Academic journal article Journal of Cognitive Psychotherapy

A Concurrent Test of the Anxiety Sensitivity Taxon: Its Relation to Bodily Vigilance and Perceptions of Control over Anxiety-Related Events in a Sample of Young Adults

Academic journal article Journal of Cognitive Psychotherapy

A Concurrent Test of the Anxiety Sensitivity Taxon: Its Relation to Bodily Vigilance and Perceptions of Control over Anxiety-Related Events in a Sample of Young Adults

Article excerpt

The present investigation evaluated the Anxiety Sensitivity (AS) taxon using the 16-item Anxiety Sensitivity Index (Reiss, Peterson, Gursky, & McNally, 1986) and its relation with two theoretically relevant cognitive processes associated with panic vulnerability: bodily vigilance and perceived uncontrollability over anxiety-related events. Taxometric analyses of 589 young adults indicated that the latent structure of AS was taxonic with an estimated base rate ranging between 13% and 14%. As predicted, an 8-item ASI Taxon Scale accounted for significant variance above and beyond that accounted for by the full-scale ASI total score in terms of bodily vigilance and perceived controllability of anxiety-related events. Moreover, after accounting for the variance explained by the full-scale ASI total score, the total score for the 8 ASI items not included in the ASI Taxon Scale was associated with significant variance in these same dependent measures, but it was in the opposite direction from that predicted by contemporary panic disorder theory. Dichotomous taxon membership accounted for significant variance above total ASI scores for bodily vigilance but not perceptions of control for anxiety-related events. These findings are discussed in terms of their theoretical implications for the study of AS and panic vulnerability.

Keywords: anxiety sensitivity; taxometrics; anxiety; bodily sensations; perceived control; panic attacks

Anxiety Sensitivity (AS), defined as the fear of anxiety and anxiety-related sensations (Reiss & McNally, 1985), is a traitlike cognitive predisposition that can theoretically increase the risk of panic and other types of anxiety problems. Since the 1980s, there has been a concerted scientific effort toward better understanding the latent structure of the AS construct (see Taylor, 1999, for a review). Whereas earlier theoretical models of the construct posited a unidimensional structure, subsequent factor analytic investigations have suggested that AS is hierarchical (Zinbarg, Mohlman, & Hong, 1999). These findings have helped researchers enhance the level of explanatory precision of AS in models of anxiety vulnerability (Zinbarg, Brown, Barlow, & Rapee, 2001; Zvolensky, Kotov, Antipova, & Schmidt, 2005).

More recently, researchers have begun to explore the latent structure of AS using taxometric procedures. Taxometrics is a branch of applied mathematics that focuses on the classification of entities (Meehl, 1995). It provides a set of statistical procedures that are used to test the latent structure of constructs (Beauchaine, 2003; Cole, 2004; Meehl, 1995, Schmidt, Kotov, & Joiner, 2004; Waller & Meehl, 1998). Although taxometric investigations have been completed for various types of psychopathology (e.g., Beach & Amir, 2003; Gleaves, Lowe, Snow, Green, & Murphy-Eberenz, 2000; Harris, Rice, & Quinsey, 1994; Haslam & Beck, 1994; Lenzenweger, 1999), the application of these analytic methods to anxiety-related constructs, including AS, is a relatively recent development (Bernstein, Zvolensky, Kotov, et al., 2006; Bernstein, Zvolensky, Weems, Stickle, & Leen-Feldner, 2005; Ruscio, Borkovec, & Ruscio, 2001; Ruscio, Ruscio, & Keane, 2002; Taylor, Rabian, & Fedoroff, 1999). This work is important on theoretical grounds in the sense that it can help explicate the extent to which AS (and other anxiety constructs) is best understood from a continuous or categorical perspective. Such knowledge, in turn, ought to help direct research and clinical activities regarding anxiety-relevant vulnerability (broadly), and panic-relevant vulnerability (specifically) in a more effective and efficient manner (Beauchaine, 2003).

Unlike earlier theoretical models positing a dimensional AS latent structure (i.e., an individual difference factor that varies only by degree; Reiss & McNally, 1985), more recent empirical investigations suggest that AS may be taxonic (i.e., discrete, naturally occurring latent classes; Meehl, 2004). For instance, although Taylor and colleagues (1999) initially reported ambiguous findings with respect to the latent structure of AS in a sample of adults using the 16-item Anxiety Sensitivity Index (ASI; Reiss et al., 1986), a subsequent taxometric study found AS to be taxonic (Schmidt, Kotov, Lerew, Joiner, & Ialongo, in press). Confidence in the taxonic structure of AS has been further strengthened by the results of other investigations. For instance, employing the 36-item Anxiety Sensitivity Index-Revised (ASI-R; Taylor & Cox, 1998), Bernstein, Zvolensky, Kotov, et al. (2006) found that AS was taxonic among adults from North America (Canada and the United States), France, Mexico, Spain, and the Netherlands (total N = 2,741). In a separate study, Bernstein, Zvolensky, Weems, et al. (2005), using the 18-item Childhood Anxiety Sensitivity Index (CASI; Silverman, Fleisig Rabian, & Peterson, 1991), found that the latent structure of AS among youth was taxonic. Collectively, the majority of taxometric studies thus far completed suggest that AS is taxonic, rather than dimensional, in nature. That is, these data indicate there may be two discrete forms of AS, which differ between individuals qualitatively or in kind (see Meehl, 2004, for a discussion of taxa). These findings depart from earlier views suggesting that AS is an individual difference variable that exists in all persons in the same form and varies between individuals only by degree along a single latent dimension (Taylor, 1999).

Although this is promising, an important next step in the taxometric study of AS is to evaluate the nomological nature of the taxon with respect to external criteria of interest. To date, only Schmidt et al. (in press) have investigated the extent to which taxon class membership is associated with increased risk for anxiety-related problems. Specifically, these investigators found that AS taxon class membership predicted the incidence of spontaneous panic attacks and the frequency of occurrence of such attacks; notably, the predictive utility of the AS taxon was above and beyond the dimensional measures of the AS construct. Such findings provide initial empirical support for the incremental (or relative) validity of the AS taxon relative to the dimensional index of the construct in terms of the onset and frequency of panic attacks. Thus, these findings provide preliminary evidence not only that the AS taxon is categorically distinct from the AS non-taxon (i.e., latent qualitative discontinuity) but that taxon class membership confers risk for panic psychopathology. The findings of the study by Schmidt and colleagues (in press), though promising, are limited in at least three key respects. First, the investigation focused largely on panic attacks as a prototypical index of panic vulnerability. Biopsychosocial models of panic disorder suggest that cognitive processes related to perceptions of limited control and bodily vigilance also are central to conferring an increased risk for panic attacks and panic disorder (Barlow, 2002; Craske, 2003; Mineka & Zinbarg, 1996). It is important to extend previous work by evaluating the explanatory utility of the AS taxon in relation to the panic-relevant cognitive processes of controllability and bodily vigilance. Second, Schmidt and colleagues (in press) evaluated the explanatory utility of the AS taxon in one analytic way; specifically, by using dichotomous (categorical) taxon class membership relative to a continuous index of the construct. This is one of a number of potential ways to evaluate the AS taxon. However it is important not to disregard the possible meaningful latent continuity within the qualitatively distinct latent classes. Accordingly, enhanced confidence in the AS taxon with respect to panic vulnerability could be attained if alternative analytic models of the taxon (e.g., derived scale of ASI items that best discriminate between taxometrically based dichotomous class membership) were employed and compared directly (e.g., Waller & Ross, 1997). Finally, the sample in the study by Schmidt and colleagues (in press) was largely composed of male military cadets. These persons may not be representative of other segments of the population, as they represent males who have sought out military experience and training. Thus, tests of the AS taxon among a nonmilitary sample of young adults will be important to enhance the generalizability of the study's findings.

The overarching purpose of the present investigation was to evaluate the AS taxon using the 16-item ASI, and, specifically, its relation to theoretically relevant cognitive processes associated with panic vulnerability. Young adults, as opposed to individuals in other age groups, were the target population, given that panic-related problems typically first emerge during this developmental period ( DSM-IV-TR ). Consistent with the promising, albeit preliminary, taxometric study of AS, we first tested the latent structure of AS prior to testing the external validity of the putative AS taxon in regard to two theoretically relevant cognitive variables for panic-related problems. Consistent with previous research, we anticipated that AS would be taxonic in the present young adult sample.

