The perception of natural contour.

The observation that natural curves and surfaces are often fractal suggests that people may be sensitive to their statistical properties. The perceptual protocols that underlie discrimination between fractals and between other types of random contour and fractals are examined. Discrimination algorithms that have precisely the same sensitivities as human observers are constructed. These algorithms do not recognize the integrated scale hierarchy intrinsic to fractal form and operate by imposing a metatheory of structure that is based on a signal-noise distinction. The success of the algorithms implies that (a) self-affinity in random fractals is not perceptually recovered and (b) people have a natural disposition to view contour in terms of signal and noise. The authors propose that this disposition be understood as a principle of perceptual organization.