The breeding of Upland cotton (Gossypium hirsutum L.) cultivars that combine high yield and fiber quality is a major challenge to the breeder. The understanding of the quantitative trait loci (QTL) contributing to agronomic and fiber quality traits offers an excellent route to solve this problem. A QTL analysis was carried out after an F2:3 population composed of 138 lines, derived from the intraspecific cross between Paymaster 54 and PeeDee 2165, was developed and a linkage map including 143 AFLP markers was constructed. The F2:3 population was grown in two locations, Alexandria and Baton Rouge in LA. The 143 linked markers were assigned to 13 major and 15 minor linkage groups, the 28 linkage groups cover a genetic distance of 1773.2 cM. This gives coverage of 37.7% of the cotton genome (4700 cM). Single-marker analysis, including simple and logistic regression, and interval marker analysis, including interval mapping (IM) and composite interval mapping (CIM), was used. Interval mapping was used to study QTL interaction effects with the environment. For the agronomic traits, the same five QTL were detected, using a significant threshold of 2 LOD, in both IM and CIM. These include two for lint weight per boll, two for seedcotton weight per plant, and one for lint percentage, which collectively, based on IM analysis, explained 32.5%, 28.6%, and 4.4% of the phenotypic variation, respectively. In total, seven and nine different QTL were detected by IM and CIM, respectively. For the fiber quality traits, the same nine QTL were detected in both IM and CIM. These include one for fiber elongation, one for length, two for uniformity, three for strength, and two for micronaire, which collectively, based on IM analysis, explained 50.9%, 18.7%, 69%, 49.6%, and 25.3% of the phenotypic variation, respectively. In total, nine and 19 different QTL were detected in IM and CIM, respectively. Eleven QTL were found to have significant interaction effects with the two locations. Future efforts in QTL mapping should focus on developing more saturated maps, using larger population sizes, and more powerful statistical algorithms and theories for identifying QTL and elucidating QTL X environment interactions.