Hostility has been pointed out as an explanatory variable of aggressive, antisocial and deliquescent behaviors (Ansell, et al., 2015; Birkley & Eckhardt, 2015; Helfritz-Sinville & Stanford, 2014; Cavalcanti & Pimentel, 2016); as well as a risk factor for the health and the psychological well-being of people (Assari, 2017; Perk et al., 2012; Rafanelli et al., 2016; Fischer et al., 2015; Stoney, 2013), being, therefore, a construct of considerable interest in psychology. About it, some of the correlates of hostility found in the literature are: depression (Fischer et al., 2015); cardiovascular disease (Perk et al., 2012); Internet addiction in adolescents (Ko et al., 2014), cigarette (Bernstein et al., 2014) and drugs use (Ansell et al., 2015), aggressive behavior (Cavalcanti & Pimentel, 2016) and partner violence (Birkley & Eckhardt, 2015).

Although hostility is a widely studied subject, its definition is still quite ambiguous and often confusing, being commonly associated with concepts such as aggression and anger, or sometimes treated as synonyms. However, it should be stressed that these are conceptually distinct terms, for instance, aggression (aggressione) means any behavior directed to another individual with the intention of causing harm (Bushman, 2016). In turn, anger refers to emotional state encompassing feelings ranging from mild annoyance to intense fury and anger, accompanied by autonomic nervous system stimulation (Spielberger & Biaggio, 1992). And as far as hostility is concerned, it is constituted as a belief, expectation and negative attitudes toward persons and things (Smith1992; Siegman & Smith, 2013).

Despite this, Barefoot (1992) emphasizes that the aggression relates to a behavioral component; anger signals an affective-emotional component and in turn, the hostility comes to a cognitive component. In a different way, Matthews, Jamison And Cottington, (1985) highlighted the hostility as a multidimensional construct composed of cognitive, affective and behavioral components The cognitive component is defined as negative beliefs and attitudes in relation to others, including cynicism and mistrust. The affective component typically labeled as anger, refers to an unpleasant emotion that varies from the anger and irritation can be assessed in relation to the frequency, intensity and target. On the other hand, the behavioral component results from an attitudinal and affective component and is an action that aims to harm others, verbally or physically.

The definition to which present study was anchored was Derogatis et al. (1976), which understands hostility as a concept encompassing cognitive, emotional and behavioral aspects. For these authors, hostility concerns thoughts, feelings or actions, characteristics of the state of negative anger, reflecting traits such as aggression, irritability, anger and resentment.

Regarding evaluation of hostility, some instruments available in the literature are: (1) Scale of Hostility of Cook-Medley (Ho, Cook & Medley, 1954); (2) the Inventory of Hostility-Guilt (Buss & Durke, 1957); (3) Questionnaire of Aggression (Buss & Perry, 1992); and (4) Scale of Assessment of Symptoms (SCL-90-R-90, Derogatis, 1976).

Derived from the Minnesota Multiphase Personality Inventory (MMPI), the Scale of the Hostility of Cook-Medley (Cook & Medley, 1954) has 50 items and evaluates such factors as: cynicism, hostile affection, aggressive response, hostile assignment, social avoidance and other (Barefoot et al.,1991). Unlike Ho, which consisted in items from the MMPI, the Inventory of Hostility-Guilty (Buss & Durke, 1957) sought to group items into sub scales representing various aspects of hostility, such as: verbal and indirect hostility. This scale has 75 items in a format that is true or false and brings the total of 8 factors: seeking to highlight the indirect violence, hostility, irritability, negativism, resentment, distrust, hostility verbal and guilt. This measure although widely used, was not fully supported by factor analytic studies and has also been criticized regarding their used practice and theoretical clarity. In addition, it received criticism regarding the response scale employed, which was the true-false, that are known only results and what motivated the construction of the Hostility-Guilt Inventory of Buss and Durkee (Fernandez & Gregory, 2014).

In a review of the Hostility-Guilt Inventory of Buss and Durkee (1957), Buss and Perry (1992) developed the Aggression Questionnaire comprising 29 items that assess four factors: physical aggression, verbal aggression, anger and hostility. These items are answered on a scale of five points, Likert scale, ranging from 1=totally disagree to 5=totally agree. Its adaptation to Brazil was presented by Gouveia, Peregrino, Branco e Gonçalves (2008). A satisfactory internal consistency was found for the hostility factor, composed of 8 items (α = 0,77).

