This article describes a quantitative method to evaluate several averaging (i.e., confusion and assimilation) theories by comparing predictions of the absolute magnitude of the Müller-Lyer (ML) illusion with results of previous studies of the composite ML figure. The magnitude of illusion was best predicted by Davies and Spencer's (1977) theory and by integrative field theory (Pressey & Pressey, 1992). Furthermore, when the ML figure was at the point of subject equality, the average of shaft and intertip distances, and the configural dimensions proposed by Davies and Spencer, were most frequently closest to being equal in the apex-in ML and apex-out ML. Results indicate that a comparison of predicted and reported absolute magnitudes of the ML illusion can provide quantitative criteria to distinguish and evaluate averaging theories of the ML illusion.