A microgenetic study of the Müller-Lyer illusion

In two experiments a decomposed Müller-Lyer pattern was used to measure the time course of the illusion. A partial report procedure was used to prevent the subjects from focusing only on parts of the pattern and to maximize visual processing. The Müller-Lyer figure was decomposed into two parts, its angles and its line. A configuration of three pairs of angles, each corresponding to a row in the usual partial report arrangement, was used. A line that did or did not fit the gap was shown with a variable delay (interstimulus interval, ISI). By this procedure the relevant row (line gap) was cued. The subject had to decide whether the line fitted or was too short/long. Two exposure times for the angles were used, 50 or 200 ms in one experiment and 50 or 500 ms in the other. The result of the first experiment, with outward-pointing fins, was a stable illusion for all values of ISI (25-400 ms) and all exposure times, with one significant exception: exposure for 50 ms with an ISI of 50 ms yielded an illusion peak. It was shown that this was not caused by a reduction in length-discrimination performance. In the second experiment, with inward-pointing fins, no such peak occurred. There was only a tendency for the illusion to vanish with zero ISI. The results are discussed with respect to 'global to local' theories of visual processing.