You are on page 1of 8

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/251531368

The factor structure of the Inventory of Parent and Peer Attachment (IPPA): A
survey of Italian adolescents

Article  in  Personality and Individual Differences · July 2011


DOI: 10.1016/j.paid.2011.03.006

CITATIONS READS

49 1,051

3 authors:

Cecilia Serena Pace Pietro San Martini


Università degli Studi di Genova Sapienza University of Rome
37 PUBLICATIONS   277 CITATIONS    37 PUBLICATIONS   283 CITATIONS   

SEE PROFILE SEE PROFILE

Giulio Cesare Zavattini


Sapienza University of Rome
119 PUBLICATIONS   577 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Risk path-ways in adolescence: attachment, alexithymia and emotional regulation in late-adoptees, residential-care and community teenagers View project

All content following this page was uploaded by Cecilia Serena Pace on 21 May 2018.

The user has requested enhancement of the downloaded file.


This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy

Personality and Individual Differences 51 (2011) 83–88

Contents lists available at ScienceDirect

Personality and Individual Differences


journal homepage: www.elsevier.com/locate/paid

The factor structure of the Inventory of Parent and Peer Attachment (IPPA):
A survey of Italian adolescents
Cecilia Serena Pace ⇑, Pietro San Martini, Giulio Cesare Zavattini
La Sapienza, University of Rome, Department of Dynamic and Clinical Psychology, Italy

a r t i c l e i n f o a b s t r a c t

Article history: Despite the intensive use of the Inventory of Parent and Peer Attachment (IPPA), a measure of attachment
Received 22 October 2010 which centres on individual differences in adolescents and their ability to evaluate their relationships
Received in revised form 18 February 2011 with parents and peers, the factor structure of the inventory requires further investigation. This study
Accepted 5 March 2011
of 1059 Italian adolescents compared the three models which are discussed in the literature: the one-fac-
Available online 6 April 2011
tor model (attachment security), the two-factor model (trust–communication and alienation) and the
three-factor model (trust, communication and alienation), and examined the influences of age and gender
Keywords:
on the IPPA’s scores. Our findings provide support for the reliability of the latest and longer version of the
Attachment
Adolescents
IPPA (75-item). Factor analysis showed that the three-factor model had the best fit, although the three
Parent dimensions are strongly interrelated. Sixteen year olds had lower attachment security to their fathers
Peer than the results of the other adolescent age groups. Males reported lower alienation scores than females
Factor structure in the paternal form, while females had higher attachment security, trust and communication scores than
males in the peers form.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction from their parental figures and allow them to form an adult
identity, but on the other hand, these changes will depend on
Attachment theory (Bowlby, 1988) advocates that children de- the adolescent’s personal history of attachment relationships (Al-
velop internal representations of attachment relationships, called len & Land, 2008). In fact, during adolescence, peer group and
internal working models (IWMs), which are formed through their friends become increasingly important and there is an overlap be-
everyday interactions with their caregivers during early child- tween new relationships with friends and previous relations with
hood. The IWMs contain expectations and strategies for managing parents.
interpersonal relationships and exploratory behaviour, regulating In order to investigate this issue further, therefore, the IPPA
emotions and dealing with distress. Differences in the quality of (Armsden & Greenberg, 1987, 1989; Greenberg, Siegel, & Leitch,
child–parent affective bonds correspond to individual differences 1984), a self-report questionnaire designed specifically for teenag-
in the IWMs of Self and Other. Secure attachments foster the ers that assesses their relationships with both their parents and
development of models in which others are viewed as available their peers, seems to be a useful method of evaluation that in re-
and trustworthy, while the self is conceptualised as worthy of cent years has been increasingly used in international research
care, love and attention. Insecure attachments develop IWMs of (Baiocco, Laghi, & Paola, 2009;Wilkinson & Walford, 2001).
the self as unworthy and unlovable, while others are considered
as unavailable and/or unreliable. According to Bowlby’s evolution-
ary theory, these patterns have been shown to be moderately sta- 1.1. Brief history of IPPA’s factor structure
ble over long periods of time under stable family and caregiving
conditions. The IWMs are viewed as the main source of continuity The IPPA was developed for adolescents aged between 12 and
between attachment in infancy and attachment in adolescence 19-years-old in order to assess their perceptions of the positive
and adulthood. and negative affective/cognitive dimensions of their relationships
Adolescence is a peculiar period in the life cycle: on the one with their parents and close friends, investigating how well these
hand, developments and changes in IWMs distance adolescents figures serve as sources of psychological security.
The original version (Greenberg et al., 1984) contained two
forms of assessing attachment security towards parents (28 items)
⇑ Corresponding author. and peers (25 items). Later, Armsden and Greenberg (1987),
E-mail address: ceciliapace@fastwebnet.it (C.S. Pace). arguing that one factor was insufficient to describe the complex

