48 (2007) 546–553www.elsevier.com/locate/comppsych

Comprehensive Psychiatry

Separation anxiety in adulthood: dimensional or categorical?Derrick Silovea,b,⁎, Tim Sladec,d, Claire Marnanec,e, Renate Wagnerc,e,

Robert Brooksa,b, Vijaya ManicavasagarfaPsychiatry Research and Teaching Unit, School of Psychiatry, University of New South Wales, Randwick NSW, 2031, Australia

bCentre for Population Mental Health Research, Sydney South West Area Health Service, Liverpool NSW, 2170, AustraliacSchool of Psychiatry, University of New South Wales, Randwick NSW, 2031, Australia

dClinical Research Unit for Anxiety and Depression at St Vincent's Hospital, Darlinghurst NSW, 2010, AustraliaeClinic for Anxiety and Traumatic Stress, Bankstown Hospital, Bankstown NSW, 2200, Australia

fThe Black Dog Institute, Randwick NSW, 2031, Australia

Abstract

Recent evidence suggests that a clinical form of separation anxiety can be observed in adults. An important question of relevance to definingthe construct of adult separation anxiety is whether there is discontinuity between that constellation and other forms of anxiety. In the presentstudy, 2 taxometric procedures—Mean Above Minus Below a Cut and Maximum Eigenvalue—were used to assess whether adult separationanxiety conformed primarily to a categorical or a dimensional pattern. The data were derived from a separation anxiety symptom questionnairecompleted by 840 consecutive adult patients attending an anxiety disorders clinic. Although some results of the analysis were ambiguous, theoverall findings suggested a dimensional pattern. The relevance of the finding to the status of adult separation anxiety is discussed.© 2007 Elsevier Inc. All rights reserved.

1. Introduction

Extensive research has been devoted to the study ofjuvenile-onset separation anxiety disorder (JSAD) [1,2], acommon disorder in the community [3,4] and the mostprevalent diagnosis presenting to childhood anxiety clinics[5]. The developmental trajectory of JSAD however remainsunclear [6-8], with some investigations suggesting that itcreates risk specifically to panic disorder in adulthood [9],whereas more recent studies have indicated that JSAD maybe a generic risk factor to a range of adult anxietysubcategories [10].

A third possibility, suggested by Manicavasagar andSilove [11], is that JSAD may persist, manifesting as an adultform of the disorder (the continuity hypothesis). Thattrajectory would be analogous to that of other early-onsetanxiety disorders, such as social phobia, that commonlyextend from adolescence into adulthood. The Diagnosticand Statistical Manual of Mental Disorders, Fourth Edition,

⁎ Corresponding author. Psychiatry Research and Teaching Unit, Level1 Mental Health Centre, Liverpool Hospital, Liverpool NSW 2170,Australia. Tel.: +61 2 9616 4311; fax: +61 2 9602 5917.

E-mail address: d.silove@unsw.edu.au (D. Silove).

0010-440X/$ – see front matter © 2007 Elsevier Inc. All rights reserved.doi:10.1016/j.comppsych.2007.05.011

Text Revision (DSM-IV-TR) [12], although classifying JSADas a disorder of childhood and stipulating onset before 18years of age, allows for the disorder to continue into later life.Nevertheless, until recently [13], the custom has been toexclude the putative subcategory of adult separation anxietydisorder (ASAD) from consideration in both epidemiologic[14,15] and clinical [16] studies focusing on the adultsubtypes of anxiety.

Yet there is growing evidence suggesting that separationanxiety can occur in later life in a form that seems equivalentto JSAD, although symptoms are modified somewhat bydevelopment [17]. The primary anxiety associated withASAD is the fear of actual or possible separation from closeattachment figures and a consequent preoccupation with thesafety and whereabouts of those persons. Anxieties extendbeyond parents to include intimate partners and children[11,17]. Whereas the criteria for JSAD highlight somaticsymptoms such as nausea and stomachaches [12], suchphysical complaints seem to be less prominent in adults whoinstead exhibit more cognitive and emotional symptoms[11]. Moreover, adulthood presents different opportunitiesfor those with separation anxiety to deal with their fears, forexample, by making frequent phone calls, by adhering torigid routines that ensure frequent contact with attachment

547D. Silove et al. / Comprehensive Psychiatry 48 (2007) 546–553

figures, or by talking excessively as a means of lengtheningcontact time with key others [17].

