Abstract Studies of genetic variation at allozyme loci, assumed to be selectively neutral, have provided valuable insights into the genetic structure of numerous populations. The degree to which population structure of allozyme variation reflects that of quantitative traits, however, is not well resolved. Here, we compare estimates of population differentiation (FST) of 11 populations for allozymes with those for nine discrete and nine continuous morphological traits. Overall, the allozymes have the lowest FST estimates, indicating relatively little population differentiation. Excepting two traits, petal width and long internode length, the continuous morphological traits have estimates similar to those from allozymes. The discrete morphological traits tend to have the highest estimates. On a single trait basis, estimates of FST for four discrete and two continuous traits are higher than those for allozymes. A more detailed (narrow-sense quantitative) genetic study of two populations suggests that these estimates of FST may underestimate the true value because of dominance. Clustering analyses show that the pattern of differentiation for the discrete morphological traits strongly reflects the geographical distribution of the populations, whereas the patterns for the continuous traits and allozymes do not. These results suggest that selection has been occurring on the discrete morphological traits, selecting toward a common optimum within each geographic group, and optima differing among geographic groups.