Physiological breeding to improve crop water-use efficiency is one of several complementary approaches to problems of agricultural water use. These problems include competition with other end-users, soil salinization, and declining or erratic water supplies for agriculture. In previous work, we developed a physiologically based model, which predicted that breeding for two specific traits might confer 10% increases in water-use efficiency (WUE) on alfalfa in large-field growth over the whole season, while only modestly depressing yield. While the predictions likely apply to other crops, we focused on alfalfa as a major crop economically and in water use in the Southwestern states. We report here more detailed tests of the hypothesis about WUE gains and their cost in reduced yield. These tests, in both controlled and field environments, also provided evidence that the two traits (Ci = CO2 concentration maintained in the leaf interior, and SLM = specific leaf mass or mass per area) are usable for breeding because they are (1) heritable, (2) stable in expression between repeated growth cycles and between modestly different growth environments, and (3) measurable for selection by accurate, cost-effective methods. We recommend that (1) primary attention should be paid to water management to conserve water in agriculture, in view of the rather modest and stringent limits on improving WUE by plant breeding, which is apparent in both comprehensive models and experiments here and elsewhere; (2) while these limits on WUE should be considered in allocating research efforts, a cultivar with low Ci and increased SLM should be developed for field trials; such breeding will be needed for future marginal gains in water conservation; other possibilities for breeding crops of superior WUE, especially selecting lower chlorophyll content of leaves will be more difficult but ultimately useful; (3) in breeding selection, Ci should be measured by mass-spectrometric determination of carbon-isotope discrimination at national facilities; (4) the design and interpretation of WUE in field trials should take careful account of effects of interplant competition and the micrometeorological conditions should be fully defined, particularly the canopy boundary-layer resistance; and (S) future breeding for WUE must account for some demonstrated trade-offs with crop drought tolerance and yield.