You are on page 1of 8

Psychological Assessment Copyright 2008 by the American Psychological Association

2008, Vol. 20, No. 3, 292–299 1040-3590/08/$12.00 DOI: 10.1037/a0012933

COMMENTS

The Psychobiological Theory of Temperament and Character: Comment on


Farmer and Goldberg (2008)

C. Robert Cloninger
Washington University

The revised Temperament and Character Inventory (TCI–R) is the third stage of development of a widely
used multiscale personality inventory that began with the Tridimensional Personality Questionnaire
(TPQ) and then the Temperament and Character Inventory (TCI). The author describes the third stage of
the psychobiological theory of temperament and character; empirical tests of its predictions from
genetics, neurobiology, psychosocial development, and clinical studies; and empirical findings that
stimulated incremental changes in theory and test construction. Linear factor analysis is an inadequate
method for evaluating the nonlinear and dynamical nature of the intrapsychic processes that influence
human personality. Traits derived by factor analysis under the doubtful assumption of linearity are
actually heterogeneous composites of rational and emotional processes that differ fundamentally in their
underlying brain processes. The predictions of the psychobiological theory are strongly validated by
extensive data from genetics, neurobiology, longitudinal studies of development, and clinical assessment.
The distinction between temperament and character allows the TCI and TCI–R to outperform other
popular personality inventories in distinguishing individuals with personality disorders from others and
in describing the developmental path to well-being in terms of dynamical processes within the individual
that are useful for both research and clinical practice.

Keywords: Temperament and Character Inventory, psychobiology, factor analysis, complex adaptive
systems

The qualitative distinction between rational cognitive processes Conventional factor analysis assumes that variables are linear.
and prelogical emotional processes is well established in psychol- That is, the measured variables are supposed to have an integral
ogy but has been essentially ignored by factor analysts working in scale with the same average effects across their range, which is
the field of personality. The critique of the revised Temperament often untrue for personality variables (Cloninger, 2005; Turner,
and Character Inventory (TCI–R) by R. F. Farmer and Goldberg Hudson, Butler, & Joyce, 2003). Furthermore, the average effects
(2008) illustrates well the general phenomenon of how adherence of differences between individuals may be almost irrelevant to
to an inappropriate analytic paradigm can lead to specious conclu- what is happening in any particular person across time (Cervone,
sions that ignore or misinterpret key observations, like the differ- 2004; Molenaar, Huizenga, & Nesselroade, 2003). “Linear” think-
ence between emotion and reason. Other analysis of the same data ing systematically neglects the interactions that are crucial for
set considered by R. F. Farmer and Goldberg has already shown understanding the development of the individual as a person over
that the TCI–R outperforms other major multiscale personality time in different situations (Bergman & Magnusson, 1997). Nev-
inventories in terms of its validity in predicting indicators of ertheless, linear methods can be a useful first approximation, as
clinical maturity versus clinical disorder (Grucza & Goldberg, when we sum scores to estimate individual traits in the TCI or
2007). As a result, my comments on factor analytic and other estimate correlations in a particular sample. However, linear meth-
alternative tests of validity may provide an instructive illustration ods often produce inconsistent results in different samples when
for students of personality assessment to begin to recognize the applied to nonlinear dynamical systems, as are found in the rela-
limitations of linear factor analysis as a way of understanding the tionships between temperament and character (Cloninger, Svrakic,
dynamical and nonlinear nature of interactions among emotional & Svrakic, 1997) or the relationships between personality traits
and rational processes in human personality. and regional brain activity (Turner et al., 2003).
The Temperament and Character Inventory (TCI) has been
developed in three stages (Cloninger, 1987, 2004; Cloninger, Svra-
kic, & Przybeck, 1993). The third stage, relevant to the TCI–R, is
C. Robert Cloninger is the copyright holder of the Temperament and
described in my book Feeling Good: The Science of Well-Being
Character Inventory—Revised.
Correspondence concerning this article should be addressed to C. Robert (Cloninger, 2004), which R. F. Farmer and Goldberg (2008) did
Cloninger, Washington University, Departments of Psychiatry, Genetics, not cite. The modifications at each stage were responses to major
and Psychology, School of Medicine, Campus Box 8134, 660 South Euclid empirical findings that forced changes in perspective for me about
Avenue, St. Louis, MO 63110. E-mail: clon@wustl.edu the nature of human personality. I briefly summarize the third

