A complete diallel cross was made among eight homozygous barley genotypes. Four of the parents were commercially grown cultivars and the remaining four consisted of isogenic derivatives of a California cultivar, Atlas. In a replicated experiment grown in two seasons, measurements were recorded on days to heading, days to maturity, plant height, grain yield and seed size. These data were then used to estimate variances (Vr) and covariances (Wr) of arrays in each season and a genetical interpretation of the relations between these statistics was sought. For days to heading, days to maturity and plant height, the Wr/Vr graphs provided an evaluation of genetic relationship among the parental genotypes. As expected on the basis of their known genetic similarity, the isogenic lines of Atlas were close together on the Wr/ Vr graph. Contrary to the expectation, however, remarkable similarity among the commercially grown genotypes was indicated by the proximity of the array points representing them. The patterns of distribution of the array points were such that the eight parents could be classified into two wellseparated groups, one comprising the isogenic lines which carried most of the dominant alleles and the other comprising the cultivars carrying largely the recessive alleles. A seasonal difference in the genetic behaviour of the parents was also observed. Although some of the observed genetic differences between the two parental groups in one season disappeared in the following season, the positions of the array points were consistent in depicting substantial differences between the groups in both seasons. In most cases a graphical analysis was possible when a number of parents were removed from the diallel table. Non-random distribution of genes among the parents did not appear to be a common cause of disturbance in the relations between Wr and Vr even when the number of parents were very small. For two economically important characters, kernel size and grain yield, a graphical analysis was not possible, presumably because of the failure of a number of hypotheses underlying the analysis.