Circular
Reasoning About Interpersonal Behavior: Evidence Concerning Some
Untested Assumptions Underlying Diagnostic Classification 
University of British Columbia, Vancouver, British Columbia, Canada
University of British Columbia, Vancouver, British Columbia, Canada
University of British Columbia, Vancouver, British Columbia, Canada
This work was supported by Social Sciences and
Humanities Research Council of Canada Grant 410871322 awarded to
Jerry S. Wiggins. We would like to thank Candace Taylor Wiggins and two
anonymous reviewers for their helpful comments on an earlier version of
this article.
Correspondence may be addressed to: Jerry S.
Wiggins, Department of Psychology, University of British Columbia, 2136
West Mall, Vancouver, British Columbia V6T 1Y7 Canada.
The Interpersonal Circle is a conceptual
representation of the domain of interpersonal behavior that depicts
interpersonal variables as vectors in a twodimensional circular space
formed by the coordinates of dominance (Dom) and love (Lov) (e. g. , Benjamin, 1974; Carson, 1969; Kiesler, 1983; Leary, 1957; Lorr & McNair, 1965; Wiggins, 1979). The Interpersonal Adjective Scales (IAS; Wiggins, 1979)
were constructed for the purpose of providing empirical markers of this
conceptual representation, thus permitting comparisons between parallel
constructs drawn from the different research traditions of clinical,
personality, and social psychology (e. g. , Wiggins, 1980, 1982; Wiggins & Broughton, 1985).
The original version of the IAS consisted of eight 16item scales whose
interrelations exhibited clear circumplex properties of the kind
illustrated in
Circumplex structure of revised Interpersonal Adjective Scales (N = 1,161).
Figure 1 (e. g. , Wiggins, Steiger, & Gaelick, 1981).
The revised Interpersonal Adjective Scales (IASR) were constructed to
provide a more convenient shortform version of the IAS, one with
improved substantive and structural characteristics (Wiggins, Trapnell, & Phillips, 1988).
The IASR comprises 64 single adjectives (e. g. , dominant) to which subjects respond by indicating the degree of selfdescriptive accuracy on an eightpoint Likert scale ranging from extremely inaccurate to extremely accurate.
Item responses are cumulated to form eight scales, each alphabetically
labeled in a counterclockwise direction around the circle:
AssuredDominant (PA), ArrogantCalculating (BC), Coldhearted (DE),
AloofIntroverted (FG), UnassuredSubmissive (HI), UnassumingIngenuous
(JK), WarmAgreeable (LM), and GregariousExtraverted (NO). As Figure 1
illustrates, these octant scales may also be designated in terms of
their expected geometric location with reference to the horizontal Lov
axis: PA (90°), BC (135°) ... NO (45°).
The circular configuration of variables in Figure 1
is subject to a variety of interpretations. One such interpretation
would assert that the eight variables exhibit a circular ordering
without beginning or end. In correlational terms, this would mean that
the ordered patterning of correlations of, for example, (PA vs. BC)
> (PA vs. DE) > (PA vs. FG) > (PA vs. HI) exists in all parts
of the matrix of intercorrelations among variables. This interpretation
is made without reference to the specific magnitudes of the
correlations, their angular locations, or their distance from the
circle's center. It is the interpretation used by Guttman (1954) when he first proposed the concept of a circumplex.
A less formal interpretation of Figure 1
would assert that it provides a convenient pictorial representation of
concepts associated with an interpersonal theory of personality. With
reference to the concepts of dominance and love, one may portray
different interpersonal types and illustrate their "movement" both
around the circle (personality change) and from the center toward the
periphery of the circle (extremity of personality type). This
nonquantitative representation of interpersonal space is reminiscent of
Lewin's (1936) use of hodological space to represent
the constructs of his system. It should be evaluated for its heuristic
potential to generate testable hypotheses, and indeed it has proven
valuable in that respect.
Figure 1 may also be taken
to be a formal geometric model of the interrelations among indicants of
constructs derived from an interpersonal theory of personality. Under
this interpretation, the specific angular location of person variables
(indicants) and their distance from the center of the circle should
provide an empirical basis for testing hypotheses derived from
interpersonal theory. Interpersonal assessment procedures that use
geometric calculations to assign individuals to typological categories
or to evaluate the relative adaptiveness of individuals within
typological categories (e. g. , Leary, 1957) are
based on strong metric assumptions. Such procedures cannot be
rationalized on the grounds that interpersonal variables exhibit a
circular ordering or by reference to the fact that individual
investigators prefer to think of interpersonal behavior in circular
terms. These procedures can be justified only by demonstrating that the
measuring instruments involved meet the geometric assumptions of the
formal model. In this article, we evaluate the extent to which one such
instrument, the IASR, meets these criteria with respect to the topic
of interpersonal diagnosis.
One of the exciting potentialities of the
Interpersonal Circle has been the possibility of representing
interpersonal tendencies with reference to the geometric properties of
a circle (LaForge, Leary, Naboisek, Coffey, & Freedman, 1954).
