Journal of Affective Disorders 92 (2006) 35 44

www.elsevier.com/locate/jad

Research report

Can personality assessment predict future depression?

A twelve-month follow-up of 631 subjects
C. Robert Cloninger , Dragan M. Svrakic, Thomas R. Przybeck
Department of Psychiatry and the Sansone Center for Well-Being, Washington University School of Medicine, St. Louis, MO, United States
Available online 26 January 2006

Abstract
Background: Personality assessment provides a description of a person's fundamental emotional needs and of the higher cognitive
processes that modulate thoughts, feelings, and behavior. Prior studies by us examined personality and mood at the same time.
Assessing personality may allow prediction of mood changes over time in a longitudinal study, as described in earlier prospective
studies by Paula Clayton and others.
Method: A group of 631 adults representative of the general population completed the Temperament and Character Inventory (TCI)
and Center for Epidemiological Studies depression scale (CES-D) at baseline and one year later.
Results: TCI scores at baseline accounted for gender differences in levels of depression. TCI personality scores were strongly stable
(range in r = .78 to .85 for each of seven dimensions) whereas mood was only moderately stable (r = .62) over the twelve-month
follow-up. Baseline personality scores (particularly high Harm Avoidance and low Self-Directedness) explained 44% of the
variance in the change in depression. Baseline levels and changes in Harm Avoidance and Self-Directedness explained 52% of the
variance in the change in depression at follow-up.
Limitations: The follow-up sample was representative of the target population except for slightly lower Novelty Seeking scores.
Clinical relevance: Observable personality levels strongly predict mood changes. Personality development may reduce vulnerability
to future depression.
2005 Elsevier B.V. All rights reserved.
Keywords: Personality assessment; Depression; Mood changes

Under the leadership of Eli Robins and Samuel B. Guze

the Department of Psychiatry at Washington University
demonstrated the utility of systematic diagnosis in patient
assessment and treatment (Goodwin and Guze, 1996). The
work at Washington University on psychiatric diagnosis
led to the adoption of explicit diagnostic criteria in the
official diagnostic and statistical manual of the American

Supported in part by grants from the National Institutes of Health

MH-60879, MH-62130, AA-840314.
Corresponding author. Department of Psychiatry, Washington
University School of Medicine, Campus Box 8134, 660 South Euclid,
St. Louis, MO 63110, United States.
E-mail address: clon@tci.wustl.edu (C.R. Cloninger).

0165-0327/$ - see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.jad.2005.12.034

Psychiatric Association in 1980 and subsequently by the

World Health Organization. Ironically, the way psychiatric
diagnosis is now done in practice following these official
systems violates the scientific principles of diagnostic
assessment that were advocated by its pioneers. In fact,
there has been a steady accumulation of knowledge about
diagnosis that now requires a fundamental paradigm shift
as a result of the careful work on differential diagnosis at
Washington University and elsewhere over the past few
decades (Cloninger, 1999a, 2000a,b, 2004).
Here we will examine the observable personality variables that strongly modulate mood changes as an illustration of the way that personality assessment is a

36

C.R. Cloninger et al. / Journal of Affective Disorders 92 (2006) 3544

necessary foundation for any rigorous differential diagnosis

of psychopathology. Paula Clayton and Robert Cloninger
carried out a systematic follow-up and family study of 500
psychiatric outpatients in collaboration with their late colleagues, Samuel Guze, Robert Woodruff, and Ronald
Martin (Clayton, 1974; Cloninger et al., 1985; Martin et al.,
1985a,b; Guze et al., 1986). Paula Clayton was engaged in
the assessment of the original patient series, and Cloninger
carried out most of the follow-up interviews six to twelve
years later. Clayton was also doing psychiatric assessments
and follow-up studies of bereavement at the same time.
Together these studies of bereaved subjects and psychiatric
outpatients revealed much about the causes and course of
depression that helped Cloninger develop a more adequate
understanding of the way that personality modulates mood
changes.
Eli Robins and Sam Guze taught that each patient had
only one fundamental diagnosis. The presence of two or
more syndromes suggested to them that the patient should
be considered undiagnosed except in particular situations
where the chronology was clear and predictable, such as a
patient with antisocial personality disorder or panic
disorder developing secondary alcoholism or secondary
depression (Feighner et al., 1972). Patients with primary
depressive disorders had no other recognized psychopathology prior to the onset of the major depressive
disorder. As a consequence, patients with primary depressive disorders must be psychologically healthy before
the onset of their first depressive episode. The Washington
University investigators also observed that the firstdegree relatives of patients with primary depression or
panic disorder with secondary depression were less likely
to have antisocial personality disorder, whereas the relatives of patients with depression secondary to somatization disorder had an increased risk of antisocial personality
disorder and substance dependence (Guze et al., 1986).
Essentially, differential diagnosis as done at Washington
University involved assessment of multiple syndromes
simultaneously to define groups of patients who were as
clinically homogeneous as possible. This required careful
attention to both inclusion and exclusion criteria for each
diagnosis. Just having sufficient criteria to diagnose a major
depression was not an adequate basis for differential
diagnosis; evidence for other psychopathology prior to the
onset of depression and in the family history had to be
considered as possible exclusion criteria for rigorous
differential diagnosis. The careful use of both inclusion
and exclusion criteria at Washington University under Eli
Robins, Sam Guze, and George Winokur is in marked
contrast to the current practice of making multiple comorbid diagnoses based largely on heterogeneous inclusion
criteria that often ignore the distinction between the primary

