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

net/publication/12440332

Inventory of Interpersonal Problems: A Three-

Dimensional Balanced and Scalable 48-Item Version

Article  in  Journal of Personality Assessment · May 2000

DOI: 10.1207/S15327752JPA7402_9 · Source: PubMed

CITATIONS READS

35 3,159

4 authors, including:

Torbjorn Moum Svein Friis

University of Oslo University of Oslo
197 PUBLICATIONS   7,258 CITATIONS    316 PUBLICATIONS   8,479 CITATIONS   

SEE PROFILE SEE PROFILE

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

Victimization in patients with first episode psychosis headed by Johannes Langeveld View project

The early detection and Intervention in Psychosis (TIPS) study View project

All content following this page was uploaded by Svein Friis on 02 February 2016.

The user has requested enhancement of the downloaded file.

This article was downloaded by: [ ]
On: 17 January 2012, At: 06:11
Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954
Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,
UK

Journal of Personality
Assessment
Publication details, including instructions for
authors and subscription information:
http://www.tandfonline.com/loi/hjpa20

Inventory of Interpersonal
Problems: A Three-Dimensional
Balanced and Scalable 48-Item
Version
Tore Gude, Torbjorn Moum, Eystein Kaldestad &
Svein Friis

Available online: 10 Jun 2010

To cite this article: Tore Gude, Torbjorn Moum, Eystein Kaldestad & Svein Friis
(2000): Inventory of Interpersonal Problems: A Three-Dimensional Balanced and
Scalable 48-Item Version, Journal of Personality Assessment, 74:2, 296-310

To link to this article: http://dx.doi.org/10.1207/S15327752JPA7402_9

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-

and-conditions

This article may be used for research, teaching, and private study purposes.
Any substantial or systematic reproduction, redistribution, reselling, loan,
sub-licensing, systematic supply, or distribution in any form to anyone is
expressly forbidden.

The publisher does not give any warranty express or implied or make any
representation that the contents will be complete or accurate or up to
date. The accuracy of any instructions, formulae, and drug doses should be
independently verified with primary sources. The publisher shall not be liable
for any loss, actions, claims, proceedings, demand, or costs or damages
 whatsoever or howsoever caused arising directly or indirectly in connection
with or arising out of the use of this material.
Downloaded by [ ] at 06:11 17 January 2012
 JOURNAL OF PERSONALITY ASSESSMENT, 74(2), 296–310
Copyright © 2000, Lawrence Erlbaum Associates, Inc.

Inventory of Interpersonal Problems: A

Three-Dimensional Balanced and
Scalable 48-Item Version

Tore Gude and Torbjørn Moum

Department of Behavioural Science in Medicine
University of Oslo, Norway
Downloaded by [ ] at 06:11 17 January 2012

Eystein Kaldestad
Department of Research
Modum Bads Nervesanatorium
Vikersund, Norway

Svein Friis
Department of Research and Education
Ullevål Hospital
Oslo, Norway

Interpersonal relating has been a focus of attention in psychiatry for decades. To address
this domain, a self-rating scale, the Inventory of Interpersonal Problems (IIP; Horowitz,
Rosenberg, Baer, Ureño, & Villaseñor, 1988), was developed. Analysis of the
psychometric properties of IIP presented in this article was performed by principal
component analysis (PCA) for the purpose of obtaining subscales with a balanced, bi-
polar dimensionality. The model was validated by the resulting dimensions’ ability to
discriminate among different categories of personality disorders (PDs). The problem of
a General Complaint factor affecting PCAs of questionnaires such as the IIP is dis-
cussed thoroughly, and ways of avoiding the problem are outlined. We present a three-
dimensional structure of the IIP with both theoretically appealing and statistically ro-
bust dimensions of Assertiveness, Sociability, and Interpersonal Sensitivity based on
48 (out of 127) items. Balanced, additive indexes using the subset of 48 items appeared
psychometrically sound by showing much lower correlations internally and less con-
founding from the General Complaint factor than extant indexes derived from the IIP.
External validity seemed to be bolstered by all subscales’ discriminating significantly
between different PDs versus no PDs, on both cluster and single diagnosis levels. Our
analysis seemed to substantiate the reliability (scalability) of three dimensions of the IIP
tapping different areas of the interpersonal relational field.
 INVENTORY OF INTERPERSONAL PROBLEMS 297

In every form of psychotherapy, there is a need for a multidimensional approach in

theoretically based clinical assessments and case formulations. Interpersonal the-
ory has gained increasing importance; Birtchnell (1993), in his book How Humans
Relate—A New Interpersonal Theory, gives an extended overview of the theories
concerning interpersonality with emphasis on the two-axial model of relating,
where the one axis represents the dominance–submissiveness or upperness–
lowerness dimension and the other, the warm–cold or closeness–distance dimen-
sion. Horowitz (1979), in his use of a multidimensional scaling of statements made
by patients in psychotherapy, identified three major dimensions in the interper-
sonal domain:

1. Degree of involvement (which can be interpreted as “intrusive versus

avoidant”).
Downloaded by [ ] at 06:11 17 January 2012

2. Intention of involvement (which can be interpreted as “dominant versus

submissive”).
3. Nature of involvement (which can be interpreted as “friendly versus hostile”).

