Because of its relationship with water‐use efficiency (W), carbon isotope discrimination in leaves (Δ) was proposed to be useful for identifying genotypes with greater water‐use efficiency. In this study we examined the relationship between W and Δ in four peanut (Arachis hypogaea L.) genotypes. The genotypes were grown in and around mini‐lysimeters embedded in soil and were subjected to two drought regimes, intermittent and prolonged water deficit conditions, by varying the irrigation timing and amount. Automated rain‐out shelters prevented any rain from reaching the experimental plots during the treatment period. The mini‐lysimeters allowed accurate measurement of water use and total dry matter (including roots) in a canopy environment. Water‐use efficiency, which ranged from 1.81 to 3.15 g kg−1, was negatively correlated with Δ, which ranged from 19.1 to 21.8%. Tifton‐8 had the highest W (3.15 g kg−1) and Chico the lowest (1.81 g kg−1, representing a variation in W of 74% among genotypes. Variation in W arose mainly from genotypic differences in total dry matter production rather than from differences in water use. It is concluded that δ is a useful trait for selecting genotypes of peanut with improved W under drought conditions in the field. A strong negative relationship existed between W and specific leaf area (SLA, cm3 g−1) and between Δ and SLA, indicating that genotypes with thicker leaves had greater W. SLA could therefore be used as a rapid and inexpensive selection index for high W in peanut where mass spectrometry facilities are not available.