An extensive programme of commercial afforestation in south-central Chile provides a unique opportunity to extend the field-based understanding of forest impact on floods to the large catchment scale (100–1000 km2). Until now, much of that understanding has been limited to small catchments (<10 km2). Data for an approximately 50-year period are analysed systematically for four catchments (434–1545 km2) with 18%–54% increases in forest cover. For each catchment, the forest impact is quantified by pairing annual maximum flood frequency curves and peak discharge/storm rainfall relationships for the pre- and post-afforestation periods. These periods are themselves defined using land use surveys and cumulative double mass curves for runoff and rainfall. Scaling the frequency curves with a discharge corresponding to mean annual rainfall normalizes for rainfall differences and thus isolates the effect of forest cover. Partial afforestation rates of up to 50%, while markedly affecting annual evapotranspiration and runoff, have less certain effects on peak discharges. Downward shifts in the flood frequency curves occur following afforestation but appear more readily explained by changes in rainfall regime. These changes include a small decline in annual rainfall and a downward shift in some of the frequency curves for the storm rainfall totals corresponding to the annual maximum discharges. Nevertheless, some forest impact cannot be ruled out and the effect of a 100% change in forest cover on flood response remains unknown. The results suggest complexity in large catchment flood response and provide only weak support for afforestation with the sole purpose of reducing flood peaks at the large catchment scale. This is a significant conclusion, given that it challenges the largely uncritical view among the public, governments and development agencies worldwide that forests prevent floods.