Barrier effect is a road-related impact affecting several animal populations. It can be caused by behavioural responses towards roads, associated emissions and/or circulating vehicles. Most studies so far have described road-effect zones along major roads, without determining the actual factor inducing the behavioural response. The purpose of the present study was to assess the factors potentially causing road-effect zones in a heterogeneous road network and eventually to estimate the reduction of habitat quality imposed by roads within the protected area of Doñana Biosphere Reserve, Spain. As model species, two ungulates were used: red deer and wild boar. The presence of both species was surveyed along 200-m transects. All transects started and were perpendiculars to reference roads, often intersecting unpaved minor roads with virtually no traffic. The presence probability of both species was mainly affected by the distance to the nearest road, but also by the proximity to reference roads. Red deer presence was also affected by the traffic volume of the nearest reference road. At a regional scale, the overall road network within the protected area imposes a reduction in presence probability of 40% for red deer and 55% for wild boar. A road network optimization, decommissioning unused and unpaved roads, would re-establish almost entirely the potential habitat quality. Both study species avoided roads regardless of their surface or traffic volume, suggesting a response due to gap avoidance which may be based on the association between linear infrastructures and the possibility of vehicles occurring along them. The overall behavioural response can substantially decrease habitat quality over large scales, including the conservation value of protected areas. For this reason, it is recommended road network optimization by road decommissioning to mitigate the impact of roads at a regional scale, with potential positive effects at ecosystem level. informacion[at]ebd.csic.es: D'Amico et al (2016) Road avoidance responses determine the impact of heterogeneous road networks at a regional scale. J Appl Ecol 53: 181–190; DOI: 10.1111/1365-2664.12572