Surface waves introduce velocity correlations that bias and often dominate Reynolds stress estimates made using the traditional variance method for acoustic Doppler current profilers (ADCPs). This analysis shows that the wave bias is the sum of a real wave stress and an error due to instrument tilt, both of which have a large uncertainty. Three alternative extensions to the variance method for calculating Reynolds stress profiles from ADCP measurements in wavy conditions are analyzed. The previously proposed variance fitting method (Variance Fit) is evaluated and two more general methods that use along- and between-beam velocity differencing with adaptive filtering (Vertical AF and Horizontal AF) are derived. The three methods are tested on datasets containing long-period monochromatic swell (Moorea, French Polynesia) and shorter-period mixed swell (Santa Barbara, California). The Variance Fit method leaves a residual wave bias in beam velocity variances, especially for intermediate waves, but gives physically reasonable Reynolds stress estimates because most of the residual wave bias cancels when the variance method is applied. The new Vertical AF method does not produce inherent wave bias in beam velocity variances, but yields comparable Reynolds stresses to the Variance Fit method. The Horizontal AF method performs poorly for all but monochromatic waves. Error remaining after one of the above methods is applied can be attributed to residual wave error, correlation of turbulence between points chosen for differencing, or correlation between waves and turbulence. A simple procedure is provided for determining the minimum bin separation that can be used.