Attachment to parents and friends as a context for development of self-concept in adolescence : The personality traits as mediators

The aim of the research is to study developmental relations between attachment to mother, father and friends and dimensions of adolescents’ self-concept. More specifically, we examine if this relation is direct or mediated by some personality construct. A sample of 878 students (13 and 16 years old) from Belgrade urban lower secondary and upper secondary schools participated in this research. Participants completed ECR for mother, father and friend; Selfperception profile for adolescents and NEOFFI. The results have indicated that dimensions of attachment are correlated with dimensions of self-concept, and that they can explain 13% of variance in dimensions of self-concept. When the personality traits are introduced as mediators, attachment dimensions explain only 4.5% of variance in self-concept. The quality of attachment relations with parents is not related to adolescents’ self-concept, while attachment to friends is correlated with social dimensions of self-concept. Hence, this study suggests that adolescents with particular parental attachment styles differ according to their self-concept profiles, but that peer attachment is important for adolescents’ social self. Beside that, this study reveals significant gender differences: relation between parental and peer attachment and adolescents’ self-concepts is not the same for boys and girls.

At the same time, cross-cultural research shows that a child develops attachment to more than one person, and every relationship is significant for the domain in which it is developed (for example, peers for socialisation, mother for health and physical development, etc.) (Howees, 1999;van Ijzendoorn & Sagi, 2008).Empirical data show that there is a relation between attachment to mother, father and friends and self-esteem, psychological health, social competencies, and adjustment during middle childhood and adolescence (Burke & Weir, 1978;Hoffman, Ushpiz, & Levy-Shift, 1988;Main, Kaplan, & Cassidy, 1985;Suess, Grossman, & Sroufe 1992).
This paper is about adolescents' self-concept development in the context of attachment relationships.Self-concept is defined as cognitive generalisation of oneself, as conscious representation of experience that a person has of oneself, which consists of evaluated perceptions, thoughts and feelings about oneself as an object and a partner in interaction (Damon & Hart, 1991;Harter, 2006).Selfconcept is a multidimensional construct.We can distinguish global evaluations of oneself as a person (usually referred to as global self-worth) from evaluations of one's own competencies and attributes in specific domains (academic, cognitive, social...).The number and content of these specific dimensions of self-concept change with age (Harter, 1988;2006).Until the end of adolescence, self-concept reaches its most complex and comprehensive form.
Self-concept is a social construct.It develops in the context of social interactions with "significant others" (Mead, 1934): parents, friends, teachers.Previous researches have shown that specific dimensions of self-concept depend on specific context and relations with different persons (Krstić, 2002).Hence the idea that the quality of attachment with parents and friends will have effects on adolescents' self-concept.

