Psychiatry Research ] (]]]]) ]]]–]]]
Contents lists available at SciVerse ScienceDirect
Psychiatry Research
journal homepage: www.elsevier.com/locate/psychres
The development of the Metacognition Assessment Interview: Instrument
description, factor structure and reliability in a non-clinical sample
Antonio Semerari a,n, Michele Cucchi b, Giancarlo Dimaggio a, Daniele Cavadini b, Antonino Carcione a,
Vittoria Battelli b, Giuseppe Nicolo a, Roberto Pedone a,c, Tomaso Siccardi b, Stefania D0 Angerio a,
Paolo Ronchi b, Cesare Maffei b, Enrico Smeraldi b
a
Terzo Centro di Psicoterapia Cognitiva—Scuola di Psicoterapia Cognitiva (SPC), Training School in Cognitive Psychotherapy, via Ravenna 9/c, 00161 Rome, Italy
Department of Clinical Neurosciences, Scientific Institute and University Vita-Salute San Raffaele, Via Stamira D0 Ancona 20, 20127 Milan, Italy
c
Department of Psychology, University of Naples II, Italy
b
a r t i c l e i n f o
abstract
Article history:
Received 7 June 2011
Received in revised form
25 June 2012
Accepted 14 July 2012
Background: Metacognition is a multi-facet psychological construct; deficits in metacognitive abilities
are associated to low social functioning, low quality of life, psychopathology, and symptoms. The aim of
this study was to describe and develop a valid and reliable interview for assessing metacognition.
Methods: The semi-structured interview, based on the author’s theory model of the metacognition
construct, is described. The Metacognition Assessment Interview (MAI) is an adaptation of the
Metacognition Assessment Scale (MAS) and evaluates how the subject is interviewed used metacognition during his own real life experiences elicited by the interviewer. A user manual was developed to
assist the interview and scoring procedure.
Results: Exploratory factor analysis and confirmatory factor analysis revealed preliminary evidence of a
two factor-hierarchical structure, with two lower-order scales, representing the two main theoretical
domains of the metacognitive function, ‘‘the Self’’ and ‘‘the Other’’, and one single higher-order scale
that we labelled metacognition. Contrary to the authors’ prediction the existence of the four distinct
dimensions under the two domains was not confirmed. The MAI and its two domains demonstrated
acceptable levels of inter-rater reliability and internal consistency.
Conclusions: The MAI appears to be a promising instrument for assessing metacognition. Future
psychometric validation steps and clinical directions are discussed.
& 2012 Elsevier Ireland Ltd. All rights reserved.
Keywords:
Metacognition
Metacognition assessment
Interview
Reliability
Factor structure
1. Introduction
Research investigating the metacognitive construct has
advanced rapidly over the past decade. Metacognition, as formulated by Semerari (Semerari et al. 2003; Carcione et al. 2008),
refers to a broad set of cognitive and affective skills which allow
people to identify mental states, reasoning about them, and
ascribing them to themselves and others. These skills allow us
to recognize the reason why a person reacts psychologically
according to some regularities and constructs personal meaning
over their lifespan.
Several authors refer to the same concept with different
meanings. Wells (2000), for example, views metacognition as a
set of beliefs about one’s own mental content that helps people to
regulate their attentive processes and that, in some cases, could
induce the maintenance of dysfunctional attentive processes such
n
Corresponding author.
E-mail address: semerari@terzocentro.it (A. Semerari).
as worry. There is a terminological confusion over metacognition
and this is also due to the fact that mind-reading abilities have
been traditionally investigated by researchers belonging to different theoretical backgrounds and research fields, each of them
with their own lexicon (Flavell, 1976; Fonagy, 1991; BaronCohen, 1995; Frith and Happé, 1999; Frith and Frith, 2006).
Metacognition, as considered by Semerari et al. (2003), partially
overlaps with similar constructs such as theory of mind (ToM)
(Baron-Cohen et al., 1985) and mentalization (Bateman and Fonagy,
2004; Allen et al., 2008), but with some differences. Compared to
ToM, as described below, metacognition usually includes more
complex mental functions. Additionally, metacognition also refers
to emotional understanding, while ToM mainly focuses on cognitive
attribution. Compared to the general definition of mentalization
given by Bateman and Fonagy (2004), Semerari’s concept differs
since it considers mind-reading to be a general ability created by
different subfunctions that interact with each other and that can be
selectively impaired. Dysfunctions in metacognition are associated
with low social functioning, low quality of life, psychopathology and
symptoms of several psychiatric and personality disorders, and
0165-1781/$ - see front matter & 2012 Elsevier Ireland Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.psychres.2012.07.015
Please cite this article as: Semerari, A., et al., The development of the Metacognition Assessment Interview: Instrument description,
factor structure and reliability in a non-clinical sample. Psychiatry Research (2012), http://dx.doi.org/10.1016/j.psychres.2012.07.015
2
A. Semerari et al. / Psychiatry Research ] (]]]]) ]]]–]]]
seem to predict worst treatment response (Semerari et al., 2007;
Lysaker et al., 2010a, 2010b; Carcione et al., 2011; Lysaker et al.,
2011a; Ogrodniczuk et al., 2011).
However, the lack of a reliable and valid instrument for
measuring metacognition has limited progress in this field,
especially for developing clinical applications. Assessments of
psychological constructs such as social cognition mostly rely on
self-report instruments, laboratory tasks and structured interviews, and each has its own pros and cons. Interviews have
unique advantages such as avoiding bias in self-ratings, especially
if one is called upon to use a skill such as self-reflection, which in
itself is supposed to be compromised; they also ask the person to
reason about mental states in the context of personally relevant
matters, i.e. the ones in which it is most important to be able to
fully and swiftly understand the mental states of oneself and
others. Therefore it is likely that metacognition as measured from
interviews provides more clinician-friendly information that can
directly affect treatment protocols (Lysaker et al., 2010a, 2011b).