292
COMMENT ON FARMER AND GOLDBERG (2008) 293

stage of the psychobiological theory of temperament and character The development of well-being and the regulation of potential
because R. F. Farmer and Goldberg did not describe the theory conflicts among prelogical emotional drives in self-aware con-
underlying the TCI–R correctly. In particular, they ignored the sciousness involve three rational cognitive processes of foresight,
findings that led to the recognition of the matrix structure of judgment, and insight, which TCI character traits are hypothesized
personality modules and the nonlinear dynamical relations among to measure. Each of these “rational” processes has modules for
these modules. adapting to each type of situation or layer of personality, which are
supposed to be measured by character subscales. For example, TCI
Psychobiological Theory of Temperament and Character Self-Directedness is hypothesized to measure the executive func-
tions of foresight, as shown by a person being self-directed in the
When I began planning functional brain imaging studies with sexual plane (i.e., responsible), in the material plane (i.e., purposeful),
the TCI, I realized that my theory of personality needed to be in the emotional plane (i.e., cheerfully self-accepting), in the intellec-
deepened to understand the modular organization of brain systems tual plane (i.e., resourceful), and in the spiritual plane (i.e., spontane-
and the dynamics of their regulation over different time frames ous so that habits and intentions are congruent). Cooperativeness
(milliseconds in a brain response paradigm or decades across the measures the legislative function of judgment, which involves making
life span). The current theory explains how the same structure of
rules that allow us to get along in a reasonable and flexible manner
human personality can describe variability in human thought regard-
with one another in each plane of our life. Self-Transcendence mea-
less of the time scale (milliseconds to decades). The invariance of the
sures the judicial function or depth of insight that allows us to know
organization of human thought regardless of time scale has strong
intuitively when our legislative rules apply in a particular situation.
implications for understanding the structure and nonlinear dynamics
The three hypothesized branches of mental self-government and their
of personality. The resulting model allowed me to recognize the
components are summarized in Table 1, along with the TCI subscales
underlying matrix structure of the subscales of temperament and
character (Cloninger, 2004) and encouraged greater precision in mea- hypothesized to measure these functions.
surement of the subscales, as is done in the TCI–R. The five layers of personality, and the perspectives that specify
I see human personality as having five layers, which I call them, can be further decomposed to define 25 modules or subplanes
planes of being and described in detail elsewhere (Cloninger, of being in a 5 ⫻ 5 matrix. Each plane of being has five subplanes that
2004). These five layers are concerned with human adaptations in deal with the sexual, material, emotional, intellectual, and spiritual
situations that are perceived to be predominantly concerned with aspects of that plane, as described in detail elsewhere (Cloninger,
reproduction and sexuality (sexual plane), practical everyday ac- 2004). For example, situations that are viewed from a sexual perspec-
tivities concerned with power and possessions (material plane), tive differ in terms of their more particular focus: Some situations
emotional bonds and social attachments (emotional plane), com- involve sexual needs (e.g., copulation), others have a material objec-
munication and culture (intellectual plane), and understanding tive (e.g., reproduction), emotional desire (e.g., sensuality), intellec-
what is beyond individual human existence (spiritual plane). The tual meaning (e.g., acts of sharing), or spiritual awareness (e.g.,
brain systems that make it possible for human beings to adopt charismatic union). Table 2 summarizes the TCI subscales hypothe-
these different perspectives have evolved in a hierarchy of steps sized to measure the emotional aspect of each subplane, with assign-
over time to allow the survival of progressively more flexible, ments based on ordered changes in content of thought as people pass
aware, and creative organisms (Cloninger, 1994, 2002, 2004; through a hierarchy of states of fear or transcendence (Cloninger,
Cloninger & Gilligan, 1987). 2004).

Table 1
Descriptions of Three Higher Cognitive Processes of Foresight, Judgment, and Insight That Regulate the Emotional Drives Relevant
to Each of the Five Layers of Human Personality

Cognitive function Characteristics of the Characteristics of the Characteristics of the Characteristics of the Characteristics of the
(character dimension) sexual layer material layer emotional layer intellectual layer spiritual layer

Foresight (Self- Responsible vs. Purposeful vs. Exuberant vs. Resourceful vs. Spontaneous vs.
Directedness [SD]) irresponsible (SD1) aimless (SD2) insecure (? SD4) inadequate (SD3) predetermined
(SD5)
Judgment Tolerant vs. prejudiced Forgiving vs. Empathic vs. Helpful vs. unhelpful Principled vs.
(Cooperativeness (C1) revengeful (C4) inconsiderate (C2) (C3) opportunistic (C5)
[C])
Insight (Self- Self-forgetful vs. alienated Patient vs. impatient Transpersonal vs. Creative vs. imitative Spiritual vs.
Transcendence [ST]) (prelogical enaction– (concrete–abstract avoiding (vocal–archetypal conventional
categorizing) (ST1) logic) (nonverbal symbols) (preverbal holistic
emotive imagery) schemas) (ST3)
(ST2)

Note. The five layers of human personality are defined by the predominant focus of the person’s perspective in a situation. Within each layer of
personality, maturation involves increasing each of the three character dimensions. Integration of the whole person requires working through these functions
in each of the layers of personality. The Temperament and Character Inventory subscales predicted to measure these processes are indicated in parentheses.
Adapted from Feeling Good: The Science of Well-Being (pp. 219 –225), by C. R. Cloninger, 2004, New York: Oxford University Press.
294 CLONINGER

Table 2
Emotional Aspects of the 25 Subplanes of Being as Described by Specific Modules of Temperament for the Nonspiritual Planes
and Modules of Character Development for the Spiritual Plane

Spiritual
Subplane Sexual plane Material plane Emotional plane Intellectual plane plane

Spiritual subplane Shy vs. beguiling Exploratory vs. Attached vs. detached Perfectionistic vs. Peaceful
(HA3) unexcitable (NS1) (RD3) pragmatic (PS4)
Intellectual subplane Pessimistic vs. optimistic Impulsive vs. rigid Sentimental vs. indifferent Determined vs. Patient
(HA1) (NS2) (RD1) ambivalent (PS2)
Emotional subplane Inhibited vs. uninhibited Irritable vs. stoic Sociable vs. distant Persistent vs. impersistent Charitable
(HA total) (NS total) (RD total) (PS total)
Material subplane Fearful vs. risk-taking Extravagant vs. frugal Warm vs. aloof Eager effort vs. lazy Respectful
(HA2) (NS3) (RD2) (PS1)
Sexual subplane Fatigable vs. vigorous Disorderly vs. regimented Dependent vs. independent Ambitious vs. Hopeful
(HA4) (NS4) (RD4) underachieving (PS3)

Note. HA ⫽ Harm Avoidance; NS ⫽ Novelty Seeking; RD ⫽ Reward Dependence; PS ⫽ Persistence. Adapted from Feeling Good: The Science of
Well-Being (pp. 98 –108), by C. R. Cloninger, 2004, New York: Oxford University Press.