Previous research on the relation between attachment and self-concept
In the literature which refers to children's or adolescents' self-descriptions and self-evaluations, a number of terms are used: self-concept, self-perception, self-worth, self-esteem, etc.They differ according to theoretical perspective and measuring instruments (Krstić, 2002), which is why it is very difficult to compare results from different studies that use different concepts and different instruments, but describe similar phenomena.In the section below we will present some researches relevant to the topic of this paper.
Generally speaking, secure attachment is found to correlate with more positive self-worth, closer relationships with parents, more positive relations with friends and peers, and better acceptance among peers in adolescence (Allen, Moore, Kuperminc, & Bell, 1998;Sroufe, Egeland, Carlson, & Collins, 2005).
Several studies aimed to investigate relations between attachment to parents and friends with different aspects of self-perception in adolescence.Armsden and Grinberg (1987) found that attachments to parents and to friends are important predictors of self-esteem and life satisfaction in adolescence.Attachment to parents accounted for about 17% of variance of self-esteem and life satisfaction, while peer attachment appeared to be more related to selfesteem (accounting for 9% of variance) then to life-satisfaction (accounting for 6% of variance).Series of studies suggest that attachment to parents is more important then attachment to friends for psychological well-being in adolescence (Raja, Mcgee, & Stanton, 1992).Close and secure relation with parents allows adolescent to experiment and to research, and those behaviours are very important in adolescence (Armsden & Greenberg, 1987;Greenberg, Siegel, & Leitch 1983).
In one study, authors (Wilkinson & Walford, 2001) found that parental attachment but not peer attachment contributed significantly to predicting adolescents' well-being.In another study, using structural equation modelling, it was discovered that the relation between peer and parental attachment and psychological health is mediated by self-esteem (Wilkinson, 2004).In yet another study, it was established that mother attachment was associated with both self-liking and self-competence, peer attachment was significant just for self-liking, and father attachment was insignificant for both measures of selfesteem (Wilkinson & Parry, 2004).
Results also differ regarding developmental changes during adolescence.In one study it was found that the attachment to parents remains important for selfesteem from early to late adolescence (Walker & Green, 1986).However in another study, results indicate that adolescents kept stable quality of affection towards their mothers, but while girls utilised mothers' support more, boys' need for mothers' support and proximity decreased with age.Quality of affection towards father decreased for both adolescent boys and girls, as well as their utilisation of father for support and proximity (Paterson, Field, & Pryor, 1994).Over time, both boys and girls increasingly rely on support and proximity of friends.
Empirical findings regarding the importance of attachment to mother, father and friends for adolescent's self-esteem are inconsistent.Several studies suggest that paternal attachment is the most important (Gecas & Schwalbe, 1986;LeCroy, 1988), especially for boys (Cubis, Lewin, & Dawes, 1989).Findings from other studies, suggest that parental attachment has larger effect than peer attachment, regardless of the age of adolescent (Armsden & Greenberg, 1987;Burke & Weir, 1978;Gecas, 1972;Greenberg et al., 1983).On the other hand, some results suggest that the correlation between attachment to friends and adolescent's adjustment (.54) and self-concept (.47) is greater than correlations of attachment to parents (.38 and .18, respectively) (Cotterel, 1992).
Findings from O' Koon (1997) suggest that parental and peer attachments correlate with different dimensions of self-perception: mother attachment with family relationships (highest obtained correlation .68),then with emotional tone (.30) and psychopathology (.26); father attachment also with emotional tone (.35), family relationships (.53) and mastery of world (.26) and educational goals (.20), while peer attachment was correlated with almost all dimensions except family relationships and educational goals, including body image, social relationships, sexual attitudes, idealism, etc. (correlations range from .24 to .39).
In several cross-cultural studies it was shown that attachment to mother affects self-esteem, well-being (anxiety and loneliness), identity foreclosure, selfconfidence, vocational attitudes, and ethnic identity, while attachment to father is important for social functioning in adolescence (Kenny, Griffiths, & Grossman 2005;Leondari & Kiosseoglou, 2000;Matos, Barbosa, De Almeida, & Costa, 1999).Attachment to parents is also correlated with better adjustment in ethnic and racial minority groups among American high-school and college students (Hinderlie & Kenny, 2002;Kenny et al., 2005).Securely attached adolescents have greater self-esteem which leads to greater self-concept clarity (Wu, 2009).
In sum, from the theoretical point of view, a child develops attachment to more than one person (Lewis, 2005;Newcombe & Reese, 2004;Takahashi, 2005;Thompson, 2005;Weisner, 2005), and the quality of those attachment relations affect different areas of psychological functioning in later life (Howees, 1999;van Ijzendoorn & Sagi, 2008).A number of researches have shown that attachment relations are significant for various psychosocial outcomes in adolescence and later life (Burke & Weir, 1978;Hoffman et al., 1988;Main et al., 1985;Suess et al., 1992).Still, results regarding the importance of attachment to mother, father and friends for different aspects of self-perception and self-evaluation are inconsistent (Cotterel, 1992;Gecas & Schwalbe, 1986;Kenny et al., 2005;LeCroy, 1988;Leondari & Kiosseoglou, 2000;Matos et al., 1999;Paterson et al., 1994;Raja et al., 1992;Wilkinson & Parry, 2004;Wilkinson & Walford, 2001).Furthermore, we were unable to find researches relating attachment with dimensions of self-concept defined as evaluations of one's own competencies and attributes in specific domains (academic, cognitive, social...) (Harter, 2006).So the "dilemma" concerning the importance of parental and peer attachment for specific dimensions of adolescents' self-concept still remains.
On the other hand, researches have indicated a correlation between attachment and different personality traits (empathy, prosocial behaviour, aggressive behaviour anxiety, etc.) and some of these traits (self-esteem, empathy, prosocial behaviour) mediate in the relation between attachment and the quality of peer relations, adjustment or psychological health (Dekovic & Meeus, 1997;Laible, Carlo, & Roesch, 2004;Wilkinson, 2004).Therefore, the relation between attachment relations and different dimensions of psychological functioning could be direct or mediated by personality traits.

