Abstract
Background
Disgust sensitivity is defined as a predisposition to experiencing disgust, which can be measured on the basis of the Disgust Scale and its German version, the Questionnaire for the Assessment of Disgust Sensitivity (QADS). In various studies, different factor structures were reported for either instrument. The differences may most likely be due to the selected factor analysis estimation methods and the small nonrepresentative samples. Consequently, the aims of this study were to explore and confirm a theorydriven and statistically coherent QADS factor structure in a large representative sample and to present its standard values.
Methods
The QADS was answered by N = 2473 healthy subjects. The respective households and participants were selected using the randomroute sampling method. Afterwards, the collected sample was compared to the information from the Federal Statistical Office to ensure that it was representative for the German residential population. With these data, an exploratory Promaxrotated Principal Axis Factor Analysis as well as comparative confirmatory factor analyses with robust Maximum Likelihood estimations were computed. Any possible sociodemographic influences were quantified as effect sizes.
Results
The datadriven and theoretically sound solution with the three highly interrelated factors Animal Reminder Disgust, Core Disgust, and Contamination Disgust led to a moderate model fit. All QADS scales had very good reliabilities (Cronbach's alpha) from .90 to .95. There were no agedifferences found among the participants, however, the female participants showed remarkably higher disgust ratings.
Conclusions
Based on the representative sample, the QADS factor structure was revised. Genderspecific standard percentages permit a populationbased assessment of individual disgust sensitivity. The differences of the original QADS, the new solution, and the Disgust Scale  Revised will be discussed.
Background
Disgust sensitivity describes an individual's timeinvariant, geneticallybased personality trait, a predisposition to reacting to specific materials with disgust [[1], p.111; [2]]. The specific materials that trigger disgust can be grouped into five theoretically proposed categories of disgust elicitors [3]: badly tasting substances can produce Distaste, which protects the body from poisons. Core Disgust can be provoked by rotten food, body products, rodents, and other small vermin in order to protect the body from disease or infection. Animal Nature/Reminder Disgust refers to sex, death, poor hygiene, and body envelope violations for protection against death and mortality. InterpersonalContamination Disgust protects the body by limiting the contact with strangers and other undesirables. Moral Disgust mainly protects the social order in case of moral offenses such as rape or murder.
Based on the 32 items of the original Disgust Scale (DS) by Haidt, McCauley, and Rozin [4] only three of the eight DS factors were found to be psychometrically stable. These three stable factors are Core Disgust, Animal Reminder Disgust, and Contamination Disgust with 25 items overall which represent the Disgust Scale  Revised (DSR) by Olatunji and colleagues [5]. The reliabilities of the DSR scales varied from (Cronbach's Alpha) .71 to .82.
Schienle, Walter, Stark, and Vaitl [1] translated all the items from the original DS into German and implemented a consistent fivepoint rating scale. Unfortunately, the translation led to insufficient psychometric properties. Therefore, 28 newly generated expertrated items  four items for each of the existing DS scales and four items for a new scale called Deformation  were included. In a second step, any item of the translated and extended DS whose Measurement of Sampling Adequacy came to > .70, was excluded. In a third step, the factor structure was tested.
The parallel analysis and an obliquely rotated Principal Axis factor analysis revealed five factors: Death/Deformation, Body Secretions, Spoilage, Poor Hygiene, and Oral Rejection. In order to adjust the unsatisfactory internal consistency of .66 of Poor Hygiene, four more items were generated. The following confirmatory factor analysis of the 39itemversion revealed a Root Mean Square Error of Approximation (RMSEA) of .06 which suggests a sufficient model fit. Two more items were excluded as their factor loadings were < .30. The final German instrument "Fragebogen zur Ekelempfindlichkeit" will be referred to as "Questionnaire for the Assessment of Disgust Sensitivity" (QADS) in the remainder of this article and can be found as Additional file 1. Even though modest to good reliabilities were reported for all the scales (.69 to .85), the calculations of the final version were based on a small nonrepresentative sample of N = 220 participants.
Additional file 1. Fragebogen zur Ekelempfindlichkeit (QADS). The original German questionnaire including the instruction.
