Genetic variation in wood properties among and within 3 provenances of balsam poplar was investigated. Between 1982 and 1984, clonal populations were sampled along the Longitude 90 degrees W in North Wisconsin (Lat. 45 degrees N to 46 degrees N); Thunder Bay, Ontario (Lat. 48 degrees N to 49 degrees N); and Pickle Lake, Ontario (Lat. 50 degrees N to 51 degrees N). Rooted cuttings were planted in a field test near Lakehead University, Thunder Bay. In 1994, 30 clones from each provenance, with 4 ramets per clone, were measured for growth characteristics, and specimen disks were cut at tree base. Ring width, relative density, percent moisture content, fiber length, and vessel element length were determined in the laboratory. Univariate analyses of variance showed significant differences among the 3 provenances in growth rate and cell length. The southern provenance had the fastest growth rate and the longest cells. Provenance differences in relative density and moisture content of the wood were not statistically significant. Canonical multivariate analysis, using growth rate, relative density, and fibre length as dependent variables, showed differences between the southern and northern provenance, with the local source in an intermediate position. Both genetic and environmental variances for a certain trait differed from provenance to provenance. Therefore the estimates of broad sense heritability were different in each provenance. Heritability was more uniform and higher for wood properties than for growth characteristics. Positive phenotypic correlation between growth rate and cell length was found. Genetic correlations and coefficients of genetic prediction showed relative genetic independence of growth characteristics from relative density and moisture content. Results justified selection based on growth characteristics, wood properties, or a combination of these two.