Using LM and PA as reference directions, it is possible to represent
the 8 (or 16) scores of an individual's interpersonal profile by a
single point in twodimensional space: where Zi is the standardized octant score in the i th category (the categories LM, NO, PA, etc. being numbered 1, 2, 3, etc. ) and θi is the angle at the center of the category, namely, (i − 1) × 45° (LaForge et al. , 1954,
pp. 138140). The cosine and sine weights result in two orthogonal
linear combinations of the octants that have maximum variance (i. e. ,
the first 2 principal components) under the assumption of
circumplexity. The expressions are multiplied by . 3 to give them unit
variance.
The two coordinate values (Lov, Dom) permit
the precise location of an individual within twodimensional space. A
computational example is provided in Table
1. The first column contains the interpersonal profile of an individual
expressed as standard scores with reference to normative data that is
described in Study 1. From the profile alone one can surmise that the
individual is an assureddominant type. The second column provides the
angular direction of each octant and the third and fourth columns the
sine and cosine of these angles.
The Lov coordinate value is
obtained by summing the products of Columns 1 and 3: (1. 45)(. 00) + (.
70)(−. 707) + ... etc. and multiplying the result by . 3. The Dom
coordinate value is obtained in similar fashion from Columns 1 and 4.
The Lov and Dom coordinates locate a single point in the interpersonal
plane. A vector from the origin to the point is characterized by its
length and its angular direction as measured by its counterclockwise
distance from the positive horizontal axis. That is, the Cartesian
coordinates Lov and Dom are expressed in polar coordinates angle and
vector length.
Angle and vector length serve
different purposes in diagnostic classification. The angular location
is used to assign category membership. In the present example, the
individual's angular location is 87. 9°, which is clearly within the
assureddominant category (i. e. , between 67. 5° and 112. 5°). The
vector length corresponding to a given individual's octant scores is
basically an index of the standard deviation of those scores.
Conceptually, individuals whose geometric
location falls far from the center of the Interpersonal Circle have
been considered deviant in a psychiatric as well as a statistical sense
(e. g. , Leary, 1957). With some possible exceptions (Chartier, 1984; Wiggins & Holzmuller, 1978, 1981),
deviance has not been clearly equated with profile variance, although
as is now shown, such an equating is completely consistent with the
psychometric implications of interpersonal theory.
For most interpersonal theorists (e. g. , Carson, 1969; Kiesler, 1983; Leary, 1957; Millon, 1981),
it is axiomatic that interpersonal behaviors fall on continua of
intensity ranging from the moderate and generally adaptive to the
extreme and often maladaptive. It has also been assumed that intensity
of expression is related to interpersonal flexibility, in the sense
that dysfunctional individuals also rely rigidly on a relatively narrow
band of extreme actions and reactions to the exclusion of other,
possibly adaptive, modes of response. When conjoined, the concepts of
intensity and flexibility imply a patterning of behavior that, in the
present instance translates into a profile of interpersonal
dispositions.
Adaptive interpersonal functioning may
therefore be construed as the moderate, flexible, and adaptive
expression of a characteristic pattern of interpersonal behaviors. The
particular pattern expressed will vary with the type of individual, as
indicated by his or her average directional tendencies with respect to
the Dom and Lov coordinates. Thus, an assureddominant type of person
(90°) is one who will often behave in a confident or assertive way,
will somewhat less frequently behave in an arrogant or calculating way,
will seldom behave in an unassured or submissive way, and so forth.
This is to emphasize that there is a characteristic pattern associated
with being an assureddominant type and that this pattern is shared by
prototypical members of that type.
Maladaptive interpersonal functioning with the
same directional tendencies may be characterized as the exaggerated,
inflexible, and dysfunctional expression of the same pattern of
interpersonal behaviors. The prototypical patterns of both
assureddominant (adaptive) and overassuredautocratic (maladaptive)
types are the same. They differ only in profile variance. This
comparability of profile configuration for adaptive and maladaptive
subjects within the same diagnostic category should reflect a more
general consistency of profile configuration for all subjects,
regardless of diagnostic group. That is, the semantic constraints of
the interpersonal circumplex (adjacent variables moderately correlated,
opposite variables negatively correlated, etc. ) should operate
uniformly for all diagnostic groups. Thus, the profile configuration of
assureddominant types and that of coldhearted types should
theoretically be identical. Profiles of the two groups would be
expected to differ only in angular orientation.
Because vector length is, psychometrically, an
index of the standard deviation of a profile, and because profile
configurations are expected to differ only in angular orientation,
there is no a priori reason to assume that vector length would, in and
of itself, be systematically related to elevations on any of the eight
variables of the interpersonal circumplex. Thus, for example, in a
random sample of subjects, a high variance profile could be equally
indicative of the elevated FG scale of an aloofintroverted type and
the elevated NO scale of a gregariousextraverted type.
