8th Alps-Adria Psychology Conference October 2-4, 2008, Ljubljana, Slovenia Local validation study of the Italian version of the Aggression Questionnaire (AQ) in Southern Italy M. Sommantico, M. Osorio Guzmàn, S. Parrello, B. De Rosa, A.R. Donizzetti Dipartimento di Scienze Relazionali G. Iacono Corso di Laurea in Psicologia dei processi relazionali e di sviluppo Università degli Studi di Napoli Federico II INTRODUCTION The Buss and Perry AQ (1992) remains even today one of the most widely used instruments in the evaluation of different levels of aggression in young adult and adolescent populations. The original version was validated across a sample of 1253 subjects and composed of 29 items, according to four factors: the first two, physical aggression (9 items) and verbal aggression (5 items) represent the functional or motor component of aggressive behavior; rage (7 items), the affective component of aggressive behavior, includes physiological arousal as well as the preparation for aggressive action; hostility (8 items) then represents the cognitive component of aggressive behavior. A strong correlation between physical aggression and verbal aggression emerges in the AQ validation study (Buss, Perry, 1992) as well as a moderate correlation with hostility; rage is significantly correlated with the remaining three factors. As noted, this version is quite reliable in evaluating the components of aggression. In terms of the AQ validation in other countries, many studies have been conducted in Europe, the Americas and Japan: - the majority of this studies (that used non-clinical subjects from student populations) support the four-factor structure of the AQ in terms of the internal coherence and the temporal stability of the single sub-scales in their reliability analysis, components and convergent validity (GarciaLéon et al., 2002; von Collani, Werner, 2005; Ando, 1999); - some of these studies underscore the AQ would be strengthened by the removal of some items, specifically the two with inverted scores (Nakano, 2001; Bryant, Smith, 2001; Vigil-Colet et al., 2005 ); - the results of studies that focused on the analysis of the psychometric properties of the AQ on specific clinical populations (psychiatric patients or aggressors), seem more contradictory in terms of being a good fit for the four-factor model or with respect to the internal consistency values of the individual sub-scales (Williams et al., 1996; Morren, Meesters, 2002; Fossati et al., 2003). Up to now the psychometric properties of the Italian version of the questionnaire of that questionnaire have been tested on three groups: a clinical sample (N = 461, average age = 33.9), and two non-clinical sample (N = 563 adolescents between the ages of 13 and 19; N = 1029 young adults between the ages of 20 and 35). All groups were taken from Northern or Central Italian populations. Essentially, the data from the validation study of the Italian AQ (Fossati et al., 2003) confirm that the instrument is statistically valid in measuring aggression, presenting sufficiently reliable internal consistency and a sufficient components validity in the sample taken into consideration. On the other hand, the adolescent and young adult population in Southern Italy presents different socio-economic characteristics (Parrello et al., 2008; SVIMEZ, 2008) than the other Italian contexts (Central and Northern). On the basis of the discussion above, the objective of the present study is to analyze the psychometric properties of the Italian version of the AQ across a representative sample of Southern Italian adolescents and young adults to verify by confirmatory factor analysis the questionnaire structure, compared in many studies. METHODOLOGY: Participants 860 subjects participated in the study (stratified sample; complete anonimous responses; free participation): males: 353 (41%) females: 507 (59%) average age: 20,10 (s.d. 3,70) city of Naples: 392 (45.6%) Naples surroundings: 468 (54.4%) Family socio-economic level: high: 6,86% medium-high: 25,81% medium: 31,86% medium-low: 16,62% low: 3,37% doesn’t answer: 15,46% Educational level: Secondary Schools students: 445 (51.7%): 238 (53.5%) Scientific training 207 (46.5%) Liberal arts training University students: 415 (48.3%): 196 (47.3%) Sciences 219 (52.7%) Liberal arts Instrument The Italian version of the AQ was used, composed of 29 items focused on evaluating physical aggression (9 items), verbal aggression (5 items), rage (7 items) and hostility (8 items), coded on