ArticlePDF Available

The content validity of the Adult Attachment Interview. An empirical investigation using text analysis.

Authors:
  • Université de Rouen Normandie

Abstract and Figures

The aim of this study was to conduct a preliminary investigation of the content validity of the Adult Attachment Interview (AAI). Content validity is a series of hypothesis control methods relating to the communication of facts and it can be tested using several procedures of analytical decomposition (Rositi, 1988). In order to evaluate the content validity of a text, Sireci (1998), Sireci and Geisinger (1992, 1995), and Ding (2005) suggested the integration within this technique of multidimensional scaling and cluster analysis. The aim of this study was to conduct a preliminary investigation of the content validity of the AAI, as proposed in the literature. We hypothesized that participants with the same attachment style would use similar semantic systems, and that people who belong to different categories of attachment would use divergent semantic systems. Overall, 80 participants volunteered to take part in this research. Of these, 62% exhibited a secure attachment style, 20% a dismissing style, 14% a preoccupied style, and 4% an unresolved style. The interviews were coded in double blind trials by two expert codifiers. We created a textual corpus composed of the interview transcripts, which was processed using T-Lab software (Lancia, 2004). For each interview, the illustrative variable “attachment style” was indicated. Specificity analysis and cluster analysis were used. Results confirmed the bond between attachment representations. In particular, we observed that secure participants resorted to metacognitive processes, while dismissive participants were prone to idealization and lack of recall, and preoccupied participants primarily exhibited anger, blame, and a desire to close themselves off from the past.
Content may be subject to copyright.
243
THE CONTENT VALIDITY
OF THE ADULT ATTACHMENT INTERVIEW:
AN EMPIRICAL INVESTIGATION
USING TEXT ANALYSIS
LUCREZIA LORITO
FABRIZIO SCRIMA
UNIVERSITY OF PALERMO
The aim of this study was to conduct a preliminary investigation of the content validity of the Adult
Attachment Interview (AAI). Content validity is a series of hypothesis control methods relating to the
communication of facts and it can be tested using several procedures of analytical decomposition
(Rositi, 1988). In order to evaluate the content validity of a text, Sireci (1998), Sireci and Geisinger
(1992, 1995), and Ding (2005) suggested the integration within this technique of multidimensional
scaling and cluster analysis. The aim of this study was to conduct a preliminary investigation of the
content validity of the AAI, as proposed in the literature. We hypothesized that participants with the
same attachment style would use similar semantic systems, and that people who belong to different
categories of attachment would use divergent semantic systems. Overall, 80 participants volunteered to
take part in this research. Of these, 62% exhibited a secure attachment style, 20% a dismissing style,
14% a preoccupied style, and 4% an unresolved style. The interviews were coded in double blind trials
by two expert codifiers. We created a textual corpus composed of the interview transcripts, which was
processed using T-Lab software (Lancia, 2004). For each interview, the illustrative variable “attach-
ment style” was indicated. Specificity analysis and cluster analysis were used. Results confirmed the
bond between attachment representations. In particular, we observed that secure participants resorted to
metacognitive processes, while dismissive participants were prone to idealization and lack of recall,
and preoccupied participants primarily exhibited anger, blame, and a desire to close themselves off
from the past.
Key words: Content validity; Content analysis; Cluster analysis; Adult attachment; Mental representa-
tions.
Correspondence concerning this article should be addressed to Lucrezia Lorito, Dipartimento di Psicologia, Universi
di Palermo, Viale delle Scienze, Ed. 15, 90100 Palermo (PA), Italy. Email: lucrezialorito@libero.it
INTRODUCTION
The concept of content validity has been controversial since its inception. The first defi-
nition was given by Lennon (1956), who described content validity as the degree of correspon-
dence between answers to items and answers from a wider universe of reference. This definition
included content validity within the area of psychometric criteria of test validity, and in 1966 it
was integrated in the Standards for educational and psychological tests and manuals published
by the American Educational Research Association (AERA), the American Psychological Asso-
ciation (APA), and the National Council of Measurement in Education (NCME).
TPM Vol. 18, No. 4, December 2011 – 243-255 – © 2011 Cises
TPM Vol. 18, No. 4, December 2011
243-255
© 2011 Cises
Lorito, L., & Scrima, F.
Content validity of the Adult Attachment
Interview
244
Standards (AERA, APA, & NCME, 1974) described content validity as a demonstration
of a specific type of content to which the test refers. This text includes the definition, representa-
tion, and relevance of the domain in psychological measures.
In line with the most recent version of Standards (AERA, APA, & NCME, 1999), which
emphasizes a unitary conceptualization of validity, we underline the importance of the content
domain. In fact, in this edition, the expression “content validity” became “content-related evi-
dence of validity” (Sireci, 1998), which refers to the degree to which samples of items, tasks, or
questions are representative of a predefined universe or content domain.