The relation between interpersonal relating styles and personality disorders

(PDs) has been focused on by Pilkonis, Kim, Proietti, and Barkham (1996), among
others, who asserted that one of the best overall markers of PDs is chronic diffi-
culty in interpersonal management. This highlights the theme of specificity among
the diagnostic categories of the Diagnostic and Statistical Manual of Mental Dis-
orders (3rd ed., rev. [DSM–III–R]; American Psychiatric Association, 1987) Axis
II system for diagnosing PDs (Ekselius, Lindstrøm, von Knorring, Bodlund, &
Kullgren 1993; Livesley, 1987; Widiger, Frances, Spitzer, & Williams, 1988).
As an instrument for assessment in the area of interpersonal functioning that
could both widen personality dimensions and act as a supplement to diagnostic
procedures and measures of symptom distress, Horowitz et al. (1988) developed
the Inventory of Interpersonal Problems (IIP). Previous research on IIP has shown
a general agreement that the instrument taps several types of interpersonal prob-
lems, but it has also shown disagreement as to exactly how many distinct dimen-
sions are represented.
Horowitz et al. (1988) studied the psychometric properties and clinical applica-
tions in a sample of relatively well-functioning outpatients (N = 103) on waiting lists
for brief psychotherapy, ending up with six factors or subscales in a principal com-
ponent analysis (PCA) with eigenvalues exceeding 3.0 (they included items with
factor loadings above .40). The six factors were H(ard to be). Assertive, H. Sociable,
H. Intimate, H. Submissive, Too Responsible, and Too Controlling. With a Varimax
rotation of the subscale means (second-order factor analysis), two factors emerged,
corresponding to five of the authors’ subscales, with internal correlations varying
from .13 to .61. In a later article, Horowitz, Rosenberg, and Bartholomew (1993)
elaborated further on the structure of IIP in a circumplex model.
 298 GUDE, MOUM, KALDESTAD, FRIIS

In their article on the investigation of the psychometric properties, validation,

and clinical relevance of IIP, Barkham, Hardy, and Startup (1994) started with a
critical evaluation of the work by Horowitz et al. (1988), focusing on its too-small
sample size (N = 103), skewed sex distribution (86% female, 14% male), and too-
high cutoff in eigenvalues (> 3.0) for factor extraction. In a PCA on raw scores
from the responses to IIP in an outpatient sample (N = 250), they obtained eight
unipolar factors, and by using ipsatized scores (every item’s deviation from the
overall mean), they obtained four bipolar factors as a basis for subscales tapping
the domains of Problems With Socializing (H. Sociable or Too Open), Problems
With Assertiveness (H. Assertive or Too Aggressive), Problems With Independ-
ence (H. Involved or Too Dependent), and Problems With Nurturance (H. Sup-
portive or Too Caring), without presenting these bipolar factors as clinically
relevant additive indexes.
Downloaded by [ ] at 06:11 17 January 2012

Barkham and Hardy (1996) also presented a short version of IIP, consisting of
32 items equally distributed on eight subscales (from their analysis of IIP in an ear-
lier publication, Barkham et al., 1994), primarily motivated by convenience in
clinical practice. They concluded, after exploratory and confirmatory analyses of
this short version, that hardly any of the psychometric properties were sacrificed
compared to the 127-item version.
Previous attempts at establishing psychometrically sound and clinically mean-
ingful dimensions and subscales of the IIP have resulted in a large number of pur-
portedly distinct factors that are hard to distinguish from one another in substantive
and theoretical terms. These purportedly distinct factors turn out to show moderate
to high correlations among themselves when one proceeds to compute subscale in-
dex scores (by adding raw scores for the appropriate items). Most of the problems
seem to be rooted in the very way in which the IIP has been constructed. In the origi-
nal 127-item version, the first 78 items were formulated to begin “It is hard for me
to,” and the remaining 49 items were formulated to begin “This I do too much.” Such
a design tends to elicit ratings that are generally high or low (on a scale of 0 to 4), re-
gardless of the content of individual items. A General Complaint factor running
through all items tends to create positive correlations between items that from a theo-
retical and clinical perspective would appear to be substantively very different and
therefore should be inversely related (e.g., the correlation between “It is hard for me
to feel like a separate person when I am in a relationship” and “It is hard for me to feel
close to other people,” which was .32, p < .001).
This pattern, in turn, makes it difficult to establish dimensionality with standard
factor-analytic models based on raw scores. In particular, it becomes almost impos-
sible to obtain balanced factor solutions, that is, factors with positive and negative
factor loadings from items tapping opposite ends of the same underlying clinically
and substantively meaningful dimension. Instead, factors with only positive load-
ings will emerge. The number of dimensions present in the set of items will also ap-
pear to be inordinately high according to standard criteria (e.g., eigenvalues).
 INVENTORY OF INTERPERSONAL PROBLEMS 299

