Anderson (1949) has reviewed in detail the currently available techniques for the study of introgression. He has indicated (Anderson 1954, 1956, 1957) that the use of metroglyphs (the "pictures" used in pictorial scatter diagrams) is the most efficient means of analyzing suspected examples of introgression. Earlier Anderson (1939) presented a genetic rationale upon which such graphical techniques may be based. To summarize Anderson's reasoning briefly (Anderson, 1939, 1949, 1953), the use of metroglyphs for introgression analysis is based upon the assumption that the presumably quantitatively inherited genes affecting taxonomic characters are linked sufficiently strongly to cause graphically detectable correlations among these characters in the offspring of species hybrids. Actually, correlations among taxonomic characters can result from several different mechanisms. The evidence that linkage is the usual cause is very limited. [ Grant (1950, 1954, 1956), Clausen and Hiesey (1958), Hiesey et al. (1963), and Nobs et al. (1964) all present similar types of data which suggest, but do not prove, that substantial taxonomic correlations can be caused by linkage.] There is both experimental evidence (Collins and Kempton, 1920; Kearney, 1923, 1926; Klekowski, 1964) and theoretical evidence (Dempster, 1949) that linkage does not account for the establishment of such correlations. Furthermore, there are theoretical reasons why linkage cannot account for the maintenance of such correlations if they should be established by some other means (Lush, 1948). At least two factors other than linkage can cause correlations among taxonomic traits. First, natural selection may effectively eliminate recombinant types, with an apparent strengthening of any linkage present. Second, if the habitat of a well-established species is subject to a limited amount of immigration from a second species with which the first species is cross-fertile, the resulting population will be in a state of disequilibrium. In such a case, plants with genotypes intermediate between or similar to those typical of the two parental species are more frequent than those with genotypes combining different features of the two parental species (recombinant types). Although Anderson has assumed that the graphical methods devised for the study of introgression are based upon correlations caused by a lack of free recombination among loci affecting taxonomic traits, the same methods should also be valid in the two cases discussed above. The purpose of this study was to reinvestigate the degree of multiple correlation (presumably caused by linkage) among taxonomic characters in the segregating generations which follow interspecific hybridization and to compare this with the degree of multiple correlation caused by continuous crossing. The validity of the 1 Contribution from the Department of Genetics, North Carolina Agricultural Experiment Station, Raleigh, North Carolina. Published with the approval of the Director of Research as Paper No. 2044 of the Journal Series. Part of a thesis submitted by the author in partial fulfillment of requirements for the Ph.D. degree. The work is part of a research program supported by National Science Foundation Grant No. G14203. The statistical computations were conducted as a part of research programs supported by National Institute of Health Grants No. GM-10494 and No. FR-0001 1. 2 National Science Foundation Cooperative Graduate Fellow. Present address: Institute of Genetics, School of Agriculture, University of So Paulo, Piracicaba, Sao Paulo, Brazil.