Finally, there is an emphasis in the Symptom Assessment Scale-90-R (SCL-90-R), which formulation was made by Leonard R. Derogatis departing from previous studies with a Hopkings Symptom Cheklist - HSCL, published on 1975 (Laloni, 2001). Composed of 90 self-assessment items ranging from 0 to 4, this scale aims to measure psychiatric and psychological symptoms of psychiatric and non-psychiatric patients, evaluating 9 symptomatic dimensions: somatization, obsessiveness, compulsiveness, interpersonal sensitivity, depression, anxiety, hostility, phobic anxiety, paranoid ideas and psychoticism. For the hostility factor, the original study pointed to an alpha of 0,84 (Derogatis, 1976), and considering its adaptation to the Brazilian context, made by Laloni (2001), presented an alpha of 0,79

Although above instruments have limitations, such as: great number of items Cook & Medley, 1954; Buss and Durkee, 1957) and lack of applicability to the clinical context (Buss and Perry, 1992; Buss and Durkee, 1957). For these reasons, the SCL-90 hostility subscale was chosen, considering that this measure is one of the instruments most used in practice and in the research focused on the evaluation of mental health, relevant both for use with clinical and non-clinical samples (Prunas et al., 2012; Prinz et al., 2013; Sereda & Dembitskyi, 2016; Carrozzino et al., 2016; Tan et al., 2015). Furthermore, its applicability was considered acceptable in different countries, such as, for example, Africa, Denmark, Norway, Italy, Thailand, Ukraine and China (Carrozzino et al., 2016; Chapman & Vines, 2012; Tan et al., 2015; Prunas et al., 2012; Sereda & Dembitskyi, 2016; Wongpakaran et al., 2011). In addition, although the SCL-90 hostility subscale had its parameters tested for the Brazilian context, it was analyzed together with the 90 items of the SCL-90, having been performed for more than a decade only for clinical sample (Laloni, 2001).

In this sense, this study aimed to know the psychometric parameters of the Hostility Subscale of SCL 90 in a non-clinical sample. Therefore, a cross-validation was performed, through two studies: the first one that used the exploratory factorial analysis and the second that started from the confirmatory factorial analysis.

Study 1: Hostility Subscale Exploratory Factor Structure

It was a cross-sectional psychometric study, whose objective was to know the evidences of its factorial validity and internal consistency, through exploratory analyzes.

Initially, to study the validity based on the internal structure of the Hostility Subscale, an exploratory factorial analysis was performed from the polyclonal correlation matrix, which was adequate [KMO = 0.75 and Bartlett’s Sphericity Test, [χ² (15) = 193.2, p <0.001)]. In order to determine the number of factors, the Hull method was chosen (Lorenzo-Seva et al., 2011), adopting the Unweighted Least Squares measurer (ULS) and the promin rotation, that detected to an unifactorial solution (Table 1), with eigenvalue equal to 2.42, explaining 40.33% of the total variance.

Table 1

Goodness-of-fit index: Method for drawing dimensions:		CFI (Comparative Fit Index). ULS
number of factors	Values of Goodness-of-fit	g.l	Values of the Scree test
0	0.00	6	0.000
1	0.97	12	22.167*
2	1.00	17	0.000

Note: *Directions about the number of common factors 1

As for the values of factorial loads, as shown in Table 2, they were higher than 0.40 and statistically different from zero (p <0.001) for all items. This version of the subscale had an internal consistency (Cronbach’s Alpha) of 0.71.

Table 2

Items	Factor
	1	h²
1. Sente-se facilmente aborrecido ou irritado?	0.56	0.31
2. Tem acessos de raiva que não pode controlar?	0.56	0.31
3. Se instiga a bater, ferir ou prejudicar alguém?	0.42	0.18
4. Se instiga a quebrar ou destruir coisas?	0.59	0.35
5. Entra em discussões frequentes?	0.57	0.32
6 Grita ou atira coisas?	0.50	0.25
Eigenvalue	2.42
Explained Variance	40,33%
Alfa de Cronbach	0.71

In general, the present study met its objective that was to know evidences of its factorial validity and internal consistency. An unifactorial structure was indicated, using the method of Hull-Comparative Fit Index, confirming the original study proposal (Derogatis., 1976). Factorial loads were higher than those indicated by Hair et al. (2009) delimiting 0.30; so that in the present study in general it was above 0.50; presenting only item 3 with a load of 0.42. In addition, there was a satisfactory internal consistency as suggested by the literature (Oviedo & Campo-Arias, 2005; Pasquali, 2010). Despite these results, it is essential to highlight that these are exploratory analyzes, and it is not known if the scores of the HS are affected by social desirability, which is common in terms of self-report measures For these reasons the Study 2 was used, as described below.

