We characterized 131 summer, diurnal bed sites of 26 elk (11 bulls and 15 cows) in Custer State Park, South Dakota, from 5 June–30 August 1994, 1995 and 1996. Overstory canopy closure, number and basal area of trees, percent litter and bare ground were greater (P < 0.05) at bed sites than at random plots. North aspects were selected (P < 0.05). Microsite air temperature and percent of grass were lower (P < 0.05) at bed sites than at random plots. Hiding cover, wind speed, percent of forbs, shrubs, rocks, and wood, slope percent, average tree dbh, elevation, distance to roads, distance to trails, and distance to water were not different between bed sites and random plots (P > 0.05). Trees were present at 128/131 (97.7%) of bed sites (0.01 ha square plot), but occurred on only 41.2% (54/131) of random plots. An average summer, diurnal elk bed site had basal area >12.4 m²/ha, >110 trees/ha, >54% canopy closure on N aspects. Overstory canopy closure, tree basal area and microsite temperature correctly classified 86.2% of the observations, suggesting thermoregulatory factors influenced CSP elk use of summer, diurnal bed sites. Although elk are successful in some unforested areas despite the lack of suitable thermal cover, our data suggest that elk in the Black Hills prefer relief sites that provide thermal bed sites when available during the summer diurnal period. Management of appropriate thermal cover should be maintained in areas in which it exists.