Existing work relevant to the nomological nature of the putative taxon (Schmidt et al., in press) suggests that the AS taxon (indexed in two different analytic ways; see section on method for details) ought to account for the association between dimensional indices of AS and panic vulnerability. Thus, we expected that the AS taxon would be significantly associated with bodily vigilance even after accounting for continuous variability in the cognitive construct (i.e., ASI total scores). This prediction was based upon conceptual models of panic vulnerability (Barlow, 2002) and research (Schmidt, Lerew, & Trakowski, 1997) suggesting that individuals at risk for panic-related problems tend to focus on, and be preoccupied with, bodily sensations. It also was hypothesized that the AS taxon would account for significant variance in perceptions of control over anxiety-related events (again, see section on method for analytic details) above and beyond the variance explained by the full-scale ASI total score. This hypothesis was based upon theory (Craske, 1999, 2003) and empirical evidence (Sanderson, Rapee, & Barlow, 1989; Zvolensky, Eifert, Lejuez, & McNeil, 1999) demonstrating that individuals at risk for panic-related problems tend to perceive aversive events as beyond their control. In addition, for both sets of taxonrelevant analyses, it was hypothesized that the full-scale ASI total score would not account for unique variance in either panic-relevant dependent variable above and beyond the AS taxon (indexed in two different analytic ways; see section on method for details).

Finally, to further test the nature of the taxon, it was hypothesized that the sum of the items most weakly correlated with AS taxon membership would not be associated with the two theoretically relevant criterion measures (i.e., greater bodily vigilance and lower levels of perceived control) after controlling for the continuous ASI total scores. That is, theoretically relevant associations between AS and panic-relevant criterion measures should not be observed for ASI items that are poor indices of the taxon if AS taxon membership categorically confers panic- relevant risk and thereby accounts for the association between dimensional indices of AS and panic vulnerability. Consistent with this view, we expected that the full-scale ASI total score would account for unique variance in both dependent variables above and beyond the sum of items most weakly correlated with AS taxon membership.

METHOD

Participants

Participants included 589 young adults (283 women; M age = 18.9, SD = 2.0), recruited from the University at Albany, SUNY, who received course credit in return for participation. The ethnic distribution was 386 (65.5%) European American, 68 (11.5%) African American, 36 (6.1%) Asian American, 53 (9.0%) Latino, 1 (.2%) Native American, 43 (7.3%) other, and 2 (.3%) did not specify. With regard to educational background, the participants included 369 (62.6%) freshmen, 142 (24.1%) sophomores, 53 (9.0%) juniors, 17 (2.9%) seniors, and 8 (1.3%) did not specify.

Measures

Anxiety Sensitivity . The ASI (Reiss et al., 1986) is a 16-item measure on which respondents indicate on a 5-point Likert-type scale (0 = very little to 4 = very much) the degree to which they are concerned about possible negative consequences of anxiety symptoms (e.g., "it scares me when I feel shaky"). The factor structure of the 16-item ASI is hierarchical. It yields three first-order factors (i.e., AS-physical concerns, AS-mental incapacitation concerns, and AS-fears of publicly observable anxiety reactions) and a single, higher order general factor (Zinbarg, Barlow, & Brown, 1997). The ASI has high levels of internal consistency (average alpha coefficient: .84) and good test-retest reliability ( r = .70 for three years; Peterson & Reiss, 1992). The factor structure and psychometric properties of the ASI have been replicated across diverse populations, testifying to its broad-based applicability (Carter, Miller, Sbrocco, Suchday, & Lewis, 1999; Schmidt & Joiner, 2002; Zvolensky, McNeil, Porter, & Stewart, 2001). The ASI is distinguishable from, and demonstrates incremental validity above and beyond, trait anxiety (Rapee & Medoro, 1994); thus, this construct is distinguishable from the frequency of anxiety symptoms (McNally, 1996).

Bodily Vigilance. The Body Vigilance Scale (BVS; Schmidt et al., 1997) was employed to assess attentional focus on interoceptive activity. The BVS is a four-item instrument on which respondents indicate on an 11-point Likert-type scale (0 = none to 10 = extreme) the degree to which they agree with a particular statement regarding attentional focus on bodily sensations. An example of one such item is, "I am the kind of person who pays close attention to internal bodily sensations." More specifically, the first three items measure attentional focus, perceived sensitivity to changes in bodily sensations, and the average duration of time spent attending to bodily sensations. A fourth item involves having participants rate their attention to 15 bodily sensations that characterize the physical symptoms of panic attacks (see Diagnostic and Statistical Manual of Mental Disorders , DSM-IV-TR ). Responses to the fourth item are averaged to yield a single score for that item. Summing the four items derives a total score for the BVS. Research suggests that the BVS has adequate internal consistency in clinical and nonclinical populations and can be used to assess changes in bodily attention resulting from cognitive-behavior therapy for panic disorder (Schmidt et al., 1997).

Perceptions of Control for Anxiety-Related Events . The Anxiety Control Questionnaire (ACQ; Rapee, Craske, Brown, & Barlow, 1996) was used to measure perceptions of control for anxiety-related events. The ACQ was initially designed to index perceived control over internal and external events/situations that are relevant to anxiety-related problems. Participants indicate their level of agreement on a 6-point Likert-type scale (0 = strongly disagree to 5 = strongly agree) for control-oriented beliefs (e.g., "When I am put under stress, I am likely to lose control"). Although the original ACQ development study found the measure to be composed of two factors (Rapee et al., 1996), subsequent work has not fully supported these earlier results for a variety of methodological reasons (Brown, White, Forsyth, & Barlow, 2004; Zebb & Moore, 1999). Brown and colleagues (2004) recently found a three-factor lower-order solution (Emotion Control, Threat Control, and Stress Control) that loaded on a single 15-item higher-order factor (global Perceived Control). This work, in turn, resulted in a 15-item revised version of the ACQ (i.e., certain items were removed from the original version of the instrument due to problematic factor loadings). In the present investigation, we utilized the revised global ACQ score to index a generalized perception of control for anxiety-related events. This decision was based on two considerations. First, previous work has not fully substantiated the lower-order ACQ factors as clinically and theoretically meaningful (Brown et al., 2004), and second, the generalized factor is most consistent with contemporary theoretical models of panic-related vulnerability (Barlow, 2002).

Procedure

After providing informed consent, participants completed the assessment instruments anonymously in a mass-testing format. The measures were randomly ordered so as to decrease the probability of response set biases. Participants were debriefed as to the study objectives prior to their departure.

Analytic Approach

Indicator Selection . Three theoretically and quantitatively based item-domains of the 16-item ASI were identified and used to build item-parcel manifest indicators of AS: (1) physical concerns, (2) mental incapacitation concerns, and (3) fears of publicly observable anxiety reactions or social concerns. These item-parcel indicators bore an item composition identical to a wellestablished factor structure of the ASI (Zinbarg et al., 1997; Zinbarg et al., 1999, for a review; i.e., physical concerns items 3, 4, 6, 8, 9, 10, 11, 14; mental incapacitation concerns items 2, 12, 15, 16; fears of publicly observable anxiety reactions or social concerns items 1, 5, 7, 13), and moreover, were consistent with the dominant theoretical model of AS (e.g., Taylor, 1999). Although there are various means of selecting manifest indicators for taxometric procedures (Haslam & Kim, 2002; Schmidt et al., 2004; Waller & Meehl, 1998), it is generally accepted that factor analytically derived indicators produce optimal taxometric results (Beauchaine, 2003; Haslam & Kim, 2002; Schmidt et al., 2004). Such factor analytically derived indicators limit artifactual nuisance correlations, meaningfully sample indicators from all facets of a construct, and optimize internal the consistency and distinctiveness of the indicators (Schmidt et al., 2004).