psychiatric illness and other secondary phenomena

(Winokur and Clayton, 1994). The current DSM systems
do require some exclusions but these are minimal in a
system with more than 300 diagnostic categories, which are
often redundant and are too numerous in practice to assess
in every patient. As a result, current diagnostic practice is
usually unreliable because the primary diagnosis is largely
determined by current presenting complaints, diagnostic
biases, and subjective impressions. These impressions can
be easily justified by listing loose inclusion criteria, which
are actually heterogeneous in their psychological and
biological basis (Cloninger, 2002b). In contrast, the original
Washington University diagnostic system was based on a
dozen or so categories that were each assessed in detail in
every patient in order to identify the primary foundation
from which other symptoms developed.
Experience with the Washington University approach
led clinicians to recognize the semi-quantitative features
of a clinical spectrum associated with each fundamental
problem. Essentially the Washington University approach
to psychiatric diagnosis required a multidimensional
assessment all patients should be systematically
assessed in each of the descriptive dimensions underlying
personality and psychopathology. People using this
system usually continued to assume that diagnoses
referred to categories of discrete diseases, but this assumption was not necessary in an empirical approach:
there has never been objective evidence for such discreteness (Kendell, 1982; Cloninger, 2002b). Nevertheless, clinicians using categorical diagnoses for descriptive
purposes frequently recognized partial expressions of
disorders that obviously varied in severity and degree of
functional impairment (Goodwin and Guze, 1996).
Longitudinal research on antisocial personality became
the exemplar at Washington University for distinguishing
the primary diagnosis upon which other syndromes developed as secondary complications (Robins, 1966;
Robins and Price, 1991; Robins et al., 1995). For example, a patient might not satisfy the full inclusion criteria for
any other major diagnosis prior to a major depression, but
they might be so aloof, asocial, or asexual that they would
be noted to have cluster A traits even if they did not
have definite schizophrenia. Alternatively, they might be
so antisocial or impulsive that they would be noted to
have cluster B traits even if they did not have antisocial
personality disorder or somatization disorder (Cloninger,
1986). They might be so anxiety-prone, fearful, shy, or
fatigable that they would be noted to have cluster C
traits even if they did not have sufficient impairment to
diagnose panic disorder or generalized anxiety disorder.
Later Cloninger systematized such observations in his
description of the structure of personality in the

C.R. Cloninger et al. / Journal of Affective Disorders 92 (2006) 3544

Temperament and Character Inventory (TCI) (Cloninger,

1986; Cloninger et al., 1993, 1994). TCI Harm Avoidance
quantifies individual differences in the extent to which a
person is anxious, pessimistic, and shy (i.e., cluster C)
versus risk-taking, optimistic, and outgoing. TCI Novelty
Seeking quantifies individual differences in the extent to
which a person is impulsive, quick-tempered, and
disorderly (i.e., cluster B) versus rigid, stoical, and
orderly. TCI Reward Dependence quantifies individual
differences in the extent to which a person is sociable,
approval seeking, and warm versus aloof, detached, and
cold (i.e., cluster A). TCI Persistence quantifies
individual differences in the extent to which a person is
overachieving versus underachieving, as is highly developed in obsessive or anankastic personalities. Hence
antisocial personality disorder provided an example of the
importance of distinguishing antecedent conditions from
their complications, and quantitative measures of personality in the TCI provided a comprehensive basis for
applying a more general biopsychosocial perspective
(Cloninger, 1986).
The TCI character dimensions also provided a way to
quantify aspects of mature mental self-government,
thereby providing a reliable way to measure the higher
cognitive processes that modulate emotional conflicts
(Cloninger et al., 1993). Anne Farmer and her colleagues have examined temperament and character as
vulnerability factors in susceptibility to major depression using family data. Studies of the sibs of psychiatric
patients with severe major depression and the sibs of
controls who have never been depressed have shown
that specific personality traits quantify the degree of
vulnerability to future depression even in individuals
who have never been depressed themselves (Farmer et
al., 2003). The familial vulnerability to major depression
is predicted most strongly by high TCI Harm Avoidance
and low TCI Self-Directedness. Novelty Seeking and
Reward Dependence also decreased risk whereas Persistence, Cooperativeness, and Self-Transcendence did
not influence the familial risk (Farmer et al., 2003). High
TCI Harm Avoidance and low TCI Self-Directedness
are also predictive of poor response to antidepressants
(Joffe et al., 1993; Joyce et al., 1994, 2003;Tome et al.,
1997; Cloninger and Svrakic, 2000).
Paula Clayton and others have compared the premorbid personalities of individuals who later develop
depressive disorders in prospective studies of individuals
sampled from the general population. In these prospective studies, cases of depressive disorder were identified
through psychiatric treatment records and compared to
untreated controls. In a study of 3000 individuals in
Sweden, 83% completed the MarkNyman Tempera-

