Three related issues concerning the theory, measurement and clinical utility of the fear of fear construct as operationalized by the Agoraphobic Cognitions and Bodily Sensations Questionnaires (Chambless, Caputo, Bright & Gallagher, Journal of Consulting and Clinical Psychology, 52, 1090-1097, 1984) were addressed: stability of ranked scores, stability of means (retest effect) and predictive ability. In addition to measures assessing moods and enduring personality traits, the relevant fear of fear scales (and introducing a companion frequency measure to the BSQ-intensity scale) were administered to somewhat over 60 clients with "panic disorder with agoraphobia" or "agoraphobia without a history of panic attacks" on two occasions, 3 months apart, with no intervention having taken place between first and second testing. As predicted, ranked fear of fear scores were highly stable. In addition, Howarth's mu index values pointed to the description of the fear of fear constructs in terms of traits. Specific fear of fear scales showed evidence of a retest effect. The short-term course of agoraphobic avoidance behaviour in untreated cases was predictable from specific fear of fear variables, even after controlling for their shared variance with pre-test measures of either state anxiety-panic or trait neuroticism. Implications of the findings are discussed.