The aim of the present study was to investigate the reliability,
internal consistency and structure of a semi-structured interview
for assessing metacognition: the Metacognition Assessment Interview (MAI). The MAI is an adaptation of the Metacognitive
Assessment Scale (MAS) (Semerari et al., 2003; Carcione et al.,
2008; Carcione et al., unpublished), a rating scale assessing the
construct as manifest in individual verbalization in psychotherapy transcripts. The common theoretical framework which underlines both the MAS and the MAI is that metacognition is made of
specific and relatively independent subfunctions, each of them
likely to be selectively damaged in clinical populations. The MAS
is therefore divided into three scales: understanding one’s own
mind, understanding others’ minds and mastery. Each scale is
further composed of different subfunctions. The subfunction
scales of understanding one’s own mind and understanding
others’ minds are included in the MAI and they will be described
in detail in Section 2. Work on an interview to investigate the
third scale, mastery, is still in progress.
The MAS and its subscales demonstrated acceptable levels of
factorial validity, inter-rater agreement, internal validity and
test–retest stability (Semerari et al., 2003, 2005; Lysaker et al.,
2005; Carcione et al., 2008; Dimaggio et al., 2009; Lysaker et al.,
2010b, 2011a, 2011b) and were related to executive functions and
treatment outcomes (Lysaker et al., 2005; Semerari et al., 2005;
Carcione et al., 2008; Lysaker et al., 2008). Studies on patients
with personality disorders showed different profiles of metacognitive impairments in patients with different diagnoses (Semerari
et al., 2003, 2005, 2007; Dimaggio et al., 2009). Furthermore,
specific function impairment correlated with different symptoms
in patients with schizophrenia (Lysaker et al., 2005).
However, the MAS does not allow metacognitive functions to
be directly stimulated through specific questions. Therefore,
when the therapist does not directly investigate one specific
subfunction, perhaps during a psychotherapy session, it is impossible to define whether the lack of use of this subfunction is due to
impairment of it or simply to a lack of use of it in that specific
circumstance. However, the direct investigation of a specific
subfunction through specific questions, along with the weak
answers that patients might give, highlights more clearly specific
impairments in the metacognitive domains.
For this reason, we aimed to develop a new tool that is capable
of directly measuring metacognition and is less time-consuming
than the MAS: the MAI (Semerari et al., 2008). Our hypotheses
were (a) that metacognition could be elicited by such an interview and reliably scored; (b) that metacognition has a two-factor
structure, with two separate domains, one for understanding the
mental states of the self and one for understanding the mental
states of others.
2. Methods
2.1. Construct description
The MAI refers to the description of emotions and cognitions, and assesses
how people are able to identify their own and others’ recurrent patterns of
thinking, feeling and dealing with social problems. The interview evaluates two
main functional skill domains of metacognition, ‘the Self’ and ‘the Other’, each one
composed of two dimensions: monitoring and integrating for the Self, differentiating and decentring for the Other.
To identify the 16 basic facets of which the dimensions are composed (four
facets for each dimension) the authors took into account the clinical literature that
describes deficit in the ability to know and regulate mental states, theoretically
based on the literature on mentalization and attachment theories (Fonagy, 1991;
Main, 1991; Fonagy and Target, 1997, 2006; Allen et al., 2008), theory of mind
(Baron-Cohen et al., 1985; Premack and Woodruff, 1978a, 1978b, 1978c; Wellman,
1990), metacognition (Wellman, 1990; Wells, 2000) and, more generally, metarepresentation (Frith, 1992; Sperber, 2000).
The Self domain comprises the ‘monitoring’ and ‘integrating’ dimensions and
describes the way in which a person has explicit access to his own mental states
(cognitive and emotional) in relation to behaviour. ‘Monitoring’ (MON) is referred
to as the ability to distinguish, recognize and define one’s own inner states
(emotions and cognitions) and following behaviours during the ‘here and now’ of
the described real-life episode. MON is composed of four facets: (a) the ability to
recognize one’s own representations (thoughts and beliefs); (b) the ability to
recognize and verbalize one’s own emotions; (c) the ability to establish relations
among the separate components of a mental state; and (d) the ability to establish
relations between the components of mental states and behaviour. MON evaluates
how a subject explains his/her own behaviour in terms of causes and/or
motivations. If there is a deficit, he/she is unable to discern the reasons for his/
her behaviour, and he/she cannot recognize or verbalize emotions or other mental
states. Examples of questions stimulating MON are ‘What do you feel?’, ‘What do
you think?’ and ‘What was your aim?’.
‘Integrating’ (INT) is the second dimension of Self domain and involves the
ability to produce coherent descriptions of people’s mental processes and states
over time. INT refers to the ability to reflect on mental states and contents, putting
them in a logical order and ranking them by relevance. Using INT abilities the
subject is able to understand the link between his/her own mental states and
different behaviours in different situations, decoding his/her mental, functional
and dysfunctional habits and forming a consistent account of how his/her mental
life has changed over his/her lifespan.
INT comprises four skills: (a) the ability to describe understandable and
coherent links among thoughts, events, actions and behaviours; (b) the ability to
describe transitions among different mental states and explain the reasons why;
(c) the ability to form generalized representations of his/her mental functioning,
taking into account continuity over time of patterns of thinking and feeling; and
(d) the ability to reconstruct and describe to the interviewer one’s own mental
functioning, providing enough information, without giving irrelevant and out-offocus details, and giving a sense of order and coherence to the discourse. Examples
of questions stimulating INT are ‘So, you have found yourself reacting byy [the
interviewer refers to the described behaviour], and feelingy [the interviewer
refers to the mentioned emotion]. Does feeling/thinking and behaving like this
happen frequently to you?’ ‘You might also react in a different way, with different
emotions or thoughts, in circumstances like the one described. Can you remember
a specific episode?’