37

ment (MNT) Inventory (Nystrom and Lindegard, 1975a,

b). The 37 depressives were low in MNT Validity scores
than controls, which corresponds to their being higher in
TCI Harm Avoidance; they did not differ in MNT Solidity
or Stability, which correspond to TCI Novelty Seeking
and Reward Dependence, respectively. In a study of 6315
Swiss men conscripted into military service, personality
was assessed at age 19 using the Freiburg Personality
Inventory (FPI) and then medical records were obtained
for the subsequent 12 years (Angst and Clayton, 1986).
Paula Clayton diagnosed unipolar depression in 19 men
based on their medical records during the follow-up
without knowledge of their earlier performance on the
FPI. The depressive cases were higher than controls in the
FPI scale Autonomic Lability, which corresponds to high
TCI Harm Avoidance and low TCI Self-Directedness, as
confirmed in subsequent studies of the personality of
depressives using the TCI itself. In addition, the depressive Swiss men were also higher in the FPI Aggression
scale, which corresponds to high TCI Novelty Seeking
and low TCI Cooperativeness. In premorbid personality,
the Swiss men with depression were similar to men who
later completed suicide or had psychopathic (antisocial or
borderline) personality disorders in that they were more
quick-tempered and revengeful than controls. This prospective finding about the role of aggressiveness in depressives and suicides has not been observed consistently in
all depressives, but may identify a subgroup of depressives with borderline personality disorder (Joyce et al.,
2003). In addition, Akiskal and colleagues found that
depressives with prominent mood lability, activity, and
daydreaming were more likely to convert to Bipolar II
disorder than other depressives (Akiskal et al., 1995).
Such Bipolar II patients are often high in TCI Cooperativeness, which is part of the TCI cyclothymic character
configuration (Cloninger et al., 1998). In summary,
prospective studies suggest that high TCI Harm Avoidance and low TCI Self-Directedness predict vulnerability
to depressive disorders, whereas TCI Novelty Seeking
and Cooperativeness may have variable roles in predicting risk of particular depressive subtypes.
Paula Clayton also showed that bereaved men had
greater morbidity and mortality than did bereaved
women (Clayton, 1974). Subsequent work has indicated
that women are usually more effective than men in
providing emotional support to others, so men often
derive social support from their spouse whereas women
derive most of their social support from their female
friends (Glynn et al., 1999). Work with the TCI showed
that women are much higher than men in TCI Reward
Dependence (i.e., sociable and warm) and in TCI
Cooperativeness (i.e., empathic and helpful) (Cloninger

38

C.R. Cloninger et al. / Journal of Affective Disorders 92 (2006) 3544

et al., 1994). In addition, women are also higher than men

on average in TCI Harm Avoidance, thereby possibly
predisposing them to an excess of cluster C psychopathology, including depressive and anxiety disorders
(Cloninger et al., 1998). More importantly, studies of
bereavement and personality indicate that mood must be
understood in terms of a person's way of processing their
experiences in a psychosocial context (Cloninger, 2004).
The availability of a reliable quantitative way of
assessing personality that was appropriate for both psychiatric patients and the general population has clarified
the question about whether mental disorders are discrete
diseases. Personality contributes substantially to vulnerability to psychopathology, and the structure of
personality is the same in psychiatric patients as it is in
the general population (Cloninger, 1999b). There are
differences in the mean values of some personality dimensions between psychiatric cases and controls
(Svrakic et al., 1993; Bayon et al., 1996), but the same
dimensions describe human behavior in both psychiatric
patients and in the general population (Cloninger,
1999b, 2004). This suggests that the dimensions of
personality measured by the TCI provide the foundation
for identification of a general set of quantitative variables that account for individual differences in personality and common forms of psychopathology (Krueger,
1999; Cloninger, 2004). If so, psychopathology can be
efficiently assessed by systematic assessment of a
relatively small number of variables that can be measured in every patient, much as was done originally in
the Washington University approach to psychiatric
diagnosis. The major difference is that assessment can
now be focused on quantitative measurements rather
than qualitative categories, which have fuzzy or arbitrary boundaries, no specific laboratory tests, and no
specific psychological or biological treatments. As in
the example of antisocial personality, it is crucial to
recognize the primary foundation upon which other
secondary phenomena arise as clinical complications.
In this article we examine the role of temperament
and character in the modulation of changes in depressed
mood in a sample from the general population. In particular, hypotheses are tested about the extent to which
personality can account for gender differences in
vulnerability to depressive disorders and the extent to
which personality can predict changes in depressed
mood over time in a longitudinal study. We examined
only the relationship between personality and mood at
the same time in earlier reports from this study
(Cloninger et al., 1998; Cloninger, 2004). Work carried
out by Paula Clayton in St. Louis and later in Minnesota
has contributed to a fuller understanding of hypotheses

about the biopsychosocial mechanisms that regulate

mood. On behalf of her many friends and colleagues at
Washington University, it is with deep personal satisfaction and professional respect for her that we contribute this paper as a part of her Festschrift.
1. Method
1.1. Sample
We attempted to study a stratified random sample of
1000 noninstitutionalized adults, 18 years of age or
older, who lived in the greater metropolitan area of St.
Louis, Missouri in June 1994 and to follow them up
twelve months later. The Washington University Human
Subjects Committee approved the study methods.
Potential participants initially were identified at random
from standard telephone lists and asked if they were
willing to participate in a questionnaire survey of depression, personality, and health sponsored by the
National Institutes of Health and Washington University
School of Medicine. Of 1740 individuals solicited for
possible participation, 243 declined (14% refusal). From
the 1497 volunteers, a final panel of 1000 was accepted
into demographic strata in numbers representative of the
1990 federal census according to age (1834, 3554,
55+), gender (male, female), ethnicity (white, black,
Hispanic, other), and geographical location of household (six counties). The 1000 selected participants were
mailed a questionnaire along with prepayment of $5 and
paid another $20 on return of the completed questionnaire. Those who did not return the questionnaire after a
few weeks were reminded with postcards and later phone
calls. Altogether 866 subjects returned the questionnaire.
Of these 804 were nearly complete and were valid based
on internal consistency checks built into the questionnaire to detect careless or inconsistent reporting.
The 804 complete and valid respondents included
slightly more women than expected (57% observed
versus 52% expected), but were otherwise representative
of the general population demographically. The respondents were also representative of the general population
in terms of their personality, as assessed in a national
area probability sample conducted as part of the 1987
General Social Survey of the National Science Foundation (Cloninger et al., 1991), and indicators of psychopathology obtained in other community samples using
interviews and questionnaires (Myers and Weissman,
1980; Robins and Regier, 1991; Cloninger et al., 1997).
Twelve months later, 631 of the 800 subjects completed the questionnaire and had valid data on the
measures of personality and depression used at time 1.