Format: PDF Size: 28KB Download file
This file can be viewed with: Adobe Acrobat Reader
The discrepancies in the factor structures of the QADS and the DS/DSR as well as the following shortcomings led to the necessity of revisiting and possibly refining the QADS factor structure.
Most of the studies on the DS factor structure applied estimation methods that did not take into account the nonnormality of item distributions [5]. The exact estimation method of the QADS' confirmatory factor analysis had not been reported by Schienle et al. [1], therefore, the validity of that factor structure is questionable at best. Furthermore, the QADS factor structure was calculated based on a small, nonrepresentative sample and had not yet been replicated. From a theoretical point of view, the scales do not clearly represent or directly refer to the groups of disgust sensitivity elicitors as proposed by Rozin, Haidt, and McCauley [3].
Consequently, the first aim of this study was to find and confirm a new, statistically coherent, datadriven and theoretically sound factor structure on a large representative German sample while applying adequate estimation methods. The second aim was to compare the fit of this structure to the fit of alternative structures in the present sample. Third, the standard percentages will be accounted for.
Methods
Sample
The data collection was conducted on behalf of the University of Leipzig and obtained in the fall of 2004 from the USUMA Berlin polling institute within a representative, multitopic survey. The participants were interviewed at their homes. The households and the participants were selected by the randomroute sampling method which ensures that the sample resembles the population in its relevant characteristics [6]. This sample was then compared with the information from the Federal Statistical Office in order to obtain a truly representative sample of the German residential population. The coverage rate was 62.3% with 2591 participants aged 14 to 99. Of these, the number of N = 2473 who were native German speakers were examined. Further details concerning the individual sociodemographic properties can be found in Table 1.
Table 1. Sociodemographic sample characteristics
All the participants volunteered and received a data protection declaration in agreement with the Helsinki Declaration. The study was approved according to the ethical guidelines of the "German Professional Institutions for Social Research" [Arbeitskreis Deutscher Markt und Sozialforschungsinstitute, Arbeitsgemeinschaft Sozialwissenschaftlicher Institute, Berufsverband Deutscher Markt und Sozialforscher].
Instrument
The Questionnaire for the Assessment of Disgust Sensitivity, QADS [1] consists of 37 items rated on a fivepoint Likert scale (0 = not disgusting, 4 = very disgusting) with six to nine items per scale to be rated on how disgusting a statement is, e.g. "you try to eat monkey meat", "you touch a dead body", or "you are about to drink a glass of milk when you smell that it's spoilage". Internal consistencies of the five scales came to: Death/Deformation = .85, Body Secretion = .74, Poor Hygiene = .78, Spoilage = .72 and Oral Rejection = .69 [1].
Statistical Analysis
First, an item analysis was computed to test for the nonnormality of the item distributions. Second, an exploratory factor analysis (EFA) was conducted to find a datadriven, statistically coherent, and theoretically sound solution factor structure. Third, a confirmatory factor analysis (CFA) tested the fit of the factor solution in a representative sample while multiple fit indices of several models were compared to find the most preferable solution. According to Olatunji et al. [5] and Brown [7], specific estimation and extraction methods had to be chosen in case of nonnormality, i.e. Principal Axis extraction in the EFA [7] and robust Maximum Likelihood estimation in the CFA. Since conducting both, the EFA and the CFA, using the same sample would lead to an artificially increased model fit, the sample was randomly divided into two partial samples, one for each upcoming procedure (n_{CFA }= 1252, n_{EFA }= 1221). No significant differences of means were found between the two partial samples for all the QADS items and the sociodemographic variables with all t (1, 2550) < 1.45 (p_{twosided }> .15).
Possible gender differences were calculated with Student ttests. As Table 1 shows, age was divided into seven groups in order to compute a oneway analysis of variance and test for possible specific cohort effects. These effects might not be assessable through simple correlation analyses. Additionally, a subsequent ttest would be susceptible to biases caused by unequal cell sizes.
Results
Descriptive item analysis
