Experiments in high Reynolds number turbulence call into question the experimental verification of Kolmogorov’s refined similarity hypothesis. Previous experimental support for the hypothesis is shown to be distorted by the severe simplifications used in estimating the turbulence energy dissipation, which has invariably been based only on the streamwise gradient of the streamwise velocity component. However, when the dissipation is estimated by the streamwise gradient of a transverse velocity component the dependence between ‖&Dgr;ur‖ and (r&egr;r*)1/3all but disappears. These results suggest that previous experimental support for the hypothesis may have been premature. ©1995 American Institute of Physics.