A statistical approach based on the modified Bartlett-Lewis rectangular pulses model is presented to disaggregate rainfall statistics from daily data. Six model parameters are estimated from 24- to 48-hour accumulated rainfall data. Based on these estimated parameters, in addition to reproducing 24- and 48-hour statistics, the model is shown to infer 1-, 2-, 6-, and 12-hour historical statistics satisfactorily. An upper limit for disaggregation scale (about 2 days) for this model has been identified. This characteristic behavior of the model is related to the power law dependence of the power spectrum for timescales smaller than 2 days. A detailed comparison between observed and modeled statistics of rainfall data is presented for two rain gages, one from central Italy and the other from the midwestern United States. ¿ American Geophysical Union 1994