The cultivated tetraploid Gossypium barbadense L. cotton produces superior natural fibers for the textile industry in the world. However, the possibility in utilization of heterosis to further increase its lint yield has not been extensively explored. In this study, two commercial US Pima cotton cultivars and three exotic G. barbadense lines, together with all of their possible hybrids in F1 and F2 progeny without reciprocals, were tested for lint yield, yield components, and fiber quality traits in four environments in 2005–2007. With a few exceptions, genotype (G), environment (E), and G × E were all significant or highly significant for all the traits studied. General combining ability (GCA) variances for all the traits in both F1 and F2 were also significant, while specific combining ability (SCA) variances were detected only for lint yield, fiber length, and micronaire in both generations and boll weight in F1. GCA × E was also detected for lint percent, seed index, and fiber length in both F1 and F2, and boll weight in F1, but none of the traits had significant SCA × E. As a group, F1 and F2 out-yielded the parent group by 20–40% and 6–10%, respectively. Mid-parent heterosis (MPH) for lint yield in F1 was generally positive, ranging from −4.7 to 116.4% with an average of 21.2–48.7%, while lint yield MPH in F2 ranged from −23.3 to 69.4% with an average of 6.4–12.4%. However, useful heterosis in lint yield was only detected in the hybrid between the two US commercial cultivars Pima S-7 and DP 340. MPH for other traits was low or not detected. MPH in F2 was lower than that in F1 but they were generally positively correlated. The genetic distances (GD) of the parents (based on 467 polymorphic RAPD and AFLP markers) between the five parents was not consistently correlated with MPH and SCA of their hybrids and dominant effects for lint yield and other traits. However, significant and positive correlations between GD of parents and the performance of their hybrids were detected for lint yield, lint percentage, and lint index in both F1 and F2 in most of the tests. GD of parents was also correlated with their GCA and additive effects in lint yield, lint percent, lint index, micronaire, plant height, and elongation. The results suggest that the close correlation between GD and hybrid performance per se was mainly due to the existence of GCA and additive effects from parents.