By the reasoning just stated for interpersonal
types, it is unlikely that there will be systematic relations between
vector length and scores on outside measures of deviance or general
psychopathology. Although nonsubstantive considerations, such as those
that led Berg (1955) to formulate the deviation
hypothesis, might lead to the expectation that all indexes of deviation
will be positively correlated, there is very little evidence to support
such an expectation (Wiggins, 1973). On the other
hand, the interpersonal theory presented earlier would predict
relations between vector length and general psychopathology within
certain diagnostic categories of the interpersonal circumplex. For
example, one might anticipate a relation between vector length and
general psychological discomfort (neuroticism) within a group of
subjects who have all been classified as aloofintroverted types.
Within the context of the interpersonal
circumplex theory outlined earlier, the mean vector direction of a
profile of interpersonal variables is assumed to provide an index of a
subject's characteristic pattern of interpersonal behaviors (e. g. ,
assureddominant), and the vector length of a subject's profile is
assumed to provide an index of the intensity, flexibility, and
adaptiveness of that characteristic pattern (e. g. ,
overassuredautocratic). For the reasons stated in the preceding two
paragraphs, one would not expect vector length to be related to
measures of interpersonal problems in a random sample of unclassified
subjects. However, to the extent that measures of interpersonal
problems that span the full universe of problems defined by the
coordinates of Dom and Lov are available, then correlations with vector
length would be predicted within all eight diagnostic categories of the
interpersonal circumplex.
The preceding introduction was meant to
provide an overview of the theoretical basis of interpersonal diagnosis
with special attention to the manner in which the constructs of the
theory might be translated into geometrically based measurement
procedures. Although it is widely believed that geometrically based
measurement procedures can be interpreted as indicants of the
constructs of interpersonal diagnosis, there is surprisingly little
hard empirical evidence that would justify such a belief. For that
reason, our introduction may be best thought of as an explication of
the assumptions that must be met before the diagnostic use of
instruments such as the IASR can be justified. These assumptions are
listed in Table
2, and as can be seen from that table, they are stated in the form of
nine hypotheses that are tested in the three empirical studies to be
reported.
The assumptions in Table 2
appear in the same order as the arguments we developed in our
introduction. They progress from fundamental issues of measurement to
more substantive issues of interpretation. Although the failure to meet
an early assumption would not preclude the possibility of empirical
support for a later assumption, it would to some extent vitiate the
theoretical interpretation that might otherwise be placed on the
empirical finding. For example, if the IASR did not have the rigorous
geometric properties required by the measurement model (Assumption 1),
there could still be a significant empirical relation between vector
length and interpersonal problems within IASR diagnostic groups
(Assumption 9). However, the empirical relation found in connection
with Assumption 9 could not be interpreted within the context of the
geometric model postulated by the theory.
In Study 1, we tested five basic hypotheses
that relate to the geometric properties of the IASR in relation to
circumplexity, vector length, and characteristic profile
configurations. In Study 2, we examined IASR vector length within
diagnostic groups in relation to measures of general psychopathology
provided by Lanyon's (1973) Psychological Screening
Inventory. In Study 3, we looked at IASR vector length within
diagnostic groups in relation to the Inventory of Interpersonal
Problems (Horowitz, Rosenberg, Baer, Ureno, & Villasenor, 1988).
Taken together, the results of these three studies provide the basis
for a preliminary evaluation of IASR as an indicant of the constructs
associated with an interpersonal theory of diagnosis.
The subjects were 1,161 University of British
Columbia students who had participated in one of nine separate studies
of the IAS conducted at UBC over the past 6 years. As can be seen from Table
3, we recruited subjects from a variety of undergraduate psychology
courses. Although the instruments administered in addition to the IAS
varied from study to study, the IAS was always the first test given.
The tests were administered in class, in our laboratory, or on a
takehome basis.
The only instrument considered in this study
was the IASR. This is a short form version of the IAS that has
improved substantive and structural characteristics and acceptable
scale reliabilities (Wiggins et al. , 1988).
We scored the IASR protocols of each subject
for the eight octants (e. g. , assureddominant [PA]) of the
interpersonal circumplex. We extracted two principal components from
the intercorrelations among these octants in order to evaluate the
circumplexity of the IASR and to compare theoretical geometric weights
for assignment to diagnostic groups with empirical factor score
coefficients used for the same purpose.
We assigned each subject to one of eight
diagnostic (type) categories on the basis of the procedure for
determining mean vector direction presented in Table 1.
The subjects were fairly evenly distributed across the eight diagnostic
categories, with a mean of 145 subjects per category and a range from
134 to 158. Vector length (distance from the center of the circle) was
also determined for each subject from the last formula presented in Table 1.
The IASR scales conform well to the theoretically expected circumplex structure. Figure 1
presents a plot of the eight IASR scales on the first 2 principal
components extracted from the intercorrelation matrix of scales in our
sample of 1,161 subjects. These two components together account for 71.
11% of the total variance among IASR scales. Equally noteworthy, the
components account for approximately the same amount of variance (36.
