Study of validation of the Portuguese version of the inventory «Experiences in Close Relationships » 1

The ‗Experiences in Close Relationships Inventory‘ permits to evaluate attachment in close relationships during adulthood based on two dimensions able to be present in this kind of relationships: the avoidance of proximity and the anxiety related with to abandonment. It is a self-report 7- points likert scale composed by 36 items. The Portuguese version was administered to a sample of 551 university students (60% female), the majority with ages between 19 and 24 years old (88%) in a dating relationship (86%). The principal components analysis with oblimin rotation was performed. The total scale has good internal consistency (α=.86), as also has the 2 subscales: anxiety (α=.86) and avoidance (α=.88). The two dimensions evaluated are significantly correlated with socio-demographics, relational characteristics (jealousy, relationship distress, and compromise), wishes (enmeshment versus differentiation) and fears (abandonment versus control) related to attitudes in significant relationships, which testify the construct validity of the instrument. The results obtained are coherent with the original version and other ECR‘s adaptations. Practitioners and researchers in the context of clinical psychology and related areas have now at their disposal the Portuguese version of the ECR inventory, which has shown its very high usefulness in the study of close relationships, and specifically attachment in adulthood.

The theoretical conceptualization of this article comprises three principal themes.It begins with: i) The conceptualization of romantic love as an attachment processwhere we present a review of the core premises of attachment theory applied to romantic love and examine the main differences between infant and adult processes; ii) Individual differenceshere we explain individual differences and the concept of internal working models associated with attachment dimensions.According with this, specify the types of models of attachment dimensions and the main classes of methodologies associated with them, and finally; iii) The processes and outcomes in intimate relationships resulting from attachment dimensions-describing the mechanisms of choice, maintenance and dissolution of relationships associated with attachment, adding also mediating variables.