0191-8869/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.paid.2011.03.006
Author's personal copy

84 C.S. Pace et al. / Personality and Individual Differences 51 (2011) 83–88

construct of attachment, expanded the number of items in both the Vivona (2000) suggested using the three sub-scales in order to as-
parental and peers forms to 31 and 29, respectively and explored sess individual differences according to the following attachment
their factor structures through exploratory factor analysis (EFA). categories:
The sample consisted of 179 college students aged between 16
and 20 years (mean age: 18.9 years). They expected to find two fac- (a) Secure attachment, when both trust and communication
tors, one related to a ‘‘positive affective/cognitive experience of levels are medium or high and the alienation level is med-
trust in the accessibility and responsiveness of attachment figures’’ ium or low. Alienation scores should always be lower than
and one linked to ‘‘negative affective/cognitive experiences of an- those of trust and communication;
ger and/or hopelessness resulting from unresponsive or inconsis- (b) Insecure-avoidant attachment, when the trust level is med-
tently responsive attachment figures’’ (Armsden & Greenberg, ium or low, the communication level is low and the alien-
1987, p. 431). However, using the criterion of eigenvalues greater ation score is high;
than 1, the authors extracted (and rotated orthogonally) three fac- (c) Insecure-ambivalent attachment, when the trust level is
tors in both the parental and peer forms. The factor structures of medium or low and the communication and alienation
the two forms, which were quite similar, were interpreted as psy- scores are medium or high.
chologically meaningful. For the parental form, the first factor
(trust) was interpreted in terms of ‘‘parental understanding and re-
spect and mutual trust’’, the second factor (communication) in 1.2. Current study
terms of ‘‘the extent and quality of verbal communication with
parents’’ and the third factor (alienation) in terms of ‘‘feelings of The current study aims to assess the following points in a sam-
alienation and isolation’’. Similarly, in the peer form, the first factor ple of adolescents:
(trust) was interpreted as ‘‘mutual respect and trust’’, the second
(communication) in terms of ‘‘perceived quality of communica- 1. The dimensional structure of the inventory using preliminarily
tion’’ and the third (alienation) as ‘‘alienation from friends, but EFAs of the three forms followed by CFAs of the three models
with the recognition of the need to be closer to them’’ (Armsden explicitly proposed or implied in the literature: the one-factor
& Greenberg, 1987, p. 433). The subscales were also strongly corre- model (attachment security), the three-factor model (trust,
lated both in the parental (|r| between .70 and .76) and peer forms communication and alienation) and the two-factor model
(|r| between .40 and .76), so that the authors themselves in further (trust–communication and alienation);
analyses used only the overall attachment security scores, which 2. The main descriptive statistics of the IPPA scores along with the
were obtained for each form by adding up the trust and communi- effects of age and gender, in order to confirm gender differences
cation scores and subtracting the alienation score. in the attachment styles as reported by Gullone and Robinson
Later, Armsden and Greenberg (1989) proposed a revised ver- (2005) and provide norms for the interpretation of the scores
sion of the inventory, dividing the parental form into maternal of individual subjects.
and paternal forms that were identical except for references to
the mother or father. This version contains 75 items, equally di-
2. Method
vided into the three forms for mother, father and peers. Greenberg
(2004) recently recommended using this version of the inventory
2.1. Participants
as it differentiates between the roles of each parent in addition
to relationships with peers.
The participants consisted of 1059 volunteers; 574 females and
In recent years, Johnson, Ketring, and Abshire (2003), as part of
402 males (83 participants did not report their gender). They were
a study mainly aimed at creating a form for parents only, tested the
middle or high school Italian students, ranging in age from 13 to
responses of a small sample of adolescents (N = 89, mean age:
18 years old (mean = 15.66, SD = 1.59). All participants belonged
14.3 years) to the paternal and maternal forms of the IPPA using
predominately to middle to upper socio-economic status and lived
confirmatory factor analysis (CFA), finding a bad fit for the three-
with both parents.
factor model in both cases. A subsequent EFA, using a scree test
to determine the number of factors and the oblique rotation of
the factor axes, found for both forms two factors that the authors 2.2. Instrument
interpreted as trust (mainly including items originally labelled as
trust and communication) and alienation. This study utilised the latest version of the IPPA (Armsden and
In another study, which aimed to validate a short French form of Greenberg ,1989), which comprises three forms for mother, father,
the IPPA (participants N = 289, mean age: 14.6 years), Vignoli and and peer, each consisting of 25 items. In each form, the item format
Mallet (2004) conducted EFAs on the items in the paternal and was a five-point Likert scale: 1 (almost always or always true), 2
maternal forms, imposing on both analyses a three-factor solution (often true), 3 (sometimes true), 4 (rarely true) and 5 (almost never
with orthogonal rotation. In both forms, the extracted factors ap- or never true). Each form yielded an overall score for attachment
peared to correspond only partially to the three subscales proposed security as well as three subscale scores: trust (example item ‘‘I
by Armsden and Greenberg (1987). trust my mother/father/friends’’), communication (example item
As Armsden and Greenberg (1987) and Johnson et al. (2003) ‘‘I tell my mother/father/friends about my problems and troubles’’)
only used a small number of subjects in their studies, it is impos- and alienation (example item ‘‘I get upset a lot more than my
sible to choose one of the proposed factor solutions, suggesting mother/father/friends knows about’’).
that further investigation using a larger sample may be useful. To
date it still remains unclear whether it would be correct or not
to distinguish one, two or three dimensions within the global con- 2.3. Procedure
struct of attachment, considering that the original purpose of the
IPPA was to capture the complexity of this construct by identifying The data were collected in classroom settings with the agree-
its different facets. ment of the teachers. Students were told that the questionnaires
Furthermore, the issue of the factorial justification of the three- were anonymous and confidential, so they could feel free to re-
factor model is relevant since both the authors of the IPPA and spond sincerely to the questions.
Author's personal copy