The Other metacognition domain comprises the ‘differentiating’ and the
‘decentring’ dimensions.
Differentiating (DIF) concerns the ability to recognize the representational nature
of one’s own and other individuals’ thoughts, the ability to differentiate between
classes of representations, such as imagination, evaluation and expectation, and to
distinguish between representation and reality. Using DIF abilities, the interviewee is
able to consider his/her own opinion as a hypothesis and not as a matter of fact; DIF
abilities allow one to consider representations as mental phenomena, separate from
but related to reality. Good DIF functioning makes people flexible in formulating
opinions and points of view, and causes mental states to change depending on the
communicative acts and on the availability of salient information. DIF comprises four
skills: (a) the ability to consider one’s own representation of the world as subjective
and questionable; (b) the ability to give plausible interpretations of events; (c) the
ability to reflect on and evaluate events (as opposed to a tendency to act impulsively);
and (d) the ability to distinguish between different modes of thoughts such as
dreaming, fantasizing and imagining. Examples of questions stimulating DIF are ‘You
said you have thought thaty [the interview refers to the reported episode]. In that
moment, how did you subjectively believe it and how did you consider other
options?’ ‘Did you take into consideration alternative interpretations of the events?’
‘During the episode how much did you feel confused or clear-headed?’ Have you ever
experienced such levels of confusion, or not been able to remember whether
something really happened, or felt dreamy, unreal, like brain fog?’.
‘Decentring’ (DEC) refers to the ability to infer others’ mental states in a
plausible manner and adopt their perspective, recognizing that it is distinct from
Please cite this article as: Semerari, A., et al., The development of the Metacognition Assessment Interview: Instrument description,
factor structure and reliability in a non-clinical sample. Psychiatry Research (2012), http://dx.doi.org/10.1016/j.psychres.2012.07.015
A. Semerari et al. / Psychiatry Research ] (]]]]) ]]]–]]]
our own. DEC leads to the realization that other people’s behaviours are understandable on the basis of their own aims, beliefs, values and principles, which
could be different from ours and independent of the relationship a person has with
the subject. It involves the ability to describe others’ psychology in a plausible,
clear way, without using stereotypes or cliché. DEC also includes the ability to
realize that basically we are not the centre of others’ intentions and goals. DEC
comprises four skills: (a) the ability to recognize, define and verbalize other
people’s emotional inner state; (b) the ability to recognize, define and verbalize
other people’s cognitive inner state; (c) the ability to establish relations among the
separate components of others’ mental state; and (d) the ability to establish
relations between the components of others’ mental state and their behaviour.
Examples of questions stimulating DEC are ‘How did you think the other person
would react emotionally during the episode?’ ‘What did you think he/she
thought?’ ‘Why did you think he/she thought that?’ and ‘What reasons did he/
she have?’.
2.2. Interview description
The MAI begins with a standardized overview about its purpose — the aim of
the interview — and collects basic demographic information. Afterwards, the
interviewee is encouraged to describe an autobiographical episode about the
worst psychological situation that he/she has experienced in the last six months.
The episode has to be autobiographical and private. In order to evaluate the
comprehension of others’ mental state, the episode has to include interaction with
another person. We decided to investigate the worst episode of the last six months
in order to be able to evaluate metacognitive function in critical circumstances
with the prospect of applying it to clinical populations. We chose six months as
the time interval to make it possible to test–retest, avoiding recalling biases, and
for evaluating change during psychotherapy. This is explicitly aimed at eliciting
affect-laden and personally relevant information. The MAI continues with four
modules, each specific to the evaluation of one metacognitive dimension as
described before. For each dimension of the metacognitive construct, the interviewer has to ask a structured list of questions.
2.3. Participants
The study protocol was approved by the local Ethical Committee. Participants
were recruited via advertisements posted around the hospital’s department of
psychiatry. Those who answered the advertisement were screened and provided
with information about the study, invited to sign the informed-consent form and
were interviewed by an expert clinician. Axis I and II diagnoses were defined
according to DSM-IV criteria (APA, 2000), using the Italian version of the
Structured Clinical Interview for DSM-IV Axis I and II disorders (First et al.,
1997). Exclusion criteria were current or past psychiatric diagnoses on DSM-IV
Axis I or II, a history of psychiatric or psychological treatment, or a history of
severe head trauma or substance-related disorders. None of the participants were
taking psychotropic drugs, or they had used them occasionally during the month
preceding the study. All six of the interviewers/raters conducted interviews and
scored (rated) the 16 questions during the course of their interview, which lasted
approximately 45 min. A sample comprising 175 non-clinical individuals (60 men
and 115 women) was recruited from the community. The mean ages for women
and men were 30.69 years (S.D. ¼ 13.51) and 30.68 years (S.D. ¼ 13.35), respectively, while the mean years of education were 14.88 (S.D. ¼2.52) for women and
14.64 years (S.D. ¼3.02) for men.
2.4. Measurement
2.4.1. Item development, interviewer training and preliminary inter-rater reliability
evaluation
As a first step, the authors formulated the structured questions that enable to
describe how the interviewee employed metacognition during the recounting of
the real-life episode. Guided by a previous theoretical work, nine questions were
written for each of the four related dimensions (MON, DIF, INT and DEC); the
interview therefore comprises 36 questions. Each question was discussed within
the research group and approved for its inclusion in the interview. For each of the
36 questions, the interviewer could use adjunctive, rather than just structured
questions to complete a more accurate information collection, by formulating
questions to fit the person’s understanding, and asking additional details to clarify
previous requests. The use of those free questions should be considered as probes
and are divided into ‘helps’ and ‘specifications’, depending on the nature of the
adjunctive question. If the interviewee needs to clarify an already detailed answer
provided to a structured question, the interviewer will call it ‘specification’; these
questions do not affect scoring. If the interviewee seems not to use well the
metacognitive facet under evaluation, the interviewer will give a ‘help question’ to
suggest the correct way to perform that task; the use of the ‘help’ will influence
the scoring procedure, determining a reduction of the score for that specific facet.