i) Romantic love as an attachment process
The investigation in the scope of intimate relationships had begun with the pioneer work of Hazan and Shaver (1987) that conceptualized the romantic love as an attachment process (Shaver, Hazan, & Bradshaw, 1988).Supporting this theory there are four core premises that both help resemble and distinguish conjugal relationships from the infant-caregiver attachment processes (for a further reading, Feeney, 2008;Fraley & Shaver, 2000).In terms of similar aspects: a) attachment in adulthood and in infancy shared the same biological system, the activation and cessation conditions and exhibited the same dynamics; b) the individual differences in adulthood resulted from the expectancies and beliefs developed during early interactions and share the same types; and finally, c) internal working models uphold the continuity of individual differences on attachment patterns, and have a role in intimate relationships' outcomes.
The main divergent aspects mentioned by Fraley e Shaver (2000) are related to the role reciprocity of the attachment system and caregiver dynamics, its sexual nature, and the integration of three social behavioural systems 5 -attachment, experiences of care, and sexual coupling, reinforcing the perspective of Hazan and Shaver (1987).According with these authors romantic love can be comprehended in terms of the mutual functioning of those behavioural systems, though each one with a different function and playing a role more or less meaningful in the developmental trajectory of intimate relationships.Hazan and Zeifman (1999) proposed a model of formation of attachment in adulthood similar to that suggested by Bowlby, strengthening that the processes can vary attending to internal models of each element of the dyad, a model that gained support for other researchers (Treboux, Crowell, & Waters, 2004) In adulthood, the behaviour of looking for help or support (e.g., expressing distress, looking for comfort or assistance in times of need), refers to a behavioural manifestation of the attachment system, and the existence and accessibility of an attachment figure (that serves as a safe haven in times of need) is associated to emotional wellness (Collins & Feeney, 2000;Feeney, 2008;Feeney & Collins, 2004), making possible the establishment of a conjugal bond.Thus, marital relationships and other significant romantic relationships accomplish the criteria of attachment behaviours, because adults look for the proximity of the partner, experience emotional distress if the partner is not accessible, build up confidence and security from this relationship and looking for the partner in times of distress and threat (Ainsworth, 1989;Feeney & Collins, 2004;Mikulincer & Shaver, 2007;Shaver et al., 1988). 5Obviously the organization of these systems in a certain person reflects the experiences in attachment relationships and has underlying an ethological function of procreation and protection of the species.
ii) Individual differences Ainsworth, Blehar, Waters and Wall (1978) considering differences in the way the individuals cope with the anxiety associated with non-optimal interactions and regulate the security feelings, that reflect the distinct -internal working models‖ (IWM), of the self (as worth or not of love and support) and the other (as being or not responsive) built in previous experiences with significant ones (Bowlby, 1973;Bretherton, 1990;Cassidy, 2000;Main, Kaplan, & Cassidy, 1985).These models determine the dynamics of intimate relationships in adulthood (Ainsworth et al., 1978;Bowlby, 1973;Feeney, 2008;Shaver et al., 1988), since the choice of the partner, maintenance and rupture of the relationship.
In the same way, Shaver and col. (1988) defend that -internal working models‖ continue to guide and determine behaviour with significant figures across the life span.
While people build new relationships, they are guided by previous expectations about the way others behave and feel toward them, and use those models to infer the behaviours or intentions of the partners.Attachment theory explains the continuity in interaction patterns with significant others across the life span, and suggest that the previous experiences of care has a significant influence, at least in a part, in the patterns of relationship in adulthood with the partner as also its possible revision (Cassidy, 2000;Crowell et al., 2008;Fraley & Shaver, 2000;Treboux, Crowell, & Waters, 2004).
When Hazan and Shaver (1987) began the study of attachment in adulthood, adopted the three categories defended by Ainsworth and col. (1978), as a frame of reference to organize the individual differences in the way the adults, think, feel, and behave in romantic relationships.These three qualitatively independent and distinct patterns of attachment (secure, anxious-ambivalent, and avoidant) were described as brief mutually exclusive questions about romantic relationships.Its reduced design associated with a scarce fidelity of the proposed instrument was subject to some criticisms, notwithstanding their precious contribute to the development of other measures (e.g., Collins & Read, 1990;Levy & Davis, 1988;Simpson, 1990).
Bartholomew for example, notices that avoidant pattern, as was described by Hazan and Shaver, includes two theoretically different ways of avoidance that they denominate as -fearful‖ and -dismissing‖ (Bartholomew, 1990;Bartholomew & Horowitz, 1991).According to the defence strategies shown by avoidant individuals, acknowledge that some of them adopt an avoidant orientation in attachment relationships to prevent being hurt or rejected by the partners (fearful), while others adopt the same orientation as a way of maintaining a defensive sense of self-reliance and independence (dismissing).Resulting from this distinction they build a tetracategorical model, in which Bartholomew maintains the classification suggested by Hazan and Shaver (1987), but splits the avoidant category in avoidant-fearful and avoidant-dismissing (this last, based on a similar category in the Adult Attachment Interview; see Main, Kaplan, & Cassidy, 1985, for a review).Moreover, adding the disposition of these categories in a bi-dimensional space defined by the positive or negative value of the representational models of self (anxiety) or other (avoidance) (see also Bartholomew & Shaver, 1998, for overviews).Consequently, from this combination results the four styles of attachment: secure, preoccupied, dismissing and fearful.Currently, this model is receiving some important remarks specifically about the independent structure of the model of self and others, and the qualitatively distinct aspects of these models underlying the known four styles (Feeney, Noller, & Hahran, 1994b;Ross, McKim, & DiTommaso, 2006).
Besides the positivity or negativity associated with models of self and models of other and its combination into a four styles of attachment, it has been valued the underlying dimensions in terms of emotional and behavioural regulation6 -anxiety and avoidance (Brennan et al., 1998;Fraley & Waller, 1998).
Attending to this state of the art, Brennan and col. (1998) exhaustively revised literature, and gathered more than 320 items collected from different questionnaires and models, with the aim of exploring the similarities and the differences between different referential frames and methodologies, and developed an inventory composed by the most relevant items.These authors concluded that the individual differences in terms of attachment can be organized in function of a bi-dimensional space, with two major functions: anxiety or vigilance about the rejection and abandonment and avoidance that corresponds to the discomfort with the closeness or dependency and the reluctance to be intimate with others.
Despite Brennan and col. (1998) did not consider a prototypical approach to measurement of adult attachment, provided strong evidences regarding anxiety dimension and Bartholomew's model of self and, avoidance dimension and others model.Considering the individual differences in attachment, individuals with a secure attachment style presented both, low anxiety and low avoidance scores: feel comfortable with intimacy, trust in others to support and believe they have self-worth.The subjects with preoccupied style (anxious-ambivalent) are identified by high anxiety and low avoidance scores: revealed strong desire of intimacy and dependency and at same time that a great preoccupation with the rejection.Individuals with avoidant-dismissing style have low anxiety and high avoidance scores: have a tendency to depreciate the importance of the intimate relationships and value independency and self-reliance.The persons with an avoidant-fearful style presented an elevation in both attachment dimensions scores: although desiring an intimate relationship and the approval of others they tend to avoid intimacy due to fear of rejection (Bartholomew & Shaver, 1998;Collins & Feeney, 2000).
In general, several types of methodologies had been employed in the study of attachment in adulthood, in particular concerning romantic relationships which consider each author's theoretical orientation and the objectives of their studies (see Bartholomew & Shaver, 1998;Feeney, 2008; for a further explanation of this issue), for example: attending to the report of the partner (Kobak & Hazan, 1991), diary registers (Feeney, Noller, & Callan, 1994), observational studies (Simpson et al., 1996), interviews (Bartholomew & Horowitz, 1991, Crowell, 1990;George, Kaplan, & Main, 1985) and self-reports (Bartholomew, 1990;Brennan et al., 1998;Collins & Read, 1990;Feeney et al., 1994b, Fraley, Brennan, & Waller, 2000;Hazan & Shaver, 1987;Simpson, 1990;Simpson, Rholes & Nelligan, 1992).However, many of those methodologies present some controversial issues specifically in terms of standardization of the studied constructs, that had justified the development of a more comprehensible, a largely applied instrument, which seems to have a clear structure such as ECR.

