Regional patterns of crustal stresses in Brazil were studied with a detailed breakout analysis performed in 541 wells distributed throughout the country: 481 from basins along the continental margin, and 60 from intracratonic basins. A total of about 591 km of cumulative well length, digitally sampled at 0.15 m intervals, was analysed. Most intervals (80 per cent) are confined between −400 and −4000 m. Only wells that deviate less than 10° from the vertical were considered. For a given well (1) the mean orientation and the standard deviation of each interval with breakouts were calculated, (2) all intervals with mean breakout orientations less than 10° from the orientation of the hole deviation were discarded, (3) all intervals with standard deviations less than 12° were selected, (4) the selected breakouts were weighted by their interval length to calculate the mean orientation and standard deviation of the whole well, and (5) the final results were classified according to the World Stress Map criteria. Thus, 16 wells were classified as A (a total of 7.2 km of breakouts), 42 as B (7.1 km), 78 as C (7.4 km), 205 as D (3.3 km), 184 as E (0.7 km) and 16 were discarded. In most basins, the breakout orientations from different wells were usually consistent for qualities A, B and C, allowing a good estimate of the regional maximum horizontal stress (SHmax); wells classified as quality D showed a large scatter. The regional SHmax determined from breakouts is generally in good agreement with the available nearby focal mechanisms. The main trends of SHmax are (1) NW-SE, parallel to the coast, along the equatorial marginal basins, (2) E-W in the Alagoas Basin, (3) NNE-SSW, parallel to the coast, in the Sergipe and Reconcavo basins, and (4) NW-SE in the Middle Amazon Basin. In the equatorial and eastern continental margins, north of 15° S, breakouts and focal mechanism measurements indicate that the SHmax orientation is remarkably parallel to the coastline, following a 90° bend of the coast in northeastern Brazil. Because the theoretically predicted intraplate stresses in northern Brazil trend about WNW to ENE (e.g. Meijer 1995; Coblentz & Richardson 1996), we interpret our observations as indicating that local sources of stress at the continental margins (e.g. flexural stresses and lateral density contrasts) dominate the plate-wide stresses and may have been underestimated in theoretical stress models of the South American plate. The new stress data (136 A-C quality out of a population of 541 wells) cover a region of mid-plate South America where only 23 data points had been compiled previously in the World Stress Map database. The more detailed observed patterns of the intraplate stress field should be helpful in better constraining future models of the forces driving the South American plate.