You are on page 1of 11

Psychological Assessment © 2009 American Psychological Association

2009, Vol. 21, No. 4, 543–553 1040-3590/09/$12.00 DOI: 10.1037/a0016899

Assessment of Dependency, Agreeableness, and Their Relationship

Jennifer Ruth Lowe, Maryanne Edmundson, and Thomas A. Widiger


University of Kentucky

Agreeableness is central to the 5-factor model conceptualization of dependency. However, 4 meta-


analyses of the relationship of agreeableness with dependency have failed to identify a consistent
relationship. It was the hypothesis of the current study that these findings might be due in part to an
emphasis on the assessment of adaptive, rather than maladaptive, variants of agreeableness. This
hypothesis was tested by using experimentally altered NEO Personality Inventory—Revised (Costa &
McCrae, 1992) items that were reversed with respect to their implications for maladaptiveness. The
predicted correlations were confirmed with the experimentally altered version with measures of depen-
dent personality disorder, measures of trait dependency (including 2 measures of adaptive dependency),
and measures of dependency from alternative dimensional models of personality disorder. The theoretical
implications of the findings and suggestions for future research are discussed.

Keywords: dependency, agreeableness, five-factor model, personality disorder

Quite a few studies have documented that the personality dis- compliance (difficulty expressing disagreement), altruism (volun-
orders of the Diagnostic and Statistical Manual of Mental Disor- teering to do unpleasant things), and modesty (needing advice and
ders (4th ed., text rev.; DSM–IV–TR; American Psychiatric Asso- reassurance from others to make everyday decisions)” (p. 96).
ciation, 2000) can be understood as maladaptive variants of the Pincus (2002) similarly conceptualized dependency from the
domains and facets of the five-factor model (FFM) of general perspective of the interpersonal circumplex (IPC; Pincus, 2005).
personality structure (Clark, 2007; Widiger & Costa, 2002). Nev- The IPC has been defined in a number of ways, but the more
ertheless, problematic findings have been obtained for some of the common conceptualization consists of two fundamental dimen-
DSM–IV–TR personality disorders, notably the dependent person- sions of agency (e.g., dominance vs. submission) and communion
ality disorder (Bornstein & Cecero, 2000; Miller & Lynam, 2008). (e.g., love vs. hate). Pincus suggested that various forms of de-
Widiger, Trull, Clarkin, Sanderson, and Costa (2002) hypothesized pendency involve combinations of low agency and high commun-
that dependent personality disorder represents, in large part, an ion, consistent with the fact that such trait terms as docile, servile,
extreme or maladaptive variant of the FFM domains of neuroti- self-sacrificing, modest, compliant, clinging, obedient, gullible,
cism and agreeableness. Empirical support for the association with submissive, self-effacing, and dependent have long been consid-
neuroticism has been obtained, but support for agreeableness has ered interpersonal in nature (Wiggins, 1982). This conceptualiza-
been weak or inconsistent. It was the purpose of the current study tion is congruent with FFM agreeableness, as agreeableness and
to explore the possible effect of the assessment instruments, with extraversion are rotational variants of IPC dominance and love
respect to the assessment of both FFM agreeableness and depen- (Pincus, 2002; Wiggins & Pincus, 1989, 2002).
dency. Studies have reported positive associations of agreeableness
Agreeableness, as assessed by the NEO Personality Inventory— with dependency using a variety of methodologies, such as clini-
Revised (NEO PI–R; Costa & McCrae, 1992), is the tendency to cians’ ratings of prototypic cases (Samuel & Widiger, 2004),
be trusting, straightforward, altruistic, compliant, modest, and researchers’ descriptions of prototypic cases (Lynam & Widiger,
tender-minded. Well before ever considering the DSM–IV–TR 2001), clinicians’ descriptions of actual cases (Blais, 1997), clini-
personality disorders as maladaptive variants of the FFM, Costa cians’ ratings of case vignettes (Sprock, 2002), informants’ ratings
and McCrae (1985) suggested that “agreeableness can also assume of persons with personality disorders (Mullins-Sweatt & Widiger,
a pathological form, in which it is usually seen as dependency” (p. 2007), behavioral indices of dependency (Rubinstein, 2007), and a
12). Widiger et al. (2002) subsequently suggested more specifi- few self-report inventory studies (e.g., Bagby et al., 2001; Costa &
cally that “the DSM–IV diagnostic criteria set includes many McCrae, 1990; Mongrain, 1993; Wiggins & Pincus, 1989; Zuroff,
explicit examples of pathological agreeableness, such as excessive 1994).
Nevertheless, most self-report inventory studies have failed to
confirm the expected relationship between agreeableness and de-
pendency. In fact, four largely independent meta-analyses of the
Jennifer Ruth Lowe, Maryanne Edmundson, and Thomas A. Widiger,
research literature have all reported no consistent relationship
Department of Psychology, University of Kentucky.
The authors express their appreciation to Brittany Stark and Leigh
between agreeableness and dependency, leading some to question
Ridings for their help with the data collection and analyses. the validity of the hypothesized association (Bornstein & Cecero,
Correspondence concerning this article should be addressed to Jennifer 2000; Miller & Lynam, 2008).
Ruth Lowe, 115 Kastle Hall, Department of Psychology, University of Saulsman and Page (2004) conducted a meta-analysis of 15
Kentucky, Lexington, KY 40506-0044. E-mail: jennifer.lowe@uky.edu independent samples correlating the domains of the FFM with

543
544 LOWE, EDMUNDSON, AND WIDIGER

measures of dependent personality disorder (DPD). They reported dimensional model of personality disorder for the forthcoming
a statistically significant effect size for the relationship of DPD fifth edition of the American Psychiatric Association’s diagnostic
with agreeableness (r ⫽ .05, weighted), but it fell well below their manual, SNAP dependency and DAPP–BQ diffidence have both
cutoff for what would constitute a meaningful effect size (i.e., r ⫽ been included within the domain of neuroticism or emotional
.20). There was, in contrast, a meaningful effect size for neuroti- instability, rather than within agreeableness (Krueger, Skodol,
cism (r ⫽ .41 weighted). The effect size for agreeableness was Livesley, Shrout, & Huang, 2008; Widiger & Simonsen, 2005).
even the lowest among all of the five domains of the FFM. This One potential explanation for the weak and/or inconsistent re-
finding was replicated more recently by Samuel and Widiger lationship of dependency with FFM agreeableness is the assess-
(2008) in their meta-analysis of 16 FFM studies that provided ment instrument. Pincus and Gurtman (1995) suggested the poten-
facet-level assessments of agreeableness (only one of which over- tial importance of measurement instrument in their comparison of
lapped with Saulsman & Page, 2004), reporting an effect size of five measures of dependency with respect to their location along
only .08 between agreeableness and DPD (the highest effect size the IPC. They distinguished between three variants of dependency
for an agreeableness facet was only .16 for modesty). (Pincus, 2002): a love dependency (confined largely to the dimen-
Bornstein and Cecero (2000) conducted a meta-analysis that sion of communion), a submissive dependency (confined largely to
focused specifically on the relationship of dependency with the the dimension of submission or low agency), and an exploitable
domains of the FFM. Their meta-analysis was broader than those dependency (representing a combination of high communion and
of Saulsman and Page (2004) and Samuel and Widiger (2008), as low agency). They suggested that the submissive dependent would
they included measures of trait dependency in addition to measures not be considered to involve much agreeableness, whereas the love
of DPD. They reported that the effect size for agreeableness was, and exploitable variants would. The love dependent would involve
at best, modest, if not weak (r ⫽ .08). One of the worst effect sizes in particular the facets of trust, altruism, and tender-mindedness,
reported by them for an individual study (i.e., Dunkley, Blankstein, whereas the exploitable would involve the facets of straightfor-
& Flett, 1997) was miscoded (it was reported to be ⫺.32 when it wardness, compliance, and modesty (Pincus, 2002).
was in fact .32), but even with this correction, the overall effect It might also be useful to distinguish between measures of
size increased to only .10. They suggested that the much-smaller- trait dependency and DPD (Bornstein, 2005, 2006). Many of the
than-expected correlation between dependency and agreeableness measures of trait dependency are concerned with a maladaptive
warranted further research. dependency (e.g., the Interpersonal Dependency Inventory;
In a fourth meta-analysis of eight studies reporting correlations Hirschfeld et al., 1977), but many researchers are now making a
between the NEO PI–R and measures of DPD, Miller and Lynam distinction between adaptive and maladaptive dependency (Born-
(2008) indicated that although Lynam and Widiger (2001) consid- stein & Huprich, 2006; Cogswell & Alloy, 2006; Dunkley, Blank-
ered dependency to involve “three agreeableness facets (i.e., trust, stein, Zuroff, Lecce, & Hui, 2006; Pincus & Wilson, 2001). Pincus
compliance, modesty), the meta-analytic findings suggest that and Wilson (2001), for instance, suggested that submissive and
none should be included” (Miller & Lynam, 2008, p. 7). They exploitable dependence are more closely associated with maladap-
proposed a revised conceptualization of dependency that excluded tivity than is love dependence, the latter concerning a healthy
any significant representation of any facet from agreeableness, and desire for obtaining and maintaining close relationships. Cogswell
they further demonstrated empirically that this revised version and Alloy (2006) and Dunkley et al. (2006) similarly distinguished
obtains greater validity than the original version, using the NEO between neediness and connectedness components of dependency
PI–R as their criterion measure of the FFM. as assessed by subscales of the Depressive Experiences Question-
Miller and Lynam (2008) are not alone in this conceptualization naire (Blatt, D’Afflittie, & Quinlan, 1976). Connectedness is per-
of dependency. An alternative to the FFM is the six-factor model haps comparable to love dependency, as it involves a valuing of
of Lee and Ashton (2004), who used as their measure of person- relationships and a sensitivity to the effect of one’s actions upon
ality the HEXACO Personality Inventory (HEXACO–PI). One of others. Dunkley et al. and Zuroff, Moskowitz, and Koestner (1996)
the HEXACO–PI facet scales is identified as Dependency, defined validated the adaptive–maladaptive distinction between neediness
as a “need for emotional support from others” (Lee & Ashton, and connectedness by demonstrating a correlation of neediness
2004, p. 334). In other words, Lee and Ashton explicitly included with NEO PI–R neuroticism and connectedness with NEO PI–R
trait dependency as a maladaptive variant of general personality agreeableness. Bornstein and Huprich (2006) likewise distin-
structure but placed the HEXACO–PI Dependency facet scale guished between destructive overdependence and healthy depen-
within the domain of emotional instability (which aligns with FFM dence using the Relationship Profile Test. Consistent with their
neuroticism), rather than within the domain of agreeableness (or expectations, destructive overdependence correlated with NEO
honesty-humility). PI–R neuroticism but not agreeableness, whereas healthy depen-
Similarly, the Schedule for Nonadaptive and Adaptive Person- dence correlated positively with agreeableness and negatively with
ality (SNAP; Clark, 1993) and Dimensional Assessment of Per- neuroticism. However, both Bacchiochi, Bagby, Cristi, and
sonality Pathology—Basic Questionnaire (DAPP–BQ; Livesley & Watson (2003) and McBride, Zuroff, Bagby, and Bacchiochi
Jackson, in press) include Dependency and Diffidence scales, (2006) failed to replicate the connectedness findings of Dunkley et
respectively, along with other maladaptive trait scales. In higher al. for agreeableness.
order factor analyses of the SNAP and DAPP–BQ, the trait de- A distinction between adaptive and maladaptive variants may
pendency and diffidence scales loaded consistently within the also occur for the assessment of FFM agreeableness. None of the
neuroticism factor, even when the factor structure could be said to four previously summarized meta-analyses obtained a moderating
include agreeableness (e.g., Clark, Livesley, Schroeder, & Irish, effect of assessment instrument on the relationship of agreeable-
1996; Clark, Vorhies, & McEwen, 2002). In proposals for a ness with dependency. However, it was also the case that there was,
DEPENDENCY AND AGREEABLENESS 545