iii) Processes and outcomes in intimate relationships
Some evidences suggest that individual's choose a partner that confirms the beliefs they have about the attachment relationships (Brennan & Shaver, 1995;Collins & Read, 1990;Frazier, Byer, Fischer, Wright, & DeBord, 1996;Holmes & Johnson, 2009;Kirkpatrick & Davis, 1994;Klohnen & Luo, 2003;Tolamcz, Goldzweig, & Guttman, 2004).For example, those with a secure attachment style tend to be pairing with secure individuals.Among the insecure ones, the more common type of pairing are avoidant with anxious individuals for whose, the differences in terms of expectancies about the relationship are mutually confirmed.
Acknowledged about the current state of the art, the diversity of evaluation methodologies of attachment processes, its limitations and advantages, and the importance of a standardized measure adapted to Portuguese context, it is our major goal to study the psychometric properties of the Portuguese version of ECR.
In the next sections we will explore ECR´s internal and external validity -and at the end, the relationship of this instrument with individual (i.e., attitudes about significant relationships) and relational variables (i.e., conflict, communication problems, dominance, relationship distress, jealous, and commitment) theoretically relevant to demonstrate its external validity.
Furthermore, it's our purpose to make a contribution to the systematic comprehension of internal and external processes related with romantic attachment in early adulthood.

Instrument
Description.The ‗Experiences in Close Relationships Inventory' (ECR, Brennan et al., 1998) permits assessing attachment in close relationships during adulthood based on two dimensions able to be present in this kind of relationships: the avoidance of proximity and the anxiety related to abandonment.It is a self-report 7point likert scale composed by 36 items.Its concise structure facilitates the answer in a short period of time, nearly 10 minutes.
The items that compose the ECR are presented in table 1. Attending to the original version, the items that correspond to each dimensions evaluated are presented in a spin order; the items with a pair number correspond to anxiety and the items with an odd number to avoidance scale.

Translation, retroversion, and semantic analysis of the items. This version is based on Portuguese spoken and written in Portugal. The instrument was translated to
Portuguese by a bilingual psychologist, and was performed a pilot study.Twelve undergraduates are inquired about the comprehension and semantic difficulty of the items, as a way of approaching this translation to the target dimensions evaluated by the instrument.Characteristics of intimate relationship.Considering the duration of more recent relationship, more than 4/5 subjects (85.7%) having or having had a relationship during more than one month in the year previous to the answer to questionnaire.More than half of the participants (60%) presented a relationship with a duration length equal or superior to a month, 26% reported that -presently have no relationship, but had had in the past‖ and 14% admitted -never having a relationship that lasted more than one month‖.At the moment of the answer to the questionnaire, a great percentage of subjects (69%) continued in the relationship.Considering the length of the more recent relationship with a partner, 41.5% confirmed a two years or more duration [for 19.0% it ranged between 1 and 2 years, for 24% the length was less than one year, with the residual 15% less than 3 months length].For those participants that finished the relationship (31%), regarding the time occurred since the relationship break up 12.2% considered it finished 1 year ago or more [6.7% reported that the relationship had finished 1 or 2 months ago, 3.9% at 3 to 5 months before, 8.2% at 6-12 months ago).
Considering the type of the more recent relationship, the great majority (84.7%) had an engaged dating relationship, 8.1% a dating relationship, and 7.2% were married or cohabiting.Almost of the respondents are sexually active with the partner (72.2%), with heterosexual orientation predominantly (98.9%).