Study 2: Hostility Subscale Confirmation Analysis

This study was based on a cross-sectional psychometric study and was aimed at gathering additional evidence of factorial validity and reliability of HS in a new sample, based on confirmatory factor analysis (CFA).

The objective was to test the one factor model found in Study 1, for this, we proceeded with a CFA assuming that the set of items were saturated in a single factor. The results pointed to good adjustment indicators: χ² (9) =4,713, p < 0,001 χ² / g.l. = 0,52; [CFI= 1,00; GFI = 0,99; TLI = 1,04; RMSEA = 0,00, p-close = 0,59(IC 90% = 0,04; 0,09)].

In Figure 1, the final confirmatory factorial structure can be seen. All HS items presented statistically different saturations (factorial weights, λ) from zero (p <0.001).

Figure 1

There was also a positive correlation between the HS and the SDS (r = 0.37, p <0.01), which indicates that high scores on hostility are accompanied by high scores on social desirability, usual in self-report measures. Finally, the Cronbach’s alpha was calculated for the set of HS items, presenting a good index (α = 0.73), as well as a good composed reliability (CC = 0.73).

In short, the results confirm the unifactorial structure of the HS, bringing together 6 items that presented good indexes of adjustment, as indicated by authors of the area (Tabachnick & Fidell, 2013). In terms of precision indexes (Cronbach’s Alpha and composed reliability) were considered adequate since they met the recommendations of the literature (Oviedo & Campo-Arias, 2005; Pasquali, 2010). Finally, it was possible to verify correlations between hostility and social desirability, suggesting that this measure should take into account the control of desirability.

The present article sought to know the psychometric parameters of the HS for a Brazilian sample. The results confirmed the unifactorial structure proposed by the measure originally developed by Derogatis (1975). In this sense, it is suggested that this measure has construct validity, since the observed factors are in line with what was expected (Pacico & Hutz, 2015). The unifactorial structure was supported by the use of the confirmatory factorial analysis in which good indexes of adjustments were verified for the parameters adopted (RMSEA, CFI, TLI, GFI), as suggested by the literature (Byrne, 2010).

The exploratory study and confirmatory study indicators showed satisfactory psychometric adequacy, with acceptable reliability indexes above 0.70, as recommended by the theorists mentioned in the area (Nunnally, 1970; Oviedo & Campo-Arias, 2005). ; Pasquali, 2010), as well as a satisfactory composite reliability in Study 2. These results converge with those obtained by Prunas et al. (2012); Sereda and Dembitskyi (2016); Laloni (2001) who found good index of internal consistency in their studies for the HS.

It was also sought to verify if the answer bias commonly present in self-report research, social desirability (Stöber, 2001), would be related to the HS. The results pointed to a moderate relationship between them, which suggests, in turn, future investigations able to use the measure of hostility controlling the social desirability influence.

Regarding the limitations of this research, it is worth pointing out the regionalization of the students, exclusively form the state of Paraíba, and it is relevant to know how this measure can be understood by other states. In addition, it is important to consider the fragility of the sample composed exclusively of high school students, and it is important to know the validity of this scale for other age groups. Finally, the convergent validity of the HS with other measures of hostility (eg, Ho, Cook-Medley, 1954; Buss & Durke’s Hostility-Blame Inventory, 1957) was not examined; therefore, these aspects are suggestions for future studies. Furthermore, studies that can verify evidences of discriminant and predictive validity are also recommended; and to prove their temporal stability (test-retest).

Even if such limitations are presented, it is hoped to offer a relevant contribution to the theme, by making available to the Brazilian context the adaptation of a short and specific measure to evaluate hostility, with satisfactory psychometric qualities. Its application can be useful for the researches that pursues to know impacts of hostility on antisocial behavior, health and well-being. In addition it can contribute in the evaluation of hostile profiles, as well as in the construction of interventions next to groups that score high in hostility.