The relevance of security: A latent domain of attachment relationships

30
1 Forthcoming paper: 1 Mannarini, S., & Boffo, M (in press). The relevance of Security: A latent domain of 2 attachment relationships. Scandinavian Journal of Psychology. 3 4 5 6 7 8 9 The relevance of Security: 10 A latent domain of attachment relationships 11 Stefania Mannarini and Marilisa Boffo 12 Department of Philosophy, Sociology, Education, and Applied Psychology 13 University of Padova 14 15 16 17 18 19 20 Correspondence concerning this article should be addressed to 21 Stefania Mannarini, 22 Department of Philosophy, Sociology, Education, and Applied Psychology 23 University of Padova 24 Via Venezia 8, 35131 Padova, Italy. 25 Phone +39 049 8276630 Fax +39 049 8276600 26 E-mail: [email protected] 27

Transcript of The relevance of security: A latent domain of attachment relationships

1

Forthcoming paper: 1

Mannarini, S., & Boffo, M (in press). The relevance of Security: A latent domain of 2

attachment relationships. Scandinavian Journal of Psychology. 3

4

5

6

7

8

9

The relevance of Security: 10

A latent domain of attachment relationships 11

Stefania Mannarini and Marilisa Boffo 12

Department of Philosophy, Sociology, Education, and Applied Psychology

13

University of Padova 14

15

16

17

18

19

20

Correspondence concerning this article should be addressed to 21

Stefania Mannarini, 22

Department of Philosophy, Sociology, Education, and Applied Psychology 23

University of Padova 24

Via Venezia 8, 35131 Padova, Italy. 25

Phone +39 049 8276630 Fax +39 049 8276600 26

E-mail: [email protected] 27

THE RELEVANCE OF ATTACHMENT SECURITY 2

Abstract 28

In the present study an adult attachment dimension, latent to the constructs of 29

security, anxiety, and avoidance, was hypothesized, wherein security was expected 30

to occupy the most relevant position. Furthermore, the reciprocal functioning of 31

attachment constructs and their interactions with self-esteem were explored. Four 32

hundreds and thirty-four Italian university students responded to two adult 33

attachment questionnaires (Attachment Style Questionnaire and Adult Attachment 34

Questionnaire) and to the Rosenberg Self-esteem scale. A Many-Facet Rasch 35

Measurement modelling approach was adopted. The main results can be summarized 36

as follows: (a) security, anxiety, and avoidance are nested under one latent 37

attachment dimension; (b) security occupies the most prominent position on the 38

dimension; (c) security is positively associated with a moderate level of attachment 39

anxiety and negatively related to avoidance; d) a positive interaction between self-40

esteem and security, and a negative relation between self-esteem and anxiety, were 41

detected. Theoretical, clinical, and empirical implications of the results are further 42

discussed. 43

44

Keywords: adult attachment; latent dimension; security; anxiety; avoidance; 45

self-esteem; self-report measure; Rasch modelling. 46

47

THE RELEVANCE OF ATTACHMENT SECURITY 3

The relevance of Security: 48

A latent domain of attachment relationships 49

Bowlby’s attachment theory (1969; 1973; 1980) is considered a milestone in 50

the study of psychological adaptation throughout the human lifespan. In the last 51

thirty years, this theory has been extensively applied in adult relationships research 52

across the social, developmental, clinical, and personality psychology realms (e.g., 53

Fraley & Shaver, 2000; Holmes & Johnson, 2009; Simpson, Rholes, & Winterheld, 54

2010; Waters, Crowell, Elliot, Corcoran, & Treboux, 2002). Despite giving a 55

promising confirmation of Bowlby’s attachment conceptualization, several studies 56

evidenced specific controversies, in particular related to the hypothesized number of 57

attachment styles and the underlying attachment dimensionality (e.g., Fraley & 58

Waller, 1998; Stein et al., 2002), and concerns about the cultural relevance of adult 59

attachment (e.g., Lee, Grossman, & Krishnan, 2008). 60

Most measures of adult attachment are based on Bowlby’s concepts (1969) 61

and have been extended to thoughts and feelings mainly related to adult romantic 62

relations (e.g., Ainsworth, Blehar, Waters, & Wall, 1987; George, Kaplan, & Main, 63

1996). Bowlby argued that whenever an individual looses trust in a significant other, 64

he/she tends to manifest an insecure attachment. Levels of insecurity and related 65

coping strategies are defined within a cognitive perspective as Internal Working 66

Models (IWMs – Bowlby, 1973, 1980). Several researchers set out that the 67

combination of these IWMs determines the individual differences in attachment 68

relationships and can be classified into different attachment styles (e.g., Kobak, Cole, 69

Ferenz-Gillies, Fleming, & Gamble, 1993; Main, Kaplan, & Cassidy, 1985). In the 70

eighties of the past century, three main attachment categories were then defined: 71

secure, anxious-avoidant, and anxious-ambivalent (e.g., Hazan & Shaver, 1987; 72

THE RELEVANCE OF ATTACHMENT SECURITY 4

Main et al., 1985). Later on, Bartholomew (1990) and Bartholomew and Horowitz 73

(1991) distinguished two aspects of avoidance: fearful-avoidant and dismissing-74

avoidant. Griffin and Bartholomew (1994) argued about the existence of two macro-75

dimensions in the IWMs, namely the model of Self and the model of Others, which 76

were supposed to be latent to four attachment prototypes: secure, fearful, dismissing, 77

and preoccupied. Brennan, Clark, and Shaver (1998) pointed out that the majority of 78

attachment self-report measures and results in most of the studies on adult 79

attachment dimensionality were partly overlapping. The application of almost every 80

attachment questionnaire existing at that time to a large number of individuals lead 81

to the identification of a two-dimensional model characterized by avoidance and 82

anxiety and to the development of the Experiences in Close Relationships 83

questionnaire (ECR), which was later revised (ECR-R – Fraley, Waller, & Brennan, 84

