Can the Factor Structure of Defense Style Questionnaire (DSQ-40) Contribute to Our Understanding of Parental Acceptance/Rejection, Bullying, Victimization and Perceived Well-Being in Greek Early Adolescents?

Defense Style Questionnaire (DSQ) is a self-report instrument designed to measure defense mechanisms. Although commonly used, the DSQ-40 has not been validated in early adolescent populations. The present study sought to determine the factor validity of the DSQ-40 in a sample of Greek primary school students (N = 265). Further, it aimed to investigate the relationship between defense mechanisms and perceived parental acceptance/rejection, the participation in bullying (either as bully or victim) as well as self-reported well being. Participants completed the Greek version of DSQ-40, adapted for use by this particular age group as well as measures in order to examine its convergent and discriminant validity. The findings support a four-factor solution as the most adequate for our data. Further, it was found that defense mechanisms are related to perceived parental acceptance and rejection. Finally, the results showed that the DSQ-40 can effectively discriminate participants with high/low bullying/victimization and perceived well-being. Our results indicate that the DSQ-40 is appropriate for use in late childhood. Implications for clinical practice and future studies that would confirm the appropriateness of the scale’s use in younger populations are also discussed.

The origins of the quantitative measurement of defense styles dates back to Vaillant's (1971Vaillant's ( , 1976 seminal studies, who used a hierarchy of defenses and reported that mature defenses were positively correlated and immature defenses negatively correlated with an objective measure of life success. Bond, Gardner, Christian, and Sigal (1983) taking into account Vaillant's hierarchical model, suggested an immaturity-maturity continuum and developed a self-report measure, the 67-item Defense Style Questionnaire (DSQ-67); the principal aim was the assessment of possible conscious aspects of defense mechanisms, with the underlying intention "to elicit manifestations of a subject's characteristic style of dealing with conflict, either conscious or unconscious, based on the assumption that persons can accurately comment on their behaviour from a distance" (Bond et al., 1983, p. 334). The DSQ-67 discriminated between psychiatric patients and normal subjects, with patients reporting more immature defense mechanisms. The self-report methodology and the construct validity of the instrument was supported by a study in which scores of the DSQ were correlated with clinical assessments . The 67-item DSQ ended up to a revised 88-item version developed by Bond and Vaillant (1986). The latest version was modified by Andrews, Pollock, and Stewart (1989) according to DSM-III-R; they suggested a 72-item version which included 20 defenses loading on three factors, labeled Mature, Neurotic and Immature. This version was further modified to a short 40-item questionnaire corresponding to 20 defense mechanisms which were organized into three second order factors: Mature, Neurotic, Immature.
The 40-item DSQ has proved to be a valid and reliable instrument with moderate to high Cronbach α values, split-half reliability, test-retest reliability and item-scale correlations and has been used in a range of studies (Carvalho et al., 2013). The factor structure of the instrument in adults suggests a four-factor solution although the findings are inconsistent (Andrews, Pollock, & Stewart, 1989). Cramer (2015) suggests that: (a) different defense mechanisms predominate at different points in the development and (b) each defense mechanism has its own developmental history, and that each defense has its own time of emerging predominance, which is followed by the decline of that defense at subsequent developmental periods. She identified significant differences in denial, projection and identification between age 11 and age 18. Although this work has indicated that defense mechanisms protect children from psychological distress, most of the studies have focused on late adolescents-emerging adults. Indeed, a great number of recent studies focus either on University students (e.g. Saint-Martin, Valls, Rousseau, Callahan, & Chabrol, 2013) or late adolescents (Lewis & White, 2009).
Studies in adolescent populations employ the DSQ-40 because of its practical use and theoretical underpinnings (Pour, Nezhad, Sabooni, & Mir Ahmadi, 2011;Ruuttu et al., 2006). Research has indicated that defense styles measured by the DSQ-40 are associated with psychological adjustment and psychopathology (Pour et al., 2011) with positive associations between immature defenses and maladjustment to dominate in adolescent samples (Lewis & White, 2009).
Studies exploring the factor structure of DSQ present altered versions revealing, three (Muris et al., 2003), four (Ruuttu et al., 2006) or five (Saint-Martin et al., 2013) factors. However, none of these versions has been used widely. Bond suggests (personal communication) the use of the DSQ-40 in late childhood and adolescence simply pointing out the need for some items to be appropriately reworded. The psychometric properties of the DSQ-40 have not been widely explored in younger populations, apart from a fine-grained study carried out by Ruuttu and colleagues (2006). In a sample of 211 adolescent psychiatric outpatients aged 13-19years and 199 Youell (2006) fits bullying into the psychological account of splitting and projection.
Further, it has been shown that bullies often use projection in an attempt to escape from their own 'painful' experienced emotions (White, 2004), whereas victimized children seem to employ more defense mechanisms such as isolation, autistic fantasy (Cilliers, 2012). Lastly, studies have linked immature defense mechanisms with the development of various psychological difficulties such as anxiety and depression (Blaya et al., 2007) as well as impaired self-esteem (Zeigler-Hill, Chadha, & Osterman, 2008).
The shortage of studies using the DSQ-40 in early adolescents indicate the need for further exploration of the appropriateness of the use of DSQ-40 for this age group (Cramer, 2007). Furthermore, a range of Greek studies have explored the psychometric properties of the DSQ-88 and its associations with health/treatment dimensions in adults with particular medical illnesses (Hyphantis, 2010). However, no study has explored the psychometric properties of the DSQ-40 in younger populations (i.e. early adolescents).
The aim of the present study is to explore (a) the factorial structure of DSQ-40 (b) to identify the internal consistency of the factors (sub-scales) emerged from the exploration of the factor structure of the questionnaire, as well as its convergent validity and (c) to test whether the defensive styles emerged from the factor analysis could differentiate individuals according to their gender and age as well as to the extent they get involved in bullying and victimization and the level of their self-reported well-being.

