Content and language integrated learning (CLIL) programmes are increasingly popular throughout Europe, but are sometimes accused of inducing a selection bias in the pupil population, both through selection mechanisms of the schools themselves and self- selection of the pupils (and/or their parents). As a result, the outcomes of the CLIL approach may be artificially promoted, and, at the same time, such a selection bias can contribute to an elitist education model, which arguably runs counter to the aims of the approach. This paper looks into a number of background variables of both English and Dutch CLIL learners in Francophone Belgium and compares them to their non-CLIL counterparts. Results from a logistic regression indicate that there is indeed evidence of selection: the socio-economic status of the pupils appears as the main predictor of whether a pupil is in a CLIL or a non-CLIL track, whereas other, more personal, variables such as non-verbal intelligence play a minor (or additional) role. Moreover, Dutch CLIL programmes appear to be more selective than English CLIL programmes in this context. We conclude that CLIL (and particularly Dutch CLIL) in French-speaking Belgium, although a priori open to anyone, is particularly attractive to a socially privileged public.