2000) and criticised for the cross-cultural validity (e.g., Lee et al., 2008). 85

In the current study, although appreciating the theoretical relevance and 86

practical advantages of the two-dimensional model based on avoidance and anxiety, 87

it was argued that this conceptualization is not in line with the core concept of 88

attachment theory, i.e. the secure/insecure attachment continuum (Bäckström & 89

Holmes, 2001). As Mikulincer and Florian (1998) pointed out, a secure attachment is 90

a resource that may help the individual to positively cope with stressful experiences. 91

Secure individuals deal with poor relational environments by activating more 92

adaptive behaviors than insecure individuals (e.g. Bowlby, 1979; Collins & Feeney, 93

2000). Further, a sense of security has been related to more positive self-views 94

(Bartholomew & Horovitz, 1991) and expectations towards the others (Collins & 95

Reads, 1990). Recent theoretical and applied research evidenced the utility of a 96

model that measures security directly and not as a lack of anxiety and avoidance, 97

THE RELEVANCE OF ATTACHMENT SECURITY 5

such as the hyperactivating/deactivating model (Mikulincer, Shaver, & Perg, 2003) 98

and the State Adult Attachment Measure (Gillath, Hart, Noftle, & Stockdale, 2009), 99

which provides a measure of the situational fluctuations in the sense of attachment 100

security/insecurity. In the present study a revised model of adult attachment 101

relationships was hypothesized, integrating the concepts of security, anxiety and 102

avoidance. This integration followed the hypothesis that individuals may react with 103

anxious or avoidant behavioural and cognitive responses to compensate for the lack 104

of security, and was aimed at a better understanding of individual differences across 105

situations (Bäckström & Holmes, 2001). 106

Research Hypotheses 107

The main research hypotheses were the following: 108

H1. Security, anxiety, and avoidance attachment representations are nested 109

within one latent dimension of global attachment (Overall, Fletcher, & 110

Friesen, 2003). A mono-dimensional model was then conceived and tested to 111

account for the performance on items representing these constructs. 112

According to Bejar (1983), the hypothesis was that, when responding to 113

items belonging to different constructs, a variety of mental and affective 114

processes are involved, but as long as the performance on each item is 115

affected by the same processes, e.g. the attachment working process, mono-116

dimensionality should hold. A subsequent hypothesis involved the constructs 117

relevance on the attachment dimension in terms of associated probability: 118

Security was indeed expected to be the most representative. In other words, 119

the probability that participants endorse security items using the highest 120

grades of the rating scale (agree and totally agree) was expected to be higher 121

when compared to the probability of endorsing anxiety and avoidance items. 122

THE RELEVANCE OF ATTACHMENT SECURITY 6

H2. A negative relation between security and anxiety and avoidance was 123

expected. 124

H3. The relation between the attachment constructs and self-esteem was also 125

examined, expecting security to be positively associated to self-esteem 126

(Bowlby, 1988). Indeed, children who experience a supportive and consistent 127

care-giving relationship develop a positive sense of self-regard, which 128

eventually becomes stable throughout the lifespan (Bartholomew & 129

Horowitz, 1991; Schmitt & Allik, 2005). A substantial link between self-130

esteem and anxiety and avoidance was also expected (e.g., Mannarini, 2010; 131

Tokar, Withrow, Hall, & Moradi, 2003). 132

A Rasch Modeling Approach to Adult Attachment 133

The research hypotheses were tested within a Many-Facet Rasch 134

Measurement modeling approach (MFRM – Linacre, 1989), which belongs to the 135

family of Rasch models. 136

Rasch models have a long tradition in the development and psychometric 137

analysis of psychological, educational, and medical assessment scales, and they have 138

been used as a template against which data collected from self-report measures may 139

be tested for measurement and construct validity, offering an elegant approach to 140

addressing several methodological and theoretical key-aspects in the scale validation. 141

The Rasch approach was chosen to accomplish, better than other methods, 142

the purposes of this study also for the probabilistic feature embedded in the Rasch 143

models: Given that all possible affectors of people’s responses to an assessment 144

measure are far not completely known, a probabilistic measurement approach 145

considers the power of a score to predict whether items are correctly endorsed and 146

evaluated by the respondents. 147

THE RELEVANCE OF ATTACHMENT SECURITY 7

To test hypothesis H1, person, items and the item rating scale were 148

parameterized in the MFRM, and the attachment constructs reciprocal relevance and 149

interactions on the latent dimension were explored in terms of probability. 150

To test hypothesis H2, anxiety and avoidance facets parameters were 151

introduced in the model, to examine their contribution to the security ratings and the 152

relations among the three constructs. 153

Eventually, to verify hypothesis H3, individual self-esteem levels were also 154

parameterized in the model to explore their relation with the attachment components. 155

Method 156

Participants 157

Participants were 434 Italian university students, 55% female, mean age 158

22.84 (SD=3.63). 25% of participants had never been in a romantic relationship and 159

75% was at least once. 160

Instruments 161

The Attachment Style Questionnaire (ASQ - Feeney, Noller, & Hanrahan, 162

1994; Italian version by Fossati et al., 2003) and the Adult Attachment Questionnaire 163

