Gene flow patterns and the genetic structure of domesticated crops like cotton are not well understood. Furthermore, marker-assisted breeding of cotton has lagged far behind that of other major crops because the loci associated with cotton traits such as fiber yield and quality have scarcely been identified. In this study, we used 19 microsatellites to first determine the population genetic structure and patterns of gene flow of superior germplasm resources in upland cotton. We then used association analysis to identify which markers were associated with 15 agronomic traits (including ten yield and five fiber quality traits). The results showed that the upland cotton accessions have low levels of genetic diversity (polymorphism information content = 0.427), although extensive gene flow occurred among different ecological and geographic regions. Bayesian clustering analysis indicated that the cotton resources used in this study did not belong to obvious geographic populations, which may be the consequence of a single source of domestication followed by frequent genetic introgression mediated by human transference. A total of 82 maker–trait associations were examined in association analysis and the related ratios for phenotypic variations ranged from 3.04% to 47.14%. Interestingly, nine SSR markers were detected in more than one environmental condition. In addition, 14 SSR markers were co-associated with two or more different traits. It was noteworthy that NAU4860 and NAU5077 markers detected at least in two environments were simultaneously associated with three fiber quality traits (uniformity index, specific breaking strength and micronaire value). In conclusion, these findings provide new insights into the population structure and genetic exchange pattern of cultivated cotton accessions. The quantitative trait loci of domesticated cotton identified will also be very useful for improvement of yield and fiber quality of cotton in molecular breeding programs.