MAXCOV-HITMAX . First, the MAXCOV procedure (Meehl & Yonce, 1996; Ruscio, 2004) was conducted. On a rotating basis, each pair of indicators served as output variables and the third remaining indicator formed the input variable. Thus, to exhaust all possible bivariate combinations of output variables, three MAXCOV plots were generated. The MAXCOV analysis was conducted in line with recent recommended guidelines (Ruscio & Ruscio, 2004; Schmidt et al., 2004; Waller & Meehl, 1998). Specifically, equal-sized nonoverlapping intervals were used to divide each input variable. A priori, the conventional default of 25 intervals (Ruscio, 2004) was not used, because the sample size was relatively small. Using the conventional default to divide the input indicators could potentially compress too many cases (i.e., complement and taxon class members) in the last intervals of the input indicators. This strategy, in turn, could result in failure to detect a relatively low base rate taxon (Bernstein, Zvolensky, Kotov, et al., 2006; Bernstein, Zvolensky, Weems, et al., 2005; Schmidt et al., in press). In the present study, this problem is of specific concern with respect to the mental incapacitation and fears of publicly observable anxiety reactions indicators, as both scales contain relatively few items. Thus, to ensure a sufficient number of intervals to detect the expected low base-rate taxon, 35 fixed-size intervals were chosen a priori. Furthermore, 50 bootstrapped replications were used to divide cases into the 35 fixed- size intervals by repeatedly resorting cases along the input indicator at random. At each replication, the correlation of the two output indicators within each interval was calculated, and these calculations were averaged across all replications. Such bootstrapped replications may be particularly useful when MAXCOV input variables are divided by fixed-size intervals and placed arbitrarily between equal-scoring cases. Thus, the use of bootstrapped replications is intended to minimize sampling error (see Ruscio, 2004, for a discussion) and bolster the reliability of the results.

Covariance plot "nose-count" (i.e., the number of taxonic, ambiguous, and nontaxonic covariance plots across indicators) and the coherency of the standard deviation of the base rate estimates served as internal consistency tests. Consistency tests, which characterize the taxometric approach (Meehl, 1992), are intended to rule out the detection of a pseudotaxon by requiring (a) convergence of data on a common conclusion and (b) specific values that are extremely unlikely to result in the absence of a taxon. Similarly, consistency tests are important to guard against pseudodimensional conclusions (Haslam & Kim, 2002). Simulation studies (e.g., Haslam & Cleland, 1996; Meehl & Yonce, 1994, 1996; Waller & Meehl, 1998) indicate that convergent evidence of taxonicity and specific estimated parameter values are improbable in the case of a latent dimension. The specific criteria used here to conduct the covariance plot "nose-count" were based on the guidelines provided by Waller and Meehl (1998) and standards used to interpret the plots in previous AS research (e.g., Bernstein, Zvolensky, Kotov, et al., 2006; Bernstein, Zvolensky, Weems, et al., 2005; Schmidt et al., in press). In addition, only completely peaked MAXCOV plots were interpreted as evidence of latent taxonicity. Although in some instances right-end cusped MAXCOV plots also may represent latent taxonicity (Schmidt et al., 2004), such cusped plots may also represent positively skewed dimensional data. Thus, to be conservative and to guard against pseudotaxonic interpretation of the results, cusped plots were interpreted in this investigation as ambiguous rather than as conclusive evidence of taxonicity. In addition, plots that could not be clearly discerned as taxonic or nontaxonic were labeled ambiguous, and all other plots were labeled nontaxonic.

After taxonicity was tested, Bayesian diagnostics (Meehl, 1973, p. 214; Ruscio, 2004; Waller & Meehl, 1998) were used to (1) assign individual participants to the taxon and complement classes; (2) estimate the taxon base rate; and (3) estimate the validity of the manifest indicators to discriminate between taxon and complement class members. Nuisance correlation (i.e., intraclass covariance within the complement and taxon classes) was used to gauge the accuracy of the Bayesian estimated parameters. When data are taxonic, the level of nuisance correlation is a robust index of the accuracy of parameter estimates because it can be used as a marker of the sensitivity and specificity of latent class membership assignment (see Schmidt et al., 2004, for a discussion).

MAXEIG-HITMAX . Using the three item-parcel indicators, MAXEIG was used as an external consistency test of the MAXCOV procedure. External consistency tests are premised on the same notion as internal consistency tests, and are specifically intended to rule out pseudotaxa or pseudodimensions that result from an artifact (or artifacts) of a particular taxometric procedure under particular methodological or statistical conditions (Schmidt et al., 2004; Waller & Meehl, 1998). Notably, MAXEIG is a multivariate extension of MAXCOV. However, in contrast to MAXCOV, which calculates the bivariate correlations between pairs of output indicators at varying levels of an input indicator, MAXEIG derives the maximum eigenvalues of the covariance matrix of a set of output indicators at varying levels of an input indicator.

Although MAXCOV and MAXEIG are mathematically similar in so far as correlations and eigenvalues are related, they are also systematically different in a number of important respects (see Waller & Meehl, 1998) that justify their concurrent utility in the present study. For example, unlike MAXCOV, MAXEIG divides the input variables with overlapping intervals (i.e., windows with a conventional default value of 90% overlap). Overlapping windows enable investigators to detect very low base rate taxa, which may be particularly important when sample size is limited. Furthermore, MAXEIG affords a powerful and unique internal consistency test. This "inchworm" test is achieved by repeatedly conducting the MAXEIG analysis while systematically increasing the number of overlapping windows, thereby systematically decreasing the size of the subsamples used to calculate maximum eigenvalues in each overlapping window (Waller & Meehl, 1998). Plots of taxonic data yield an increasingly better defined unimodal peak, as the number of windows is increased, resembling the head of an inchworm. In the instance of dimensional data, in contrast, the systematic increase of the number of overlapping windows will produce plots that do not systematically peak (e.g., flat or cusped plots). Fifty bootstrapped replications were used to divide cases into overlapping windows by repeatedly re- sorting cases along the input indicator at random. At each replication, the maximum eigenvalue of the two output indicators within each interval was calculated, and these calculations were averaged across all replications. In the present investigation, we conducted the "inchworm" test by generating four MAXEIG plots, from 25 windows to 175 windows, in 50 window increments. These specific values were selected on an a priori basis to produce four sets of MAXEIG analyses starting at 25 windows (Ruscio, 2004), and to allow for a sufficiently wide range of windows to permit the detection of what was expected to be a low base rate taxon.

After taxonicity was tested, Bayesian diagnostics (Meehl, 1973, p. 214; Ruscio, 2004; Waller & Meehl, 1998) were used to (1) assign individual participants to the taxon and complement classes; (2) estimate the taxon base rate; and (3) estimate the validity of the manifest indicators to discriminate between taxon and complement class members. The level of nuisance correlation was used to gauge the accuracy of the Bayesian estimated parameters.

Parameter-Matched Monte Carlo Simulations. Following recently recommended guidelines for taxometric analyses (Ruscio, 2004; Ruscio & Ruscio, 2004), parameter-matched Monte Carlo simulated dimensional and taxonic data were derived. Simulated dimensional data were matched to the number of indicators, sample size, and importantly, the observed indicator correlation matrix, and the distributions of all indicators including their skew, kurtosis, and discrete values (Ruscio, 2004). Simulated taxonic data were similarly matched to the parameters of the research data. These Monte Carlo simulated data were derived for two primary purposes. First, for the purpose of a priori suitability testing (Ruscio & Ruscio, 2004), all proposed taxometric analyses were tested with the simulated dimensional and taxonic data before taxometric analyses of the research data were interpreted. By examining the degree to which we can distinguish between the taxometric plots of the simulated dimensional and taxonic data under the parameter-matched conditions of the research data, the simulations demonstrate the capacity of the research data to afford meaningful taxometric analyses. Thus, only if each taxometric procedure could distinguish between the simulated dimensional and taxonic data could the research data be interpreted meaningfully. 1 If the simulated data pass the suitability tests, then the simulated data permit a second primary function. Specifically, the simulations allow investigators to compare the shape of the research data plots to the plots of the simulated taxonic and dimensional data. Accordingly, if simulated dimensional and taxonic data are discernible, then a nontaxonic pattern of findings in the research data may be reliably interpreted as a marker of latent continuity or nontaxonicity. In addition, if simulated dimensional and taxonic data are discernible, then a taxonic pattern of findings in the research data can be reliably interpreted as support for taxonicity that is unlikely to be artifactual or pseudotaxonic. Thus, in addition to contrasting the taxometric plots of the research data to predefined criteria (Waller & Meehl, 1998), these parameter-matched simulations are useful because they provide a comparative idiographic benchmark for interpreting the research data with respect to their unique parameters and thereby preclude pseudotaxonic or pseudodimensional conclusions.

