Aims. We have re-examined the published rotation measures (RMs) of extragalactic point sources and pulsars with |b| < 3 • to study the magnetic field in the fourth Galactic quadrant. Methods. We reduced the influence of structure in electron density as much as possible by excluding objects for which Hα-data indicate large fluctuations in n e somewhere along the line of sight. We also excluded objects for which the RM may have been significantly "corrupted" by an intervening supernova remnant. We modeled RM(l), the longitude dependence of RM of the unaffected extragalactic sources and pulsars. We assumed several geometries for the large-scale field. All but one of those are based on logarithmic spiral arms (with various pitch angles and widths), while one has circular symmetry. We also made different assumptions about the large-scale n e -distribution. Results. The data suggest the following generic behaviour of the large-scale field in the 4th Galactic quadrant. The field is most likely organized along logarithmic spiral arms and shows two significant reversals: from the Norma arm (CCW field) to the Norma-Crux interarm region (CW field), and from the Norma-Crux interarm region to the Crux arm (CCW field). The present data do not constrain the field in and beyond the Crux-Carina interarm region. Although the models give a good description of the global character of RM(l), individual RM-estimates deviate by typically 15 times their measurement errors. We argue that these large deviations are most likely due to the "small-scale" field that dominates on scales of up to several hundred pc. Conclusions. The picture that emerges is thus of a field that has significant structure on smaller scales, but for which the average values in arms and interarm regions are nevertheless well-defined. In addition, this smaller-amplitude large-scale field appears to reverse at each arm-interarm boundary that we can study with the present data. We briefly discuss the link between these results and theoretical predictions.