Background/Question/Methods Once a major component of the Upper Great Lakes forests, contemporary eastern hemlock (Tsuga canadensis) now occupies a mere fraction of its former abundance and is often found in small isolated patches in a matrix of sugar maple (Acer saccharum). These hemlock stands, along with other conifer patches in the region, have been shown to be important winter habitat for white-tailed deer (Odocoileus virginianus) on the leeward side of Lake Superior, where lake-effect snowfall can range from 2.5 to 6 meters annually. High winter deer densities, however, are thought to be the primary bottleneck to hemlock recruitment, though across the landscape some hemlock stands appear to be experiencing disproportionately higher deer densities and more intense browse pressure than others. The objective of this research was to determine how snow depth, hemlock stand structure, and landscape composition and configuration influence habitat selection by wintering deer. We used a stratified random sampling design to select 39 remnant hemlock stands from across the western Upper Peninsula of Michigan. For this three-year study deer fecal pellet group counts were used to quantify relative deer densities for each stand. We quantified hemlock stand structure and composition, and landscape composition metrics were derived for varying buffer widths surrounding each stand. Linear regression analyses were used to identify factors, both locally and across the landscape, that influence deer winter habitat selection. Results/Conclusions In areas of highest snow depths, deer used stands of all sizes (0.4-60 ha), whereas only larger stands were selected in relatively lower snow depths. Percent hemlock overstory and basal area had little influence on yard selection; however, there was a significant correlation when combined with the availability of non-hemlock browse: sugar maple (Acer saccharum) and red maple (Acer rubrum). Winter severity appears to be influential to yarding behavior; 75% of hemlock stands were utilized during the course of this study. These results suggest that deer response to the availability of sugar maple as browse is likely a short-term positive feedback that enhances hemlock stand selection, but will ultimately result in the loss of winter cover due to a negative feedback on hemlock regeneration.