(AAQ - Colosso, Barbon, & Cusinato, 2006) were administered and expected to 164

concur to the validation of the attachment mono-dimensionality hypotheses. The 165

selection of two instruments aimed at verifying whether the research hypotheses hold 166

across the two cultural contexts of development. Both instruments were developed 167

according to Bartholomew’s attachment theorization (Bartholomew, 1990; 168

Bartholomew & Horowitz, 1991). The ASQ is a worldwide well-known measure of 169

adult attachment, whereas the AAQ was developed for the Italian context. 170

The Attachment Style Questionnaire. 171

THE RELEVANCE OF ATTACHMENT SECURITY 8

The ASQ consists of 40 items rated on a 6-point Likert scale from 1 (totally 172

disagree) to 6 (totally agree). According to Stein et al. (2002), the five subscales 173

were treated as three attachment constructs: Confidence (8 items), Avoidance (17 174

items; Discomfort with Closeness and Relationships as Secondary), and Anxiety (15 175

items; Need for Approval and Preoccupation with Relationships). 176

The Adult Attachment Questionnaire. 177

The AAQ is a 56-item scale derived for the Italian population from the 178

Relationship Questionnaire (RQ - Bartholomew & Horowitz, 1991). Items are rated 179

on a 4-point Likert scale, from 1 (strongly disagree) to 4 (strongly agree). According 180

to Bäckström and Holmes (2001), the four subscales were treated as three attachment 181

constructs: Security (28 items; Secure and Fearful), Preoccupation (14 items), and 182

Dismissing (14 items). 183

Since the same attachment constructs are worded differently in the two 184

questionnaires (e.g., self-confidence in the ASQ and security in the AAQ), from now 185

on the three constructs will be univocally denominated security, anxiety, and 186

avoidance, to simplify the reader’s comprehension. When needed, the constructs will 187

be referred to the ASQ and AAQ scales, respectively. 188

The Rosenberg Self-Esteem Scale. 189

The RSES (Rosenberg, 1965; Italian version by Prezza, Trombaccia, & 190

Armento, 1997) consists of 5 positively and 5 negatively worded items. The items 191

are rated on a 4-point Likert scale from 1 (strongly disagree) to 4 (strongly agree). 192

The Model 193

The MFRM (Linacre 1989) derives from the Simple Logistic Model (SLM 194

– Rasch 1980), which is the traditional and most basic Rasch model for the 195

transformation of ordinal observations into interval linear measures. The SLM is 196

THE RELEVANCE OF ATTACHMENT SECURITY 9

meant for dichotomous data and expresses the probability of a response as a function 197

of the respondent’s ability and the item’s difficulty , as formalized in the logit 198

scale (Rasch 1980), i.e., the logarithmic transformation of the probability 199

of giving a particular response given certain conditions. These two elements can be 200

considered as two facets that interact with each other to produce the response to an 201

item, and that can be modelled to operate independently, so that their parameter 202

estimates can be combined additively on a latent variable. However, in the 203

measurement contexts other aspects may interfere with the person and item’s 204

attributes, such as specific situational, social, and personality factors. Within the 205

mono-dimensionality requirement and properties of Rasch theory (for a review see, 206

Bond & Fox 2007), the MFRM takes into account these multifaceted situations, by 207

introducing new parameters -or facets-, into the model, to consider other sources 208

of systematic variability. 209

In the present study, besides the respondents’ parameter , describing the 210

attachment orientation (facet 1), and item parameter , which stands for the 211

representativeness of the attachment contents on the latent trait (facet 2), additional 212

facets were subsequently introduced in the model to account for the anxiety, 213

avoidance and self-esteem effects on the attachment scores. 214

In the MFRM model, facet parameters are estimated simultaneously and 215

independently from the distribution of participants and items, by means of the Joint 216

Maximum Likelihood Estimation procedure (Linacre 2010), and are located on the 217

same latent trait, allowing comparison among them. All facet parameter estimates, 218

except for the item parameter estimates δi, were positively scaled to indicate a secure 219

attachment orientation. The item parameter estimates δi were centred around the 220

origin of the logit scale (.00), with negative values indicating items easily endorsed 221

THE RELEVANCE OF ATTACHMENT SECURITY 10

by the respondents using the highest grades of the rating scale and strongly 222

representative of the underlying dimension, and positive measures indicating the 223

opposite. 224

To evaluate the goodness-of-fit of the parameter estimates, the MFRM 225

presents two main fit statistics which show how much the data for each estimate 226

adhere to the model requirements: the mean square Outfit and mean square Infit 227

statistics. These statistics are calculated for each participant, item and any other facet 228

parameter, and express the relationship between observed and model-derived 229

expected scores, ranging from zero to infinity. Statistics equal to or near 1 indicate 230

perfect correspondence between observed and expected values, whereas statistics 231

above or below 1 indicate the presence of greater or lower variance in the data than 232

that modelled (underfit or overfit). A range of .70–1.30 indicates a satisfactory fit of 233

the observed data to the model requirements (Bond & Fox, 2007). Infit and Outfit 234

statistics are both derived from the squared standardized residuals for each 235

item/participant interaction (for details, see Myford & Wolfe, 2003). The Outfit 236

statistic is the average of the standardized residuals squares and is unweighted, 237

meaning that it is more sensitive to outlier observations. For the Infit statistic the 238

residuals are information-weighted by their individual variance, thus relatively more 239

affected by inlying response patters. 240

A Chi-square statistic – the Fixed (all same) χ2 – is also provided for each 241

facet entered in the model and helps to reject the hypothesis that there is no 242

difference between elements of a model facet on the latent trait (i.e., the elements of 243

a facet have the same logit in relation to the measurement error), sketching out the 244

discriminating power of the facet. 245

Results 246

THE RELEVANCE OF ATTACHMENT SECURITY 11

The MFRM analyses were performed using FACETS 3.60.0 software 247

(Linacre, 2006). 248

Hypothesis H1: Definition of the Attachment Latent Dimension 249

The MFRM model specified only the person ( v ) and item ( i ) facet 250

parameters and was separately applied to the two questionnaires. An additional 251

parameter ( k ) accounting for the calibration of the Likert-type rating scale k={1, .., 252

m}was embedded in the model. 253

ASQ results. 254

The item parameter. A common latent dimension was recovered and 255

composed of 37 well-fitting items out of the 40 original ones. Three items presented 256

a misfit to the model requirements, with Infit and Outfit statistics values exceeding 257

the goodness-of-fit range [.70, 1.30], hence not well representing the latent 258

dimension. The three misfitting items were then dropped out and a second MFRM 259

analysis was conducted on the remaining 37 items, re-assuring the existence of the 260

latent trait, with fit statistics in the range .70 – 1.30. 261

The 37 item parameter estimates ranged between -1.11 – 1.00 ( ES = .04) 262

and were satisfactorily distributed along the latent dimension, indicating that they 263

can well represent different aspects of the attachment dimension (Fixed 2 (36, 434) = 264