Data collection
Participants were informed of the aim of the study (to better understand the processes underlying intimate relationships), confidentiality was guarantied and the researcher's contact was given to them.Then, they filled out a social-demographic questionnaire and the ECR.They also filled out two other self-report measures: Personal and Relationships Profile (PRP, Straus, Hamby, Boney-McCoy, Sugarman, 1999, Paiva & Figueiredo, 2006) and Attitudes about Significant Relationships (ASR, Henry, 1995;Paiva & Figueiredo, 2003).This two measures, intend to demonstrate the construct validity of ECR (Campbell & Fiske, 1959), evaluating along with it some constructs that theoretically are associated with attachment dimensions, that is, the quality and attitudes with intimate relationships.To accomplish this aim the following analysis are performed: 1. Correlates of anxiety and avoidance and characteristics of the relationship (existence, type, duration, terminus, presence of a sexual component); 2.
Correlates of anxiety, avoidance and variables of the relationship (conflict, communication problems, dominance, jealous, commitment and relationship distress); According with theoretical frame supporting the ECR, it was performed a confirmatory factor analysis, to test the hypothesis of the underlying constructs of anxiety and avoidance (see Brennan et al., 1998, for a review), and confirm their orthogonal bi-dimensionality.
We realized a PCA to establish the number of latent constructs that better explain the relationship of the variables in analysis and its underlying factor structure.
To obtained factorial matrix we applied an oblique rotation (oblimin) 7 .
For the internal consistency it was considered the coefficient Cronbach's alpha.
The descriptive statistics for each item, and the mean differences attending to the gender as tested through t test for independent samples.The computation of criteria validity as performed between ECR dimensions and the different categories of socialdemographics by one-way ANOVA, applying the criteria post-hoc Bonferroni.Pearson correlation was performed for test the association of ECR dimensions with other external criteria (PRP), and the linear regression (stepwise method) for the ASR scales was also performed.

Internal validity
Confirmatory factor analysis.A confirmatory factor analysis using a Maximum Likelihood Estimation Method in AMOS (version 17.0) was performed to test the model.An adequate model fit is found for the tested model (χ2(593)=2605.614,p<.001;RMSEA=.078,90% CI [.075-.081];cut-off value <.80).RMSEA index is considered an adequate index to test goodness of fit of the model (Raykov, 1998;Riggdon, 1996), especially with large sample sizes.We also tested a second model, considering the 36 items in each attachment dimension as considered by the authors, adding a constraint of original model, permitting the correlation of two factors, resulting in a slight improvement.So we consider the orthogonality of attachment dimensions, as the authors of original ECR defended.
Determined the adjustment of the theorical model to the data, we had performed an exploratory factor analysis to test the internal validity of the instrument, the PCA. 7Despite the theoretical considerations presented by Brennan and col. (1998) in the original version of ECR, concerning the independence of two ECR dimensions, we used an oblique rotation (allowing the correlation of factors), following the same procedure as these authors in the original version (opcit) and by others that had adapted the same instrument in different cultures (e.g., Alonso-Arbiol et al., 2007;Picardi et al., 2000).Other revisions of ECR suggest an oblique structure for anxiety and avoidance dimensions (Wei, Russel, Mallinckrodt, & Vogel, 2007).Also reinforcing this rotation procedure, a recent meta-analytic study, which considered 244 studies, stated the correlation between anxiety and avoidance ranged from -.36 to .68, and noticing the inconsistencies observed in orthogonal dimensionality of attachment for heterogeneous cultures and contexts (see Finnegan & Cameron, 2009, for a review).
Principal components analysis.The 36 ECR´s items were submitted to PCA.
The value of Kaiser-Meyer-Oklin is .888,and the value of Bartlett´s sphericity test=((630)= 5912.115,p=.000) obtained a significant statistical value strengthening the factorability of the correlation matrix.
The presence of 8 components which exceeded the eigenvalue 1 (the first two with values 6.9 and 5.6 respectively, and the rest with values ranging between 1.7 and 1.0), explaining 56.6% of the total variance.An inspection of the screeplot reveals an accentuated cut after the second component, that added to the obtained analysis considering the scree test of Catell, reinforces the retention of the 2 components 8 , to which was applied the oblimin rotation as the same way as the original version, since recent research has shown that the attachment dimensions may be significantly correlated (e.g., Finnegan & Cameron, 2009;Wei et al., 2007).
The rotated solution (as we can see in table 2) shows a concise structure with strong saturations and high communalities values (h 2 ), congruent with the theoretical rational underlying the original scale.It explains 34.8% of the total variance, with the first component contributing with 19.1% and the second with 15.7% of the total variance.On the other hand, as was theoretically expected, the two components do not correlate strongly among themselves; the value of the correlation between each other is only .06,despite the oblique rotation, suggesting that the two dimensions are essentially orthogonal (Brennan et al., 1998).When the correlation value of factors is reduced, the differences between orthogonal and oblique rotation procedures are also insignificant, so the generalizability of the results is assured (Rennie, 1997).
The Table 2 also shows that for each item, the values of the communalities and the saturation, also for each component the explained variance and the coefficient Cronbach´s alpha. 8Parallel analysis criteria for the number of main components indicate the possibility of retaining 4 factors based on empirical and randomized eigenvalues (first randomized value 1.6 < empirical value 1.7; second randomized value 1.5<1.6)computed through RanEigen program (Enzmann, 1997).However, we opted for the retention of 2 factors, based on theoretical and methodological (statistical) reasons, which we will refer later in this article.Forcing the PCA to three factors, the third factor will be composed exclusively with items 21 and 29, independent of other overall ECR constructs (low communalities, see Table 2), adding by the fact that in other ECR validation studies they are described as problematic (see Alonso-Airbirol et al., 2007;Conradi, Gerlsma, van Duijn, & de Jonge, 2006;Picardi et al., 2000;for overviews).
Despite structural similarities with the original version, we point out some variations in the composite of the items that constitute each of the dimensions, as we will analyse in the next section.An important note should also be made concerning item 21 and 29, in addition to their reduced saturation values, its inversion seems to reflect more the comfort than the discomfort with the dependency.

