This work uses the Hodrick and Prescott filter (HP), the Band-Pass filter (BP), and a theoretical decomposition (TD), to show how the cyclical regularities characterized by the model of exogenous growth of King, Plosser and Rebelo (1988) depend on the filter technique used. This article exhibits statistically significant differences between properties of the cyclical components isolated with the HP, BP and TD filters; in particular between the HP and BP filters, which differs from the results of Canova (1998). It also shows that the TD filter technique is not consistent with the definition of economic cycle given by the National Bureau of Economic Research (NBER).