An important question, nevertheless, is whether theproposed novel category of ASAD should be accorded aseparate status equivalent to established subtypes of anxietysuch as panic disorder and social phobia. Several factorsdetermine the addition of disorders to the classificationsystem, including inter alia, the consistent clustering ofsymptoms, a discriminating risk factor profile, familyaggregation, a distinct period of onset, a predictable courseand outcome, and a demonstrated response to specificinterventions [18,19]. If the putative disorder is clearlydemarcated from related symptom patterns by conforming toa categorical construct, that observation may add to otherevidence supporting the distinctiveness of the syndrome(although the converse is not the case, that is, a dimensionalpattern does not on its own preclude recognition of a disorderas warranting a separate status in the diagnostic system). Thepresent study aims to examine whether symptoms of adultseparation anxiety conform to a categorical or dimensionalpattern. The present study builds on existing researchundertaken by our group [17,20-22] that has sought toexamine the phenomenology, developmental trajectory, andfamilial clustering of adult separation anxiety.

In pursuing these investigations [8,11,17], we haveestablished 2 methods for assessing the putative category ofASAD: a structured interview [17] and a self-report ques-tionnaire [23], each containing the same items. Items werederived from modified JSAD criteria and from clinicalobservations of patients thought to have the adult disorder.The structured interview allows trained clinicians to make aglobal clinical judgment about the presence or absence ofASAD, with past studies yielding high levels of interraterreliability [17,22]. The Adult Separation Anxiety Self-ReportQuestionnaire (ASA-27), previously called the Adult Separa-tion Anxiety Self-Report Checklist [21], is rated directly byrespondents on a 4-point symptom frequency scale. Receiveroperation characteristic analysis has shown a close concor-dance between the structured interview and the checklist [23].

Demonstrating continuities between a recognized child-hood disorder (JSAD) and a putative adult form (ASAD)offers support for the nosologic status of the latter. In acommunity sample, Manicavasagar et al [17] found that mostpersons with ASAD reported high levels of separationanxiety in their early years. A subsequent study undertakenat an anxiety clinic (N = 70) [21] confirmed thisdevelopmental association, with those classified as havingASAD reporting differentially high levels of early separationanxiety compared with other anxiety patients. A third,community-based study of adults with histories of schoolanxiety [8] showed a close association between ASAD andpast JSAD. Those reporting past JSAD had an 8-fold risk ofbeing assigned a current diagnosis of ASAD.

Since psychiatric disorders often cluster within families,demonstrating a pattern of aggregation adds indirect evidencein support of the status of a disorder. A study undertaken at a

juvenile anxiety clinic [22] found that parents of children withJSAD had an 11-fold increased risk of ASAD but no higherrate of other anxiety disorders when compared with parents ofchildren with other juvenile anxiety disorders. The resultssuggest a high level of specificity in the family clustering ofadult and juvenile separation anxiety.

As yet, however, no studies have assessed whether adultseparation anxiety conforms to a categorical or dimensionalpattern, that is, whether there is a “point of rarity”demarcating that pattern from other forms of anxiety. Thepresent study aimed to examine this issue by applyingtaxometric analyses to a large data set obtained from ananxiety clinic. Taxometric techniques were formulated byMeehl and colleagues [24-26] to determine the latentstructure of a particular phenomenon. Specifically, theanalysis applied herein aimed to assess whether the latentstructure of ASAD is best represented as an extreme point ona continuum of separation anxiety (ie, a continuous/dimensional structure) or whether the symptom patternclearly polarized respondents into 2 groups, hence indicatinga taxonic structure.

2. Methods

2.1. Participants

Participants comprised 840 consecutive patients attend-ing a public outpatient anxiety disorders clinic covering adefined catchment area in Sydney, Australia. Services areprovided free of charge, and there are no other specialistclinics for anxiety in the geographical area. Previousstudies [27] undertaken at the clinic have shown that thediagnostic and demographic profile of attending patientsare typical of those documented in similar anxiety clinicsworldwide [28].