6418.9, p < .001). Given that in Rasch analyses the items with low negative 265

parameter estimates are rated on the higher grades of the rating scale, the results 266

evidenced that the item with the lowest measure (δ = -1.11), “In general I feel as a 267

valuable person”, was a security item, which was the most representative of the 268

underlying attachment dimension. On the other hand, an avoidance item presented 269

the highest parameter measure (δ = 1.00) ,“To achieve one’s own objectives is more 270

THE RELEVANCE OF ATTACHMENT SECURITY 12

important than caring for the relationships with the others”, showing to be the less 271

representative of the general attachment latent dimension. 272

The person parameter. The mean person parameter estimate βv was equal 273

to -.33 (range = [-1.46, .83], ES = .15), with both fit statistics equal to 1.00. 274

AAQ results. 275

The item parameter. Also for the AAQ dataset, the existence of a common 276

dimension, latent to the three attachment constructs, was verified. A set of 52 items 277

out of the 56 original ones fitted properly the model requirements. Four items 278

presented unsatisfactory Infit and Outfit statistics and were discarded from the 279

analyses. A second MFRM analysis on the remaining 52 items confirmed the latent 280

trait, with fit statistics in the range .70 - 1.30. The 52 item parameter estimates were 281

in the range -1.78 – 1.90 ( ES = .06) and resulted to be satisfactorily distributed 282

along the latent trait, hence well representing different attachment orientations 283

(Fixed 2 (51, 434) = 12037.1, p < .001). Also for the AAQ the most representative 284

item of the attachment dimension (δ = -1.78) pertained to the construct of security: 285

“When I find myself in an embarrassing situation, I defend myself by running away” 286

(item score-reversed), whereas the least representative and endorsed item (δ = 1.90) 287

belonged to the avoidance construct: “I would renounce to any of my demands in 288

place of feeling loved by others”. 289

The person parameter. The person parameter estimate βv mean value was -290

.11 ( ES = .18), with Outfit and Infit fit statistics equal to 1.00 and 1.02, 291

respectively. 292

Attachment constructs relevance on the latent dimension. 293

To compare the relevance of the different constructs on the common latent 294

dimension, a MFRM probability value for each construct was computed according to 295

THE RELEVANCE OF ATTACHMENT SECURITY 13

the model specified to test hypothesis H1: the mean person ( v ) and item ( i ) 296

parameter estimates were calculated for each attachment construct and questionnaire 297

and equated into the model to compute the MFRM probability value for each grade 298

of the rating scale ( k ) (see Bond & Fox, 2007). The obtained values were then 299

averaged across the rating scale grades providing a mean MFRM probability value 300

for each attachment construct and self-report. Further, the attachment probability 301

proportion was computed, to evidence the relative construct relevance on the latent 302

dimension (see Table 1). 303

[insert Table 1 about here] 304

The ASQ and AAQ security constructs presented the lowest mean parameter 305

measures and the lowest upper and lower range limits. This result indicates that the 306

security items generally received higher scores on the rating scale in both measures. 307

The mean probability values reported in Table 1 further confirmed the previous 308

results. Both ASQ and AAQ security constructs presented higher values when 309

compared with the other constructs, consistently throughout the rating scale, 310

supporting the hypothesis that they are endorsed by the participants with higher 311

rating values. In other words, security contents are “easier” to be shared by the 312

respondents, given the relevance they play in characterizing the attachment latent 313

domain. 314

The mean probability proportions displayed in Table 1 further highlighted the 315

previous findings. Considering the ASQ, the 41% of the attachment domain is 316

occupied by security (e.g., .41 = .58/(.58 + .44 + .38)), whereas for the AAQ the 317

security proportion reaches the 53% of the attachment domain. Altogether, these 318

results show that both questionnaires evidenced the prominent position of security 319

THE RELEVANCE OF ATTACHMENT SECURITY 14

over the other constructs, which nonetheless are substantial components of the 320

attachment domain. 321

Hypothesis H2: Attachment Constructs Interrelations 322

To test hypothesis H2 a second MFRM model was applied to the subset of 323

security items for each questionnaire (see Figure 1). The constructs of anxiety and 324

avoidance were equated as two model facets contributing to the probability of giving 325

a score x to the security items, by discretizing their score distributions over the 33rd

326

and 66th

percentiles to obtain three ordered categories: Low, Middle and High 327

(Blanton & Jaccard, 2006). This discretization procedure complies with the Rasch 328

modelling requirement of entering only discrete variables in the model. 329

[insert Figure 1 about here] 330

Security-anxiety interrelation. 331

For the ASQ (circled line), Figure 1(a) shows that a Middle level of anxiety 332

corresponded to the highest measure of security (.19), whereas a High level of 333

anxiety corresponded to the lowest measure of security (-.27). Further, a Low level of 334

anxiety was associated to a security measure slightly over the mean (.08). In other 335

words, when attachment security tends to be high, a middle level of anxiety is also 336

present. The ASQ anxiety levels evidenced a satisfactory discriminating property for 337

the security estimates on the latent attachment trait (Fixed 2 (2, 434) = 57.4, p < .001). 338

