We critically analyze the results on seismic intensity attenuation in Italy derived by Albarello and D’Amico (2004) and Gasperini (2001). We demonstrate that, due to the inadequacy of certain underlying assumptions, the empirical relationships determined in those studies did not best reproduce the decay of intensity as the distance from the source increases. We reconsidered some of the relevant concepts and assumptions used in these intensity-attenuation studies (macroseismic epicenter, epicentral intensity, data completeness) to suggest some useful recipes for obtaining unbiased estimates. In particular, we suggest that (1) data for distances from the source at which an intensity below the limit of diffuse perceptibility (≤IV) is expected should be excluded from attenuation computations because such data are clearly incomplete, (2) attenuation equations that include a term proportional to the epicentral intensity I0 with a coefficient different from 1.0 must not be used because they imply a variable offset between I0 and the intensity expected at the epicenter, and (3) epicentral intensities must be recomputed consistently with the attenuation equation because those reported by the Italian catalog do not generally correspond with the intensity predicted at the epicenter by the attenuation relationships so far proposed. Following these suggestions produces a significant reduction in the standard deviation of the model that might lead to a corresponding reduction of the estimates of seismic hazard.