A prominent feature of the matrix structure of personality mod- Linear factor analysis considers the average differences between
ules is that individual items descriptive of personality may be groups at one point in time, which may have little to do with
related to one another as a result of overlaps in the situation, basic patterns of change over time. For example, the extent to which
emotional processes, higher cognitive and spiritual processes, or subscales of IQ are increasing across generations (the Flynn effect)
combinations of these variables (Cloninger, 2004). All personality does not correspond to their factor loadings on general intelligence
modules involve person by situation interactions that are regulated (Flynn, 2007). Likewise, the TCI character subscales develop over
by a set of dynamical nonlinear systems, which allows human time as a nonlinear dynamical system with a spiral structure having
beings to be purposeful, flexible, and self-aware in their adaptation five layers (Cloninger et al., 1997). R. F. Farmer and Goldberg
to life. Human personality is not adequately characterized as a set (2008) did not acknowledge the nonlinear dynamics of personality
of linear traits because the components of personality are nonlin- development, which requires an account of both the correlations
ear in their functional effects and relationships with one another. between individuals at one time and the patterns of change over
Human thought involves frequent rapid adaptation of a person to time, as we have done (Cloninger et al., 1997).
different situations or to different perspectives on the same experi- Although human thought is highly dynamic and has a wide
ence. The connections between different brain modules change sud- range, we can use averages or modes of our thoughts and relation-
denly and synchronously with changes in thought. Human beings ships as central indicators of states of being that are meta-stable, as
have 10 thoughts per second on average, which requires sudden, is typical of equilibrium states in nonlinear dynamical systems.
reversible, nonlinear changes in connections throughout the whole
Frequent meta-stable states of being are often described as per-
brain, as is characteristic of scale-free networks (Cloninger, 2004).
sonality traits, even though they are neither fixed nor linear in their
Most of the information about personality can be captured in the
functions and relationships.
5 ⫻ 5 matrix just described, so personality inventories that are
As a result, personality is a complex expression of nonlinear
designed to be thorough, like the TCI, TCI–R, and NEO Person-
interactions among a whole hierarchy of learning systems that
ality Inventory—Revised, have around 25 subscales. Less detailed
have evolved and that develop over time as a complex adaptive
description is likely to have less useful information, whereas more
process, as described in detail elsewhere (Cloninger, 2002, 2004;
detailed assessments are likely to compromise validity by tiring
subjects. The full three-dimensional structure of personality can be Kaasinen, Maguire, Kurki, Bruck, & Rinne, 2005; Turner et al.,
visualized as a three-dimensional spiral with five planes and as 2003). Each personality module has three aspects, each regulated
increasing in height, width, and depth from the sexual to the by dissociable but overlapping systems of learning and memory: a
spiritual plane (Cloninger, 2004). The height of the spiral indicates somatic or emotional aspect regulated by learning of habits and
a person’s foresight, as measured by their level of TCI Self- skills, a cognitive aspect regulated by learning of facts and prop-
Directedness. The width of the spiral indicates a person’s broad- ositions, and a spiritual aspect regulated by self-aware learning of
ness of social concern and flexibility of judgment, as measured by one’s continuity with and participation in a whole greater than
their level of TCI Cooperativeness. The depth of the spiral indi- oneself (Cloninger, 2004; Tulving, 2002). These three systems can
cates a person’s depth or shallowness of insight, as measured by be dissociated but normally interact through their extensive inter-
their level of TCI Self-Transcendence (Cloninger, 2004). In the connections. For example, a person can know how to do something
fourth dimension of time, a person’s thoughts can shift rapidly (using procedural memory) but not know the fact that he or she
throughout their possible range, spiraling down with increasing knows the skill (a deficit in semantic memory). A person can know
negative emotions and other indicators of ill-health when stressed a particular fact and report it to others (using semantic memory)
(e.g., preoccupied with fears, threats, or desires) or spiraling up with but not have memory of when and where he or she learned the fact
increasing positive emotions and other indicators of well-being when (a deficit in episodic memory). The systems evolve and mature in
calm and secure (e.g., feeling hopeful, kind, and aware). sequence but operate simultaneously from early in childhood.
COMMENT ON FARMER AND GOLDBERG (2008) 295