The main procedure for evaluating content validity is the expert judge technique
(Crocker, Miller, & Franks, 1989; Osterlind, 1989). This technique is employed in studies in
which experts evaluate items, and rate them according to their relevance and representativeness
of the content domain. Sireci and Geisinger (1992, 1995) proposed the integration within this
technique of the methods of multidimensional scaling and cluster analysis, as they can reveal the
structure of a content domain, when a text composed of items is used (Ding, 2005; Sireci, 1998).
Sireci (1998) believed that these methods may reveal the areas of content that are present in
the text through clusters located in a multidimensional space, where groups of items for different con-
tent areas are discriminated. The level of similarity between items emphasizes the convergence or dif-
ferentiation of different portions of the text content. Deville (1996) proposed including the answers to
items and their importance within Sireci and Geisinger’s (1992, 1995) method.
Since a content analysis of these Adult Attachment Interview (AAI) transcripts has not
yet been performed, the general aim of this study was to conduct a preliminary investigation of
content validity of the AAI (Main, Goldwyn, & Hesse, 2002). In particular, content analysis is a
survey method based on a categorical scale; it is composed of a series of techniques, designed to
classify the information contained in oral or written material (Holsti, 1969; Krippendorf, 1980).
The central aspect of this procedure is the decomposition into classification units coinciding with
the units of context. Known as a thematic analysis in McClelland’s (1961) definition, content
analysis is described as a multi-stage process requiring the development of categories to codify
the thematic content, and also requiring the identification of materials and suitable statistical data
analysis. Therefore, message content is conceived as a window, allowing researchers to under-
stand the characteristics of a participant or a group (Krippendorf, 1980; Weber, 1990). For these
reasons, content analysis is considered a procedure halfway between qualitative and quantitative
analysis.
Adult Attachment Theory
Bowlby’s (1979) attachment theory provides a unique and comprehensive account of the
normative and individual differences in the processes that generate emotions in close relation-
ships. The individual difference component articulates how an individual’s personal history of
receiving care and support from attachment figures across his/her life span shapes the goals,
working models, and coping strategies that he/she will use when emotion-eliciting stimuli or
events occur in the context of relationships. Following Bowlby’s formulation, most research on
the significance of early attachment for later relationships relies on the distinction between secure
and insecure attachment experiences (Waters & Cummings, 2000).
TPM Vol. 18, No. 4, December 2011
243-255
© 2011 Cises
Lorito, L., & Scrima, F.
Content validity of the Adult Attachment
Interview
245
An enormous amount of research has been conducted since Ainsworth and her students
(e.g., Ainsworth, Blehar, Waters, & Wall, 1978) first identified individual differences between
infants in the use of various attachment strategies (Cassidy & Shaver, 2008; Mikulincer &
Shaver, 2007). Ainsworth and colleagues named three main categories of attachment with regard
to infants: free, dismissing, and entangled. This terminology was applied to adults by Main et al.
(2002), who used the terms secure, dismissing, and preoccupied.
Kobak and Sceery (1988) suggested that the way in which individuals perceive and manage
emotions in relationships depends on the nature of the working models formed in response to their
specific attachment histories. Secure attachment is organized according to rules that allow individu-
als to acknowledge distress and to turn to others for support; avoidant attachment is organized ac-
cording to rules that restrict the acknowledgement of distress and the associated attempts to seek
comfort and support; for preoccupied attachment, rules direct individuals’ attention toward distress
and attachment figures in a hypervigilant manner that inhibits the development of autonomy and
self-confidence.
In an extension of these ideas, Mikulincer and Shaver (2003) proposed a process model
that outlines the conditions under which the attachment system should be activated and termi-
nated in individuals who are securely attached. When potential threats are perceived, secure indi-
viduals should remain confident that their current attachment figures will be attentive, responsive,
and available to meet their needs and mitigate their distress. These beliefs should increase their
feelings of security, thereby deactivating their attachment systems and allowing them to use con-
structive, problem-focused coping strategies. Insecurely-attached individuals (dismissing and
preoccupied), on the other hand, should be more likely to experience attachment system activa-
tion, which motivate them to adopt interpersonal self-focused strategies, in order to compensate
for uncertainty about their partners’ responses. Recently, Mikulincer and Shaver (2007) stated
that the major insecure attachment patterns are associated with relatively poor adjustment and, in
some cases, psychopathology at various phases of life span.
Different aspects of attachment theory and adult relationships have led to the develop-
ment of a variety of assessment methods. These methods, developed from attachment theory, are
self-reports or interviews. Hazan and Shaver (1987) created the first questionnaire to measure at-
tachment in adults. It was designed to classify adults according to the three attachment styles
identified by Ainsworth et al. (1978). The questionnaire consisted of three sets of statements,
each describing an attachment style. An important advance in attachment questionnaires was the
use of independent items to assess attachment. Investigators have created several questionnaires
using this strategy to measure adult attachment. Two popular measures of this type are the Ex-
periences in Close Relationships questionnaire (Brennan, Clark, & Shaver, 1998) and the Experi-
ences in Close Relationships Revised questionnaire (Fraley, Waller, & Brennan, 2000).
