This paper is about a thorough study aiming to identify the components of the modelling of physical phenomenon by means of differential equations from their emergences to the present. Two very important question (Q1 and Q2) relative respectively to process of the modelling are examined: Q1: How do build a model? In the light of researches - Guillon (1995), Winther (1993) - in physics related to the modelling of real systems, it is interesting to distinguish two methods of construction of a model: the first one, that can be called experimental process, requires to put in equation the system in an existing theory by basing itself on the theoretical knowledge. The second that is called theoretical process necessitates the acquisition of experimental data. Q2: What is the validity of the built model? To decide on the validity of the model, generally it is necessary to compare the data obtained from the equation to those collected by the real system. If these two sets of value is conform to each other, it is possible to say that set of equations represents correctly the system (Blanchard P. and al., on 1996). To analyse these two questions, we propose to study the modelling of physical events by differential equations. This study is based on two detailed analysis: the first one aims to understand the reasoning of scientists while dealing with classical problems from the XVIIth century ("Differential Equations Theory" being introduced to solve those problems) and especially how they built the model that enables to solve them. The second one deals with the "modelling of real systems" problem, encountered during the Terminal S* and DEUG Sma 1st year** classes of the actual French educational system. At the end of two analyses, we first determined the epistemological and current characteristics of modelling and we second tried to identify the role of formalisation and generalisation in modelling. The preceding analysis enables to answer questions Q1 and Q2 (see the beginning of the article) concerning the problems encountered while using differential equations for modelling, in mathematics and physics. - Modelling process is theoretical. - Only a few study has been dedicated to model validation, in particular concerning the comparison of real data with predicted behaviour. It should be added that the theoretical process is based on some well-known laws. Consequently, the steps of resolution of these problems are rather well-known and established. We believe that this way of doing contribute to the devalue of the concept of differential equation. That's why, we recommend, using an example, the experimental process in the teaching of mathematics and physics.