Development of Temperament Model detail elsewhere (Cloninger, 1986, 1987). Eventually harm avoid-
ance, novelty seeking, reward dependence, and persistence were
Linear models like factor analysis evaluate the average relations shown to be independently inherited in large-scale twin studies
among variables between individuals in groups, which can be an
(Heath, Cloninger, & Martin, 1994; Stallings, Hewitt, Cloninger,
informative preliminary approximation for exploring personality
Heath, & Eaves, 1996).
structure, as was done by Eysenck and others. Under certain
In addition to the evidence for unique genetic determination of
conditions, a linear model can be a useful approximation, but its
the four temperaments, the construct validity of these dimensions
application requires insight and judgment, not just blind faith in
has been strongly supported by their correlation with other mea-
factor analysis as a gold standard. Jeffrey Gray (1981) made a
sures of normal personality, measures of personality disorder sub-
seminal contribution to the field of personality assessment when he
types, and related psychopathology, as described in detail else-
pointed out that learning responses within individuals did not
where (Cloninger, Przybeck, Svrakic, & Wetzel, 1994). For
correspond to these average effects—rather, neurotic introverts
example, typical findings of relations to other biologically based
were more anxiety-prone and responsive to conditioned signals of
models of personality are summarized in Table 3 (Zuckerman &
punishment than stable extraverts, whereas neurotic extraverts
Cloninger, 1996).
were more impulsive in approaching conditioned signals of reward
than stable introverts. Gray recommended a rotation of factors to The TCI temperament dimensions are nonlinear in their effects
provide a better account of the processes taking place within on emotional responses and behavior. For example, only individ-
individuals as they learned. Subsequently social– cognitive and uals with high scores in Harm Avoidance have an augmented
developmental psychologists have emphasized the need for a startle response when they are previewing unpleasant images. In
model of learning processes within the person as he or she adapts contrast, only individuals with low scores in Harm Avoidance have
to different situations (Bergman & Magnusson, 1997; Cervone, a reduced startle response when they are previewing pleasant
2004; Mischel & Shoda, 1998) and regard the factor-analytically images (Corr, Kumari, Wilson, Checkley, & Gray, 1997; Corr,
derived models of between-persons differences as unacceptable for Pickering, & Gray, 1995; Corr, Wilson, et al., 1995). Likewise, the
understanding emotional and cognitive processes (Cervone, 2004). activation of blood flow in a brain circuit involving the ventral
However, factor analysts have missed the point of such critiques striatum, anterior cingulate cortex, and orbitofrontal cortex in
and continued to rely on models of the average relations among response to increasing frequency of frustrative nonreward has a
variables between individuals rather than trying to understand positive slope for individuals high in TCI Persistence and a neg-
personality as a set of dynamic intrapsychic processes by which a ative slope for those with low TCI Persistence (Gusnard et al.,
person learns to adapt to different life situations. 2003). Such nonlinearity is a fundamental characteristic of the
Following Gray’s (1981) observations about learning differ- functional relationships of all the TCI dimensions of temperament
ences within individuals, I synthesized genetic, pharmacological, and character, not just particular dimensions (Cloninger, 2004;
and psychosocial data that also pointed to the need for a model of Cloninger et al., 1997; Turner et al., 2003).
learning processes occurring within the person (Cloninger, 1986, There are strong relationships of TCI temperament dimensions
1987). For example, extraversion measures were known to have to psychomotor functions in specific brain systems related to
differentially heritable components of venturesome risk-taking, processing of negative emotions and habits. However, the average
impulsivity, and sociability (Eaves & Eysenck, 1975) that I later relations among temperament, neurotransmitters, and brain func-
described as low harm avoidance, high novelty seeking, and high tions are usually weak or inconsistent unless there is careful
reward dependence. I formulated the processes in terms of pre- attention to the specific temperament, specific type of neurotrans-
logical emotional processes of associative conditioning, as was mitter receptor, and specific location and connections among the
properly described by R. F. Farmer and Goldberg (2008). My brain regions. For example, carriers of a gene that influences the
initial development of the temperament dimensions is described in activity of the serotonin transporter have increased harm avoid-

Table 3
Correlations (r ⫻ 100) Between the Temperament and Character Inventory (TCI) Scales and Those of the Eysenck Personality
Questionnaire (EPQ—Revised) and the Zuckerman–Kuhlman Personality Questionnaire (ZKPQ)

Harm Novelty Reward Self- Self-


Scale avoidance seeking dependence Persistence directedness Cooperativeness transcendence

EPQ Neuroticism 59 ⴚ45


EPQ Extraversion ⴚ53 44 23 18
EPQ Psychoticism 41 ⴚ45 ⫺29 ⫺31 ⴚ42
EPQ Lie ⫺21 25 34
ZKPQ Neuroticism 66 ⴚ49
ZKPQ Impulsive Sensation ⫺39 68 ⫺20 28
ZKPQ Hostility ⫺27 ⫺32 ⴚ60
ZKPQ Sociability ⫺38 37 31
ZKPQ Activity ⫺29 46 36

Note. Correlations over .40 are in bold. Only significant correlations are shown, N ⫽ 207. Adapted from “Relationships Between Cloninger’s,
Zuckerman’s, and Eysenck’s Dimensions of Personality,” by M. Zuckerman and C. R. Cloninger, 1996, Personality & Individual Differences, 21, p.
283–285.
296 CLONINGER