Creating subscales based on unbalanced sets of items (typically, with a majority

of items with positive loadings) will confound the subscales with the General
Complaint factor and create very strong correlations among the various subscales.
Typically, correlations between subscales in previous research have not been re-
ported. To avoid the factor-analytic complications arising from the General Com-
plaint factor, a rather effective strategy is to rely on ipsatized scores. This tends to
produce factors with a relatively balanced distribution of positive and negative
factor loadings. In other words, ipsatized scores allow dimensionality to emerge
with greater clarity.
However, for other purposes, such as comparison of patient groups and analy-
ses of change at the individual level, the use of ipsatized scores to create subscale
indexes has some limitations because ipsatized scores are based on deviations
from the individual patient’s own overall item mean, so that differences between
Downloaded by [ ] at 06:11 17 January 2012

groups of patients and change over time tend to be suppressed.

Ipsatized scores appear to be the preferable input in PCAs aimed at establishing
a clinically meaningful factor structure in a series of items using the response for-
mat of the IIP. Working with the resulting factor structure, one should then try to
identify a balanced set of items for each dimension, that is, a set containing an
equal number of items with negative and positive loadings. This balancing of sets
is necessary for the next step, in which one wishes to use raw scores to create sub-
scale indexes. If an unequal number of negatively and positively keyed items is
used to create the subscale indexes, confounding by the General Complaint factor
will necessarily ensue and most subscales will show high intercorrelations. The
General Complaint factor will make it more difficult to obtain satisfactory alpha
reliability for the subscale index, however, when half of the items (i.e., items with
negative factor loadings) have been reversed for substantive content. Within each
set of items (i.e., within the positively keyed and negatively keyed halves, respec-
tively), the average correlations between individual items will be high, but across
sets, correlations between items will be rather low (because the General Complaint
factor in this case pulls in opposite directions within each pair of items). Thus,
when half of the items have been reversed before item scores are added, the size of
the overall average correlation between all the items that make up the subscale
score will tend to be reduced. It is, of course, this overall average interitem (inter-
indicant) correlation that is used to compute Cronbach’s alpha. Therefore, a proce-
dure reversing some of the items when all items are infected with a bias (e.g.,
General Complaint) will result in low estimates of reliability in additive indexes.
It is therefore very unlikely that balanced raw-score subscales can include all
the items of the IIP. A selection of items has to be made to satisfy standard
psychometric criteria. This procedure would also give us the opportunity to test
whether the original IIP form (127 items), with its six unbalanced, unipolar sub-
scales, taps general psychopathology or specific interpersonal dimensions. Based
on the previous considerations, the following research questions are addressed:
 300 GUDE, MOUM, KALDESTAD, FRIIS

1. How many dimensions with clinical meaningfulness should be extracted in a

PCA of IIP in a sample of hospitalized patients with a marked impact of PDs?
2. Can these dimensions (components from PCA) be represented with suffi-
cient internal consistency by balanced, bipolar subscales based on sets of raw item
scores?
3. Will the dimensional structure show cross-sample stability?
4. Will the balanced subscales be able to discriminate between groups of pa-
tients, for example, between different categories of PDs?
5. To what degree will balanced bipolar subscales (derived from a subset of IIP)
be more “robust,” that is, show lower correlations with the General Complaint fac-
tor and closer statistical associations with PDs (controlled for the General Com-
plaint factor) than unbalanced unipolar subscales?
Downloaded by [ ] at 06:11 17 January 2012

MATERIAL AND METHOD

The patients in our sample (N = 646) were treated in Modum Bads Nervesanatorium,
a national psychiatric hospital, during the years 1990 to 1994. For characteristics of
the sample, see Table 1. The DSM–III–R diagnostic system was used, and all pa-
tients were interviewed with the corresponding Structured Clinical Interview for
DSM–III–R Axes I and II-1 (SCID I and II; Spitzer & Williams, 1988) during the
first few weeks. The Axis I interview took place immediately, and the Axis II inter-
view took place after the most distressing symptom conditions had been treated.
Diagnostic reliability was tested and found satisfactory for both Axis I (Hoffart &
Martinsen, 1992) and Axis II (Gude, Dammen, & Friis, 1997).
The 127-item version IIP, with a 5-point scale ranging from 0 (not at all) to 4
(extremely) was administered on four occasions, at preassessment (3 to 9 months
before admission), pretreatment (on admission), posttreatment (close to dis-
charge), and 1-year follow-up (by means of mailed forms). Because the question-
naire was not introduced at the very beginning of the study, patients were exposed
to it for the first time at different points during the course of the study. To get the
patients’ immediate responses to the questions, we selected every patient’s “first-
time” items (from the very first time they filled in the form); for 30.2% of the pa-
tients (n = 195), this was at the time of preassessment; for 18.7% (n = 121), this
was at pretreatment; for 3.4% (n = 22), this was at posttreatment; and for 47.7% (n
= 308), this was at follow-up. These first-time items were ipsatized, and to enable
statistical investigation of the factor structure of this form, a PCA with listwise de-
letion of missing values was applied initially to establish dimensionality. Based on
the PCA, we selected a balanced subset of positively and negatively keyed items,
loading above .30 or below –.30, to represent the hypothesized dimensions. If an
item showed loadings above .30 or below –.30 on more than one factor, the differ-
ence was not to be less than .15. These subsets of items were then used to create
 INVENTORY OF INTERPERSONAL PROBLEMS 301