When considering the AAQ (squared line in Figure 1(a)), a similar relation 339

between security and anxiety emerged (the security parameter estimates were .06, 340

.07, and -.14 for Low, Middle, and High anxiety levels, respectively). The 341

discriminating power of the three AAQ anxiety levels was also satisfactory (Fixed342

2 (2, 434) = 50.7, p <.001). 343

Security-avoidance interrelation. 344

THE RELEVANCE OF ATTACHMENT SECURITY 15

Figure 1(b) shows that for the ASQ scale security and avoidance are 345

consistently negatively linked together (the security estimates were -.26, .03, and .25 346

for High, Middle and Low avoidance levels, respectively).The discriminating 347

property of the ASQ avoidance levels for the security latent values was satisfactory (348

2 (2, 434) = 64.1, p < .001). A similar result characterized the interaction between 349

security and avoidance for the AAQ (Fixed 2 (2, 434) = 18.6, p < .001). 350

Hypothesis H3: Self-Esteem and Attachment Constructs relations 351

To test hypothesis H3 a third MFRM model was separately applied to the 352

items of each attachment construct, for each questionnaire(see Figure 2). The 353

construct of self-esteem was equated as a model facet, accounting for the probability 354

of giving any score x to the attachment constructs items. Like in the previous 355

analysis, the same discretization procedure was applied to the self-esteem score 356

distribution. 357

[insert Figure 2 about here] 358

Security-self-esteem interrelation. 359

Considering the ASQ security construct (circled line), Figure 2(a) 360

demonstrates that the highest value of security (.20) was associated to the Middle 361

level of self-esteem, whereas the lowest measure (-.28) to the Low level of self-362

esteem. A security value near the mean ( .07) was linked to the High self-esteem 363

level. These results did not totally confirm the expected positive relation between 364

self-esteem and security, as they suggest that a strong secure attachment might be 365

better off associated to the middle than to the high level of positive self-evaluation. 366

On the other hand, the analysis identified an association between low security and 367

Low level of self-esteem, in accordance with the expectations. The discriminating 368

property of the self-esteem levels for the security measures on the latent trait was 369

THE RELEVANCE OF ATTACHMENT SECURITY 16

satisfactory (Fixed 2 (2, 434) = 66.2, p < .001). Also for the AAQ, the self-esteem and 370

security constructs (squared line in Figure 2(a)) showed a similar interaction and an 371

adequate self-esteem discriminating power (Fixed 2 (2, 434) = 5.5, p < .05). 372

Anxiety-self-esteem interrelation. 373

Considering the ASQ anxiety construct (circled line), Figure 2(b) shows that 374

the Middle and High levels of self-esteem corresponded to the lowest measures of 375

anxiety (-.35 and -.26, respectively). Further, a Low self-esteem corresponded to the 376

highest measure of anxiety (.61). The discriminating power of self-esteem for the 377

anxiety latent measures was indeed confirmed (Fixed 2 (2, 434) = 1199.7, p < .001). 378

The AAQ anxiety construct (squared line in Figure 2, panel b)) showed similar 379

results, and a sufficient self-esteem’s discriminating power in differentiating the 380

anxiety latent measures (Fixed 2 (2, 434) = 5.6, p < .05). 381

These results suggested that one’s positive self-evaluation can have a strong 382

negative relation with anxious attachment relationships. 383

Avoidance-self-esteem interrelation. 384

No significant relation between self-esteem and avoidance was found. 385

Discussion 386

Altogether, the findings of this study can be summarized within the existing 387

theoretical and empirical literature, as follows: 388

As hypothesized (H1), security, anxiety, and avoidance representations of 389

attachment are nested within one latent attachment dimension, wherein 390

security occupies the most relevant position and anxiety and avoidance are 391

present to a less extent. From a theoretical point of view, these findings seem 392

to support the interpretation of security as a primary factor within a 393

developmental attachment process (e.g., Ainsworth & Bowlby, 1991). 394

THE RELEVANCE OF ATTACHMENT SECURITY 17

Following the suggestion about the strong importance of attachment security 395

(e.g., Bäckström & Holmes, 2001; Gillath et al., 2009; Shaver & Mikulincer, 396

2002), this construct should be then measured and conceptualized 397

independently of the other attachment constructs. It is not sufficient to define 398

security as a lack of anxiety and avoidance, but rather the three constructs are 399

part of the same space. Indeed, participants’ responses to both measures 400

indicated that security presents the highest attachment representative 401

probability. 402

Considering hypothesis H2, the results demonstrated that security is 403

significantly related to anxiety and avoidance. A Middle level of anxiety 404

corresponded to the highest measure of security, indicating that feeling 405

secure in attachment relationships does not completely exclude feelings of 406

relational anxiety. On the other hand, security presented an opposite relation 407

with avoidance, i.e. a person may identify him/herself as feeling insecure in 408

relationships, but at the same time he/she may use an avoidant behaviour to 409

negate or escape his/her problems. In line with previous findings (e.g., 410

Baldwin, Keelan, Fehr, Enns, & Koh-Rangarajoo, 1996; Ross & Spinner, 411

2001), it seems that constructs such as anxiety and avoidance act differently, 412

at least to some extent, in their relations with security, and can be deemed as 413

a sort of “strategy” to better interpret and deal with secure/insecure 414

relationships. 415

The hypothesis H3 about the significant interaction between the attachment 416

constructs and self-esteem was confirmed, except for the avoidant 417

component. This last result might be partially explained by the fact that the 418

respondents are university students, whose self-esteem is mostly related to 419

THE RELEVANCE OF ATTACHMENT SECURITY 18

feeling successful in school, which may not be related to avoiding close 420

relationships. On the other hand, a middle level of self-esteem was associated 421