Fidelity: Internal Consistency
Concerning fidelity, the total scale and the sub-scales present high internal consistency values.The total scale composed by 34 items has a value of =.86 and the sub-scales anxiety (18 items) .86 and avoidance (16 items) .88 (see Table 2).
Correlation inter-dimensions.The correlation between the anxiety and avoidance scales for the total sample, although statistically significant is reduced (r(521)=.179,p<.001).Conversely, the correlation between each of the dimensions and 9 When a more detailed analysis was done on the factorial structure enforcing it to a 3 component, items 21 and 29 grouped separately from the others and exclusively on a third component, explaining only 4.7% of the variance, and presenting a value of internal consistency of .45,and saturation values respectively .554 e -.607 considering the reduced variance explained by this possible third factor, we decided to maintain the bi-dimensional original structure and remove items 21 and 29 of the analysis.Besides, alpha increases with this removal and these items don't evaluate the dimensions that they are designated theoretically.
Normative values and gender differences.Table 3 presents the normative values for the items, the total scale, and the sub-scales for the studied sample according to gender.Half of the ECR´s items showed significant gender differences.For anxiety 8/18 items had significant differences: items 2, 4, 14, 18, 22, 32 higher for women than for men; and items 6, 26 higher for men than women.For avoidance 11/16 items suggested gender differences: items 13, 5, 7, 3 higher for women and items 1, 27, 33, 35, 31 higher for men.
Considering the total sub-scales, there are significant differences only for anxiety; women reported it more than men.For avoidance, there are no significant gender differences; however men predominantly have a higher mean than women.In general the values found are clearly higher in the items of anxiety versus avoidance independently of the gender.
External validity: to demonstrate the construct validity of ECR the evaluated differences were correlated with other variables which should be theoretically associated (Campbell & Fiske, 1959).The characteristics were considered along with, the quality, and the attitudes with the intimate relationship.For the total scale, there are no significant gender differences.

External validity
To demonstrate the ECR´s construct validity the evaluated dimensions by the instrument were correlated with other variables with which are theoretically associated (Campbell & Fiske, 1959).We considered the characteristics, the quality and the attitudes in intimate relationships.

Correlates of anxiety and avoidance and characteristics of the intimate
relationship: existence, type, duration, terminus, presence of a sexual component.
The variance analysis (one-way ANOVA) considering the status of the relationship (have currently, had in the past, never had an intimate relationship) recognize significant statistical differences [F(2, 510)=81.39,p=.000], by the Bonferroni post-hoc analysis we can see the differences are specifically among the categories -never had‖ (M=3.57,SD=.74) and: -have currently‖ (M=2.45,SD=.74; dif. M=1.2, SD=.12, p=.000), -had in the past‖ (M=3.17,SD=.79; dif.M=.42, SD=.13, p=.004); those whom never had a relationship showed higher avoidance values than those who have currently or had in the past.