TABLE 1
Characteristics of the Sample

Characteristic % n M SD Range

Sex
Female 67 430
Male 33 216
Age (years) 38.2 10.5 17–69
Treatment time (days) 86 40 2–221
Any diagnosis on Axis I 80.1 515
Any depressive diagnosis 59.4 384
Any anxiety diagnosis 54.6 353
Axis I diagnoses 2.8 1.7 1–10
Any diagnoses on Axis II 55.9 361
Cluster A 9.8 63
Downloaded by [ ] at 06:11 17 January 2012

Cluster B 21.4 138

Cluster C 45.4 293
Schizo PD (schizoid and schizotypal PD) 1.7 11
Paranoid PD 8.2 53
Borderline PD 15.3 99
Histrionic PD 5.9 38
Narcissistic PD 2.5 16
Antisocial PD 1.4 9
Avoidant PD 26.6 172
Dependent PD 16.3 105
Obsessive compulsive PD 11.9 77
Passive aggressive PD 3.9 25
PD not otherwise specified 6.3 41
Axis II diagnoses 1.8 1.0 1–6

Note. N = 646. PD = personality disorder.

raw score subscales, and we stipulated that these subscales should achieve satis-
factory reliability as indexed by internal consistency (Cronbach’s αs around .70).
To establish the robustness of the dimensions suggested by the PCAs, multiple
cross-sample tests (i.e., independent PCAs) were performed on 10 randomly se-
lected subsamples (50%) of the entire material, using the same selection criteria as
in the initial PCA on the 127-item version.
Axis II diagnostic categories (on cluster-level, hierarchically selected, and on
single-diagnosis level, computed as “pure” conditions, i.e., no other Axis II diag-
noses present) were coded as dichotomous variables and used as external validity
criteria for the dimensions presumably tapped by the raw score subscales.
We wished to check our model against the models presented by Horowitz et al.
(1988) and Barkham et al. (1994), concerning not only the pureness of the subscales
developed but also the strength of discriminative ability among diagnostic catego-
ries in the three models. The influence of the General Complaint factor could be in-
 302 GUDE, MOUM, KALDESTAD, FRIIS

vestigated through control for the raw score means of the 127-item IIP. For this
purpose, we used bivariate correlations and partial correlations by partialling out the
127-item IIP’s overall mean of the raw score items. We also used analyses of vari-
ance (simple factorial) for testing discriminative ability to see if our model was less
influenced by the General Complaint factor and overall symptom distress than were
previous solutions. PCA with Varimax rotation was used to investigate factor struc-
ture. SPSS (Version 6.1.3) was used for all analyses.

RESULTS

The scree plot from the PCA performed on the ipsatized item scores did not clearly
identify a specific number of factors. Two-, three-, and four-factor solutions ap-
Downloaded by [ ] at 06:11 17 January 2012

peared as possible candidates. We therefore decided to investigate all of these op-

tions more thoroughly. The substantive content of the items emerging from the
two-factor solution of the PCA seemed to lack intuitive appeal, with too-heteroge-
neous subscales. This solution was therefore excluded from further analyses. The
stability of the dimensions in the three- and four-factor solutions was checked in
several ways. The first step was to obtain a balanced subset with an equal number of
items from each factor loading positively and negatively. Because only 8 items
loaded positively above .30 on one of the factors in the three-factor solution, we had
to select 8 negatively loading items from the same factor and 8 positively and nega-
tively loading items from each of the other factors. With our selection criteria
(items with highest factor loadings > .30 or lowest < –.30; if loading on more than
one factor > .30 or < –.30, the difference was not to be < .15), a subset of 48 items re-
mained. These items were entered in a second PCA (Varimax rotation, three factors
extracted), replicating the same balanced dimensional structure according to item
distribution as in the PCA on all 127 items. The factor scores from this last analysis
were correlated to the factor scores from the first PCA on all 127 items, yielding
values from .92 to .95. The three bipolar subscales’ additive indexes (from the 48-
item design), which were based on raw scores, showed internal consistency as mea-
sured by Cronbach’s alpha ranging from .69 to .80 (see Table 2).
When repeating this procedure for the four-factor solution, we had to reduce the
number of items to 30, because only 5 items had positive loadings above .30 on
one of the factors. With the subset of 30 items in the next PCA (Varimax rotation,
four factors extracted), we were not able to replicate the solution; this solution was
therefore excluded from further analyses. We then checked the three-factor model
in 10 multiple, randomly selected cross-sample PCA tests, obtaining an average of
78% of the best loading 48 items in each PCA (using the same selection criteria as
described before) overlapping with the items selected in the initial PCA.
To check the validity of the raw score first-time additive indexes for each of the
dimensions—Assertiveness, Sociability, and Interpersonal Sensitivity—we corre-
 Downloaded by [ ] at 06:11 17 January 2012

