Abstract
Maximum likelihood difference scaling (MLDS; Maloney & Yang, 2003) is a straightforward and intuitive method to measure perceptual scales. We adopted MLDS to measure lightness scales in different contexts, such as plain view and various transparency conditions, and we showed that it was possible to predict lightness matches across different contexts. One great benefit of MLDS is that it only ever asks observers to compare entities that are seen under identical conditions. However, to compare perceptual scales across conditions they need to be anchored to a common origin, which by default is chosen to be zero and which might not always be the best choice. Maximum likelihood conjoint measurement (MLCM; Knoblauch & Maloney, 2012) asks observers to do within and across context comparisons of targets and estimates various scales whereby one of the scales serves as a reference and all other scales are expressed relative to the first scale. Here, we compared scales estimated with MLDS and MLCM. We measured scales of perceived lightness in rendered images of variegated checkerboards seen in plain view or through a transparent medium. The transparency compresses and shifts the luminance range with respect to plain view. For MLCM we used the method of paired comparisons, and for MLDS we used the method of triads where observers indicate which of two pairs of checks appears more different in lightness. We simulated responses from theoretical lightness scales assuming varying degrees of lightness constancy. MLCM and MLDS were both able to recover the theoretical lightness scales. MLCM was able to detect cases of partial lightness constancy for which MLD scales might be confounded. Experimentally we found that scales derived with MLCM were consistent with those from MLDS. MLCM might be a valuable alternative to MLDS when the perceptual magnitudes under study exhibit less constancy than perceived surface lightness.
Meeting abstract presented at VSS 2018