This paper is devoted to a dynamic model for observational study,which does not depend on statistical concept nor on any a priori theoretical model of the object.Consequently,it pocesses great generality.Recursive similarity is the central concept of the research,resulting from which are neighborhood function,jumping,evolution similarity and calculable index,responsible for describing the evolution properties of the observational sequence.As a target,the reconstruction of a dynamic system based onan observational(vectorial or functional)series and the article is only part of the research towards the goal.