The state of stress in the vicinity of old continental rifts is examined to investigate the possibility that crustal structure associated with ancient rifts (specifically a dense rift pillow in the lower crust) may modify substantially the regional stress field. Both shallow (2.0-2.6 km depth) breakout data and deep (20-45 km depth) crustal earthquake focal mechanisms indicate a N to NNE maximum horizontal compression in the vicinity of the Paleozoic Amazonas rift in central Brazil. This compressive stress direction is nearly perpendicular to the rift structure and represents a ???75?? rotation relative to a regional E-W compressive stress direction in the South American plate. Elastic two-dimensional finite element models of the density structure associated with the Amazonas rift (as inferred from independent gravity modeling) indicate that elastic support of this dense feature would generate horizontal rift-normal compressional stresses between 60 and 120 MPa, with values of 80-100 MPa probably most representative of the overall structure. The observed ???75?? stress rotation constrains the ratio of the regional horizontal stress difference to the rift-normal compressive stress to be between 0.25 and 1.0, suggesting that this rift-normal stress may be from 1 to 4 times larger than the regional horizontal stress difference. A general expression for the modification of the normalized local horizontal shear stress (relative to the regional horizontal shear stress) shows that the same ratio of the rift-normal compression relative to the regional horizontal stress difference, which controls the amount of stress rotation, also determines whether the superposed stress increases or decreases the local maximum horizontal shear stress. The potential for fault reactivation of ancient continental rifts in general is analyzed considering both the local stress rotation and modification of horizontal shear stress for both thrust and strike-slip stress regimes. In the Amazonas rift case, because the observed stress rotation only weakly constrains the ratio of the regional horizontal stress difference to the rift-normal compression to be between 0.25 and 1.0, our analysis is inconclusive because the resultant normalized horizontal shear stress may be reduced (for ratios >0.5) or enhanced (for ratios <0.5). Additional information is needed on all three stress magnitudes to predict how a change in horizontal shear stress directly influences the likelihood of faulting in the thrust-faulting stress regime in the vicinity of the Amazonas rift. A rift-normal stress associated with the seismically active New Madrid ancient rift may be sufficient to rotate the horizontal stress field consistent with strike-slip faults parallel to the axis of the rift, although this results in a 20-40% reduction in the local horizontal shear stress within the seismic zone. Sparse stress data in the vicinity of the seismically quiescent Midcontinent rift of the central United States suggest a stress state similar to that of New Madrid, with the local horizontal shear stress potentially reduced by as much as 60%. Thus the markedly different levels of seismic activity associated with these two subparallel ancient rifts is probably due to other factors than stress perturbations due to dense rift pillows. The modeling and analysis here demonstrate that rift-normal compressive stresses are a significant source of stress acting on the lithosphere and that in some cases may be a contributing factor to the association of intraplate seismicity with old zones of continental extension.