ance, increased reactivity of the amygdala to fear, and increased In particular, low TCI Self-Directedness strongly specified the
risk of depression compared to others (Pezawas et al., 2005). Harm presence and severity of all personality disorders diagnosed by
Avoidance scores explain 30% of the variance in the functional experts using reliable structured interviews (Svrakic et al., 1993).
connectivity between the amygdala and perigenual cingulate, Low TCI Cooperativeness also contributed to the detection of
which are involved in perceptual processing of fearful stimuli. milder personality disorders (Strack, 2006; Svrakic et al., 1993).
More generally, the TCI temperament dimensions have strong Various configurations of the TCI dimensions differentiated the
nonlinear relations with the specific parts of the brain involved in symptoms of the full range of personality disorder subtypes. At the
processing of negative emotions, habits, and motor skills (Turner same time high scores on all three character dimensions are
et al., 2003). In contrast, the TCI character dimensions have strong strongly associated with well-being, as indicated by resilient pos-
nonlinear relationships with the parts of the neocortex involved in itive emotions, satisfying quality of life, and virtuous conduct
the processing of facts and propositions, not negative emotions and (Cloninger, 2004).
habits (Turner et al., 2003). The TCI character dimensions differ qualitatively from the
temperament dimensions, as expected for processes that are ratio-
Differentiation of Temperament nal rather than emotional. The TCI temperament dimensions can
and Character Dimensions be roughly described as bipolar: Both extremes of each of the four
temperaments have some practical advantages as well as some
Despite the success of the Tridimensional Personality Question- practical disadvantages, depending on the situation. For example,
naire (TPQ) as a description of susceptibility to negative emotions being high in Harm Avoidance leads to anxiety that can be socially
and subtypes of personality disorders, something important was crippling, but it can also promote caution that can be life-
missing from it as a description of maturity and well-being in preserving. On the other hand, being low in Harm Avoidance is
human personality. I found that groups with different configura- associated with risk-taking, which can lead to death or success
tions of the temperament dimensions did differ in their average depending on the situation. The temperaments describe a person’s
probability of having a personality disorder, as reviewed elsewhere emotional style without any prejudice about what is socially de-
(Cloninger, 2004, pp. 40 – 44). However, to my surprise, these sirable or undesirable about a particular individual.
average effects in groups did not allow accurate measurement of In contrast, the character dimensions can be roughly described
whether a particular individual had a personality disorder or how as unipolar variables related to a person’s voluntary goals and
mature they were (Cloninger et al., 1997)! My colleagues and I values. They are skewed toward high values by the tendency of
studied personality variation in both the general population and in people to recognize and endorse socially desirable behaviors. For
samples of people with varying degrees of mental disorder. The assessing the validity of TCI self-reports, we have derived subtle
full range of possible temperament configurations occurred in measures of TCI dimensions based on properties of specific re-
both mature and immature people, even though the probabilities of sponse sets, such as the frequency of endorsement of rare items
the configurations differed on average. The average associations regardless of their content (Cloninger et al., 1994; Cloninger &
allowed no reliable statements to be made about the maturity or Svrakic, 1997). The character scales, but not the temperament
health of any individual person on the basis of their temperament scales, are strongly influenced by specific variables related to
configuration alone. In response to this observation, I checked frequent rare responses, conformity to intermediate cultural norms,
what variables were measured in other personality tests but were or acquiescence. The temperament scales are influenced by se-
not accounted for by the TPQ (Cloninger et al., 1994). These quence contingencies related to the emotional impact of prior item
variables included three sets of variables that I concluded were endorsements.
related to self-directedness (e.g., low repression in the Minnesota
Multiphasic Personality Inventory [MMPI], internal locus of con- Findings About the TCI–R
trol, purposefulness), cooperativeness (e.g., agreeability in the
NEO Personality Inventory, low hostility), and self-transcendence The findings about the TCI–R dimensions reported by R. F.
(e.g., absorption in the Multidimensional Personality Question- Farmer and Goldberg (2008) are based on a preliminary version of
naire [MPQ], transpersonal experiences, spirituality). the test with 295 items from which we derived the test now called
The TCI was developed to measure the four temperaments and the TCI–R with 240 items. Subsequently we conducted an inde-
the three character dimensions of Self-Directedness, Cooperative- pendent normative study with 917 subjects in Missouri using the
ness, and Self-Transcendence. My colleagues and I found that each final version of the test and are preparing a report for its publica-
of the seven dimensions had good internal consistency and sub- tion elsewhere. Although R. F. Farmer and Goldberg did not report
stantial unique variability in descriptive content (Cloninger et al., this result for the Oregon sample, the internal consistency of the
1993, 1994). In addition, to summarize briefly a literature of over four TCI–R temperament dimensions ranged from .84 to .92, and
2,000 peer-reviewed articles (see https://psychobiology.wustl for the three character dimensions from .89 to .90 in our Missouri
.edu), each of the seven TCI dimensions has been shown to have sample. We expect that these measures are similar to that in the
differential relations with personality disorder subtypes (Goldman, Oregon sample because the psychometric properties of the 29 sub-
Skodal, McGrath, & Oldham, 1994; Mulder & Joyce, 1997; Svra- scales and 240 items as reported by R. F. Farmer and Goldberg are
kic, Whitehead, Przybeck, & Cloninger, 1993), other mental dis- similar to what we have observed with the current test.
orders (Cloninger et al., 1994; A. Farmer et al., 2003), unique The major specific findings about the structure of the TCI–R
patterns of functional brain activity (Turner et al., 2003), and stated by R. F. Farmer and Goldberg (2008) were that factor
unique genetic antecedents (Gillespie, Cloninger, Heath, & Martin, analyses suggested that (a) there was evidence for six or seven
2003). personality factors in the TCI–R, including a distinct factor for
COMMENT ON FARMER AND GOLDBERG (2008) 297

