Creativity:

Theories, Prediction, and Etiology

Diploma Thesis
for the Final Examination
in the Department of Psychology
of the Faculty for Psychology and Sports Sciences
at the University of Bielefeld

Presented by
Lars Penke
in 2003

First Assessor and Supervisor:

PD Dr. F. M. Spinath

Second Assessor
Prof. Dr. A. Angleitner

Department of
Individual Differences and Personality Research
Acknowledgements I

Acknowledgements

I would like to thank Dr. Frank M. Spinath and Prof. Dr. Alois Angleitner for their
support during this study, good cooperation during the last two years and, together
with Prof. Dr. Rainer Riemann and Prof. Dr. Peter Borkenau, for giving me access to
the extraordinary BiLSAT and GOSAT twin data sets. This study owes much more
than a simple ‘thank you’ can express to Anna-Karen Stürmer and my fiancé Simone
Sprenger for their unbelievable stamina during the scoring of the T-88 data. The latter
one also deserves all my gratitude for giving the only meaning to it all. Finally, I would
like to thank my parents, Hermann and Sigrid Penke, for supporting me along the
whole way that led me to where I am now.

I want to dedicate this thesis to the memorial of Prof. Dr. Linda Mealey (1955-2002),
who substantially encouraged my path during my undergraduate years by taking me
and my early ideas serious.

Parts of this thesis were already presented as a poster at the 7. Fachgruppentagung

für Differentielle Psychologie, Persönlichkeitspsychologie und Psychologische
Diagnostik (DPPD), Halle, Germany, 29.-30. 09. 2003.
Table of contents II

TABLE OF CONTENTS

1. INTRODUCTION ................................................................................................ 1
1.1 Background .................................................................................................. 1
1.2 Cognitive influences on creativity ................................................................. 2
1.3 Creativity and personality ............................................................................. 6
1.4 Models and theories of a disposition towards creativity.............................. 12
1.5 The etiology of creativity............................................................................. 17
1.6 Aims and hypotheses of the present study................................................. 21

2. PILOT STUDY: THE T-88 AS A MEASURE OF CREATIVITY........................ 23

2.1 Method ....................................................................................................... 23
2.1.1 Sample................................................................................................. 23
2.1.2 Materials .............................................................................................. 24
2.1.3 Procedure ............................................................................................ 27
2.2 Results ....................................................................................................... 28
2.2.1 Reliability analyses............................................................................... 28
2.2.2 Correlational analyses.......................................................................... 29
2.3 Discussion .................................................................................................. 31

3. MAIN STUDY: CREATIVITY IN BiLSAT AND GOSAT ................................... 33

3.1 Method ....................................................................................................... 33
3.1.1 Sample................................................................................................. 33
3.1.1.1 Total sample .................................................................................. 33
3.1.1.2 Core sample .................................................................................. 34
3.1.2 Materials and procedure ...................................................................... 34
3.1.2.1 Measurement point 1: II. BiLSAT wave.......................................... 34
3.1.2.1 Measurement point 2: GOSAT ...................................................... 35
3.1.2.3 Measurement point 3: III. BiLSAT wave......................................... 37
3.1.3 Zygosity determination ......................................................................... 38
3.2 Results ....................................................................................................... 38
3.2.1 Reliability analyses............................................................................... 38
3.2.2 Aggregation of measures ..................................................................... 42
3.2.3 Age and sex effects.............................................................................. 46
Table of contents III

3.2.4 Phenotypic relations of creativity with intelligence and personality ...... 47

3.2.4.1 Relations of creativity with intelligence .......................................... 47
3.4.2.2 Relations of creativity with personality ........................................... 50
3.4.2.3 Regression analyses ..................................................................... 52
3.4.2.4 The structure of creativity after controlling for intelligence and
personality ..................................................................................... 57
3.2.5 Genetic and environmental influences on creativity and its
correlations with personality and intelligence....................................... 60
3.2.5.1 Univariate behavioral genetic analyses ......................................... 60
3.2.5.2 Bivariate behavioral genetic analyses............................................ 68
3.3 Discussion .................................................................................................. 74

4. GENERAL DISCUSSION AND CONCLUSION ............................................... 81

5. REFERENCES ................................................................................................. 85

6. APPENDIX ..................................................................................................... 100

6.1 The T-88................................................................................................... 100
6.2 Rater instructions for the T-88 .................................................................. 103
6.3 Response baselines for the T-88.............................................................. 105