a 5-point Likert scale where 1 represented entirely false for me and 5 represented entirely true for me. Statistical Analyses To verify the four-factor correlated model, proposed by Buss and Perry (1992) and then confirmed (Garcia-León et al., 2202; Fossati et al., 2003; Gerevich, Bácskai, Czobor, 2007), a Confirmatory Factor Analysis was run relying on the Lisrel 8.51 software (Jöreskog e Sörbom, 1993). To evaluate the gap between the reproduced matrix and the observed matrix, the research relied on the rapport between both the Chi-square test and degrees of freedom, since the former is not reliable when used with numerous samples as in the present study. The following indexes were also used: Mean Square Error of Approximation (RMSEA), Normed Fit Index (NFI), Non-Normed Fit Index (NNFI), Comparative Fit Index (CFI), Standardized Root Mean square Residual (SRMS), Goodness of Fit Index (GFI) and Adjusted Goodness of Fit Index (AGFI). An Exploratory Factor Analysis was subsequently conducting with SPSS.16 software, using the Principal Axis Factorization method (Oblimin rotation with Kaiser Normalization) and a Principle Components Analysis (Oblimin rotation with Kaiser Normalization). Using the same software, internal coherence was verified with Cronbach’s α values, and Variance Analyses were run to verify the existence of statistically significant differences in mean scores among sub-traits in relation to subjects’ gender and age – the latter defined based on enrollment in Secondary School or University. RESULTS: Validation study Confrontation between confirmative factor analysis on different models Model χ2/df RMSEA NFI NNFI CFI SRMR GFI AGFI four dimensions 5.10 (Buss e Perry, 1992) .07 .71 .74 .76 .07 .87 .85 two dimensions 8.48 (Williams et al., 1996) .09 .60 .61 .64 .085 .80 .76 The results of the analysis showed unsatisfactory values for each of the indexes considered therefore verifying the two-dimensional correlated model developed by Williams et al. (1996), a study that considers physical aggression together with rage and verbal aggression with hostility. The results derived from the application of this model indicate a deterioration of all indices taken into consideration. Confirming the inadequacy of models in the current literature with respect to the data collected in our study, we continued with an Exploratory Factor Analysis (Principle Axis Factorization method, Oblimin rotation with Kaiser Normalization). The scree test reading indicated the need to extract three factors and the saturation analysis led to the subsequent elimination of six items with saturations inferior to ±30 or with elevated saturations of more latent traits. The final result was composed from 23 items that collectively explain the overall 33.02% variance. As showed, the first factor (20.61% variance) mirrors the functional or motor subcomponent of aggressive behavior and can be entirely assimilated to the “physical aggression” trait; the second factor (8.0% variance) captures cognitive competence with the “hostility” trait items; the third factor (4.41% variance) is primarily composed of a mixture of items – negative signs – originally directed at the functional or motor subcomponent (verbal aggression) and at the emotional subcomponent (rage), therefore read as the “unsuccessful verbalization of rage.” Moreover, considering the correlation among the latent structures is became clear that the unsuccessful verbalization of rage trait is significantly correlated (.01, bilateral test) with physical aggression (.48) and hostility (.35), the latter being significantly correlated (.01, bilateral test) with physical aggression (.23). The resulting internal coherence was satisfactory considering: physical aggression (α .80), hostility (α .72) and the unsuccessful verbalization of rage (α .77). It should be highlighted that the exploratory procedure adopted by some researchers (Bryant, Smith, 2001; Garcia-Léon, et al., 2002; Gerevich, et al., 2007) is slightly but significantly different from that just described. In particular, appeal is generally made to Principle Component Analysis (PCA), with some exception (Buss, Perry, 1992). This is not considered a Exploratory Factor Analysis since it is a technique whose goal is summarizing information contained a totality of observed variables and not arriving at the identification of latent constructs. This is achieved by applying the method of principle factors or Principle Axis Factorization, regression analysis or maximum