Correlates of anxiety, avoidance and variables of the relationship (PRP):
conflict, communication problems, dominance, jealous, commitment and relationship distress.
The association of the dimensions evaluated by ECR within external criteria was equally determined.For this purpose, it was considered the concurrent reports in the following dimensions of the intimate relationships: conflict (α=.76), communication problems (α=.50),dominance (α=.63), relationship distress (α=.84) (for a more detailed description of these scales of PRP in Portuguese see Paiva & Figueiredo, 2006), wishes (α=.86), fears (α=.58) and power tactics (α=.80) (for a full description of the Portuguese version of ASR, see Paiva & Figueiredo, 2003).
As we can see in Table 4, both attachment dimensions of ECR are significantly positively correlated with the relational scales of PRP: conflict, communication problems, dominance and relationship distress.An increase in anxiety or avoidance is associated with higher conflict, more communication problems, more dominance over the partner, and more relationship distress.Even though the values of the correlation raise the statistical significant criteria for both dimension, there is a clear strong association between avoidance and the indicators of dysfunction of the relationship.
Moreover, anxiety dimension is significantly and positively correlated with jealous scale and avoidance is negatively correlated with commitment.These results seems to also confirm the literature reports (Brennan et al., 1998;Collins & Read, 1990;Hazan & Shaver, 1987, Lafontaine, & Lussier, 2003;Picardi et al., 2000;Simpson, 1990) and seems to certify the validity of construct for the two dimensions of attachment evaluated by this version of ECR.

Predictors of anxiety and avoidance
We performed a linear regression (stepwise method) for the different scales of ASR (wishes-enmeshment, protection and care versus differentiation and autonomy, fears-rejection and abandonment versus over-enmeshment and control, and power tactics-benign affiliative autonomy, protective /affiliative control, and hostile control) and each dimension of ECR, as a way to test the evaluated constructs.Table 5 showed the main 3 predictors of anxiety related to abandonment, explaining 28% of the total variance: fear of rejection and abandonment, wish of enmeshment, protection and care, and as a power tactic over the other-hostile control.Concerning the avoidance dimension, five predictors were found that explained 24% of the total variance: fear of control and over-enmeshment, fear of rejection and abandonment, and in inverted manner the desire of enmeshment, protection and care, in terms of power tactics the hostile control and the benign autonomy in inversion mode.
These results allowed strengthening the construct validity for the sub-scales of avoidance and anxiety of ECR.

Norms and reference group criteria-quality of intimate relationship
The values presented in Table 6, constituted the norms for the studied sample comparing the two groups, with an high versus a low quality of intimate relationship with the partner.As reference values, we consider the values with a standard deviation above the mean and the percentil note (20 and 80); notwithstanding the detailed analysis of each item and its variability according with the gender.
As we can see in Table 6, there are significant differences when the groups with high and low quality of intimate relationship are compared for each ECR´s items and attachment dimensions (anxiety and avoidance) with important specificities for each gender.Individuals with high quality of intimate relationship presented lower values of anxiety and avoidance and total scale scores when compared with low quality of intimate relationships´ group.This result happens both for men, women and total sample, as it is shown in Table 6.

