1. Numbers of gypsy moth larvae feeding on each of 922 randomly sampled trees in a Quercus-Acer-Fraxinus forest in southwestern Quebec, Canada were counted in 1979 and in 1980 to quantify the larval feeding preferences as observed in the field for eighteen deciduous and one coniferous tree species at the northern range limit of the gypsy moth.2. Both the diameter at breast height (dbh) and the estimated foliage biomass of the sampled trees were used to calculate the relative proportions of foliage represented by each of the nineteen tree species in the forest canopy. With these data on availability and utilization of the tree species by the gypsy moth larvae an Ivlev-type electivity index was used to quantify the larval feeding preferences. These preferences observed in the field define the susceptibility of a tree species to attack by the gypsy moth.3. The feeding preferences calculated using estimated foliage biomass were comparable to the simpler calculation based on dbh (Spearman's rho = 0.79; P= 0.0001). The dbh-based feeding preferences remained almost unchanged in 1979 and 1980 (Spearman's rho=0.83;P=0.0001).4. The composite 1979-80, dbh-based feeding preferences show Quercus rubra, Populus grandidentata, Ostrya virginiana, Amelanchier spp. and Acer saccharum were preferentially attacked by gypsy moth. Prunus serotina, Betula lutea, Acer rubrurn, A.pensylvanicum, Fraxinus americana, Urnus rubra, P.pensylvanicum and B.papyrifera were avoided. All nineteen tree species were, however, utilized to at least some degree by gypsy moth larvae. 5. These results quantitatively affirm and clarify earlier reports of gypsy moth feeding preferences in North America and Eurasia. The advantages and limitations of using an electivity index to estimate the susceptibility of different tree species to attack by folivores like the gypsy moth are discussed.