Research-based information on yield stability of cotton varieties is helpful to Tennessee producers in selecting cultivars to plant. Relatively stable yields allow producers to plan their crop budgets more accurately. This study assessed the relative yield stability of older and newer cotton varieties, and investigated possible relationships between varietal yield stability, yield potential, and earliness of maturity. Earliness was measured as the percent of total yield that was spindle-picked at first harvest. Data on lint yield and earliness of varieties were compiled from replicated variety trials conducted at three West Tennessee locations (Milan, Jackson, and Ames Plantation) from 1991 through 1999. All trials were planted in loess-derived silt loam soils in terrace or creek bottom settings. Supplemental irrigation was used at one location. Balanced data for varieties tested during successive 3-year intervals (1991-93, 1994-96, and 1997-99) were analyzed by a stability variance procedure, SAS-Stable (Magari and Kang, 1997). It is based on the principle that varieties that contribute relatively little to genotype-environment interaction (GxE) are relatively stable. Stability variance ( ² GE) estimates the magnitude of varietal contributions to GxE, and z-tests categorize varieties by significance of their contributions. Varieties differed in F ² GE for each 3-year period in this study. Of 24 varieties tested in 1991-93, two (Delcot 344 and MD 51ne) did not contribute significantly to GxE. They were categorized as relatively stable (Table 1). Four varieties made highly significant (P<0.001) contributions to GxE in 1991-93, and were categorized as relatively unstable. The remaining varieties were intermediate in yield stability. All 13 varieties evaluated in 1994-96 contributed to instability (Table 2). Of these, five were highly significant (P<0001), and were categorized as relatively unstable. Of 9 varieties common to the 1997-99 tests, four did not contribute significantly to GxE. They were categorized as relatively stable (Table 3). Yield stability was not significantly correlated with yield potential or earliness of maturity of older or newer cultivars, but there were weak correlations between F ² GE and earliness in 1991-93 and 1997-99 (Table 4). These trends were more attributable to instability of a few late maturing varieties than to higher levels of stability in early maturing cultivars. These results demonstrate that both older and newer varieties can be categorized in terms of relative yield stability. There is no indication from these data, however, that newer varieties are less stable than those grown in the early 1990s. Because of the difficulty of predicting yield stability from yield potential or earliness data, producers may find F ² GE useful to consider alongside yield and earliness when selecting varieties to plant.