Abstract: In order for development of new parents for development of new hybrids，the parental materials of popularized hybrids including Xiangzamian 2, Wanza 40 and Zhongmiansuo28 were used to construct recurrent populations for Pooling High Yield Genes in Upland cotton, and two cycles of selection were completed. The objective of this experiment was to investigate the results of phenotypic recurrent selection and MAS. The phenotypic recurrent selection was effective at both individual and cross level, and in different environments for the same trait. The tendency of mean variation in initial population and all populations of two cycles showed no difference, and the means of the most traits including lint yield, seed cotton yield and boll numbers per plant of populations in the second cycle were significantly higher than that of populations in the initial cycle while the genetic variances had no significant difference among them. The effects of MAS using S1495210, S167280 and S3994100, which identified by interval mapping in our institute were closely related to the QTLs for lint yield, lint percent and seed size respectively, were significant. MAS were also efficient using the marker S3452180, identified by individual maker mapping, related to lint yield and lint percent. The results of MAS for lint percent with two markers were more efficient than those with one of two markers.

Key words: Upland cotton, Recurrent selection, GCV, Genetic variance