On the basis of the interviewee’s responses, the interviewer should determine
the scores for each one of the 16 facets. The output of the MAI is made of 16 items.
3
For each facet the interviewer will assign a score ranging from one to five, on the
basis of a Likert scale, to describe how well the interviewee employed that aspect
of the metacognitive function during the real-life episode; a score of one stands for
‘always deficit functioning, while a score of five stands for ‘always functioning
well’. For each question, a set of prompts and/or probes was written to elicit
information assisting in the accurate scoring of the item.
After the item selection procedure, 20 audiotaped interviews were conducted
by the two senior interviewers after obtaining written informed consent from the
interviewees. These interviews were then discussed in plenary sessions, creating
the basis for the practical part of the user’s manual, which contains instructions
for administration and scoring.
After completing the user’s manual, a preliminary inter-rater reliability
evaluation was carried out.
The MAI was administered to 11 outpatients (f ¼5 and m¼ 6) and 3 healthy
control subjects (f ¼ 2 and m¼ 1), recruited consecutively. The subjects were
volunteers and fulfilled DSM-IV criteria for different disorder categories. The 11
interviews were conducted by the two senior interviewers, while all six of the staff
rated each interview, using audio tapes. The interviewers and the raters were
blinded to the clinical characteristics of the sample recruited.
2.5. Statistical analysis
2.5.1. Preliminary inter-rater reliability
A blind rating procedure (blinded for diagnosis and others’ rating) was used in
an ANOVA design. In order to estimate the correlation for every single facet rated
by different judges, the Intraclass Correlation Coefficient (ICC) was used. A twoway mixed absolute agreement model was applied to carry out the Intraclass
Correlation Coefficients for each dimension of the metacognitive construct,
evaluated through the MAI.
2.5.2. Explorative factor analysis
Firstly, we explored the factorial structure of the MAI with a factor analysis
(based on the principal component extraction procedure) carried out using the
Statistica 7.0 software package for Windows (StatSoft Inc., Tulsa, Oklahoma, USA).
We used the scree test, the eigenvalue criteria (i.e. 41.00) and additional ‘rules of
thumb’ for determining the number of factors for extraction and rotation (Horn,
1965; Cattel, 1966; Gorsuch, 1983; Floyd and Widaman, 1995).
2.5.3. Confirmatory factor analysis (CFA) and model comparisons
We used CFA techniques in order to examine a number of different models
that could fit the matrix of the 16 items of the interview and that could be
considered conceptually and theoretically plausible, clarifying which latent
structure best represents the data. On the basis of the results from the explorative
factor analysis, we considered three competitive models: a two-factor nonhierarchical model, a two-factor hierarchical model and a one-factor model.
Multiple statistical fit indices were used in order to assess goodness of fit of
the proposed models: absolute fit indices (Chi Square/d.f. ratio and Goodness of Fit
Index [GFI]); relative fit indices (Comparative Fit Index [CFI] and Non-Normed Fit
Index [NNFI]); non-centrality based index (Root Mean Square Error of Approximation [RMSEA]). A good-fitting model produces consistent results on many different
indices, according to statistical criteria reported elsewhere by different authors
(Ullman, 1996; Kline, 1998; Netemeyer et al., 2004). Comparisons between
adequate models with different degrees of freedom have been tested with the
Chi Square test. The goodness-of-fit summary statistics from the CFA is displayed
in Table 2.
2.5.4. Mean scores and reliability estimate and intercorrelations of the MAI and its
scales
Descriptive statistics for the MAI dimensions and total score were calculated;
differences between men and women were explored using a series of T-tests. In
order to analyze the internal consistency of the scale, we carried out Cronbach0 s
alphas, average inter-item correlations (AICs) and corrected item-total correlations (CITCs) for the two domains and the global metacognitive function as
measured with the MAI and Pearson correlations among the MAI scales.
3. Results
3.1. Preliminary inter-rater reliability
The ICC for M facets ranges from 0.54 to 0.69; for DIF facets
from 0.44 to 0.76; for I facets from 0.59 to 0.64; and for D facets
from 0.41 to 0.57. The significance level is p o0.0001 for all
analyses.
Please cite this article as: Semerari, A., et al., The development of the Metacognition Assessment Interview: Instrument description,
factor structure and reliability in a non-clinical sample. Psychiatry Research (2012), http://dx.doi.org/10.1016/j.psychres.2012.07.015
4
A. Semerari et al. / Psychiatry Research ] (]]]]) ]]]–]]]
Table 1
Results from the factor analysis of the 16-item set (two-factor solution).
