The effect of depth rotation on object identification

Five experiments are reported in which the time to verify the name of different three-dimensional common objects shown rotated in depth was investigated. Views of computer-generated images of elongated objects rotated in steps of 30 degrees along six axes of rotation were used as stimuli. A significant main effect of view was found in all experiments. This effect was initially attributed to the relatively slower verification times to the end-on views of objects but further analysis revealed that views 30 degrees off the end-on views were significantly slower to verify than other views. Objects with gravitational uprights yielded the same effects as objects without gravitational uprights. The results were not dependent on practice with the stimuli prior to the experiment or on repeated exposure of the views during the experiment. Also, there was no benefit found for the identification of shaded over silhouetted images of objects when shown in more-conventional views but unconventional views were more recognisable when shaded than when silhouetted. Last, initial verification times for familiar views of a set of novel objects were faster than for unfamiliar views even when the views were unconventional. With practice on unfamiliar views, however, the same function relating view to verification time found for familiar objects was found for the novel objects. The results suggest that for recognition purposes visual memory stores discrete views of objects but it characteristically favours a canonical range of views of elongated objects that are based on the salient geometry of the objects so that more unconventional or foreshortened views are less readily recognised.