Almost all studies investigating the cognitive processes underlying peoples’ judgments and estimates rely on simple artificial stimuli with predefined cue structures. One reason for the use of these kinds of stimuli is that the cognitive models used in this area of research require that the cues and cue values are known. This limitation makes it difficult to apply the models to research questions with complex stimuli with an unknown cue structure. Drawing on early categorization research, in two studies we demonstrate how cues and cue values of complex stimuli can be extracted from pairwise similarity ratings with a multidimensional scaling analysis. These extracted cues can then be used in a state-of-the-art hierarchical Bayesian model of quantitative judgments. Using this procedure, we replicate previous findings from multiple-cue judgment literature but use complex stimuli with an a priori unknown cue structure to demonstrate that MDS-based attributes can be used as cues in a cognitive model of numerical judgments.