site index of natural jack pine (pinus banksiana lamb.) in the thunder bay area was related to features of soil and topography using multiple regression analyses. site index (si, height of jack pine trees at 50 years from breast height) was used as the dependent variable and 129 soil and topographic variables were considered for the analyses. due to problems associated with numerous independent variables, multicollinearity, and highly variable soil groups, the four preliminary regression equations computed from 95 plots could not be accepted as valid equations. the final analyses limited the number of independent variables considered for analyses and landform types were more precisely defined. the "best" final regression equations explain 83, 65, 65, and 75 percent of the variation in si for bedrock, morainal, glaciofluvial, and lacustrine landforms, respectively. the variables included in the bedrock equation are depth to bedrock (dbr), and coarse fragment content of the a horizon (cfraga). the variables included in the moraine equation are depth to a restricting layer (drl), percent clay in the a horizon (cla), and coarse fragment content of the c horizon (cfragc). the variables included in the glaciofluvial equation are depth to a moist restricting layer (dmrl), and percent slope (slope). the variables included in the lacustrine equation are thickness of the a horizon (tha), and ph of the bc horizon (phbc). the final regression equations for bedrock, moraine, and glaciofluvial sites predict si reasonably well for 15 check plots.