In contrast, the AAI is designed to generate inferences about defenses associated with an
adult’s current state of mind, regarding one’s childhood relationships with one’s parents. The
central hypothesis suggests that parents’ mental representations of childhood attachment experi-
ences — as manifested in language — strongly influence the quality of their child’s attachment to
them. It is hypothesized that an adult’s evaluation of childhood experiences and their influence
on current functioning will become organized into a relatively stable state of mind with regard to
attachment (Main, Kaplan, & Cassidy, 1985).
The AAI is based on two assumptions: first, that autobiographical memory is the ongoing
reconstruction of one’s own past in light of new experiences, and second, that the idealization of
TPM Vol. 18, No. 4, December 2011
243-255
© 2011 Cises
Lorito, L., & Scrima, F.
Content validity of the Adult Attachment
Interview
246
the past, particularly of negative experiences, can be traced by studying the form and content of
the autobiographical narrative (van Ijzendoorn, 1995). As highlighted by Simonelli and Sironi
(2004), narrative methods of attachment evaluation are constructed in order to confront the par-
ticipant with mental processes related to attachment.
The possibility of classification is based upon the existing correspondence between the
organization of the mental world and some characteristics of narration (Main, 1995). Unlike ob-
servational methods, the perspective based on mental representations provides a representa-
tion of attachments, constructed after childhood, in terms of thought processes and the ability to
revise one’s past experiences (Simonelli & Sironi, 2004). Individual differences can be redefined
as individual differences in the representational and meta-representational elaboration of attach-
ment experiences, in light of the observation that there is a distinction between the mental proc-
esses of secure and insecure participants.
As observed by Ortu, Dazzi, De Coro, Pola, and Speranza (1992), secure participants
provide fluid and integrated narratives of their childhood experiences, and know how to appraise
the influence of these experiences on their mental state. In contrast, dismissing participants
devalue their attachment experiences and provide a narrative that is poor with regard to memo-
ries, while preoccupied participants are involved in precocious attachment experiences and pro-
vide a confused and untruthful narrative.
Narrative ability has the fundamental role of leading to the creation of shared histories,
from which particular forms of behavior derive. Narrative competence influences the internal
world and has an effect on the modulation of emotions and self-organization (Siegel, 2001).
AIM AND HYPOTHESES
The aim of this study was to conduct a preliminary investigation of the content validity of
the AAI, as proposed by Sireci (1998) and Ding (2005). The emergence of similar words and
similar semantic structures may indicate both content validity and that specific jargon is typical
of different attachment styles. In accordance with Main’s (1995) work, we suggested that the
bond between attachment behavior and its representation resides in this narrative style.
Specifically, we hypothesized that:
H1 = distinctive elements would emerge in the responses of participants, belonging to a
specific attachment style;
H2 = different semantic content would emerge in the responses of participants with dif-
ferent attachment styles.
METHOD
Participants
A total of 80 Italian employees volunteered to take part in this study. Participants’ mean
age was 39 (SD = 10.34). Sixty percent of participants were men, and 40% were women. Their
average seniority was 12.37 years (SD = 8.42). Fifty-five percent of participants described them-
selves as public sector employees, and 45% as private sector employees. The mean number of
TPM Vol. 18, No. 4, December 2011
243-255
© 2011 Cises
Lorito, L., & Scrima, F.
Content validity of the Adult Attachment
Interview
247
years of education was 14.5 (SD = 1.8). Of the participants, 50 (62%) showed a secure attach-
ment style, 16 (20%) an avoidant style, 11 (14%) a preoccupied style, and three (4%) an unre-
solved style. As there were so few of them, participants classified as unresolved were not in-
cluded in the analysis. In our group of 77 participants, none had a “cannot classify” (CC) style of
attachment.
Measures
In order to measure the attachment style, we used the AAI (Main et al., 2002), which is
composed of 20 questions taking an average of 60 minutes to answer. It evaluates participants’
mental representation of attachment, and their perceived family experiences in relation to attach-
ment. Participants were asked to give an overview of their childhood relationships with their par-
ents and to provide sets of five adjectives, describing their childhood relationship with each par-
ent. They were then invited to cite incidents or experiences from childhood that could illustrate or
explain the choice of each adjective. Next, feelings of rejection, experiences of being upset, ill or
hurt, separations, losses, and abuse were investigated. Participants were also encouraged to dis-
cuss changes in relationships with parents since childhood, to describe current relationships with
them, and to explain their understanding of their parents’ behavior. Finally, participants were
prompted to consider the effects of early childhood experiences on their adult personality and
parenting, as well as concerns and hopes for their children.
The psychometric properties of the AAI were evaluated by George, Kaplan, and Main
(1984). Today, the AAI is one of the most reliable and valid tools for the measurement of adult at-
tachment (van Ijzendoorn, 1995), and it allows attachment to be appraised through the analysis of nar-
ratives produced by adults and the identification of internal working models of the self and attach-
ment. The relevant literature has provided evidence that the AAI has satisfactory reliability (Baker-
mans-Kranenburg & van Ijzendoorn, 1993; Cassidy & Shaver, 2008; Sagi, Donnell, van Ijzendoorn,
Mayseless, & Aviezer, 1994; van Ijzendoorn, 1995) and discriminant validity. The categories of at-
tachment used in the interview do not differ in their scores for levels of memory, intelligence, and so-
cial desirability (Bakermans-Kranenburg & van Ijzendoorn, 1993; Cassidy & Shaver, 2008).
