In this paper, an empirical evaluation is presented of two competing flexible labour supply models. The first is a standard unitary model, while the second is based on the collective approach to household behaviour. The evaluation focuses on the testing of the models' theoretical implications, on their ability to identify structural information, like preferences and on their empirical performance. Models are estimated on Belgian microdata from 1992 and 1997. The unitary model cannot be rejected for single person households, while it is rejected for a sample of two person households. The alternative collective model cannot be rejected for the same sample. However, since the crucial assumption of egoistic or Beckerian caring individual preferences is rejected, the comparative advantage of the collective model as basis for intrahousehold welfare evaluations cannot be fully exploited. Finally, the collective model has the best empirical fit.