The South Western Sydney Area Health Service EthicsCommittee provided ethics approval for the study, andrespondents completed signed consent forms.

2.2. Measures

Participants completed the ASA-27 [23]. The instrumentcontains 27 questions assessing separation anxiety symptomsoccurring after the age of 18 years. Each item is rated on a 4-point frequency scale: “This happens very often,” “Thishappens often,” “This happens occasionally,” and “This hasnever happened.” Items are assigned ratings of 3, 2, 1 and 0,respectively, yielding total scores ranging from 0 to 81.Questionnaire items have shown high levels of internalconsistency (Cronbach α = .89) and test-retest reliability (r =.86, P b .001) [17]. In a comparison with the structuredinterview, receiver operation characteristic analysis yielded ahigh area under the curve coefficient (0.9), indicating a closecorrespondence for the construct measured by the 2 instru-ments [23]. Another team has since validated a separatemeasure against the ASA-27, reporting a correlation coeffi-cient of 0.84 [29].

548 D. Silove et al. / Comprehensive Psychiatry 48 (2007) 546–553

2.3. Statistical analysis

Taxometric techniques developed by Meehl and Yonce[24,30] were applied to the data. The output from thetaxometric analyses is in the form of several graphs, theshapes of which indicate whether the structure of theunderlying construct represents discontinuity (ie, suggeststhe identification of a taxon) or is dimensional. Twotaxometric techniques were used in accordance withcontemporary recommendations [31]: Mean Above MinusBelow a Cut (MAMBAC) [30] and Maximum Eigenvalue(MAXEIG) [26].

The MAMBAC procedure calculates the difference inmean scores of one indicator above and below all possiblecutoff points on a second indicator. These individual valuesare then plotted on a graph. Dish-shaped or monotonicallyrising graphs indicate a latent dimension, whereas peakedgraphs indicate a latent taxon. For the MAXEIG procedure,the eigenvalues of the first factor derived from a factoranalysis of 2 or more indicators are plotted in subsamplesdefined by scores on another indicator. The MAXEIG graphsthat are indicative of a dimensional latent structure remainrelatively flat or, in the case of skewed indicators, rise towardone extreme. A taxonic latent structure produces noticeablypeaked MAXEIG graphs. Comparing the mean base rate

Table 1Specification of ASA-27 questionnaire items and the indicator set to which they b

1. More secure at home when with people who are close2. Difficulty staying away from home for several hours3. Carrying around something that gives security or comfort4. Extreme stress before leaving home to go on a trip5. Nightmares or dreams about being separated from someone close6. Extreme stress before leaving someone close to go on a trip7. Upset when usual daily routines are disrupted8. Worrying about the intensity of relationships with people close9. Headaches, stomachaches, or nausea before leaving for work10. Talk a lot to keep people close11. Concerned about where people close to you are going12. Difficulty in sleeping alone at night13. Better able to go to sleep if hearing the voices of people close to you14. Very distressed when thinking about being away from people close to you15. Nightmares or dreams about being away from home16. Worrying about people close to you coming to serious harm17. Upset about changes to routine that interfere with contact with people close to18. Worrying a lot about people close to you leaving19. Sleep better if the lights are on in the house or bedroom20. Tried to avoid being at home alone21. Sudden bouts of anxiety when thinking about being separated from those clos22. Get anxious if not speaking to people close to you on the telephone regularly23. Would not be able to cope or go on if someone you cared about left you24. Sudden bouts of anxiety when separated from people close to you25. Worrying a lot about possible events that may cause separation from attachme26. People said that you talk a lot27. Worrying that relationships are so close that they cause problems

estimates derived from taxometric analyses to a base ratesuggested either by the extant literature or clinical experiencepresents another source of data from which to derive con-clusions about the likely latent structure. If the mean base ratecontains little variation and is close to the proposed base rate,then this suggests evidence for taxonicity. If the mean baserate is highly variable and/or deviates markedly from theproposed base rate, then this is suggestive of dimensionality.

