In this paper we describe an experiment to obtain information about the perceptual potential of the human visual system regarding shadow perception. Shadows play an important part for communicating spatial structures of objects to the observer. They are also essential for the overall realism of the rendered image. Unfortunately, most algorithms in computer graphics which are capable of producing realistic shadows are computationally expensive. The main idea behind the experiment is to use a simplified version of the shadow caster to generate hard and soft shadows, which would rapidly increase performance and to evaluate up to which degree a simplification is possible, without producing noticeable errors. Therefore, an experiment is performed, in which the test persons should mark the point of the just noticeable difference. First results show, that a mesh simplified to only 1% of its original complexity is capable to cast soft shadows that satisfy 90% of the test persons. CR Catego...