Edited by: Pietro Cipresso, Istituto Auxologico Italiano (IRCCS), Italy
Reviewed by: Harsha Perera, University of Nevada, Las Vegas, United States; Donald Sharpe, University of Regina, Canada
*Correspondence: Silvia Testa
This article was submitted to Quantitative Psychology and Measurement, a section of the journal Frontiers in Psychology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Efforts to evaluate adult attachment have led to the production of a vast number of tools, both narrative and self-report. These instruments examine a diversity of contents (early attachment experience, current representations of attachment, romantic relationships, representations of caregiving) and are used in different fields of research: self-report tools have mainly been developed in social research contexts, and are generally administered to large samples (Crowell et al.,
The AAI is generally viewed as the benchmark tool for evaluations of adult attachment, although it is difficult and time consuming to administer, like all narrative instruments. Self-report tools are easier and more economical to use, but can be over-simplistic and therefore fail to access distinctive features of the different attachment styles, especially in relation to unconscious representations as in the case of dismissing attachment (Maier et al.,
The aim of our study is to evaluate the factorial structure of the CA-MIR (CArtes-Modèles Individuels de Relations: Pierrehumbert et al.,
The CA-MIR is a tool that shares aspects of both the above-mention research lines. In fact, one of its strengths lies in the opportunity it gives to be administered in a Q-Sort version that can solve at least some social desirability problems and be more effective in capturing even unaware aspects of individual representations. Nevertheless, in this study, only the Likert format is considered, since responses need to be independent in order to apply factor analysis tools.
Informed by attachment theory, the CA-MIR assumes that a self-and-other model drives personal relationships throughout life and adulthood: this model is proposed to have been generated by early experience with caregivers, although it may be modified by subsequent relational experiences. The questionnaire is intended for use with people of different ages (from adolescence to old age), and of both genders, regardless of family background and current family circumstances: parents or currently childless; currently or no longer living with family of origin; single or with a partner; coming from families with two parents, a single parent, or an adoptive family; with or without siblings.
The CA-MIR comprises 72 items designed to explore three different areas of representation: the present (questions about respondents' relationships with their current family unit), the past (questions concerning aspects of respondents' past experience with both parents, or with one parent in particular) and their
Representation of
Theoretical structure of the instrument and sample items.
| As a child I was afraid of being abandoned. (39) | When I go away from those who are close to me, I feel ill at ease. (56) | I feel that I did not have the opportunity to assert myself in the environment I grew up in. (52) | ||||
| Even if it is not true, I feel that I had the best parents in the world. (53) | I trust those who are close to me. (36) | I enjoy thinking about my childhood. (25) | ||||
| When I was a child, my wishes did not count much for the adults around me. (29) | I hate feeling dependent on others. (12) | Every time I try to think of my parents' good sides, I end up recalling their bad sides. (60) | ||||
| As a child, I was afraid of my parents. (33) | I have difficulty accurately recalling childhood events. (51) | |||||
| When I was a child, I would set the grownups against one another to get what I wanted. (44) | In a family, it is very important that there be respect for the parents. (8) |
For each of the three areas of representation, there are items characterizing three different attachment styles, namely
In previous studies that adopted the Q-Sort format, the instrument showed good convergent and discriminant validity
