Anxiety reactivity and anxiety perseveration represent dissociable dimensions of anxiety vulnerability: A replication and extension

Abstract

Trait anxiety is a unitary construct reflecting individual differences in the tendency to experience anxious symptomatology, typically measured with questionnaires such as the Spielberger Trait Anxiety Inventory (STAI‐T). Recent research by Rudaizky, Page, and MacLeod has found evidence that two different dimensions of trait anxiety account for independent variance in trait anxiety scores. These dimensions are anxiety reactivity (AR), reflecting the probability of experiencing an anxious reaction, and anxiety perseveration (AP), reflecting the persistence of anxious symptoms once elicited. There are two key issues addressed in this study: first, the replicability of Rudaizky et al.'s findings and second, the ability of the measures of AR and AP developed by Rudaizky et al. to predict independent variance in STAI‐T scores after statistically controlling for variance shared with a measure of depression. Regression analysis determined that AR and AP do account for independent variance in STAI‐T trait anxiety scores even after statistically controlling for depression. The implications of these findings for the understanding of anxiety vulnerability are discussed.

Key words:

• anxiety
• anxiety perseveration
• anxiety reactivity
• anxiety vulnerability
• trait anxiety

This work was partly supported by Australian Research Council Grant DP0879589 and partly by a grant from the Romanian National Authority for Scientific Research, CNCS‐UEFISCDI, project number: PNII‐ID‐PCCE‐2011‐2‐0045.

Trait anxiety is a construct reflecting individual differences in the tendency for an individual to experience anxiety symptoms when exposed to a stressor. One of the ways in which trait anxiety is commonly assessed is through questionnaire instruments, for example, the Spielberger Trait Anxiety Inventory (STAI‐T; Spielberger, Gorsuch, Lushene, Vagg, & Jacobs, 1983), which requires respondents to specify how often they experience particular anxiety symptoms. High scores on the STAI‐T are thought to reflect a higher level of trait anxiety that has been shown to predict the development of a clinical anxiety disorder (Indovina, Robbins, Núñez‐Elizalde, Dunn, & Bishop, 2011) and poorer prognosis of recovery from such conditions (Chambers, Power, & Durham, 2004).

Recent research has provided preliminary evidence for the dissociation of two dimensions of anxiety vulnerability that have been confounded in traditional measures of trait anxiety such as the STAI‐T. These dimensions are termed anxiety reactivity (AR) and anxiety perseveration (AP). The former reflects individual differences in the probability of experiencing an anxious reaction after exposure to a stressor, and the latter reflects individual differences in the persistence of an anxiety response once such a response has been elicited. Rudaizky, Page, and MacLeod (2012) have shown that measures of AR and AP, provided by the newly developed Anxiety Reactivity and Perseveration Scale (ARPS), although correlated, each account for independent variance in STAI‐T scores.

While these findings support the hypothesis that AR and AP represent dissociable dimensions of trait anxiety, we believe that there are two key issues that need to be resolved in order to validate this distinction. The first issue concerns the replicability of the findings from the previous study by Rudaizky et al. (2012) showing that measures of AR and AP each account for independent variance in trait anxiety scores. Demonstrating the reliability of this finding through replication will bolster the conclusion that these are indeed separate facets of anxiety vulnerability.

The second issue concerns whether the ARPS measures of AR and AP are predicting variance in anxiety vulnerability specifically or negative emotionality in general. Rudaizky et al. (2012) found that measures of AR and AP contributed to independent variance in the STAI‐T. While the STAI‐T is the most widely used measure of trait anxiety (Grös, Antony, Simms, & McCabe, 2007), scores on this instrument have consistently been found to correlate highly with measures of depressive disposition (Bieling, Antony, & Swinson, 1998). Therefore, it remains uncertain whether or not the AR and AP scales of the ARPS predict independent variance in vulnerability to anxiety specifically. It is possible that the independent variance accounted for in STAI‐T scores by the measures of AR and AP may be related to variation in depression tapped by the STAI‐T questionnaire rather than the variation in anxiety vulnerability tapped by this instrument. To exclude this possibility, it would be necessary to show that AR and AP scores continue to predict independent variance in STAI‐T trait anxiety scores after accounting for the variance the STAI‐T scores share with a measure of depressive disposition such as the Beck Depression Inventory (BDI; Beck, Brown, & Steer, 1996a). The BDI is perhaps the most widely used measure to assess depression; elevated scores on this questionnaire are thought to reflect a range of dispositional qualities (Kendall, Hollon, Beck, Hammen, & Ingram, 1987).

In this study, we sought to address these issues by having participants complete the ARPS, the STAI‐T, and the BDI. Regression analysis was employed in order to assess (1) whether the present study replicates the findings reported by Rudaizky et al. (2012) by showing that these measures of AR and AP each account for significant independent variance in STAI‐T trait anxiety scores, and (2) whether these measures of AR and AP continue to predict independent variance in STAI‐T scores when controlling for the variance in STAI‐T scores that is shared with the BDI measure of depressive disposition.