The interview allows five classifications of attachment: secure, preoccupied, dismissing,
unresolved, and cannot be classified (Main et al., 2002). The coding process is composed of two
separate parts: first, the content and text form are analyzed using nine continuous scales, then a
final classification of the participant’s attachment style is assigned. Scales are separated into sub-
jective experience scales (loving, rejecting, neglecting, role reversal, and pressure to achieve) and
state of mind scales (idealizing, involving anger, derogation, insistence upon lack of recall, meta-
cognitive processes, passivity of thought processes, fear of loss, unresolved loss/unresolved
trauma, coherence of transcript, and coherence of mind). The dimensions of idealization, deroga-
tion and insistence upon lack of recall characterize the dismissing attachment style, while the an-
ger and passivity scales characterize preoccupied participants, and the metacognitive and coher-
ence dimensions define secure participants.
The coding was carried out by two expert codifiers and we applied Cohen’s kappa index
(Landis & Koch, 1977) in order to appraise the reliability of coding. Interrater agreement was
TPM Vol. 18, No. 4, December 2011
243-255
© 2011 Cises
Lorito, L., & Scrima, F.
Content validity of the Adult Attachment
Interview
248
calculated, and data revealed a high level of agreement between judges ( =.85). The coders
achieved 82% agreement over the four classifications (secure, dismissing, preoccupied, cannot be
classified). Disagreements between coders were settled through discussion.
Procedure
First, we individually administered the AAI, with average length of 63 minutes per inter-
view (range: 42-78). The interviews were audiotaped, and the verbatim transcripts were rated on
15 9-point scales measuring adult attachment style. Then, we assigned an attachment category to
each interview and created a textual corpus, comprising the interviews transcripts. Such corpus
was composed of 80 texts, processed using T-Lab software (Lancia, 2004). For each interview,
the illustrative variable, or “attachment style,” was indicated.
Specifically, after the corpus was processed, we performed a preliminary preparation of
the text. We carried out an examination of poliforms, or multiple words (i.e., attachment_style).
This procedure allowed the presence of phrases with one semantic value to be highlighted. Sub-
sequently, we proceeded with a disambiguation of lexical forms in order to point out the semantic
differences between homographic words. Finally, we identified the structures and the headwords
of the corpus and chose a list of keywords. In T-Lab systems, the keywords list is based on the
total number of occurrences (quantitative criterion) and on the qualitative importance of items
(Lancia, 2004).
Statistical Analysis
The analysis resulted in 235 keywords. Specificity analysis and cluster analysis were
used. Specificity analysis allowed us to extract the words typical of each attachment category as
well as those which were used exclusively in one of the three subsets considered (secure,
avoidant, preoccupied). In T-Lab, this method of analysis allowed the identification of the exclu-
sive lexical units of the subsets of the corpus (defined by the attachment style” variable, or
rather by words present in the subset in question and not in the others). To the extracted words ²
test with p < .05 (one degree of freedom) was applied.
Cluster analysis was carried out using Ward’s (1963) method. Words were then included
into clusters, represented on a Cartesian space, and related to each category of attachment. Clus-
ter analysis allowed words and variables with the greatest similarity and the greatest difference to
be grouped and explained. Ward’s method starts treating each individual observation as a cluster.
These clusters are gradually agglomerated into one large cluster on the basis of a proximity
measure, using a predefined fusion algorithm (based on the analysis of variance approach). In or-
der to enable the identification of robust groups of observations, we stopped using the fusion al-
gorithm at the point in which clusters were as homogenous as possible internally, and as hetero-
geneous as possible in relation to all the other clusters. The parsimony criterion and the relative
increment of the agglomeration coefficient were used to determine the optimal number of clusters
to retain (Dillon & Goldstein, 1984).
TPM Vol. 18, No. 4, December 2011
243-255
© 2011 Cises
Lorito, L., & Scrima, F.
Content validity of the Adult Attachment
Interview
249
Results
As shown in Table 1, participants with a secure attachment style used terms that can be
included in the metacognitive processes dimension, such as “to believe,” “to decide,” “aware-
ness,” and “choice.” The fact that “mentality” is among the specificities for defect reveals the
presence of room for dialogue and negotiation, that seems to prevent secure people from adher-
ing rigidly to norms. Words such as “point of reference” (e.g., “My mom’s sister is a point of
reference for me, I felt as if she was my mom”), “protective” (e.g., “It has never been a distress-
ing relationship, she was protective of me”) and “tied up” (e.g., “My attitudes toward my aunt
and uncle have always been very tied up”) underline the fact that attachment is valued and high-
light the existence of feelings expressed to describe attachment experiences. Furthermore, the
word “independent” (e.g., “She has had a very positive influence on my personality, always
making sure that her child could be independent”) suggests the presence of autonomy in rela-
tionships. To conclude, the negative use of the term “to get mad” is evidence of the low pres-
ence of anger.