Considering the complexity of the CA-MIR theoretical model and the only partially overlapping results of the analyses just reviewed, additional analysis of the Likert-format items is necessary. The aim of the current study was to contribute to clarifying the factorial structure of the instrument, by performing confirmatory factor analysis (CFA) to directly verify the goodness-of-fit of the original 13-factor structure and assessing the concurrent and discriminant validity of the CA-MIR latent dimensions. According to the attachment literature and the item content, the following expectations were formulated about the latent dimensions' correlation. Due to a certain persistence of relational modalities across the life span, a moderate positive correlation between
We recruited 979 participants during the spring and fall of 2005, using a snowball sampling method. About one third of the sample (30.1%) was made up of people related to each other, such as siblings, partners, parents, and children. To reduce the potentially distorting effect of family ties among participants, we removed one member from each interrelated pair of respondents, two members from each interrelated group of three or four and three members from each interrelated group of five
Sample (
| ISTAT, |
||
| 15-19 | 3.5 | 5.7 |
| 20-29 | 47.0 | 13.8 |
| 30-39 | 11.9 | 18.7 |
| 40-49 | 20.3 | 17.4 |
| 50-59 | 14.9 | 15.1 |
| 60 and over | 2.4 | 29.3 |
| ISTAT, |
||
| Percentage of females | 61.2 | 51.9 |
| ISTAT, |
||
| Primary and middle school | 23.1 | 64.3 |
| High school | 59.6 | 26.5 |
| University | 17.3 | 9.1 |
| ISTAT, |
||
| Self-employed | 13.2 | 11.9 |
| In third party employment | 45.8 | 32.7 |
| Housewife | 7.1 | 17.7 |
| Student | 29.3 | 8.9 |
| Retired | 3.1 | 21.9 |
| Other (seeking employment and other categories of unemployed) | 1.5 | 7.0 |
| ISTAT, |
||
| Single | 51.0 | 43.9 |
| Cohabiting | 5.6 | |
| Married | 36.8 | 52.4 |
| Divorced | 5.2 | 1.2 |
| Widowed | 1.5 | 2.5 |
| ISTAT, |
||
| Having one or more children: | 79.0 | 74.3 |
| Number of children: | ||
| 1 | 43.4 | 47.8 |
| 2 | 46.3 | 41.3 |
| 3 or more | 10.3 | 10.9 |
Other sample characteristics (comparison data for the Italian population not available).
| 15-81; 34.2; 13.4 | |
| Northwestern Italy | 80.2 |
| Other | 19.8 |
| Yes | 40.4 |
| No | 59.6 |
| Mother | 84.2 |
| Father | 72.1 |
| Early bereavement | 1.0 |
| Parental separation/divorce | 1.3 |
| Social disadvantage | 0.2 |
| Childhood adoption | 0.1 |
| Child of single parent | 0.2 |
| Parent-child separation | 0.2 |
| Illness or death of a loved one | 1.7 |
| Illness | 0.1 |
| Trauma | 0.4 |
| Separation/divorce | 0.1 |
| Psychosocial issues | 0.4 |
Most of the respondents were from Northwestern Italy (80.2%). Their ages ranged from 15 to 81 years (M = 34 years;
Fifty-one percent of the sample were single (vs. the lower level of 43.9% in the Italian population); 5.6% cohabited with a partner, 36.8% were married (vs. the higher level of 52.4% in the general population); the percentage of divorcees was higher than in the Italian population while the percentage of widowed persons was lower (ISTAT,
With a view to test and cross-validate the factor structure of CA-MIR, we randomly divided the sample into two sub-samples
The present study involved human beings in a data collection process aimed at gaining information about attitudes related to close interpersonal relationships and socio-demographical data from people aged 15 years old or more. The research conformed to the Codice Etico AIP (ethical code for research in psychology, Italian Psychological Association) and the provisions of the Italian laws on privacy and data protection (L. 196/2003). The participants volunteered to participate in the research, and signed an informed consent form in which they agreed to anonymously complete a Q-sort questionnaire and allow the researchers to use the data for scientific purposes. In case of minors, also parents' consent was collected.
The questionnaires were administered by undergraduate students who had received 20 h of
The first author translated the questionnaire from French to Italian, discussing and revising the translation with the Swiss authors and with a colleague native speaker both in Italian and French, and producing Italian versions of the instrument for both male and female respondents
Respondents' socio-demographic data was collected concurrently, and each participant's family ties with other respondents, such as siblings, partners, parents, or children, were recorded. In addition, we collected information regarding non-normative life experiences that could potentially influence participants' attachment representations (e.g., juvenile grief, illness, divorce).
