Abstract
Singh and Anderson (2002) proposed that the perceived transmittance of a transparent filter seen in front of a sinewave background is determined by the ratio of its Michelson contrast to that of its background seen directly. In their experiment, observers matched the perceived transmittances of target and comparison filters by adjusting the comparison filter's luminance range, keeping its mean luminance constant. The two filters had different mean luminances, but identical backgrounds. Assuming that the visual system parameterized the filter and background image regions (sinewaves) in terms of mean luminance and contrast, then this setup meant that observers could only match on the basis of contrast. The possibility that mean luminance also influences perceived transmittance was not considered. Thus, observers might have simply been matching the best they could under the restriction that they could only adjust contrast.
To address this issue, observers in my experiment matched perceived transmittance by adjusting the comparison filter's (simulated) physical transmittance. The target and comparison backgrounds had different mean luminances and Michelson contrasts, thereby unconfounding filter contrasts from filter-to-background contrast ratios. Also, adjusting the physical transmittance of the comparison filter changed both its mean luminance and contrast, so that their relative importance for matching perceived transmittance could be investigated.
Both mean luminance and contrast strongly affected perceived transmittance, contrary to Singh and Anderson's theory. For example, their theory incorrectly predicts that all filters with zero reflectance will appear equally transmissive. Also, Singh's (2004) claim that their theory predicts lightness matching through transparency is incorrect.