Method Participants
Two hundred and sixty-five children and early adolescents took part in this study, of which 133 (50.2%) were boys and 132 (49.8%) were girls. The sample includes 105 (39.5%) fifth grade students, and161 (61.5%) sixth grade students (M age = 11.19, SD = .66). The students come from different schools of Athens, Crete and Ioannina, Greece.

Measures
The Defense Style Questionnaire (DSQ; Andrews et al., 1993) is a self-report inventory that measures specific defense mechanisms in terms of DSM-IV concepts. DSQ comprises of 40 items in a 9-point Likert format that derive scores for 20 defense mechanisms, two items for each. These mechanisms are organized in four subfactors (Immature, Mature, Image-distorting and Neurotic), often referred to as defense styles. Scores for defense mechanisms and defense styles are formed by averaging the ratings for relevant items. All DSQ items were translated in Greek, and blindly back-translated by native English speakers.
We measured psychological adjustment with the Psychological Well-being scale (Diener & Biswas-Diener, 2008

Procedure
The administration of the questionnaires took place in participants' classrooms, during ordinary class sessions and lasted approximately twenty minutes. Issues of anonymity and confidentiality were particularly highlighted, along with the voluntary participation in the study. Permission to conduct the research was granted by the Greek Ministry of Education. There was no payment or other incentive to participate in the study. After the completion of the questionnaires, children were debriefed and thanked for their participation. The data were collected during the 2013 spring semester. Eventually, they were encoded, transferred, and analyzed with SPSS 21, at the University of Crete.

Data Analyses
Exploratory Factor Analyses (EFA) was utilized instead of Principal Components Analysis (PCA). Previous studies used PCA to reduce correlated observed variables to a smaller set of independent composite variables (Park, Dailey, & Lemus, 2002), however we chose EFA for the following reasons: a. We wanted to test a previous theoretical model of latent factors underlying the variables observed b. We presumed correlated latent factors. Therefore, a geomin (oblique) instead of a quartimax (orthogonal) rotation criterion was chosen c. EFA is conceptually more realistic in identifying common factors.
Regarding the examination of other psychometric characteristics, we used correlations coefficient to test the DSQ internal consistency. We also used Pearson's correlations to examine both the construct validity (i.e. correlation coefficients between each of the four DSQ-40 defense styles) and the convergent validity of the measure with other constructs (i.e. parental acceptance-rejection, psychological well-being and bullying/ victimization). We also examined the measure's discriminant validity by testing gender and age differences (independent samples t-test) in the use of defenses. Moreover, we investigated the ability of DSQ to distinguish different groups of individuals, using logistic regression in order to predict participants' self-reported well-being as well as their participation in bullying acts, either as bullies or victims. The allocation of participants in the high/low groups was made using median split analysis (high group = +1SD, low group = -1SD, respectively).