C.S. Pace et al. / Personality and Individual Differences 51 (2011) 83–88 85

3. Results the range of .08–.10 suggest a ‘‘mediocre’’ model fit (MacCallum,


Browne, & Sugawara, 1996). CFI values in the range of .90–.95
3.1. Exploratory factor analysis (EFA) are generally considered to be indicative of an acceptable model
fit (e.g. Bentler, 1990).
To assess the factor structure of the inventory, three EFAs, using Among these indices, we considered the RMSEA to be particu-
principal axis factoring (PAF) as the method of extraction, were larly relevant because it compensates for the effect of the model’s
preliminarily conducted on the items in the three forms. complexity by conveying the discrepancy in fit for each degree of
The three uni-factorial solutions (Table 1) showed acceptable freedom in the model. As a rough guideline, for the difference in
levels of accounted variance and substantial loadings (r > .30) with fit to be sufficiently large to reject a more parsimonious model in
the expected signs for all the items on the extracted factors. Only favour of a more complex model, it has been suggested that there
two items on the peer form belonging to the alienation subscale must be a change in RMSEA of less than .015 (Chen, 2007). A more
did not present high loadings: these were item 22 (‘‘I get upset a conservative guideline is that a more parsimonious model is sup-
lot more than my friends know about’’, r = .27) and item 9 (‘‘I feel ported if the RMSEA is as good (or better) than that of the more
the need to be in touch with my friends more often’’, r = .15). complex model (Marsh, 2007). We used also the Akaike informa-
tion criterion (AIC), a fit index which also takes into account the
parsimony of the model and that is often used in the comparison
3.2. Comparing the factor models: confirmatory factor analysis (CFA)
of two or more non-nested models, with smaller values represent-
ing a better fit of the hypothesised model.
The results cited above justify the use of global attachment
The results for the maternal form are reported in Table 2a. The
security scores for the three forms of the inventory, as most
three-factor model showed the best fit, with a CFI close to .90 and
authors have done so. They do not, however, deny that dividing
an ‘‘adequate’’ RMSEA, and according to the Chen criterion it was
the construct of attachment into two (Johnson et al., 2003) or three
clearly preferable to the one or the two-factor models which did
(Armsden & Greenberg, 1987) correlated sub-dimensions may ex-
not show a clear-cut difference in fit. However, the internal consis-
plain the correlations between the items better than the uni-facto-
tencies (a) of the model were high (>.80), except for alienation and
rial model. We decided therefore to also adopt a confirmative
the latent dimensions of the three-factor model were highly corre-
perspective and to compare the three models proposed in the
lated: communication vs. trust: r = .84; communication vs. alien-
literature.
ation: r = .72; trust vs. alienation: r = .81, as they were in the
The fit of the models were assessed with the v2 reflecting the
two-factor model (trust–communication vs. alienation r = .80).
absolute discrepancy between actual and model-generated data,
This suggests that the sub-dimensions are poorly differentiated,
and also by the root mean square error of approximation (RMSEA),
casting some doubt on their practical usefulness.
the comparative fit index (CFI). These indices are based on the non-
The results for the paternal form were similar to those of the
centrality parameter; that is, on the assumption that the model
maternal form (Table 2b): once again, the three-factor model ap-
may only provide a reasonable approximation of the population
peared to be the most appropriate, with the RMSEA at the upper
covariance matrix rather than a perfect reproduction. This assump-
limit of the ‘‘adequate range’’ and the CFI just below .90. The other
tion appears to be particularly realistic in the field of personality,
two models presented considerably worse fits with decidedly low-
where it is easy to argue that any model is probably incorrect even
er CFI and RMSEA values which were beyond acceptability. Accord-
before it is fitted to the data (Raykov, 1998). RMSEA values less
ing to the Chen criterion, the three-factor model was clearly
than .08 and .05 suggest an ‘‘adequate’’ and ‘‘close’’ model fit,
preferable to the one and the two-factor models which did not
respectively (Marsh, Hau, & Wen, 2004), while RMSEA values in
show a clear-cut difference in fit. The reliability of the scales and
subscales was high, except for alienation. The latent dimensions
Table 1 of the three and two-factor model were highly correlated: commu-
EFA: matrixes of loadings of the uni-factorial solutions. nication vs. trust: r = .82; communication vs. alienation: r = .87;
trust vs. alienation: r = .84, trust–communication vs. alienation:
Item Mother form Father form Item Peer form
r = .86.
trust1 .702 .725 com1 .461
Also for the peer form (Table 2c) the fit of the three-factor mod-
trust2 .669 .761 com2 .656
trust3(r) .624 .657 com3 .598 el was substantially better than the fit of the one-factor or the two-
trust4 .571 .626 alien4 .371 factor models. The gain in fit of the three-factor model was greater
com5 .582 .664 trust5(r) .590 than .015, suggesting that it was preferable to the one-factor mod-
com6(r) .379 .446 trust6 .683 el. High levels of internal consistency were obtained, except for
com7 .624 .681 com7 .654
alien8 .448 .442 trust8 .636
alienation. The correlations between the latent dimensions of the
trust9(r) .378 .328 alien9 .155 three-factor model were rather high, except between communica-
alien10 .514 .598 alien10 .412 tion and alienation (r = .61): communication vs. trust: r = .91;
alien11 .400 .460 alien11 .516 trust vs. alienation: r = .77. The trust–communication vs. alien-
trust12 .680 .705 trust12 .682
ation correlations (r = .72) in the two-factor model was lower
trust13 .631 .667 trust13 .738
com14(r) .343 .364 trust14 .696 than in the other two forms, but was still higher than desirable.
com15 .702 .723 trust15 .713
com16 .678 .684 com16 .603 3.3. Descriptive statistics
alien17 .588 .655 com17 .663
alien18 .482 .630 alien18 .484
com19 .760 .729 trust19 .693
Table 3 presents the descriptive statistics of the overall scores
trust20 .810 .810 trust20 .713 for attachment security and the subscale scores for trust, commu-
trust21 .703 .730 trust21 .717 nication and alienation, given separately for males and females.
trust22 .664 .693 alien22 .271 It may be noted that the skewness and kurtosis indices were
alien23 .473 .464 alien23 .488
quite close to normal (<1.5), with the exception of the trust sub-
com24 .733 .754 com24 .723
com25 .600 .663 com25 .594 scale in the peer form which showed a marked positive kurtosis,
especially for females. Overall, the mean values were sufficiently
Acc. Variance % 38.54 43.02 38.22
far from the extremes of the scales and the SDs were large enough.
Author's personal copy

