Two hypotheses regarding mental rotation were contrasted. If subjects rotate each stimulus image to the upright (the image rotation hypothesis), then response time should depend solely on the extent of angular deviation from the upright. But if subjects rotate their frame of reference to match that of the disoriented stimulus (the frame rotation hypothesis), then response time should vary with the angular deviation between the current stimulus and the preceding stimulus. In four experiments, one involving normal and reflected letters (Experiment 1) and the other three involving lexical decisions on Hebrew letter strings (Experiments 2, 3A, and 3B), much stronger evidence for the image rotation hypothesis was found, though weak but systematic effects of frame rotation were also obtained. Increased likelihood that the same orientation would be repeated (Experiment 4) did not yield any stronger frame rotation effects. Also there was no indication of consistent individual differences in the preference for the frame rotation strategy (Experiment 3B). Additional findings pertinent to the application of the mental rotation paradigm to word recognition were discussed.