This paper computes the effective duration of callable corporate bonds, using a contingent-claims model that incorporates both default risk and call risk. The model generates empirical implications regarding the cross-sectional variation and the firm-specific determinants of duration, and demonstrates that the effect of the call feature is to shorten duration (except for low-grade bonds). The effective duration is also estimated empirically for a large sample of long-term corporate bonds, using monthly bond price and interest rate data. Cross-sectional regression analysis is used to test the empirical implications of the model regarding the determinants of effective duration, and the empirical results are quite supportive of the model’s predictions.