29% and 34. 82%, respectively); a difference of only 1. 47 percentage
points. The latter is important for the circularity of the solution, as
discrepancies in magnitude between the two latent roots will yield an
elliptical structure. The eigenvalue corresponding to the 3rd principal
component was . 78, and this component accounted for less than 10% of
the total variance.
The formulas for calculating Lov and Dom are
theoretical in the sense that the weights by which each of the
standardized octant scores is multiplied are based on the sines and
cosines of ideal angular locations. The empirical counterparts of these
theoretical vector weights would be the empirical factor score
coefficients associated with the two principal components extracted
from the intercorrelations among octants. The extent to which the two
sets of formulas are in agreement provides an index of the goodness of
fit of empirical IASR subject assignment to that specified by the
geometric model.
For each subject we computed Dom and Lov coordinate estimates by the geometric formulas given in Table 1
and by factor score estimates derived from the principalcomponents
analysis of the IASR scales. In the total subject sample, the
correlation between theoretical and empirical estimates was . 999 for
the Dom coordinate and . 999 for the Lov coordinate. When subjects were
assigned to one of eight diagnostic groups on the basis of the two sets
of formulas, there was agreement in assignment for 97. 2% of the
subjects.
We computed the standard deviation of each
eightoctant profile of interpersonal variables for each subject. The
empirical correlation between profile standard deviation and vector
length in this sample was . 946.
We ranked subjects within each of the eight
diagnostic categories in terms of their vector lengths and arbitrarily
defined the upper 10% as extreme and the remainder as moderate. On
average, there were 14 extreme and 131 moderate subjects in each of the
eight diagnostic categories. Within each category we correlated the
mean profiles of extreme and moderate groups. These correlations
approached unity in all categories; the average correlation was . 989
and the range was from . 981 to . 997. This finding is compatible with
our assumption that moderate variance and extreme variance subjects
within the same diagnostic category have the same characteristic
profile pattern. It thus provides justification for our interpretation
of vector length as a homogeneous index of extremity or deviance.
Mean profile of interpersonal variables for subjects classified as coldhearted (DE; n
= 135). (FG = AloofIntroverted, HI = UnassuredSubmissive, JK =
UnassumingIngenuous, LM = WarmAgreeable, NO = GregariousExtraverted,
PA = AssuredDominant, and BC = ArrogantCalculating).
Figure 2 presents the mean
profile of interpersonal variables for the 135 subjects classified as
falling within the DE diagnostic category. This profile exhibits a
characteristic shape that we refer to informally as the "interpersonal
spaceship": the highest elevation occurs on the defining octant (DE),
followed by adjacent octants (BC and FG), and diminishing to the highly
truncated opposite octant (LM). Given the semantic constraints of the
interpersonal circumplex, this configuration should occur in other
diagnostic groups that differ in angular orientation.
To test the uniformity of
semantic constraints around the circumplex, we compared the mean
profiles of the eight diagnostic groups in the total sample. Lag
correlations were computed between all combinations of diagnostic group
profiles. In comparing profiles of the PA and BC diagnostic groups, for
example, the mean scores for the former group, PA, BC, DE, ... etc. ,
were correlated with the mean scores for the latter group, BC, DE, FG,
... etc. The mean of the 28 lag correlations thus computed was . 978;
the standard deviation was . 01.
Vector length was correlated with scores on each of the eight IASR octants in the total sample. As can be seen from Table
4, the correlations ranged from −. 02 (HI) to . 13 (DE), with a mean of
. 04. Although there was a slight tendency for vector length to be
correlated with scores on undesirable IASR octants, it was neither
strong nor consistent. The highest significant correlation obtained
suggests a shared variance of 2%.
The strong geometric
assumptions underlying the interpretation of the structural relations
among IASR scales as a circumplex and the assignment of subjects to
diagnostic groups based on that circumplex are more than adequately
met. The finding that the theoretical sine and cosine weights for
determining the angular locations of interpersonal profiles are
equivalent to the empirical IASR factor coefficients is especially
noteworthy. This result suggests that in large samples of subjects,
IASR factor estimates of Dom and Lov will converge on the theoretical
geometric formulas. Because no instrument can be considered errorfree,
we recommend the theoretical formulas presented in Table 1 over empirical IASR factor estimates in assigning subjects to diagnostic groups, especially in samples of moderate size.
Profile configurations of IASR variables were
remarkably comparable both for moderate and extreme subjects within the
same diagnostic groups and for subjects in different diagnostic groups.
It is still possible, of course, that subjects with the same profile
angular location and vector length score could have different profile
configurations, or that subjects assigned to different diagnostic
groups could have profile configurations that differed in ways other
than angular location. However, these possibilities are empirically so
unlikely that any profile that deviates markedly from semantic
expectations should be viewed with suspicion.