M3
DIF4
I3
I4
I1
M1
DEC3
DIF2
I2
DEC1
DIF3
M2
M4
DEC2
DIF1
DEC4
Eigenvalues
Variance (%)
Cumulative variance (%)
Factor 1
Factor 2
Communalities
0.23
0.00
0.23
0.15
0.24
0.33
0.83
0.63
0.55
0.86
0.29
0.29
0.40
0.82
0.69
0.81
7.02
43.90
43.90
0.69
0.52
0.65
0.72
0.74
0.54
0.22
0.29
0.37
0.17
0.51
0.67
0.63
0.19
0.23
0.20
1.74
10.92
54.83
0.53
0.27
0.48
0.55
0.61
0.40
0.74
0.48
0.44
0.77
0.35
0.54
0.56
0.71
0.54
0.71
3.3. Confirmatory factor analysis and model comparisons
Factor loadings higher than 0.40 are marked. M1 ¼Monitoring, item 1;
M2¼Monitoring, item 2; M3 ¼ Monitoring, item 3; M4 ¼ Monitoring, item 4;
I1 ¼Integrating, item 1; I2 ¼ Integrating, item 2; I3 ¼ Integrating, item 3;
I4 ¼Integrating, item 4; DIF1¼ Differentiating, item 1; DIF2 ¼Differentiating, item
2; DIF3 ¼ Differentiating, item 3; DIF4 ¼ Differentiating, item 4; DEC1 ¼Decentring, item 1; DEC2 ¼Decentring, item 2; DEC3 ¼ Decentring, item 3;
DEC4 ¼Decentring, item 4.
Table 2
CFA goodness of fit results for various factor structure models for the 16-item set.
Models
Goodness of fit statistics
w2 (d.f.)
w2/d.f. GFI
self-oriented metacognitive domain. The ‘describes transitions
across different mental states and explains the reasons’ item
(integration) has a less clear conceptual attribution, because it
concerns one0 s own mental state but loads on the other-oriented
factor.
CFI
NNFI
RMSEA
ratio
Two-factor models
2a
2b
198.48 (103)
123.36 (87)
1.93
1.42
0.87
0.91
0.92
0.97
0.91
0.96
0.07
0.05
One-factor model
1a
394.68 (104)
3.80
0.70
0.78
0.74
0.16
3.2. Explorative factor analysis
The first solution, obtained without imposing any constriction
on the analysis, comprised three factors. The third factor, however, comprised only two significant loading items (i.e. 40.40)
(Netemeyer et al., 2004) and was considered trivial (Gorsuch,
1983). The two-factor solution comprised at least three significant
loading items, with each extracted factor explaining more than 5%
of the unrotated variance (see Table 1) and the whole model
explaining 54% of the total variance, as recommended by
Netemeyer et al. (2004). We interpret this solution as providing
interesting information about the theoretical hypothesis concerning the organization of the metacognitive function.
All decentring items, two differentiation items (‘considers his/
her own representation of reality as subjective and questionable’
and ‘makes plausible interpretations of events’) and one integration item (‘describes transitions across different mental states and
explains the reasons’) load on the first factor; the remaining
items—all monitoring, two differentiating (‘evaluates and reflects
on events’ and ‘distinguishes between different kinds of mental
representations’) and three integrating (‘demonstrates a comprehensible and coherent link between thoughts, emotions and
behaviours’, ‘generalizes on mental functioning and detects
aspects of continuity in his/her own way of thinking and feeling’
and ‘reconstructs and describes relevant aspects of his/her narration’) items—loaded on the second one. These two dimensions
could be conceptualized respectively as an other-oriented and a
Model 2a was a two-factor, non-hierarchical model constituted by two independent factors. NNFI and CFI show adequate fit
(0.92 and 0.93 respectively) as well as RMSEA (0.076), while GFI is
0.87, consistent with an inadequate fit. The Chi Square/d.f. ratio is
consistent with a good fit (1.93). Model 2b was a two-factor
hierarchical model constituted by two partially dependent factors, nested under a single higher-order factor. NNFI and CFI show
good fit (0.96 and 0.97 respectively) for this model, while GFI is
0.92, consistent with an adequate fit as well as RMSEA (0.051).
The Chi Square/d.f. ratio is consistent with a good fit (1.42). Model
1a was a one-factor model, with all items loading on a single
factor. GFI and relative fit indices produce inconsistent results
(o0.90) as well as RMSEA (0.164) while the Chi Square/d.f. ratio
is consistent with a just sufficient adequate fit (3.80).
Because models 2a and 2b have both shown consistent results
on most or all fit indices, they have been compared with the Chi
Square test of difference. Model 2b produced better fit than model
2a: model 2a vs. model 2b, Chi Square (16)¼ 75.126, po0.00000.
Based on these results, we could conclude that a two-factor
hierarchical model is the best way to describe the 16-item matrix.
In particular, the MAI seems to be composed of two lower-order
scales, representing the two main domains of the metacognitive
function, which has the Self and the Other as object of its
processes, and one single higher-order scale that we labelled
‘metacognition’. Parameter estimates for model 2b are displayed
in Table 3.
3.4. Mean scores and reliability estimate and intercorrelations of the
MAI and its scales
Table 4 reports Cronbach’s alphas, AICs and CITCs for the two
domains and the global metacognitive function as measured with
the MAI. Cronbach’s alphas range between 0.85 of the second
Table 3
Item-to-facet parameter estimates from the CFA. Two-factor hierarchical model
(n ¼175).
Factor/item
Parameter estimate
Prob. level
(2)-1-4[M3]
(2)-2-4[DIF4]
(2)-3-4[I3]
(2)-4-4[I4]
(2)-5-4[I1]
(2)-6-4[M1]
(2)-7-4[DIF3]
(2)-8-4[M2]
(2)-9-4[M4]
(1)-10- 4[DEC3]
(1)-11-4[DIF2]
(1)-12-4[I2]
(1)-13-4[DEC1]
(1)-14-4[DEC2]
(1)-15-4[DIF1]
(1)-16-4[DEC4]
0.38
0.42
0.28
0.57
0.48
0.34
0.36
0.38
0.30
0.62
0.37
0.10
0.62
0.68
0.40
0.61
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.18
0.00
0.00
0.00
0.00
M1¼Monitoring. item 1; M2 ¼Monitoring. item 2; M3 ¼Monitoring. item 3;
M4¼Monitoring. item 4; I1 ¼Integrating. item 1; I2 ¼Integrating. item 2;
I3 ¼Integrating. item 3; I4 ¼Integrating. item 4; DIF1¼ Differentiating. item 1;
DIF2¼ Differentiating. item 2; DIF3¼ Differentiating. item 3; DIF4¼ Differentiating.
item 4; DEC1¼ Decentring. item 1; DEC2¼ Decentring. item 2; DEC3¼ Decentring.
item 3; DEC4 ¼ Decentring. item 4.