86 C.S. Pace et al. / Personality and Individual Differences 51 (2011) 83–88

Table 2
CFA of the maternal, paternal and peer forms: v2, RMSEA, CFI and AIC of the one-factor, two-factor and three-factor models.

Model d.f. v2 RMSEA RMSEA CFI AIC


accurate estimate confidence interval (90%)
2a-Maternal form
One-factora 275 2119 .094 .091–.097 .839 2.719
Two-factorb 274 1932 .089 .086–.092 .855 1.935
Three-factorc 272 1472 .069 .066–.073 .895 1.502
Reliability index (Cronbach’s alpha)
Attachment security Trust–com Trust Com Alien
a = .93 a = .92 a = .88 a = .86 a = .70
2b-Paternal form
One-factora 275 2754 .115 .112–.118 .821 2.721
Two-factorb 274 2506 .110 .107–.113 .839 2.485
Three-factorc 272 1786 .080 .077–.083 .891 1.805
Reliability index (Cronbach’s alpha)
Attachment security Trust–com Trust Com Alien
a = .94 a = .93 a = . 90 a = .88 a = .77
2c-Peer form
One-factora 275 2012 .091 .088–.094 .843 2.012
Two-factorb 274 1716 .080 .077–.083 .870 1.732
Three-factorc 272 1543 .071 .068–.074 .885 1.571
Reliability index (Cronbach’s alpha)
Attachment security Trust–com Trust Com Alien
a = .92 a = .93 a = .90 a = .87 a = .65
2a, attachment security (25 items); 2b, trust–communication (19 items) and alienation (6 items); 2c, trust (10 items), communication (9 items) and alienation.

Table 3
Means, SD, skewness and kurtosis of the scores for attachment security, trust, communication and alienation.

