Long range dependence and regime switching are very intimately related effects. In this paper we consider the problem of spuriously detecting breaks point in hypothesis of long memory data generating processes. For this purpose, we address the issue of estimating the number of breaks using several techniques, namely, the information criteria, Bai and Perron’s sequential selection procedure (Econometrica 66:47–78, 1998), and the automatic procedure of Lavielle (J Financ Econ 2:290–318, 2004). By means of Monte Carlo experiments, we investigate the effect of increasing the long memory parameter on selecting the number of breaks and their locations, and show that the Lavielle’s method is the best technique since its frequency of choosing the true number of changes is the highest particularly when the order of integration is close to 0.5. As it seems that inflation rates contains long memory and structural breaks, an application to the U.S. inflation process is presented to illustrate the usefulness of these procedures. The results show that the Lavielle’s method (J Financ Econ 2:290–318, 2004) selects only 2 breaks, however, the number of breaks detected by the information criteria and the sequential procedure of Bai and Perron (Econometrica 66:47–78, 1998) are superior or equal to three.