^{*}These authors were equally contributed.

The aim of the present study was to examine the psychometric properties of the Korean Verbal Abuse Questionnaire (K-VAQ) that consists of 15 items related with life-time verbal aggression exposure.

A total of 5814 university students who agreed to take part in the study completed the K-VAQ, the Korean version of the Life Event CheckList (LEC-K) and Impact of Event Scale-Revised (K-IES-R). Internal consistency was checked by using item-total item correlation and Cronbach's alpha coefficient. Exploratory and confirmatory factor analyses were performed, and convergent and concurrent validity levels were examined. Finally, a cluster analysis was conducted to verify the validity of the cutoff point of the K-VAQ.

The Cronbach's alpha correlation coefficient was 0.9. The K-VAQ showed a single factor structure which explained 55.34% of the total variance. The K-VAQ was significantly associated with the LEC-K (r=0.24) and K-IES-R (r=0.28), indicating good convergent validity and concurrent validity. The cluster analysis provided four clusters of trauma experiences: high, moderate, low, and minimal, with K-VAQ ranges of 43-81, 20-42, 7-19, and 0-6, respectively. In a further investigation, a K-VAQ score of 40 was found to be the appropriate cutoff point to delineate the highly verbally abused group, as used in the previous studies. A sum of 36.5% of the highly verbally abused group reported to show substantial symptoms of PTSD (K-IES-R score >22).

The present findings suggest that the K-VAQ has good psychometric properties for assessing verbal aggression among the Korean population.

Verbal aggression is one of the most common forms of emotional abuse. The fraction of 31.3% of Korean women report to have had parental verbal violence history,

The harmful effects of VA, compared to physical and sexual abuses, are frequently ignored by those individuals who significantly influence children.

The verbal abuse questionnaire (VAQ) covers various domains of VA and is a useful tool for detecting a history of VA and measuring the severity of VA experiences. Also, the VAQ is widely applicable regardless of whom the perpetuator is in any specific case. Previous studies have shown that it has adequate internal consistency for both parental VA and peer VA.

The data in this study was obtained from the online annual medical and psychiatric checkups of 5,814 unselected college students, both undergraduates and graduates, of the Korea Advanced Institute of Science and Technology (KAIST). The subjects were assessed between April of 2013 till January of 2014, and all participants voluntarily agreed to complete the questionnaires. The mean age of the subjects was 23.9 [standard deviation (SD)=4.49], and the majority of the sample subjects (77.5%) were male. This study was approved by the Ethics Committee of KAIST.

The Verbal Abuse Questionnaire (VAQ) can evaluate 15 types of verbal abuse, including scolding, yelling, swearing, blaming, threatening, demeaning, ridiculing, insulting, criticizing, screaming, belittling, and devaluating, among other types. This questionnaire uses a nine-point scale to measure the frequency of each item (0=not at all to 8=everyday) during life-time, and the total score is the sum of the 15 items (range, 0-120). The VAQ can also be applied to various types of VA. The VAQ shows adequate internal consistency for both parental VA (Cronbach's alpha=0.98 for maternal and Cronbach's alpha=0.94 for paternal) and peer VA (Cronbach's alpha=0.95 in females and Cronbach's alpha=0.96 in males).

The Life Events Checklist (LEC) is the most widely used scale to evaluate general traumatic events with adequate levels of reliability and validity.

Horowitz et al.

After obtaining the permission from the original author of the VAQ scale, a forward-backward translation process was conducted. First, one psychiatrist and one bilingual student fluent both in Korean and English jointly translated the original VAQ into Korean. Secondly, a clinical psychologist, who had completed a doctoral degree in England and another psychiatrist who was fluent in English translated the Korean version back to English. Finally, a committee consisting of three psychiatrists, one psychologist, one neurologist, and one bilingual student reviewed and adjusted the K-VAQ to make it more apprehensible and adaptable to the Korean culture.

Participants' demographic characteristics were assessed using descriptive statistics. For internal consistency including item-total item correlation, the Cronbach's alpha coefficient was calculated. Exploratory factor analysis was performed using principle axis factoring (PAF) extraction methods with varimax rotation. Convergent and concurrent validity were assessed using Pearson's correlation coefficient. The aforementioned analyses were conducted using SAS software (version 9.3; SAS Institute, Cary, NC, USA). Confirmatory factor analyses were performed using the Mplus 7.1 software package. Five model-fit indices were chosen to determine the goodness-of-fit of the model: 1) the chi-squared statistic, 2) the non-normed fit index (NNFI, 0.95 or greater), 3) the root mean-square error of approximation (RMSEA, 0.06 or less); 4) the standardized root mean residual (SRMR, 0.08 or less), and 5) the comparative fit index (CFI, 0.95 or greater).

Additionally, a cluster analysis was conducted to verify the validity of the cutoff point of 40 of the VAQ; this value was also used in previous studies.

The Cronbach's alpha correlation coefficient for the K-VAQ was 0.93. The corrected item-total correlation coefficients ranged from 0.49 to 0.78, indicating an acceptable range.

