Tarr and Pinker (1989, 1990) delineated the conditions in which the orientation of a shape affects the time subjects take to recognize it. Relatively large effects of orientation were found (1) when misoriented asymmetrical shapes were unfamiliar, and (2) when familiar asymmetrical shapes appeared in unfamiliar orientations. In contrast, relatively small effects or no effect of orientation were found (1) when familiar asymmetrical shapes appeared at familiar orientations, (2) when unfamiliar mirror-reversals of familiar asymmetrical shapes appeared at unfamiliar orientations, (3) for misoriented symmetrical shapes, whether they were unfamiliar or familiar, and (4) for misoriented bilaterally redundant shapes, whether they were unfamiliar or familiar. To account for these data, two questions must be addressed: What mental processes underlie large effects of orientation or their absence; and why do these two patterns of data depend on manipulations of familiarity, handedness, and shape geometry? In Tarr and Pinker (1989, 1990) we proposed a theory that simultaneously answered both of these questions {what processes are used when very small effects of orientation are observed, and when are these processes used). The theory, multiple-views-plus-transformation, suggests that large effects of orientation in shape recognition are due to mental rotation. Crucially, this conclusion is based not only on the presence of an orientation effect, but on the similarity between our rotation rates and the rates observed in experiments by Shepard and Cooper (1982). These experiments used independent evidence to demonstrate the existence of an incremental mental rotation transformation, not confined to the effects of orientation, but depending on converging manipulations such as response time to probes at intermediate orientations and presentation points, effects of advance information and preparation time, and other techniques. We suggested that the absence of such large effects of orientation may result from orientation-invariant mechanisms of three kinds: orientation-specific representations of familiar shapes at familiar orientations; a 180° rotation in depth to align unfamiliar mirror-reversals with their familiar standards; and \Vi D orientation-independent descriptions for symmetrical and bilaterally redundant shapes. Together these hypotheses not only explain the causes of the absence of mental-rotation-size orientation