply was constrained. Greater distances between the first and second closest maternity unit were strongly associated with increasing preferences for proximity; when these distances were ≥ 30 km, over 85% of women selected the closest unit (revealed preference) and over 70% reported that proximity was the reason for their choice (expressed preference). Women living at a short distance to the closest maternity unit appeared to be more sensitive to increases in distance between their first and second closest available maternity units. The preference for proximity, expressed and revealed, was related to demographic and social characteristics: women from households in the manual worker class chose a maternity unit based on its proximity more often and also went to the nearest unit when compared with women from professional and managerial households. These sociodemographic associations held true after adjusting for supply factors, maternal age and socioeconomic status. Conclusions: Choice seems to be arbitrated in both absolute and relative terms. Taking changes in supply into consideration and how these affect choice is an important element for assessing the real impact of maternity unit closures on pregnant women's experiences. An indicator measuring the proportion of women for whom the distance between the first and second maternity unit is greater than 30 km can provide a simple measure of choice to complement indicators of geographic accessibility in evaluations of the impact of maternity unit closures.