Self-Transcendence; (b) Harm Avoidance and Self-Directedness psychological or etiological equivalence and can be dissociated in
were moderately and negatively correlated, and they loaded to- individuals because of their nonlinear functional relationship
gether on one factor that is usually called Neuroticism; (c) Reward (Cloninger et al., 1997). For example, the so-called “love” hor-
Dependence and Cooperativeness were moderately and positively mone oxytocin is released into blood after sexual orgasm regard-
correlated, and some subscales of both dimensions loaded together less of gender and is moderately correlated with Reward Depen-
on one factor; (d) the other three factors identified as interpretable dence (r ⫽ .43) but not with Cooperativeness (r ⫽ .09; Bell,
by R. F. Farmer and Goldberg were defined by a factor with three Nicholson, Mulder, Luty, & Joyce, 2006). However, such biolog-
of four Novelty Seeking subscales, a factor with all four Persis- ical findings can be difficult to replicate in nonlinear systems, as
tence subscales, and a factor with two of four subscales of Reward reviewed in depth elsewhere (Cloninger, 2000, 2004).
Dependence; and (e) a seventh factor was described by R. F. In addition, R. F. Farmer and Goldberg (2008) pointed out that
Farmer and Goldberg as uninterpretable; however, it is defined by the 140-item short form of the TCI–R has nearly the same structure
low disorderliness (NS4) and highly principled behavior (C5), and reliability as the longer form for measuring the seven higher
which can be understood as respect for order and principle rather order dimensions of temperament and character. I agree that the
than preference for disorder and opportunistic behavior. short form works well for its purpose. However, the short form
The findings of R. F. Farmer and Goldberg (2008) are similar to does not provide the precision needed for clinical work with
what we have usually observed with linear factor analysis of the individuals using the TCI subscales. The main motivation for the
TCI and TCI–R in several different countries. It is known that the development of the TCI–R was to increase precision in measuring
structure of interindividual differences does not correspond to that its subscales by using a 5-point Likert response set rather than
of intraindividual differences in nonlinear dynamical systems. As true–false and to provide subscales for Persistence to increase its
a result, since 1997, I have recommended that linear factor anal- reliability.
yses of the structure of the TCI consider temperament separately
from character because of their strongly nonlinear relationship Conclusions
(Cloninger, 2000; Cloninger et al., 1997; Hansenne, Delhez, &
Cloninger, 2005). Alternatively, some may wish to apply non- It is important to underscore the important falsification of one of
linear factor analytic methods appropriate for intraindividual the early predictions I made about personality. Based on early
development (Molenaar et al., 2003). Nevertheless, this recom- work by Cattell and Loehlin on the differential heritability of
mendation has not been heeded by some investigators, who personality traits (Loehlin, 1986), I expected that the character
have found results with linear models just like R. F. Farmer and dimensions would be expressions of higher cognitive processes
Goldberg did. As expected for nonlinear systems, the linear involving the propositional learning system and less heritable than
interindividual models do not conform to the hypothesized the temperament dimensions (Cloninger, 1994; Cloninger et al.,
intraindividual structure of the TCI (Ball, Tennen, & Kranzler, 1993). I have learned that many people really want these two
1999; Herbst, Zonderman, McCrae, & Costa, 2000). hypotheses to be linked because they like to think that we can
It is noteworthy that R. F. Farmer and Goldberg (2008) identi- separate biology and psychology, at least the psychology of higher
fied six or seven interpretable factors in the TCI–R rather than only cognitive and social processes. In fact, we confirmed the hypoth-
five factors. I do not disagree with the data presented, but the esis that character is related to higher cognitive processing in more
appropriateness of their method for structural analysis and psycho- recently evolved parts of the brain than is temperament (Cloninger,
metric evaluation of the TCI–R is questionable for the fundamental 2002; Turner et al., 2003), but to my surprise, we found that the
reasons I have presented earlier. First, the functional relationships heritability of character was as great as that of temperament
of the TCI dimensions are not linear as assumed by factor analysis. (Cloninger, 2004; Gillespie et al., 2003). This important finding
Second, the average relationships among variables between indi- suggests that the popular dichotomy in science between the neu-
viduals do not provide a model of the relationships within indi- robiological and psychosocial paradigms of human personality is
viduals that corresponds to a dynamical intrapsychic model of not well justified.
personality as the TCI–R is intended to provide. Third, the items of On several occasions I have presented data showing the unique
the TCI are elements of a nonlinear matrix structure that allows for genetic determinants and functional brain processes characteristic
simultaneous overlap due to common situations, emotional pro- of TCI Harm Avoidance and Self-Directedness to personality trait
cesses, rational cognitive influences, and variable combinations of theorists who are accustomed to using factor analytic measures of
these variables; consequently, factor analysis at the item level does neuroticism. I have also pointed out the genetic, neurobiological,
not provide a valid way to identify the nonlinear structure of the and psychological heterogeneity of the multiple temperament di-
intrapsychic processes measured by the TCI because the data mensions that contribute to factor analytic measures of extraver-
violate the multivariate normal distributions assumed by factor sion. However, factor analysts persist in ignoring these findings
analysis. Fourth, moderate or even strong correlations among about emotional and rational processes in human personality de-
variables do not ensure psychological or etiological equivalence. spite the inability of statistical methods alone to define the struc-
In particular, it is questionable that high scores on Harm Avoid- ture of personality unambiguously.
ance should be combined with low scores on Self-Directedness Premature preoccupation with creating the impression of a sim-
because these qualitatively distinct emotional and rational vari- ple linear factor structure for the complex architecture of human
ables have different genetics (Gillespie et al., 2003), neurobiology personality is likely to give the appearance of precision while
(Turner et al., 2003), development (Cloninger, 2003), psychomet- actually reducing the predictive validity of the instruments. A
rics, and meaning, as discussed earlier. Likewise, the correlations recent study (Grucza & Goldberg, 2007) set out to compare the
among Reward Dependence and Cooperativeness do not imply predictive validity of 11 popular multiscale personality inventories,
298 CLONINGER