to the lowest measure of anxiety, whereas medium and high security levels 422

were related to high and middle levels of self-esteem, respectively. These 423

results seem to support what Bowlby (1988) said about the positive relation 424

between supportive care-giving in childhood and sense of self-regard in 425

adulthood. Furthermore, recent research recovered that negative measures of 426

security and low self-esteem mediates the generation of depression and 427

anxiety symptoms, shedding light on the potential clinical implications of 428

attachment constructs in the development of psychological disorders (e.g., 429

Hankin, Kassel, & Abela, 2005; Lee & Hankin, 2009; Riggs & Hans, 2009; 430

Roberts, Gotlib, & Cassel, 1996) and inspiring further research within the 431

personal relationships and psychopathological domains. 432

From a measurement perspective, both attachment scales supported the 433

hypotheses, yielding the suggestion of the communalities of adult attachment 434

processes across two different cultural contexts of scale development. 435

Within a confirming and replicating perspective, further studies are needed 436

targeting participants differently aged and with varying cultural backgrounds. 437

Moreover, not only other attachment questionnaires can be compared to verify 438

whether this theoretical perspective holds consistently across measures, but also 439

indirect measures, tapping those processes that are not directly measurable and work 440

outside the individual conscious control, might be used to assess the cognitive and 441

affective components of attachment working models at an automatic level (e.g., 442

Dewitte, De Houwer, & Buysse, 2008; Zayas & Shoda, 2005). This suggestion 443

follows the idea of extending the present research to the next step, by drawing a 444

THE RELEVANCE OF ATTACHMENT SECURITY 19

bridge between the implicit nature of the attachment working processes and more 445

indirect measurement techniques which do not rely on subjective explicit evaluations 446

and possible self-biases. 447

448

449

450

451

452

453

454

455

456

457

458

459

460

461

462

463

464

465

466

467

468

469

THE RELEVANCE OF ATTACHMENT SECURITY 20

References 470

Ainsworth, M.D.S., Blehar, M.C., Waters, E., & Wall, S. (1978). Patterns of 471

attachment: Assessed in the strange situation and home. Hillsdale, NJ: 472

Lawrence Erlbaum. 473

Ainsworth, M.S., & Bowlby, J. (1991). An ethological approach to personality 474

development. American Psychologist, 46(4), 333-341. doi: 10.1037/0003-475

066X.46.4.333 476

Bäckström, M., & Holmes, B.M. (2001). Measuring adult attachment: A construct 477

validation of two self-report instruments. Scandinavian Journal of 478

Psychology, 42(1), 79-86. doi: 10.1111/1467-9450.00216 479

Baldwin, M.W., Keelan, J.P.R., Fehr, B., Enns, V., & Koh-Rangarajoo, E. (1996). 480

Social-cognitive conceptualization of attachment working models: 481

Availability and accessibility effects. Journal of Personality and Social 482

Psychology, 71, 94-109. doi: 10.1037/0022-3514.71.1.94 483

Bartholomew, K. (1990). Avoidance of intimacy: An attachment perspective. 484

Journal of Social and Personal Relationships, 7, 147-178. doi: 485

10.1177/0265407590072001 486

Bartholomew, K., & Horowitz, L.M. (1991). Attachment styles among young adults: 487

A test of a four category model. Journal of Personality and Social 488

Psychology, 61, 226-244. doi: 10.1037/0022-3514.61.2.226 489

Bejar, I.I. (1983). Achievement in testing: recent advances. Beverly Hills: Sage 490

Publications. 491

Blanton, H., & Jaccard, J. (2006). Arbitrary metrics redux. American Psychologist, 492

61, 62-71. doi: 10.1037/0003-066X.61.1.62 493

THE RELEVANCE OF ATTACHMENT SECURITY 21

Bond, T.G., & Fox, C.M. (2007). Applying the Rasch Model. Fundamental 494

measurement in the human sciences. 2nd

Edition. Mahwah, NJ: Lawrence 495

Erlbaum Associates Publishers. 496

Bowlby, J. (1969). Attachment and loss, Vol.1 Attachment. New York: Basic Books. 497

Bowlby, J. (1973). Attachment and loss, Vol. 2 Separation: Anxiety and anger. New 498

York: Basic Books. 499

Bowlby, J. (1979). The making and breaking of Affectional bonds. New York: 500

Routledge. 501

Bowlby, J. (1980). Attachment and loss, Vol. 3 Loss: Sadness and depression. New 502

York: Basic Books. 503

Bowlby, J. (1988). A secure base. New York: Routledge. 504

Brennan, K.A., Clark, C.L., & Shaver, P.R. (1998). Self report measurement of adult 505

attachment: An integrative overview. In J.A. Simpson, & W.S. Rholes (Eds.), 506

Attachment theory and close relationships (pp. 46-76). New York: The 507

Guilford Press. 508

Collins, N.L., & Feeney, B.C. (2000). A safe heaven: An attachment theory 509

perspective on support seeking and care-giving in intimate relationships. 510

Journal of Personality and Social Psychology, 78(6), 1053-1073. doi: 511

10.1037/0022-3514.78.6.1053 512

Collins, N.L., & Read, S.J. (1990). Adult attachment, working models, and 513

relationship quality in dating couples. Journal of Personality and Social 514

Psychology, 58(4), 644-663. doi: 10.1037/0022-3514.58.4.644 515

Colosso, W., Barbon, M., & Cusinato, M. (2006). Questionario di attaccamento fra 516

adulti (QAA). [The Adult Attachment Questionnaire (AAQ)]. Famiglia, 517

Interdisciplinarietà, Ricerca, 11, 267-292. 518

THE RELEVANCE OF ATTACHMENT SECURITY 22

Dewitte, M., De Houwer, J., & Buysse, A. (2008). On the role of the implicit self-519

concept in adult attachment. European Journal of Psychological Assessment, 520

24, 282-289. doi: 10.1027/1015-5759.24.4.282 521

Feeney, J.A., Noller, P., & Hanrahan, M. (1994). Assessing adult attachment. In 522

