A local-magnitude scale is derived for northwestern Italy from wave-form data recorded at 17 stations from 2600 local earthquakes, ranging in distance from 10 to 300 km. By averaging the horizontal components in a single measure, we used 8127 zero-to-peak amplitudes from synthetic Wood–Anderson seismograms to determine, in a least-squares sense, the appropriate –log A0 attenuation function, the event local magnitude, and the station corrections. Both a parametric and a non-parametric description of –log A0 is considered while performing the inversion. In both cases, the constraint of 1-mm motion recorded at 100 km for a magnitude 3.0 earthquake was used. The resulting parametric distance correction is given by –log A0 = 1.144log(r/100) + 0.00476(r – 100) + 3. The remarkable agreement between the parametric and nonparametric results confirms that the assumptions on the attenuation function that we made for deriving the parametric distance correction are reasonable. Moreover, inversion of bootstrap replications of our data set furnished stable solutions. Station magnitude corrections range between –0.28 and 0.48, suggesting a variable and significant effect of station site properties on recorded amplitudes. Finally, the local-magnitude scale has been used to evaluate the magnitude values for the whole catalog of earthquakes recorded by the IGG–University of Genoa network from 1996 until March 2001 (about 3200 events) and for a subset relevant to the southwestern Alps (about 1900 events). Statistics performed using the new local-magnitude scale and the usual duration magnitude demonstrate the influence of this choice on both completeness and b-value estimation. For example, in the southwestern Alps, the completeness threshold decreases from 2.0 (Md) to 1.7 (ML) and the b-value from 1.38 (Md) to 1.08 (ML).