Transient state-dependent fluctuations in anxiety measured using STAI, POMS, PANAS or VAS: a comparative review

Abstract

Several psychometric instruments can be used to measure state-dependent variations in anxiety, including the State-Trait Anxiety Inventory (STAI), the Profile of Mood States (POMS), the Positive and Negative Affect Schedule (PANAS) and the Visual Analog Scales (VAS). Each of these instruments rests on specific theoretical assumptions about the construct of state anxiety, and has been widely used for this purpose in different research domains. However, it remains difficult to determine what may be the specificities of these four instruments, when the goal is to measure transient state-dependent variations in anxiety. In this work, we provide a systematic and comparative literature review of studies which have explored rapid fluctuations (i.e., test-retest intervals not exceeding 24 hours) in state anxiety by means of these specific instruments. Almost 200 studies were eventually included in our review. This comparative review confirms that, despite some disparities and specificities, each of these four instruments provides a reliable measure to capture rapid state-dependent variations in anxiety, although they have been used in non-overlapping research domains or experimental contexts.

Keywords:

• state anxiety
• self-report
• STAI
• POMS
• PANAS
• VAS

Acknowledgements

This work is supported by a grant from the European Research Council (Starting Grant #200758). We would like to thank Dr. Ernst Koster and two anonymous reviewers for their constructive comments on earlier drafts of this manuscript.

Notes

1. When compared to the four dominant self-report measures (STAI: 315 total hits; PANAS: 188; POMS: 196; VAS: 190), it becomes evident that the use of the SUDS (23 hits in Isi Web of Knowledge, 0 hits in PsychInfo and PsychArticles), EAS (0 hits in Isi Web of Knowledge, 5 hits in PsychInfo and PsychArticles) and the TESI (1 hit in Isi Web of Knowledge, 6 hits in PsychInfo and PsychArticles) remains limited.

2. Note that for the STAI-S the use of multiple research keys was required to overcome many inconsistencies in the way this specific part of the inventory has been commonly referred to in the literature, while for the three other questionnaires, a more widely accepted consensus regarding the use of the acronyms was found. In general, we adhered to the standard acronym used by the developers of the scales and reported in the published manuals.

3. “Short” has been formally operationalized as within the 24 hours. Although more longitudinal studies are certainly equally valid from a scientific viewpoint, we wished to focus in this review on “rapid” changes in anxiety and assess the ability of these four instruments to capture these fast and short-lived fluctuations.

4. In specific cases, when an experimental design combined anxiety down-regulation (e.g., via drugs) with experimentally increased anxiety (by means of stress-inducing tasks, for example), the preferred label was “Treatment”, because the experimental induction of tension or stress was meant to be only functional to the testing of the regulatory factor. On the other hand, when the positive outcome of an experimental manipulation was not expected or hypothesized a priori, or when both the directions of the effect were tested (for example, when exploring the effect of positive and negative interpretive bias training onto subsequent neutral tasks), the “Experimental” label has been chosen consistently.

5. Since scores on STAI Form Y (Spielberger, 1983; Vagg, Spielberger, & O'Hearn, 1980) and STAI Form X (Spielberger et al., 1970) are highly correlated (r correlation coefficients ranging from .96 to .97 in American students), the use of the unrevised form may, in some cases, be accepted. Nevertheless, the use of the Y version is highly recommended because of its more stable and replicable underlying factor structure, obtained after having identified and replaced some items providing only poor psychometric properties (Spielberger, 1983).

6. Simple modifications of the instructions enable flexibility, and yield similar assessments based either on previous (and still recent) moments in time (e.g., during the last block of the task, or during the therapeutic session), or for a future or hypothetical situation (just before or during a very difficult exam, for example).

7. Since the STAI-S was developed chronologically before other instruments measuring state anxiety, this may explain why no concurrent validity scores are provided in the Manual. In turn, the STAI-S has been used as reference for the validation of other instruments: for this reason, the correlation coefficients with the POMS, PANAS and VAS are reported in the sections concerning their respective psychometric properties (sections “General description, main use and psychometric properties”).

8. A less stable factor of Confusion-Bewilderment and an additional unscored Friendliness dimension were isolated in a subset of populations, but were not always replicated.

9. Between these two extremes, a wide variety of time frames can be tested, depending on the modifications made to the original task instructions (for details on this topic, see the technical Manual; Watson et al., 1988).

10. The internal validity coefficients provided for the PANAS-X were obtained with correlations between the NA and PA scales (PA scores non-reported here) and regression-based scores on the first two Varimax Factors in the samples assessed with the original 60 PANAS-X Mood Descriptors.

11. Note that in contrast with the other psychometric properties reported here, which were calculated with present instructions, the external validity coefficients of the PANAS-X have been calculated with past few weeks instructions. For further details on the PANAS-X psychometric properties with different timeframe instructions, see Watson and Clark, (1991, 1992, 1994).

12. Both the VAS-A and the STAI-S presumably provide an estimate of the same, very specific, construct (namely, anxiety), such that these two instruments should normally share a high portion of variance. Accordingly, correlation coefficients ranging between .34 (Hornblow & Kidson,1976 ) and .52 (Cella & Perry, 1986) may be considered moderately low in this context.