Method

Participants

Participants were recruited in the same manner as in Rudaizky et al. (2012). A sample of undergraduate students from the University of Western Australia took part in the study. The sample consisted of 547 students (186 male) with a mean age of 18.56 years (standard deviation (SD) = 3.36, range 16–50). Trait anxiety scores ranged from 21 to 72 (mean (M) = 40.65, SD = 9.39), comparable with the distribution of scores reported in Rudaizky et al.

Measures

As in Rudaizky et al. (2012), participants completed the ARPS and the STAI‐T, but in the present study, the BDI was included as well; further details on these instruments are presented below.

ARPS

AR and AP were assessed with the ARPS as developed by Rudaizky et al. (2012). This scale consists of two questions asked about each of the 20 symptoms presented on the STAI‐T: an AR question and an AP question. The AR question asked, ‘When exposed to moderately stressful situations, what is the probability that you will experience this particular feeling?’. The response options were probability categories ranging from 1 = extremely improbable to 4 = extremely probable. The AP question was concerned with assessing the persistence of these thoughts and behaviours once they have been elicited. Participants were asked, ‘When you are exposed to situations sufficiently stressful to evoke this particular feeling, how long is this feeling likely to persist?’. The response options reflected symptom longevity ranging from 1 = extremely transient to 4 = extremely persistent. Response options were summed across each of the two questions resulting in two scores: an AR score and an AP score, each ranging between 20 and 80.

STAI‐T

Trait anxiety was assessed with the STAI‐T. The STAI‐T has been shown to have good reliability and validity (Beck, Stanley, & Zebb, 1996b); further details on the items and scoring of the STAI‐T are presented in Spielberger et al. (1983).

BDI

The primary measure of depressive symptomatology was the BDI. The BDI has been shown to have good reliability and validity (Beck, Steer, & Carbin, 1988; Wiebe & Penley, 2005); further details on the items and scoring of the BDI are presented in Beck et al. (1996a).

Procedure

The participants first completed the conventional set of STAI‐T questions followed by the ARPS and BDI, respectively. Each questionnaire was presented on a separate page and was completed in a group setting.

Results

Multiple regression was employed to assess whether AR (M = 43.66, SD = 12.05) and AP (M = 40.86, SD = 11.55) scores predicted significant independent variance in trait anxiety scores on the STAI‐T (M = 40.65, SD = 9.39). Once again, as in the original study, AR and AP scores were positively correlated, r = .73, n = 547; however, the variance inflation factor (VIF) was inspected and found to be less than 10 (VIF = 2.15), indicating that multicollinearity was not high enough to cause instability in the estimates. Both AR (r = .77, p < .001) and AP (r = .72, p < .001) scores were positively correlated with STAI‐T scores. When AR and AP scores were entered simultaneously as predictors, the regression model was found to be significant (F(2, 544) = 493.21, p < .001, R² = .64). Importantly, both AR (β = .52, t(544) = 14.04, p < .001) and AP (β = .33, t(544) = 8.88, p < .001) were found to account for significant independent variance in STAI‐T trait anxiety scores. These findings replicate those reported by Rudaizky et al. (2012) and support the reliability of the distinction between these two facets of anxiety vulnerability.

A hierarchical multiple regression was next conducted in order to determine the capacity of the measures of AR and AP to predict independent variance in STAI‐T scores when controlling for the variance in STAI‐T scores that is shared with the BDI. Significant positive correlations were found between BDI scores and scores on all three of the other measures (with STAI‐T scores r = .69, p < .001, with AR scores r = .58, p < .001, and with AP scores r = .58, p < .001). When BDI scores (M = 10.07, SD = 7.83) were entered into the first step, the regression model was found to be significant (F(1, 545) = 487.44, p < .001, β = .69, t(545) = 22.08, p < .001, R² = .47), reflecting the fact that BDI scores did, as expected, account for significant variance in STAI‐T scores. In the second step, both AR and AP scores were entered simultaneously. The regression model was found to be significant (F(3, 543) = 429.67, p < .001, R² = .70), and BDI scores were again found to predict variance in STAI‐T scores (β = .31, t(543) = 10.40, p < .001). Furthermore, and most importantly, both AR and AP scores were found to predict independent variance in STAI‐T trait anxiety scores in the second step despite the fact that any variance that STAI‐T shared with the depression measure was now statistically controlled for (AR β = .42, t(543) = 11.88, p < .001 and AP β = .23, t(543) = 6.41, p < .001). These findings support the hypothesis that the ARPS measures of AR and AP predicted independent variance that was specific to STAI‐T scores, rather than variance that STAI‐T scores shared with the measure of depression.

