The gambling attitudes and beliefs scale (GABS: Breen & Zuckerman, 1999) was designed to assess a latent affinity for gambling. Using methods based in item response theory we demonstrated how a reduced set of GABS items maintained their relative severity and discriminated similarly when used among nonproblem gambling students selected to represent low levels of gambling behavior (n = 487) and when used clinically among treatment seeking pathological gamblers (n = 234). This stability increases confidence both in the construct measured by the GABS and in the ability to assess levels of gambling affinity across disparate ranges of gambling-problem severity. The GABS also demonstrated incremental validity in predicting increases in the frequency of gambling behavior among non-problem gambling students beyond that explained by an index of gambling-problem severity. Implications for assessment of gambling affinity across pathological and nonpathological gamblers are discussed.