The role of personality traits
"The five-factor model" is an empirically well based framework for explaining large part of individual differences in non-cognitive domain of personality (John & Srivastava, 1999;Knežević, Džamonja-Ignjatović, & Đurić-Jočić, 2004).Consequently, attachment theorists and researchers had to show that attachment variables are not just "clones" of already existing personality traits.It is empirically proven that individual differences in attachment cannot be reduced to differences in personality traits (Noftle & Shaver, 2006;Picardi, Caroppo, Toni, Bitetti, & Di Maria, 2005;Shaver & Brennan, 1992).Shaver and Brennan (1992) were the first to study correlations between dimensions of attachment and Big five personality traits.They found that attachment anxiety is in positive correlation with neuroticism, but not to the extent that would suggest redundancy of two concepts (r = .33).After their study, a number of researchers studied correlations of attachment and Big five personality traits.According to the summary of past findings from 14 studies on attachment dimensions and the Big five, by Noftle & Shaver (2006), results suggest that secure attachment has a moderate negative correlation with Neuroticism, moderate positive correlation with Extraversion and Agreeableness, a small positive with Conscientiousness, and no correlation with Openness to experience (Becker, Billings, Eveleth, & Gilbert 1997;Carver, 1997;Shafer, 2001;Shaver & Brennan, 1992;Wilkinson & Walford, 2001).Results also suggest that attachment anxiety is in moderate to strong correlation with Neuroticism, and in no correlation with Openness (Baeckstroem & Holmes, 2001;Becker et al., 1997;Shaver & Brennan, 1992;Shaver et al., 1996).Evidence about relations with the rest of the Big five traits are inconsistent.Attachment avoidance has a small to moderate negative correlation with Extraversion and Agreeableness, and has no correlation with Openness (Baeckstroem & Holmes, 2001;Becker et al., 1997;Gallo, Smith, & Ruiz, 2003;Shafer, 2001;Shaver & Brennan, 1992).In some, but not all, studies a positive correlation with Neuroticism and negative correlation with Conscientiousness were established (Noftle & Shaver, 2006).
Additionally, researches have also shown a relation between self-perception and personality traits.However, even though both concepts have been subjects of numerous studies, there are just a few focusing on the relation between personality traits as core, basic dimensions and dimensions of self-perceptions as peripheral, surface personality characteristics.Studies of these relations suggest that selfworth is in negative correlation with Neuroticism, and in positive correlation with Extraversion, and slightly less with Openness to experience, Agreeableness and Conscientiousness.These traits account for 34% of variance of self-worth (Judge, Erez, & Bono, 1998;Marsh, Trautwein, Lüdtke, Köller, & Baumert, 2006;Robins, Tracy, Trzesniewski, Potter, & Gosling, 2001;Watson, Suls, & Haig, 2002).Results of rare studies of correlations between personality traits and self-perceptions as multidimensional constructs support the idea of existence of multiple dimensions of self-perception and personality traits as moderately correlated but still different dimensions of personality (Marsh et al., 2006).Results suggest that dimensions of self-perception measure something different from personality traits.
Therefore, previous studies indicate that attachment and self-concept are correlated with personality traits, but also that they describe individual differences that cannot be reduced to differences in personality traits.Baron and Kenny (1986) formulated the criteria that must be met in order to consider a variable as a mediator.According to that criteria applied to our example, there are three requirements: attachment and self-concept are correlated, attachment is correlated with personality traits, and self-concept is correlated with personality traits.Research findings mentioned previously imply that all three requirements are met in our case, so the important research question is whether the relation between attachment relationships and dimensions of self-perception is mediated by personality traits.