Females Males
Mean S.D. Skewness Kurtosis Mean S.D. Skewness Kurtosis
Maternal form
Att. security 3.83 0.75 .93 .39 3.82 0.62 .76 .58
Trust 4.00 0.80 1.23 1.12 4.06 0.68 1.13 1.11
Communication 3.24 0.81 .67 .28 3.15 0.71 .46 .06
Alienation 1.60 0.56 .75 .27 1.57 0.48 .71 .94
Paternal form
Att. security 3.46 0.85 .62 .15 3.56 0.73 .68 .52
Trust 3.83 0.93 1.09 .63 3.86 0.83 1.24 1.49
Communication 2.71 0.86 .09 .69 2.84 0.78 .20 .37
Alienation 1.89 0.67 .57 .13 1.74 0.57 .69 .45
Peer form
Att. security 4.04 0.62 1.15 1.38 3.80 0.62 .69 1.01
Trust 4.22 0.73 1.41 2.16 4.00 0.73 1.07 1.57
Communication 4.05 0.73 .94 .72 3.59 0.77 .38 .12
Alienation 2.21 0.63 .92 1.28 2.61 0.68 .84 1.13

3.4. Effects of gender and age on the scores for the attachment security
Table 4
scale
Means and SDs for attachment security for the maternal, paternal and peer forms at
different ages.
To evaluate the effects of age and gender on the attachment
Age Maternal form Paternal form Peer form
security scores of the three forms of the inventory, three two-factor
variance analyses (ANOVAs) were conducted. P-values <.01 were Mean S.D. Mean S.D. Mean S.D.

considered statistically significant in all the analyses. 13 3.86 0.62 3.63 0.78 3.91 0.66
The ANOVA for the maternal form did not indicate any signifi- 14 3.94 0.59 3.60 0.75 3.99 0.64
15 3.87 0.71 3.60 0.77 3.98 0.61
cant effects of age by gender interaction (F5,955 < 1), age
16 3.71 0.79 3.27 0.95 3.97 0.62
(F5,955 = 2.16, p = .056), or gender (F1,955 < 1). 17 3.77 0.69 3.52 0.77 3.94 0.63
The ANOVA for the paternal form showed no effects of age by 18 3.79 0.74 3.38 0.76 3.83 0.62
gender interaction (F5,954 = 1.13, p = .339) or gender (F1,954 = 2.70,
p = .100), but a significant effect of age (F5,954 = 4.31, p = .001,
g2 = .022). Post-hoc comparisons (Tukey test) showed that this ef-
fect was due to the group of 16 year olds scoring significantly low- 3.5. Effects of gender and age on the scores for the trust,
er than all the other age groups except the 18 year olds (Table 4). communication and alienation subscales
The ANOVA for the peer form revealed no effects of age by gen-
der interaction (F5,955 = 1.07, p = .373) or age (F5,955 = 1.23, The effects of gender and age on the subscale scores were exam-
p = .293), but a significant effect of gender (F1,955 = 34.02, p = .000, ined by multivariate analysis of variances (MANOVAs) followed by
g2 = .034), with males scoring lower than females (Table 3). univariate analyses of the multivariate significant effects.
Author's personal copy