Discussion
The comprehension of attachment in adulthood originally derived from Bowlby and Ainsworfth theories had received broad attention and advances in the last 30 years.
Whilst universally recognized, attachment theory permits to understand human development in adulthood, and the main processes and dynamics underlying romantic relationships.The pioneer work of Hazan and Shaver (1987), research on adult attachment has indeed flourished.
However, even though emerging from a common reference frame, the agreement concerning attachment types or dimensions had been described in different ways considering each author's theoretical inclination (Bartholomew & Shaver, 1998;Crowell, Fraley & Shaver, 2008;Feeney, 2008, for overviews).The studied version of ECR, obtained through the principal components analysis with oblique rotation (oblimin), revealing the presence of a concise factorial structure, composed by two components congruent with the original version for almost all the items.It presents high consistency values either for the total (86) as for dimensions, the anxiety related to the abandonment (86) and avoidance of the proximity (.88).The items 21 and 29 presented argumentative performance and lower saturations in each of the two studied dimensions, even that similar to other adaptations of the instrument (e.g., Alonso-Arbiol et al., 2007;Conradi et al., 2006;Picardi et al., 2000).This justified the possibility of its removal, a main difference with the original version.For this decision accounted methodological analysis (statistical), a PCA with 3 factors showed them aggregated together in a same factor and presented low commonalities values (<.30).Apart from which, the reader can also utilize them, with the remark they seem to be indicative of the (dis)comfort with the dependency of the partner, as Collins and Read (1990) considered with the dimension "depend".More recently, some evidence are strengthening a three structure for attachment dimensions suggesting a better fit that the usual bi-dimensional (see Bäckström & Holmes, 2007, for a review).With awareness of this heterogeneity among studies in conceptualization of attachment dimensions, we recommend the inclusion of these items (21 and 29) in future studies, in particular transcultural ones [the value of Cronbach´s alpha of the scale with 18 items remains with a high value (.85)].Furthermore, items 12, 16, 20 and 26, although the high saturation in the theoretical corresponding dimension (anxiety) they presented high saturations in avoidance dimension.The explanation for this fact can be found in the argument of Bartholomew (1990), regarding the distinction of the pattern dismissing and fearful, advancing for the last one a strong desire of proximity and simultaneously the avoidance for fear of rejection (which seem to justify the sharing of high saturation values for these items in both dimensions of attachment).Nowadays, some authors are questioning the orthogonal dimensionality of Bartholomew tetra-dimensional prototypes, and suggesting that models of the self and models of the other enclose different components for each of the four prototypes (see, Ross, McKim, & DiTommaso, 2006, for overviews).The dilemma of independence versus complementarity of anxiety and avoidance dimensions of attachment in adulthood has been in the scope of current literature (e.g., Finnegan & Cameron, 2009).The two subscales have a marginal correlation value, indicating that attachment dimensions are essentially orthogonal, also confirmed by the confirmatory factor analysis (Conradi et al., 2006).
In terms of the normative values for each sub-scale, the anxiety dimension reveals higher values than avoidance dimension (e.g., Brennan et al., 1998;Picardi et al., 2000).Considering gender, women showed higher mean values than men for anxiety, similar to what Bartholomew and Horowitz (1991) founded for preoccupied style and other authors had found considering gender differences in attachment (Picardi et al., 2000;Paiva, Figueiredo, & Henry, 2008;Ross et al., 2006).
Regarding external validity, the relationship observed between the dimensions of ECR mainly the avoidance, and the characteristics of the intimate relationship, pointed to an evidence of concurrent validity.The participants that presented high avoidance, reported more that they never had an intimate relationship, have a relationship with no commitment, of a reduced length of duration (less than 1 month), without a sexual component, or finished the relationship confirming the individual differences in intimate relationships as described in literature (Brennan et al., 1998;Feeney et al., 1994a;Feeney, 2008;Simpson, Rholes & Phillips, 1996).
In addition, confirming the external validity, attachment dimensions evaluated by ECR are associated to the quality/dysfunction of intimate relationships.As a way, the positive correlation for both the insecure attachment dimensions and the indicators of relational dysfunction (i.e., conflict, communication problems, relationship distress) and the positive correlation of the sub-scale of anxiety with the jealous scale, and the negative correlation with the sub-scale of avoidance with the commitment scale are also similar to the reported by the original version authors.
The evaluation of the main wishes, fears and power tactics underlying the intimate relationships permits adding empirical evidence to the construct validity.
Consequently, how higher are the anxiety, the higher the fear of abandonment, the wish of enmeshment protection and care and the hostile control as a power tactic as predictors of anxiety related to abandonment.The higher the fear of control, the lower desire of enmeshment, protection and care and higher fear of control and the power tactics characterized by the absence of benign autonomy and the presence of hostile control configure the avoidance of the proximity.The obtained results reinforced that the sub-scale of anxiety evaluate in first place the fear of abandonment while the avoidance sub-scale the fear of control, also adding other constructs that can help explain the anxiety and the avoidance in the relationship (see also, Ross et al., 2006).