Tests of External Validity . After a taxon is identified by means of taxometric tests of internal latent structure, it is important to determine whether the taxon has import with respect to nomologically relevant external criterion variables (Lenzenweger, 2004; Schmidt et al., 2004). Specifically, it is important to ascertain whether relations observed between responses on the ASI and panic-related external criterion measures are consistent with the taxonic latent class model. Although there may be a number of possible analytic means to test this core conceptual question, arguably the most stringent test is to test whether the measurement of AS based on taxon membership offers any incremental validity above and beyond the continuous measure (i.e., fullscale ASI) from which the taxon was derived (Schmidt et al., in press), and concurrently to test whether the continuous measure offers any incremental validity beyond the measurement of AS based on taxon membership. Our approach to evaluating the explanatory utility of the AS taxon involved a number of systematically sequenced and conceptually interrelated analytic steps, as explicated below.

First, as in previous taxometric investigations that tested the external validity of taxa (Gangestad & Snyder, 1985; Schmidt et al., 2004; Waller, Putnam, & Carlson, 1996; Waller & Ross, 1997), we selected a subset of ASI items that best discriminated between taxometrically based dichotomous class membership (i.e., r > .40). For the remainder of this article, this subset of ASI items will be referred to collectively as the "ASI Taxon Scale." We then evaluated whether the ASI Taxon Scale accounts for the relations observed between the full-scale ASI total score and theoretically relevant panic-related external criterion measures (i.e., BVS and ACQ scores). Specifically, we tested whether the ASI Taxon Scale offers incremental validity beyond the full-scale ASI from which it was derived (i.e., after controlling for its constituent parts and for the additional 8 items not included in the ASI Taxon Scale). It is important to note that the BVS and ACQ were not used in the taxometric analyses to derive the taxon or to assign cases to the latent classes. Thus, only if the taxometrically derived taxon bore systematic import with respect to panic-related criterion variables could the ASI Taxon Scale demonstrate incremental validity. Such a pattern of results can come about only if the taxometrically based ASI Taxon Scale reflected systematic variance delimited by latent taxonic structure and thereby eliminated error inherent in indices of AS (e.g., ASI total scores) that include items not associated with the AS taxon. Thus, two additional hierarchical multiple regression analyses were conducted. The criterion variables for these analyses were the BVS and ACQ total scores. The full-scale ASI total score was entered into the first step of the model and the ASI Taxon Scale score was entered into the second step of the model. Two additional hierarchical multiple regression analyses were conducted to test whether the full-scale ASI total scores accounted for unique variance beyond the ASI Taxon Scale scores. Here, the ASI Taxon Scale score was entered into the first step of the model and the full-scale ASI total score was entered into the second step of the model. The same criterion variables were used.

Second, to further test the explanatory power of the ASI Taxon Scale, we evaluated whether the 8 ASI items not used to compose the ASI Taxon Scale (i.e., the items most weakly correlated with taxon membership) also demonstrated incremental validity beyond the full-scale ASI total score. Finding incremental validity of the ASI Taxon Scale while failing to find such incremental validity for the sum of the other 8 ASI items would further demonstrate the systematic import of the taxon. In contrast, finding that the 8 ASI items not used to compose the ASI Taxon Scale do, in fact, demonstrate incremental validity beyond the full-scale ASI total score would threaten the systematic importance of the ASI Taxon Scale, and thereby the AS taxon, in regard to panic vulnerability. Accordingly, two regression analyses were conducted, identical to those already described, except that instead of the ASI Taxon Scale score, the sum of the 8 ASI items not used to compose the ASI Taxon Scale was entered into the second step of the model. As before, two additional hierarchical multiple regression analyses were conducted to test whether the full-scale ASI total scores accounted for unique variance beyond the sum of the 8 ASI items not used to compose the ASI Taxon Scale. Here, the sum of the 8 ASI items not used to compose the ASI Taxon Scale was entered into the first step of the model and the full-scale ASI total score was entered into the second step of the model with bodily vigilance and perceived control serving as criterion variables.

Finally, we conducted a test of the AS taxon by evaluating whether dichotomous (categorical) taxon membership offers enhanced explanatory power relative to the dimensional model in terms of predicting bodily vigilance and perceptions of control for anxiety-related events. Specifically, we tested whether dichotomous taxon membership had incremental validity beyond the full-scale ASI total score with respect to predicting scores on the panicrelated criterion measures. Such a pattern of results would indicate that the taxometrically based dichotomization of AS reflected unique systematic variance delimited by latent taxonic structure and thereby eliminated error inherent in the dimensional measurement of AS. Thus, two regression analyses were conducted, identical to those already described, except that dichotomous taxon membership was entered into the second step of the hierarchical regression model. Again, two additional hierarchical multiple regression analyses were conducted to test whether the full-scale ASI total scores accounted for unique variance beyond dichotomous taxon membership. Here, dichotomous taxon membership was entered into the first step of the model and the full-scale ASI total score was entered into the second step of the model. The criterion variables were, again, bodily vigilance and perceived control for anxiety-related events.

RESULTS

Suitability Testing

First, the suitability of the data for MAXCOV analyses was tested (see Figure 1). Specifically, MAXCOV analyses were conducted with the parameter-matched simulated dimensional and taxonic data. The covariance plots of the simulated dimensional data were clearly cusped, whereas the covariance plots of the simulated taxonic data were distinctly and differentially peaked. Thus, the research data are likely suitable for MAXCOV analysis. Second, the suitability of the data for MAXEIG analyses was tested (see Figure 2). Specifically, MAXEIG analyses were conducted with the parameter-matched simulated dimensional and taxonic data. The "inchworm consistency" test of the simulated dimensional data failed to yield unimodally peaked plots as the number of overlapping windows was increased, whereas the MAXEIG plots of the taxonic data yielded unimodal peaks (Waller & Meehl, 1998). Thus, the research data are likely suitable for MAXEIG analysis. Consequently, because the parameter-matched data passed suitability tests, MAXCOV and MAXEIG analyses of the research data are likely to produce meaningful and reliable conclusions with respect to the latent structure of AS.

AS Taxonicity and Parameters

To ease interpretation, Table 1 provides a summary of the taxometric analyses. Consistent with prediction, MAXCOV analysis (Meehl & Yonce, 1996; Ruscio, 2004) of the three item-parcel indicators produced two unimodally peaked taxonic plots, one ambiguous plot and no nontaxonic plots (see Figure 1). This pattern of covariance plots provides evidence of AS taxonicity, as a ratio of greater than 1:1 between taxonic and nontaxonic plots is evidence of latent taxonicity (Schmidt et al., 2004). Furthermore, the MAXCOV covariance plots of the research data were not cusped like the parameter-matched simulated dimensional data, and in support of latent taxonicity, were more similar to the peaked parameter-matched simulated taxonic data (e.g., see Figure 1). Also consistent with prediction, base rate estimates from all indicator combinations were similar; the standard deviation of base rate estimates was .03 ( SD < .10 is desirable; Schmidt et al., 2004). The mean base rate estimate of the taxon was 14.3%. It is important to note that the overall nuisance correlation (i.e., within-class indicator correlations) was somewhat high ( r = .38), but within the tolerable range (Schmidt et al., 2004; Waller & Meehl, 1998). Thus, the Bayesian estimated parameters of the latent distributions are likely only approximations and not precise estimates. Furthermore, all three indicators discriminated between latent classes well (average indicator validity = 2.2 SD ). An effect size between taxon and complement class members of 1.2 SD or greater is desirable for Coherent Cut Kinetic taxometric procedures (Beauchaine & Beauchaine, 2002; Waller & Meehl, 1998).

MAXEIG analysis and the "inchworm" consistency test of the research data produced further strong evidence in support of the taxonic structure. Consistent with prediction and evidentiary of latent taxonicity, cusped plots became peaked for all three indicators as the number of overlapping windows was increased (see Figure 2; Ruscio & Ruscio, 1998; Waller & Meehl, 1998). In addition, the MAXEIG plots of the research data were dissimilar to the matched simulated dimensional data, and more similar to the parameter-matched simulated taxonic data in that they produced peaks consistent with the head of an inchworm (see Figure 2). Furthermore, the base rate estimates were similar across the three MAXEIG plots, yielding an average base rate estimate of 12.6% ( SD = .03; see Table 1). Furthermore, the overall nuisance correlation was somewhat high ( r = .35), but within the tolerable range (Schmidt et al., 2004; Waller & Meehl, 1998). Thus, the Bayesian estimated parameters of the latent distributions are likely only approximations and not precise estimates. Furthermore, all three indicators discriminated between latent classes well (average indicator validity was 1.8 SD ). Overall, the convergence of evidence between MAXCOV and MAXEIG analyses of the Monte Carlo simulated and research data supports the taxonic structure.