likelihood estimation (cf. Luccio & Paganucci, 2007; Giannini & Panocchia, 2006). Despite the inadequacy of PCA to reach our objectives, for the purposes of comparison, we chose to repeat the analysis adopting the PCA along with Oblimin rotation and Kaiser Normalization. On the basis of the scree test analysis, the four-trait structure was confirmed even with the necessary elimination of six items (7, 9, 11, 20, 23, 25) on the basis of saturation analysis. The final solution, including 23 items, explains the 44.59% total variance, but the distribution of the items among the four factors is distinct from Buss and Perry (1992): while the first two traits (physical aggression and hostility) are completely represented by relative items, the other two are composed from a mixture of items that originally referred to verbal aggression and rage. Moreover, the four factors remain significantly interrelated. On the basis of this factor solution, we evaluated the reliability of the scale in relation to the exactness or precision with which the entire scale and the individual sub-scales could estimate the different levels of the various components making up aggressive behavior. Although the first three sub-scales present a satisfactory internal coherence (respective α values of .78, .67 and .77), the final sub-scale presents an unsatisfactory internal coherence index (α .44). On the basis of our results, considering indications from the methodological literature (Pannocchia & Giannini, 2007; Fabrigar, Mac-Callum, Wegener & Strahan, 1999), the solution obtained by applying Exploratory Factor Analysis with Principle Axis Factorization was considered more pertinent. We consequently proceeded to confirm and verify this structure. The indexes to adjust the model to the data can be considered satisfactory [χ2/df = 4.19; RMSEA = .06; NFI = .82; NNFI = .84; SRMR = .06; GFI = .91; AGFI = .89]. Model graph 0.67 5 1.00 21 0.62 1 1.50 9 0.95 27 1.11 17 1.36 29 1.78 24 1.12 16 1.10 26 1.27 8 0.99 12 1.24 11 1.48 20 1.39 28 1.64 4 1.04 18 1.14 14 0.75 13 1.09 19 1.22 15 0.92 6 1.22 22 1.02 1.01 0.74 0.85 0.64 0.68 0.73 0.46 Physical Aggression 1.00 0.26 0.76 0.75 0.77 0.67 0.69 0.56 0.55 0.47 Hostility 1.00 0.59 0.41 0.89 0.74 0.80 0.76 0.66 0.42 0.77 Unsuccessfull Verbalisation of Rage 1.00 RESULTS: Subjects’ differentiation on the basis of age and gender The descriptive analyses for this three-dimensional model demonstrate mean scores not particularly elevated across the scale, as with the sub-scales. Mean scores Scale Total 2.54 PA 2.30 HOS 2.54 UVR 2.81 M F TOT F Sig. Univ. S.S. TOT F Sig. PA 2.73 1.99 2.29 195,716(1,858) 0.000 PA 2.11 2.46 2.29 36,751(1,858) 0.000 HOS 2,61 2.49 2.54 4,635(1,858) 0.032 HOS 2.48 2.59 2.54 3,957(1,858) 0.047 UVR 2.69 2.89 2.81 14,925(1,858) 0.00 UVR 2.72 2.89 2.81 10,956(1,858) 0.001 The variance analysis shows that mean scores of physical aggression and hostility are higher in males, while among females the unsuccessful verbalization of rage was higher. While mean scores of physical aggression, hostility and the unsuccessful verbalization of rage are higher in younger subjects, frequently Secondary School students, these scores are lower in older subjects. DISCUSSION In terms of the components validity, the results of the confirmatory factor analyses show that for the Italian version of the AQ, the four-factor structure, which was confirmed in other countries (Ando et al., 1999; Bernstein, Gesn, 1996; Bouchard, 2007; Gallardo-Pujol et al., 2006; Garcia-Léon et al., 2002; Meesters et al., 1996; Morren, Meesters, 2002; Rodríguez, Peña, Graña, 2002; von Collani, Werner, 2005), does not hold true with our sample. The quality of this instrument is proven by its clear and distinct evaluation of the functional or motor component of aggressive behavior (physical aggression); the emotional component (rage) remains linked to the other functional or motor sub-component (verbal aggression), similarly to the cognitive component (hostility). Such contextual specificity leads us to consider that in Southern Italy the emotional component of aggressive behavior is rooted in its verbal expression, therefore giving rise to a single trait and bringing back an item originally ascribed to physical aggression (“I get into fights a little more than the average person”). This last data in particular leads us to consider that the subjects’ reading of this item reveals an ulterior contextual specificity; among our subjects, picking a fight is read as a verbal-expressive