Exploratory Factor Analysis (EFA)
The DSQ-40 items were submitted to EFA, to determine factor structure using the mean scores for each defense. Geomin oblique rotation was employed. We used the Mean-and Variance-adjusted Weighted Least Square (WLSMV) estimator in Mplus 6 because both the univariate and multivariate normality of the data were not met (Muthén & Muthén, 2007). The goal was to see how the 20 individual defenses loaded onto factors, commonly referred to as defense styles. To determine the number of factors, we examined the goodness of fit for each proposed model based on criteria of goodness of fit, interpretability of the solution, and strength of parameter estimators.
First, we examined the goodness of fit (root mean square of approximation-RMSEA; root mean squared residual-SRMR; comparative fit index-CFI), for each proposed model (one-to four-factor models). The findings supported a four-factor solution as most adequate for our data (Table 1). Second, we examined the eigenvalues. The Kaiser (1960) criterion, which recommends that the number of factors be equivalent to the number of eigenvalues greater than 1, suggested five factors. Eigenvalues were 4.20 for the first and 1.12 for the fifth factor.
Next, we examined the scree plot as proposed by Cattell (1966)  Lastly, we considered the interpretability of the results by examining the factor pattern loadings (Table 2). Undoing .50 Idealization .54 Pseudo-altruism .65 Denial .75 Dissociation .60 Devaluation .42 Isolation .34 Splitting .66 Note. In parentheses we present the cross-loadings and in italics are the retained loadings on Factors 1 and 2.
With regard to item selection for the four factors, factor loadings >.20 were considered meaningful. Some defenses appeared to have double loadings on one or two factors (i.e., loadings of greater than .20) and were not interpretable on its factor and were excluded. On the basis of those criteria, two defenses were excluded; sublimation (mature factor), and realization (mature factor) had loadings on different factors than the original and it was difficult to interpret them. However, we retained two defenses: Suppression (mature), and autisticfantasy (immature). The first had a double loading on two factors (with loadings of .18 and .22 on Factors 1 and 3 respectively), but the first loading was sensible because of the defense's nature. The second defense style had also a cross-loading more than .20, but an acceptable correlation with Factor 2 (.40) and it was more interpretable to use it on this factor (immature).

Internal Consistency
DSQ-40 psychometric characteristics (Cronbach's alpha values and defenses correlations are presented in   Table 4 presents the correlation coefficients between each of the four DSQ-40 defense styles. The intercorrelations range from .20 (between Image-distorting and Neurotic style) to .62 (between Immature and Image-distorting style).

Convergent Validity
The results in Table 4 show that the Mature factor was positively associated with Psychological Well-being whereas the Neurotic factor was positively associated with paternal acceptance and Psychological Well-being.
Finally, the Immature factor was negatively associated with both paternal and maternal acceptance and with Psychological well being.

Discriminant Validity
We also compared use of defenses between genders; girls reported less use of mature and image-distorting styles than did boys. On the individual defenses, girls reported more use of reaction formation, but less anticipation, denial, and dissociation.
Further, we compared the use of defenses between participants of the 5 th and the 6 th grade. We found that 5 th graders scored higher in the Neurotic factor and less use in the Immature factor. With regards to specific defenses, it was found that 5 th graders scored higher in anticipation, pseudoaltruism and idealization, whereas 6 th graders scored higher in passive aggression, displacement, dissociation, splitting and rationalization. As discriminant validity indicates also the ability of a scale to distinguish different groups of individuals, we performed three separate logistic regression analyses (see Table 5) to investigate whether individual scores in the four Defense Styles (Immature, Mature, Image-distorting and Neurotic) could predict: a) the levels of selfreported psychological well-being, b) the participation in bullying behaviours and c) the experience of victimization. In the first analysis, it was found that Immature, Mature, Image-distorting and Neurotic factors were entered as predictor variables and Psychological Well-being as predicting variable. A test of the full model with the four predictors was statistically reliable, [χ 2 (4) = 26.20, p < .001], indicating that, as a set, the predictors reliably distinguished between high-level (HL) and low-level (LL) individuals in perceived psychological well-being. The variance in the HL status accounted for was large, with Nagelkerke R 2 = .39. Prediction was good, with 62.1% of HL individuals and 78.3% of LL individuals correctly being predicted, for an overall success rate of 71.1%, while predictive values positive (sensitivity) and negative (sphericity) were 73% and 69%, respectively. Table 5 shows the regression coefficient, Wald statistic, and odds ratio for each of the three predictors. According to the Wald criterion, the Immature, the Mature and the Neurotic scores predicted HL status. This finding implies that the odds ratios of these predictors (.95, 1.06 and 1.04, respectively) show a great change in the likelihood of the individual demonstrating high levels of well-being on the basis of one unit change in these variables (i.e.,