TABLE 1
Output table of specifity analysis for the secure attachment category
Secure
Lemmas ² Subtotal Total
To believe 17.58 187 255
To decide 17.77 38 45
Point of reference 10.33 29 33
Awareness 9.74 22 24
Tied up 8.74 43 48
Protective 8.61 34 41
Choice 6.70 55 73
Independent 6.53 25 30
Feeling 6.12 13 14
To separate 6.12 45 59
Dialogue 5.72 33 42
Safety 5.22 47 63
To express 4.87 11 12
Present 4.77 131 192
Attached 4.73 33 43
To get mad –35.11 3 32
Mentality –22.75 7 34
Support –9.76 4 17
Tragedy –5.75 5 16
Note. All ² values are significant, p < .05.
As shown in Table 2, avoidant adults used terms such as “I don’t remember” (e.g., “The
table was to be always set, because it was the best moment, being at the table, and I do not clearly
remember those times”) and “not remembering” (e.g., “I find it normal not to remember those
TPM Vol. 18, No. 4, December 2011
243-255
© 2011 Cises
Lorito, L., & Scrima, F.
Content validity of the Adult Attachment
Interview
250
times”) which point out a lack of memory. On the other hand, “loving” (e.g., “Where loving my
parents was the most important thing”), “thoughtful,” “respectful” (e.g., “My father was thought-
ful and respectful with women”), and “beauty” (e.g., “Mom was generous and affectionate, my
mom was a beautiful woman. I remember that she was a beautiful young woman and a good
mother. Then, I do not remember much”) emphasize the dimension of idealization.
TABLE 2
Output table of specifity analysis for the avoidant attachment category
Avoidant
Lemmas ² Subtotal Total
To impart 52.03 10 12
Loving 45.66 9 11
Thoughtfully 40.46 14 25
Respectfully 33.23 12 22
Do not remember 30.41 200 1070
Beautiful 24.09 77 352
Refuse 13.66 11 31
Dissatisfaction 13.61 10 27
I don’t remember 12.31 23 90
To educate 8.59 10 33
To improve 8.38 6 16
Normal 6.81 25 118
To play 6.72 30 148
Note. All ² values are significant, p < .05.
The exclusive use of the words “refusal” (e.g., “…is an innate feeling for children. I can
say it was a refusal and felt neglected because there was a little sister”) and “dissatisfied” (e.g., “I
felt dissatisfied toward myself, as there was my sister”) reveal the significant influence of the re-
jection dimension. The use of the term “normal” (e.g., “…because I think that is normal between
father and son, but I’ve never been threatened”) points to a tendency toward normalization and
minimization of distress, that is typical of avoidant attachment, as confirmed by the negative use
of “mother” and “important.” Similarly, the term “to play” suggests the appeal to avoidant people
of an emphasis on activities and fun (e.g., “…my mother complained to me because I was always
playing with my friends. It was the only way to have fun”), while the word “to impart” indicates
their preference for rigid educational norms (Mikulincer, 1998) (for instance: “I think it is right
that the father imparts the rules. My father made the general rules; the rules of a practical nature
were imparted by my mother”).
As regards preoccupied participants, as shown in Table 3, the expressions “to get mad”
(e.g., “…and it hurts my son and then I get mad! Then there are arguments because of this”),
“contradiction” (e.g., “…is a contradiction because I felt protected, I was happy, but I was always
fearful”), “to rub in,” “to cry,” “to bother” (e.g., “It bothers me to stand there in front [of them] to
cry, but not because I don’t want to cry but because I feel weak”), and “to go nuts” underline the
involvement of the dimension of anger and the tendency to blame others. This is also confirmed
TPM Vol. 18, No. 4, December 2011
243-255
© 2011 Cises
Lorito, L., & Scrima, F.
Content validity of the Adult Attachment
Interview
251
by the terms “fear” and “anxiety” (e.g., “Maybe everything I had inside, what I was in childhood,
my anxiety, my fear, came out later, after 30 years”). On the other hand, “slap” (e.g., “…Now I
do not think you can raise a child without giving him or her a slap on the ass”), “fear,” “death,”
“to suffer”, and “funeral” (e.g., “First, my grandmother died, and then, after a few years, my
grandfather died. I was 20 and I suffered a lot. I was very shocked, because I saw them rou-
tinely”) underline the need to break with the past and traumatic experiences.