C.S. Pace et al. / Personality and Individual Differences 51 (2011) 83–88 87

The MANOVA for the maternal form showed no significant ef- which to develop a more complex multidimensional inventory’’
fects of age by gender interaction or age, and a significant effect (p. 251).
of gender (F3,953 = 4.310, p = .005). However, the ANOVAs of the last Age had no effect on IPPA scores, except for 16-year-old adoles-
effect did not reveal any significant effects, either for trust (F1,965 < 1), cents who showed an overall global score of attachment security
communication (F1,965 = 3.07, p = .080) or alienation (F1,965 < 1). towards their fathers which was lower than for the younger and
The MANOVA for the paternal form presented no significant ef- older participants. We do not have a reasonable explanation for
fects of age by gender interaction and age, but a significant gender this result and will not discuss it further.
effect (F3,952 = 6.53, p = .000). The ANOVAs of the latter effect re- In terms of gender, two effects were observed. Males scored
vealed that alienation had a significant gender effect lower than females on the alienation subscale of the paternal form.
(F1,954 = 7.618, p = .006, g2 = .008) with a higher mean score for fe- This finding may reveal that males develop less problematic rela-
males (see Table 3), and no significant effect of communication tionships with their fathers than females, who could be more in-
(F1,954 = 5.43, p = .020) or trust (F1,954 < 1). volved in struggles for independence and autonomy. On the
The MANOVA for the peer form showed no significant effects of other hand, males, compared to females, appeared to be more inse-
age by gender interaction and age, but a significant effect of gender cure around their peers, particularly in terms of the trust and com-
(F3,953 = 35.08, p = .000). The ANOVAs of the latter effect revealed munication subscales. These results seem to be consistent with the
significant effects of communication (F1,955 = 78.71, p = .000, results of Gullone and Robinson (2005), indicating that males re-
g2 = .076) and trust (F1,955 = 17.75, p = .000, g2 = .018), with a high- ported more positive attachments to their parents than females,
er mean score for females for both (see Table 3). No significant ef- while females reported more positive attachments to their peers
fect of alienation emerged (F1,955 < 1). compared with males. One possible interpretation is that females,
during adolescence, tend to disengage earlier from parental bonds
4. Discussion and conclusions and invest more in their relationships with their friends.
In conclusion the findings from this study give further support
In terms of the dimensional structure of the three forms of the to the factorial validity and the reliability of the questionnaire
inventory, the CFAs show that the model best fitted to the data is when it is used to evaluate overall attachment security. As for
the model with three correlated dimensions. This is the conceptual the assessment of the three sub-dimensions originally proposed
equivalent to a hierarchical model with three first order factors by Armsden and Greenberg (1987), however, our results suggest
functionally dependent on a second order factor and supports both to reduce the excessive correlations between the three subscales
the use of the overall scores for attachment security and the sub- by searching for other items more specifically linked to the con-
scale scores for trust, communication and alienation. In all three structs they refer to. An EFA in the pool of the old and new items
forms, however, the high correlations between the latent variables may then help improve the subscales by eliminating cross-load-
indicate that the constructs are poorly differentiated and this casts ing/ambiguous items and retaining only items that appear as pure
some doubt over whether the segmentation of the inventory into markers of the sub-dimensions. For future research it may also be
three subscales is useful at a practical level. For instance, it is un- useful to integrate the study of the IPPA into the broader context of
clear whether this segmentation is useful in terms of the incremen- adult attachment psychometrics, for example, defining the rela-
