In Table 1 of our original paper we include the coordinates of the centers of the 3.7' long slits obtained in the observations. However, these do not correspond to the centers of the slits drawn in our Figure 1, which represents the extension of the portions of the slits that we finally analyze in the paper (we remove some sections, as indicated in Table 2). The central coordinates of these ``effective'' slits are given in Table E1 below. In § 3 we indicate that the values of the reddening coefficient we obtain are somewhat low, lower than the values found in most of the literature (e.g. Esteban et al.1998 ApJ, 675, 389 [2008]; O'Dell & Yusef-Zadeh 2000 ApJ, 675, 389 [2008]), but we do not discuss this fact in the paper. In particular, we find that our c(Hβ) values are in some regions about 0.3 or 0.4 dex lower than those found by O'Dell & Yusef-Zadeh for the same areas. We have investigated the reason of that discrepancy, and find that the data do not apparently suffer for flux calibration problems. The method we used involved measuring the Hβ and Hγ flux ratios and comparing them with the theoretically expected values. This method demands very precise spectrophotometry, because one is using observations over a short wavelength base to determine reddening corrections over a much longer wavelength range. We have not found any error in our analysis, and hence cannot explain the differences in the extinction when compared with earlier results. However, the results of our paper are not particularly sensitive to the reddening correction. We have calculated the effect on the T([N II]) and T([O III]) distributions of increasing the c(Hβ) by a constant value (0.3 and 0.7 dex) in all our slit positions, and the changes are not relevant for our conclusions. Obviously, the density maps are largely unaffected. In the case of an increment of 0.3 dex in c(Hβ), the temperatures change somewhat-about 100 K in T([N II]), and considerably less for T([O III])-for a given point, but the point-to-point variations are identical. The spatial behavior the abundance discrepancy factor, the main concern of the paper, is also not substantially affected, and its variation remains within the observational errors. We are very grateful to C. R. O'Dell for his careful reading of our paper and for indicating the points discussed and corrected here, as well as the anonymous referee of this erratum.