An experimental study of the role of howling in wolf territory maintenance was conducted in the Superior National Forest, Minnesota. Vocal replies and behaviour of radio-collared wolves in response to human howls were analyzed for eight packs and 10 lone wolves during a 2-year period. Reply rate varied significantly throughout the year. A mid-winter increase was correlated with the breeding season, especially for groups containing breeding animals (alpha male or alpha female). A second, longer increase in reply rate started in midsummer, peaked about August, and declined to a low in early winter. The decline in autumn howling response occurred sooner in a pack whose pups developed faster. Through the year, the howling reply rate was significantly higher among all packs and lone wolves attending prey kills. The more food remaining at a kill, the higher the reply rate was. For wolves separated from their pack, the howling reply rate was dependent on their age and social role. Among adults, only alpha males ever replied alone, and their reply rate, and number of howls per session, exceeded those of other animals. Alpha males sometimes approached during howling sessions, whereas other adults usually retreated. Younger animals replied more often as pups than as yearlings, and then only during their first 7 months, after which they replied little more than most adults. Finally, larger packs replied more often than smaller packs. Specific behaviours noted during howling sessions, including movements away from the howler, indicated that howling was related to interpack agonism. In addition, three of the major factors influencing reply rate also significantly affect the level of agonism toward pack strangers : pack size, social role, and breeding season. The other two factors, kills and pups, are both important pack resources necessitating exclusive occupancy of a site. The high reply rates at sites containing kills or pups constitute strong circumstantial evidence that howling is important in territory maintenance. During howling sessions, wolves usually remained near their original site after replying, or retreated if they remained silent. This difference apparently was related to the problem of avoiding both accidental and deliberate encounters, and to cost/benefit considerations at the wolves' location. Howling was considered most effective in mediating avoidance in two situations : when two packs approached a common area of overlap, and when a pack returned to an area little used for weeks, in which scent posts would have lost effectiveness in deterring strangers. Both scent-marking and howling apparently are important in spacing. However, they differ in their roles and are complementary, with scent-marking being long-term and site-specific, and howling being immediate and long-range. Finally, lone wolves which do not possess territories, rarely replied, sharing the 'low-profile' behaviour expected of surplus animals in a territorial population.