Two major models investigate the relationship between conceptions and knowledge: the KVP model of Clement and the structural model of Balacheff. In order to prove the necessity to shift the conceptions as claimed by Giordan’s allosteric model, it is important to study the evolution and the relationship between conceptions and knowledge. This research deals with the following questions: what sort of relationship can be established between conceptions and knowledge? What factor can make these two parameters evolve? This study was carried out on 45 pupils (two classrooms). It intended to collect conceptions and knowledge before and after science learning training. Individual and collective evolution path were analyzed. The results show a similar improvement of conceptions and knowledge for most of the pupils after learning training. However, some pupils present an imbalance between the level of their conceptions and knowledge after training. These data could mean that the phase of deconstruction is not yet achieved. However, balance and imbalance between conceptions and knowledge reveal the existence of a relationship. These results give evidence in favour of the transformation of conception as claimed by allosteric model. They also confirm the existence of relationship between conceptions and knowledge as indicated in Clement’s model, but so far no data allows confirming Balacheff’s hypothesis claiming that knowledge can be described as a set of conceptions.