including the TCI–R. The possibility of such a comparison of validity Cloninger, C. R. (2000). Biology of personality dimensions. Current Opin-
is a tribute to the foresight and persistence of Lew Goldberg and to the ion in Psychiatry, 13, 611– 616.
trust of many investigators that data about their instruments would be Cloninger, C. R. (2002). Functional neuroanatomy and brain imaging of
handled in a collegial and professional manner. personality and its disorders. In H. D’haenen, J. A. den Boer, & P.
Three types of criteria were used to compare the predictive Willner (Eds.), Biological psychiatry (Vol. 2, pp. 1377–1385). Chi-
validity of the various inventories: clusters of behavioral acts (drug chester, United Kingdom: Wiley.
Cloninger, C. R. (2003). Completing the psychobiological architecture of
use, undependability, friendliness, erudition, communication, and
human personality development: Temperament, character, & coherence.
creativity), descriptions of the five factor domains by knowledge-
In U. M. Staudinger & U. E. R. Lindenberger (Eds.), Understanding
able informants, and clinical indicators potentially associated with human development: Dialogues with lifespan psychology (pp. 159 –182).
various types of psychopathology (sociopathy, magical ideation, Boston: Kluwer.
obsessive compulsion, depression, dissociation, and borderline Cloninger, C. R. (2004). Feeling good: The science of well-being. New
personality). The TCI–R predicted these criteria as well as, and in York: Oxford University Press.
some cases better than, every other major inventory. At the level of Cloninger, C. R. (2005). How does personality influence mortality in the
the higher order constructs like the seven TCI–R dimensions, the elderly? Psychosomatic Medicine, 67(6), 839 – 840.
range of mean predictive validity coefficients ranged from .37 to Cloninger, C. R., & Gilligan, S. B. (1987). Neurogenetic mechanisms of
.45, with the TCI at .45. At the level of lower order constructs like learning: A phylogenetic perspective. Journal of Psychiatric Research,
the 29 TCI–R subscales, the mean predictive validity coefficients 21(4), 457– 472.
ranged from .39 to .45, with the TCI–R at .44. The TCI–R was Cloninger, C. R., Przybeck, T. R., Svrakic, D. M., & Wetzel, R. D. (1994).
particularly strong in the prediction of acts of friendliness with The Temperament and Character Inventory: A guide to its development
Reward Dependence, indicators of borderline personality with and use. St. Louis, MO: Washington University, Center for Psychobi-
ology of Personality.
Self-Directedness, indicators of sociopathy with Cooperativeness,
Cloninger, C. R., & Svrakic, D. M. (1997). Integrative psychobiological
and indicators of magical ideation with Self-Transcendence. The
approach to psychiatric assessment and treatment. Psychiatry, 60, 120 –141.
TCI–R outperformed all other personality inventories overall in
Cloninger, C. R., Svrakic, D. M., & Przybeck, T. R. (1993). A psychobi-
the prediction of indicators of clinical maturity: The range of ological model of temperament and character. Archives of General
multiple correlations was .27 to .45 for other tests, whereas the Psychiatry, 50, 975–990.
TCI–R had the best mean coefficient of .53. Cloninger, C. R., Svrakic, N. M., & Svrakic, D. M. (1997). Role of
These findings identify two facts worthy of consideration by personality self-organization in development of mental order and disor-
factor analysts in general. First, traditional linear psychometrics der. Development and Psychopathology, 9, 881–906.
for evaluating personality inventories (such as linear factor struc- Corr, P. J., Kumari, V., Wilson, G. D., Checkley, S., & Gray, J. A. (1997).
ture and coefficient alpha) may have little bearing on actual Harm avoidance and affective modulation of the startle reflex: A repli-
predictive validity. Second, the three character scales of the TCI–R cation. Personality & Individual Differences, 22, 591–593.
allowed it to outperform all other popular tests in the prediction of Corr, P. J., Pickering, A. D., & Gray, J. A. (1995). Personality and
measures of maturity versus clinical disorder. This suggests that reinforcement in associative and instrumental learning. Personality &
the psychobiological model of temperament and character may Individual Differences, 19, 47–71.
offer important insights and predictive validity for understanding Corr, P. J., Wilson, G. D., Fotiadou, M., Kumari, V., Gray, N. S., Checkley,
S., et al. (1995). Personality and affective modulation of the startle
the interaction of the rational and emotional in human personality.
reflex. Personality & Individual Differences, 19, 543–553.
We have much to learn beyond what has been established about
Eaves, L. J., & Eysenck, H. J. (1975). The nature of extraversion: A
linear models of interindividual differences.
genetical analysis. Journal of Personality and Social Psychology, 32(1),
102–112.
References Farmer, A., Mahmood, A., Redman, K., Harris, T., Sadler, S., & McGuffin,
P. (2003). A sib-pair study of the Temperament and Character Inventory
Ball, S. A., Tennen, H., & Kranzler, H. (1999). Factor replicability and
in major depression. Archives of General Psychiatry, 60(5), 490 – 496.
validity of the Temperament and Character Inventory in substance abuse
Farmer, R. F., & Goldberg, L. R. (2008). A psychometric evaluation of the
patients. Psychological Assessment, 11(4), 514 –524.
revised Temperament and Character Inventory (TCI–R) and the TCI–
Bell, C. J., Nicholson, H., Mulder, R. T., Luty, S. E., & Joyce, P. R. (2006).
140. Psychological Assessment, 20, 281–291.
Plasma oxytocin levels in depression and their correlation with the
Flynn, J. R. (2007). What is intelligence? Cambridge, United Kingdom:
temperament dimension of reward dependence. Journal of Psychophar-
macology, 20(5), 656 – 660. Cambridge University Press.
Bergman, L. R., & Magnusson, D. (1997). A person-oriented approach in Gillespie, N. A., Cloninger, C. R., Heath, A. C., & Martin, N. G. (2003).
research on developmental psychopathology. Development and Psycho- The genetic and environmental relationship between Cloninger’s dimen-
pathology, 9, 291–319. sions of temperament and character. Personality & Individual Differ-
Cervone, D. (2004). The architecture of personality. Psychological Review, ences, 35, 1931–1946.
111(1), 183–204. Goldman, R. G., Skodal, A. E., McGrath, P. J., & Oldham, J. M. (1994).
Cloninger, C. R. (1986). A unified biosocial theory of personality and its Relationship between the Tridimensional Personality Questionnaire and
role in the development of anxiety states. Psychiatric Developments, 3, DSM–III–R personality traits. American Journal of Psychiatry, 151,
167–226. 274 –276.
Cloninger, C. R. (1987). A systematic method for clinical description and Gray, J. A. (1981). A critique of Eysenck’s theory of personality. In H. J.
classification of personality variants: A proposal. Archives of General Eysenck (Ed.), A model of personality (pp. 246 –276). New York:
Psychiatry, 44, 573–587. Springer-Verlag.
Cloninger, C. R. (1994). The genetic structure of personality and learning: Grucza, R. A., & Goldberg, L. R. (2007). The comparative validity of 11
A phylogenetic perspective. Clinical Genetics, 46, 124 –137. modern personality inventories: Predictions of behavioral acts, infor-
COMMENT ON FARMER AND GOLDBERG (2008) 299