This step was conducted using SPSS 16.0 as well as LISREL 8.80s/PRELIS 2.80s. Table 2 presents the item analyses for the QADS items. All 37 items were considered to be disgusting by at least 77% of the participants while the appraisal to 21 items was above the mean overall appraisal (M = 2.50). Consistent with the findings from the DS [5], most of the item distributions demonstrated significant univariate skewness and kurtosis as well as significant univariate nonnormality in the ShapiroWilk test of normality [8] with W > .77 (p < .01). Further, significant multivariate nonnormality was found with Mardia's multivariate skew (β_{1, p }= 83.9, χ^{2 }= 33417.6, p = .000) and Mardia's multivariate kurtosis (β_{2, p }= 1804.4, N (β_{2, p}) = 164.4, p = .000) [9]. Therefore, appropriate estimation procedures should be applied in the EFA [5,7].
Table 2. QADS item characteristics (N = 2349 to 2457)
Exploratory Factor Analysis (EFA)
The EFA (n_{EFA }= 1221) was computed using the Principal Axis extraction method in SPSS 16.0 in order to adjust to nonnormal item distributions [7]. A Parallel Analysis provided the most convincing argument for determining the number of factors to retain [10,11] and led to a fourfactorsolution (EFA eigenvalues: 14.53, 2.57, 1.52, 1.39, 1.04; parallel analysis average eigenvalues: 1.34, 1.30, 1.27, 1.25, 1.23). As all the items tended to measure disgust sensitivity in related domains (i.e. disgust elicitors), the factors were obliquely rotated with Promax. The userdefined Promax power parameter Kappa was systematically altered between 1, 2, 4, 5 and 6 to find the most appropriate solution (see [12] pp.190197). Exclusively, the items showing a substantial loading of > .30 on at least one factor were interpreted, i.e. nonhyperplane items (see [5], p.285; [7] p.130).
An EFA with four factors revealed one hyperplane item in several solutions (item 37) while Kappa = 5 led to the least number of complex items (items 5, 17, 27, 33, 36). In every one of the rotated factor solutions, at least two of the three items that had salient loadings on factor four, also loaded considerably on another factor. Factor four undermined the simplicity of a latent structure (see [13] p.335), was difficult to interpret, and accounted for less than five items, thus it was probably an obsolete factor. Furthermore, the literature on disgust sensitivity prefers threefactorsolutions whenever "moral disgust" is not to be considered [3,5].
Consequently, a similar EFA with three factors was computed. Kappa = 5 led to five complex items, but no hyperplane item. Table 3 presents factor loadings for the rotated threefactor solution including the complex items mentioned. The highly correlated factors (r_{12 }= .60; r_{13 }= .76; r_{23 }= .71) accounted for 11.3% of the total variance for factor one, and 10.8% and 12.3% for factors two and three, respectively. Due to factor intercorrelations, these sums of squared factor loadings need not be added up.
Table 3. QADS communalities, factor loadings (principal axis analysis, Promaxrotated, Kappa = 5, N = 1221) and itemscalecorrelations
The properties of the QADS factors can be seen in Table 4. The factor distributions did not tend to be significantly skewed, except for Animal Reminder Disgust (γ_{1 }= 47). All the distributions were significantly flatter than the Gaussian distribution (γ_{2 }= .23 to .71).
Table 4. QADS genderspecific descriptives (N = 2349)
Confirmatory Factor Analysis (CFA)
In order to test the proposed threefactor model and alternative models for their fit in the respective partial sample (n_{CFA }= 1252), several CFAs were computed using Mplus 5.1. The input data consisted of intervalscaled raw values. Missing data was excluded listwise, resulting in n = 1206 examined cases. The robust Maximum Likelihood Method of estimation was used in order to account for the significant nonnormality of the data [7]. The item assignments of the threefactor model can be seen in Table 2. In order to maximize the simplicity of the structure, no crossloadings between items were allowed, and errors were specified as random and uncorrelated. Items 3, 13 and 34 were chosen to be marker indicators since these items had the highest salient loadings on their respective latent factors. The unrestricted onefactor model and the fivefactor model proposed by Schienle and colleagues [1] constituted the alternative models.
For the threefactor model, all observed and standardized factor loadings were significant with p = .000. For factor one, loadings varied between .54 and .71 (standardized errors were .015 to .020), for factor two between .63 and .79 (.011 to .018), and for factor three between .52 and .74 (.014 to .020). The standardized residual variances of the items varied between .38 and .72. Similar to the EFA results, the three factors were highly and significantly intercorrelated (r_{12 }= .71; r_{13 }= .88; r_{23 }= .80; all p = .000).