Tests of External Validity and Predictive Utility

First, we constructed the ASI Taxon Scale by examining the correlation matrix between taxometrically based dichotomous class membership and each of the 16 ASI items. Eight items were most strongly correlated with class membership. The other 8 items demonstrated smaller correlations ( r < .40) with class membership. The 8 items that were most strongly correlated with class membership (range of r = .40-.70) included all 4 items from the mental incapacitation concerns factor (i.e., items 2, 12, 15, 16), 3 of the 8 items from the physical concerns factor (i.e., items 9, 11, 14), and 1 item from the fears of publicly observable anxiety reactions factor or social concerns (i.e., item 13). Although somewhat arbitrary, 8 items were identified because they demonstrated stronger statistical associations with taxometrically derived taxon membership than the other 8 ASI items. It is important to note that in identifying these 8 items, we are not proposing that these items serve as a new subscale or revision of the ASI, particularly at this preliminary juncture in the taxometric study of AS. Instead, the purpose of this preliminary set of analyses is to provide a novel means to evaluate the AS taxon via bootstrapped taxometric analyses (Schmidt et al., 2004). It is certainly possible that slightly fewer or slightly more ASI items might serve as the best possible set of items to use in ultimately constructing an ASI-based taxon scale. Future study using Receiving Operating Curve (ROC) and Item Response Theory (IRT) strategies may be promising in so doing, but such a task is beyond the limits of the present investigation. As expected, the scores on the ASI Taxon Scale for cases assigned to the taxon class ( M = 17.9, SD = 4.2) were higher than for the cases assigned to the complement class ( M = 4.8, SD = 4.0). Furthermore, although half the length of the full-scale ASI and composed of items from all three factors, the internal consistency of the ASI Taxon Scale (α = .86) was very similar to the internal consistency of the full-scale ASI (α = .89). Finally, as predicted by previous factor analytic studies of the ASI (see Zinbarg et al., 1999, for a review), the ASI Taxon Scale score and the sum of the other 8 ASI items not used to compose the ASI Taxon Scale were strongly correlated at the zero-order level ( r = .72, p < .05).

Following the derivation of the AS Taxon Scale, we conducted the first set of hierarchical multiple regression analyses. The criterion variables were the total scores of the BVS and ACQ. The full-scale ASI-total score was entered at level one of the model, and the ASI Taxon Scale was entered at level two. For BVS, the full-scale ASI total score accounted for 21.8% of the variance ( p < .05; step 1 standardized β = .46). As hypothesized, the ASI Taxon Scale entered at level two in the model added a significant amount of unique variance (1.2%, p < .05; step 2 standard ized β = .29). For the ACQ, the full-scale ASI total score accounted for 10.7% of the variance ( p < .05; step 1 standardized β = -.32). As predicted, the ASI Taxon Scale added a significant amount of unique variance to the ACQ scores (.7%, p < .05; step 2 standardized β = -.20). Next, we conducted these two hierarchical multiple regression analyses again, except that we reversed the order of the predictor variables. That is, the ASI Taxon Scale was entered at level one of the model, and the full-scale ASI-total score was entered at level two. For the BVS, the ASI Taxon Scale score accounted for 23% of the variance ( p < .05; step 1 standardized β = .48). As hypothesized, the full-scale ASI total score entered at level two in the model did not add a significant amount of unique variance to BVS scores (.5%, n.s.; step 2 standardized β = .19). For the ACQ, the ASI Taxon Scale score accounted for 11.1% of the variance ( p < .05; step 1 standardized β = -.33). As predicted, the full-scale ASI total score entered at level two in the model did not add a significant amount of unique variance to the ACQ scores (.3%, n.s.; step 2 standardized β = -.13).

Next, we conducted the second set of hierarchical multiple regression analyses. The full-scale ASI-total score was entered at level one of the model, and the sum of the 8 ASI items not used to compose the ASI Taxon Scale (i.e., items 1, 3, 4, 5, 6, 7, 8, 10) was entered at level two. For BVS, the full-scale ASI total score accounted for 21.8% of the variance ( p < .05; step 1 standardized β = .46), whereas the sum of the 8 ASI items not used to compose the ASI Taxon Scale was associated with a significant amount of unique variance (1.2%, p < .05; step 2 standardized β = -.31). Here, it is important to note that the step 2 standardized β for the sum of the 8 other ASI items (cf. ASI Taxon Scale) was negative. For the ACQ, the full-scale ASI total score accounted for 10.7% of the variance ( p < .05, step 1 standardized β = -.32); the sum of the 8 ASI items not used to compose the ASI Taxon Scale was associated with a significant amount of unique variance (.7%, p < .05; step 2 standardized β = .21). The step 2 standardized β for the ASI Taxon Scale was negative, whereas the step 2 standardized β for the sum of the other 8 ASI items was positive for perceptions of control for anxiety-related events. In addition, we conducted these two hierarchical multiple regression analyses again, except that we reversed the order of the predictor variables. That is, the sum of the 8 ASI items not used to compose the ASI Taxon Scale was entered at level one of the model, and the full-scale ASI-total score was entered at level two. For BVS, the sum of the 8 ASI items not used to compose the ASI Taxon Scale accounted for 15.5% of the variance ( p < .05; step 1 standardized β = .39). As predicted, the full-scale ASI total score entered at level two in the model added a significant amount of unique variance to the BVS scores (7.4%, p <. 05; step 2 standardized β = .75). For the ACQ, the sum of the 8 ASI items not used to compose the ASI Taxon Scale accounted for 7.6% of the variance ( p < .05; step 1 standardized β = -.27). As predicted, the full-scale ASI total score entered at level two in the model added a significant amount of unique variance to the ACQ scores (3.7%, p < .05; step 2 standardized β = -.53).

For the third set of hierarchical multiple regression analyses, the ASI total score was entered at the first level of the model and dichotomous taxon membership was entered into the second level. For the BVS, the full-scale ASI total score accounted for 21.8% of the variance ( p < .05; step 1 standardized β = .46). As predicted, the dichotomous taxon membership added a significant amount of unique variance (.8%; p < .05; step 2 standardized β = .11). For the ACQ, the fullscale ASI total score accounted for 10.7% of the variance ( p < .05; step 1 standardized β = -.32). In contrast to prediction, the dichotomous taxon membership did not add a significant amount of unique variance (.1%, n.s., step 1 standardized β = -.03) for the ACQ. In addition, we conducted these two hierarchical multiple regression analyses again, except that we reversed the order of the predictor variables. That is, dichotomous taxon membership was entered at level one of the model, and the full-scale ASI total score was entered at level two. For the BVS, dichotomous taxon membership accounted for 11.5% of the variance ( p < .05; step 1 standardized β = .33). In contrast to prediction, the full-scale ASI total score entered at level two in the model added a significant amount of unique variance to the BVS scores (11.1%, p < 05; step 2 standardized β = .40). For the ACQ, dichotomous taxon membership accounted for 4.5% of the variance ( p < .05; step 1 standardized β = -.21). In contrast to prediction, the full-scale ASI total score entered at level two in the model added a significant amount of unique variance to the ACQ scores (6.3%, p < .05; step 2 standardized β = -.30).

DISCUSSION

Recent research suggests that AS is taxonic (Bernstein, Zvolensky, Kotov et al., 2006; Bernstein, Zvolensky, Weems, et al., 2005; Schmidt et al., in press). Yet, there has been little work evaluating the nomological nature of the AS taxon. The purpose of the present investigation, therefore, was to evaluate the AS taxon in relation to theoretically relevant cognitive processes associated with panic vulnerability among a sample of young adults.