mant reports and clinical indicators. Journal of Personality Assessment, Mulder, R. T., & Joyce, P. R. (1997). Temperament and the structure of
89(2), 167–187. personality disorder symptoms. Psychological Medicine, 27, 99 –106.
Gusnard, D. A., Ollinger, J. M., Shulman, G. L., Cloninger, C. R., Price, J. L., Pezawas, L., Meyer-Lindenberg, A., Drabant, E. M., Verchinski, B. A.,
Van Essen, D. C., et al. (2003). Persistence and brain circuitry. Proceedings Munoz, K. E., Kolachana, B. S., et al. (2005). 5-HTTLPR polymorphism
of the National Academy of Sciences, USA, 100(6), 3479 –3484. impacts human cingulate–amygdala interactions: A genetic susceptibil-
Hansenne, M., Delhez, M., & Cloninger, C. R. (2005). Psychometric ity mechanism for depression. Nature Neuroscience, 8(6), 838 – 834.
properties of the Temperament and Character Inventory—Revised Stallings, M. C., Hewitt, J. K., Cloninger, C. R., Heath, A. C., & Eaves,
(TCI–R) in a Belgian sample. Journal of Personality Assessment, 85(1), L. J. (1996). Genetic and environmental structure of the Tridimensional
40 – 49. Personality Questionnaire: Three or four temperament dimensions?
Heath, A. C., Cloninger, C. R., & Martin, N. G. (1994). Testing a model for Journal of Personality and Social Psychology, 70, 127–140.
the genetic structure of personality: A comparison of the personality Strack, S. (Ed.). (2006). Differentiating normal and abnormal personality
systems of Cloninger and Eysenck. Journal of Personality and Social (2nd ed.). New York: Springer.
Psychology, 66, 762–775.
Svrakic, D. M., Whitehead, C., Przybeck, T. R., & Cloninger, C. R. (1993).
Herbst, J. H., Zonderman, A. B., McCrae, R. R., & Costa, P. T., Jr. (2000).
Differential diagnosis of personality disorders by the seven-factor model
Do the dimensions of the Temperament and Character Inventory map a
of temperament and character. Archives of General Psychiatry, 50,
simple genetic architecture? Evidence from molecular genetics and
991–999.
factor analysis. American Journal of Psychiatry, 157, 1285–1290.
Tulving, E. (2002). Episodic memory: From mind to brain. Annual Review
Kaasinen, V., Maguire, E. A., Kurki, T., Bruck, A., & Rinne, J. O. (2005).
of Psychology, 53, 1–25.
Mapping brain structure and personality in late adulthood. NeuroImage,
24, 315–322. Turner, R. M., Hudson, I. L., Butler, P. H., & Joyce, P. R. (2003). Brain
Loehlin, J. C. (1986). Are California Psychological Inventory items dif- function and personality in normal males: A SPECT study using statis-
ferently heritable? Behavior Genetics, 16(6), 599 – 603. tical parametric mapping. NeuroImage, 19, 1145–1162.
Mischel, W., & Shoda, Y. (1998). Reconciling processing dynamics and Zuckerman, M., & Cloninger, C. R. (1996). Relationships between Clon-
personality dispositions. Annual Review of Psychology, 49, 229 –258. inger’s, Zuckerman’s, and Eysenck’s dimensions of personality. Person-
Molenaar, P. C. M., Huizenga, H. M., & Nesselroade, J. R. (2003). The ality & Individual Differences, 21, 283–285.
relationship between the structure of interindividual and intra-individual
variability: A theoretical and empirical vindication of developmental
systems theory. In U. M. Staudinger & U. E. R. Lindenberger (Eds.), Received November 13, 2007
Understanding human development: Dialogues with lifespan psychology Revision received February 5, 2008
(pp. 339 –360). Boston: Kluwer. Accepted February 7, 2008 䡲

You might also like