Advances in the ⁴⁰Ar/³⁹Ar method using new techniques have led to considerable efforts to improve the accuracy of the calibration of interlaboratory standards. However, the accuracy of the method ultimately relies on the measurement of ⁴⁰Ar*/³⁹Ar_K ratios on standards that have been calibrated with the K–Ar method. Usually a ⁴⁰Ar/³⁹Ar total gas age is assumed to equate to a K–Ar age, but this assumes that there is zero loss of ³⁹Ar due to recoil. Traditional ⁴⁰Ar/³⁹Ar total gas ages are Ar retention ages and are not strictly comparable to K–Ar ages. Efforts to estimate the importance of this effect on standards have relied on indirect evidence for ³⁹Ar recoil. We report direct measurements of ³⁹Ar recoil for primary and secondary standards using the vacuum-encapsulation technique and show that adjustments to some standard ages may be needed. Revised ages corrected for recoil are given for hornblendes MMhb-1 and Hb3gr, biotites GA1550 and FCT-3 and sanidines FCT-2 and TCR-2. The results show that, in most cases, recoil loss exceeds that which would be expected from grain size and geometry. Internal defects within mineral grains are likely the dominant control on the fraction of recoiled ³⁹Ar lost from these standards.