The Kaiser-Meyer-Olkin Measure (KMO) value for sampling adequacy was 0.952, and the chi-value of Bartlett's test of sphericity was 59551.220 (p<0.001), which indicates that the data in the present study is satisfactory for use in a further exploratory factor analysis. The eigenvalue (>1) and a scree plot showed that two factors explained 60.32% of the total variance; Factors 1 and 2 accounted for 55.34% (eigenvalue: 8.30) and 9.80% (eigenvalue: 1.47), respectively. Factor 1 included eleven items reflecting swearing, insults, humiliation, threats, screaming and blaming. Factor 2 consisted of the four items of raising ones' voice, yelling, scolding, and blaming. All rotated factor loadings exceeded 0.50. However, we noted that, while items 1, 3 and 15 generally reflected the contents of raising voices, item 12 ("scream at you for no apparent reason") of Factor 1 also implies a raised voice. The contents of item 4 classified in Factor 2 ("blame you for things") also overlap those of item 9 ("blame you for things you didn't do") of Factor 1. Furthermore, the variance of Factor 2 appeared to be small compared to that of Factor 1. In light of these findings, alternatively, we performed the same analysis using a fixed number of factors as a single factor. As a result, a single factor accounted for 55.34% (eigenvalue: 8.30) of the total variance. All factor loadings were over 0.50, confirming that all items could be classified as a single factor.

Exploratory factor analyses were conducted for single-factor and two-factor models. The single-factor model showed a very good fit for the data, [χ^{2}(43)=809.897, p=0.000, NNFI= 0.987, CFI=0.969, RMSEA=0.055, and SRMR=0.02]. The two-factor model showed a poor fit for the data, [χ^{2}(87)= 3987.609, p=0.000, NNFI=0.934, CFI=0.921, RMSEA=0.088, and SRMR=0.49] (^{2}=3177.712 > χ^{2}._{05} (40)= 55.76 ~ χ^{2}._{05} (50)=67.50. Δd.f=44]. These results confirmed that the single-factor model is more suitable for verbal aggression compared to the two-factor model.

The Pearson correlations were calculated in order to determine convergent and concurrent validity between the K-VAQ, the LEC-K and the K-IES-R. These results are shown in

The subjects were divided into four groups according to the bend point in the SSE plot with an appropriate number of subjects in each group (^{2}(3)=212.42, p<0.001]. Also, in a post-hoc analysis, the highly verbally abused group showed a higher PTSD symptoms than the moderately verbally abused group [χ^{2}(1)= 12.16, p<0.001].

In our study, the K-VAQ showed adequate psychometric properties when used to measure the participants' parental VA experiences. The internal consistency of the K-VAQ (Cronbach's alpha=0.93) showed an acceptable value, similar to previous studies using the original version of the VAQ (Cronbach's alpha=0.98 for maternal and Cronbach's alpha=0.94 for paternal). Also, the item-total item correlation coefficients were within the optimal range (r>0.3).

In the exploratory factor analysis, two factors were originally extracted, accounting for 55.34% (eigenvalue: 8.30) and 9.80% (eigenvalue: 1.47) of the total variance, respectively. These results are consistent with the study conducted by Teicher et al.

To determine the dimensions of the K-VAQ scale, both one-factor and two-factor models were verified using structural equation modelling (SEM). This concluded that the single-factor model is more suitable than the two-factor model. The original version of the VAQ also used the total score of all questions to screen for the presence of a history of VA. This also did not categorize into two subscales. Therefore, our analysis support that the single-factor model can be useful to evaluate VA.

The convergent validity of the K-VAQ was confirmed with significant results from the correlation analysis between the K-VAQ and the LEC-K, a general scale used to evaluate traumatic life events.14 Although the LEC-K did not contain those items that were directly associated with verbal abuse, our results imply the existence of common characteristics in both measures. To substantiate, existing evidences show that verbal abuse leads to elevated psychiatric symptoms, similar to those from physical abuse or catastrophic natural disasters (e.g., earthquake, war).

Moreover, we evaluated the validity of original VAQ's cut-off score. To measure the appropriate cutoff score of the K-VAQ, we conducted a cluster analysis. In the previous research, a cutoff point of 40 was utilized, representing the top 10% of scores in randomly selected participants. This cutoff point showed face validity.

There are several limitations to this study. First, the study was performed with university students from a top-ranked university in Korea. Most of the participants were young, male and highly educated. Our subjects might have been raised in a good environment, including high level of social and economic status (SES). Given that childhood maltreatment are influenced by low SES,

Despite these limitations, this study suggests that the K-VAQ has good psychometric properties. As our sample size was quite large and young adults were suitable subjects, given that VA from parents or peers mainly occurred during their childhoods, our results are likely to be reliable. The K-VAQ can be widely administered to measure the severity of various types of VA and can be used as a screening tool for the presence of VA. This scale can be useful for teachers, clinicians, health care workers, psychologists, or anyone in close contact with young students.

Thanks to H.C. Park's help to translate the English version of the VAQ to Korean one.

This work was supported by grants NRF-2012R1A1A2001 and NRF-2006-2005372 (to B. Jeong) from National Research Foundation, and in part by grant for KAIST Future Systems Healthcare Project from the Ministry of Education, Science and Technology (N01130009, N01130010, G04110085 to B. Jeong).

^{*}rotated factor loadings exceeded 0.50

1=single factor model, 2=two factor model. NNFI: the non-normed fit index, CFI: Comparative Fit Index, RMSEA: the root mean-square error of approximation, SRMR: the standardized root mean residual

All p values<0.001. K-VAQ: Korean version of the Verbal Abuse Questionnaire, LEC-K: Life Event Checklist-Korean version, K-IES-R: Korean version of the Impact of the Event Scale-Revised