The present study
On the whole, a large number of researches investigated the relation between attachment and self-concept, self-esteem, self-perception, selfevaluation and similar concepts at various ages.The results are not clear and consistent, leaving some questions unanswered.How are attachment relations with different people (mother, father, friends) related to specific dimensions of self-perception profile?Which dimensions of self-concept are related to specific attachment relations (H1)?Do these complex relations change over adolescence (H2)?Are those relations direct or mediated by personality traits (H3)?
The main aim of the research is to study relations between attachment to mother, father and friends and dimensions of adolescents' self-concept.More specifically, we examine if personality traits mediate this relation between attachment and self-concept in adolescence.Empirical evidence suggests that during adolescence the impact of parents and peers on adolescent's self-concept change (Krstić, 2002).For that reason, the relation between attachment relations and self-concept in this study was examined at the age of early and middle adolescence. Hypotheses.
1) Attachment styles to mother, father and friend are correlated with different dimensions of self-concept: peer attachment will be correlated with social dimensions of self-concept; mother attachment with physical appearance, behaviour and global self-worth; and father attachment with academic and athletic competencies) (Harris, 1998;Thompson, 2005;van Ijzendoorn & Sagi, 2008); 2) At older adolescent age attachment to friends will have more effects on selfconcept than parental attachment (Krstić, 2002;Bretherton, 1985;Harter, 2006); 3) Personality traits will mediate a part of relation between attachment and selfperception (Marsh et al., 2006;Shaver & Brennan, 1992;Picardi et al., 2005).

Method Sample
This research was conducted on a sample of 878 13 and 16 year-old students (53.4% females) attending lower and upper secondary education in five urban schools in Belgrade.

Variables
Attachment to mother, father and friend; measured as 1) attachment-related dimensions of anxiety and avoidance; and 2) attachment style to each figure: secure attachment style, or insecure dismissing, preoccupied and fearful attachment style.
Adolescent Self-perception profile measured through eight dimensions: Physical Appearance, Scholastic and Athletic Competence, Social Acceptance, Behavioural Conduct, Close Friendship, Romantic Appeal and Global Self-Worth.
Personality traits, according to "the five-factor model": Neuroticism, Extraversion, Openness to Experience, Agreeableness and Conscientiousness.
Instruments I) Self-Perception Profile for Adolescence (Harter, 1988).This instrument taps nine self-concept domains, of which eight were used in this research1 : Physical Appearance, Scholastic and Athletic Competence, Social Acceptance, Behavioural Conduct, Close Friendship, Romantic Appeal and Global Self-Worth.There are a total of 40 items, five for each sub-scale.Each item has "a structured alternative format" (Harter, 1982) which consists of two opposite descriptions.The adolescent is asked to decide which description fits him/her better and then he/she decides whether the description is "Really True for Me" or "Sort of True for Me".Each item is scored on a four-point scale, where score 1 indicates the lowest perceived competence and the score of 4 indicates the highest level of competence (Harter, 1988).II) A modified version of Experience in Close Relationships (Brennan, Clark, & Shaver, 1998), is a self-report measure of attachment designed to assess attachment-related avoidance and anxiety, and on the basis of their scores, adolescents can be classified into a specific attachment category (secure, dismissing, preoccupied and fearful.)Modified versions used in this research consist of parallel scales designed to measure attachment relationship with mother, father and friend (Kamenov & Jelić, 2003).Each scale consists of 18 Likert-type items (9 for each dimension), and responses range on seven-point scale, where 1 means "Strongly Disagree" and 7 means "Strongly Agree".This shortened and modified version of questionnaire has kept the two-factorial structure and has a satisfactory reliability (Cronbach's alpha ranged from .73 to .87).

Procedure
After a parental approval and adolescents' informed consent were obtained, participants were administered questionnaire booklets, which consist of six questionnaires in balanced order (ECR for mother, father and friend, SPP and NEOFFI).The questionnaires were completed in group sessions, during regular school classes, with average testing duration of 60 min for younger and 45 min for older students.

Results
At the beginning we analysed the frequency of different attachment styles in relation to mother, father and friend in our sample of adolescents.As Table 1 shows, our adolescents are predominantly securely attached to all three figures; almost all of them to a friend.Distribution of attachment styles is similar to those obtained in previous studies on adolescents (Hanak, 2009;van Ijzendoorn & Bakermans-Kranenburg, 1996), with a somewhat more secure attachment in our sample.

