Under low signal-to-noise conditions, measurements of the acoustic backscattering cross-section σbs depend upon the position of the target in the beam. Split- and dual-beam measurements of σbs increase as the target moves away from the acoustic axis, leading to substantial overestimates of acoustic size for off-axis targets. We show how this phenomenon can be largely explained by measurement error in split-beam estimates of position, and by the effect of a minimum voltage threshold. We demonstrate a two-step correction to remove most of the bias. First, smoothed estimates of position are used to obtain an improved estimate of the beam directivity effect for each echo. Second, the statistical bias induced by the minimum voltage threshold is estimated and subtracted.