Carbon isotope discrimination (CID) usually correlates with the dry matter to transpiration ratio (water use efficiency) in C3 species, but its heritability and relationship to forage production is less clear. In a four year study of tall fescue (Festuca arundinacea Schreb.) at Pullman, WA and Ardmore, OK we determined, 1) if CID differences in divergently selected populations made on single plants were maintained in solid seeded plots, and, 2) how CID in selected populations and a set of four cultivars were related to forage production, leaf N, C, and the C:N ratio. Consistent with a heritable trait, differences in CID for low and high CID populations selected on spaced plants were maintained in solid seeded plots at both Pullman and Ardmore. At Pullman, the low Δ selection had consistently higher production than the high CID selection with the base population intermediate, a pattern consistent with low CID promoting production. That pattern was observed at Ardmore in one of six cuttings. Leaf N was positively correlated, and the C:N ratio negatively correlated with Δ at both locations, but N, C, and C:N were not consistently related to forage production. Across cuttings CID was negatively correlated with forage production at Ardmore (r = -0.26**) and positively correlated with production at Pullman (r= 0.71**). However, partial correlations of CID with forage production, controlling for cutting effects, were not significant. Partial correlation between CID and forage production on the selected and base populations was significant (r= -0.36*, n=32) at Pullman, but not at Ardmore. The data show selection for low CID may improve forage production, but not consistently. For breeding, one cycle of mass selection in advanced material is recommended.