One of the important features of the taxometricmethodology is the use of multiple, semi-independentstatistical procedures. Confidence in the underlying structureis achieved through consistent findings across these multipletaxometric analyses. Another innovation of a recent programdesigned to carry out taxometric analysis is the generationand analysis of simulated data sets that mirror all theproperties of the research data but differ in their underlyingstructure [32]. If the research data are closer in structure tothe simulated dimensional data, then greater confidence isplaced on a dimensional solution. If, on the other hand, theresearch data are closer to the simulated taxonic data, then ataxon is supported. It has recently been shown that thesesample-specific simulations do not perform optimally whenthe base rate of the construct is low (ie, b10%) [33].However, as described below, the base rate for the currentstudy is 15% and therefore is safe from interpretational

elong

Theoreticallyderivedindicator set(n = 8 indicators)

Empiricallyderivedindicator set(n = 6 indicators)

Factoranalysis–derivedindicator set(n = 8 indicators)

54

27 14 4 2

35 8

84

1 5 56 3 66 61 3 77 12 5

you 6 31 2 7

65 6

e to you 1 78 6 71 2 7

1 7nt figures 3 4 7

48

549D. Silove et al. / Comprehensive Psychiatry 48 (2007) 546–553

difficulties brought about by examination of low baserate constructs.

A relative fit statistic can also be calculated that isrepresented as a positive value when the research data arecloser to the simulated dimensional data and as negativevalues when the research data are closer to the simulatedtaxonic data. Hence, traditional MAXEIG and MAMBACgraphs and the calculation of mean base rate estimates, alongwith the analysis of simulated data sets, provide acomprehensive suite of interpretative tools for the determi-nation of latent structure. The MAMBAC and MAXEIGanalyses of the research data, the simulated taxonic data, andthe simulated dimensional data along with estimation ofbase rates were all generated using programs developed byRuscio [31].

2.4. Selection and construction of indicator sets

In accordance with contemporary guidelines for conduct-ing taxometric analysis [34], 3 sets of indicators wereconstructed using different combinations of items from theASA-27. The first set of indicators was theoretically derived,being based on the structure of DSM-IV diagnostic criteriafor JSAD. Because formal diagnostic criteria for a putativecategory of ASAD currently do not exist and there is theforegoing evidence that the juvenile form extends intoadulthood, the 8 diagnostic criteria for JSAD were matchedto relevant items in the ASA-27 (see Table 1 for specificationof items and the indicator sets to which they belong).Indicator validity, as measured by the separation in standarddeviation units between the taxon and complement groups,ranged from 1.41 to 2.47, with all indicator validities beingabove the recommended cutoff score of 1.25 [25].

Two empirically based approaches were applied toconstruct the second and third sets of indicators. The secondindicator set identified items that correlated most highly withthe total ASA-27 questionnaire score, assuming that thoseitems would be the most discriminating. The 12 items withthe greatest correlations were selected and then pairedaccording to the largest correlations between items within apair, yielding a set of 6 indicators (see Table 1 for itemspecification). Indicator validities for this second set ofindicators ranged from 1.95 to 2.28, again all above therecommended cutoff of 1.25.

The third set of indicators was based on the results of anexploratory factor analysis of all ASAD items. A principal-components analysis indicated that between 5 and 11 factorsaccounted for the structure of the item pool. The large samplesize and number of factors examined made χ2 analysis anunreliable measure of fit for this model [35]. Instead,confirmatory factor analysis was applied that supported an8-factor model (see Table 1 for item specification) with asingle second-order factor producing a root mean squarederror of approximation of .047 [35]. Indicator validities forthis third set of indicators ranged from 1.31 to 2.25, onceagain all above the recommended cutoff of 1.25.

