Auditory-visual integration efficiency (IE) is a presumed skill employed by subjects independently from their ability to extract information from auditory and visual speech inputs (Grant, 2002, JASA). However, currently there are no established methods for determining a subject’s IE. One approach is based on employing models of auditory-visual integration to predict optimal AV performance. Differences between model predictions and obtained scores are then used to estimate IE. However, the validity of these derived estimates of IE are necessarily based on the accuracy of the model fit. Here we present a novel measurement technique to address this issue without requiring explicit assumptions about the underlying audiovisual processing. It is based on a version of the Theory of Fechnerian Scaling developed by Dzhafarov and Colonius (Reconstructing distances among objects from their discriminability, Psychometrika, 2006, 71: 365-386) that permits the reconstruction of subjective distances among stimuli of arbitrary complexity from their pair wise discriminability. We demonstrate the approach on various data sets including a same-different experiment with phoneme-grapheme pairs from our lab.