TABLE 2
Factor Loadings for the 48-Item Subset Selected From the PCA on 127 Items

48-Item Solution 127-Item Solution

Dimension and Item Key Words Factor 1 Factor 2 Factor 3 Factor 1 Factor 2 Factor 3

Assertivenessa
90 Too aggressive .71 .18 –.06 .63 .24 –.01
79 Fight too much .66 .24 .12 .60 .29 .02
112 Loose temper too much .64 .28 .08 .54 .26 .16
116 Argue too much .63 .36 –.04 .60 .37 .02
96 Too critical .55 .31 –.08 .51 .31 –.04
82 Too irritated .52 .09 –.11 .43 .08 .19
127 Too revengeful .45 .25 .14 .45 .23 –.08
123 Feel competitive .39 .21 –.20 .42 .26 –.12
14 It is hard for me to reveal anger –.64 –.06 –.14 –.58 –.05 –.19
68 It is hard for me to feel anger –.59 –.03 –.01 –.56 .05 –.08
20 It is hard for me to be aggressive –.58 –.04 –.03 –.54 –.04 –.10
33 It is hard for me to be firm enough –.56 –.01 .05 –.53 .02 .01
6 It is hard for me to tell people off –.48 .06 –.25 –.51 .02 .10
36 It is hard for me to set limits –.48 .06 .14 –.53 .07 .05
2 It is hard for me to say “no” –.48 .26 .13 –.49 .23 .10
73 It is hard for me to care to own needs –.44 .16 .19 –.51 .10 .12
Sociabilityb
88 Too open .21 .55 .01 .22 .53 .07
113 Too much self-disclosure .24 .53 .03 .24 .54 .11
109 Give too much compliments .17 .49 .12 .21 .41 .16
97 Too trustful .14 .41 .06 .15 .39 .08
103 Too gullible –.04 .40 –.14 –.03 .34 .17
120 Too exploitable –.21 .39 .24 –.28 .33 .28
104 Too generous .01 .39 .09 .05 .31 .14
4 It is hard for me to keep things private .11 .33 –.11 .15 .30 .04

303
27 It is hard for me to feel comfortable around others .01 –.68 –.14 –.06 –.64 –.18
(Continued)
 Downloaded by [ ] at 06:11 17 January 2012

TABLE 2 (Continued)

304
48-Item Solution 127-Item Solution

Dimension and Item Key Words Factor 1 Factor 2 Factor 3 Factor 1 Factor 2 Factor 3

23 It is hard for me to socialize –.02 –.66 –.15 .06 –.59 –.14

42 It is hard for me to relax in company –.08 –.61 .02 –.05 –.54 –.02
78 It is hard for me to be self-confident –.09 –.60 –.22 –.10 –.57 .17
99 Avoiding others too much .04 –.57 .01 .09 –.53 .03
105 Too afraid of others .03 –.57 –.23 .01 –.59 .22
3 It is hard for me to join in .02 –.56 .04 –.01 –.54 .01
7 It is hard for me to introduce myself .01 –.53 .01 –.02 –.52 .05
Interpersonal Sensitivityc
122 Too guilty about failures .01 .00 .57 –.07 –.10 .52
93 Too guilty about doings –.05 –.12 .55 –.09 –.20 .46
110 Too much worry about disappointing –.11 –.00 .51 –.10 –.06 .48
106 Too much worry about others’ reactions –.03 –.23 .47 –.04 –.26 .47
114 Too much blame myself for causing others’ problems .07 .17 .40 .03 .09 .40
87 Too sensitive to rejection .10 .00 .40 .04 –.02 .37
80 Too sensitive to criticism .09 –.04 .37 .04 –.03 .37
100 Too sensitive to others mood .08 .07 .36 .03 .08 .37
65 It is hard for me to feel good about others .20 .02 –.57 .33 .09 –.47
48 It is hard for me to give gifts .15 .09 –.58 .22 .14 –.48
62 It is hard for me to give credit to others .16 –.02 –.58 .27 .01 –.50
11 It is hard for me to express my admiration for others .07 –.16 –.57 .14 –.13 –.47
40 It is hard for me to really care about others .16 .08 –.57 .23 –.05 –.49
26 It is hard for me to show affection –.01 –.31 –.50 .01 –.31 –.52
38 It is hard for me to be supportive to others .24 .07 –.50 .29 .13 –.44
46 It is hard for me to keep up friendships .12 –.21 –.42 .17 –.16 –.37