To evaluate model fits, the following thresholds appear to be appropriate. The χ^{2}/df index should be < 2.0 as mentioned by Bollen [14]. According to recommendations by Hu and Bentler [15], the Comparative Fit Index (CFI) as well as the TuckerLewisIndex (TLI) must be above .95 and the Standardized Root Mean Square Residual (SRMR) under .08. Browne and Cudeck [16] found a RMSEA of less than .05 to be good and of .05 to .079 to be adequate. The Akaike Information Criterion (AIC) is used in the comparison of the model fit for those nonnested models (see [7], pp.175181), giving preference to the model with the lowest AIC.
The overall fit statistics suggest poor fit for the onefactor solution, χ^{2 }(629) = 5,296.048, p < .001, SRMR = .068, RMSEA = .079, TLI = .736, CFI = .750, and moderate fit for both the threefactor model, χ^{2 }(626) = 3,989.355, p < .001, SRMR = .061, RMSEA = .067, TLI = .81, CFI = .82 and the fivefactor model, χ^{2 }(619) = 6,967.598, p < .001, SRMR = .066, RMSEA = .055, TLI = .816, CFI = .829. The model comparison fit statistics revealed that the threefactor model (AIC = 122,571.362) has a slightly better fit than the onefactor model (AIC = 122,751.471) and definitely a better fit than the fivefactor model (AIC = 243,078.898). Therefore, the threefactor model seems to be superior to the alternative models. Modifications to this preferred threefactor model, e.g. by reassigning several or all crossloading items to other factors, or leaving these items out, were tested. However, at best, the fit indices resulted either in being inflated or in being equal.
Sociodemographic influences
The stability of the threefactorstructure was examined within several groups of the sample (N = 2743) using SPSS 16.0. Table 4 shows that on all QADS factors the female participants reported a significantly higher disgust sensitivity with all t (1, 2348) > 9.6 (p < .001), producing significant mediumsized effects on the Main Score (Cohen's d = 0.50), Core Disgust (Cohen's d = 0.44), Animal Reminder Disgust (Cohen's d = 0.45) and Contamination Disgust (Cohen's d = 0.51).
The oneway analysis of variance for age differences yielded significant but meaningless ageeffects on the factors such as all F (6, 2342) = 3.04 to 4.52 (p < .01) and all η^{2}_{part }< .02, making possible posthoc tests unnecessary. No factorspecific patterns occurred for gender or age.
Standard percentages of QADS
Since distributions tended to be nonGaussian and no meaningful age specificities but gender specificities had been found for QADS factors, standard percentages are presented separately for male and female participants in Tables 5 and 6.
Discussion
The present study aims at analyzing the factor structure of the Questionnaire for the Assessment of Disgust Sensitivity (QADS), a generalized selfreport instrument that assesses the individual disposition of 'disgust sensitivity'. For the first time, this analysis was based on a large representative sample.
The adjustment of the statistical procedures during the exploratory and confirmatory factor analyses to the nonnormality of item distributions led to a model with three statistically coherent, datadriven, theoretically sound, and highly correlated factors. This model moderately fits the presented data from a representative German sample, and fits better than an unrestricted onefactor model or the fivefactor model proposed by Schienle and colleagues [1].
Factor one of the preferred threefactor model assembles 15 items: all but one Oral Rejection item, five items of Body Secretions, three Spoilage items and two of Poor Hygiene. The idea behind this factor can be considered as Core Disgust according to Rozin and colleagues [3], i.e. disgust triggered by the threat of disease through mostly oral contact and a sense of offensiveness, including stimuli such as rotten food and body secretions.
Factor two consists of nine items: all the Death/Deformation items as well as two Spoilage items that overlap with Death as they point to dead creatures. This factor clearly reminds people of their animal origin and mortal nature as Animal Reminder Disgust suggests [3].
Factor three incorporates 13 items: seven Poor Hygiene items, three Spoilage items, two Body Secretions items and one Oral Rejection item. The major concept behind this factor can be referred to as Contamination Disgust, representing reactions toward the perceived threat of contagion through mainly nonoral contact, i.e. inhalation or skin contact (see [3,5] p.285).
It can be assumed that the three identified factors overlap mainly in content with the three factors of the Disgust Scale  Revised (DSR) [5]. Although 15 QADS items (items 8, 10 to 15, 27 to 33, 35, 36) are also included in the 25 DSR items, the QADS is more than a mere German DS translation. It contains several items that strongly contribute to the respective factor but are not included in the original DS, e.g. items 3, 13, 16, 19, 23, and 34. Future research could investigate a combined item pool of both the DSR and the QADS while examining the convergent validity of these measurement instruments.