3. Results

The average age of the sample was 37 years (SD = 12.9).Most subjects were female (70%, n = 629) and born inAustralia (73%, n = 658), and approximately half wereemployed (51%, n = 463). Almost half were cohabiting withan intimate partner (49%, n = 442), with 36% being single(n = 320), 12% divorced or separated (n = 112), and 3%widowed (n = 27). Psychologists assigned the followingDSM-IV adult subcategories for anxiety (excluding ASAD):229 (26%) panic disorder with agoraphobia; 178 (20%)generalized anxiety disorder; 159 (18%) panic disorder; 118(13%) social phobia; and 28 (3%) each for posttraumaticstress disorder, obsessive compulsive disorder, and specificphobia. The remaining 14% comprised a broad range ofassignments including subthreshold anxiety disorders, mooddisorders, and anxiety disorder not otherwise specified.

In addition to these conventional diagnoses, a clinicaldiagnosis of ASAD was made where 2 psychologists (one ofwhom was the director of the clinic) concurred about itspresence. The overall prevalence estimate for ASAD at theclinic was 15%, the figure used in the taxometric analysis.

3.1. MAXEIG analyses

The MAXEIG analyses were carried out on all 3 sets ofindicators. The 3 panels in Fig. 1 show the averageMAXEIG graphs using each of the 3 sets of indicators forthe research data (left graphs), the simulated taxonic data(middle graphs), and the simulated categorical data (rightgraphs) (individual MAXEIG and MAMBAC graphs areavailable upon request). Average graphs for the theoreticallyderived and the empirically paired indicator sets all showedclear concordance in structure between the research data andthe simulated dimensional data. Average graphs for thefactor analysis–derived indicator set demonstrated lessclear, yet still evident, concordance between the researchdata and the simulated dimensional data. The relative fitstatistics were +3.76 for the theoretically derived indicatorset, +4.81 for the empirically paired indicator set, and +0.92for the factor-analyzed indicator set. The fit statistics for theempirically derived and theoretically derived indicator setswere strongly positive, supporting the concordance betweenthe research data and the simulated dimensional data. The fitstatistic for the factor-analyzed indicator set was moreambiguous, precluding a definitive conclusion being drawnabout whether a dimensional or taxonic structure waspreferred. The overall results, however, favored a dimen-sional latent structure.

A dimensional latent structure was further supported bythe highly variable mean (and SD) estimates of the base rate(theoretically derived, 0.33 [0.23]; empirically derived, 0.28[0.12]; and factor analysis derived, 0.21 [0.11]). The baserates produced by the analysis diverged substantially fromthe actual base rate of 15%, adding further evidence in favorof a dimensional latent structure.

Fig. 1. Average MAXEIG graphs for research data, simulated taxonic data, and simulated dimensional data from theoretically derived indicators (top panel),empirically paired indicators (middle panel), and factor-analyzed indicators (bottom panel).

550 D. Silove et al. / Comprehensive Psychiatry 48 (2007) 546–553

3.2. MAMBAC analyses

The MAMBAC analyses were carried out on all 3 sets ofindicators. Average MAMBAC graphs for the research data,the simulated taxonic data, and the simulated dimensionaldata are shown in Fig. 2. As with the MAXEIG analyses,MAMBAC analyses based on the theoretically derivedindicator set showed clear evidence of a dimensionalstructure, with the research data more closely resemblingthe simulated dimensional rather than the simulated taxonicmodel. The graphs from the empirically derived indicator setdemonstrate a right-hand cusp at the very end of thedistribution. This provides some evidence suggesting acategorical latent structure. Although the graphs from the

factor analysis–derived indicator set are more ambiguous,the research data showed greater similarity to the simulateddimensional data than the simulated taxonic data. Therelative fit statistics were +11.05 for the theoreticallyderived indicator set, +7.48 for the empirically pairedindicator set, and +6.63 for the factor-analyzed indicator set.As with the MAXEIG analysis, all the fit statistics werehighly positive, thus providing additional support for theconcordance between the research data and the simulateddimensional data.

A dimensional latent structure was also suggested by theresults of mean base rate estimation (theoretically derived,0.44 [0.09]; empirically derived, 0.42 [0.04]; and factoranalysis derived, 0.37 [0.06]). Although these mean base

Fig. 2. Average MAMBAC graphs for research data, simulated taxonic data, and simulated dimensional data from theoretically derived indicators (top panel),empirically paired indicators (middle panel), and factor-analyzed indicators (bottom panel).