On the data of Sample 1, the original 13-factor structure of CA-MIR was tested by a confirmatory factor model in which all cross-loadings were constrained to be zero, i.e., by means of an independent clusters model (ICM-CFA, Marsh et al.,
In the present work, it was not possible to apply the ESEM approach because the high number of latent factors (13) would have required the estimation of more parameters than the number of participants (930 free parameters); and neither it was feasible to identify the subset of items affected by multidimensionality on theoretical grounds. We thus performed an exploratory factor analysis with target rotation and Minres as a method of estimation, followed by a CFA in which all the EFA secondary loadings >0.30, alongside the target loadings, were estimated (called complex-CFA model in the following)
In the second sample, we cross-validated the model emerged from the analyses of the data from the first sample.
All analyses were conducted using SPSS 24; R 3.4; Prelis and LISREL 8.72 (Jöreskog and Sörbom,
The theoretical 13 factor model showed good overall fit to the data as all the fit indices, RMSEA, CFI, and SRMR, fulfiled the cut-off values (Table
Goodness of fit of the CFA models on Sample 1.
| 13 factor ICM-CFA | 4186.8 | 2406 | 0.042 (0.040–0.044) | 0.95 | 0.067 | |||
| 13 factor Complex-CFA model | 3635.8 | 2376 | 389.9 | 30 | <0.0001 | 0.036 (0.033–0.038) | 0.96 | 0.063 |
| 12 factor Complex-CFA model | 3694.9 | 2392 | 54.5 | 16 | <0.0001 | 0.036 (0.034–0.038) | 0.96 | 0.064 |
Based on the results of the explorative target rotated factor analysis reported in
In the light of these results, a third model with 12 latent variables in which scales D and F belongs to the same latent factor was estimated, in order to assess whether the two highest correlated latent factors could be collapsed. As shown in Table
In order to gain further understanding of, and further clarify, the factor structure of the CA-MIR, we replicated the complex-CFA model in the second sample.
The overall fit statistics were good: SB-Chi-Square (2376) = 4130.6,
Complex-CFA model: Standardized parameter estimates (Sample 2,
| 7 | - | – | – | – | – | – | – | – | – | – | – | ||
| 35 | – | – | – | – | – | – | – | – | – | – | – | ||
| 39 | – | – | – | – | – | – | – | – | – | – | – | ||
| 48 | – | – | – | – | – | – | −0.07 | – | – | – | – | – | |
| 54 | – | – | 0.03 | – | – | – | – | – | – | – | – | – | |
| 62 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 2 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 26 | – | – | – | – | – | – | – | – | – | – | – | ||
| 41 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 52 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 55 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 64 | – | – | – | – | – | – | – | – | – | – | – | ||
| 20 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 22 | – | – | – | – | – | – | – | – | – | — | – | – | |
| 32 | – | – | – | – | – | – | – | – | – | – | – | ||
| 56 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 68 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 72 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 9 | – | – | – | – | – | −0.14 | – | – | – | – | – | – | |
| 21 | – | – | – | – | 0.11 | – | – | – | – | – | – | – | |
| 40 | – | – | – | – | – | – | – | – | – | – | – | ||
| 53 | – | – | – | – | – | – | – | – | – | – | – | ||
| 58 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 66 | – | – | – | – | – | – | – | – | – | – | – | ||
| 10 | – | – | – | – | 0.14 | – | – | – | – | – | – | – | |
| 11 | – | – | – | – | – | – | – | – | – | – | – | ||
| 19 | – | – | – | – | – | – | – | – | – | – | – | ||
| 25 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 28 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 6 | – | – | – | – | 0.12 | – | – | – | – | – | – | – | |
| 1 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 4 | – | – | – | – | – | – | – | – | – | – | – | ||
| 18 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 27 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 36 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 69 | – | – | – | – | – | – | – | 0.00 | – | – | – | – | |
| 15 | – | – | – | – | – | – | 0.03 | – | – | – | – | – | |
| 29 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 30 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 31 | – | – | – | – | – | – | – | – | – | – | – | ||
| 38 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 71 | – | – | – | – | – | – | – | – | – | – | – | ||
| 13 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 47 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 50 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 57 | – | – | 0.18 | – | – | – | – | – | – | – | – | – | |
| 60 | – | – | – | – | – | – | – | – | – | – | – | ||
| 67 | – | – | – | – | – | – | −0.01 | – | – | – | – | – | |