Itemscalecorrelations were medium to high, replicating the findings by Olatunji and colleagues [5]. The QADS factors inherited very good reliabilities of .90 to .95, which were slightly higher than those reported by Schienle and colleagues [1] for the five original QADS factors (.69 to .85). They were also higher than those presented for the DS and DSR with a reliability of .84 [4] and an internal consistency of < .70 [5], respectively. This might have been due to diverse item scaling in the DS and its revised version or too few items per factor in the original QADS.
The QADS factors were demonstrated as unaffected by the age of the participants. This might be due to the probably timestable nature of the traitlike concept. Gender influences, however, were found. Women were moderately more sensitive to disgust than men. Similar gender effects were reported earlier [4,17] and are of importance as they mediate gender differences in disgustrelated disorders, e.g. contamination fear in individuals with an obsessive compulsive disorder [18].
Construct validity on a conceptual level should be given as the items were partly derived from established disgust sensitivity measures, i.e. the DS and the DSR. Construct validity on a factorial level is supported by recent findings by Olatunji and colleagues [19]. Convergent validity was also tested by van Overveld and colleagues ([20], see also [5]) using the Disgust Propensity and Sensitivity Scale by Cavanagh and Davey [21]. A comparison of QADS and DSR to this or other generalized disgust instruments such as Wronska's questionnaire [22] and the Disgust Emotion Scale by Kleinknecht and colleagues [23] still need to be carried out in large representative samples and with adequate statistical methods adjusted to possibly nonnormal item distributions.
Another aspect is that the concept of disgust sensitivity might need to be refined since some of the popular definitions rely on the empirical basis of now revised findings. The DSR, for example, explicitly relies on the theoretical concepts by Rozin and colleagues [3] instead of several more or less independent domains as the DS did. The underlying threefactor model of disgust sensitivity is further supported by recent findings as all three factors were demonstrated to be distinct in relation to personality characteristics, behavioral, and physiological reactions as well as clinical fears [19]. Olatunji and colleagues [19] speak of "kinds of disgusts" while referring to what we call "disgust sensitivity", hereby not precisely distinguishing between the emotion of disgust and the underlying traitlike preparedness to experience disgust, Van Overveld and colleagues [20] postulate an even more sophisticated view on disgust sensitivity. They argue that the concept itself should be renamed disgust propensity, i.e. how easily a person becomes disgusted in opposition to the tendency of experiencing something as 'horrid', or rather how strongly a person is bothered by it, which they call disgust sensitivity. It seems important to note that the QADS factors are named according to the disgust sensitivity concept by Rozin and colleagues [3] since both the QADS items and DSR items have not yet been divided into propensity items and sensitivity items.
Conclusion
Developed as a paperandpencil test with as few as 37 fivepoint scaled items, the QADS can be applied easily and quickly. In sum, the QADS is ageindependent, objective, reliable, and now standardized on a representative sample with tabulated standard percentages for both genders. On account of its advantageous design being comprised of three different factors, it can be used to assess specific disgust domains as well as dispositional disgust sensitivity via the Main Score. Limitations are due to unexplored overlaps with other disgust sensitivity measures, their items, and factor structures. To be specific, a comparison of the QADS, DSR and other generalized disgust sensitivity instruments still needs to be conducted. For theoretical reasons, including "moral disgust" items could be considered [3].
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
EB and CA were responsible for the conception and the design of the study as well as the acquisition of the data. GS and MR performed the statistical analysis and substantially contributed to the interpretation of the data. KP and SP contributed to the interpretation of the data, wrote the first and final version of the manuscript, and critically revised the manuscript for intellectual content.
All the authors read and approved the final version of the manuscript for publication.
Acknowledgements
The authors would like to thank all the participants of the study as well as the University of Leipzig for making the survey possible.
References

