In the present study, errors in using Taylor’s hypothesis to transform measurements obtained in a temporal (or phase) frame onto a spatial one were evaluated. For the first time, phase‐averaged (‘‘real’’) spanwiseandstreamwise vorticity data measured on a three‐dimensional grid were compared directly to those obtained using Taylor’s hypothesis. The results show that even the qualitative features of the spanwise and streamwise vorticity distributions given by the two techniques can be very different. This is particularly true in the region of the spanwise roller pairing. The phase‐averaged spanwise and streamwise peak vorticity levels given by Taylor’s hypothesis are typically lower (by up to 40%) compared to the real measurements. ©1995 American Institute of Physics.