Rotating-frame relaxation rates, R-1rho, are often measured in NMR studies of protein dynamics. We show here that large systematic errors can be introduced into measured values of heteronuclear R-1rho rates using schemes which are usually employed to suppress cross-correlation between dipole-dipole and CSA relaxation mechanisms. For example, in a scalar-coupled two-spin X-H spin system the use of H-1 WALTZ16 decoupling or H-1 pulses applied at regularly spaced intervals leads to a significant overestimation of heteronuclear R-1rho values. The problem is studied experimentally and theoretically for N-15-H-1 and C-13-H-1 spin pairs, and simple schemes are described which eliminate the artifacts. The approaches suggested are essential replacements of existing methodology if accurate dynamics parameters are to be extracted from spin-lock relaxation data sets.