Abstract The Manteigas-Bragança fault is a major, 250-km-long, NNE-striking, sinistral strike-slip structure in northern Portugal. This fault has no historical seismicity for large earthquakes, although it may have generated moderate (M5+) earthquakes in 1751 and 1858. Evidence of continued left horizontal displacement is shown by the presence of Cenozoic pull-apart basins as well as late Quaternary stream deflections. To investigate its recent slip history, a number of trenches were excavated at three sites along the Vilariça segment, north and south of the Douro River. At one site at Vale Meão winery, the occurrence of at least two and probably three events in the past 14.5 ka was determined, suggesting an average return period of about 5–7 ka. All three events appear to have occurred as a cluster in the interval between 14.5 and 11 ka, or shortly thereafter, suggesting a return period of less than 2 ka between events within the cluster. In the same area, a small offset rill suggests 2–2.5 m of slip in the most recent event and about 6.1 m after incision below a c . 16 ka alluvial fill event along the Douro River. At another site along the Vilariça River alluvial plain, NE of the Vale Meão site, several trenches were excavated in late Pleistocene and Holocene alluvium, and exposed the fault displacing channel deposits dated to between 18 and 23 ka. In a succession of closely spaced parallel cuts and trenches, the channel riser was traced into and across the fault to resolve c . 6.5 m of displacement after 18 ka and c . 9 m of slip after c . 23 ka. These observations yield a slip rate of 0.3–0.5 mm/a, which is consistent with earlier estimates. Combining the information on timing at Vale Meão winery and displacement at Vilariça argues for earthquakes in the M7+ range, with coseismic displacements of 2–3 m. This demonstrates that there are potential seismic sources in Portugal that are not associated with the 1755 Lisbon earthquake or the Tagus Valley, and, although rare, large events on the Vilariça fault could be quite destructive for the region. This work provides an analogue for the study of active faulting in intracontinental settings and supports the view that earthquakes within intracontinental settings tend to cluster in time. In addition, this study highlights the usefulness and application of multiple field, remote sensing and geochronological techniques for seismic hazard mitigation.