Immature, Mature and Neurotic factors).
In the second analysis, we entered again the Immature, Mature, Image-distorting and Neurotic factors as

Discussion
The study suggests that the DSQ-40 is a valid and reliable instrument for use in late childhood/early adolescence. This instrument which was initially developed for adults seems, with minor phrasing changes, to be able to provide researchers and clinicians with valid information about children emotional maturityimmaturity styles, shedding light on their psychological adjustment. This suggestion is supported by appropriate scores found in different aspects of validity, tested in a non-clinical sample of Greek children/early adolescents: construct validity, convergent validity and discriminant validity.
To our knowledge, this is the first study exploring the psychometric properties of the DSQ-40 in an early-aged sample. The present study failed to support the three factor solution suggested in the original study (Andrews et al., 1993) with adults. The four factor structure comes out well, in spite of cross-loadings. Cross-loadings also appear in studies with both adults (Hyphantis, 2010) and adolescents (Hayashi, Miyake, & Minakawa, 2004).
The first factor consisted of three mature defenses: suppression, humor, anticipation often employed to reduce anxiety and distress. Two defenses, sublimation and rationalization, failed to contribute to this factor. Not having reached abstract thinking at this particular age possibly restricts the development of complex defense mechanisms.
The second factor consists of immature defenses: projection, passive aggression, autistic fantasy, somatisation, displacement, and acting out. This factor involves maladaptive defense mechanisms, reported also by Ruuttu et al. (2006), which very often appear in object relationships and dominate in clinical work, especially in psychoanalytic psychotherapy (e.g. Youell, 2006). Moreover, in consistency with previous studies (e.g. Andrews et al., 1993;Hyphantis, 2010;Steiner et al., 2001), splitting gives a remarkable loading on this factor, although it is conceptually consistent with the third factor (image distorting). This factor differentiates from the third one, image-distorting, although they both depict immature personality patterns, in terms of actionorientation and image-orientation, respectively. The third factor, also in line with Ruuttu et al. (2006), depicts an image-distorting defense style. It comprises denial, dissociation, devaluation, isolation and splitting. The image distorting defenses, associated with borderline and narcissistic defenses (Conte & Plutchik, 1995) indicate coping that aims to distort self and object images to conform totally with a particular meaning or emotional state. The fourth factor comprises neurotic defenses indicating an intermediate defense style that appears more private comparing to immature (action-oriented) and image-distorting (image oriented) defenses. Possibly, this is the reason of no cross-loadings (except the reaction formation) in this factor. The defense mechanisms loading on this factor are: reaction formation, undoing, idealization, pseudo-altruism. These are in line with those reported in the original study (Andrews et al., 1993), and also in a range of studies in adolescents from different nationalities (Hayashi et al., 2004). The four factors described above and the correlations amongst them support Vaillant's (1971) continuum from maturity to immaturity.
Concerning the internal consistency coefficients of the scale, the study supports previous findings reporting relatively low reliability coefficients for the mature defense style in both adults and adolescents but sufficient to high for the neurotic, immature and the image-distorting defense styles (Karagiannopoulou, Athanasopoulous, & Hyphantis, 2015). Perhaps, the lower alpha coefficients are affected by the fewer items comprising these two styles (Zeigler-Hill et al., 2008). Nevertheless, all Cronbach alphas were greater than 0.5, justifying the factorial structure of the four styles.
Overall, although our sample consists of children in late latency/early adolescence, our findings about the factor structure are consistent with the adult frontline studies (Andrews et al., 1993;Bond et al., 1983) and also with later studies in both adolescents and adults (e.g. Sammallahti, Holi, Komulainen, & Aalberg, 1996). Our suggestion for two immature components corresponding to (immature/action-oriented and image-distortion/ image-oriented) possibly depicts the complexity of the states of mind associated with the latter part of latency (ten or eleven years) and those associated with the early part of puberty (twelve or thirteen years) (Waddell, 2002). To our knowledge, the only recent study focusing on the factor structure of the DSQ-40 (Saint-Martin et al., 2013) suggests a five factor structure depicting discrepancies with defense theory. The consistency of our findings with the only, comprehensive study on the exploration of the factor structure of the DSQ-40 in adolescents (Ruuttu et al., 2006), lends support to the appropriateness of this instrument for early adolescents.
