We present a new hierarchical clustering criteria which canbe applied to data set. This is done after generating an initialpartition byusing a Topological Self Organizing Map. This criteriacontains two terms which take into account two dierent errorssimultaneously: the square error of the entire clustering (as theWard criteria) and the topological structure given by the SelfOrganizing Map. A parameter T allows to control the corresponding inuence of these two terms. Results on simulated data are presentedwhich show the eect of this criteria for different values of T.