TABLE 3
Output table of specifity analysis for the preoccupied attachment category
Preoccupied
Lemmas ² Subtotal Total
To get mad 146.23 28 32
Contradiction 54.81 11 13
To get used to 50.73 23 46
Fear 37.98 58 202
To admire 32.11 8 11
To cry 31.22 43 140
Slap 19.79 17 47
To bother 14.57 22 50
To suffer 14.25 12 22
To rub in 13.60 9 22
Anxiety 13.21 12 34
To get angry 9.15 11 35
Death 7.82 18 54
To refuse 7.56 7 20
Funeral 7.54 14 52
To go nuts 6.74 8 17
To bear 5.77 6 18
To strike 5.05 6 19
Violence 4.71 4 11
Respect –11.11 2 96
Present –9.26 12 192
Trauma –6.02 1 51
To disturb –5.86 1 50
Disturbed –4.29 2 59
Note. All ² value are significant, p < .05.
Cluster Analysis
In our data, an abrupt change in the relative increment of the agglomeration coefficient
occurred when four clusters were merged into three. Thus, we obtained three clusters, each char-
acterized by a category of attachment (Figure 1). Cluster 1 (explained variance = 25.69%) repre-
sented secure participants (² = 3.85, p < .05), cluster 2 (explained variance = 40.36%) repre-
sented preoccupied participants (² = 15.02, p < .01), and cluster 3 (explained variance = 33.95%)
represented avoidant participants (² = 13.30, p < .01).
TPM Vol. 18, No. 4, December 2011
243-255
© 2011 Cises
Lorito, L., & Scrima, F.
Content validity of the Adult Attachment
Interview
252
Cluster 1
Cluster 2
Cluster 3
Secure
Preoccupied
Avoidant
-1.0
-1.5
-0.5
0.5
1.0
1.5
2.0
0-1.0-1.5 -0.5 0.5 1.0 1.5 2.0
Note. Circles represent clusters extracted from the analysis; squares represent the position of the explanatory
variables in the plot. A smaller distance between cluster and the explanatory variable indicates a higher de-
gree of association between them.
FIGURE 1
Dimensional space of cluster analysis.
As shown in Table 4, in accordance with specificity analysis, the first clusters (secure)
encompassed the words which referred to metacognitive activity (to believe, to decide, dialogue)
and to the fact of valuing attachment (availability, trust, to love). In the second cluster (preoccu-
pied), the presence of anger (to get mad) and blame (guilt, to desire, desire, to be convinced) was
observed. The presence of the term “attachment” suggests the use of psychological jargon (psy-
chobabble). The third cluster encompassed the dimension of idealization and indicated the ex-
perience of refusal. This cluster was associated with variables of the avoidant attachment modal-
ity. In particular, the presence of the words “happiness” and “love” suggests the tendency to
minimize negative experiences, as it can also be deduced by the presence of the terms “separa-
tion,” “pain” and “lack,” which suggest negative experiences.
DISCUSSION
Content validity is based on the correspondence between the content being investigated
by the measure and a given domain. The absence of a robust analysis of content validity of the
AAI stimulated our interest in verifying how semantic speaking modalities can be used in order
to explain the specificities of attachment styles. Our contribution must be understood as a first
attempt to study content validity as well as an alternative way of analyzing the AAI.
As demonstrated by results, specificity analysis and cluster analysis allowed the words
used by members of every attachment category to be identified. Results show that two distinct
methods of analysis (judges and software) on the same content are coherent. In fact, T-Lab analy-
TPM Vol. 18, No. 4, December 2011
243-255
© 2011 Cises
Lorito, L., & Scrima, F.
Content validity of the Adult Attachment
Interview
253
TABLE 4
T-Lab output table of cluster analysis
Cluster 1 Cluster 2 Cluster 3
Lemmas Occurrence Lemmas Occurrence Lemmas Occurrence
Mother 1112 Home 554 Parent 559
Child 417 Beautiful 329 Brother 308
To believe 238 Child 285 To grow up 206
Different 153 Friend 216 Happiness 70
Education 108 Affection 151 Detachment 61
Sorrow 48 Change 146 Love 60
To decide 45 To know 128 Pain 58
Closed 44 Ugly 105 Too frequent
43
To accompany 41 Companion 66 Generosity 41
To wish 41 Teaching 58 Separation 40
Happy 41 Attachment 76 Lack 37
Availability 40 To be convinced 45
Dialogue 37 To accept 40
Good 36 Desire 34
Trust 30 To get mad 31
To love 28 Distance 30
Guilt 24
To desire 21
Variable ² Variable ² Variable ²
Secure 3.85 Preoccupied 15.02 Avoidant 13.30
Note. ² values of variables are significant, p < .05; ² values of words are significant, p < .01.
sis provided evidence for reliability of AAI coding system and the constructs that it measures.
This coherence may be used to evaluate the quality of judges’ coding process; we think that it is a
possible indicator of the reliability of using computer-based analysis when coding the AAI.
Therefore, we suggest that this procedure is used to support the AAI classical coding system and
the construct it measures. Despite the relatively small size of the sample, the analysis of the AAI
text partially confirmed the dimensional structure of the attachment categories.