551D. Silove et al. / Comprehensive Psychiatry 48 (2007) 546–553

rates are less variable than those derived from the MAXEIGanalysis, they are still highly divergent in relation to theproposed base rate of 15%.

4. Discussion

The present study sought to investigate the putativeconstruct of ASAD further by using taxometric analysis.Although some of the graphs from the MAXEIG andMAMBAC analyses were ambiguous with regard to latentstructure, on balance, the results suggested that adultseparation anxiety is best represented as dimensional ratherthan a categorical construct.

This finding is consistent with taxometric analysesundertaken on a range of psychiatric disorders thatconsistently reveal dimensional patterns to the underlyingconstructs. These include depression [36,37], obsessive-compulsive disorder [38], posttraumatic stress disorder [39],and pathological worry associated with generalized anxietydisorder [40]. Together with past research, the presentfindings support the view that the various forms of anxietymay be best conceptualized as interrelated dimensions [41].Yet the categorical (present-absent) approach remains anentrenched tradition in psychiatric classification because itsimplifies and facilitates clinical practice as well as therepresentation of prevalence estimates in epidemiologicstudies [42,43]. Recognizing the true underlying dimen-

552 D. Silove et al. / Comprehensive Psychiatry 48 (2007) 546–553

sional structure of a disorder nevertheless may help guide amore nuanced study of several aspects of the phenomenon,for example, in trying to identify its etiology and itsresponsiveness to treatment [44]. For example, where aconstruct is dimensional, it may be more meaningful tomeasure improvement after treatment using continuousmeasures. Symptoms of ASAD may diminish with treatmentbut they may not disappear altogether. This treatmentoutcome might be obscured if a categorical approach isapplied [45]. Similarly, in large-scale community studies,instruments producing a broad range of scores rankingpeople on severity may provide a more fine-grainedrepresentation of the phenomenon than would a present/absent diagnostic approach [40].

The strengths of the study are the large sample size, theuse of a psychometrically tested measure of adult separationanxiety, and the application of multiple taxonic indicators. Inparticular, a range of theoretically and empirically derivedcriteria was used, offering ample opportunity for one or moresets of criteria to emerge as a taxon if a categorical structureexisted. Nevertheless, it needs to be conceded that the 3indicator sets were derived from the same questionnaire sothat they represent overlapping conceptualizations of ASAD.A stricter test of the latent structure of ASAD may beachieved through analysis of entirely independent sets ofindicators derived from related yet distinct measures.Another limitation relates to the nature of the sample. Allparticipants had high levels of anxiety, a factor that mighthave reduced the potential to distinguish a category ofASAD. As such, the test applied might be regarded asconservative when compared with applying a taxometricanalysis to a general population study or a more hetero-geneous clinical population not constituted entirely ofanxiety patients.

5. Conclusions

This study indicates that although there were ambiguitiesin a minority of the results, the predominant findings suggestthat separation anxiety in adulthood may be best representedas a continuously distributed construct. The findings areconsistent with those emerging from taxometric analyses ofmost other adult anxiety subcategories [38,39]. The resultstherefore suggest that the symptom pattern alone may not besufficient to make a diagnosis of ASAD, but that othercriteria such as onset, course, family history, salience ofseparation anxiety compared with other symptoms, andassociated disability should all be taken into account inreaching a final diagnosis.

References

[1] Gittelman R, Klein DF. Relationship between separation anxiety andpanic and agoraphobic disorders. Psychopathology 1984;17(Suppl 1):56-65.

[2] Holmes J. Attachment theory: a biological basis for psychotherapy? BrJ Psychiatry 1993;163:430-8.

[3] Bowen RC, Offord DR, Boyle MH. The prevalence of overanxiousdisorder and separation anxiety disorder: results from the Ontario ChildHealth Study. J Am Acad Child Adolesc Psychiatry 1990;29(5):753-8.

[4] Moreau D, Weissman M. Anxiety symptoms in non-psychiatricallyreferred children and adolescents. In: Last CG, editor. Anxiety acrossthe lifespan: a developmental perspective. New York (NY): SpringerPublishing Company; 1993. p. 37-62.

