A characterization of the modelling process in science is proposed for science education, based on Mario Bunge's ideas about the construction of models in science. Galileo's "Dialogues" are analysed as a potentially fruitful starting point to implement strategies aimed at modelling in the classroom in the light of that proposal. It is argued that a modelling process for science education can be conceived as the evolution from phenomenological approaches towards more representational ones, emphasizing the role of abstraction and idealization in model construction. The shift of reference of theories--from sensible objects to conceptual objects--and the black-box models construction process, which are both explicitly presented features in Galileo's "Dialogues", are indicated as highly relevant aspects for modelling in science education.