| 12 | – | – | – | – | – | – | – | – | – | – | – | ||
| 14 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 17 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 3 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 33 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 45 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 59 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 61 | – | – | – | – | – | – | – | – | −0.02 | – | 0.16 | – | |
| 63 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 37 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 46 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 51 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 5 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 16 | – | – | – | – | – | – | – | – | – | – | – | ||
| 23 | – | – | – | – | – | – | −0.05 | – | – | – | – | – | |
| 42 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 44 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 70 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 8 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 24 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 34 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 43 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 49 | – | – | – | – | – | – | – | – | – | – | – | – | |
| 65 | – | – | – | – | – | – | – | – | – | – | – | – | |
Items 7, 35, 39 resulted to be poor indicators of factor A (Preoccupied-Past); their loadings, albeit statistically significant, were low in value. More specifically, item 7 (“I' d like my children to be more autonomous than I have been”) and item 35 (“My parents didn 't really realize that a child who is growing up needs to have a life of his own”) seemed to be indicators of the State of Mind dimension instead of the Past, whereas item 39 (“As a child, I was scared of being abandoned”) loaded heavily (and negatively) on Autonomous-State of Mind factor.
Items 26 (“As a teenager, no one around me has ever really understood my worries”) and 64 (“In my family, we lived isolated from the rest of the world”) loaded both on the expected factor (scale C), and on a second factor: factor A (item 26), and factor I (item 64).
As regard autonomy factors (D, E, and F), item 40 (“As a child, I was encouraged to share my feelings”) loaded also on scale C, but with negative sign, and item 53 (“Even if it's not true, I have the feeling l've had the best parents in the world”) showed a secondary positive loading on factor F. Moreover, item 18 (“I spend a lot of time talking to the people who are close to me”) was a poor indicator of factor E (Autonomous-Present), without showing relevant secondary loadings, and item 10 (“I believe l've known how to give back to my parents the love they've given me”) was more an indicator of the Past than of the State of Mind dimension.
Moving to the Dismissing dimensions (factors G, H, and I), item 15 (“When I was a child, the people close to me were often impatient and irritable”) resulted to be related to the J factor instead of the G factor and item 31 (“When I was a child, it was difficult for us to make decisions in the family”) to factor L. As regard item 12 (“I hate feeling I depend on others”), an unexpected negative secondary loading on factor I was observed.
In the remaining four scales (factors J, K, L, and M), only two items showed poor fit: item 61 (“I have the feeling I was a rejected child”), that resulted to be an indicator of factor I instead of factor J, and item 24 (“Adults have to control their emotions with their children, whether it concerns pleasure, love or anger”), that emerged as a weak indicator of scale M without showing any secondary loadings.
In summary, four items (17, 18, 24, and 44) were weak indicators of the intended factor and did not load on secondary factors. Others four items (10, 15, 61, and 6) did not load on the expected factor but on a plausible related factor; and finally, five items (64, 11, 19, 71, and 12) resulted in a substantial loading (>0.30) on two factors.
As reported in Table
Complex-CFA model: factors correlations (Sample 2,
| A | 1 | ||||||||||||
| B | 1 | ||||||||||||
| C | 1 | ||||||||||||
| D | 0.12 | 1 | |||||||||||
| E | 0.12 | 1 | |||||||||||
| F | 0.09 | 1 | |||||||||||
| G | 0.20 | 1 | |||||||||||
| H | −0.01 | 1 | |||||||||||
| I | 0.03 | 1 | |||||||||||
| J | 0.04 | 1 | |||||||||||
| K | 0.04 | 0.13 | 0.10 | −0.01 | 1 | ||||||||
| L | 0.15 | 0.14 | 1 | ||||||||||
| M | 0.14 | −0.10 | 1 | ||||||||||
With regard to Scales J, K, L, and M, the only correlations above 0.30 were that between J and L (0.38), and L and M (−0.36). Factors J and L were also strongly correlated with the attachment style scales (especially the Autonomous and Dismissing scales), whereas scales K and M were those least strongly correlated with all the other scales.