Vector length is an index of profile
variability (standard deviation), and this, together with the
abovementioned findings on the constancy of profile shapes within
diagnostic categories, supports the idea of deviance as an exaggeration
of a characteristic profile configuration, or type. Although three of
the correlations of vector length with IASR octant scores were
statistically reliable in the present large sample, the overall pattern
of correlations was not supportive of a moderate or consistent
relation. The critical next stage of investigation explores the
relation between vector length and general measures of psychopathology
within IASR diagnostic categories.
The subjects were 139 University of British
Columbia undergraduates (58 men and 81 women) who were enrolled in the
first term of introductory psychology. The instruments were
administered in class as part of an introduction to personality tests.
In addition to the IASR, the subjects were given Lanyon's (1973)
Psychological Screening Inventory (PSI). This 130item, truefalse
personality inventory was designed for use as a brief mental health
screening device. In addition to its brevity, one of this instrument's
principal advantages is the fact that the scales are substantially
correlated with MMPI clinical and validity scales, even though the
scales include a minimum of objectionable items.
Four nonoverlapping pathology scales and a
validity scale may be scored from the PSI. The Alienation scale was
designed to indicate the similarity of a respondent to hospitalized
psychiatric patients. The MMPI correlates include Psychasthenia (Pt), Paranoia (Pa), and Schizophrenia (Sc).
The Social Nonconformity scale was designed to indicate the similarity
of a respondent to incarcerated prisoners. The highest MMPI correlate
is Psychopathic Deviate (Pd). The Discomfort scale was
designed to assess the dimension of perceived maladjustment, anxiety,
or neuroticism. The highest correlate is Welsh's A scale, a
marker of the first factor of the MMPI. The Expression scale was
designed to measure the dimension of undercontrol, impulsivity, or
extraversion. The highest MMPI correlate is Hypomania (Ma). The Defensiveness scale was designed to assess defensiveness in testtaking attitude. The highest MMPI correlates are Lie (L) and Testtaking Attitude (K).
We scored the IASR protocols of each subject
for the eight octants of the interpersonal circumplex. We assigned
subjects to one of eight diagnostic categories on the basis of the
procedure for determining mean vector direction discussed in Study 1.
The mean number of subjects per diagnostic category was 17 and the
range was from 12 to 30. We also determined vector length for each
subject by the method discussed in Study 1 and scored the four
pathology scales and the validity scale of the Psychological Screening
Inventory for each subject.
Vector length was correlated with scores on the five PSI scales in the total sample of 139 subjects. As can be seen from Table
5, these correlations ranged from −. 09 to . 13, with a mean of −. 01.
Perhaps the most important of these nonsignificant correlations is that
between vector length and discomfort (−. 01), as the latter has been
demonstrated to be substantially correlated with the first factor of
the MMPI, a general measure of psychopathology or deviance.
Within each of the eight diagnostic groups
derived from IASR profiles, vector length was correlated with scores
on the five PSI scales. We obtained significant correlations within
five of the eight IASR diagnostic groups and with three of the five
PSI scales. These correlations are displayed in Table 6.
Within the arrogantcalculating
category, the correlation between vector length and the Social
Nonconformity scale is understandable because of the relation between
BC and Machiavellianism (Wiggins & Broughton, 1985) and the relation between BC and the deployment of manipulation tactics (Buss, Gomes, Higgins, & Lauterbach, 1987).
Within the aloofintroverted category, the correlation between vector
length and discomfort suggests a notunexpected relation between FG and
the dimension of anxiety or perceived maladjustment. The negative
correlation between vector length and discomfort within the
assureddominant category is compatible with the selfassured
egoresiliency (Block, 1965) of the PA variable.
Finally, the pattern of correlations between vector length and the PSI
Expression scale, within the arrogantcalculating,
unassuredsubmissive, and gregariousextraverted categories, suggests
that the Expression scale includes a fairly broad band of content
reflecting undercontrol, impulsivity, and extraversion.
These interpretations must be tempered by the
realization that they were obtained in small samples of nonclinical
subjects. These data, although limited, nonetheless serve to underscore
a point regarding the relation between interpersonal dispositions and
psychopathology that has hitherto been neglected in the empirical
literature. The general psychopathology component of the MMPI (PSI
Discomfort scale) is not general in the sense that it is associated
with deviance in all or even most interpersonal dispositions: The
correlation between general deviance (vector length) and discomfort was
essentially zero in the present study. However, when vector length was
considered within those dispositional types captured by the relatively
restricted MMPI item pool (Wiggins, 1987), we
obtained significant correlations. These correlations, although forming
a substantively coherent pattern, did not reflect a circumplex ordering
(moderate correlations appearing within adjacent diagnostic groups,
negative correlations appearing within opposite diagnostic groups). The
first principal component of the MMPI reflects a simplex of
psychopathology, ranging from relatively low (Ma) to relatively high (Sc) impairment (Wiggins et al. , 1981).
The dispositions measured by the IASR reflect a circumplex of
personality characteristics that are presumably associated with a
broader range of interpersonal problems. We investigate this last
proposition in Study 3.
