Abstract: Background: In censuses and demographic surveys, religion is recorded as a variable of state, assuming that individuals' religious affiliation is unique and definitive. However, in sub-Saharan Africa, pluralism are commonplace. In this paper, we discussed the relevance and feasibility of a statistical approach to religious practices, taking into consideration their complexity and variability over an individual's lifetime. Data Source and methodes: We used longitudinal data collected since 25 years in the south-east of Mali, among a population where traditional and Christian religions coexist. We compared the results of a classic cross-sectional approach with those obtained via a longitudinal approach that tal‹es into consideration individuals' religious trajectories. Résultats: Plurality and variability in religious practices are confirmed. Most individuals, at some point in their lives, become affiliated with different religions. Mobility and reversibility in religious affiliation are common. Conclusion: The relevance of cross-sectional data on religious affiliation for demographic analysis is therefore questionable. Keywords: religion, religious pluralism, religious affiliation, Africa, Mali