Note. PCA = principal component analysis. Boldface indicates the highest factor loadings on the respective factor to which each item belongs. The key words
in this table are reprinted from the Inventory of Interpersonal Problems. Copyright © 1979 by The Psychological Corporation. Reproduced by permission. All
rights reserved.
aα = .80. bα = .74. cα = .69.
 INVENTORY OF INTERPERSONAL PROBLEMS 305

lated them with the factor scores from the 48-item version of the first-time ipsatized
items (rs = .95, .95, and .96, respectively). The intercorrelations among the three
subscales did not exceed .24 (see Table 3). A series of logistic regression analyses
was performed, with the dichotomous diagnostic categories as dependent variables
(three hierarchical diagnostic clusters [A, B, and C] and six of the most frequent sin-
gle diagnoses in pure form [borderline, avoidant, dependent, obsessive–compul-
sive, and passive–aggressive being comorbid within Cluster C, and schizoid and
schizotypal PDs being comorbid within all clusters]). With the three first-time raw
score indexes as independent variables, two of the factors (Assertiveness and Socia-
bility) discriminated significantly between three of the diagnostic categories and the
no PDs, whereas the third, Interpersonal Sensitivity, discriminated between only
one diagnostic category and the no PDs (see Figures 1 and 2).
Downloaded by [ ] at 06:11 17 January 2012

TABLE 3
Correlations Between Subscales in Our Three-Dimensional Model and the
First Time Overall Means From 127-Item IIP (General Complaint Factor), Between Our
Subscales and the Subscales of Horowitz et al. (1988) and Barkham, Hardy, and
Startup (1994), and Between Their Subscales and the General Complaint Factor

Our Dimension

Interpersonal First-Time IIP

Assertiveness Sociability Senitivity Overall M

Our dimension
Assertiveness .24 (.12) –.20 (–.10) –.38
Sociability .08 (.22) –.35
Interpersonal Sensitivity .31
Horowitz’ subscales
Hard to Be Submissive .24 (.68) –.10 (.18) –.09 (–.15) .64
Hard to Be Assertive –.70 (–.81) –.34 (.01) .35 (.16) .87
Too Controlling .19 (.61) .15 (.53) .26 (.08) .65
Hard to Be Sociable –.34 (–.03) –.68 (–.82) .25 (–.04) .87
Too Responsible –.40 (–.17) –.17 (.28) .58 (.63) .86
Hard to Be Intimate –.15 (.25) –.30 (–.05) –.15 (–.65) .78
Barkham’s subscales
Hard to Be Assertive –.65 (–.74) –.33 (–.03) .39 (.26) .89
Too Aggressive –.34 (–.04) –.71 (–.83) .11 (–.31) .85
Hard to Be Supportive .02 (.35) –.21 (.02) –.26 (–.63) .62
Too Caring –.36 (–.12) .08 (.57) .50 (.43) .76
Hard to Be Sociable –.26 (.19) –.24 (.19) .51 (.56) .90
Too Open .67 (.88) .09 (.20) .09 (.00) .29
Hard to Be Involved –.29 (.00) –.33 (–.10) .06 (–.29) .76
Too Dependent .30 (.24) .61 (.58) .08 (.16) –.21

Note. IIP = Inventory of Interpersonal Problems. The General Complaint factor was partialled out
of the values in parentheses. Corresponding subscales are in bold.
Downloaded by [ ] at 06:11 17 January 2012

FIGURE 1 Raw score differences between patients with diagnosed personality disorders
(PDs) at cluster level versus higher level clusters or no PD. Differences between groups were
significant (by t test) at *p < .05, **p < .01, and ***p < .001. Values above zero mean dominant
on Assertiveness, open intrusiveness on Sociability, and overconscientious on Interpersonal
Sensitivity.

FIGURE 2 Raw score differences between patients with a specific personality disorder (PD)
versus no PD. Differences between groups were significant (by t test) at *p < .05, **p < .01, and
***p < .001. Values above zero mean dominant on Assertiveness, open intrusiveness on Socia-
bility, and overconscientious on Interpersonal Sensitivity (schizo PD may be comorbid within
all clusters; passive–aggressive PDs may be comorbid within Cluster C).