The results of the MAXCOV test, the internal consistency tests, and the MAXEIG test and comparisons to Monte Carlo simulated taxonic and continuous data all indicated that AS is taxonic in this sample of young adults. Specifically, the MAXCOV and MAXEIG plots were either characteristically categorical or characteristic of latent discontinuity in shape. The MAXCOV base rate estimate was .143 and the three MAXEIG base rate estimates averaged .126. These results are consistent with the findings of previous taxometric work on AS using the 16-item ASI in a predominately male cadet sample (Schmidt et al., in press), the CASI among a mixed sample of youth (Bernstein, Zvolensky, Weems, et al., 2005), and the 36-item ASI-R among adults from six different countries (Bernstein, Zvolensky, Kotov, et al., 2006). Thus, the present findings, in conjunction with related AS taxometric work, suggest that the initial ambiguous structure of AS reported by Taylor et al. (1999) may have been due to methodological features specific to that investigation. For example, there may have been limitations related to the indicators employed (such as low validities), which were rationally grouped into pairs of 8 ASI items. Overall, the present results add to past work by replicating and extending the empirical evidence for the existence of a categorical AS taxon among young adults.

As predicted, the ASI Taxon Scale (composed of the 4 items from the mental incapacitation factor, 3 out of 8 items from the physical concerns factor, and 1 item from the publicly observable anxiety reactions or social concerns factor) accounts for significant variance in both bodily vigilance and perceptions of limited control over anxiety-related events. These significant effects were above and beyond the variance accounted by the ASI total score. It is important to note that interest in these analyses is naturally not in the absolute effect size (i.e., the step 2 variance accounted for by the ASI Taxon Scale was 1.2% and .7% for bodily vigilance and perceptions of control, respectively), but rather to make an initial attempt to gauge whether the taxon might account for the association between the full-scale ASI and panic-related external criterion measures. Specifically, the continuous index of AS (ASI total score) may be tapping taxonic variance in the construct and may thereby account for the association between the continuous index of AS and panic-relevant criterion indices. Thus, the issue is not that the measurement based on the taxonic latent structure of AS is necessarily better than the continuous index of the construct. Rather, the issue is that the latent taxon may help explain key aspects of the relationship between the AS and panic-relevant processes. To further evaluate the notion that the continuous index of AS (full-scale ASI total score) may be associated with panic-relevant criterion indices only because it taps taxonic AS variance, additional regression analyses were conducted. These analyses were designed to test whether the full-scale ASI total score accounts for unique variance in BVS and ACQ scores beyond the ASI Taxon Scale. The results here showed that the full-scale ASI total score did not account for unique variance beyond the ASI Taxon Scale in predicting BVS and ACQ scores. This pattern of findings across both sets of regression analyses suggests that the well-established continuous index of AS (i.e., 16-item ASI total score) is associated with panicrelevant criterion variables because it taps AS taxonic variance. In other words, the AS taxon is associated with panic vulnerability. This account, of course, needs to be further evaluated with additional analyses that include sets of items that discriminate poorly between the taxon (as we describe below).

Confidence in the ASI Taxon Scale findings is, in fact, strengthened by the complementary analyses testing an alternative hypothesis using the 8 items not used to compose the taxon scale. Here, the results showed that after controlling for the ASI total score, the sum of the 8 ASI items not used to compose the ASI Taxon Scale accounted for significant variance in both bodily vigilance and perceptions of control for anxiety-related events. After accounting for the variance explained by the full-scale ASI total score, the variance accounted for at the second step in these hierarchical regression equations, however, was in the opposite direction compared to the ASI Taxon Scale results. Specifically, whereas the ASI Taxon Scale, as would be predicted by theoretical models of panic disorder (Barlow, 2002), was positively associated with bodily vigilance and negatively associated with perceptions of control (i.e., higher ASI taxon, less control), the exact opposite pattern of findings was observed for the items not comprising the ASI taxon scale. These data suggest that the ASI items that are less strongly associated with taxon class membership do not relate to bodily vigilance and perceptions of control for anxiety-related events in a manner that would be predicted by biopsychosocial theories of panic disorder. That is, these non-taxon- relevant ASI items are associated with lower levels of vigilance for bodily sensations and greater degrees of control for anxiety-related events (i.e., associations in the opposite direction predicted by panic disorder theoretical models). This finding is particularly noteworthy in that the the sum of these items and the sum of items that composed the ASI Taxon Scale are strongly related to one another at the zero-order level ( r = .72, p < .001). The regression analyses testing whether the full-scale ASI total score accounts for unique variance in criterion measures beyond the sum of the 8 items not used to form the ASI Taxon Scale further support the AS taxon. Specifically, unlike the pattern of results observed for the analyses of the incremental or relative validity of the full-scale ASI total score beyond the ASI Taxon Scale, the full-scale ASI total score did, in fact, account for a significant amount of unique variance beyond the sum of the 8 items not used to form the ASI Taxon Scale for both BVS and ACQ scores. These findings, in conjunction with the ASI Taxon Scale findings reported above, increase confidence in the hypothesis that one reason ASI total scores are associated with panic-relevant outcomes may be the fact that some (but not all items) tap taxonic variance in the construct. The well-established continuous index of AS may therefore be associated with panic-relevant criterion variables because it taps AS taxonic variance.

The third set of regression analyses followed the approach used by Schmidt et al. (in press) using dichotomous taxon membership as the predictor variable. As expected, dichotomous taxon membership predicted a significant amount of variance in bodily vigilance after controlling for the continuous index of the construct (ASI total score). These findings extend past work by Schmidt et al. (in press), which found that dichotomous taxon membership was related to the incidence of spontaneous panic attacks and the frequent occurrence of such attacks in a largely male cadet sample. In contrast to prediction, however, taxon membership did not predict perceived control for anxiety-related events above and beyond the variance accounted for by ASI total scores. On one level, these data may suggest (1) that dichotomous taxon membership may relate to certain panic-related cognitive processes but not all (i.e., bodily vigilance but not perceptions of control for anxiety-related events); or (2) that within-taxon class variance is systematically important for some but not all panic-related processes. Consequently, dichotomization alone, even based on latent class structure, may discard some meaningful variance accounted for by continuous measurement. Additional regression analyses conducted in the present investigation testing whether full-scale ASI total scores accounted for unique variance above and beyond dichotomous taxon membership in predicting BVS and ACQ scores may help clarify these issues. These regression analyses demonstrated that the dimensional index of AS (full-scale ASI total scores) accounted for a significant amount of unique variance in both dependent variables beyond dichotomous taxon membership. In light of the findings supporting the explanatory utility of the ASI Taxon Scale beyond the dimensional index of AS (i.e., full-scale ASI total scores), the observed pattern of findings suggests that although the latent taxon is important, simple dichotomization based on taxon membership eliminates important within-class quantitative variability in AS, at least with respect to the assessed panic-relevant cognitive processes. In other words, the detection of latent taxonicity in AS does not mean that psychopathologists interested in panic vulnerability must disregard the likely meaningful latent continuity within the qualitatively distinct latent classes. Indeed, the existence of a taxon does not imply the absence of latent within-class quantitative gradations. It has been previously theorized that, when taxonicity is detected, a nonarbitrary qualitative difference is superimposed on within-class quantitative variability, and to ensure that theoretical development is meaningful, both these latent qualitative and quantitative forms of difference need to be addressed (Bernstein, Zvolensky, Kotov, et al., 2006; Pickles & Angold, 2003; Waller & Meehl, 1998). The present findings, therefore, provide empirical support for the importance of concurrent consideration of quantitative differences within qualitative forms of difference. Overall, these findings underscore the utility of taxometrics in refining our theoretical understanding of AS and its measurement. Taxometrics is not limited to forwarding categorical measurement in and of itself. Rather, it is useful in refining current measurement that reflects the latent structure of constructs.

There are a number of interpretative caveats to the current investigation and directions for future work. First, we should make clear that the samples used here were not epidemiologically defined. As a consequence, we recommend that the specific cutoff values, including the ASI Taxon Scale, not be used to determine taxon and nontaxon class membership beyond the present investigation (i.e., at a broad-based level). Thus, future taxometric study of AS could derive an ASI Taxon Scale and categorical cutoff scores in an epidemiological sample that could then be employed in other work. A second limitation, similar to that of previous work examining the structure of AS, is that a unimethod approach was used to index the construct. In terms of the taxometric study of AS, such unimethod assessment approaches may inflate intraclass covariance (i.e., nuisance correlation) due to shared method variance between indicators. Such issues, in turn, create nonoptimal conditions for taxometric tests to detect latent taxonicity and can disturb the accuracy of parameter estimates. In such cases, the typical outcome is a decreased proba bility of detecting taxonicity. Future investigations should replicate the present findings using a multimethod approach. Evaluating the predictive utility of the AS taxon using laboratory-based procedures such as biological challenge paradigms may be a useful next step in this regard.