Our study supports previous findings indicating links between defense mechanisms and adjustment problems, anxiety (e.g. Blaya et al., 2007) parental acceptance/rejection (Dubois-Comtois & Moss, 2008) and forms of children's interaction with their parents (von der Lippe, 2000). Greater maturity of defense styles was associated with lower scores in bullying behaviours and with higher scores of psychological well-being and parental acceptance. These findings indicate meaningful strong associations mainly for the immature factor that was negatively related with parental and maternal acceptance and also with psychological well-being. It is suggested that mature defenses protect against bullying behaviours supported by a positive perception of psychological well-being and parental acceptance (Fonagy, 2008). Strength to this suggestion is lent by the negative relationship between immature defenses with well-being and parental acceptance.
The study supports previous findings concerning gender differences in defense use. Boys appear to use more mature and image-distorting defense mechanisms than girls. However, when focusing on particular defenses this was the case only for anticipation (previous studies indicate a higher score in suppression) while for two image-distorting defense mechanisms, namely denial and dissociation. Possibly, boys fail to acknowledge some aspects of external reality (denial, dissociation) while girls seem to reside to less pathological defense Defense Style Questionnaire (DSQ-40) 278 mechanisms. This is possibly indicating either lower levels of stress for females of this age-range or a feminine defense in terms of changing the inner affect (Cramer, 2007).
Interestingly, age differences were significant for defense styles, although the age range in the current sample was very small. Latency (5 th graders) implies a "quite" period without new upheavals but early adolescencepuberty (6th graders) seems to involve regression in the direction of infancy (Waddell, 2002); internal conflicts and anxieties challenge internal structures and are met by a range of immature defense strategies. In particular, the 5 th graders' scored higher in two neurotic defenses: pseudoaltruism and idealization to sustain the self and on one mature defense: anticipation, indicating tolerance to future discomfort. Moreover, the 6th grade students, in a possible attempt to bypass the complex and painful task of working through depressive anxieties (Waddell, 2002), scored higher in three immature and one image-distorting defense, namely, passive aggression, displacement, splitting and dissociation, respectively. Perhaps, intense anxieties and impulses stir in an intense need to displace and reside with passive aggression and splitting possibly strongly reinforced by the underlying Oedipal fears (Waddell, 2002); 6 th graders avoid the recognition of some aspects of external reality to protect themselves from excessive anxieties, using dissociation.
Moreover, the study supports clinical findings suggesting emotional immaturity to be considered as a factor underlying both patterns of behavior; children's involvement in high levels of both bullying behaviours and victimization was associated with the use of more immature defense mechanisms. Also, mature defenses were associated with lower risk of victimization and neurotic defenses with lower risk of bullying possibly indicating that victims might be considered more emotionally mature than (Fonagy, 2008). The study also supports associations between emotional maturity and psychological adjustment (Benson & Elder, 2011); children with higher levels of psychological well-being used more mature and less immature defense mechanisms.

Limitations -Future Directions
In the present study, valid and widely used instruments provide evidence for the appropriateness of the adult version of the DSQ-40 for use in late childhood-early adolescents. Overall, this study supports the appropriateness of the use of the DSQ-40 in early adolescence and provides important evidence to support the link of defense mechanisms with parental acceptance, psychological adjustment issues (both in the form of ivolvement in bullying behaviours and impaired well-being) as well as gender and age differences in the use of specific mechanisms.
However, the study has some limitations.

23
If my boss bugged me, I might make a mistake in my work or work more slowly so as to get back at him.

24
There is someone I know who can do anything and who is absolutely just and fair.