Some time between 30 and 40 thousand years ago modern humans began creating the first "signs" ever produced on Earth. The cave art of the Upper Palaeolithic is a remarkable record of the very earliest representational artistic expression. These were scratches, colored blotches and figurative art painted, drew on or engraved into bones or walls of the cave shelters. These paleolithic graphics (PG) consist largely of large animals (presumed prey of the creators) with some abstract geometric forms and very few human figurines. Developing an understanding of these early forms of art requires an understanding of the human perception of the images. The present paper is focused on the question what perceptual capacities do these graphics reflect?
I will argue that PG are representations based on basic perceptual elements, the same elements that can be found in the modern human and already in macaques and that are reflected also in the neuronal activity of the macaque brain. PG are not abstract symbolic images, but are concrete, perceptually motivated representations that emphasize the distinctive features of the depicted objects. What are the perceptual elements that were most widely used in PG?
We live in a 3D world and the depiction of a 3D object on a 2D surface is not an easy task. The first computational step of many recent models of object recognition is the extraction of contours. Thus, it is not surprising that most of the PG consists of the occluding contours (i.e. the outer boundaries or outlines) of the figures and many PG are simply line drawings or silhouettes of the figures. In addition to the use of occluding contours sometimes texture elements and color were used to enhance contrast between the background pattern of stones and the depicted shape. In a few rare cases parts of the shapes are depicted as illusory contours, suggesting that creators of PG (just like modern human, macaques and certain neurons of the macaque brain) already considered them equivalent to real contours.
However , (1) these depictions are not equally informative from every view and (2) both view-based and component-based theories of object recognition suggest that some views of a given object are easier to recognize than others. Indeed, animals on PG are usually depicted from profile; a view, very informative for silhouettes and also close to the typical canonical viewpoint.
Although depth representation is not obviously present in the PG (application of perspective e.g. was not found) there are signs of the use of other, monocular depth cues, such as relative size, partial overlap or occlusion, that are informative cues of depth for both monkeys and human observers.
To help recognition and perceptual categorization on the basic and subordinate levels certain distinctive (non accidental) features are helpful and humans, monkeys (as well as single neurons of the macaque brain) are very sensitive to changes of these features of the images. Indeed, it is common in PG that certain features (horns, antlers) of the animals that are distinctive to the given species are exaggerated (are drawn larger or carved deeper into the stone) or even depicted in isolation.
In summary, there is a style to PG that is consistent with the view that they represent the construction of perceptual images that stand individually and collectively as signs of the phenomena of nature of which they are elements.