Relation between attachment and self-concept -canonical correlation
To determine the associations between the attachment and dimensions of self-concept, canonical correlation between attachment-related anxiety and avoidance in relation to mother, father and friend, and eight dimensions of self-concept were used.Previous studies have shown that there are gender differences, so canonical correlations were calculated for four subsamples: younger and older, boys and girls.Canonical roots of attachment dimensions explain 9% to 14% of variance of adolescents' self-concept scores.Results show that dimensions of attachment are correlated with dimensions of self-concept, but percentages of explained variance are small.That means that based on the data on attachment we cannot conclude much about adolescents' self-concept.Moreover, we can see that with the age, the percentage of explained variance decreases (and more for boys), which means that correlations between attachment and self-concepts are smaller at older than on younger adolescent age.Analysis of the structure of canonical roots shows that in all four subsamples dimensions defining the first canonical root are peer attachmentrelated Avoidance and the Friendship and Social acceptance dimensions in the self-concept profile (Appendix).Besides, there are no clear patterns of relations between parental attachments and specific dimensions of self-concept.Comparing attachment to mother and father, we can conclude that attachment to father is less important for self-concept of older adolescents, especially for boys.These results suggest that the quality of attachments to friends affects adolescents' social self-concept while the quality of attachments to parents, and especially to father, does not make much difference in the self-perception profiles of our adolescents.

Relation between attachment and self-concept -multiple regression analysis
In order to perform a thorough study of the relations between given variables, a multiple regression analysis (Enter method) was conducted.When the attachment-related anxiety and avoidance were introduced as predictors, they accounted for only 13% of total variance in dimensions of self-concept (Table 4) (total variance is calculated as the average of all regression coefficients).Percentages of variance which dimensions of attachment explain in every dimension of self-concept range from 0.6% to 3.5% (Table 5).Thus, these results confirm once again that there is a relation, albeit weak, between attachment styles and self-concept profiles in adolescence.

Age differences in relation between attachment and self-concept
Taking all the above results in consideration, we can also analyse age differences in relation between attachment and self-concept.Firstly, as previously stated, the results from canonical correlation show that the percentage of explained variance decreases with age, which means that attachment and selfconcept correlations are smaller at older than at younger adolescent age (Table 3).Furthermore, at the older age canonical factors are defined by fewer dimensions of attachment; dimensions of attachment to father do not participate in the structure of correlations between attachment and self-concept of older boys and in the structure of the first factor for older girls (Appendix).Furthermore, we can analyse age differences in regression analysis for attachment dimensions as predictors of self-concept (Table 5).Still, based on the inspection of the number and values of standardised regression coefficients (β) for attachment dimensions as predictors of self-concept for younger and older adolescents, we cannot draw conclusions regarding age differences.There is no clear pattern which would suggest that with age the importance of peer attachment for adolescents' selfconcept increases, as we anticipated.

Relation between attachment and self-concept -personality traits as mediators
The next aim of this study was to analyse whether this relation between attachment and self-concept is mediated by personality traits.The question is which dimensions of attachment are correlated with dimensions of self-concept independently of personality trait?In order to analyse that, a hierarchical regression analysis with three sets of variables was performed.Firstly, personality traits were introduced, followed by dimensions of attachment as predictors of self-concept.Thus it can be concluded what added percentage of variance of self-concept can be explained by the dimensions of attachment.
When personality traits are introduced in the prediction, they account for 23% of variance of self-concept dimensions.Subsequently, attachment dimensions explain additional 4.3% of variance in adolescents' self-concept over and above personality traits (Table 4).The largest part of the variance of self-concept dimensions mediated by personality traits comes from attachmentrelated Anxiety in relations with mother and friend.The correlation between selfconcept dimensions and attachment-related Anxiety in relations with mother and friend can be explained by personality traits, especially neuroticism.Interestingly, the dimension of Anxiety with friends remains significant, independently of personality traits, for younger boys, for dimensions Friendship, Self-worth and Athletic competence, and for Social acceptance and Self-worth for older boys.However, that is not the case when it comes to the Anxiety with father.This dimension has significant contribution in five cases and is not mediated by personally traits.Attachment-related Anxiety with father is correlated with dimensions Behavioural Conduct and Friendship for older girls and with Social acceptance, Romantic relations and Physical appearance for younger boys.Still, girls who are less anxious with their fathers have more positive concepts of themselves, while boys who are less anxious in relation with their fathers have more negative concepts of themselves on those three dimensions.The contribution of attachment-related Avoidance in relation to mother, father and friends to total correlations between attachment and self-concept proved to be mostly independent of personality traits.Dimensions of Avoidance in relations with parents have little contribution to total examined correlations.The most significant contribution that one attachment dimension has to total correlation with self-concept is a contribution of attachment-related Avoidance in relation with friends.This dimension contributes to the correlation with Social Acceptance and Friendship for all four adolescents groups, independently of personality traits.Interestingly, the relation of peer Attachment-related Avoidance with Romantic relations for all and Self-worth for boys is totally mediated by personality traits.