C.R. Cloninger et al. / Journal of Affective Disorders 92 (2006) 3544

Those who returned the follow-up did not differ from

the others in terms of demographic or psychometric
variables except that they were older (mean ages 46
versus 40 years, t = 4.97, df = 802, P b.0001) and slightly
lower in Novelty Seeking (17.5 versus 18.8, t = 2.69,
df = 802, P b.01). The testretest reliability of the
personality ratings was high (i.e., correlations of .78 to
.85 for each of the TCI dimensions), which supports the
reliability and validity of the self-reports.
1.2. Assessment questionnaires
Subjects completed a questionnaire booklet with
separate sections for general demographic information,
the TCI (Cloninger et al., 1994), the NIMH Center for
Epidemiological Studies depression scale (CES-D)
(Radloff, 1977), and some other measures of psychopathology and health described elsewhere (Cloninger et al.,
1997, 1998). The TCI, version 9, is a set of 226 questions
answered true or false, measuring four dimensions of
temperament and three dimensions of character. Each of
the seven TCI dimensions are about 50% heritable according to large scale twin studies (Gillespie et al., 2003).
The temperament dimensions measure individual differences in emotional responses to associatively conditioned stimuli. The four temperaments are Harm
Avoidance (i.e., anxious versus risk-taking), Novelty
Seeking (i.e., impulsive versus rigid), Reward Dependence (i.e., approval seeking versus aloof), and Persistence (i.e., overachieving versus underachieving). The
character dimensions measure individual differences in
higher cognitive processes that modulate emotional
conflicts to satisfy a person's goals and values. The
character dimensions quantify the three branches of
mental self-government: Self-Directedness (executive
functions, such as being responsible, purposeful, and
resourceful), Cooperativeness (legislative functions,
such as being tolerant, forgiving, and helpful), and
Self-Transcendence (judicial functions, such as being
intuitive, judicious, and aware) (Cloninger, 2004). Individual differences in TCI character dimensions measure the presence and severity of personality disorder,
whereas the temperament dimensions indicate membership in the anxious cluster if high in Harm Avoidance, the
impulsive cluster if high in Novelty Seeking, and the
aloof cluster if low in Reward Dependence (Svrakic et al.,
1993; Cloninger, 2000b). The CES-D asks about 20
depressive symptoms, which are rated on a four-point
frequency scale for the past week. In a community sample, a CES-D score of 21 or higher had a sensitivity (i.e.,
proportion of true cases detected) of 54% and a specificity
(i.e., proportion of noncases correctly classified) of 96%

39

for major depression diagnosed by structured interview;

scores of 16 or higher had a sensitivity of 64% and a
specificity of 94% (Myers and Weissman, 1980). Scores
of less than 16, 1620, and 21 or more define cut-offs
suggesting a person is not depressed, has a minor depression, or a major depression, respectively.
1.3. Statistical analysis
All analyses were carried out with SAS, version 6.12
statistical software (Institute, 2002). Our analysis
focused on the seven personality and CES-D scores at
index and follow-up, in the 631 subjects who completed
both questionnaires. We also considered the effects of
age, gender and history of psychiatric treatment. We
examined gender differences in personality and CES-D
scores at time 1 and then used the personality and
demographic variables in multiple regression to predict
CES-D score. To assess stability of mood we first examined the distribution of CES-D score groups (b 16,
1620, 21+) at time 1 versus time 2, and then computed
a series of multiple and step-wise regressions to find the
set of variables which best explained CES-D at time 2.
With CES-D at time 1 as a covariate, we sequentially
considered models that first included time 1 CES-D, the
personality variables, age and sex; next adding change
in the personality dimensions; and finally adding
evidence of psychiatric treatment. The psychiatric
treatment variables included ever taking medicine for
a psychiatric condition, ever being treated by a psychiatrist, and ever having a psychiatric hospitalization.
2. Results
The levels of depressive symptoms, as measured by
CES-D scores, are summarized for men and for women
in Table 1. Women more often had CES-D scores of 21
or more than did men (18% versus 13%), as expected
Table 1
Levels of depressive symptoms as measured by the Center for
Epidemiological Studies-Depression (CES-D) scale in men and
women in a sample of the general population of St. Louis (1994 St.
Louis Health Survey)
Gender

Row % with CES-D score at time 1

(No
depression)
b16

Men
267 76
Women
364 72
Total
631 (n = 467)
number

(Minor
depression)

Total

(Major
depression)

1620

21+

11
10
(n = 66)

13
18
(n = 98)

100
100

40

C.R. Cloninger et al. / Journal of Affective Disorders 92 (2006) 3544

Table 2
Gender differences in TCI personality dimensions
TCI dimension

Temperament
Harm Avoidance
Novelty Seeking
Reward Dependence
Persistence
Character
Self-Directedness
Cooperativeness
Self-Transcendence

Mean (SD)

Men

Women

11.7(7.1)
17.7(6.0)
14.7(4.2)
5.4(2.0)

15.5(6.9)
17.3(5.8)
17.3(3.8)
5.5(1.9)

6.7
0.9
8.0
0.2

.0001
NS
.0001
NS

32.9(7.4)
33.9(6.4)
10.7(8.5)

33.1(6.9)
36.4(4.4)
12.1(10.2)

0.5
5.5
1.9

NS
.0001
.0642

from previous community surveys. There were also

strong differences in personality between men and
women: women were higher than men in Harm Avoidance, Reward Dependence, and Cooperativeness (Table 2).
The observed gender differences in personality
accounted for the observed differences in CES-D scores.
CES-D scores were regressed on demographic variables
and TCI variables. Only the TCI variables of character and
Harm Avoidance contributed to the prediction of CES-D
scores. Age and gender did not add significantly to
depression scores when personality was taken into account
(Table 3).
TCI personality scores were strongly correlated over
the twelve-month follow-up period (.78 to .85 for each
of the seven TCI dimensions). CES-D scores were
moderately correlated over the twelve-month follow-up
(r = .62, P = .0001). The variation in CES-D scores is
summarized in Table 4 for the cut-offs used to indicate
no depression (less than 16), minor depression (1620),
and major depression (more than 20). As in studies of
the treatment of depression with placebos (Walsh et al.,
2002), more than 30% of individuals with CES-D scores
greater than 20 at time 1 had scores of less than 16 one
Table 3
Regression of CES depression scores on TCI personality scores at the
same time (standardized regression coefficients estimated in general
linear model for time 1 for 631 subjects with follow-up)
TCI dimension
Temperament
Harm Avoidance
Novelty Seeking
Reward Dependence
Persistence
Character
Self-Directedness
Cooperativeness
Self-Transcendence

Partial regression

Significance

+0.41

.0001

0.42
0.26
+0.22

.0001
.0001
.0001

Total multiple R squared = .38; gender and age had no significant

effects.

Table 4
Mood swings measured by changes in CES-D scores by follow-up
testing one year later
CES-D
score at
time 1

Row % with CES-D

score at time 2

Total

b16

1620

21+

b16
16-20
21+
Total

467
66
98
631

90
63
33

6
15
12

4
22
55

100
100
100

year later, but 55% still had scores greater than 20. 4% of
individuals with initial scores lower than 16 had scores
greater than 20 at follow-up. Those with intermediate
scores rarely stayed at the same level; they usually
improved (i.e., 62% had scores lower than 16) but 22%
worsened (i.e., had scores greater than 20).
Initial CES-D scores explained 39% of the variance
in CES-D scores at follow-up so we considered the
predictive value of personality controlling for initial
CES-D scores (Table 5). CES-D scores at follow-up
were regressed on the 7 TCI personality scores along
with initial CES-D scores, age, and gender. The initial
TCI measures increased the explained variance to 44%,
providing a moderate prediction of depression at followup. Depression at follow-up was predicted by high
Harm Avoidance (i.e., fatigable, fearful, anxious, shy),
by low Self-Directedness (i.e., irresponsible, aimless,
inept), and by high Persistence (i.e., overachieving),
even taking initial CES-D scores into account. Age and
gender did not contribute to the prediction of depression
at follow-up when personality was taken into account.
The addition of change scores in TCI Self-Directedness and TCI Harm Avoidance explained another 8% of
the variance (Table 5), providing a strong prediction of
depression at follow-up (multiple r = .72, P = .0001). The
Table 5
Regression of CES depression scores at follow-up (time 2) on initial
TCI personality and CES depression scores (time 1) (standardized
regression coefficients and partial r-squares estimated stepwise for 631
subjects with follow-up)
Predictor
variable

Partial
regression

Cumulative
model
R-squares

Significance

CES-D at time 1
TCI Harm
Avoidance
TCI SelfDirectedness
TCI Persistence

+0.46
+0.21

.39
.42

.0001
.0001

0.22

.43

.0001

+0.42

.44

.0060

Age, gender, other personality variables, and treatment variables had

no significant effects (P N.05).

C.R. Cloninger et al. / Journal of Affective Disorders 92 (2006) 3544

Table 6
Regression of CES depression scores at follow-up (time 2) on initial
TCI personality and CES depression scores (time 1) plus measures of
change in TCI scores (partial r-squares estimated stepwise for 631
subjects at time 2)
Predictor variable

Partial
regression

Cumulative
model
R-squares

Significance

CES-D at time 1
Change in
Self-Directedness
TCI Self-Directedness
TCI Harm Avoidance
Change in Harm
Avoidance
TCI Persistence

+0.48
0.45

.39
.43

.0001
.0001

0.31
+0.24
+0.34

.49
.50
.52

.0001
.0006
.0001

+0.34

.52

.0193

Age, gender, other personality and treatment variables had no

significant effects (P N.05).

order of entry of variables and the cumulative model

explained variance (i.e., r-squares) are shown in Table 6.
Age, gender, and other personality and treatment
variables (i.e., outpatient or inpatient treatment, use of
psychotropic medications) had no significant effects in
the prediction of CES-D scores at follow-up once
personality and initial depression scores were taken into
account. The model predicting depression scores at
follow-up is summarized in Table 7. Baseline TCI SelfDirectedness and change in TCI Self-Directedness
provided the most robust prediction of CES-D scores
regardless of age, gender, treatment variables, or
baseline CES-D scores. Self-Directedness remained a
strong predictor of CES-D scores at follow-up regardless of initial CES-D score; even among those with
initial CES-D scores above 16 or above 20, TCI SelfDirectedness remained a strong predictor of depression
at follow-up. Baseline TCI Harm Avoidance and change
in TCI Harm Avoidance also contributed strongly
regardless of age, gender, or treatment variables. However, Harm Avoidance primarily distinguished individuals whose CES-D scores were above or below 16 and
did not predict severity at follow-up among those with
initial scores above 16. TCI Persistence contributed only
weakly to the prediction of depression at follow-up.
3. Discussion
Our findings indicate that TCI Harm Avoidance is a
marker of emotional vulnerability to depression. In
contrast, TCI Self-Directedness is a marker of executive
functions that protect a person from depression. As a
result, individuals with depression are likely to be both
anxiety-prone (i.e., high in Harm Avoidance) and immature (i.e., low in Self-Directedness and Cooperative-

41

ness) (see Table 7). In prior work, we showed that

depression and these same personality variables were
correlated at the same time (Cloninger et al., 1998), and
here extend that to show that the TCI personality
variables measure vulnerability that predict future depression. In other words, individuals with cluster C
personality disorders are most likely to be depressed.
Such individuals are emotionally anxiety-prone and
have limited ability to recover from depression. These
observations confirm and extend findings from prospective and family studies showing that TCI Harm
Avoidance and Self-Directedness are substantial predictors of vulnerability to major depressive disorders
(Nystrom and Lindegard, 1975a; Angst and Clayton,
1986; Farmer et al., 2003). We also confirm that low
Cooperativeness is associated with the current state of
depression but does not predict later depression, as
observed in family studies of vulnerability to depression
(Farmer et al., 2003). We were unable to confirm that
high aggressiveness, measured as high TCI Novelty
Seeking and/or low TCI Cooperativeness, are predictive
of depression (Angst and Clayton, 1986), suggesting
that this finding from prospective studies is relevant to a
particular depressive subgroup with features of borderline personality disorder (Joyce et al., 2003).
However, vulnerability to develop depression is not the
same as inability to recover from depression. SelfDirectedness measures a person's degree of cognitive
coherence and reality testing, which facilitate a person to
be responsible, purposeful, resourceful, self-accepting,
and hopeful. We found that mood changes were strongly
predicted by baseline levels and change in TCI SelfTable 7
Regression of CES depression scores at follow-up on initial depression
and personality scores plus change in personality scores (standardized
regression coefficients estimated in general linear model for 631
subjects at time 2)
TCI dimension

Partial regression

Significance

CES-D at time 1
Temperament at time 1
Harm Avoidance
Novelty Seeking
Reward Dependence
Persistence
Character at time 1
Self-Directedness
Cooperativeness
Self-Transcendence
Personality Change
Harm Avoidance
Self-Directedness

+0.48

.0001

+0.24

+0.34

.0001

.0193

0.31

.0001

+0.34
0.45

.0001
.0001

Total multiple R squared = .52; gender, age, and other personality and
treatment variables had no significant effects.

42

C.R. Cloninger et al. / Journal of Affective Disorders 92 (2006) 3544

Directedness. To our knowledge, this is the first time a

strong and reliable basis for predicting the longitudinal
course of depression has been described. One of the most
remarkable deficiencies in psychiatric practice has been
the inability of psychiatrists to predict relapse and
recovery in individuals with a depressive episode. Individuals with depression may respond acutely to
treatment, but nearly 80% relapse over time because
their vulnerability has not been reduced by past treatment.
Who will relapse and who will maintain recovery? Our
findings, combined with other results, indicate that
increased Self-Directedness indicates an increase in
coherence of personality that protects a person from
depression. Fortunately, both antidepressants and cognitive therapies have been shown to enhance SelfDirectedness providing a means to reduce long-term
vulnerability to depression and related psychopathology
(Bulik et al., 1998; Cloninger, 2004).
Functional brain imaging has shown that individual
differences in TCI Self-Directedness, Harm Avoidance,
and Persistence are each modulated in specific brain
networks. TCI Self-Directedness is strongly correlated
(r = .74) with activation of the medial prefrontal cortex
(Brodmann areas 9/10) during executive function tasks
in healthy volunteers (Cloninger, 2002a; Gusnard et al.,
2003). This brain region plays a key role in the brain
networks allowing self-aware thinking. TCI Harm
Avoidance is moderately correlated (r = .49) with the
volume of the right anterior cingulate cortex (Brodmann
area 24), which is involved in modulating the salience of
emotional stimuli (Pujol et al., 2002). Persistence is
strongly correlated (r = .8) with individual differences
reward-seeking behavior modulated by a well-known
network involving the ventral striatum, orbitofrontal
cortex (Brodmann area 47), and anterior cingulate cortex (Brodmann area 24) (Gusnard et al., 2003).
Depression was correlated here with high TCI Persistence (i.e., overachieving) whereas euthymic bipolar
patients are low in Persistence (i.e., easily discouraged,
underachieving) (Osher et al., 1996, 1999). Feelings of
well-being depend on coherent orchestration of brain
networks that regulate discrete mind-brain states in
human thought, as is described in depth elsewhere
(Cloninger, 2004). The heritable personality traits of
Harm Avoidance and Self-Directedness that modulate
mood have a heritability of about 50% and also account
for about 50% of the variance in depression at follow-up.
Nevertheless, our results confirm other work showing
that the mood changes in human beings are unpredictable
to a substantial degree. Nearly 50% of the variance in
mood is unexplained by any variable we could measure.
Likewise, antidepressants and empirical defined psy-

chotherapies for depression have only moderate effect

sizes (Lipsey and Wilson, 1993; Westen and Morrison,
2001; Walsh et al., 2002). Electroconvulsive therapy has a
strong effect on depressed mood acutely, but no influence
on the subsequent course of depression (Lipsey and
Wilson, 1993). Consequently, the human capacity for
creative self-organization appears to have an unpredictable influence on mood and personality development.
This capacity can be facilitated systematically by mental
exercises described in detail elsewhere (Cloninger, 2004).
Our findings suggest the hypothesis that personality
development can strongly reduce future vulnerability to
depressive disorders. This hypothesis needs to be tested in
longitudinal treatment trials to determine whether personality development in randomized controlled trials of
medication and/or psychotherapy can predictably reduce
the subsequent risk of depression.
Acknowledgment
We thank Dr. Richard Hudgens for his valuable comments on an earlier draft.

References
Akiskal, H.S., Maser, J.D., Zeller, P.J., Endicott, J., Coryell, W., Keller,
M., Warshaw, M., Clayton, P.J., Goodwin, F., 1995. Switching
from unipolar to bipolar II. An 11-year prospective study of
clinical and temperamental predictors in 559 patients. Archives of
General Psychiatry 52, 114123.
Angst, J., Clayton, P.J., 1986. Premorbid personality of depressive,
bipolar, and schizophrenic patients with special reference to
suicidal issues. Comprehensive Psychiatry 27, 511532.
Bayon, C., Hill, K., Svrakic, D.M., Przybeck, T.R., Cloninger, C.R.,
1996. Dimensional assessment of personality in an outpatient
sample: relations of the systems of Millon and Cloninger. Journal
of Psychiatric Research 30, 341352.
Bulik, C.M., Sullivan, P.F., Joyce, P.R., Carter, F.A., McIntosh, V.V.,
1998. Predictors of 1-year treatment outcome in Bulimea Nervosa.
Comprehensive Psychiatry 39, 206214.
Clayton, P.J., 1974. Mortality and morbidity in the first year of
widowhood. Archives of General Psychiatry 30, 740750.
Cloninger, C.R., 1986. A unified biosocial theory of personality and its
role in the development of anxiety states. Psychiatric Developments 3, 167226.
Cloninger, C.R., 1999a. A new conceptual paradigm from genetics and
psychobiology for the science of mental health. Australian and
New Zealand Journal of Psychiatry 33, 174186.
Cloninger, C.R. (Ed.), 1999b. Personality and Psychopathology.
American Psychiatric Press, Washington, D.C.
Cloninger, C.R., 2000a. Biology of personality dimensions. Current
Opinion in Psychiatry 13, 611616.
Cloninger, C.R., 2000b. A practical way to diagnose personality
disorder: a proposal. Journal of Personality Disorders 14, 99108.
Cloninger, C.R., 2002a. Functional neuroanatomy and brain imaging
of personality and its disorders. In: D'haenen, H., den Boer, J.A.,

C.R. Cloninger et al. / Journal of Affective Disorders 92 (2006) 3544

Willner, P. (Eds.), Biological Psychiatry, vol. 2. John Wiley &
Sons, Ltd., Chichester, England, pp. 13771385.
Cloninger, C.R., 2002b. Implications of comorbidity for the
classification of mental disorders: the need for a psychobiology
of coherence. In: Maj, M., Gaebel, W., Lopez-Ibor, J.J., Sartorius,
N. (Eds.), Psychiatric Diagnosis and Classification. John Wiley &
Sons, Ltd., Chichester, England, pp. 79106.
Cloninger, C.R., 2004. Feeling Good: the Science of Well Being.
Oxford University Press, New York.
Cloninger, C.R., Svrakic, D.M., 2000. Personality disorders. In:
Sadock, B.J., Sadock, V.A. (Eds.), Comprehensive Textbook of
Psychiatry. Lippincott Williams & Wilkins, New York,
pp. 17231764.
Cloninger, C.R., Martin, R.L., Guze, S.B., Clayton, P.J., 1985.
Diagnosis and prognosis in schizophrenia. Archives of General
Psychiatry 42, 1225.
Cloninger, C.R., Przybeck, T.R., Svrakic, D.M., 1991. The tridimensional personality questionnaire: U.S. normative data. Psychological Reports 69, 10471057.
Cloninger, C.R., Svrakic, D.M., Przybeck, T.R., 1993. A psychobiological model of temperament and character. Archives of General
Psychiatry 50, 975990.
Cloninger, C.R., Przybeck, T.R., Svrakic, D.M., Wetzel, R.D., 1994.
The Temperament and Character Inventory: a Guide to its
Development and Use. Washington University Center for
Psychobiology of Personality, St. Louis, MO.
Cloninger, C.R., Bayon, C., Przybeck, T.R., 1997. Epidemiology
and Axis 1 comorbidity of antisocial personality disorder.
In: Stoff, D.M., Breiling, J., Maser, J.D. (Eds.), Handbook
of Antisocial Behavior. John Wiley & Sons, New York,
pp. 1221.
Cloninger, C.R., Bayon, C., Svrakic, D.M., 1998. Measurement of
temperament and character in mood disorders: a model of
fundamental states as personality types. Journal of Affective
Disorders 51, 2132.
Farmer, A., Mahmood, A., Redman, K., Harris, T., Sadler, S.,
McGuffin, P., 2003. A sib-pair study of the Temperament and
Character Inventory in major depression. Archives of General
Psychiatry 60, 490496.
Feighner, J., Robins, E., Guze, S.B., Woodruff, R., Winokur, G.,
Munoz, R., 1972. Diagnostic criteria for use in psychiatric
research. Archives of General Psychiatry 26, 5763.
Gillespie, N.A., Cloninger, C.R., Heath, A.C., Martin, N.G., 2003.
The genetic and environmental relationship between Cloninger's dimensions of temperament and character. Personality
and Individual Differences 35, 19311946.
Glynn, L.M., Christenfeld, N., Gerin, W., 1999. Gender, social
support, and cardiovascular responses to stress. Psychosomatic
Medicine 61, 234242.
Goodwin, D.W., Guze, S.B., 1996. Psychiatric Diagnosis. Oxford
University Press, New York.
Gusnard, D.A., Ollinger, J.M., Shulman, G.L., Cloninger, C.R., Price,
J.L., Van Essen, D.C., Raichle, M.E., 2003. Persistence and brain
circuitry. Proceedings of the National Academy of Sciences of the
United States of America 100, 34793484.
Guze, S.B., Cloninger, C.R., Martin, R.L., Clayton, P.J., 1986. A
follow-up and family study of Briquet's syndrome. British Journal
of Psychiatry 149, 1723.
Institute, S., 2002. SAS System for Statistical Analysis. SAS Institute,
Cary NC.
Joffe, R.T., Bagby, R.M., Levitt, A.J., Regan, J.J., Parker, J.
D., 1993. The tridimensional personality questionnaire in

43

major depression. American Journal of Psychiatry 150,

959960.
Joyce, P.R., Mulder, R.T., Cloninger, C.R., 1994. Temperament
predicts clomipramine and desipramine response in major
depression. Journal of Affective Disorders 30, 3546.
Joyce, P.R., Mulder, R.T., Luty, S.E., McKenzie, J.M., Sullivan, P.F.,
Cloninger, C.R., 2003. Borderline personality disorder in major
depression: differential drug response, and 6-month outcome.
Comprehensive Psychiatry 44, 3543.
Kendell, R.E., 1982. The choice of diagnostic criteria for biological
research. Archives of General Psychiatry 39, 13341339.
Krueger, R.F., 1999. The structure of common mental disorders.
Archives of General Psychiatry 56, 921926.
Lipsey, M.W., Wilson, D.B., 1993. The efficacy of psychological,
educational, and behavioral treatment: confirmation from metaanalysis. American Psychologist 48, 11811209.
Martin, R.L., Cloninger, C.R., Guze, S.B., Clayton, P.J., 1985a.
Frequency and differential diagnosis of depressive syndromes in
schizophrenia. Journal of Clinical Psychiatry 46, 913.
Martin, R.L., Cloninger, C.R., Guze, S.B., Clayton, P.J., 1985b.
Mortality in a follow-up of 500 psychiatric out-patients. II. Causespecific mortality. Archives of General Psychiatry 42, 5866.
Myers, J.L., Weissman, M.M., 1980. Use of a self-report symptom
scale to detect depression in a community sample. American
Journal of Psychiatry 137, 10811084.
Nystrom, D., Lindegard, B., 1975a. Depression: predisposing factors.
Acta Psychiatrica Scandinavica 51, 7787.
Nystrom, S., Lindegard, B., 1975b. Predisposition to mental syndromes:
a study comparing predisposition to depression, neurasthenia and
anxiety state. Acta Psychiatrica Scandinavica 51, 6976.
Osher, Y., Cloninger, C.R., Belmaker, R.H., 1996. TPQ in euthymic
manic-depressive patients. Journal of Psychiatric Research 30,
353357.
Osher, Y., Lefkifer, E., Kotler, M., 1999. Low persistence in euthymic
manic-depressive patients: a replication. Journal of Affective Disorders
53, 8790.
Pujol, J., Lopez, A., Deus, J., Cardoner, N., Vallejo, J., Capdevila, A.,
Paus, T., 2002. Anatomical variability of the anterior cingulate
gyrus and basic dimensions of human personality. NeuroImage 15,
847855.
Radloff, L.S., 1977. The CES-D scale: a self-report depression scale
for research in the general population. Applied Psychological
Measurement 36, 749760.
Robins, L.N., 1966. Deviant Children Grown Up: a Sociological and
Psychiatric Study of Sociopathic Personality. Williams and
Wilkins, Baltimore.
Robins, L.N., Price, R.K., 1991. Adult disorders predicted by
childhood conduct problems: results from the NIMH Epidemiological Catchment Area Project. Psychiatry 54, 116132.
Robins, L.N., Regier, D.A. (Eds.), 1991. Psychiatric Disorders in
America. Free Press, New York.
Robins, L.N., Helzer, J.E., Davis, D.H., 1995. Narcotic use in
southeast Asia and afterward. An interview study of 898 Vietnam
returnees. Archives of General Psychiatry 32, 955961.
Svrakic, D.M., Whitehead, C., Przybeck, T.R., Cloninger, C.R., 1993.
Differential diagnosis of personality disorders by the seven factor
model of temperament and character. Archives of General
Psychiatry 50, 991999.
Tome, M.B., Cloninger, C.R., Watson, J.P., Isaac, M.T., 1997.
Serotonergic autoreceptor blockade in the reduction of antidepressant latency: personality and response to paroxetine and pindolol.
Journal of Affective Disorders 44, 101109.

44

C.R. Cloninger et al. / Journal of Affective Disorders 92 (2006) 3544

Walsh, B.T., Seidman, S.N., Sysko, R., Gould, M., 2002. Placebo
responses in studies of major depression: variable, substantial, and
growing. Journal of the American Medical Association 287,
18401847.
Westen, D., Morrison, K., 2001. A multidimensional meta-analysis of
treatments for depression, panic, and generalized anxiety disorder:
an empirical examination of the status of empirically supported

therapies. Journal of Consulting and Clinical Psychology 69,

875899.
Winokur, G., Clayton, P.J., 1994. The Medical Basis of Psychiatry.
WB Saunders Co., Baltimore.