The present study examined Reiss and McNally's expectancy model in the prediction of claustrophobic fear, measured across three domains. Non-clinical subjects (N = 117) reporting claustrophobic concerns were administered a behavioral approach test to a claustrophobic chamber. Consistent with the expectancy model, danger expectancy, anxiety expectancy and the interaction of anxiety sensitivity and anxiety expectancy accounted for unique portions of behavioral performance, with other variables partialled out. The expectancy model variable set, however, did not meaningfully relate to subjective fear or heart-rate reactivity. The formulation of anxiety sensitivity as a measure of the salience of anxiety is discussed. These findings lend support to the theory as a model for the behavioral dimension of pathological fear, but not the subjective or physiological facets.