aspect of rage without physical externalization. The results also show that the validity of the AQ overall is satisfactory as are the three sub-scales that make it up, although each in a different measure. In particular the functional or motor sub-component of aggressive behavior (physical aggression), has the highest Cronbach’s α values while the cognitive component (hostility) has the least elevated although still satisfactory. In line then with the original study by Buss and Perry (1992), and just as the studies conducted in other countries which evaluated the psychometric characteristics of the AQ (Bouchard, 2007; Bernstein, Gesn, 1997; Harris, 1995, 1997; Nakano, 2001; Rodríguez, Peña, Graña 2002; Vigil-Colet et al., 2005), it is possible to confirm that the Italian version of the AQ also shows sufficient validity. As the authors of the original version have confirmed, however, we can therefore demonstrate that, although with necessary differences, even with the southern Italian sample, the AQ is a psychometric instrument that provides sufficient empirical evidence and bases its precision on theoretical validity in the evaluation of different types of aggression (Buss, Perry, 1992). Moreover, as has come out of previous works (Bryant, Smith, 2001; Gerevich, Bácskai, Czobor, 2007; Morales-Vives, Codorniu-Rada, VigilColet, 2005; Nakano, 2001; Ramirez, Andreu, Fujihara, 2001; VigilColet et al., 2005), our analyses also show that the exclusion of some items from the overall scale produces an improved factor model. Regarding the effects of the age and gender variables on different types of aggression, we must underscore how the results obtained partially mirror those obtained by other authors, even in different contexts (Buss, Perry, 1992; Sommantico et al., in press; Fossati et al., 2003; Harris, 1995, 1997; Meesters et al., 1996; Nakano, 2001; Rodríguez, Peña, Graña 2002; von Collani, Werner, 2005). In general terms, even though mean scores were not very high, men tend to show greater levels of aggression, specifically physical and proportional to the verbal expression of rage. In line with the cited studies and the literature on this specific phase of the life (Greenspan, Pollock, 1991; Bergeret et al., 1985), the increased age of the subject contributes to a decrease in the different types of aggression. On the other hand, women show lower levels in the verbal expression of rage. In conclusion, in our subjects, hostility, generally ascribed to women, demonstrates higher average means in males. References • Ando, A., Soga, S., Yamasaki, K., Shimai, S., Shimada, H., Utsuki, N., Oashi, O., Sakai, A. (1999). Development of the Japanese version of the Buss-Perry AQ(BAQ). Shinrigaku Kenkyu, 70, 384-392. • Ang, R. (2007). Factor structure of the 12-item aggression questionnaire: Further evidence from Asian adolescent samples. Journal of Adolescence, 30 (4), 671-685. • Bazán, R. G. y Osorio, G. M. (2007). Manual de muestreo. En M. Chávez Becerra. Métodos Cuantitativos. México: Ediciones FES Iztacala UNAM. • Bergeret, J., Cahn, R., Diatkine, R., Jeammet, P., Kestenberg, E., Lebovici, S. (1985). Adolescence terminée, adolescence interminable, Paris: PUF. • Bernstein, I.H., Gesn, P.R. (1997). On the dimensionality of the Buss/Perry Aggression Questionnaire. Behaviour Research and Therapy, 35, 563-568. • Bouchard, J. (2007). Validation de la version française du AQauprès de deux échantillons: étudiants universitaires (étude1) et adultes non-recrutés en milieu universitaire (étude 2). Mémoire de Maitrise en Psychologie, Université du Québec à Chicoutimi. • Bryant, F.B., Smith, B.D. (2001). Refining the architecture of aggression: a measurement model for the Buss-Perry aggression questionnaire. Journal of Research on Personality, 35, 138-167. • Buss, A.H., Durkee, A. (1957). An inventory for assessing different kinds of hostility. Journal of Consulting and Clinical Psychology, 21. 