M.B. Sperling & W.H. Barman (Eds.), Attachment in adults: Clinical and 523

developmental perspectives (pp.128-152). New York: Guilford Press. 524

Fossati, A., Feeney, J.A., Donati, D., Donini, M., Novella, L., Bagnato, M., ... 525

Maffei, C. (2003). On the dimensionality of the Attachment Style 526

Questionnaire in Italian clinical and non-clinical participants. Journal of 527

Social and Personal Relationships, 20(1), 55-79. doi: 528

10.1177/02654075030201003 529

Fraley, R.C., & Shaver, P.R. (2000). Adult romantic attachment: Theoretical 530

developments, emerging controversies, and unanswered questions. Review of 531

General Psychology, 4, 132-154. doi: 10.1037/1089-2680.4.2.132 532

Fraley, R.C., & Waller, N.G. (1998). Adult attachment patterns: A test of the 533

typological model. In J.A. Simpson, & W.S. Rholes (Eds.), Attachment 534

theory and close relationships (pp. 77-114). New York: The Guilford Press. 535

Fraley, R.C., Waller, N.G., & Brennan, K.A. (2000). An item response theory 536

analysis of self-report measures of adult attachment. Journal of Personality 537

and Social Psychology, 78, 350-365. doi: 10.1037/0022-3514.78.2.350 538

George, C., Kaplan, N., & Main, M. (1996). Adult Attachment Interview Protocol. 539

Unpublished manuscript. 540

Gillath, O., Hart, J., Noftle, E.E., & Stockdale, G.D. (2009). Development and 541

validation of a state adult attachment measure (SAAM). Journal of Research 542

in Personality, 43(3), 362-373. doi: 10.1080/14616730010024771 543

THE RELEVANCE OF ATTACHMENT SECURITY 23

Griffin, D.W., & Bartholomew, K. (1994). The metaphysics of measurement: The 544

case of adult attachment. In K. Bartholomew & D. Perlman (Eds.), Advances 545

in personal relationships, Vol. 5: Attachment processes in adulthood (pp.17-546

52). London: Jessica Kingsley. 547

Hankin, B.L., Kassel, J.D., & Abela, J.R.Z. (2005). Adult attachment dimensions 548

and specificity of emotional distress symptoms: Prospective investigations of 549

cognitive risk and interpersonal stress generation as mediating mechanisms. 550

Personality Social Psychology Bulletin, 31, 136–151. doi: 551

10.1177/0146167204271324 552

Hazan, C., & Shaver, P. (1987). Romantic love conceptualized as an attachment 553

process. Journal of Personality and Social Psychology, 52, 511-524. doi: 554

10.1037/0022-3514.52.3.511 555

Holmes, B.M., & Johnson, K.R. (2009). Adult attachment and romantic partner 556

preference: A review. Journal of Social and Personal Relationships, 26, 833-557

852. doi:10.1177/0265407509345653 558

Kobak, R.R., Cole, H.E., Ferenz-Gillies, R., Fleming, W.S., & Gamble, W. (1993). 559

Attachment and emotion regulation during mother-teen problem solving: A 560

control theory analysis. Child Development, 64, 231-245. doi: 561

10.1111/j.1467-8624.1993.tb02906.x 562

Lee, A., & Hankin, B.L. (2009). Insecure attachment, dysfunctional attitudes, and 563

low self-esteem predicting prospective symptoms of depression and anxiety 564

during adolescence. Journal of Clinical Child & Adolescent Psychology, 38, 565

219-231. doi: 10.1080/15374410802698396 566

Lee, Y-S., Grossman J., & Krishnan, A. (2008). Cultural relevance of adult 567

attachment. Rasch modeling of the revised Experiences in Close 568

THE RELEVANCE OF ATTACHMENT SECURITY 24

Relationships in a Korean sample. Educational and Psychological 569

Measurement, 68, 824-844. doi: 10.1177/0013164407313367 570

Linacre, J.M. (1989). Many-Facet Rasch measurement. Chicago, IL: MESA Press. 571

Linacre, J.M. (2006). Facets computer program for many-facet Rasch measurement, 572

version 3.60.0. Beaverton, Oregon: Winsteps.com. 573

Linacre, J. M. (2010). FACET Rasch Model Computer Program: Users Guide 574

[Computer software manual]. Chicago: Winsteps.com. 575

Main, M., Kaplan, N., & Cassidy, J. (1985). Security in infancy, childhood, and 576

adulthood: A move to the level of representation. In I. Bretherton & E. 577

Waters (Eds.), Growing points of attachment theory and research 578

(Monographs of the Society for Research in Child Development, 50 (1-2, 579

Serial No. 209, pp. 66-106). Chicago, IL: University of Chicago Press. 580

Mannarini, S. (2010). Assessing the Rosenberg Self-Esteem Scale dimensionality 581

and items functioning in relation to self-efficacy and attachment styles. 582

Testing, Psychometric, Methodology in Applied Psychology, 17(4), 229-242. 583

Mikulincer, M., & Florian, V. (1998). The relationship between adult attachment 584

styles and emotional and cognitive reactions to stressful events. In J.A. 585

Simpson & W.S. Rholes (Eds.), Attachment theory and close relationships 586

(pp. 143-165). New York: Guilford Press. 587

Mikulincer, M., Shaver, P.R., & Pereg, D. (2003). Attachment theory and affect 588

regulation: The dynamics, development, and cognitive consequences of 589

attachment-related strategies. Motivation and Emotion, 27(2), 77-102. doi: 590

10.1023/A:1024515519160 591

THE RELEVANCE OF ATTACHMENT SECURITY 25

Myford, C.M., & Wolfe, E.W. (2003). Detecting and measuring rater effects using 592

many-facet Rasch measurement: Part I. Journal of Applied Measurement, 4, 593

386-422. 594

Overall, N.C., Fletcher G.J.O., & Friesen, M.D. (2003). Mapping the intimate 595

relationship mind: Comparisons between three models of attachment 596

representations. Personality and Social Psychology Bulletin, 29, 1479-1493. 597

doi:10.1177/0146167203251519 598

Prezza, M., Trombaccia, F.R., & Armento, R. (1997). La scala dell’autostima di 599