In accordance with Main et al.’s description (2002), individual differences in the repre-
sentational and meta-representational elaboration of attachment experiences (metacognitive proc-
esses, lack of memory, valuing of attachment, involvement of anger, etc.) can be used to explain
the distinction between the mental processes of secure and insecure participants. As hypothe-
sized, our results suggest that it is possible to identify different styles of attachment on the basis
of the lexicon used (Fyffe & Waters, 1997; Main, 1995; Simonelli & Sironi, 2004). Secure par-
ticipants displayed metacognitive processes and valuing of attachment; dismissing participants
showed lack of memory, rejection, and normalization; and preoccupied participants presented
closing in one’s past and involvement of anger. This draws attention to the correspondence be-
tween the content of the interviews and a defined universe or content domain (Sireci, 1998).
Moreover, no content was theoretically extraneous to the investigated domain.
TPM Vol. 18, No. 4, December 2011
243-255
© 2011 Cises
Lorito, L., & Scrima, F.
Content validity of the Adult Attachment
Interview
254
Due to the characteristics of the analyses carried out, it has not been possible to trace two
salient characteristics for the attribution of attachment categories: speech passivity (e.g., the abil-
ity of using vague phrases or nonsensical words, and that of wandering over irrelevant topics) and
the coherence of the transcript. Finally, because respondents categorized as unresolved and CC
were absent, we suggest repeating the study with a larger group of participants, including clinical
respondents.
REFERENCES
Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, S. (1978). Patterns of attachment. A psychological
study of a Strange Situation. Hillsdale, NJ: Erlbaum.
American Educational Research Association (AERA), American Psychological Association (APA), & Na-
tional Council of Measurement in Education (NCME). (1966). Standards for educational and psy-
chological tests and manuals. Washington, DC: Author.
AERA, APA, & NCME (1974). Standards for educational and psychological tests and manuals. Washin-
gton, DC: Author.
AERA, APA, & NCME (1999). Standards for educational and psychological testing. Washington, DC:
AERA.
Bakermans-Kranenburg, M. J., & van Ijzendoorn, M. H. (1993). A psychometric study of the Adult At-
tachment Interview: Reliability and discriminant validity. Developmental Psychology, 29, 870-879.
Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998). Self-report measurement of adult attachment. In J. A.
Simpson & W. S. Rholes (Eds.), Attachment theory and close relationships (pp. 46-76). New York:
Guilford Press.
Cassidy, J., & Shaver, P. R. (2008). Handbook of attachment. Theory, research, and clinical applications
(2nd ed.). New York: Guilford Press.
Crocker, L. M., Miller, D., & Franks, E. A. (1989). Quantitative methods for assessing the fit between test
and curriculum. Applied Measurement in Education, 2, 179-194.
Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests. Psychological Bulletin,
52, 281-302.
Deville, C. W. (1996). An empirical link of content and construct validity evidence. Applied Psychological
Measurement, 20, 127-139.
Dillon, W. R., & Goldstein, M. (1984). Multivariate analysis. Methods and applications. New York:
Wiley.
Ding, C. S. (2005). Determining the significance scale values from multidimensional scaling profile analy-
sis using a resampling method. Behavior Research Methods, 37, 37-47.
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 Psychology, 78, 350-365.
Fyffe, C. E., & Waters, E. (1997, April). Empirical classification of the adult attachment status: Predicting
group membership. Paper presented at the biennial meeting of the Society for Research in Child
Development, Washington, DC.
George, C., Kaplan, N., & Main, M. (1984). Adult Attachment Interview protocol. Unpublished manu-
script, University of California at Berkeley.
Holsti, O. R. (1969). Content analysis for the social sciences and the humanities. Cambridge, UK: Addi-
son-Wesley.
Kobak, R., & Sceery, A. (1988). Attachment in late adolescence: Working models, affect regulation, and
representations of self and others. Child Development, 59, 135-146.
Krippendorf, K. (1980). Content analysis: An introduction to its methodology. London: Sage.
Lancia, F. (2004). Strumenti per l’analisi dei testi. Introduzione all’uso di T-Lab [Instruments for text
analysis. Introduction to the use of T-Lab]. Milano: FrancoAngeli.
Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical data. Biomet-
ric, 33, 159-174.
Lennon, R. T. (1956). Assumptions underlying the use of content validity. Educational and Psychological
Measurement, 16, 294-304.
Main, M. (1995). Recent studies in attachment overview with implications for clinical works. In S. Gold-
berg, R. Muir, & I. Kerr (Eds.), Attachment theory: Social development and clinical perspective (pp.
407-474). Hillsdale, NJ: Analytic Press.
Main, M., Goldwyn, R., & Hesse, E. (2002). Adult attachment scoring and classification system. Unpub-
lished manuscript, University of California at Berkeley.
TPM Vol. 18, No. 4, December 2011
243-255
© 2011 Cises
Lorito, L., & Scrima, F.
Content validity of the Adult Attachment
Interview
255
Main, M., Kaplan, N., & Cassidy, J. (1985). Security in infancy, childhood and adulthood: A move to the
level of representation. Monographs of the Society for Research in Child Development, 50(1-2, Se-
rial No. 209).
McClelland, C. (1961). The achieving society. Princeton, NJ: Van Nostrand.