Please cite this article as: Semerari, A., et al., The development of the Metacognition Assessment Interview: Instrument description,
factor structure and reliability in a non-clinical sample. Psychiatry Research (2012), http://dx.doi.org/10.1016/j.psychres.2012.07.015
A. Semerari et al. / Psychiatry Research ] (]]]]) ]]]–]]]
Table 4
CITCs. Cronbach’s a and AICs for the M.A.I. scales.
Factor 1
M3
DIF4
I3
I4
I1
M1
DIF3
M2
M4
DEC3
DIF2
I2
DEC1
DEC2
DIF1
DEC4
a/AIC
0.80
0.61
0.54
0.80
0.76
0.65
0.76
0.90/0.57
Factor 2
Total
0.61
0.38
0.58
0.60
0.67
0.55
0.50
0.63
0.65
0.57
0.30
0.55
0.53
0.62
0.55
0.71
0.60
0.59
0.70
0.51
0.61
0.66
0.68
0.61
0.69
0.91/0.40
0.85/0.40
M1¼ Monitoring. item 1; M2 ¼ Monitoring. item 2; M3 ¼ Monitoring. item 3;
M4¼ Monitoring. item 4; I1 ¼ Integrating. item 1; I2 ¼ Integrating. item 2;
I3 ¼ Integrating. item 3; I4 ¼ Integrating. item 4; DIF1 ¼Differentiating. item 1;
DIF2 ¼Differentiating. item 2; DIF3 ¼Differentiating. item 3; DIF4 ¼Differentiating.
item 4; DEC1 ¼Decentring. item 1; DEC2 ¼Decentring. item 2; DEC3 ¼Decentring.
item 3; DEC4 ¼Decentring. item 4.
Table 5
Pearson correlations among the MAI scales.
Factor 1
Factor 2
Total
Factor 1
Factor 2
Total
–
0.62
0.90
0.62
–
0.89
0.90
0.89
–
All po 0.001.
factor (self-oriented metacognition) and 0.91 of the global scale
(total metacognition), exceeding the 0.70 criterion (Clark, 1995;
Ullman, 1996; Kline, 1998; Netemeyer et al., 2004). For the three
scales, no item, if deleted, produces higher Cronbach’s alphas.
CITCs for all the items are largely higher than 0.20 (Clark, 1995;
Ullman, 1996; Kline, 1998; Netemeyer et al., 2004). Pearson
correlations among the MAI scales are reported in Table 5; all
correlations are statistically significant. Taken together, these
results are consistent with the hypothesis of a homogeneous,
multidimensional construct.
4. Discussion
Our study aimed first to investigate the chance of successfully
and reliably assessing metacognitive skills through the use of a
semi-structured interview about a real-life episode. Results
regarding the preliminary statistical steps for the validation
of a psychometric instrument are encouraging. The MAI can
provide good inter-rater reliability, factorial validity and internal
consistency.
Our second aim was to test the idea that two partially different
and separate domains, i.e. self-reflection and understanding of
others’ minds, can be identified in the overall human capacity to
detect and reason about mental states. Findings only partially
supported the hypothesis. Specifically, the hypothesized domains
were self-reflection and understanding others’ minds. The CFA
yielded a two-factor solution but did not confirm the fit of the 16
facets into the domains as theoretically expected. On Factor 1,
which could resemble the Other domain, were loaded all four
items of the DEC dimension, two items of the DIF dimension,
5
‘consider one’s own representation of the world as subjective and
questionable’ and ‘give plausible interpretations of events’, and
one item from the INT dimension, ‘describes transitions among
different mental states and explains the reasons why’. On Factor
2, which could resemble the Self domain, were loaded: all four
items of the monitoring domain and three items of the integration
domain.
A possible explanation for the items of the differentiation
domain is that this subfunction measures two different levels of
metacognitive processing. One level concerns the ability to
distinguish between imagination or fantasy and reality. As
observed by Bateman and Fonagy (2004), when this function is
damaged, imagination and fantasy tend to be considered as real,
reducing consequent reflective functioning and increasing the
tendency for action. The ability to discriminate between fantasy
and reality seems to be strongly related to the monitoring in the
Self domain (as ‘I am aware’, ‘I am imagining’) and its dysfunction
will come along with a tendency to act out.
The two items of differentiation which fall into others’ domain
could be considered as a manifestation of the false belief knowledge; in other words, they require some awareness that behaviour could be guided by beliefs, which are also sometimes
wrong. Such awareness needs to take into account that others
might hold beliefs and perspectives and make evaluations different from ours. Therefore it is not surprising that these two items
are located in the Other domain. If these results are confirmed by
further studies, a revision and distinction between the two types
of differentiation would also be necessary for the MAS.
All the D facets were loaded on the same factor, as well as two
facets of the DIF dimension. These results fit into out theoretical
model. With regard to the integration facet unexpectedly loaded
on Factor 1 — ‘describes transitions among different mental
states and explains the reasons why’ — a possible explanation
is that this ability requires more complex inferential load in order
to integrate individual mental representations with the representations of others. The reason why this item falls into the Other
domain could be that the understanding of others’ mental states,
without a direct access to these states as in monitoring, requires
more inferential effort.