306
 INVENTORY OF INTERPERSONAL PROBLEMS 307

The comparison of Horowitz et al.’s (1988) six subscales and Barkham et al.’s
(1994) eight subscales with our three-factor solution was based on the correlations
among dimensions or subscales derived from the different models (see Table 3).
Our model and Horowitz et al.’s model overlapped to a degree that gave evidence
of congruence between the two. There was a clear tendency for substantively op-
posite pairs of Horowitz et al.’s subscales to correlate positively and negatively,
respectively, with each one of our dimensions that was closest in conceptual mean-
ing. This tendency changed to clear evidence when we partialled out the General
Complaint factor (represented by the overall item raw score mean). With Barkham
et al.’s (1994) eight subscales, we found a somewhat more mixed pattern, both in
bivariate analyses and when partialling out the General Complaint factor.
The correlations between our subscales (dimensional scores) and the General
Complaint factor from the 127-item IIP yielded values below .40, whereas in
Downloaded by [ ] at 06:11 17 January 2012

Horowitz and Barkham’s models, the correlation values were markedly higher (see
Table 3).
To test the difference in the explained variance in diagnostic categories from the
subscales in our study compared with the subscales from Horowitz and Barkham, we
ran a series of multiple linear regression analyses with the subscales as the dependent
variable, forcing the General Complaint factor into the equation before all single di-
agnostic categories in Axis II were entered as independent variables. For our three
subscales together, we obtained an additional 12.2% explained variance over and
above the variance explained by the General Complaint factor, whereas the six sub-
scales from Horowitz et al. (1988) obtained 8.2% (the three with the greatest vari-
ance giving 6.3%) and the eight subscales from Barkham et al. (1994) obtained
10.6% (the three with the greatest variance giving 8.2%).

DISCUSSION

The model in this presentation, with its three-axial solution concerning interper-
sonal relating measured with IIP, shows both differences and similarities with arti-
cles presented earlier. The main difference lies in the method by which the three-
dimensional model with balanced subscales is developed, whereas the similarities
are found in the considerable overlap with our own findings and those of other au-
thors concerning the dimensionality (factor structure) of IIP.
Based on our sample of patients, we argue for the three-dimensional solution
for the following reasons:

1. There is a very good match between empirically derived results, on the one
hand, and theoretically and clinically meaningful dimensions, on the other.
2. There is satisfactory internal consistency (reliability) for the subscales
(Cronbach’s α).
 308 GUDE, MOUM, KALDESTAD, FRIIS

3. There is structural invariance in tests involving multiple randomly selected

cross-samples.
4. There is overlap with other models, in particular with the six subscales of
Horowitz et al. (1988).
5. Our balanced, bipolar subscales have the ability to discriminate between
different categories of PDs.
6. There is more explained variance from diagnostic categories in our three
subscales than in the unbalanced, unipolar subscales of Horowitz et al.
(1988) and Barkham et al. (1994).

Concerning the match between theory and clinical concepts, we found evidence
that the concepts of degree, intention, and nature of psychological involvement
corresponded with the dimensions emerging in our three-axial model. We propose
Downloaded by [ ] at 06:11 17 January 2012

that degree of involvement in our model is represented by intrusive openness ver-

sus social avoidance on the Sociability dimension, intention is represented by
dominance versus submissiveness on the Assertiveness dimension, and nature is
represented by overconscientiousness versus neglect on the Interpersonal Sensi-
tivity dimension. The internal consistency of the three dimensions, even with the
reversed raw score values on items loading negatively in PCA, were surprisingly
high and increased our confidence that we managed to avoid the statistical prob-
lems caused by the General Complaint factor in principal component analyses and
scale construction.
Earlier we discussed possible sources of errors in procedures like the ones we
chose. In our opinion, these errors have been eliminated because we obtained suffi-
cient overlap in the cross-sample validation procedures and we were able to replicate
the model by retesting the factor structure of the 127-item version by selecting 48
items from the three factors in a new PCA. With the replicated, balanced model, we
thus avoided the problem of unequal weighting of either the positive or the negative
side of the dimensions. Because we argue for the necessity of using raw scores and
not ipsatized values for measuring change, the ability of our model in this respect de-
pends on the correlations between our raw-score indexes for each dimension and the
factor scores from the PCA based on the ipsatized item values. With Pearson’s r in
the range of .76 to .92, we found this test satisfactory. The structural invariance is
shown by the high correlation values (.95–.96) between the factor scores from the
ipsatized first-time 127-item version and the factor scores from the next PCA on the
48-item version, as well as the overlap resulting from the randomly selected multiple
cross-sample tests.
The striking overlap between our own results and the models of Horowitz et al.
(1988) and Barkham et al. (1994) was especially interesting to observe. Our results
from the bivariate and partial correlation analyses indicate that the subscales of this
previous research, when applied to our patient sample, were to a large extent influ-
enced by the General Complaint factor, thus boosting internal consistency but at the
 INVENTORY OF INTERPERSONAL PROBLEMS 309

