The state-trait anxiety inventory, trait version: does it really measure anxiety?

To clarify what is actually measured by the trait version of the State-Trait Anxiety Inventory (STAI; Spielberger, Gorsuch, & Lushene, 1970), we conducted a confirmatory factor analysis of various models and evaluated convergent and discriminant validity. The best fit was obtained with both a bifactor model, comprising 2 specific factors plus a general factor, and a 1-construct, 2-method model. The total score and the 2 method subscales of the STAI trait version were more strongly correlated with depression than with anxiety. In the bifactor model with 2 specific factors, the depression subscale showed stronger correlations with measures of depression than with measures of anxiety. The correlation of the hypothetical anxiety subscale with measures of depression was equivalent to or higher than its correlation with measures of anxiety. These results suggest that the questionnaire does not strictly evaluate anxiety but, rather, negative affect.