Mikulincer, M., & Shaver, P. R. (2003). The attachment behavioral system in adulthood: Activation, psy-
chodynamics, and interpersonal processes. In M. P. Zanna (Ed.), Advances in experimental social
psychology (pp. 53-152). New York: Academic Press.
Mikulincer, M., & Shaver, P. R. (2007). Attachment in adulthood: Structure, dynamics, and change. New
York: Guilford Press.
Osterlind, S. J. (1989). Constructing test items. Hingham, MA: Kluver.
Ortu, F., Dazzi, N., De Coro, A., Pola, M., & Speranza, A. M. (1992). Un contributo di ricerca sugli stili di
attaccamento in preadolescenza: la coerenza della verbalizzazione nell’intervista [A research contri-
bution to attachment styles in preadolescence: verbalization coherence in the interview]. Adolescen-
za, 3, 269-287.
Rositi, F. (1988). Analisi del contenuto [Content analysis]. In F. Rositi & M. Livolsi (Eds.), La ricerca
sull’industria culturale (pp. 59-94). Roma: Nis.
Sagi, A., Donnell, F., van Ijzendoorn, M. H., Mayseless, O., & Aviezer, O. (1994). Sleeping out-of-home
in a kibbutz communal arrangement. It makes a difference for infant-mother attachment. Child De-
velopment, 65, 992-1004.
Siegel, D. J. (2001). La mente relazionale [The relational mind]. Milano: Raffaello Cortina.
Simonelli, A., & Sironi, A. (2004). L’attaccamento come narrazione [Attachment as narrative]. In G. Fava
Viziello & A. Simonelli (Eds.), Adozione e cambiamento (pp. 122-143). Torino, Italy: Bollati Bo-
ringhieri.
Sireci, S. G. (1998). The construct of content validity. Social Indicator Research, 45, 83-117.
Sireci, S. G., & Geisinger, K. F. (1992). Analyzing test content using cluster analysis and multidimensional
scaling. Applied Psychological Measurement, 16, 17-31.
Sireci, S. G., & Geisinger, K. F. (1995). Using subject matter experts to assess content representation. Ap-
plied Psychological Measurement, 19, 241-255.
van Ijzendoorn, M. H. (1995). Adult attachment representations, parental responsiveness, and infant at-
tachment: A meta-analysis on the predictive validity of the Adult Attachment Interview. Psycho-
logical Bulletin, 117, 387-403.
Ward, J. H. (1963). Hierarchical grouping to optimize an objective function. Journal of American Statisti-
cal Association, 58, 236-244.
Waters, E., & Cummings, E. M. (2000). A secure base from which to explore close relationships. Child
Development, 71, 164-172.
Weber, R. P. (1990). Basic content analysis. London: Sage.
Article
-The goal was to compare three-factor and two-factor solutions and construct validity of the Adult Attachment in the Workplace (AAW) questionnaire. Participants were 660 volunteers from three countries (France, Italy, and Great Britain). The two-factor model of Neustadt, Chamorro-Premuzic, & Furnham (2006) and the three-factor theoretical model of Collins and Read (1990) were compared. Construct validity was assessed by calculating correlations among the two- and three-factor AAW, the Workplace Attachment Scale, and the Organizational Commitment Scale. The three-factor structure differentiated between the three attachment styles, i.e., secure, preoccupied, and avoidant. There were moderate, significant correlations between AAW, workplace attachment, and affective commitment. Confirmatory factor analysis indicated that the three-factor structure fit the data better. Furthermore, the AAW, the Workplace Attachment Scale, and the Organizational Commitment Scale can be considered independent. In line with previous empirical evidence, a further distinction is noted between avoidant and preoccupied styles in the workplace.
Article
Full-text available
Our model outlines the cognitive operations, response strategies, and dynamics of the attachment system in adulthood. It also describes the goals of each attachment strategy and their psychological manifestations and consequences. Whereas the goals of security-based strategies are to form intimate relationships, to build a person's psychological resources, and to broaden his or her perspectives and capacities, the goal of secondary attachment strategies is to manage attachment-system activation and reduce or eliminate the pain caused by frustrated proximity-seeking attempts. Hyperactivating strategies keep the person focused on the search for love and security, and constantly on the alert for threats, separations, and betrayals. Deactivating strategies keep the attachment system in check, with serious consequences for cognitive and emotional openness. This framework serves as our "working model" for understanding the activation and functioning of the attachment system in adulthood. It also provides a framework for reviewing our research findings, which is the mission of the next section.
Chapter
Constructing test items for standardized tests of achievement, ability, and aptitude is a task of enormous importance—and one fraught with difficulty. The task is important because test items are the foundation of written tests of mental attributes, and the ideas they express must be articulated precisely and succinctly. Being able to draw valid and reliable inferences from a test’s scores rests in great measure upon attention to the construction of test items. If a test’s scores are to yield valid inferences about an examinee’s mental attributes, its items must reflect a specific psychological construct or domain of content. Without a strong association between a test item and a psychological construct or domain of content, the test item lacks meaning and purpose, like a mere free-floating thought on a page with no rhyme or reason for being there at all.