We tracked 12 mule deer (Odocoileus hemionus hemionus) between February 1999 and April 2003 by using global positioning system (GPS) radiotelemetry in southeastern British Columbia to provide detailed information on migration and habitat use to local managers. We tested winter resource selection at the home-range and within-home-range scale to test a hypothesis that ungulate resource selection is scale-dependent. All sampled mule deer in this population migrated from low-elevation winter ranges to high-elevation summer ranges, supporting a hypothesis that migration is obligatory in mountainous, heavy-snow areas. We found little consistent selection at the within-home-range scale, but considerable selection at the home-range scale, supporting a scale-dependent hypothesis. Potential mule deer winter range could be predicted from 2 biophysical attributes, elevation and solar duration. Currently suitable winter habitat can then be further delineated on the basis of amount of mature coniferous forest within this zone. Use of GPS radiotelemetry increased sample intensity of individual deer, and thereby accuracy of individual parameter estimates. However, because of high equipment costs and failure rates, increased sample intensity occurred at the expense of sample size, and therefore illustrates a trade-off consideration for future work.