We found that the attenuation functions proposed for southern California by Richter (1958) and Hutton and Boore (1987) are not entirely suitable for local magnitude calculation for earthquakes in northern Baja California. In this study, we used zero-to-peak amplitudes from synthetic Wood-Anderson seismograms to determine appropriate − log A0 attenuation functions and station magnitude corrections through a multiple linear regression analysis. Our data were grouped so that each energy path was entirely contained in either the granitic Peninsular Ranges of Baja California (PRBC) or the sedimentary environment of the Mexicali-Imperial Valley (MIV). The best solutions of the regression analysis are -log A0 = (1.0134 ± 0.0288) log (r/100) + (0.0025 ± 0.0005) (r − 100) + 3.0 for MIV and −log A0 = (1.1319 ± 0.0197) log (r/100) + (0.0017 ± 0.0002) (r − 100) + 3.0 for PRBC, where r is hypocentral distance. A comparison of these functions with those for southern California revealed the following: (a) a remarkable agreement between Richter's function and the function for MIV at distances between 40 and 450 km, (b) Hutton and Boore's function and the function for PRBC are quite similar in the whole distance range considered (10 to 450 km), and (c) for both areas, at distances shorter than 50 km, the −log A0 Richter function was as much as 0.5 units lower than the others. Thus, as for other regions, the use of Richter's function results in understimation of local magnitudes at near-source distances. Additionally, the attenuation function for PRBC was 0.30 units lower than Richter's at 400 km, which means that at such longer distances the Richter function overestimates the magnitude of PRBC earthquakes. Finally, station magnitude corrections from −0.42 to 0.26 units were obtained, suggesting significant effects due to station site properties.