[5] Last CG, Hersen M, Kazdin AE, Finkelstein R, Strauss C. Comparisonof DSM-III separation anxiety and overanxious disorders: demo-graphic characteristics and patterns of comorbidity. J Am Acad ChildAdolesc Psychiatry 1987;26(4):527-31.

[6] Silove D, Manicavasagar V, Curtis J, Blaszczynski A. Is earlyseparation anxiety a risk factor for adult panic disorder? A criticalreview. Compr Psychiatry 1996;37(3):167-79.

[7] Silove D, Manicavasagar V. Early separation anxiety and its relation-ship to adult anxiety disorders. In: Vasey MW, Dadds MR, editors. Thedevelopmental psychopathology of anxiety. New York (NY): OxfordUniversity Press; 2001. p. 459-80.

[8] Silove D, Manicavasagar V, Drobny J. Associations between juvenileand adult forms of separation anxiety disorder: a study of volunteerswith histories of school refusal. J Nerv Ment Dis 2002;190(6):413-4.

[9] Yeragani VK, Meiri PC, Balon R, Patel H, Pohl R. History ofseparation anxiety in patients with panic disorder and depression andnormal controls. Acta Psychiatr Scand 1989;79(6):550-6.

[10] Lipsitz JD, Martin LY, Mannuzza S, Chapman TF, Liebowitz MR,Klein D, et al. Childhood separation anxiety disorder in patients withadult anxiety disorders. Am J Psychiatry 1994;151(6):927-9.

[11] Manicavasagar V, Silove D. Is there an adult form of separation anxietydisorder? A brief clinical report. Aust N Z J Psychiatry 1997;31(2):299-303.

[12] American Psychiatric Association. Diagnostic and statistical manualof mental disorders, fourth edition, text revision. Washington (DC):American Psychiatric Association; 2000.

[13] Kessler RC, ChiuWT, Demler O,Walters EE. Prevalence, severity, andcomorbidity of 12-month DSM-IV disorders in the NationalComorbidity Survey Replication. Arch Gen Psychiatry 2005;62(6):617-27.

[14] Kessler RC, McGonagle K, Zhao S. Lifetime and 12-month prevalenceof DSM-III-R psychiatric disorders in the United States. Arch GenPsychiatry 1994;51:8-19.

[15] Kendler KS, Davis CG, Kessler RC. The familial aggregation ofcommon psychiatric and substance use disorders in the NationalComorbidity Survey: a family history study. Br J Psychiatry 1997;170:541-8.

[16] Solvason HB, Ernst H, Roth W. Predictors of response in anxietydisorders. Psychiatr Clin North Am 2003;26(2):411-33.

[17] Manicavasagar V, Silove D, Curtis J. Separation anxiety in adulthood:a phenomenological investigation. Compr Psychiatry 1997;38(5):274-82.

[18] Feighner JP, Robins E, Guze SB, Woodruff Jr RA, Winokur G, MunozR. Diagnostic criteria for use in psychiatric research. Arch GenPsychiatry 1972;26(1):57-63.

[19] Kendell RE. Clinical validity. Psychol Med 1989;19(1):45-55.[20] Manicavasagar V, Silove D, Hadzi-Pavlovic D. Subpopulations of

early separation anxiety: relevance to risk of adult anxiety disorders.J Affect Disord 1998;48(2-3):181-90.

[21] Manicavasagar V, Silove D, Curtis J, Wagner R. Continuities ofseparation anxiety from early life into adulthood. J Anxiety Disord2000;14(1):1-18.

[22] Manicavasagar V, Silove D, Rapee R, Waters F, Momartin S. Parent-child concordance for separation anxiety: a clinical study. J AffectDisord 2001;65(1):81-4.

[23] Manicavasagar V, Silove D, Wagner R, Drobny J. A self-reportquestionnaire for measuring separation anxiety in adulthood. ComprPsychiatry 2003;44(2):146-53.