A third limitation is that, given evidence that the predisposing factors for AS may be different for men and women (Jang, Stein, Taylor, & Livesley, 1999; Stein, Jang, & Livesley, 1999), future studies might usefully explore the extent to which taxometric results vary as a function of gender. We did not complete such analyses in the present study because of limitations in overall sample size. A fourth limitation is that we conducted a cross-sectional test. Although this methodological approach provides an understanding of the associations between outcome and AS at one point in time, it cannot be used to explicate the direction of association. Future work would benefit by using prospective methodologies to better understand the nature of the AS taxon in relation to panic-related vulnerability processes. These tests will be most powerful when they incorporate outcome variables indexing panic risk across levels of analysis (e.g., information-processing biases, self-regulation processes). Other work may find it useful to evaluate the association between the AS taxon and other risk-related processes often associated with panic-related problems such as drug use (e.g., cigarette smoking; see Zvolensky, Feldner, Leen-Feldner, & McLeish, 2005, for a review) and certain health behaviors (e.g., exercise frequency; see Salmon, 2001, for a review). This type of work would serve to meaningfully extend the taxometric study of AS beyond that of emotional vulnerability factors and integrate it with other theoretically relevant biobehavioral processes. Finally, it should be noted that in the construction of the ASI Taxon Scale, we utilized correlations with taxon membership to create the scale. This is one approach to bootstrapping taxometric analyses in constructing a "taxon scale." Future work may seek to integrate other analytic tactics in the construction of similar scales, including ROC and IRT.

Overall, the present study replicates and extends past work on AS taxonicity among a young adult sample using the 16-item ASI. The results suggest that the AS taxon, indexed in two analytic ways, is significantly associated with certain panic-related processes even after controlling for variability explained by the well-established dimensional index of AS. These results, in conjunction with the observation that the items most weakly correlated with AS taxon membership were associated with the panic-related dependent measures in directions opposite to those predicted by theory, suggest that the taxon may serve to explain the well-documented relations between AS and the studied panic-relevant cognitive processes.

Acknowledgments. This article was supported by National Institute on Drug Abuse research grants (R03 DA16307-01 and 1 R21 DA016227-01) awarded to Dr. Zvolensky. This project also was supported by a National Research Service Award predoctoral fellowship (F31 MH073205-01) awarded to Amit Bernstein.

NOTE

1. It is noteworthy that additional external consistency tests, MAMBAC and MAXSLOPE, were tested on the parameter-matched simulated data. MAMBAC analyses of simulated dimensional and taxonic data both yielded cusped plots, which in the case of taxonic data may be markers of a low base rate taxon, but may alternatively be a product of positively skewed dimensional data. Similarly, MAXSLOPE plots of the simulated dimensional and taxonic data were not distinguishable. Thus, these additional external consistency tests did not pass suitability testing and therefore were not used to test the latent structure of AS in the research data. Furthermore, we know very little about the consequences of elevated levels of nuisance correlation for the meaningfulness of the L-MODE output. It has been argued, however, that high nuisance correlations can distort the factor scores on which the L-MODE output is based, and thereby markedly distort the output (see Schmidt et al., 2004; Waller & Meehl, 1998). Thus, L-MODE may be best suited for taxometric studies using multimethod manifest indicators in which there is a need to combine and optimally weight the varied manifest indicators to yield the single most powerful indicator possible (see Schmidt et al., 2004, for a discussion). Because of the present study's unimethod source of manifest indicators (i.e., a single measurement tool), sufficient interindicator correlations, and factor analytically based item-parcel indicator construction, we opted not to use L-MODE.

[Reference]

REFERENCES

American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders (4th ed., text revision). Washington, DC: Author.

Barlow, D. H. (2002). Anxiety and its disorders (2nd ed.). New York: Guilford.

Beach, S. R. H., & Amir, N. (2003). Is depression taxonic, dimensional, or both? Journal of Abnormal Psychology, 112, 228-236.

Beauchaine, T. P. (2003). Taxometrics and developmental psychopathology. Development and Psychopathology, 15, 501-527.

Beauchaine, T. P., & Beauchaine, R. J. (2002). A comparison of maximum covariance and k-means cluster analysis in classifying cases into known taxon groups. Psychological Methods, 7, 245-261.

Bernstein, A., Zvolensky, M. J., Kotov, R., Arrindell, W. A., Taylor, S., Sandin, B., et al. (2006). Taxonicity of anxiety sensitivity: A multi-national analysis. Journal of Anxiety Disorders, 20, 1-22.

Bernstein, A., Zvolensky, M. J., Weems, C., Stickle, T., & Leen-Feldner, E. (2005). Taxonicity of anxiety sensitivity: An empirical test among youth. Behaviour Research and Therapy, 43, 1131-1155.

Brown, T. A., White, K. S., Forsyth, J. P., & Barlow, D. H. (2004). The structure of perceived emotional control. Psychometric properties of a revised anxiety control questionnaire. Behavior Therapy, 35, 75-99.

Carter, M. M., Miller, O., Sbrocco, T., Suchday, S., & Lewis, E. (1999). Factor structure of the Anxiety Sensitivity Index among African American college students. Psychological Assessment, 11, 525-533.

Cole, D. A. (2004). Taxometrics in psychopathology research: An introduction to some of the procedures and related methodological issues. Journal of Abnormal Psychology, 113, 3-9.

Craske, M. G. (1999). Anxiety disorders: Psychological approaches to theory and treatment. Boulder, CO: Westview Press.

Craske, M. G. (2003). Origins of phobias and anxiety disorders: Why more women than men? New York: Elsevier.

Gangestad, S., & Snyder, M. (1985). To carve nature at its joints: On the existence of discrete classes in personality. Psychological Review, 92, 317-349.

Gleaves, D., Lowe, M., Snow, A. C., Green, B. A., & Murphy-Eberenz, K. P. (2000). Continuity and discontinuity models of bulimia nervosa: A taxometric investigation. Journal of Abnormal Psychology, 109, 56-68.

Harris, G. T., Rice, M. E., & Quinsey, V. L. (1994). Psychopathy as a taxon: Evidence that psychopaths are a discrete class. Journal of Consulting and Clinical Psychology, 62, 387-397.

Haslam, N., & Beck, A. T. (1994). Subtyping major depression: A taxometric analysis. Journal of Abnormal Psychology, 103, 686-692.

Haslam, N., & Cleland, C. (1996). Robustness of taxometric analysis with skewed indicators: II. A Monte Carlo study of the MAXCOV procedure. Psychological Reports, 79, 1035-1039.

Haslam, N., & Kim, H. C. (2002). Categories and continua: A review of taxometric research. Genetic, Social, and General Psychology Monographs, 128, 271-320.

Jang, K. L., Stein, M. B., Taylor, S., & Livesley, W. J. (1999). Gender differences in the etiology of anxiety sensitivity: A twin study. Journal of Gender Specific Medicine, 2, 39-44.

Lenzenweger, M. F. (1999). Deeper into the schizotypy taxon: On the robust nature of maximum covariance analysis. Journal of Abnormal Psychology, 108, 182-187.

Lenzenweger, M. F. (2004). Considerations of the challenges, complications, and pitfalls of taxometric analysis. Journal of Abnormal Psychology, 113, 10-23.

McNally, R. J. (1996). Anxiety sensitivity is distinct from trait anxiety. In R. M. Rapee (Ed.), Current controversies in the anxiety disorders (pp. 214-227). New York: Guilford.

Meehl, P. E. (1973). MAXCOV-HITMAX: A taxometric search method for loose genetic syndromes. In Psychodiagnosis: Selected papers (pp. 200-224). Minneapolis: University of Minnesota Press.

Meehl, P. E. (1977). Specific etiology and other forms of strong influence: Some quantitative meanings. Journal of Medicine and Philosophy, 2, 33-53.

Meehl, P. E. (1992). Factors and taxa, traits and types, differences of degree and differences of kind . Journal of Personality, 60, 117-174.

Meehl, P. E. (1995). Bootstrap taxometrics. American Psychologist, 50, 266-275.

