Few studies have been made on the applicability of classical quantitative genetic theory to sugar cane, and few studies have examined the efficiency of selection systems used in sugar cane. For this thesis, five experiments were conducted to investigate various aspects involved in the estimation of genetic variance components in sugar cane, Including the applicability of the classical theory. These experimentsalso provided information on the efficiency of selection systems in original seedlings. A sixth experiment examined selection in 30-sett plots, with particular reference to the benefits conferred by weighing the plots.
Many of the assumptions underlying classical genetic theory are unlikely to be valid in sugar cane. The implications of non-validity were tested in Experiment II which allowed comparison of independent estimates of genetic variances derived under different genetic assumptions. The character Brix was measured with good statistical precision, and the estimates of genetic variances suggested that the effect of non-validity of the assumptions was limited. For other characters, measurement was less precise, and greater differences occurred between estimates of genetic variances. However, the differences were not statistically significant.
Additive genetic variance was more important than dominance genetic variance for m s t characters. However, substantial dominance genetic variance was shown for tons cane per acre and tons Brix per acre in two experiments. Epistatlc genetic variance was apparently small.
In general, differences between reciprocal crosses were small. Maternal effects were unimportant except in Experiment IV in which they could be attributed to the effect of accidental self-pollination.
Cross X environment interactions were generally small in all experiments in which they were estimated. However, dominance genetic variance interacted with years for several characters in Experiment 1. The male parent x year interaction was significant for all characters ln Experiment II, but this was attributed to years being completely confounded with crop class (plant or ratoon); that is, the interaction was due to differential behaviour of male parents in the plant and ratoon crops.
Estimates of variance components from plant and ratoon crops were similar for Brlx, but differed for the other characters. The differences were attributed to the Interaction between male parents and crop class. Estimates of variance components from seed-planted and sett-planted material were found to be comparable for most characters. These data suggested that estimates from seed-planted material could be applied to sett-planted populations, whereas estimates from plant crops may not be applicable to ratoon crops.
Withln-family competition had a major effect on the estimation of both environmental and genetic variance components, particularly in the ratoon crop of Experiment II. Many negative estimates of variance components were obtained, and most of these could he attributed to the effect of competition. In particular, many estimates of dominance genetic variance were biased negatively.
Herltability and degree of genetic determination on an Individual basis in original seedling populations were low ( <0.3) for all characters except Brix which had a value of approximately 0.5. For 30-sett plots, the degree of genetic determination for yield relative to a standard was 0.5, Indicating that selection for yield in these plots should be more effective than in original seedlings.
Few characters exhibited a high degree of genetic variability as measured by the genetic coefficient of variability. For most characters, the coefficient was less than 20 per cent and, for Brix, it was only six to eight per cent. Thus, although Brix has a high degree of genetic determination, progress from selection will be slight due to the restricted genetic variability in the population.
Genetic, phenotypic, and environmental correlations between important characters were estimated, and their implications in selection were discussed. The important genetic correlation between Brix andyyield of cane was variable, but was concluded to be small and negative, in general.
A simulated selection experiment on a population of original seedlings was conducted using the results of Experiments I and II. Mass selection based on a phenotypi: seore (called a grade) was the most effective method tested.
In Experiment VI, weighing of 30-sett plots was compared with visual estimation of yield. Weighing of plots resulted in a gain from selection for grade of approximately five per cent. Many elite clones that would have been discarded on the basis of visual estimation of yield in 30-sett plots were retained by selection based on weights.