Discussion
The aim of the research was to study relations between attachments to mother, father and friends and dimensions of adolescents' self-concept profiles, and the role personality traits have in that relation.Results show that there is a correlation, albeit weak, between attachment-related Anxiety and Avoidance and dimensions of self-concept in adolescence.Based on that it can be concluded that attachment influences adolescents' self-concepts, but based on attachment style we cannot say much about the profiles of adolescents' self-concept.The correlation of those two sets of dimensions is too small, hence specific self-concept profiles cannot be described based on the attachment styles of adolescents.
That implies that individual differences between self-concept profiles are too large even in a group of securely attached adolescents.This results could be a consequence of a large number of secure attachment styles in our sample in all three relations; consequently it is difficult to determine specific profiles.On the other hand, although attachment styles indicate specific internal working models of self and others, and results suggest that attachment styles have impact on self-concept, there are numerous domains of psycho-social functioning where different competencies could be displayed, or sources of pleasure and selfrealisation found.This is particularly true in adolescence, when self-concept is developing and changing from one moment to another, from one context to another (Elliot & Feldman, 1990;Kimmel & Weiner, 1995;Noller & Callan, 1991).In previous study we found that adolescents' self-concept depends on representations of how significant others (parents and friends) see them (Krstić, 2008), but its seems it is independent of the quality of attachment with those significant others, per se.
The percentage of the explained variance of self-concept decreases with age, so the influence of attachment is smaller at the age of 16-17.The adolescents' self-concept is less dependent on the quality of relations with others in middle adolescence, and especially for boys.Besides, we could not find other age differences in relation between adolescents' self-concept and their attachment relations with different people.Previous studies suggest there are age differences in the importance of representations of how peers see them (Krstić, 2008), or utilisations, reliance on support and proximity of friends (Paterson et al., 1994), which is not in accordance with findings from this study.It seems that although importance of friends and of peer attachment increase over age, adolescents' self-concept does not depend on the quality of relations with friends but more on the cognitive aspects of social feedback they receive from peers.
The relation between attachment styles and self-concept in adolescence is mainly based on the relation between attachment-related Avoidance with friends and social dimensions of self-concept.This finding once again stresses the importance of friendships and peer relations as a secure base in adolescence (Ainsworth, 1989;Allan, 2008;Cassidy, 2001;Waters & Cummings, 2000).Friends in adolescence could serve as a secure base and as a base for developing a positive model of self, but also for experimentation and research which are immensely important for identity development in adolescence.Adolescents who are securely attached to friend, and especially if they are also securely attached to one of their parents, tend to have higher self-worth and more positive perception of themselves on all or most dimensions of self-concept.
Although researchers have found specific effects of attachment to mother or to father on different aspects of psycho-social functioning in adolescence (Howees, 1999;van Ijzendoorn & Sagi, 2008), these differences were not confirmed in this study.The relation between parental attachment and particular dimensions of self-concept was not discovered in this research.This discrepancy may be due to the different instruments used for measuring either self-concept or attachment, or due to the age and gender differences neglected in previous studies.Attachment to mother is shown to be more important for adolescents' self-concept, especially in early adolescence, while attachment to father has an effect only on self-concept of girls.Therefore, these results do not contribute to the dilemma regarding the importance of attachment to different persons and the importance of parental and peer attachment for specific dimensions of selfconcept (Lempers & Clark-Lempers, 1992;O'Koon, 1997;Paterson et al., 1994;Suess et al., 1992).
One of the findings from this study deserves special attention: the relation between attachment to parents and friends and adolescents' self-concepts is not the same for boys and girls and these gender differences should be studied more profoundly.Studies on gender differences in attachment revealed very mixed results (Kenny & Rice, 1995;Laible et al., 2004;Song et al., 2009;Wilkinson, 2004), particularly regarding the relation between parental and peer attachment and adolescents' self-concept (Cubis et al., 1989;Song et al., 2009).These findings were not elaborated here in detail, but they indicate that we could not speak about a relationship between attachment and adolescents' self-concept independently of gender.Gender differences in attachment and self-concept represent an important topic which deserves special attention in further research.
Regarding the role of personality traits, results of this study suggest that the relations between attachment and self-concept are largely mediated by personality traits.Attachment-related Anxiety in relation to mother and friend can be explained almost completely by neuroticism, but Anxiety in relation to father is dependent on the real quality of interaction with a father.Attachmentrelated avoidance is independent of personality traits and it is a reflection of dyadic relation.Our findings indicate that there is a small correlation between attachment-related dimensions and self-concept in adolescence that is independent of personality traits.
These results go in line with the ideas about the significance of social relation for the self-concept of adolescence.This study is in concordance with the assumptions about importance of "the looking-glass self" (Cooley, 1902;Mead, 1934;Opačić, 1995) for adolescents' self-concept, especially for girls.During adolescence the relations with and feedback from peers and friends become particularly important (Krstić, 2002;2008;Harter, 2006).The attachment to parents and friends has an impact on adolescents' self-concept, but that influence is largely mediated by personality traits.Neuroticism as a personality trait explains the effect of attachment-related Anxiety with mother and friend on adolescents' self-concept.On the other hand, Anxiety with father is independent of personality traits.It seems that the nature and meaning of attachment-related Anxiety varies in different relations.While Anxiety in relation with mother and friends, as an internal working model of oneself in those relations, can be explained with more basic tendency towards negative emotions, Anxiety in relation with father is less based on basic personality traits and more a reflection of specific experience of oneself in everyday interactions with the father.It might be assumed that tendency towards negative emotions will lead to more frequent activation of attachment activity, and that adolescents with high scores on neuroticism will need a safe haven or a secure base more often and probably will worry more about availability of mother or a friend (Crawford, Shaver, & Goldsmith, 2007).Mother is a primary attachment figure, but in the course of adolescence friends become more and more important; adolescents need their closeness, support, security, hence those who are prone to negative emotions will worry more about their availability and will react more negatively when experiencing an inadequate response.On the other hand, father has less importance as an attachment figure.In attachment to father, we have got the least securely attached adolescents, and the most dismissing styles, in comparison to attachments to mother and friends.It can be assumed that those who are prone to negative emotions will not be anxious in relation with father but will avoid closeness with him in order to avoid to be hurt or left alone.

