This paper describe some problems related to the evaluation of hu- man communicative behaviour. Along these Eines, a solution of sorts is sought to the subject through un attempt ut combining the two existingpa- radigms. This new íype is called referential-ecological analysis and its most original contribution is the notion of including the guidance of the adult experimenter as part of the communicative exchange. A detailed ca- tegorization system is presented for the comrnunicative behaviour of the speaker, the receiver and the adult, justified by the previous contributions of various authors. A study of the reliability of the encoders is also presented. Finally, un example of this new encoding system of the communi- cative exchange is displayed.