Indirect selection for cotton yield using yield component variables may provide improved yield, yield stability, and fiber quality. Data from irrigated and non-irrigated tests conducted in 1999 through 2006 at four Arkansas locations were evaluated. Parameters included lint yield (LY), basic yield components of number of seed per area (SPA) and lint index (LI), secondary yield components of seed index (SI), number of fibers per seed (FPS), and fiber density (FDEN), lint percentage (LP), and seed percentage (SP), and fiber quality traits of micronaire (MIC), length (LEN), uniformity index (UI) and strength (STR). Seed index and LP were dropped from further consideration due to collinearity. Locations and genotypes varied in most of the tests for most of the variables. Genotype × location (G×L) was most frequently found for LY and SPA and least frequently found for LEN and UI. Path coefficients from PathSAS were used to identify the direct effects and relationships among variables. Higher instability was found in non-irrigated tests and in tests of more recently developed genotypes. Lint yield was influenced by SPA, but SPA exhibits low heritability. Lint index was influenced by FPS, MIC and LEN with FPS having the strongest influence. Among the fiber quality parameters, MIC and LEN had a negative relationship while UI and STR had strong direct effects on LEN. Fiber density via FPS tended to be related to lower MIC. These data suggest that FDEN could serve as a selection criterion for increased LY without negatively impacting fiber quality traits.