553D. Silove et al. / Comprehensive Psychiatry 48 (2007) 546–553

[24] Meehl PE, Yonce L. Taxometric analysis: II. Detecting taxonicity usingcovariance of two quantitative indicators in successive intervals of athird indicator (MAXCOV procedure). Psychol Rep 1996;78(3, Pt 2):1091-227.

[25] Meehl PE. Bootstraps taxometrics solving the classification problem inpsychopathology. Am Psychol 1995;50(4):266-75.

[26] Waller NG, Meehl PE. Multivariate taxometric procedures: distin-guishing types from continua. Thousand Oaks (Calif): Sage; 1998.

[27] Wagner R, Silove D, Joukhador J, Manicavasagar V, Marnane C,Kirsten LT, et al. Characteristics of the first 1000 clients attending ananxiety clinic in South West Sydney. Aust J Psychol 2005;57(3):180-5.

[28] Swinson RP, Cox BJ, Kerr SA, Kuch K, Fergus KD. A survey ofanxiety disorders clinics in Canadian Hospitals. Can J Psychiatry 1992;37:188-91.

[29] Cyranowski JM, Shear M, Rucci P, Fagiolini A, Frank E, GrochocinskiVJ, et al. Adult separation anxiety: psychometric properties of a newstructured clinical interview. J Psychiatr Res 2002;36(2):77-86.

[30] Meehl PE, Yonce L. Taxometric analysis: I. Detecting taxonicity withtwo quantitative indicators using means above and below a sliding cut(MAMBAC procedure). Psychol Rep 1994;74(3, Pt 2):1059-274.

[31] Ruscio J. Taxometric software documentation. Retrieved 2004 fromhttp://www.etown.edu/psychology/faculty/ruscio.htm.

[32] Ruscio J, Ruscio AM, Keane TM. Using taxometric analysis todistinguish a small latent taxon from a latent dimension with positivelyskewed indicators: the case of involuntary defeat syndrome. J AbnormPsychol 2004;113(1):145-54.

[33] Beach SR, Amir N, Bau JJ. Can sample-specific simulations helpdetect low base-rate taxonicity? Psychol Assess 2005;17(4):446-61.

[34] Ruscio J, Ruscio AM. A conceptual and methodological checklist forconducting a taxometric investigation. Behav Ther 2004;35:403-47.

[35] Schumacker R, Lomax R. A beginner's guide to structural equationmodelling. New Jersey (NJ): Lawrence Erlbaum Associates; 1996.

[36] Ruscio J, Ruscio AM. Informing the continuity controversy: ataxometric analysis of depression. J Abnorm Psychol 2000;109(3):473-87.

[37] Slade T, Andrews G. Latent structure of depression in a communitysample: a taxometric analysis. Psychol Med 2005;35(4):489-97.

[38] Haslam N, Williams BJ, Kyrios M, McKay D, Taylor S. Subtypingobsessive-compulsive disorder: a taxometric analysis. Behav Ther2005;36(4):381-91.

[39] Ruscio AM, Ruscio J, Keane TM. The latent structure of posttraumaticstress disorder: a taxometric investigation of reactions to extremestress. J Abnorm Psychol 2002;111(2):290-301.

[40] Ruscio AM, Borkovec TD, Ruscio J. A taxometric investigationof the latent structure of worry. J Abnorm Psychol 2001;110(3):413-22.

[41] Brown TA. Validity of the DSM-III-R and DSM-IV classificationsystem for anxiety disorders. In: Rapee RM, editor. Currentcontroversies in the anxiety disorders. New York (NY): The GuilfordPress; 1996. p. 21-45.

[42] Kessler RC. The categorical versus dimensional assessment contro-versy in the sociology of mental illness. J Health Soc Behav 2002;43(2):171-88.

[43] Widiger TA, Samuel DB. Diagnostic categories or dimensions? Aquestion for the Diagnostic and Statistical Manual of MentalDisorders–fifth edition. J Abnorm Psychol 2005;114(4):494-504.

[44] Meehl PE. Factors and taxa, traits and types, differences of degree anddifferences in kind. J Pers 1992;60(1):117-74.

[45] Strube M. Evidence for the type in type A behavior: a taxometricanalysis. J Pers Soc Psychol 1989;56(6):972-87.