Abstract
<p>The State–Trait Anxiety Inventory (STAI) is widely used in health and well-being research to distinguish enduring anxiety proneness from transient anxiety responses, yet the temporal composition of its trait and state subscales remains unclear. Data were drawn from 1,177 participants in the 16-wave COVID-Dynamic dataset. Item-level confirmatory factor analyses, longitudinal measurement invariance testing, and trait–state–occasion (TSO) models were used to evaluate the measurement structure and temporal variance composition of the STAI-T and STAI-S. Different wave combinations spanning intervals of 21–294 days were examined to assess variation in trait and occasion-specific variance across temporal scales. For both scales, a one-construct, two-method-factor model provided the best fit, and subsequent analyses supported essential unidimensionality and longitudinal scalar invariance. TSO analyses showed that STAI-T variance was almost entirely trait-like. Although the STAI-S contained more occasion-specific variance, most of its variance was also attributable to stable individual differences. The variance composition of the STAI-S was relatively consistent across shorter time spans, whereas the occasion-specific variance proportion was higher for the longest span examined. No statistically detectable differences emerged between wave combinations covering the same time span. Thus, the two scales appear to differ quantitatively in their temporal variance composition rather than representing categorically distinct temporal constructs. In health, clinical, and intervention research, STAI-S scores should therefore not be interpreted as pure indicators of state anxiety, and measurement interval should be treated as a substantive design consideration.</p>