Meehl, P. E. (2004). What's in a taxon? Journal of Abnormal Psychology, 113, 39-43.

Meehl, P. E., & Yonce, L. J. (1994). Taxometric analysis: I. Detecting taxonicity with two quantitative indicators using means above and below a sliding cut (MAMBAC procedure). Psychological Reports, 74, 1059-1274.

Meehl, P. E., & Yonce, L. J. (1996). Taxometric analysis: II. Detecting taxonicity using covariance of two quantitative indicators in successive intervals of a third indicator (MAXCOV procedure). Psychological Reports, 78, 1091-1227.

Mineka, S., & Zinbarg, R. (1996). Conditioning and ethological models of anxiety disorders: Stress-indynamic context anxiety models. In D. Hope (Ed.), Nebraska symposium on motivation (pp. 135-210). Lincoln: University of Nebraska Press.

Peterson, R. A., & Reiss, S. (1992). Anxiety Sensitivity Index manual (2nd ed.). Worthington, OH: International Diagnostic Systems.

Pickles, A., & Angold, A. (2003). Natural categories or fundamental dimensions: On carving nature at the joints and the rearticulation of psychopathology. Development and Psychopathology, 15, 529-551.

Rapee, R. M., Craske, M. G., Brown, T. A., & Barlow, D. H. (1996). Measurement of perceived control over anxiety-related events. Behavior Therapy, 27, 279-293.

Rapee, R., & Medoro, L. (1994). Fear of physical sensations and trait anxiety as mediators of the response to hyperventilation in nonclinical subjects. Journal of Abnormal Psychology, 4, 693-699.

Reiss, S., & McNally, R. J. (1985). Expectancy model of fear. In S. Reiss & R. R. Bootzin (Eds.), Theoretical issues in behavior therapy (pp. 107-121). San Diego: Academic Press.

Reiss, S., Peterson, R. A., Gursky, M., & McNally, R. J. (1986). Anxiety, sensitivity, anxiety frequency, and the prediction of fearfulness. Behaviour Research and Therapy, 24, 1-8.

Ruscio, J. (2004). Taxometric programs in the R language. Retrieved from http://www.etown.edu/psychology/Documents/TaxProgDocR.PDF

Ruscio, A. M., Borkovec, T. D., & Ruscio, J. (2001). A taxometric investigation of the latent structure of worry. Journal of Abnormal Psychology, 110, 413-422.

Ruscio, J., & Ruscio, A. M. (2000). Informing the continuity controversy: A taxometric analysis of depression. Journal of Abnormal Psychology, 109, 473-487.

Ruscio, J., & Ruscio, A. M. (2004). Clarifying boundary issues in psychopathology: The role of taxometrics in a comprehensive program of structural research. Journal of Abnormal Psychology, 113, 24-38.

Ruscio, A. M., Ruscio, J., & Keane, T. M. (2002). The latent structure of posttraumatic stress disorder: A taxometric investigation of reactions to extreme stress. Journal of Abnormal Psychology, 111, 290-301.

Salmon, P. (2001). Effects of physical exercise on anxiety, depression, and sensitivity to stress-A unifying theory. Clinical Psychology Review, 21, 33-61.

Sanderson, W. C., Rapee, R. M., & Barlow, D. H. (1989). The influence of an illusion of control on panic attacks induced via inhalation of 5.5% carbon dioxide-enriched air. Archives of General Psychiatry, 46, 157-162.

Schmidt, N. B., & Joiner, T. E., Jr. (2002). Structure of the Anxiety Sensitivity Index: Psychometrics and factor structure in a community sample. Journal of Anxiety Disorders, 16, 33-49.

Schmidt, N. B., Kotov, R., & Joiner, T. E., Jr. (2004). Taxometrics: Toward a new diagnostic scheme for psychopathology. Washington, DC: American Psychological Association.

Schmidt, N. B., Kotov, R., Lerew, D. R., Joiner, T. E., & Ialongo, N. S. (in press). Evaluating latent discontinuity in cognitive vulnerability to panic: A taxometric investigation. Cognitive Therapy and Research.

Schmidt, N. B., Lerew, D. R., & Trakowski, J. H. (1997). Body vigilance in panic disorder: Evaluating attention to bodily perturbations. Journal of Consulting and Clinical Psychology, 65, 214-220.

Silverman, W. K., Fleisig, W., Rabian, B., & Peterson, R. A. (1991). Child Anxiety Sensitivity Index. Journal of Clinical Child Psychology, 20, 162-168.

Stein, M. B., Jang, K. L., & Livesley, W. J. (1999). Heritability of anxiety sensitivity: A twin study. American Journal of Psychiatry, 156, 246-251.

Taylor, S. (1999). Anxiety sensitivity. Mahwah, NJ: Erlbaum.

Taylor, S., & Cox, B. J. (1998). An expanded anxiety sensitivity index: Evidence for a hierarchic structure in a clinical sample. Journal of Anxiety Disorders, 12, 463-483.

Taylor, S., Rabian, B., & Fedoroff, I. (1999). Anxiety sensitivity: Progress, prospects, and challenges. In S. Taylor (Ed.), Anxiety sensitivity: Theory, research, and treatment of the fear of anxiety (pp. 339-353). Mahwah, NJ: Erlbaum.

Waller, N. G., & Meehl, P. E. (1998). Multivariate taxometric procedures. Thousand Oaks, CA: Sage.

Waller, N. G., Putnam, F. W., & Carlson, E. B. (1996). Types of dissociation and dissociative types: A taxometric analysis of dissociative experiences. Psychological Methods, 3, 300-321.

Waller, N. G., & Ross, C. A. (1997). The prevalence and biometric structure of pathological dissociation in the general population: Taxometric and behavior genetic findings. Journal of Abnormal Psychology, 106, 499-510.

Zebb, B. J., & Moore, M. C. (1999). Another look at the psychometric properties of the anxiety control questionnaire. Behaviour Research and Therapy, 37, 1091-1103.

Zinbarg, R. E., Barlow, D. H., & Brown, T. A. (1997). Hierarchical structure and general factor structure saturation of the Anxiety Sensitivity Index: Evidence and implications. Psychological Assessment, 9, 277-284.

Zinbarg, R. E., Brown, T. A., Barlow, D. H., & Rapee, R. M. (2001). Anxiety sensitivity, panic, and depressed mood: A reanalysis teasing apart the contributions of the two levels in the hierarchical structure of the Anxiety Sensitivity Index. Journal of Abnormal Psychology, 10, 372-377.

Zinbarg, R. E., Mohlman, J., & Hong, N. N. (1999). Dimensions of anxiety sensitivity. In S. Taylor (Ed.), Anxiety sensitivity: Theory, research, and treatment of the fear of anxiety (pp. 83-114). Mahwah, NJ: Erlbaum.

Zvolensky, M. J., Eifert, G. H., Lejuez, C. W., & McNeil, D. W. (1999). The effects of offset control over 20% carbon dioxide-enriched air on anxious responding. Journal of Abnormal Psychology, 108, 624-632.

Zvolensky, M. J., Feldner, M. T., Leen-Feldner, E. W., & McLeish, A. (2005). Smoking and panic attacks, panic disorder, and agoraphobia: A review of the empirical literature. Clinical Psychology Review, 25, 761-789.

Zvolensky, M. J., Kotov, R., Antipova, A. V., & Schmidt, N. B. (2005). Diathesis-stress model for panic-related distress: A test in a Russian epidemiological sample. Behaviour Research and Therapy, 43, 521-532.

Zvolensky, M. J., McNeil, D. W., Porter, C. A., & Stewart, S. H. (2001). Assessment of anxiety sensitivity in young American Indians and Alaska natives. Behaviour Research and Therapy, 39, 477-493.

[Author Affiliation]

Michael J. Zvolensky , PhD

University of Vermont, Burlington

John P. Forsyth , PhD

University at Albany, SUNY

Amit Bernstein, PhD

Ellen W. Leen-Feldner , PhD

University of Vermont, Burlington

[Author Affiliation]

Correspondence regarding this article should be directed to Michael J. Zvolensky, PhD, University of Vermont, Department of Psychology, 2 Colchester Avenue, John Dewey Hall, Burlington, VT 05405-0134. E-mail: Michael.Zvolensky@uvm.edu

Author Advanced search

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.