The aim of this paper is mainly a methodological one. We used individual-level retrospective data from the Italian Fertility and Family survey, based on the information provided by respondents. Linking individual-level data with geographical characteristics of the community, we …rstly used non-parametric methods, and in particular the bi-dimensional LOESS (Locally weighted regression) method, to build smoothed maps of the transition to …rst sexual intercourse. Secondly, we used a discrete-time multilevel event-history model (with logit speci…cation), allowing for the presence of unobserved heterogeneity at the municipality level. We used 'latitude', 'longitude', and dimension of the municipality as aggregate-level explanatory variables in the model. Thus, we could build spatial pro…les and a map, using predicted values for the logodds-ratio, and we could study cohort dynamics too. Our results con…rm that Italy is a heterogeneous country; when looking at age at the …rst sexual intercourse, women are more in ‡uenced by their context than men.