^{1}. The subjects
were 264 University of British Columbia undergraduates (122 men and 142
women) who were enrolled in the first term of lower division psychology
courses. Testing materials were distributed in classes for completion
on a takehome basis.
In addition to the IASR, the subjects were given the Inventory of Interpersonal Problems (IIP) developed by Horowitz (1979) and his associates (Horowitz et al. , 1988).
The initial item pool of the IIP consisted of verbatim statements of
problems made in the course of a videotaped intake interview by
patients about to undergo psychotherapy. Interpersonal problem
statements were identified by judges and classified under the two
general categories of "I can't" (e. g. , "trust other people") and "I
have to" (e. g. , "avoid other people"). When meaningsimilarity
sortings of these statements were subjected to a multidimensional
scaling analysis, the dimensions of control (dominantsubmissive) and
affiliation (friendlyhostile) were clearly identified.
The most recent version of the IIP is composed
of 127 items, 78 of which have the stem "It is hard for me to" (e. g. ,
"be assertive with another person") and 49 of which are presented as
"things that you do too much" (e. g. , "I fight with other people too
much"). The items are answered on a fivepoint Likert scale that ranges
from not at all to extremely. Responses to these
items are cumulated on a set of 12 nonoverlapping interpersonal
problems scales. In addition, a set of 8 nonoverlapping interpersonal
problems circumplex scales have recently become available (Alden, Wiggins, & Phillips, 1987).
These scales exhibit a clear circumplex structure in both normal and
psychiatric populations and have acceptable testretest reliabilities.
Although the scales have been interpreted as reflecting dysfunctional
aspects of interpersonal dispositions measured by the IASR circumplex,
they were developed independently of that instrument. In
counterclockwise order around the circumplex, these scales are
Autocratic, Competitive, Cold, Introverted, Subassertive, Exploitable,
Overly Nurturant, and Overly Expressive.
We scored the IASR protocols of each subject
for the eight octants of the interpersonal circumplex. Subjects were
assigned to one of eight diagnostic categories on the basis of the
procedure for determining mean vector direction presented in Study 1.
The mean number of subjects per diagnostic category was 33 and the
range was from 19 to 45. We also determined vector length for each
subject by the method described in Study 1. The eight interpersonal
problems circumplex scales of the IIP were also scored for each
subject.
Vector length was correlated with scores on the eight IIP circumplex scales in the total sample. As can be seen from Table 7, these correlations ranged from −. 16 to . 18, with a mean of . 07. Of the three reliable correlations (p
< . 05), two were positive and one was negative. As was true of the
correlations between vector length and IASR octant scores (Table 4),
there is a slight tendency for vector length to be correlated with
scores on undesirable IIP circumplex scales, but again, this tendency
is neither strong nor consistent. The highest significant correlation
obtained suggests a shared variance of 3%.
Within each of the eight diagnostic groups
derived from IASR profiles, vector length was correlated with scores
on the eight IIP circumplex scales. We obtained significant
correlations within all of the IASR diagnostic groups and with all of
the IIP circumplex scales. These correlations are displayed in Table 8.
We consider the two hypotheses
of this study to be confirmed because whereas vector length is only
slightly and inconsistently correlated with IIP circumplex scales in
the total sample of subjects, it is significantly although not
substantially correlated with IIP circumplex scales within all of the
IASR diagnostic categories. Interpretation of this pattern of obtained
correlations is complex because, unlike the expected relations between
vector length and the general measures of psychopathology provided by
the PSI scales in Study 2, the expected relations between vector length
and IIP circumplex scales in our study involve circumplex
considerations in which vector length is anticipated to be positively
related to IIP circumplex scales within comparable dispositional
categories, negatively related to opposite scales, and unrelated to
orthogonal scales.
The expected circumplex pattern of correlations may be illustrated with reference to the first column of Table 8.
Within the assureddominant diagnostic category (PA) of IASR, vector
length would be expected to be positively related to its comparable
Autocratic scale (. 36), negatively related to its opposite
Subassertive scale (−. 27), positively related to the adjacent scales
of Competitive (. 27) and Expressive (a miss), negatively related to
its opposite offquadrant scales of Introverted (−. 27) and Exploitable
(−. 35), and unrelated to the orthogonal scales of Cold (ns) and Nurturant (ns).
Taking into account the signs of significant
correlations and the occurrence of anticipated nonsignificant
correlations, 38 of the 64 possible correlations, or lack of same, in Table 8
are in accord with circumplex expectations. Of the expected significant
correlations that did or did not occur, only three were not in accord
with predictions. Although there are many instances in which the
relative magnitude of the significant correlations departed from
expectations, the overall pattern of correlations in Table 8
strongly suggests the operation of two circular systems. Consideration
of the substantive convergences, or lack of same, between the IIP and
the IAS is well beyond the scope of this article. This rather complex
topic is treated elsewhere (Alden et al. , 1987).