same time reducing the ability of the subscales to discriminate between PDs. We ob-
tained a good match between the two most obvious opposite pairs of Horowitz et
al.’s subscales and our corresponding dimensions; this match became even clearer
when we partialled out the General Complaint factor. We interpret this finding as a
further indication of the construct validity of our model. We did not achieve the same
obvious match between subscales with Barkham et al.’s (1994) model.
Alden, Wiggins, and Pincus (1990) also managed to keep both satisfactory in-
ternal consistency and discriminative power in their subscales, but due to their reli-
ance on ipsatized scores when computing subscale means, the measurement of
change over time will be jeopardized.
When categories of PDs were used as external criteria to tentatively validate
our three-dimensional solution, we found that our dimensions did indeed do a
better job at discriminating among various PDs than previously constructed sub-
Downloaded by [ ] at 06:11 17 January 2012

scales derived from the IIP. These results indicate that we were able to effectively
purge our subscales of the influence of the General Complaint factor. This we con-
sider to be substantively important in both theoretical and practical terms. Because
of the size of our sample, we computed hierarchical groupings on the cluster level,
that is, Cluster B should have no Cluster A conditions and Cluster C should have
neither Cluster A nor Cluster B conditions. On the single-diagnosis level, we com-
puted pure categories to avoid the confounding effects of comorbidity.
In this article, we presented analyses supporting the construct and concurrent
validity of a three-dimensional structure in a 48-item selected subset of the IIP; As-
sertiveness, Sociability, and Interpersonal Sensitivity emerged as separate dimen-
sions that can be tapped by subscales with high reliability (internal consistency)
and low internal correlations according to a parsimonious, balanced, and theoreti-
cally appealing model with simple scoring rules for computing subscale scores in
practical applications and good discriminative ability among patients having dif-
ferent PDs.
Further research on data from clinical use of the proposed 48-item version of
IIP is needed to test its stability and utility in everyday applications.

ACKNOWLEDGMENT

This study was supported by the Haldis and Josef Andresen’s Foundation, Norway.

REFERENCES

Alden, L. E., Wiggins, J. S., & Pincus, A. L. (1990). Construction of circumplex scales for the Inventory
of Interpersonal Problems. Journal of Personality Assessment, 55, 521–536.
 310 GUDE, MOUM, KALDESTAD, FRIIS

American Psychiatric Association. (1987). Diagnostic and statistical manual of mental disorders (3rd
ed., rev.). Washington DC: Author.
Barkham, M., & Hardy, G. E. (1996). The IIP–32: A short version of the Inventory of Interpersonal
Problems. British Journal of Clinical Psychology, 35, 21–35.
Barkham, M., Hardy, G. E., & Startup, M. (1994). The structure, validity and clinical relevance of the In-
ventory of Interpersonal Problems. British Journal of Medical Psychology, 67, 171–185.
Birtchnell, J. (1993). How humans relate—A new interpersonal theory. Westport, CT: Praeger.
Ekselius, L., Lindstrøm, E., von Knorring, L., Bodlund, O., & Kullgren, G. (1993). Personality disorders
in DSM–III–R as categorical or dimensional. Acta Psychiatrica Scandinavica, 88, 183–187.
Gude, T., Dammen, T., & Friis, S. (1997). Clinical vignettes in quality assurance—An instrument for
evaluating therapists’ competence in diagnosing personality disorders. Nordic Journal of Psychia-
try, 51, 207–212.
Hoffart, A., & Martinsen, E. W. (1992). Personality disorders in panic with agoraphobia and depression.
British Journal of Clinical Psychology, 31, 213–214.
Horowitz, L. M. (1979). On the cognitive structure of interpersonal problems treated in psychotherapy.
Journal of Consulting and Clinical Psychology, 47, 5–15.
Downloaded by [ ] at 06:11 17 January 2012

Horowitz, L. M., Rosenberg, S. E., Baer, B. A., Ureño, G., & Villaseñor, V. S. (1988). Inventory of Inter-
personal Problems: Psychometric properties and clinical applications. Journal of Consulting and
Clinical Psychology, 56, 885–892.
Horowitz, L. M., Rosenberg, S. E., & Bartholomew, K. (1993). Interpersonal problems, attachment
styles, and outcome in brief dynamic psychotherapy. Journal of Consulting and Clinical Psychol-
ogy, 61, 549–560.
Livesley, J. W. (1987). A systematic approach to the delineation of personality disorders. American
Journal of Psychiatry, 144, 772–777.
Pilkonis, P. A., Kim, Y., Proietti, J. M., & Barkham, M. (1996). Scales for personality disorders devel-
oped from the Inventory of Interpersonal Problems. Journal of Personality Disorders, 10, 355–369.
Spitzer, R. L., & Williams, J. B. W. (1988). Structured Clinical Interview for DSM–III–R (SCID 6/1/88).
New York: New York State Psychiatric Institute.
Widiger, T. A., Frances, A., Spitzer, R. L., & Williams, D. S. W. (1988). The DSM–III–R personality
disorders: An overview. American Journal of Psychiatry, 145, 786–795.

Tore Gude
Department of Behavioural Science in Medicine
University of Oslo
N-0317 Oslo
Norway
E-mail: tore.gude@basalmed.uio.no

Received July 9, 1998

Revised November 16, 1999

View publication stats