in fact, very little variation in the assessment of FFM agreeableness. studies included in the meta-analyses of Bornstein and Cecero
Eleven of the 15 samples considered by Saulsman and Page (2004), (2000), Miller and Lynam (2008), Saulsman and Page (2004), or
17 of the 18 studies considered by Bornstein and Cecero (2000), all Samuel and Widiger (2008) was there a report of differences in the
eight of the studies considered by Miller and Lynam (2008), and relationship of FFM agreeableness with trait dependency or DPD
all 16 studies considered by Samuel and Widiger (2008) used the because of gender (e.g., see Ball, Tennen, Poling, Kranzler, &
NEO PI–R or a measure modeled closely after the NEO PI–R. Rounsaville, 1997; Dunkley et al., 1997; Mongrain, 1993).
Haigler and Widiger (2001) suggested that some relationships of Participants from the student population were enrolled in an
FFM domains with measures of personality disorder, including introductory undergraduate psychology course that awarded class
DPD, might be affected by the fact that the NEO PI–R was credit for study participation. Participants from the substance-
constructed largely for the purpose of providing a measure of dependence residential treatment program were recruited either by
normal personality functioning, particularly items assessing high flyer (containing a brief description of the study) or verbally by
conscientiousness; low neuroticism; high extraversion; high open- members of the research team at one of two locations. Participants
ness; and, most important for the current study, high agreeable- received a packet containing the study materials and consent forms
ness. They indicated that 83% of the NEO PI–R agreeableness and were allowed to complete the study materials at the time and
items, when keyed in the direction of high agreeableness, were location of their choosing. Upon completion of the study, partic-
assessing an adaptive, rather than a maladaptive, trait. They con- ipants from the clinical sample received payment of $15.00 and a
structed an experimentally altered version of the NEO PI–R by debriefing form.
inserting words into the items to change the direction of the Of the 105 college student participants, 94% reported their ages
maladaptivity without otherwise altering its content. For example, as being 18 –23 years, 1 reported her age as being 30 –35, and 5 did
the agreeableness items of “I think of myself as a charitable not report their ages. With respect to ethnicity, 84% were identi-
person,” “I believe that most persons are basically well- fied as Caucasian/White, 5% as African American/Black, 4% as
intentioned,” and “I would rather cooperate with others than com- Hispanic/Latina, and 3% as Asian. Five student participants did
pete with them” (Costa & McCrae, 1992, pp. 71–72) became “I am not indicate their ethnic identification. Of the 69 women of the
so charitable that I give more than I can afford,” “I tend to be residential substance-dependence treatment programs, 7% reported
gullible regarding the intentions of others,” and “I cooperate with their ages as 18 –23 years old, 28% as 24 –29, 19% as 30 –35, 16%
others even when it would be better to competitive” (Haigler & as 36 – 41, and 16% as over 41. Ten participants from the clinical
Widiger, 2001, p. 347). In sum, 83% of the items contained within population did not report their ages. Regarding ethnicity, 59%
the experimentally altered version of NEO PI–R agreeableness from the clinical population were identified as Caucasian/White,
described maladaptive, dysfunctional behavior. They reported in a 23% as African American/Black, and 1 participant identified her-
college-student sample that correlations of agreeableness with self as Native American/Pacific Islander/other. Eleven participants
three DPD scales were significant for the experimentally altered (16%) did not identify their ethnicity.
scale but not for the original NEO PI–R agreeableness.
It is the purpose of the current study to determine whether the
weak and inconsistent relationship of measures of trait dependence Materials
and/or DPD with agreeableness might be due in part to the instru-
ments of assessment. The current study considers 13 alternative The current study included 13 alternative measures of depen-
measures of dependency, including three measures of DPD, eight dency: three measures of DPD, eight measures of trait dependency,
measures of trait dependency (two of which are scales assessing an and two measures of dependency from alternative dimensional
adaptive dependency), and two measures of dependency from models of personality disorder.
alternative dimensional models of personality disorder, along with Millon Clinical Multiaxial Inventory—III (MCMI–III). The
the original and the experimentally altered versions of NEO PI–R MCMI–III (Millon, Millon, & Davis, 1997) is a 175-item true/false
agreeableness (as well as the other four NEO PI–R domain scales). self-report inventory, developed in accordance with the DSM–IV–TR,
The study samples from both a student and a clinical population. which assesses 14 personality disorders as well as 10 other clinical
syndromes. The MCMI–III is among the most frequently used per-
sonality disorder self-report inventories in clinical practice (Widiger
Method
& Boyd, in press) and has been used in prior FFM agreeableness
Participants included, in separate data collections, female col- research (Bornstein & Cecero, 2000; Miller & Lynam, 2008;
lege students and women residing at a substance-dependence res- Samuel & Widiger, 2008; Saulsman & Page, 2004). The DPD
idential treatment program. The current study was confined to scale of the MCMI–III consists of 16 items.
women because DPD has been a disorder that has traditionally Personality Diagnostic Questionnaire— 4 (PDQ– 4). The
(rightly or wrongly) been closely associated with women and PDQ– 4 (Bagby & Farvolden, 2004) is a 99-item true/false self-
feminine gender (Morey, Alexander, & Boggs, 2005), a perception report inventory that assesses 12 personality disorders according to
consistent with a recent meta-analysis of gender differences in the the DSM–IV personality disorder criteria. The PDQ– 4 is perhaps
prevalence of personality disorder (Lynam & Widiger, 2007). As the most commonly used self-report measure of personality disor-
Bornstein (1996) indicated in his review of sex differences in the der within clinical research (Widiger & Boyd, in press) and has
prevalence of DPD, a woman is 40% more likely to be said to have been used in prior FFM agreeableness research (Samuel & Widi-
the disorder when this assessment is provided by self-report in- ger, 2008; Saulsman & Page, 2004). The DPD scale from the
ventories, which are the focus of the current investigation. It is also PDQ– 4 has eight items, each of which corresponds to a DSM–
worth noting though that in none of the prior FFM dependency IV–TR diagnostic criterion.
546 LOWE, EDMUNDSON, AND WIDIGER

Wisconsin Personality Disorders Inventory (WISPI). The study, we administered only the dependency scale, consisting of 21
WISPI (Klein et al., 1993) is a 214-item self-report inventory that items.
assesses the DSM–IV–TR personality disorders. Each item is rated NEO Personality Inventory—Revised (NEO PI–R). The NEO
on a 1–10-point Likert scale. The WISPI has not been used in PI–R (Costa & McCrae, 1992) is a measure of the FFM of
previous FFM agreeableness research. Items of the WISPI corre- personality and contains 240 items that are rated on a Likert scale
spond to DPD diagnostic criteria, albeit written in a manner that ranging from 1 (strongly disagree) to 5 (strongly agree). This
emphasizes an object-relational, interpersonal perspective (Klein instrument is composed of five broad domain scales, which are
et al., 1993). The DPD scale consists of 28 items. each in turn assessed by six underlying facet scales. There are 48
Depressive Experiences Questionnaire (DEQ). The DEQ items for each of the five FFM domain scales. Internal consistency
(Blatt et al., 1976) is a 66-item inventory designed to assess two coefficients have ranged from .86 (Agreeableness) to .92 (Neurot-
psychoanalytically derived dimensions of depression experience: icism), and 7-year test–retest reliability coefficients ranged from
dependency and self-criticism. Each item is rated on a 1–7-point .63 to .81 (Costa & McCrae, 1992).
scale. DEQ dependency assesses “concerns about being rejected, Experimentally altered NEO PI–R Agreeableness scale (EXP–A).
hurting or offending people, and having difficulty in managing The development of the 48 EXP–A items proceeded through two
anger and aggression for fear of losing someone” (Blatt et al., stages. In the first stage, the direction of the adaptiveness versus
1976, pp. 384 –385). The DEQ includes subscales assessing Con- maladaptiveness of each of the 240 NEO PI–R items was deter-
nectedness (29 items) and Neediness (29 items), which represent mined; in the second stage, each item was revised to reverse the
(in part) adaptive and maladaptive dependency, respectively direction of maladaptiveness. For example, Haigler and Widiger
(Dunkley et al., 2006; Zuroff et al., 1996). The findings for the (2001) judged whether it was more adaptive to agree or disagree
DEQ are provided for only the clinical sample. with the NEO PI–R agreeableness item “I have a good deal of faith
Interpersonal Dependency Inventory (IDI). The 48-item IDI in human nature” (Costa & McCrae, 1992, p. 72). In all but a few
(Hirschfeld et al., 1977) consists of three scales, with each item instances, this judgment was straightforward (e.g., it is generally
rated on a 4-point Likert scale. The IDI is one of the most widely preferable to have a good deal of faith in human nature). Items for
used self-report dependence measures over the past 25 years which there was a perceived ambiguity by either judge, or for
(Bornstein, 2005). Two scales assess dependency: Emotional Re- which a disagreement occurred, were discussed until a consensus
liance on Another Person (IDI–ER; sample item, “I do my best decision was reached.
work when I know it will be appreciated”) and Lack of Social To assess the reliability of the coding decisions, three graduate
Self-Confidence (IDI–SSC; sample item, “It is hard for me to ask students in clinical psychology were also asked to indicate (inde-
someone for a favor”). A third scale is a measure of autonomy (i.e., pendently and blindly) whether it would be more adaptive to agree
Assertion of Autonomy; “I prefer to be by myself”). All three or disagree with each of the 240 NEO PI–R items. All three of
scales were administered in the current study, but for the sake of these independent coders agreed unanimously with the authors’
conserving space, only the findings for the IDI–ER (14 items) and judgments on 75% (181) of the 240 items; at least two of the three
IDI–SSC (16 items) are reported, because Assertion of Autonomy coders agreed with the authors on 90% (217). Haigler and Widiger
is not generally considered to be a measure of dependency traits. (2001) reevaluated their decisions for the remaining 23. The final
3 Vector Dependency Inventory (3VDI). The 27-item 3VDI coding indicated that 83% of the NEO PI–R agreeableness items
(Pincus & Wilson, 2001) consists of three scales, with each item were considered to involve a more adaptive, desirable behavior
rated on a 1– 6-point Likert scale. The three scales (each with nine when the person responded in the direction of agreeableness rather
items) concern different variants of dependency, including Love than antagonism.
(3VDI–L), Exploitable (3VDI–E), and Submissive (3VDI–S), cor- The second stage was to alter each item to reverse the direction
responding to segments along the friendly–submissive quadrant of of adaptiveness (or maladaptiveness) without changing the direc-
the IPC (Pincus & Gurtman, 1995). All three subscales were tion in which the item was keyed or otherwise altering the content
included in the current study. of the item. For example, the agreeableness item “I have a good
Dimensional Assessment of Personality Pathology—Basic deal of faith in human nature” was revised to “I have too much
Questionnaire (DAPP–BQ). The DAPP–BQ (Livesley & Jack- faith in the goodness of human nature.” These experimentally
son, in press) contains 560 statements to which an individual altered items were then submitted to three additional graduate
responds on a 5-point Likert-type scale. The DAPP–BQ contains students in clinical psychology to independently and blindly judge
18 scales of personality pathology (e.g., affective lability, social the adaptiveness of each item to provide an estimate of the success
avoidance, conduct problems, compulsivity). The DAPP–BQ is of the item reversals. At least two of the three raters agreed with
among the leading contenders for a dimensional model of person- the alterations for 89% of the items. The wording of the remaining
ality disorder (Clark, 2007; Krueger et al., 2008; Widiger & items was reevaluated. Further details regarding the development
Simonsen, 2005). In the current study, we administered only the of the experimentally altered NEO PI–R agreeableness scale are
diffidence scale, consisting of 16 items. provided in Haigler and Widiger (2001). The experimentally al-
Schedule for Nonadaptive and Adaptive Personality (SNAP). tered agreeableness scale correlated .53 ( p ⬍ .001) with the
The SNAP (Clark, 1993) is a 375-item true/false instrument that original NEO PI–R agreeableness scale in Haigler and Widiger.
assesses a dimensional model of personality disorder containing
three broad temperament and 12 maladaptive personality trait Results
scales (e.g., detachment, exhibitionism, workaholism). The SNAP
is among the leading contenders for a dimensional model of Table 1 provides the mean scores and alpha coefficients for each
personality disorder (Widiger & Simonsen, 2005). In the current measure of dependency and the NEO PI–R domain scales for both
DEPENDENCY AND AGREEABLENESS 547

Table 1
Descriptive Statistics for Dependency and Five-Factor Model Measures

Student Clinical

Measure M SD ␣ M SD ␣ t Cohen’s d

MCMI–III 7.5 5.8 .80 8.9 6.4 .83 ⫺1.48 (df ⫽ 164) ⫺0.23
PDQ–4 1.8 2.2 .73 2.2 2.3 .80 ⫺1.16 (df ⫽ 164) ⫺0.18
WISPI 74.9 26.8 .81 110.1 37.2 .87 ⫺6.38ⴱⴱⴱ (df ⫽ 101) ⫺1.09
DEQ–D — ⫺0.61 0.97 .89
DEQ–N — 0.74 0.74 .83
DEQ–C — ⫺0.66 1.15 .83
IDI–ER 45.1 9.9 .81 47.6 10.3 .83 ⫺1.48 (df ⫽ 152) ⫺0.25
IDI–SSC 33.4 8.1 .82 38.3 8.4 .77 ⫺3.65ⴱⴱⴱ (df ⫽ 152) ⫺0.59
3VDI–L 32.9 10.3 .87 36.6 10.0 .83 ⫺2.21ⴱ (df ⫽ 156) ⫺0.36
3VDI–E 25.0 9.4 .87 30.5 8.2 .68 ⫺3.74ⴱⴱⴱ (df ⫽ 156) ⫺0.62
3VDI–S 38.4 8.2 .83 37.5 8.2 .73 .76 (df ⫽ 156) 0.11
DAPP–BQ 42.3 12.8 .93 40.5 15.2 .94 .81 (df ⫽ 163) 0.13
SNAP 7.1 5.0 .85 10.0 4.2 .76 ⫺3.95ⴱⴱⴱ (df ⫽ 150) ⫺0.63
NEO–N 96.6 25.1 .92 99.6 17.7 .78 ⫺.82 (df ⫽ 153) ⫺0.14
NEO–E 117.7 20.1 .88 99.3 17.8 .80 5.45ⴱⴱⴱ (df ⫽ 145) 0.97
NEO–O 110.1 20.3 .87 100.5 18.1 .81 2.87ⴱⴱ (df ⫽ 145) 0.50
NEO–A 113.3 22.4 .91 99.3 15.9 .74 4.38ⴱⴱⴱ (df ⫽ 142) 0.72
EXP–A 91.2 17.0 .83 88.8 18.1 .79 1.23 (df ⫽ 160) 0.14
NEO–C 113.2 23.8 .93 96.4 16.5 .76 5.02ⴱⴱⴱ (df ⫽ 144) 0.82

Note. MCMI–III ⫽ Millon Clinical Multiaxial Inventory–III (Millon, Millon, & Davis, 1997); PDQ– 4 ⫽ Personality Diagnostic Questionnaire– 4 (Bagby
& Farvolden, 2004); WISPI ⫽ Wisconsin Personality Disorders Inventory (Klein et al., 1993); DEQ–D, DEQ–N, and DEQ–C ⫽ Depressive Experiences
Questionnaire (Dependency, Neediness, and Connectedness subscales, respectively; Blatt, D’Afflittie, & Quinlan, 1976); IDI–ER and IDI–SSC ⫽
Interpersonal Dependency Inventory, Emotional Reliance and Social Self-Confidence subscales, respectively (Hirschfeld et al., 1977); 3VDI–L, 3VDI–E,
and 3VDI–S ⫽ 3 Vector Dependency Inventory, Love, Exploitation, and Submissive subscales, respectively (Pincus & Wilson, 2001); DAPP–BQ ⫽
Diffidence subscale from the Dimensional Assessment of Personality Pathology—Basic Questionnaire (Livesley & Jackson, in press); SNAP ⫽
Dependency subscale from the Schedule for Nonadaptive and Adaptive Personality (Clark, 1993); NEO–N, NEO–E, NEO–O, NEO–A, and NEO–C ⫽ the
Neuroticism, Extraversion, Openness, Agreeableness, and Connectedness subscales, respectively, of the NEO Personality Inventory—Revised (Costa &
McCrae, 1992); EXP–A ⫽ experimentally altered NEO Personality Inventory—Revised Agreeableness scale. There were 105 students and 69 patients.
Degrees of freedom for t values vary because not all persons completed all items on each inventory. Analyses for WISPI and SNAP involved unequal
variances.

p ⬍ .05. ⴱⴱ p ⬍ .01. ⴱⴱⴱ p ⬍ .001.

the student and patient samples. No differences were obtained dependency (Klein et al., 1993; Pincus & Wilson, 2001), and
between the clinical and student samples for neuroticism. This is consistent with their common interpersonal emphasis, they were
perhaps due to the fact that the members of the 6-month residential all highly convergent. In contrast, the SNAP failed to correlate
treatment program had been in treatment for quite a period of time significantly with the WISPI or with any one of the three 3VDI
(an average of 4 months). There is a considerable body of research dependency scales within the clinical sample. The correlations
indicating that scores on neuroticism decrease during the course of within the student sample were generally higher. In fact, within the
treatment, with much of this decrease attributed to changes in student sample, SNAP dependency obtained consistent convergent
mood state (Widiger, in press). The patient group did obtain validity with all other measures of dependency.
significantly higher levels of trait dependency and/or DPD, as Table 3 provides the correlations of the measures of trait de-
assessed by the WISPI, IDI–SSC, 3VDI–L, 3VDI–E, and SNAP. pendency and DPD with the NEO PI–R domain scales and the
In a complementary fashion, the students also obtained signifi- experimentally altered assessment of agreeableness (EXP–A) for
cantly higher scores on NEO PI–R agreeableness, extraversion, both the student and clinical samples (EXP–A correlated .49 and
openness, and conscientiousness. .47 [both p ⬍ .001] with NEO PI–R agreeableness for the student
Table 2 provides the correlations among the trait dependency and clinical samples, respectively). It is evident from Table 3 that
and/or DPD measures for the clinical and student samples. It is measures of trait dependency and DPD correlated more highly
evident from Table 2 that there is quite a range of convergent with EXP–A than with the original version. Only one of the
validity among the measures of trait dependency and DPD, from a measures of dependency correlated significantly with NEO PI–R
low of .03 between the SNAP and 3VDI–Exploitable to .76 be- agreeableness within the student sample, and only two within the
tween the MCMI–III and PDQ– 4 within the clinical sample. None clinical sample. None of the adaptive trait measures (i.e., 3VDI–L
of the trait and DPD measures of dependency correlated signifi- and DEQ-C) correlated positively with agreeableness. When the
cantly with all of the other scales, although both the DAPP–BQ experimentally altered version of the NEO PI–R agreeableness
and IDI–ER failed to correlate significantly with only one other scale was used, only 1 of the 13 measures of dependency failed to
scale (i.e., DEQ Connectedness and 3VDI–E, respectively) within correlate significantly with agreeableness in the clinical sample
the clinical sample. The WISPI and the 3VDI scales are the four (i.e., the SNAP dependency) and only one failed in the student
with the greatest emphasis on an interpersonal conceptualization of sample (i.e., 3VDI–S). In 75% of the 20 instances in which a
548 LOWE, EDMUNDSON, AND WIDIGER

measure of trait dependency or DPD failed to correlate signifi-

3VDI–E, and 3VDI–S ⫽ 3 Vector Dependency Inventory, Love, Exploitation, and Submissive subscales, respectively (Pincus & Wilson, 2001); DAPP–BQ ⫽ Diffidence subscale from the Dimensional
Personality Disorders Inventory (Klein et al., 1993); DEQ–D, DEQ–N, and DEQ–C ⫽ Depressive Experiences Questionnaire (Dependency, Neediness, and Connectedness subscales, respectively; Blatt,
D’Afflittie, & Quinlan, 1976); IDI–ER and IDI–SSC ⫽ Interpersonal Dependency Inventory, Emotional Reliance and Social Self-Confidence subscales, respectively (Hirschfeld et al., 1977); 3VDI–L,
Note. MCMI–III ⫽ Millon Clinical Multiaxial Inventory–III (Millon, Millon, & Davis, 1997); PDQ– 4 ⫽ Personality Diagnostic Questionnaire– 4 (Bagby & Farvolden, 2004); WISPI ⫽ Wisconsin

Assessment of Personality Pathology—Basic Questionnaire (Livesley & Jackson, in press); S ⫽ student; C ⫽ clinical; SNAP ⫽ Dependency subscale from the Schedule for Nonadaptive and Adaptive
.41ⴱ
DAPP–BQ

C
cantly with NEO PI–R agreeableness, this correlation was also
statistically significantly lower than the correlation with maladap-

.66ⴱⴱ
S
tive agreeableness (see Table 3).
Table 4 provides partial correlations between the measures of

Personality (Clark, 1993). There were 105 students and 69 patients. The number for each analysis varies because not all persons completed all items on each inventory (see Table 1).
.30ⴱ
trait dependency and DPD with EXP–A, neuroticism, and consci-

.21
C
3VDI–S

entiousness in both the clinical and student samples. In the student


sample, when variance that could be accounted for by neuroticism

.52ⴱⴱ
.50ⴱⴱ
S

was partialed out, EXP–A continued to correlate significantly with


all of the trait-dependency and DPD measures (a significant cor-

.43ⴱ
relation with 3VDI–S was now also obtained). In the clinical
.10

⫺.03
C
3VDI–E

sample, it continued to correlate with the WISPI, IDI–SSC,


3VDI–L, 3VDI–E, and DAPP–BQ. However, significant correla-
.31ⴱⴱ
.58ⴱⴱ
.65ⴱⴱ
tions with the MCMI–III, PDQ– 4, IDI–ER, and SNAP were lost.
S

On the other hand, in a complementary fashion, when the variance


that could be explained by EXP–A was partialed out, neuroticism
.51ⴱ
.53ⴱ
.32ⴱ
.12
C

failed to correlate significantly with the PDQ– 4, IDI–ER,


3VDI–L

3VDI–L, 3VDI–E, and DAPP–BQ in the clinical sample. The


.49ⴱⴱ
.41ⴱⴱ
.68ⴱⴱ
.52ⴱⴱ

correlations for conscientiousness remained for the IDI–SSC and


S

3VDI–E when we controlled for neuroticism in the clinical sample


and remained for all but one of the measures (WISPI) within the
.39ⴱ
.56ⴱ
.29ⴱ
.50ⴱ
.13
C
IDI–SSC

student sample.
.53ⴱⴱ
.78ⴱⴱ
.36ⴱⴱ
.72ⴱⴱ
.70ⴱⴱ
S

Discussion
The results of the current study suggest that the failure to
.59ⴱ
.35ⴱ

.68ⴱ
.34ⴱ
.32ⴱ
.12
C
IDI–ER

recognize in past research the potentially important role of agree-


ableness in conceptualizing and understanding dependency may
.67ⴱⴱ
.42ⴱⴱ
.46ⴱⴱ
.49ⴱⴱ
.58ⴱⴱ
.57ⴱⴱ

reflect in part how FFM agreeableness is being assessed. The


S

assessment of the FFM in the four published meta-analyses was


confined largely, if not solely, to the NEO PI–R or a measure
DEQ–C

.55ⴱⴱ

.63ⴱⴱ

.73ⴱⴱ

modeled after the NEO PI–R. In the current study, only 1 of the 13
C

.24

.15

.15
.19

measures of trait dependency or DPD within the student sample


and only 2 of 13 within the clinical sample correlated significantly
DEQ–N

with NEO PI–R agreeableness, consistent with the prior research.


.56ⴱⴱ
.55ⴱⴱ
.41ⴱⴱ
.43ⴱⴱ
.43ⴱⴱ
.61ⴱⴱ
.26ⴱ
C

.24

However, when the experimentally altered version of the NEO


PI–R agreeableness scale was used, only 1 of the 13 measures of
trait dependency or DPD failed to correlate significantly with
DEQ–D

.73ⴱⴱ
.80ⴱⴱ
.69ⴱ
.41ⴱ
.56ⴱ
.33ⴱ
.67ⴱ
.45ⴱ
.34ⴱ

agreeableness within the clinical sample and only 3 of 13 failed to


C

do so within the student sample.


Theoretical models should be revised in response to empirical
.55ⴱ
.59ⴱ
.46ⴱ
.51ⴱ
.53ⴱ
.68ⴱ
.42ⴱ
.54ⴱ
.48ⴱ
.11
C

refutation, but failures to confirm hypotheses do at times reflect


WISPI

methodological limitations of the existing research (Cronbach &


.72ⴱⴱ
.71ⴱⴱ
.56ⴱⴱ
.62ⴱⴱ
.37ⴱⴱ
.67ⴱⴱ
.65ⴱⴱ

Meehl, 1955). As suggested in the classic article by Cronbach and


Correlations Among Measures of Dependency




S

Meehl (1955), no single measure should be understood as provid-


ing an operational definition of a construct, and Costa and McCrae
.34ⴱ

.26ⴱ

.32ⴱ
.37ⴱ

.30ⴱ

.43ⴱ
.35ⴱ
.25

.12

.15

.16
C

(1995) acknowledged some time ago that “the NEO PI–R is not the
PDQ–4

gold standard by which other conceptualizations of the FFM


.58ⴱⴱ

.46ⴱⴱ
.57ⴱⴱ
.48ⴱⴱ
.59ⴱⴱ
.40ⴱⴱ
.57ⴱⴱ
.78ⴱⴱ

should be judged” (p. 36). The NEO PI–R is justifiably the





S

predominant measure of the FFM, given its very extensive con-


struct validation (Costa & McCrae, 1992). However, it was con-
.76ⴱ
.28ⴱ
.35ⴱ
.32ⴱ

.39ⴱ
.40ⴱ

.25ⴱ

.45ⴱ
.61ⴱ
.19

.22

.20
MCMI–III

structed largely for the purpose of assessing the FFM within the
p ⬍ .05. ⴱⴱ p ⬍ .01.

general population, and it may lack some fidelity in the assessment


.77ⴱⴱ
.67ⴱⴱ

.57ⴱⴱ
.69ⴱⴱ
.61ⴱⴱ
.64ⴱⴱ
.41ⴱⴱ
.67ⴱⴱ
.82ⴱⴱ

of abnormal variants of some of the FFM domains.





S

A reasonable concern regarding the methodology of the current


study is that the authors might have simply replaced NEO PI–R
DAPP–BQ
Measure

IDI–SSC

agreeableness items with dependency items. However, the exper-


Table 2

3VDI–L
3VDI–E
3VDI–S
DEQ–D
DEQ–N
DEQ–C
IDI–ER
PDQ–4
WISPI

SNAP

imental manipulation was instead confined to reversing the direc-


tion of the maladaptivity of each respective item without otherwise

DEPENDENCY AND AGREEABLENESS 549

Table 3
Correlations of Measures of Dependency With Five-Factor Model Scales

NEO–N NEO–E NEO–O NEO–A EXP–A NEO–C

Measure S C S C S C S C S C S C

MCMI–III .43ⴱⴱ .53ⴱⴱ ⫺.22 .12 ⫺.14 .09 ⫺.04 .44ⴱⴱ .43ⴱⴱa .44ⴱⴱ ⫺.31ⴱⴱ .05
PDQ–4 .38ⴱⴱ .41ⴱⴱ ⫺.16 .04 ⫺.11 .06 ⫺.14 .32ⴱ .31ⴱⴱa .39ⴱⴱ ⫺.25ⴱ .05
WISPI .36ⴱⴱ .13 ⫺.24ⴱ ⫺.06 ⫺.23ⴱ ⫺.21 ⫺.15 ⫺.06 .28ⴱa .42ⴱⴱa ⫺.28ⴱ .20
DEQ–D — .51ⴱⴱ — .07 — .11 — .20 — .35ⴱⴱ — .12
DEQ–N — .34ⴱ — .00 — .03 — .00 — .43ⴱⴱa — ⫺.08
DEQ–C — .45ⴱ — .11 — .15 — .25 — .31ⴱⴱ — .10
IDI–ER .39ⴱⴱ .34ⴱⴱ ⫺.11 .29ⴱ ⫺.04 ⫺.06 .23ⴱ .18 .23ⴱ .30ⴱ ⫺.30ⴱ
IDI–SSC .45ⴱⴱ .20 ⫺.38ⴱⴱ .10 ⫺.22ⴱ .08 ⫺.09 .16 .45ⴱⴱa .42ⴱⴱa ⫺.41ⴱⴱ ⫺.37ⴱⴱ
3VDI–L .33ⴱⴱ .28ⴱ ⫺.31ⴱⴱ .04 ⫺.20 .14 .03 .24 .49ⴱⴱa .43ⴱⴱa ⫺.18 ⫺.05
3VDI–E .48ⴱⴱ .23ⴱ ⫺.33ⴱⴱ ⫺.06 ⫺.23ⴱ .17 ⫺.12 .19 .40ⴱⴱa .34ⴱⴱa ⫺.43ⴱⴱ ⫺.40ⴱⴱ
3VDI–S .35ⴱⴱ .43ⴱⴱ .00 .11 ⫺.08 .10 .01 .24 .20 .37ⴱⴱa ⫺.03 .03
DAPP–BQ .38ⴱⴱ .44ⴱⴱ ⫺.23ⴱ .12 ⫺.31ⴱⴱ .10 .05 .18 .49ⴱⴱa .52ⴱⴱa ⫺.29ⴱ ⫺.08
SNAP .49ⴱⴱ .43ⴱⴱ ⫺.19 .12 ⫺.14 ⫺.02 ⫺.02 .12 .38ⴱⴱa .13 ⫺.30ⴱⴱ .23

Note. NEO–N, NEO–E, NEO–O, NEO–A, and NEO–C ⫽ the Neuroticism, Extraversion, Openness, Agreeableness, and Connectedness subscales,
respectively, of the NEO Personality Inventory—Revised (Costa & McCrae, 1992); EXP–A ⫽ experimentally altered NEO Personality Inventory—Revised
Agreeableness scale; S ⫽ student; C ⫽ clinical; MCMI–III ⫽ Millon Clinical Multiaxial Inventory–III (Millon, Millon, & Davis, 1997); PDQ– 4 ⫽
Personality Diagnostic Questionnaire– 4 (Bagby & Farvolden, 2004); WISPI ⫽ Wisconsin Personality Disorders Inventory (Klein et al., 1993); DEQ–D,
DEQ–N, and DEQ–C ⫽ Depressive Experiences Questionnaire (Dependency, Neediness, and Connectedness subscales, respectively; Blatt, D’Afflittie, &
Quinlan, 1976); IDI–ER and IDI–SSC ⫽ Interpersonal Dependency Inventory, Emotional Reliance and Social Self-Confidence subscales, respectively
(Hirschfeld et al., 1977); 3VDI–L, 3VDI–E, and 3VDI–S ⫽ 3 Vector Dependency Inventory, Love, Exploitation, and Submissive subscales, respectively
(Pincus & Wilson, 2001); DAPP–BQ ⫽ Diffidence subscale from the Dimensional Assessment of Personality Pathology—Basic Questionnaire (Livesley
& Jackson, in press); SNAP ⫽ Dependency subscale from the Schedule for Nonadaptive and Adaptive Personality (Clark, 1993). There were 105 students
and 69 patients. The number for each analysis varies because not all persons completed all items on each inventory (see Table 1).
a
The correlation between dependency measure and EXP–A was significantly higher ( p ⫽ .05) than the correlation between dependency measure and
NEO–A.

p ⬍ .05. ⴱⴱ p ⬍ .01.

altering its content. For example, the NEO PI–R altruism items “I 2008), as anger contributes to the occurrence of antagonistic be-
try to be courteous to everyone I meet,” “Some people think of me havior (Costa & McCrae, 1992). However, just as it is the case that
as cold and calculating,” “I think of myself as a charitable person,” not all antagonistic behavior is driven by the negative affect of
“Some people think I’m selfish and egotistical,” and “I go out of anger (e.g., manipulation, exploitation, arrogance, and deception),
my way to help others if I can” (Costa & McCrae, 1992, p. 72) all the disposition to be docile, servile, self-effacing, compliant, obe-
describe behavior for which it would be preferable (or adaptive) to dient, gullible, and submissive may not necessarily be driven by
endorse the item in the altruistic direction. The experimentally anxious insecurity. There is a rich clinical, empirical, and theoret-
altered versions were “I am overly courteous to everyone I meet,” ical literature on interpersonal models of personality and person-
“I can be cold and calculating when it’s necessary,” “I am so ality disorders (Benjamin, Rothweiler, & Critchfield, 2006; Pin-
charitable that I give more than I can afford,” “Most people think cus, 2005) that should be brought to bear on one’s understanding
that I take good care of my own needs,” and “I have sacrificed my and treatment of maladaptive dependent behavior. This clinical
own needs to help others” (respectively). The correlation of NEO understanding will be sorely handicapped if trait dependency and
PI–R agreeableness with the experimentally altered version was DPD are understood solely in terms of neuroticism and negative
not so high that one could interpret the two measures as assessing affect.
the same construct (the median convergent validity of any two The relationship of the experimentally altered measure of agree-
measures of DPD, as reported in 24 studies, is only .59; Widiger ableness with the WISPI, 3VDI–E, 3VDI–L, IDI–SSC, and
& Boyd, in press), nor should they be interpreted as such. The DAPP–BQ held up for both the student and clinical samples when
NEO PI–R can be said to be assessing adaptive agreeableness variance due to neuroticism was removed (as well as the MCMI–
while the EXP–A would be assessing maladaptive agreeableness. III, PDQ– 4, and SNAP within the student sample). The WISPI,
In sum, an adequate clinical understanding of dependency 3VDI–E, 3VDI–L, and IDI–SSC are among the more interperson-
should probably include the excessive and extreme gullibility, ally oriented measures of trait dependency and/or DPD (Hirschfeld
self-effacement, self-sacrifice, and compliance of persons with this et al., 1977; Klein et al., 1993; Pincus & Wilson, 2001). Never-
personality disorder. There is much that is interpersonal about theless, those instances where significant correlations were lost within
dependency (Wiggins, 1982), and clinicians may fail to appreciate the clinical sample when variance because of neuroticism was first
this point if it is understood to be just a component of neuroticism, removed could suggest that a degree of neuroticism was uninten-
as in the models of Clark (1993), Krueger et al. (2008), Lee and tionally injected into the assessment of maladaptive agreeableness.
Ashton (2004), and Livesley and Jackson (in press). The self- On the other hand, correlations of neuroticism with measures of
consciousness, vulnerability, and insecurity of neuroticism con- trait dependency and DPD were also lost when we controlled for
tribute to the occurrence of dependent behavior (Miller & Lynam, maladaptive agreeableness within the clinical sample. In other
550 LOWE, EDMUNDSON, AND WIDIGER

Table 4 petent and needing support, assistance, and guidance from others).
Partial Correlations of Measures of Dependency With EXP–A, It should be noted, however, that this negative correlation with
NEO–N, and NEO–C conscientiousness was not replicated well within the clinical sam-
ple, with the exception of the IDI–SSC and 3VDI–E.
EXP–A NEO–N NEO–C The current study also found, at best, only a modest effect for
controlling for controlling for controlling for
NEO–N EXP–A NEO–N
the measure of dependency. The effect of the experimental ma-
nipulation of agreeableness items was largely consistent across 12
Measure S C S C S C of the 13 measures of dependency, including measures of DPD,
adaptive and maladaptive trait dependency, and one of the two
MCMI–III .43ⴱⴱ .17 .49ⴱⴱ .39ⴱⴱ ⫺.26ⴱ .02
alternative dimensional models of personality disorder. The sole
PDQ–4 .32ⴱⴱ .23 .42ⴱⴱ .16 ⫺.25ⴱ .01
WISPI .26ⴱ .41ⴱⴱ .36ⴱⴱ ⫺.17 ⫺.22 ⫺.17 exception occurred for the SNAP within the clinical sample. SNAP
DEQ–D — .20 — .23 — .00 dependency did correlate significantly with the experimentally
DEQ⫺N — .38ⴱⴱ — .02 — ⫺.08 manipulated version of NEO PI–R agreeableness within the stu-
DEQ–C — .04 — .32ⴱ — .07 dent sample but not within the clinical sample. The failure to
IDI–ER .29ⴱⴱ .27 .45ⴱⴱ .12 ⫺.25ⴱ .11
IDI–SSC .51ⴱⴱ .37ⴱⴱ .50ⴱⴱ ⫺.07 ⫺.31ⴱⴱ ⫺.44ⴱⴱ obtain a significant correlation for the SNAP, however, could be
3VDI–L .52ⴱⴱ .29ⴱ .33ⴱⴱ .02 ⫺.14 ⫺.02 consistent with the convergent validity findings, which suggested
3VDI–E .38ⴱⴱ .33ⴱ .50ⴱⴱ ⫺.07 ⫺.35ⴱⴱ ⫺.41ⴱⴱ that the SNAP assessment of trait dependency is the least inter-
3VDI–S .33ⴱⴱ .14 .36ⴱⴱ .29ⴱ .02 .04 personal of all 13 measures, failing to correlate as well with the
DAPP–BQ .54ⴱⴱ .32ⴱ .39ⴱⴱ .22 ⫺.26ⴱ ⫺.11
WISPI or with any one of the three 3VDI measures in the clinical
SNAP .44ⴱⴱ ⫺.15 .53ⴱⴱ .43ⴱⴱ ⫺.24ⴱ .24
sample.
Note. NEO–N and NEO–C ⫽ the Neuroticism and Connectedness sub- The current study found weak support for the distinction be-
scales, respectively, of the NEO Personality Inventory—Revised (Costa & tween adaptive and maladaptive trait dependency. The findings
McCrae, 1992); EXP–A ⫽ experimentally altered NEO Personality Inven- were consistent across DEQ connectedness and neediness, incon-
tory—Revised Agreeableness scale; S ⫽ student; C ⫽ clinical; MCMI–
III ⫽ Millon Clinical Multiaxial Inventory–III (Millon, Millon, & Davis, sistent with Dunkley et al. (2006) and Zuroff et al. (1996) but
1997); PDQ– 4 ⫽ Personality Diagnostic Questionnaire– 4 (Bagby & Far- consistent with Bacchiochi et al. (2003) and McBride et al. (2006).
volden, 2004); WISPI ⫽ Wisconsin Personality Disorders Inventory (Klein Specific relationships were obtained with the 3VDI scales but not
et al., 1993); DEQ–D, DEQ–N, and DEQ–C ⫽ Depressive Experiences in the precise manner expected. The latter findings are perhaps best
Questionnaire (Dependency, Neediness, and Connectedness subscales, re-
understood as reflecting limitations of the NEO PI–R facet scales
spectively; Blatt, D’Afflittie, & Quinlan, 1976); IDI–ER and IDI–SSC ⫽
Interpersonal Dependency Inventory, Emotional Reliance and Social Self- (original and experimentally altered) in providing truly specific
Confidence subscales, respectively (Hirschfeld et al., 1977); 3VDI–L, assessments of octants of the IPC, representing instead complex
3VDI–E, and 3VDI–S ⫽ 3 Vector Dependency Inventory, Love, Exploi- blends of agency and communion.
tation, and Submissive subscales, respectively (Pincus & Wilson, 2001);
DAPP–BQ ⫽ Diffidence subscale from the Dimensional Assessment of
Personality Pathology—Basic Questionnaire (Livesley & Jackson, in Limitations
press); SNAP ⫽ Dependency subscale from the Schedule for Nonadaptive
and Adaptive Personality (Clark, 1993). There were 105 students and 69 The current study sampled both a student and a clinical popu-
patients; the numbers vary because not all persons completed all items on lation, providing an ability to directly compare findings across two
each inventory (see Table 1).
ⴱ different populations that have been studied separately in prior
p ⬍ .05. ⴱⴱ p ⬍ .01.
research. No prior FFM dependency study included both a clinical
and student sample. However, the sampling was confined solely to
words, their shared variance may simply reflect that both neurot- female participants. Dependency scores are much higher in
icism and maladaptive agreeableness were assessing some com- women, particularly when the assessment is conducted with self-
mon impairments in functioning. report inventories, which were the focus of the current investiga-
The current study found partial support for the suggestion of tion. Nevertheless, the relationship of dependency to female sex
Miller and Lynam (2008) to revise the FFM conceptualization of and gender is controversial, as the relationship could reflect biases
dependency to include low facets of conscientiousness. Most of the in the conceptualization or assessment of trait dependency or DPD
measures of dependency did correlate negatively with conscien- (Bornstein, 1996; Morey et al., 2005; Widiger & Bornstein, 2001).
tiousness in the student sample, consistent with the meta-analysis Sex differences tend not to be found when free-response or per-
of Miller and Lynam (2008). Low conscientiousness was not formance tests are used (i.e., tests previously identified as projec-
included in the FFM conceptualization of DPD by Widiger et al. tive tests; Bornstein, 1996, 2005). No prior FFM dependency study
(2002), nor was it evident in researchers’ or clinicians’ descrip- has used free response or performance tests and it would be of
tions of prototypic cases of DPD (Lynam & Widiger, 2001; Sam- interest to extend this research to other approaches for assessing
uel & Widiger, 2004). Low conscientiousness may in fact be an dependency, including behavioral indices (Rubinstein, 2007) as
important trait of dependency that was not adequately recognized well as free response or performance tests (Bornstein, 2005;
in prior FFM conceptualizations that has become apparent through Ganellen, 2007).
empirical research. A valid description of dependency should It would also be useful for future studies to explore whether the
perhaps recognize the presence of (for instance) low competence, results generalize to men. We expect that men will obtain lower
low self-discipline, immaturity, and poor impulse control (low scores on the self-report measures of dependency but that the
constraint) in persons high in dependency (reflecting in part their correlations between the measures of the FFM and dependency
belief and presentation of themselves as being helpless and incom- will be largely replicated, consistent with all of the prior studies
DEPENDENCY AND AGREEABLENESS 551

that considered the effect of gender on the relationship of trait and DPD and with the original and experimentally altered assess-
dependency and DPD with agreeableness (Bornstein & Cecero, ments of FFM agreeableness.
2000; Miller & Lynam, 2008; Samuel & Widiger, 2008; Saulsman
& Page, 2004). On the other hand, studies using trait measures of Conclusions
dependency (e.g., the IDI), rather than measures of DPD (e.g., the
MCMI–III), have obtained elevated scores for men who were Dependency has long been considered to be a disorder primarily
violently aggressive toward women (Bornstein, 2006). It would of interpersonal relatedness (Pincus, 2002). From the perspective
then be of interest to extend the current research to include samples of the FFM, this would imply a relationship with agreeableness
of men who are violently aggressive toward women. We hypoth- (Wiggins & Pincus, 1989, 2002). Nevertheless, a relationship of
esize that a positive relationship of dependency will still be ob- dependency with FFM agreeableness has not been observed in
tained with maladaptive agreeableness, although the dependent quite a number of studies. In fact, four largely independent meta-
behavior in violent men could indeed be due more to elevated analyses have all failed to report a consistent relationship, leading
neuroticism than to maladaptive agreeableness (Widiger & some to suggest a reformulation of an FFM conceptualization of
Mullins-Sweatt, 2004). dependency. The results of the current study suggest instead that
Sampling of persons within a residential treatment program for the absence of an empirical relationship may reflect a methodolog-
substance abuse may also not have been optimal. The ideal sample ical limitation of the existing research, most notably the reliance on
might be persons in treatment for DPD or, perhaps more realistic, measures that are failing to provide an adequate assessment of
persons in treatment for a personality disorder. Only one of the maladaptive variants of agreeableness.
prior FFM dependency studies sampled persons in treatment for a
personality disorder (Soldz, Budman, Demby, & Merry, 1993), References
and even in that instance, only 5% of the persons met diagnostic
criteria for DPD. American Psychiatric Association. (2000). Diagnostic and statistical man-
Another possibility is to sample persons in treatment for depres- ual of mental disorders (4th ed., text rev.). Washington, DC: Author.
Bacchiochi, J. R., Bagby, R. M., Cristi, C., & Watson, J. (2003). Validation
sion, given the considerable interest in the study of the relationship
of connectedness and neediness as dimensions of the dependency con-
of trait dependency, DPD, and neuroticism with depression. The struct. Cognitive Therapy and Research, 27, 233–242.
authors of the current study preferred a sample without elevated Bagby, R. M., & Farvolden, P. (2004). The Personality Diagnostic
scores on neuroticism that could be attributed to depressed mood Questionnaire– 4 (PDQ– 4). In M. J. Hilsenroth, D. L. Segal, & M.
(Widiger, in press). However, Costa, Bagby, Herbst, and McCrae Hersen (Eds.), Comprehensive handbook of psychological assessment:
(2005) argue that elevated scores on neuroticism because of de- Vol. 2. Personality assessment (pp. 122–133). New York: Wiley.
pressed mood should not be understood as simply a reflection of Bagby, R. M., Gilchrist, E. J., Rector, N. A., Dickens, S. E., Joffe, R. T.,
this depressed mood, because these scores do correlate with other Levitt, A., et al. (2001). The stability and validity of the sociotropy and
variables unrelated to depression, and they have incremental va- autonomy personality dimensions as measured by the Revised Personal
lidity in the prediction of personality-relevant criteria above and Style Inventory. Cognitive Therapy and Research, 25, 765–779.
Ball, S. A., Tennen, H., Poling, J. C., Kranzler, H. R., & Rounsaville, B. J.
beyond the effects of severity of depression. Costa et al. also
(1997). Personality, temperament, and character dimensions and the
demonstrated empirically no significant changes in NEO PI–R DSM–IV personality disorders in substance abusers. Journal of Abnor-
profiles over the course of treatment (however, there was substan- mal Psychology, 106, 545–553.
tially more change in the depression and vulnerability facets of Benjamin, L. S., Rothweiler, J. C., & Critchfield, K. L. (2006). The use of
neuroticism than for any other NEO PI–R facet scores). In any Structural Analysis of Social Behavior (SASB) as an assessment tool.
case, the concern is perhaps moot in the current study, given the Annual Review of Clinical Psychology, 2, 83–109.
relatively low levels of neuroticism for the persons within clinical Blais, M. A. (1997). Clinician ratings of the five-factor model of person-
treatment. The clinical sample did have elevated scores on mea- ality and the DSM–IV personality disorders. Journal of Nervous and
sures of trait dependency and/or DPD and lower scores on NEO Mental Disease, 185, 388 –393.
PI–R agreeableness (as well as extraversion, conscientiousness, Blatt, S. J., D’Afflittie, J. P., & Quinlan, D. M. (1976). Experiences of
depression in normal young adults. Journal of Abnormal Psychology, 85,
and openness).
383–389.
It would be of interest for future research to consider additional Bornstein, R. F. (1996). Sex differences in dependent personality disorder
measures of the FFM, perhaps in particular the HEXACO–PI (Lee prevalence rates. Clinical Psychology: Science and Practice, 3, 1–12.
& Ashton, 2004), as it includes a scale for dependency within its Bornstein, R. F. (2005). The dependent patient: Diagnosis, assessment, and
assessment of general personality structure. However, the treatment. Professional Psychology: Research and Practice, 36, 82– 89.
HEXACO–PI assessment of dependency is concerned with feel- Bornstein, R. F. (2006). The complex relationship between dependency
ings of emotional instability or neuroticism, and its items are then and domestic violence: Converging psychological factors and social
perhaps confined to feelings of self-consciousness, vulnerability, forces. American Psychologist, 61, 595– 606.
and insecurity. Although it may include items that describe being Bornstein, R. F., & Cecero, J. J. (2000). Deconstructing dependency in a
docile, servile, self-sacrificing, modest, compliant, clinging, obe- five-factor world: A meta-analytic review. Journal of Personality As-
sessment, 74, 324 –343.
dient, gullible, submissive, and self-effacing (i.e., trait terms typ-
Bornstein, R. F., & Huprich, S. K. (2006). Construct validity of the
ically associated with dependent behavior), the items are perhaps Relationship Profile Test: Three-year retest reliability and links with
coupled with feelings of insecurity and self-consciousness with core personality traits, object relations, and interpersonal problems.
regard to this manner of interpersonal relatedness. In any case, it Journal of Personality Assessment, 86, 162–171.
might be of interest in future research to explore the relationship of Clark, L. A. (1993). Manual for the Schedule for Nonadaptive and Adap-
HEXACO–PI dependency with other measures of trait dependency tive Personality. Minneapolis: University of Minnesota Press.
552 LOWE, EDMUNDSON, AND WIDIGER

Clark, L. A. (2007). Assessment and diagnosis of personality disorder: personality to understand sex differences in the personality disorders.
Perennial issues and an emerging reconceptualization. Annual Review of Journal of Personality Disorders, 21, 583– 602.
Psychology, 58, 227–257. McBride, C., Zuroff, D. C., Bagby, R. M., & Bacchiochi, J. (2006).
Clark, L. A., Livesley, W. J., Schroeder, M. L., & Irish, S. L. (1996). Depressive Experiences Questionnaire: Does it measure maladaptive
Convergence of two systems for assessing personality disorder. Psycho- and adaptive forms of dependency? Social Behavior and Personality, 34,
logical Assessment, 8, 294 –303. 1–16.
Clark, L. A., Vorhies, L., & McEwen, J. L. (2002). Personality disorder Miller, J. D., & Lynam, D. R. (2008). Dependent personality disorder:
symptomatology from the five-factor perspective. In P. T. Costa & T. A. Comparing an expert generated and empirically derived five-factor
Widiger (Eds.), Personality disorders and the five-factor model of per- model personality disorder count. Assessment, 15, 4 –15.
sonality (2nd ed., pp. 125–148). Washington, DC: American Psycholog- Millon, T., Millon, C., & Davis, R. (1997). Millon Clinical Multiaxial
ical Association. Inventory: MCMI–III. Upper Saddle River, NJ: Pearson Assessments.
Cogswell, A., & Alloy, L. B. (2006). The relation of neediness and Axis II Mongrain, M. (1993). Dependency and self-criticism located within the
pathology. Journal of Personality Disorders, 30, 16 –21. five-factor model of personality. Personality and Individual Differences,
Costa, P. T., Bagby, R. M., Herbst, J. F., & McCrae, R. R. (2005). 15, 455– 462.
Personality self-reports are concurrently reliable and valid during acute Morey, L. C., Alexander, G. M., & Boggs, C. (2005). Gender and person-
depressive episodes. Journal of Affective Disorders, 89, 45–55. ality disorder. In J. Oldham, A. Skodol, & D. Bender (Eds.), Textbook of
Costa, P. T., & McCrae, R. R. (1985). The NEO Personality Inventory personality disorders (pp. 541–554). Washington, DC: American Psy-
manual. Odessa, FL: Psychological Assessment Resources. chiatric Association.
Costa, P. T., & McCrae, R. R. (1990). Personality disorders and the Mullins-Sweatt, S. N., & Widiger, T. A. (2007). The Shedler-Westen
five-factor model of personality. Journal of Personality Disorders, 4, Assessment Procedure from the perspective of general personality struc-
362–371. ture. Journal of Abnormal Psychology, 116, 618 – 623.
Costa, P. T., & McCrae, R. R. (1992). Revised NEO Personality Inventory Pincus, A. L. (2002). Constellations of dependency within the five factor
(NEO PI–R) and NEO Five-Factor Inventory (NEO–FFI) professional model of personality. In P. T. Costa & T. A. Widiger (Eds.), Personality
manual. Odessa, FL: Psychological Assessment Resources. disorders and the five factor model of personality (pp. 203–214). Wash-
ington, DC: American Psychological Association.
Costa, P. T., & McCrae, R. R. (1995). Domains and facets: Hierarchical
Pincus, A. L. (2005). The interpersonal nexus of personality disorders. In
personality assessment using the Revised NEO Personality Inventory.
S. Strack (Ed.), Handbook of personology and psychopathology (pp.
Journal of Personality Assessment, 64, 21–50.
120 –139). New York: Wiley.
Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological
Pincus, A. L., & Gurtman, M. B. (1995). The three faces of interpersonal
tests. Psychological Bulletin, 52, 281–302.
dependency: Structural analyses of self-report dependency measures.
Dunkley, D. M., Blankstein, K. R., & Flett, G. L. (1997). Specific
Journal of Personality and Social Psychology, 69, 744 –758.
cognitive-personality vulnerability styles in depression and the five-
Pincus, A. L., & Wilson, K. R. (2001). Interpersonal variability in depen-
factor model of personality. Personality and Individual Differences, 23,
dent personality. Journal of Personality, 69, 223–251.
1041–1053.
Rubinstein, G. (2007). The behavioural indicators of dependency and the
Dunkley, D. M., Blankstein, K. R., Zuroff, D. C., Lecce, S., & Hui, D.
five-factor model of personality. Psychology and Psychotherapy: The-
(2006). Neediness and connectedness and the five factor model of
ory, Research, and Practice, 80, 333–342.
personality. European Journal of Personality, 20, 123–136.
Samuel, D. B., & Widiger, T. A. (2004). Clinicians’ descriptions of
Ganellen, R. J. (2007). Assessing normal and abnormal personality func-
prototypic personality disorders. Journal of Personality Disorders, 18,
tioning: Strengths and weaknesses of self-report, observer, and perfor-
286 –308.
mance-based methods. Journal of Personality Assessment, 89, 30 – 40. Samuel, D. B., & Widiger, T. A. (2008). A meta-analytic review of the
Haigler, E. D., & Widiger, T. A. (2001). Experimental manipulation of relationships between the five-factor model and DSM–IV–TR personality
NEO PI–R items. Journal of Personality Assessment, 77, 339 –358. disorders: A facet level analysis. Clinical Psychology Review, 28, 1326 –
Hirschfeld, R. M. A., Klerman, G. L., Gough, H. G., Barret, J., Korchin, 1342.
S. J., & Chodoff, P. (1977). A measure of interpersonal dependency. Saulsman, L. M., & Page, A. C. (2004). The five-factor model and
Journal of Personality Assessment, 41, 610 – 619. personality disorder empirical literature: A meta-analytic review. Clin-
Klein, M. H., Benjamin, L. S., Rosenfeld, R., Treece, C., Husted, J., & ical Psychology Review, 23, 1055–1085.
Greist, J. H. (1993). The Wisconsin Personality Disorders Inventory: I. Soldz, S., Budman, S., Demby, A., & Merry, J. (1993). Representation of
Development, reliability, and validity. Journal of Personality Disorders, personality disorders in circumplex and five-factor space: Explorations
7, 285–303. with a clinical sample. Psychological Assessment, 5, 41–52.
Krueger, R. F., Skodol, A. E., Livesley, W. J., Shrout, P. E., & Huang, Y. Sprock, J. (2002). A comparative study of the dimensions and facets of the
(2008). Synthesizing dimensional and categorical approaches to person- five-factor model in the diagnosis of cases of personality disorder.
ality disorders: Refining the research agenda for DSM–V. In J. E. Helzer, Journal of Personality Disorders, 16, 402– 423.
H. C. Kraemer, R. F. Krueger, H.-U. Wittchen, P. J. Sirovatka, & D. A. Widiger, T. A. (in press). Neuroticism. In M. Leary & R. Hoyle (Eds.),
Regier (Eds.), Dimensional approaches in diagnostic classification (pp. Handbook of individual differences in social behavior. New York:
85–99). Washington, DC: American Psychiatric Association. Guilford Press.
Lee, K., & Ashton, M. C. (2004). Psychometric properties of the HEXACO Widiger, T. A., & Bornstein, R. F. (2001). Histrionic, narcissistic, and
Personality Inventory. Multivariate Behavioral Research, 39, 329 –358. dependent personality disorders. In H. E. Adams & P. Sutker (Eds.),
Livesley, W. J., & Jackson, D. (in press). Manual for the Dimensional Comprehensive handbook of psychopathology (3rd ed., pp. 507–529).
Assessment of Personality Pathology—Basic Questionnaire. Port Huron, New York: Plenum Press.
MI: Sigma Press. Widiger, T. A., & Boyd, S. (in press). Personality disorder assessment
Lynam, D. R., & Widiger, T. A. (2001). Using the five-factor model to instruments. In J. N. Butcher (Ed.), Oxford handbook of personality
represent the DSM–IV personality disorders: An expert consensus ap- assessment (3rd ed.). New York: Oxford University Press.
proach. Journal of Abnormal Psychology, 110, 401– 412. Widiger, T. A., & Costa, P. T., Jr. (2002). Five-factor model personality
Lynam, D. R., & Widiger, T. A. (2007). Using a general model of disorder research. In P. T. Costa & T. A. Widiger (Eds.), Personality
DEPENDENCY AND AGREEABLENESS 553

disorders and the five-factor model of personality (2nd ed., pp. 59 – 87). Wiggins, J. S., & Pincus, A. L. (1989). Conceptions of personality disor-
Washington, DC: American Psychological Association. ders and dimensions of personality. Psychological Assessment, 1, 305–
Widiger, T. A., & Mullins-Sweatt, S. (2004). Typology of maritally violent 316.
men: A discussion of Holtzworth-Monroe and Meehan (2003). Journal Wiggins, J. S., & Pincus, A. L. (2002). Personality structure and the
of Interpersonal Violence, 19, 1396 –1400. structure of personality disorders. In P. T. Costa & T. A. Widiger (Eds.),
Widiger, T. A., & Simonsen, E. (2005). Alternative dimensional models of Personality disorders and the five-factor model of personality (2nd ed.,
personality disorder: Finding a common ground. Journal of Personality pp. 103–124). Washington, DC: American Psychological Association.
Disorders, 19, 110 –130. Zuroff, D. C. (1994). Depressive personality styles and the five-factor
Widiger, T. A., Trull, T. J., Clarkin, J. F., Sanderson, C. J., & Costa, P. T., model of personality. Journal of Personality Assessment, 63, 453– 472.
Jr. (2002). A description of the DSM–IV personality disorders with the Zuroff, D. C., Moskowitz, D. S., & Koestner, R. (1996, August). Differ-
five-factor model of personality. In P. T. Costa & T. A. Widiger (Eds.), entiating the construct of dependency: Rude and Burnheim’s neediness
Personality disorders and the five-factor model of personality (2nd ed., and connectedness factors. Paper presented at the annual meeting of the
pp. 89 –99). Washington, DC: American Psychological Association. American Psychological Association, Toronto, Ontario, Canada.
Wiggins, J. S. (1982). Circumplex models of interpersonal behavior in
clinical psychology. In P. Kendall & J. Butcher (Eds.), Handbook of Received August 21, 2008
research methods in clinical psychology (pp. 183–221). New York: Revision received June 3, 2009
Wiley. Accepted June 11, 2009 䡲

You might also like