The Interpersonal Circle is a conceptual
representation of the domain of interpersonal behavior that depicts
interpersonal variables as vectors in a twodimensional circular space
formed by the coordinates of dominance and love. Within that circular
space, the vectors that emanate from the center of the circle are
interpreted as continua of intensity ranging from the moderate and
generally adaptive to the extreme and often maladaptive. Interpersonal
diagnosis involves the assignment of subjects to typological categories
defined by the average directional tendencies of their interpersonal
behaviors with reference to the coordinates of dominance and love.
Subjects classified as falling within the same typological sector of
the Interpersonal Circle exhibit the same characteristic pattern of
interpersonal behaviors, but differ from one another in terms of the
intensity or adaptiveness of that pattern, as indicated by their
distance from the center of the circle.
The properties of the Interpersonal Circle
just described may be viewed as simply a convenient pictorial
representation of concepts associated with an interpersonal theory of
personality. However, to the extent that these properties are
transformed into concrete assessment procedures that assign subjects to
diagnostic groups on the basis of geometric calculations, the strong
metric assumptions underlying such assignments must be justified. We
identified nine such assumptions that ranged in scope from fundamental
issues of measurement to more substantive issues of interpretation.
These assumptions were stated in the form of hypotheses concerning
properties of the IASR that were tested in three empirical studies.
The eight octant scales of the IASR exhibited
a clear circumplex structure in accord with theoretical expectations.
Factor score coefficients for the two principal components underlying
this circumplex structure were virtually identical to the theoretical
sine and cosine weights for ideal angular locations. This latter
finding provides strong support for IASR diagnostic assignments based
on geometric principles. The theorybased hypothesis that subjects with
extreme and moderate vector lengths within the same IASR diagnostic
groups would exhibit the same characteristic profile pattern was also
strongly supported. More generally, we found the characteristic shape
of interpersonal profiles (the interpersonal spaceship) in all
diagnostic groups, regardless of angular orientation. Overall, it may
be concluded that the geometric properties of the IASR meet the basic
assumptions of measurement associated with the assignment of subjects
to diagnostic groups.
Properties of the heretofore neglected
variable of vector length were generally in accord with theoretical
expectations. On both conceptual and psychometric grounds, vector
length should be an index of characteristic profile variability
(standard deviation), and indeed we found this to be the case. The
hypothesis that vector length is, in itself, unrelated to IASR octant
variables was clearly supported in a large and presumably
representative sample of university students.
The expected relations between vector length
and measures of psychopathology within IASR diagnostic groups depends
very much on the psychopathology measures used. As Horowitz (1979)
has emphasized, the relation between symptoms (e. g. , depression) and
interpersonal problems (e. g. , intimacy) has not been clearly
explicated, and one would certainly not expect a onetoone
correspondence. Psychiatric screening devices, such as Lanyon's PSI,
assess both symptoms and a limited variety of interpersonal problems.
Within our sample of university students, PSI scales were unrelated to
vector length, as expected. Moreover, within some IASR diagnostic
groups we found relations between vector length and PSI scales that
formed a substantively coherent pattern. Although these findings were
in accord with expectations, they serve primarily to justify further
explorations of relations between IASR diagnosis and traditional
general measures of psychopathology such as the MMPI.
The interpersonal problems circumplex scales from the Horowitz et al. (1988)
IIP were generally unrelated to vector length, as expected. Also as
expected, vector length was significantly although not substantially
correlated with interpersonal problems scales within all IASR
diagnostic groups. We interpret these findings as supporting our
hypotheses concerning the relation between IASR profile vector length
and interpersonal problems and as justifying further research on the
theoretically important topic of the relation between interpersonal
dispositions and interpersonal problems.
The sound geometric and psychometric
properties of the IASR suggest that it may be applied fruitfully to
traditional clinical research topics such as the psychodiagnosis of
personality disorders. More generally, these properties of the IASR
should increase measurement precision in experimental studies involving
the preselection of subjects on individual difference measures of
interpersonal behavior. In such experimental studies, subjects are
typically assigned to high or low groups on the basis of scores on a
single personality inventory scale without reference to their possible
scores on other (unmeasured) interpersonal variables. Our findings with
the IASR in these studies suggest that assignment of subjects to
diagnostic categories results in homogeneous groupings of subjects with
respect to characteristic patterns of interpersonal behavior, and that
vector length may be used as an index of the extremity of these
characteristic patterns.
Assignment of subjects to groups on the basis
of multiscale configural patterns results in increased substantive
homogeneity within groups, even with scales of less than desirable
psychometric properties, such as MMPI scales (Payne & Wiggins, 1972).
The precise and theoretically based assignment of subjects to
preselection groups on the basis of IASR may serve to advance
currently active areas of experimental inquiry such as interpersonal
perception and interpersonal construal style, complementarity and
similarity in dyadic transactions, and competition and cooperation in
experimental games.
1
We are grateful to Lynn Alden for providing these data, which were collected for another purpose.
