Optimum number of lines to retain during inbred line development in maize

The objectives of this investigation were to: 1) compare several stochastic genetic models and select one that describes the response of grain yield to inbreeding and selection in maize. 2) and compare selection schemes to determine one which maximizes genetic gain. A FORTRAN program was written to compare nine genetic models containing various combinations of gene effects. The genotype of each line was represented by two vectors, each with a dimension of 100, representing 100 loci. Allelic values of one, or, zero were allocated as the elements of the vectors at random. An environmental effect was generated assuming a heritability of 0.20 and allocated to genotypes at random. Among models compared a model consisting of 30, 40, and 30% additive, dominant, and over-dominant gene effects, respectively gave the response expected from inbreeding. The selected model was used to study the effects of varying combinations of selection intensities and family sizes during inbreeding. The experiment was replicated five times. Average frequency of more favorable alleles increased as selection pressure was intensified over four generations of inbreeding. In general, percent heterozygosity increased as selection intensity was relaxed. Intense selection reduced heterozygosity and increased average percentage of loci fixed with more favorable allele. The genetic mean consistently increased with higher selection intensities. Retaining fewer but larger families resulted in more genetic gain. Differences in percent loci fixed with more favorable allele and differences among genetic means resulting from several differential selection intensities were statistically significant. Interactions among selection schemes and generations were not statistically significant, indicating that selection can be planned at any generation of line development regardless of selection schemes applied in earlier generations. Evidence in support of the need for early generation testing was indicated as 5% of the S$\sb0$ parents appeared in the pedigree of selected S$\sb4$ lines and only three of S$\sb0$ parents produced nearly half the selected S$\sb4$ lines. Resources could be saved by applying high selection intensities and early generation testing