In the evaluation of the factorial structure of the CA-MIR, both the ICM-CFA model and a model with the presence of several secondary loadings were estimated. The former was aimed at directly assessing the goodness of fit with the Italian data of the 13-factor theoretical model; the latter, here called complex-CFA model, was considered to eventually take into account the presence of item multidimensionality (Marsh et al.,
The comparison of the two above-mentioned models in the first sample showed that several secondary loadings were present. The same model, replicated in the second sample, confirmed the presence of some multidimensionality: in a few circumstances, secondary loadings were greater than the loadings in the expected factors and others few items showed not negligible cross-loadings. The content of these unfitting items justifies these findings: in most cases, the items ambiguity was connected to the distinction between Past and State of mind scales of the same attachment style or to a negative relation with a scale of a different attachment style (indicators of Autonomous style with negative loadings on one of the unsecure styles, and vice versa). Four items emerged as poor indicators of the expected latent dimension, without exhibiting secondary loadings. One of these, item 17, belongs to the
The internal consistency of each of the remaining scales was satisfactory and comparable to that found in the original Swiss validation study. Only
The CFA approach allowed us to directly test and confirm the 13-factor model and contributes to advancing our understanding of why exploratory tool like that applied in Garrido et al. (
If we examine the correlations among the scales in light of the theoretical model, we find that the CA-MIR can differentiate between the three main styles of attachment. The signs of the correlation coefficients are theoretically coherent: specifically, the three scales associated with autonomous attachment are negatively correlated with those assessing preoccupied and dismissing attachment, which in turn are positively correlated with one another. The last-mentioned result invites further analysis: indeed, the relationship between the two insecure styles is puzzling and requires cautious interpretation.
Within each attachment style, the correlations among individual scales are positive, but vary in strength.
The correlations between the dimension of the
The last four Scales concern two further constructs.
Scales J and K are theoretically related to traumatic experience and are designed to tap into
Scales L and M assess respondents' general views about the role of parents and the importance of authority: theoretically, these scales are not linked to attachment but to “ideological”/“abstract” assumptions about parents and adults. In the current dataset, these scales clearly loaded onto distinct factors, and the pattern of correlations confirms their independent status.
Scales L and M were moderately and negatively correlated, as expected: the former concerns the absence of parental discipline in respondents' childhood experience, and the latter, on the contrary, their views regarding the importance of authority.
Scales J (
The Scale L correlations are lower, but in the same direction: this provides support for the idea that “sensitive discipline” (Van Zeijl, et al., 2006) is a key aspect of secure experience in childhood.
Scale M (
This study is part of a broader research project aimed at completing Italian standardization of the CA-MIR. The objective of the current study was to evaluate the factor structure of the instrument. Indeed, the CA-MIR is an interesting tool precisely because of the richness and complexity of its theoretical model. Nevertheless, this model had never been tested with CFA, whereas previous exploratory analysis had only partially reproduced the original theoretical model (Garrido et al.,
Our CFA enabled us to confirm the originally proposed structure of the CA-MIR: when used with its Likert response format it adequately differentiates between the three attachment styles, taking into account both past and present experiences of attachment relationships, and providing a complex and rich assessment of multiple features of attachment.
From a cross-cultural point of view, our results support the original theoretical model in a novel cultural context, namely Italy. The importance of this generalization is linked to one of the core hypotheses of the attachment theory: the broad cross-cultural stability of attachment representations. Therefore, the 13 scores obtained by administering the CA-MIR may be taken as valid indexes providing key information about subjective experience and conscious representations of attachment.
Taken as a whole, the Italian internal consistency values are acceptable and comparable to those obtained in the original Swiss validation study. As mentioned earlier, Scale H displays a number of weakness, including both the low number of items composing the scale (three items vs. the six items forming most of the other scales) and the formulation of Item 17. The last-mentioned item (“
The outcomes of the correlation analysis provide some pointers regarding appropriate interpretation of CA-MIR scores. Firstly, we must be cautious in differentiating between respondents' evaluations of past experience and their states of mind about attachment: these two dimensions are strongly correlated for all attachment styles, and individual cases of difference between these scores must be carefully assessed. The dimension of present experience is relatively independent of past experience and state of mind, particularly in relation to dismissing and preoccupied attachment styles, and this is in line with the theoretical expectations. Furthermore, due to the strong correlation between dismissing and preoccupied attachment scales, we must be cautious in viewing them as distinct.
