In this paper, I develop a composite methodology for handling and visualizing measures of significant attraction between lexical items in a set of near-synonyms. I aim to uncover revealing aspects of the conceptual structure of four English moderators (rather, quite, fairly, and pretty) and shed new light on previous studies on degree modifiers. This methodology combines univariate and multivariate statistics. Collexemes are used as input for hierarchical agglomerative cluster analysis, and multiple distinctive collexemes are used as input for correspondence analysis.