Schienle A, Walter B, Stark R, Vaitl D: Ein Fragebogen zur Erfassung der Ekelempfindlichkeit (FEE) [A questionnaire for the assessment of disgust sensitivity].
Zeitschrift für Klinische Psychologie und Psychotherapie: Forschung und Praxis 2002, 31:110120. Publisher Full Text

Olatunji BO, BromanFulks JJ: A taxometric study of the latent structure of disgust sensitivity: Converging evidence for dimensionality.
Psychol Assess 2007, 19(4):437448. PubMed Abstract  Publisher Full Text

Rozin P, Haidt J, McCauley CR: Disgust. In Handbook of Emotions. 3rd edition. Edited by Lewis M, HavilandJones JM, Feldman Barrett L. New York: Guilford; 2008:757775.

Haidt J, McCauley C, Rozin P: Individual differences in sensitivity to disgust: A scale sampling seven domains of disgust elicitors.
Pers Indiv Diff 1994, 16:701713. Publisher Full Text

Olatunji BO, Williams NL, Tolin DF, Abramowitz JS, Sawchuck CN, Lohr JM, Elwood LS: The disgust scale: Item analysis, factor structure, and suggestions for refinement.
Psychol Assess 2007, 19(3):281297. PubMed Abstract  Publisher Full Text

Koch A: ADMDesign und EinwohnermelderegisterStichprobe. Stichproben bei mündlichen Bevölkerungsumfragen [ADMDesign and samples from resident registries. Samples in oral population surveys]. In Stichproben in der Umfragepraxis [Samples in Survey praxis]. Edited by Gabler S, HoffmeyerZlotnik JHP. Opladen: Westdeutscher Verlag; 1997:99116.

Brown TA: Confirmatory factor analysis for applied research. New York: Guilford; 2006.

Shapiro SS, Wilk MB: An analysis of variance test for normality (complete samples).

Mardia KV: Measures of multivariate skewness and kurtosis with applications.
Biometrika 1970, 57:519530. Publisher Full Text

Horn JL: A rationale and test for the number of factors in factor analysis.
Psychometrika 1965, 30:179185. PubMed Abstract  Publisher Full Text

Ledesma RD, ValeroMora P: Determining the Number of Factors to Retain in EFA: an easytouse computer program for carrying out Parallel Analysis. [http://pareonline.net/getvn.asp?v=12&n=2] webcite

Gorsuch RL: Factor analysis. Hillsdale, NJ: Lawrence Erlbaum Associates; 1983.

Thurstone LL: Multiple factor analysis. Volume 335. Chicago: University Press; 1947.

Bollen KA: Structural equations with latent variables. New York: Wiley; 1989.

Hu LT, Bentler PM: Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives.
Structural Equation Modeling 1999, 6:155. Publisher Full Text

Browne MW, Cudeck R: Alternative ways of assessing model fit. In Testing structural equation models. Edited by Bollen KA, Long JS. Newbury Park: Sage; 1993:136162.

Templer DI, King FL, Brooner RK, Corgiat MD: Assessment of body elimination attitude.
J Clin Psychol 1984, 40:754759. PubMed Abstract  Publisher Full Text

Olatunji BO, Sawchuk CN, Arrindell WA, Lohr JM: Disgust sensitivity as a mediator of the sex differences in contamination fears.
Pers Indiv Diff 2005, 38:713722. Publisher Full Text

Olatunji BO, Haidt J, McKay D, David B: Core, animal reminder, and contamination disgust: Three kinds of disgust with distinct personality, behavioral, physiological, and clinical correlates.
J Res Pers 2008, 42:12431259. Publisher Full Text

van Overveld M, de Jong PJ, Peters ML, Cavanagh K, Davey GCL: Disgust propensity and disgust sensitivity: Separate constructs that are differentially related to specific fears.
Pers Indiv Diff 2006, 41:12411252. Publisher Full Text

Cavanagh K, Davey GCL: The development of a measure of individual differences in disgust. Paper presented to the British Psychology Society, Winchester, UK; 2000.

Wronska J: Disgust in relation to emotionality, extraversion, psychoticism and imagery abilities. In European perspectives in psychology: Theoretical, psychometrics, personality, developmental, educational, cognitive, gerontological. Volume 1. Edited by Drenth PJD, Sergeant JA, Takens RJ. Oxford, England: John Wiley & Sons; 1990::125138.

Kleinknecht RA, Kleinknecht EE, Thorndike RM: The role of disgust and fear in blood and injectionrelated fainting symptoms: A structural equation model.
Behav Res Ther 1997, 35:10751087. PubMed Abstract  Publisher Full Text
Prepublication history
The prepublication history for this paper can be accessed here: