Two different turbulent flows, Langmuir supercells and unstable convection, have been sampled with a VADCP, an acoustic Doppler current profiler (ADCP) with an additional vertical (V) beam. Direct measurements of the profile of vertical velocity variance provided by the vertical beam are used to calculate observational response functions for algorithms used to derive vertical velocity from the 4 beams of a standard ADCP. A theoretical response function derived for the vertical velocity estimate from a single pair of opposed slant beams illustrates the importance of large-scale quasi-coherent flow structures, as well as effects of different angles of slant beams from vertical. Different large-eddy characteristics for Langmuir supercells and unstable convection yield different theoretical response: however in both cases, the theoretical response agrees qualitatively with that derived from observations. For Langmuir super-cells, there is additional agreement with numerical response functions generated by using the geometry of a VADCP to sample three-dimensional flow fields available from large eddy simulations (LES). The results from all three approaches show that there can be significant error in vertical velocity inferred from slant beam velocities. The error may be either over- or under-estimation, depending upon (usually unknown) features of the large eddies of the turbulent field, such as vertical/horizontal anisotropy, phase coherence, and orientation of horizontally anisotropic turbulent structures relative to the instrument. Given only a standard ADCP, the “best” estimate of vertical velocity variance is not the usual 4-beam estimate, but the larger of the two pair estimates.