tal validity of the subscales in predicting attachment-related tions between the three factors of the IPPA – trust, communication
variables or in building sufficiently differentiated attachment and alienation – and the two dimensions of romantic attachment –
categories. anxiety and avoidance – evidenced by Fraley, Waller, and Brennan
Regarding this issue, it may be relevant to note that in Armsden (2000).
and Greenberg’s study (1987) the content of some items is not
clearly referable to the sub-scales to which they belong. For in-
stance, ‘‘My friend listens to what I have to say’’, which belongs References
to the peer subscale of trust (item 12) could also be considered a Allen, J. P., & Land, D. (2008). Attachment in adolescence. In J. Cassidy & P. R. Shaver
communication item. Item 12 of the mother and father forms: (Eds.), Handbook of attachment. Theory, research and clinical applications (2nd ed.,
‘‘When we discuss things, my father/my mother cares about my pp. 363–381). New York, NY: The Guilford Press.
Armsden, G., & Greenberg, M. T. (1987). The inventory of parent peer attachment:
point of view’’ is included in the subscale of trust, but could also
Individual differences and their relation to psychological well-being in
be interpreted as an aspect of communication. In fact, in the peer adolescence. Journal of Youth and Adolescence, 16, 427–454.
form the corresponding item (item 3) was included in the subscale Armsden, G. & Greenberg, M. T. (1989). The inventory of parent peer attachment
of communication. In a future study, it would be interesting to see (IPPA). Unpublished report. WA, USA: University of Washington.
Baiocco, R., Laghi, F., & Paola, R. (2009). Le scale IPPA per l’attaccamento nei
the resulting reliabilities and subscale correlations if the cross- confronti dei genitori e del gruppo dei pari in adolescenza: un contributo alla
loading/ambiguous items were excluded. validazione italiana [IPPA scales for evaluating attachment to parents and peers
As the original purpose of the IPPA was to provide researchers in adolescence: A contribution to the Italian validation]. Psicologia Clinica dello
Sviluppo, 13(2), 355–383.
and clinical practitioners with a multidimensional measure which Bentler, P. M. (1990). Comparative fit indices in structural models. Psychological
takes into account the complexity of the attachment system, it may Bulletin, 107, 238–246.
be useful to improve the inventory by focusing more on the three Bowlby, J. (1988). A secure base: Clinical applications of attachment theory. New York,
NY: Basic Books.
sub-dimensions and identifying items that reflect these dimen- Chen, F. F. (2007). Sensitivity of goodness of fit indexes to lack of measurement
sions more specifically, so that better differentiated and internally invariance. Structural Equation Modeling, 14, 464–504.
homogeneous subscales may be obtained. Fraley, R. C., Waller, N. G., & Brennan, K. A. (2000). An item response theory analysis
of self-report measures of adult attachment. Journal of Personality and Social
From this point of view the results obtained by Vignoli and Psychology, 78, 350–365.
Mallet (2004) are encouraging. They selected, on the base of EFA, Greenberg, M. T. (2004). Personal communication.
the items that best represented, both conceptually and statistically, Greenberg, M. T., Siegel, J. M., & Leitch, C. J. (1984). The nature and importance of
attachment relationships to parents and peers during adolescence. Journal of
the three sub-dimensions for both parents, and developed a short
Youth and Adolescence, 12, 373–386.
form of the parental part of the inventory with six items for com- Gullone, E., & Robinson, K. (2005). The inventory of parent and peer attachment
munication, four for trust and four for alienation. The three sub- revised (IPPA-R) for children: A psychometric investigation. Clinical Psychology
scales that in terms of content adhered best to their respective and Psychotherapy, 12, 67–79.
Johnson, L. N., Ketring, S. C., & Abshire, C. (2003). The revised inventory of parent
constructs showed inter-correlations lower than the original sub- attachment: Measuring attachment in families. Contemporary Family Therapy,
scales, and according to the authors may be ‘‘a reliable basis on 25(3), 333–349.
Author's personal copy

