Reduced-impact logging (RIL) and conventional techniques (CNV) were compared in a mixed dipterocarp hill forest in East Kalimantan in three blocks of about 100 ha each. Damage was evaluated using pre- and post-harvesting assessments in 24 one-hectare sample plots. RIL techniques nearly halved the number of trees destroyed (36 vs 60 trees/ha). RIL's main benefit was in the reduction of skidding damage (9.5% of the original tree population in RIL vs 25% in CNV). Before logging, mean canopy openness in CNV (three plots only) and RIL (9 plots) was similar (3.6 and 3.1%). and not significantly different (y2 = 2.73, P = 0.254). After logging, the mean canopy openness was 19.2% in CNV (n = 9 plots) and 13.3% in RIL (n = 8 plots), and the distributions of the canopy class in RIL and CNV significantly different (y2 = 43.56, P < 0.001). CNV plots showed a higher proportion of measurements in the most open class >30% than in RIL. At a larger scale, the area of skidtrail per unit timber volume extracted was halved in the RIL compartment (15 m2 vs 27 m2 m-3 for CNV). However, under high felling intensity (>8 trees/ha), both stand damage and canopy disturbance in RIL approached those recorded in CNV under low or moderate felling regime. Over this felling intensity threshold the effectiveness of RIL in reducing tree damage is limited. In mixed dipterocarp forest where harvestable timber density generally exceeds 10 trees/ha, a minimum diameter felling limit is clearly insufficient to keep extraction rates below 8 trees/ha. Based on these new results and previous studies in Borneo, we suggest three silvicultural rules: (1) to keep a minimum distance between stumps of ca. 40 m, (2) to ensure only single tree gaps using directional felling, (3) to harvest only stems with 60-100 cm dbh. Foresters, policy makers and certifiers should consider these as criteria for sustainable forest management. We emphasise the need to expand harvesting studies to look at impacts and trade-off