343-349. • Buss, A.H., Perry, M. (1992). The Aggression Questionnaire. Journal of Personality and Social Psychology, 63, 452-459. • Fabrigar, L., Maccallum, R.C., Wegener, D., Strahan, E. (1999). Evaluating the use of exploratory factor analysis in psychological research. Psychological Methods, 4 (3), 272-299. • Fossati, A., Maffei, C., Acquarini, E., Di Ceglie, A. (2003). Multigroup confirmatory component and factor analyses of the Italian version of the Aggression Questionnaire. European Journal of Psychological Assessment, 19 (1), 54-65. • Gallardo-Pujol, D., Kramp, U., García-Forero, C., Pérez-Ramírez, M., AndrésPueyo, A. (2006). Assessing aggressiveness quickly and efficiently: the Spanish adaptation of Aggression Questionnaire-Refined version. European Psychiatry, 21 (7), 487-494. • Garcia-León, A., Reyes, G.A., Vila, J., Pérez, N., Robles, H., Ramos, M.M. (2002). The Aggression Questionnaire: a validation study in student samples. The Spanish Journal of Psychology, 5, 45-53. • Gerevich, J., Bácskai, E., Czobor, P. (2007). The generalizability of the BussPerry Aggression Questionnaire. International Journal of Methods in Psychiatric Research, 16 (3), 124-136. • Giannini, M., Panocchia, L. (2006). L’analisi fattoriale esplorativa in psicologia. Una guida pratica per la ricerca. Firenze: Giunti O.S. • Greenspan, S.I., Pollock, G. (Eds) (1991). The Course of Life. Vol. IV. Adolescence. Madison: International Universities Press. • Harris J.A. (1995). Confirmatory factor analysis of the Aggression Questionnaire. Behaviour Research and Therapy, 33 (8), 991-993. • Harris, J.A. (1997). A further evaluation of the Aggression Questionnaire: Issues of validity and reliability. Behaviour Research and Therapy, 35 (11), 10471053. • Jöreskog, K.G., Sörbom, D. (1993). Structural Equation Modeling with the SIMPLIS Command Language. Chicago: Scientific Software International, Inc. • Luccio, R., Paganucci, C. (2007). Analisi fattoriale esplorativa e analisi delle componenti principali: un tentativo di guida dei perplessi. Bollettino di Psicologia Applicata, 251, 3-12. • Maffei, C. (2008). Versione italiana dell’AQ(AQ). In, C. Maffei, Borderline. Struttura, categoria, dimensione. Milano: Cortina. • Meesters, C., Muris, P., Bosma, H., Chouten, E., Beuving, S. (1996). Psychometric evaluation of the Dutch version of the Aggression Questionnaire. Behavioral Research and Therapy, 34 (10), 839-843. • Morales-Vives, F., Codorniu-Raga, M.J., Vigil-Colet, A. (2005). Caracteriricas psicométricas de las versiones reducidas del Cuestionario de Agresividad de Buss y Perry. Psicothema, 17 (1), 96-100. • Morren, M., Meesters, C. (2002). Validation of the Dutch version of the AQin adolescent male 405 offenders. Aggressive Behavior, 28, 87-96. • Nakano, K. (2001). Psychometric evaluation of the Japanese adaptation of the Aggression Questionnaire. Behavioral Research and Therapy, 39, 853-858. • Panocchia L., Giannini, M. (2007). Valutazione dell’uso dell’analisi fattoriale esplorativa nella ricerca psicologica in Italia. Giornale Italiano di Psicologia, 34, 391-406. • Parrello, S., Aleni Sestito, L., Nasti, M.G., Sica, L.S. (2008). Redefinition of self-others system in the transition to emerging adulthood: a narrative approach. EARA, European Association for Research on Adolescence. Torino, Maggio 2008. • • • • • • • • • Ramirez, J.M., Andreu, J.M., Fujihara, T. (2001). Cultural sex differences in aggression: a comparison between Japanese and Spanish students using two different inventories. Agrressive Behaviour, 27, 313-322. Rodríguez, J.M., Peña, E., Graña, J.L. (2002). Adaptación psicométrica de la version española del Cuestionario de Agresión. Psicothema, 14 (2), 476-482. Santisteban, C., Alvarado, J.M., Recio, P (2007). Evaluation of a Spanish version of the Buss and Perry Aggression Questionnaire: Some personal and situational factors related to the aggression scores of young subjects. Personality and Individual Differences, 42 (8), 1453-1465. Sierra, J.C., Gutiérrez, Q.J.R. (2007). Validación de la versión española del Cuestionario de Agresión de Buss-Perry en Estudiantes Universitarios Salvadoreños. Psicología y Salud. 17, 1, 303-313. Sommantico, M., Osorio Guzmán, M., Parrello, S., De Rosa, B. (in press). Significado de las funciones familiares en adolescentes italianos. nuevas fronteras entre paterno y materno. Revista Colombiana de Psicologia, 17. SVIMEZ (2008). Rapporto 2008 sull'economia del Mezzogiorno. Bologna, Il Mulino. Vigil-Colet, A., Lorenzo-Seva, U., Codorniu-Raga, M.J., Morales, F. (2005). Factor structure of the AQamong different samples and languages. Aggressive Behavior, 31 (6), 601-608. von Collani, G., Werner, R. (2005). Self-related and motivational constructs as determinants of aggression. An analysis and validation of a German version of the Buss-Perry Aggression Questionnaire. Personality and Individual Differences, 38, (7), 1631-1643. Williams, T.Y., Boyd, J.C., Cascardi, M.A., Poythress, N. (1996). Factor Structure and Convergent Validity of the AQin an Offender Population. Psychological Assessment, 8 (4), 398-403.