For example, high control and dominance in interpersonal dynamics associated to attachment styles with high anxiety (see Bartholomew & Horowitz, 1991), seem to be evident in the straight association between this dimension and the power tactic hostile control (potentially dysfunctional, based on the dominance, threat of separation, blaming the other, etc.).The characterization of interpersonal distance and coldness underlining the patterns with high avoidance, can justify the negative relationship with the power tactic benign affiliative autonomy, that is expressed through behaviours such as calming down, listening emphatically, reinforcing the autonomy and associated to the security of attachment (see Paiva, Figueiredo, & Henry, 2008).The collected data are congruent with the literature that associates the insecure attachment to the presence of dysfunctional intimate relationships (Bartholomew & Horowitz, 1991;Brennan & Shaver, 1995;Collins & Read, 1990;Feeney, 1994;Kirkpatrick & Davis, 1994;Simpson, 1990;Simpson et al., 1996;Treboux et al., 2004).
Not questioning the internal validity of the Portuguese version of ECR, some limitations are inherent to the present study.Firstly, the age of the participants and the marital status, the great majority are young adults with a dating relationship, their position considering the intimate relationships can be different from other older age groups (Hazan & Shaver, 1987), so it is convenient to take into account this limitation for the generalization of the results.Associated with this limitation, there is a need to enlarge the sample to other age groups and marital status, as a way to confirm the obtained results.Secondly, in terms of methodology the bias associated with self report measures, namely recovering past memories.Consequently it is necessary to pair with other measures, for example the interview to test the equivalence of the constructs (e.g., Bartholomew & Horowitz, 1991;Crowell, Fraley & Shaver, 2008).Third, the cultural specificity related to the evaluated constructs is a relevant aspect too, despite the unquestionable universality of the two dimensions of attachment in adulthood as its orthogonal structure in many countries, including Portugal and the Southern European countries (Schmitt et al., 2004), it is important to know more about the underlying constructs of anxiety and avoidance, an issue that rises importance when some items are not well fitted in some cultural adaptations of ECR, like item 21.A valuable contribution of this study is that it encompasses a measure of wishes, fears and power tactics that may permit to understand better the underlying dimensions of anxiety and avoidance related with attachment.
Forth, also in conceptual terms, it is important to mention that the duration and the valuation of the relationships as relevant factors on the attachment study, considering the different stages of establishing a conjugal bond as it is reported by Hazan and Zeifman (1999).The determination of individual differences in attachment can also benefit in future studies to compare the responses of a person focused on a particular relationship with a partner (the more recent) or in the general romantic relationships (ever relationships), this last was considered in the original and the present version of ECR, but can also be a remark to pay attention in comparative studies.
Lastly, to add that the romantic love conceptualized as Shaver and col. (1988) described integrated three systems: attachment, sexual and experiences of care, which are not considered in the present study.So, in future studies, it might be important to include the stability of relationships and the strategies of maintenance of them.
In general, the analysis of the psychometrics characteristics of the Portuguese version of ECR is satisfactory, and allows its applicability in the study of experiences with intimate relationships in the Portuguese context.
In clinical or investigation settings the use of ECR may be a helpful tool.In clinical settings it can be used with different purposes, especially as an orientation guide for the evaluation of close relationships in their different stages and processes (establishment, maintenance and dissolution), as an element of diagnosis in marital therapy, suggesting the meaning, the origin and the processes of change of relationship problems.As a research tool, it can be valuable for understanding social networks dimensions underlying romantic relationships, accessing the self and other concepts in romantic relationships, acceding to cognitive structures that guide behaviour and emotions with a partner, a way to guide conjugal outcomes, to understand deeply psychosocial functioning and develop more comprehensive and general models.
3. Main predictors of anxiety and avoidance; and 4. Norms and reference group criteria based on the quality of intimate relationship.The procedure occurred during the school period 45-60 minutes, in different classes in public and private universities in the North of Portugal, excluding psychology courses.Data analysis SPSS 17.0 and AMOS 17.0 were used for the computation of the data.The internal validity of the instrument was tested through a Confirmatory Factor Analysis (CFA) added by a Principal Components Analysis (PCA).

Table 1 .
Portuguese version of Experiences in Close Relationships Inventory N

Table 2 .
Factor structure, communalities and correlation item-total scale Differences with the original version.Considering the avoidance and the anxiety dimensions, almost all the discriminated items in the original version saturate preferably in the respective component (see Table2).Nevertheless, the items12, 16, 20 and 26 Alonso-Arbiol et al., 2007;Conradi et al., 2006;Picardi et al., 2000;scribed in other ECR validation studies as problematic (seeAlonso-Arbiol et al., 2007;Conradi et al., 2006;Picardi et al., 2000;for overviews), for that reason, we consider that it was preferable to removing them.

Table 3 .
Descriptives, correlation item-total and alpha Cronbach for ECR items and scales who don't have sexual relations with the partner.

Table 6 .
Differences in ECR´s items and scales considering participants relationship quality