The scaling properties of the time dynamics of self-potential data (in the frequency range between 0 and 0.125Hz) observed in 1995 at Acapulco station in Mexico were investigated. On the basis of detrended fluctuation analysis (DFA), which is a powerful method to detect scaling in nonstationary time series, deviations from uniform power-law scaling were identified and quantified. Our results suggest that an evident non-uniform scaling behavior in self-potential data could be related to the occurrence on September 14, 1995 of the strongest earthquake. These first results could be useful in the framework of earthquake prediction studies.