1. Alden, L., Wiggins, J. S. & Phillips, N. (1987). Interpersonal dispositions and interpersonal problems. Unpublished manuscript, Department of Psychology, University of British Columbia, Vancouver, BC, Canada.
2. Benjamin, L. S. (1974). Structural analysis of social behavior. Psychological Review, 81, 392425.
3. Berg, I. A. (1955). Response bias and personality: The deviation hypothesis. Journal of Psychology, 40, 6171.
4. Block, J. (1965). The challenge of response sets: Unconfounding meaning, acquiescence and social desirability in the MMPI. New York: AppletonCenturyCrofts.
5. Buss, D. M., Gomes, M., Higgins, D. S. & Lauterbach, K. (1987). Tactics of manipulation. Journal of Personality and Social Psychology, 52, 12191229.
6. Carson, R. C. (1969). Interaction concepts of personality. Chicago: Aldine.
7. Chartier, B. M. (1984). The meaning and measurement of interpersonal flexibility. Unpublished doctoral dissertation, University of Saskatchewan, Saskatoon, Sask. , Canada.
8. Guttman, L. (1954). A new approach to factor analysis: The radex.In P. R. Lazarsfeld (Ed. ),Mathematical thinking in the social sciences(pp. 258348). Glencoe, IL: Free Press.
9. Horowitz, L. M. (1979). On the cognitive structure of interpersonal problems treated in psychotherapy. Journal of Consulting and Clinical Psychology, 47, 515.
10. Horowitz, L. M., Rosenberg,
S. E., Baer, B. A., Ureno, G. & Villasenor, V. S. (1988). The
Inventory of Interpersonal Problems: Psychometric properties and
clinical applications. Journal of Consulting and Clinical Psychology, 56, 885895.
11. Kiesler, D. J. (1983). The 1982 Interpersonal Circle: A taxonomy for complementarity in human transactions. Psychological Review, 90, 185214.
12. LaForge, R., Leary, T. F.,
Naboisek, H., Coffey, H. S. & Freedman, M. B. (1954). The
interpersonal dimension of personality: II. An objective study of
repression. Journal of Personality, 23, 129153.
13. Lanyon, R. I. (1973). Psychological Screening Inventory manual. Goshen, NY: Research Psychologists Press.
14. Leary, T. (1957). Interpersonal diagnosis of personality. New York: Ronald Press.
15. Lewin, K. (1936). Principles of topological psychology. New York: McGrawHill.
16. Lorr, M. & McNair, D. M. (1965). Expansion of the interpersonal behavior circle. Journal of Personality and Social Psychology, 2, 823830.
17. Millon, T. (1981). Disorders of personality: DSM III, Axis II. New York: Wiley.
18. Payne, F. D. & Wiggins, J. S. (1972). MMPI profile types and the selfreport of psychiatric patients. Journal of Abnormal Psychology, 79, 18.
19. Wiggins, J. S. (1973). Personality and prediction: Principles of personality assessment. Reading, MA: AddisonWesley.
20. Wiggins, J. S. (1979). A psychological taxonomy of traitdescriptive terms: The interpersonal domain. Journal of Personality and Social Psychology, 37, 395412.
21. Wiggins, J. S. (1980). Circumplex models of interpersonal behavior.In L. Wheeler (Ed. ),Review of personality and social psychology(Vol. 1, pp. 265294). Beverly Hills, CA: Sage.
22. Wiggins, J. S. (1982).
Circumplex models of interpersonal behavior in clinical psychology.In
P. C. Kendall & J. N. Butcher (Eds. ),Handbook of research methods in clinical psychology(pp. 183221). New York: Wiley.
23. Wiggins, J. S. (1987,September). How interpersonal are the MMPI personality disorder scales?In R. L. Greene (Chair),Current research on MMPI personality disorder scales. Symposium conducted at the annual meeting of the American Psychological Association, New York.
24. Wiggins, J. S. &
Broughton, R. (1985). The Interpersonal Circle: A structural model for
the integration of personality research.In R. Hogan & W. H. Jones
(Eds. ),Perspectives in personality: Vol. 1(pp. 147). Greenwich, CN: JAI Press.
25. Wiggins, J. S. & Holzmuller, A. (1978). Psychological androgyny and interpersonal behavior. Journal of Consulting and Clinical Psychology, 46, 4052.
26. Wiggins, J. S. & Holzmuller, A. (1981). Further evidence on androgyny and interpersonal flexibility. Journal of Research in Personality, 15, 6780.
27. Wiggins, J. S., Steiger, J. H. & Gaelick, L. (1981). Evaluating circumplexity in personality data. Multivariate Behavioral Research, 16, 263289.
28. Wiggins, J. S., Trapnell, P.
& Phillips, N. (1988). Psychometric and geometric characteristics
of the revised Interpersonal Adjective Scales (IASR). Multivariate Behavioral Research, 23
Received: April 24, 1987. Revised: December 11, 1987. Accepted: July 11, 1988.
