The article by Tarr and Pinker explores how humans recognize objects despite differences in their retinal projections when seen at different orientations. They present four experiments to investigate the mechanisms of shape recognition, focusing on mental rotation and orientation dependence. The authors consider three main hypotheses: (1) objects are represented in memory as structural descriptions in object-centered coordinate systems, (2) objects are represented in a single canonical orientation with mental rotation operations, and (3) objects are stored in multiple representations corresponding to different orientations. In the experiments, subjects studied several objects at a single orientation and practiced naming or classifying them at various orientations. Initially, response times increased with departure from the study orientation, suggesting mental rotation. With practice, subjects recognized objects quickly at all familiar orientations, but response times increased again for novel orientations, indicating that subjects stored representations at each practice orientation and recognized shapes by rotating them to a stored orientation. The results support a hybrid model combining aspects of the second and third hypotheses. Input shapes are transformed to a stored view, either the nearest or canonical orientation. Interestingly, mirror-image shapes were recognized equally quickly at all orientations, suggesting that mental transformations align shapes along the depth axis in the picture plane. The authors also discuss previous studies on shape recognition, including evidence for mental rotation and the rotation-for-handledness hypothesis. They address three problems for the rotation-for-handledness hypothesis: the presence of local cues, persistent small orientation effects, and the interaction with familiarity. The experiments show that mental rotation is used for handedness judgments but not for recognizing shapes, and that orientation effects diminish with practice for familiar shapes. Overall, the findings suggest that with increasing experience, subjects form representations that enable handedness judgments only at specific orientations, indicating that these representations are more concrete and orientation-specific.The article by Tarr and Pinker explores how humans recognize objects despite differences in their retinal projections when seen at different orientations. They present four experiments to investigate the mechanisms of shape recognition, focusing on mental rotation and orientation dependence. The authors consider three main hypotheses: (1) objects are represented in memory as structural descriptions in object-centered coordinate systems, (2) objects are represented in a single canonical orientation with mental rotation operations, and (3) objects are stored in multiple representations corresponding to different orientations. In the experiments, subjects studied several objects at a single orientation and practiced naming or classifying them at various orientations. Initially, response times increased with departure from the study orientation, suggesting mental rotation. With practice, subjects recognized objects quickly at all familiar orientations, but response times increased again for novel orientations, indicating that subjects stored representations at each practice orientation and recognized shapes by rotating them to a stored orientation. The results support a hybrid model combining aspects of the second and third hypotheses. Input shapes are transformed to a stored view, either the nearest or canonical orientation. Interestingly, mirror-image shapes were recognized equally quickly at all orientations, suggesting that mental transformations align shapes along the depth axis in the picture plane. The authors also discuss previous studies on shape recognition, including evidence for mental rotation and the rotation-for-handledness hypothesis. They address three problems for the rotation-for-handledness hypothesis: the presence of local cues, persistent small orientation effects, and the interaction with familiarity. The experiments show that mental rotation is used for handedness judgments but not for recognizing shapes, and that orientation effects diminish with practice for familiar shapes. Overall, the findings suggest that with increasing experience, subjects form representations that enable handedness judgments only at specific orientations, indicating that these representations are more concrete and orientation-specific.