Even the loading of all the MON facets and the three of the four
I facets on factor 2 fits into our theoretical model. These data are
still consistent with the idea that, in order to be skilfully aware of
ourselves, on one hand we need to describe what we feel and
think in order to form a consistent picture of ourselves as time
and context change, and on the other hand we need to successfully form a coherent self-narrative that helps us identify basic
aspects of our experience and realize whether we acted consistently or not with our recurrent patterns of behaviour (Dimaggio
et al., 2008).
In conclusion as theoretically hypothesized, the metacognitive
construct, as measured by the MAI, seems to be supported by one
multifaceted latent variable. Our results confirm that two higherorder factors (domains), the Self and the Other, exist but some
elements of self-reflection are deeply intertwined with reflection
in the mind of the others. These findings may yield relevant
information for clinical practice, as they are likely to inform the
clinician where to target the interventions as different aspects of
the metacognitive system are compromised.
We think that the initial findings from the present study are
promising. In this study we found that in a sample of healthy
subjects the MAI showed good inter-rater reliability, factorial
validity and internal consistency. The MAI takes approximately
40–50 min, as well as lots of clinical interviews; with its prompts
and probes to elicit further information and to appraise and judge
any metacognitive deficits, the MAI may help to overcome specific
problems of self-report measurements for assessing ability
Please cite this article as: Semerari, A., et al., The development of the Metacognition Assessment Interview: Instrument description,
factor structure and reliability in a non-clinical sample. Psychiatry Research (2012), http://dx.doi.org/10.1016/j.psychres.2012.07.015
6
A. Semerari et al. / Psychiatry Research ] (]]]]) ]]]–]]]
regarding social cognition such as not being aware of deficits in
one’s own mental state. Moreover, the interview evaluates how
the subject actually uses metacognition or fails to cope with
stress, because of metacognitive deficits, within troubling real-life
situations. Thanks to this feature the MAI resembles a psychometric tool that helpfully supports clinical practice.
There are many limitations to our study. For example, further
examinations of the psychometric properties are required, and in
particular concurrent validity, test–retest stability, inter-rater
reliability with a larger sample and predictive validity should be
tested. Samples of 200 and larger are typically recommended for
factor analysis procedures, and results from CFA have to be
considered provisional for the same reason. Data about the
concurrent validity of the instrument with the above listed
variables in a non-clinical sample are now under evaluation.
Furthermore, additional studies are needed to further confirm
the construct validity of the MAI in other samples, particularly
those from psychiatric populations.
References
Allen, J.G., Bateman, A.W., Fonagy, P., 2008. Mentalizing in Clinical Practice.
American Psychiatric Publishing Inc., Washington DC.
American Psychiatric Association, 2000. Diagnostic and Statistical Manual of
Mental Disorders. Text Revision. DSM IV—TR, 4th ed. APA Press, Washington,
DC.
Baron-Cohen, S., 1995. Mindblindness. An Essay on Autism and Theory of Mind,
Cambridge MA. MIT Press.
Baron-Cohen, S., Leslie, A.M., Frith, U., 1985. Does the autistic child have a ‘theory
of mind’? Cognition 21, 37–46.
Bateman, A., Fonagy, P., 2004. Psychotherapy for Borderline Personality Disorder.
Oxford University Press, Oxford, UK.
Carcione, A., Dimaggio, G., Fiore, D., Nicolo, G., Procacci, M., Semerari, A., 2008. An
intensive case analysis of client metacognition in a good-outcome psychotherapy: Lisa’s case. Psychotherapy Research 18 (6), 667–676.
G., Fiore, D., Procacci, M., Semerari, A.
Carcione, A., Dimaggio, G., Conti, L., Nicolo,
Metacognition Assessment Scale (MAS) V.4.0-Manual. Terzocentro, Rome,
unpublished.
G., Pedone, R., Popolo, R., Conti, L., Fiore, D., Procacci, M.,
Carcione, A., Nicolo,
Semerari, A., Dimaggio, G., 2011. Metacognitive mastery dysfunctions in
personality disorder psychotherapy. Psychiatry Research 190 (1), 60–71.
Cattel, R.B., 1966. The scree test for the number of factors. Multivariate Behavioural Research 1, 245–276.
Clark, L.A., Watson, D., 1995. Constructing validity: basic issues in objective scale
development. Psychological Assessment 7, 309–319.
G., Semerari, A., 2008. Know
Dimaggio, G., Lysaker, P.H., Carcione, A., Nicolo,
yourself and you shall know the othery to a certain extent: multiple paths of
influence of self-reflection on mindreading. Consciousness and Cognition 17
(3), 778–789.
G., Conti, L., Fiore, D., Pedone, R., 2009. Impaired
Dimaggio, G., Carcione, A., Nicolo,
decentration in personality disorder: a series of single cases analysed with the
metacognition assessment scale. Clinical Psychology and Psychotherapy 16
(5), 450–462.
First, M.B., Gibbon, M., Spitzer, R.L., Williams, J.B.W., Benjamin, L.S., 1997.
Structured Clinical Interview for DSM-IV Axis II Personality Disorders, (SCIDII). American Psychiatric Press Inc, Washington DC.
Flavell, J.H., 1976. Metacognition aspects of problem solving. In: Resnick, L. (Ed.),
The Nature of Intelligence. Erlbaum, Hillsdale, NJ.
Floyd, F.J., Widaman, K.F., 1995. Factor analysis in the development and refinement of clinical assessment instruments. Psychological Assessment 7,
286–299.
Fonagy, P., 1991. Thinking about thinking: some clinical and theoretical considerations in the treatment of a borderline patient. International Journal of
Psychoanalysis 72, 639–656.
Fonagy, P., Target, M., 1997. Attachment and reflective function: their role in selforganization. Development and Psychopathology 9 (4), 679–700.
