Computer programming is a core area in computer science education that involves practical as well as conceptual learning goals. The literature in programming education reports however that novice students have great problems in their learning. These problems apply to concepts as well as to practise. The empirically based research presented in this thesis contributes to the body of knowledge on students' learning by investigating the relationship between conceptual and practical learning in novice student learning of programming. Previous research in programming education has focused either on students' practical or conceptual learning. The present research indicates however that students' problems with learning to program partly depend on a complex relationship and mutual dependence between the two. The most significant finding is that practise, in terms of activities at different levels of proficiency, and qualitatively different conceptual understandings, have dimensions of variation in common. An analytical model is suggested where the dimensions of variation relate both to concepts and activities. The implications of the model are several. With the dimensions of variation at the center of learning this implies that when students discern a dimension of variation, related conceptual understandings and the meaning embedded in related practises can be discerned. Activities as well as concepts can relate to more than one dimension. Activities at a higher level of proficiency, as well as qualitatively richer understandings of concepts, relate to more dimensions of variation. Concrete examples are given on how variation theory and patterns of variation can be applied in teaching programming. The results can be used by educators to help students discern dimensions of variation, and thus facilitate practical as well as conceptual learning.