Abstract: It is well known that survival data arising out of prevalent cohort and current duration sampling schemes are length-biased and are respectively the forward and the backward recurrence time versions of the core event time variable. In presence of covariates, the Proportional Hazards model may not be valid for such recurrence time data since the proportionality structure is not preserved when going from the core to the recurrence time versions. However, the AFT model can be an useful alternative to model failure time in terms of covariates due to its invariance under length-biased and cross-sectional sampling. Here we present a semiparametric approach to estimating the regression parameter in the AFT model for Forward and Backward recurrence time data.