In his Vision Science : pp.421-24, Stephen Palmer discusses his work and that of Eleanor Rosch's group on preferred camera positions when identifying pictures of common objects. Overall, people most readily identify pictures of objects when the camera is elevated 30-45 degrees above the object and rotated 30-45 degrees to the left or right of the object, in contrast to dead-level straight on front or side views. This view down a diagonal gives maximum information about the depth dimension.
The limitations of experimental study of perception are quite in evidence here. You cannot, of course, photograph objects just by themselves--there is always a context or a viewpoint. Palmer reproduces an array of small black and white photos of such objects as irons, houses, cameras, cars, shoes, pianos, and the like. All of these objects are photographed on a flat surface with a wall behind it with diffuse light and rather uninformative shadows. To make scale relatively uniform, toy models of the larger objects were used, so that the experiment provides data about the recognition of toy cars, houses, etc., rather than the real thing. The result of these various adjustments is that the viewpoint is pretty constantly that of an adult standing before a table with things or small models of things on it. That would seem to have to do with why subjects preferred these images as recognizable presentations of the objects. Note that if the car and house were not toy models, one would have to imagine oneself in an entirely different viewing position--well above ground level, for one thing. So it is not exactly clear how much we know about canonical recognition perspectives, or how we might find out more.
Nicholas Wade examines a bent stem pipe like Magritte's to determine why the side-on view is the canonical ("stereotypical") one and concludes that, for pipes at least, the side view is the one that does not subject the asymmetrical axis of the pipe to perspectival foreshortening. He does not attempt to generalize beyond the pipe. (Nicholas Wade, Visual Allusions: Pictures of Perception, Lawrence Erlbaum Associates, 1990, pp. 6-11.)back to Photomontage