Rosenberg: traduzione e validazione italiana. [The Rosenberg self-esteem 600

scale: Italian translation and validation]. Bollettino di Psicologia Applicata, 601

223, 35-44. 602

Rasch, G. (1980). Probabilistic models for some intelligence and attainment tests. 603

Chicago: University of Chicago Press. (First published in 1960) 604

Ross, L.R., & Spinner, B. (2001). General and specific attachment representations in 605

adulthood: Is there a relationship? Journal of Social and Personal 606

Relationships, 18, 747-766. doi: 10.1177/0265407501186001 607

Riggs, S.A., & Han, G. (2009). Predictors of anxiety and depression in emerging 608

adulthood. Journal of Adult Development, 16, 39-52. doi: 10.1007/s10804-609

009-9051-5 610

Roberts, J.E., Gotlib, I.H., & Kassel, J.D. (1996). Adult attachment security and 611

symptoms of depression: The mediating roles of dysfunctional attitudes and 612

low self-esteem. Journal of Personality and Social Psychology, 70, 310–320. 613

doi: 10.1037/0022-3514.70.2.310 614

Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ: 615

Princeton University Press. 616

THE RELEVANCE OF ATTACHMENT SECURITY 26

Schmitt, D.P., & Allik, J. (2005). Simultaneous administration of the Rosenberg 617

Self-Esteem Scale in 53 nations: Exploring the universal and culture-specific 618

features of global self-esteem. Journal of Personality and Social Psychology, 619

89, 623-642. doi: 10.1037/0022-3514.89.4.623 620

Shaver, P.H., & Mikulincer, M. (2002). Attachment-related psychodinamics. 621

Attachment and Human Development, 4(2), 133-161. doi: 622

10.1080/14616730210154171 623

Simpson, J.A., Rholes, W.S., & Winterheld, H.A. (2010). Attachment working 624

models twist memories of relationship events. Psychological Science, 21, 625

252-259. doi: 10.1177/0956797609357175 626

Stein, H., Koontz, A.D., Fonagy, P., Allen, J.G., Fultz, J., Brethour, J.R. Jr.,...Evans, 627

R.B. (2002). Adult attachment: What are the underlying dimensions? 628

Psychology and Psychotherapy: Theory, Research and Practice, 75, 77-91. 629

doi: 10.1348/147608302169562 630

Tokar, D.M., Withrow, J.R., Hall, R.J., & Moradi, B. (2003). Psychological 631

separation, attachment security, vocational self-concept crystallization, and 632

career indecision: A structural equation analysis. Journal of Counseling 633

Psychology, 50, 3-19. doi: 10.1037/0022-0167.50.1.3 634

Waters, E., Crowell, J., Elliot, M., Corcoran, D., & Treboux, D. (2002). Bowlby’s 635

secure base theory and the social/personality psychology of attachment 636

styles: Work(s) in progress. Attachment and Human Development, 4, 230-637

242. doi: 10.1080/14616730210154216 638

Zayas, V., & Shoda, Y. (2005). Do automatic reactions elicited by thoughts of 639

romantic partner, mother, and self relate to adult romantic attachment? 640

THE RELEVANCE OF ATTACHMENT SECURITY 27

Personality and Social Psychology Bulletin, 31, 1011-1025. 641

doi:10.1177/0146167204274100 642

THE RELEVANCE OF ATTACHMENT SECURITY 28

Table 1. Relevance of ASQ and AAQ constructs on the latent attachment dimension: Item

parameter estimates range and mean values ( i range and i ), probability value for each

rating scale category ( k ), mean probability, and attachment probability proportion, for

each attachment construct.

Item Rating Scale

Probability up to

Category k

i range i 2 3 4 5 6

Mean

Probability

Probability

Proportion

ASQ

Security -1.11 – -.41 -.72 .83 .76 .67 .39 .26 .58 .41

Anxiety -.84 – .63 -.02 .71 .61 .51 .24 .15 .44 .31

Avoidance -1.02 – 1.00 .27 .65 .54 .43 .19 .10 .38 .27

AAQ

Security -1.78 – .41 -.90 .85 .71 .44 .67 .53

Anxiety .34 – 1.9 .99 .46 .27 .11 .28 .22

Avoidance - .13 – 1.32 .72 .46 .32 .13 .30 .24

THE RELEVANCE OF ATTACHMENT SECURITY 29

Figures

Figure 1. ASQ and AAQ security measures at the three levels of Anxiety (a) and

Avoidance (b).

Low Medium High

-0.3

-0.1

0.1

0.3 ASQ

AAQ

Attachment Anxiety

Attachm

ent S

ecuri

ty

Low Medium High

-0.4

-0.2

0.0

0.2

0.4 ASQ

AAQ

Attachment Avoidance

Attachm

ent S

ecuri

ty

a) b)

THE RELEVANCE OF ATTACHMENT SECURITY 30

Figure 2. ASQ and AAQ security (a) and anxiety (b) measures at the three levels of

Self-esteem.

Low Medium High

-0.4

-0.2

0.0

0.2

0.4 ASQ

AAQ

Self-esteem

Attachm

ent S

ecuri

ty

Low Medium High

-0.5

0.0

0.5

1.0 ASQ

AAQ

Self-esteem

Attachm

ent A

nxie

ty

a) b)