Abstract Importance sampling is proposed as a technique for estimating log volume that eliminates the bias of conventional methods arising from irregular log taper. In a trial of the technique, the average relative sampling error for individual logs, based on a measurement of one cross-section per log, was 5.2%. To estimate an aggregate volume, a two-stage sampling procedure is suggested in which importance sampling constitutes the second stage. Optimal sampling strategy is discussed briefly. Forest Sci. 32:1073-1078.