88 C.S. Pace et al. / Personality and Individual Differences 51 (2011) 83–88

MacCallum, R. C., Browne, M. W., & Sugawara, H. M. (1996). Power analysis and Raykov, T. (1998). On the use of confirmatory factors analysis in personality
determination of sample size for covariance structure modeling. Psychological research. Personality and Individual Differences, 24(2), 291–293.
Methods, 1, 130–149. Vignoli, E., & Mallet, P. (2004). Validation of a brief measure of adolescents’ parent
Marsh, H. W. (2007). Application of confirmatory factor analysis and structural attachment based on Armsden and Greenberg’s three-dimension model.
equation modeling in sport/exercise psychology. In G. Tenenbaum & R. C. European Review of Applied Psychology, 54(4), 251–260.
Eklund (Eds.), Handbook of on sport psychology (3rd ed., pp. 774–798). New York, Vivona, J. M. (2000). Parental attachment styles of late adolescents: Qualities of
NY: Wiley. relationships and consequences for adjustment. Journal of Counselling
Marsh, H. W., Hau, K. T., & Wen, Z. L. (2004). In search of golden rules: Comment on Psychology, 47(3), 316–329.
hypothesis testing approaches to setting cut-off values for fit indexes and Wilkinson, R. B., & Walford, W. A. (2001). Attachment and personality in the
dangers in overgeneralising Hu & Bentler’s (1999) findings. Structural Equation psychological health of adolescents. Personality and Individual Differences, 31,
Modeling, 11, 320–341. 473–484.

View publication stats

You might also like