Using three-dimensional Monte Carlo radiative transfer simulations, we examine the effect of beam transmissometer geometry on the relative error in the measurement of the beam-attenuation coefficient in an aquatic environment characterized by intense light scattering, especially within submerged bubble clouds entrained by surface-wave breaking. We discuss the forward-scattering error associated with the detection of photons scattered at small angles (< 1 degrees) and the multiple-scattering error associated with the detection of photons scattered more than once along the path length of the instrument. Several scattering phase functions describing bubble clouds at different bubble void fractions in the water are considered. Owing to forward-scattering error, a beam-attenuation meter (beam transmissometer) with a half-angle of receiver acceptance of 1.0 degrees and a path length of 0.1 m can underestimate the true beam attenuation within the bubble cloud by more than 50%. For bubble clouds with a beam attenuation of as much as 100 m(-1), the multiple-scattering error is no more than a few percent. These results are compared with simulations for some example phase functions that are representative of other scattering regimes found in natural waters. The forward-scattering error for the Petzold phase function of turbid waters is 16% for a typical instrument geometry, whereas for the Henyey-Greenstein phase function with the asymmetry parameter of 0.7 and 0.9 the error range is 8-28%.