Predicting species' responses to habitat loss is a significant challenge facing conservation biologists. We examined the response of both European three-toed woodpecker subspecies Picoides tridactylus tridactylus and P. tr. alpinus to different amounts of dead wood in a boreal and a sub-Alpine coniferous forest landscape in central Sweden and Switzerland, respectively. Habitat variables were measured by fieldwork in forests with breeding woodpeckers (n=10+12) and in control forests without breeding woodpeckers (n=10+12) in the same landscape. Logistic regression analyses revealed steep thresholds for the amount of dead standing trees and the probability of three-toed woodpecker presence in both Sweden and Switzerland. The probability of the presence of three-toed woodpeckers increased from 0.10 to 0.95 when snag basal area increased from 0.6 to 1.3 m²ha⁻¹ in Switzerland and from 0.3 to 0.5 m²ha⁻¹ in central Sweden. In Switzerland, a high road network density was negatively correlated to the presence of woodpeckers (r=−0.65, p=0.0007). The higher volumes of dead wood in Switzerland, where population trends are more positive, than in central Sweden, where the population is declining, would suggest that the volumes of dead wood in managed forests in Sweden are too low to sustain three-toed woodpeckers in the long-term. In terms of management implications, we suggest a quantitative target of at least 5% of standing trees in older forests being dead over at least 100 ha large forest areas. This corresponds about to ⩾1.3 m²ha⁻¹ (basal area) or ⩾15 m³ha⁻¹ (volume), still depending on site productivity.