Discussion

The present findings demonstrate that the ARPS measures of AR and AP predict significant independent variance in trait anxiety scores. These results replicate those reported previously by Rudaizky et al. (2012), demonstrating the reliability of this finding and confirming that it is meaningful to distinguish between these two facets of anxiety vulnerability. A second important finding of the current study was that AR and AP continue to predict independent variance in STAI‐T trait anxiety scores after the variance that STAI‐T scores shared with the BDI measure of depression were statistically controlled for.

We believe that there are two key issues that could usefully be addressed by future research on this topic. These concern the underlying causes, and the experiential consequences, of the distinction between AR and AP. Future research could usefully investigate the differing mechanisms that give rise to these two facets of individual difference. The particular mechanisms of interest would depend on the theoretical framework adopted by the researchers. However, findings from diverse research domains such as cognitive and neuropsychology already suggest candidate mechanisms that may be causally related to the distinction between these dimensions of anxiety vulnerability. For example, emotion regulation researchers have previously made the distinction between primary and secondary emotional responses, with primary emotional responses being related to the onset and magnitude of an emotional reaction, while secondary responses are thought to be associated with the factors that moderate the subsequent emotional response and determine its offset (Koole, 2008). Researchers adopting a neuropsychological framework have looked at different brain regions that could be associated with these different aspects of the anxiety response. There is evidence implicating the amygdala in individual differences in the intensity of an anxious reaction, while the prefrontal cortex has been demonstrated to be associated with the relative persistence of such a reaction (Davidson, 2002).

Cognitive theorists have identified distinctions between alternative types of information processing bias that may differentially contribute to individual differences in AR and AP. For example, it is now widely accepted that anxious individuals show an attentional bias that selectively favours the processing of negative information (MacLeod & Mathews, 2012). Recently, researchers have differentiated two components of this attentional bias for negative information. Specifically, a distinction has been made between biased attentional engagement with negative information (Koster, Crombez, Verschuere, & De Houwer, 2006), which reflects an increased tendency for attention to be captured by initially distal negative information, and biased attentional disengagement from such information (Cisler & Olatunji, 2010), which reflects an increased tendency for attention to be held by initially proximal negative information. We suggest that future research could evaluate the quite plausible hypothesis that individual differences in biased attentional engagement and disengagement with negative information could respectively underpin individual differences in AR and AP. Further evaluation of this type of hypothesis may serve to potentially shed light on the mechanisms that underlie the distinction between these two facets of anxiety vulnerability.

Illuminating the potentially differing patterns of emotional dysfunction that may result from excessive AR and from excessive AP represents another important avenue for research to consider. While high levels of trait anxiety powerfully predicts a range of dysfunctional emotional outcomes (Mineka & Oehlberg, 2008), the ability to differentially predict alternative manifestations of anxiety dysfunction using a general trait anxiety measure is limited (Reiss, 1997). The deconstruction of this superordinate construct of trait anxiety into the component dimensions of AR and AP may further enhance our ability to predict the likelihood of anxiety pathology. It is possible that one of these constructs may predict such pathology more strongly than the other. Alternatively, it may be that the separate consideration of both dimensions of anxiety vulnerability may enhance our capacity to predict the differential likelihood of developing specific anxiety disorders. For example, the prospect of developing panic disorder, which is characterised by short bursts of intense anxiety with an abrupt onset, may be best predicted by elevated levels of AR rather than AP. In contrast, the likelihood of developing post‐traumatic stress disorder, which instead involves extended anxiety symptomatology following a stressful experience that has passed, may be better predicted by elevated AP than by elevated AR. Conditions such as generalised anxiety disorder, which involves anxiety symptoms that are both frequent and enduring, may best be predicted by heightened levels of both AR and AP, although the precise patterns of symptomatology may depend on which disposition is most strongly evident. Thus, not only could the deconstruction of trait anxiety into the dimensions of AR and AP potentially enhance the prediction of clinical anxiety dysfunctions, but it may also permit the development of testable hypotheses concerning the differing dispositional underpinnings of alternative types of anxiety disorder.

While these remain issues for future research, we can now confidently conclude that AR and AP differentially contribute to individual differences in trait anxiety. These two facets of anxiety vulnerability not only each account for independent variance in STAI‐T trait anxiety scores, but continue to do so even after statistically controlling for the variance that this trait anxiety measures share with depression.

Additional information

Funding

Australian Research Council Grant

Romanian National Authority for Scientific Research

CNCS‐UEFISCDI

Notes

This work was partly supported by Australian Research Council Grant DP0879589 and partly by a grant from the Romanian National Authority for Scientific Research, CNCS‐UEFISCDI, project number: PNII‐ID‐PCCE‐2011‐2‐0045.