Conclusion
The focus of the research was to study relations between attachment to mother, father and friends and dimensions of adolescents' self-concept, and the role of Big five personality traits as mediators.Findings suggest that there is a correlation, albeit weak, between dimensions of attachment and dimensions of self-concept, and than correlation is independent of personality traits and decreases during adolescence.The quality of attachments to parents cannot predict the profile of adolescents' self-concept, while attachment to friends is correlated with social dimensions of self-concept.There are significant gender differences in the relations between attachment and self-concept.
Limitations of this study involve methodological issues.Although we utilised the most commonly used questionnaires, operationalisation and measuring of both attachment and self-concept represents just one of numerous possible ways.As discussed earlier, this makes comparison of results difficult.Therefore, the question of relations between attachment and self-concept remains unanswered.Results suggest that in investigating these relations we should be sensitive to gender and age differences.In this study we got a very large number of adolescents securely attached to friends, which raises the question of whether this questionnaire, method and design of investigating peer attachment are sensitive enough to reveal small and fine differences in the quality of peer attachment relations.
short version of NEO-P-R(Costa & McCrae, 1985; adaptation Đurić-Jočić,  Džamonja-Ignjatović, & Knežević, 2004)  was used for personality trait assessment.It provides a quick and reliable measure of the five domains of personality: Neuroticism, Extraversion, Openness to Experience, Agreeableness and Conscientiousness.It is a 60-item inventory (12 per domain), answered on a five-point Likert scale.

Table 1
Frequency of attachment styles in relation to mother, father and friend

Table 2
Statistics for the canonical functions

Table 3
Canonical roots and percentage of explained variance

Table 4
Standardised regression coefficients (β) for attachment (Att) and personality traits (NEO) as predictors of self-concept * only significant regression coefficients are shown

Table 5
Standardised regression coefficients (β)for attachment dimensions as predictors of selfconcept before and after personality traits (NEO) are introduced