Fonagy, P., Target, M., 2006. The mentalization-focused approach to self pathology.
Journal of Personality Disorders 20 (6), 544–576.
Frith, C., 1992. The Cognitive Neuropsychology of Schizophrenia. Erlbaum, Hillsdale, NJ.
Frith, C., Frith, U., 2006. The neural basis of mentalizing. Neuron 50, 531–534.
Frith, U., Happé, F., 1999. Theory of mind and self consciousness: what is it like to
be autistic? Mind & Language 4, 1–22.
Gorsuch, R.L., 1983. Factor Analysis. Lawrence Erlbaum, Hillside.
Horn, J.L., 1965. A rationale and test for the number of factors in factor analysis.
Psychometrika 30, 179–185.
Kline, R.B., 1998. Principles and Practice of Structural Equation Modeling. Guilford
Press, New York.
G., Procacci, M.,
Lysaker, P.H., Carcione, A., Dimaggio, G., Johannesen, J.K., Nicolo,
2005. Metacognition amidst narratives of self and illness in schizophrenia:
associations with neurocognition, symptoms, insight and quality of life. Acta
Psychiatrica Scandinavica 112, 64–71.
G., 2010a. MetacogniLysaker, P.H., Dimaggio, G., Buck, K.D., Carcione, A., Nicolo,
tion and schizophrenia: the capacity for self-reflectivity and prospective
assessments of work performance over six months. Schizophrenia Research
122, 124–130.
Lysaker, P.H., Dimaggio, G., Daroyanni, P., Buck, K.D., LaRocco, V.A., Carcione, A.,
2010b. Assessing metacognition in schizophrenia with the metacognition
assessment scale: associations with the social cognition and object relations
scale. Psychology and Psychotherapy 83, 303–315.
Lysaker, P.H., Warman, D.M., Dimaggio, G., Procacci, M., LaRocco, V.A., Clark, L.K.,
2008. Metacognition in schizophrenia associations with multiple assessments
of executive function. Journal of Nervous and Mental Disease 196 (5),
384–389.
Lysaker, P.H., Erickson, M.A., Buck, B., Buck, K.D., Olesek, K., Grant, M., 2011a.
Metacognition and social function in schizophrenia: associations over a period
of five months. Cognitive Neuropsychiatry 16 (3), 241–255.
Lysaker, P.H., Kyle, L., Olesek, L., Warman, D.M., Martin, J.M., Salzman, A.K., 2011b.
Metacognition in schizophrenia: correlates and stability of deficits in theory of
mind and self-reflectivity. Psychiatry Research 190 (1), 18–22.
Main, M., 1991. Metacognitive knowledge, metacognitive monitoring, and singular
(coherent) vs. multiple (incoherent) models of attachment: findings and
directions for future research. In: Stevenson-Hinde, J., Marris, P. (Eds.),
Attachment Across the Life Cycle. Routledge, pp. 127–159.
Netemeyer, R.G., Bearden, W.O., Sharma, S., 2004. Scaling Procedures: Issues and
Applications. Sage Publications, Thousand Oaks.
Ogrodniczuk, J.S., Piper, W.E., Joyce, A.S., 2011. Effect of alexithymia on the process
and outcome of psychotherapy: a programmatic review. Psychiatry Research
190 (1), 43–48.
Premack, D., Woodruff, G., 1978a. Chimpanzee problem-solving—test for comprehension. Science 202 (4367), 532–535.
Premack, D., Woodruff, G., 1978b. Does chimpanzee have a theory of mind.
Behavioral and Brain Sciences 1 (4), 515–526.
Premack, D., Woodruff, G., 1978c. Cognition and consciousness in nonhuman
species—reply. Behavioral and Brain Sciences 1 (4), 616–628.
G., Procacci, M., 2007. UnderSemerari, A., Carcione, A., Dimaggio, G., Nicolo,
standing minds: different functions and different disorders? The contribution
of psychotherapy research. Psychotherapy Research 17 (1), 106–119.
G., Procacci, M., 2003.
Semerari, A., Carcione, A., Dimaggio, G., Falcone, M., Nicolo,
How to evaluate metacognitive functioning in psychotherapy? The metacognition assessment scale and its applications. Clinical Psychology & Psychotherapy 10, 238–261.
G., Pedone, R., Procacci, M., 2005.
Semerari, A., Carcione, A., Dimaggio, G., Nicolo,
Metarepresentative functions in borderline personality disorder. Journal of
Personality Disorders 19 (6), 609–710.
Semerari, A., D0 Angerio, S., Popolo, R., Cucchi, M., Ronchi, P., Maffei, C., Dimaggio,
G., Nicoló, G., Carcione, A., 2008. L0 Intervista per la Valutazione della Metacognizione (IVaM): Descrizione dello Strumento [The Metacognition Assessment Inerview (MAI): descriptionn of the instrument]. Cognitivismo Clinico 5,
174–192.
Sperber, D., 2000. Metarepresentations. Oxford University Press, Oxford.
Ullman, J., 1996. Structural equation modeling. In: Tabachnik, B., Fidell, L. (Eds.),
Using Multivariate Statistics. Harper Collins, New York, pp. 709–811.
Wellman, H.M., 1990. The Child’s Theory of Mind. Bradford Books. MIT Press,
Cambridge, MA..
Wells, A., 2000. Emotional Disorders and Metacognition, Innovative cognitive
therapy. John Wiley & Sons Ltd., UK.
Please cite this article as: Semerari, A., et al., The development of the Metacognition Assessment Interview: Instrument description,
factor structure and reliability in a non-clinical sample. Psychiatry Research (2012), http://dx.doi.org/10.1016/j.psychres.2012.07.015