Scales J and K are not correlated and must therefore be taken as two separate indicators rather than as forming a single scale of unresolved attachment, but they can nonetheless be of assistance in interpreting a respondent's clinical profile. Although the L and M scales are not envisaged as directly linked to attachment profiles, they offer useful additional information that can help to build up clinical profiles, e.g., when evaluating parenting skills.
These characteristics, combined with the advantages of the self-report format, make the CA-MIR a useful and powerful instrument, capable of discriminating between representations of secure and insecure attachment, and of providing information about attachment representations that is of both clinical and research value.
The sampling procedure adopted in the current study is one of its limitations. We chose a snowball sampling method in order to recruit a large sample. However, starting out from a northern region and a young adult population, this procedure reduced the representativeness of the sample because of the high proportion of participants from North-western Italy and aged between 20 and 29 years. Nevertheless, we have provided a detailed comparison of the demographic characteristics of our sample with those of the general Italian population, highlighting both differences and similarities, to facilitate contextualization of our findings. Although not completely representative, our sample is comparable to that of the original Swiss-led validation study, and its principal limitation (the over-representation of young adults) may also be viewed as an interesting feature, given that individuals in this age range typically face the key attachment challenges of adult life such as developing romantic relationships, growing more independent of their parents and becoming parents themselves. More in-depth analysis of socio-cultural variables is one of our future research objectives. Moreover, sample size was not big enough to employ the ESEM approach. The comparison between the ICM-CFA and complex-CFA models showed that some “construct relevant item multidimensionality” (Morin et al.,
In addition, an interesting direction for future research could be to test the concurrent and discriminant validity of the Italian version of the test, by comparing the CA-MIR with other instruments assessing adult attachment. In particular, it would be useful to compare CA-MIR self-report data with outcomes obtained using narrative instruments, such as the AAI. Comparison of AAI categories with CA-MIR scores would be a further test of the CA-MIR's effectiveness in investigating attachment representations, and can be useful to further analyse the behavior of the few unfitting items observed in this study. Finally, the instrument could be administered to clinical groups in order to assess its ability to detect specific aspects of insecure attachment in particular sub-samples, as has already been done in previous studies (Molina et al.,
PM: Conceived the study and wrote the manuscript; MC and MNS: Contributed to data collection, to analyze the data and to revise the manuscript; ST: Analyzed the data and contributed to revise the manuscript.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The Supplementary Material for this article can be found online at:
1In terms of convergent validity, Roskam et al. (
2Given that we had not followed any particular criterion for order of data entry, we removed from the data set the first subject entered for each interrelated pair, the first and third subjects entered for interrelated groups of three or four and the first, third and fifth subjects entered for groups of five.
3To enable comparison with the general population, we followed the ISTAT (National Institute of Statistics) definition of family unit, taking into account participants who were married, cohabiting, divorced or widowed.
4We checked the equivalence of the two groups, finding no significant differences in relation to the following variables: age, gender, highest level of education completed, birthplace, marital status, parenthood status, issues encountered during development, issues encountered recently and semester in which the instrument was completed (all questionnaires were administered during the same calendar year, 2005, but across two different academic years: 2004/2005—II semester, and 2005/2006—I semester).
5The original French version and the Italian and English versions of the questionnaire are available at
6We tested models with less liberal thresholds (0.29, 0.25) but they resulted in some items with suspiciously high standard error of estimates compared to the others.
7In presence of ordinal data, polychoric correlations and WLS or DWLS estimators are usually suggested. As a control analysis we performed a DWLS estimation based on polychoric correlations on Sample 1 obtaining comparable values for global fit indices but some Heywood cases emerged, whereas MLR resulted in proper solutions; for this reason we considered the MLR results as more reliable.