7. EIDESSTATTLICHE ERLÄRUNG (AFFIDAVIT) ........................................... 107

Introduction 1

1. Introduction

1.1 Background

Creativity is a phenomenon that has always fascinated lay people as well as scientists.
It is for example valued as a property of pieces of art or literature, musical compositions,
scientific works, narrations, witty comments, decorations, and technical or social
inventions. In all these various facets, creativity is a prime source of cultural progress
and responsible for a multitude of small contributions to our everyday enjoyment and
well-being. Because of these effects, it is also appreciated as an attribute of employees,
artists, entertainers, scientists, friends and mates. Its social and cultural importance led
just about every major personality psychologist of the 20th century – be it Freud, Jung,
Adler, Skinner, Maslow, Murray, Rogers, Kelly, Guilford, Cattell, or Eysenck – to write
about creativity (Woodman, 1981).
Mayer (1999) reviewed seven definitions given by authors contributing to the 1999
‘Handbook of Creativity’ (Sternberg, 1999), and summarized them as “[…] creativity
involves the creation of an original and useful product.” (italics in original). This definition
can account as representative for those traditionally used in psychology (see also Stein,
1953; MacKinnon, 1962). Under different perspectives, creativity can be studied as a
property of a person, process, product or even situation. In the light of the important, yet
often ignored, hedonistic function of creativity, I want to emphasize the necessity of a
subjective aspect in this definition. Its second part, ‘useful’, should explicitly include the
potential of a creative product to satisfy subjective aesthetic needs. Without doubt, there
are areas of creative endeavor where more objective criteria can and should be used. In
evaluating the creativity of a scientific theory, for example, criteria like parsimony or
practical and explanatory value should be given precedence. But most products valued
for creativity – especially in the arts and in everyday life - do not belong to such areas.
Generally excluding the aesthetic aspect for the mere sake of simplicity, as suggested
by Runco (1993), would make it hard to find a reason why things like paintings, poems,
songs or jokes should be valued by people at all. Indeed, attempts to validate a more
objective definition of usefulness (or appropriateness) as an aspect of creativity have
failed (Runco & Charles, 1993). Contrary to common prejudices, subjective evaluations
of aesthetic appeal are by no means purely idiosyncratic or arbitrary. After all, aesthetic
preferences stem from cognitive and emotional adaptations which are evolved solutions
Introduction 2

to recurring problems in our phylogenetic history (Voland & Grammer, 2003). Thornhill
(2003) put it this way: “I am arguing that intellectual aesthetic value represents a
functionally based way of dealing with a cultural environment that is full of diverse ideas.
In this perspective, cultural learning of values is not arbitrary. Learning mechanisms, in
conjunction with feeling mechanisms and mechanisms of self-awareness that allow us
to test how our ideas and behaviors are perceived by others, guide us through a maze
of ideas towards intellectual beauty. Appropriate values will often differ between
societies and within societies between social strata and individuals.” Not surprisingly,
Amabile (1982, 1983a) found across several studies that people do show considerable
interjudge agreement (around .80) on the creativity of products, even when they use
their own definition of what is creative. She suggested a consensual definition of
creativity, stating that “[a] product or response is creative to the extent that appropriate
observers independently agree it is creative. Appropriate observers are those familiar
with the domain in which the product was created or the response articulated” (Amabile,
1982, p. 1001). The present thesis will use Amabile’s definition as a necessary
supplement to the traditional one.
Over 50 years after Guilford’s (1950) APA presidential address ignited the intensive
scientific study of creativity, it is still much less understood than other psychological
constructs (Brown, 1989). Many questions, some of them fundamental ones, must be
regarded unanswered (Mayer, 1999). Several of these questions concern the person
perspective on creativity (Simonton, 1999a). Despite the opinion of some early
sociologists and anthropologists, who attributed the advent of creative products solely to
environmental factors (e.g. Sorokin, 1937-1941; White, 1949), a position that has been
proven as too parsimonious (Simonton, 1988, 1999a), the mere existence of individual
differences in a disposition towards creativity (Barron, 1955) is now widely accepted. Its
very nature, however, the answer to the central question “What makes a person
creative?”, remains unclear. Especially urgent issues include (1) how creativity relates
to cognitive variables such as general intelligence, (2) which personality traits facilitate
or are necessary for a creative thinking style, (3) the etiology of creativity, and (4) the
heterogeneity of creativity. All of these issues will be dealt with in this thesis.
Introduction 3

1.2 Cognitive influences on creativity

The relationship of individual differences in creativity to individual differences in

cognitive abilities, such as intelligence, is still discussed. Indeed, any possible
relationship between creativity and intelligence has been proposed (Sternberg &
O'Hara, 1999). In one extreme version, creativity and intelligence are regarded as totally
independent, a position especially taken by some cognitive psychologists with strong
focus on the creative process (Hayes, 1989; Weisberg, 1992, 1999; Klahr & Simon,
1992). These authors see creativity as a mental operation accessible to everyone, only
dependent on domain-specific knowledge (i.e. the amount of exposure to and expertise
in a given field) and deliberate practice. Their position therefore denies not only the
influence of intelligence, but of any individual difference beyond knowledge and
motivational factors, on creativity. While the results from research on problem solving
and insight are exhaustive and surely noteworthy, such a strong position might be too
neglectful and mainly reflective of the unfortunate, but still existing intellectual borders
between psychological subdisciplines in general (Gigerenzer, 1998) and particularly
experimental and correlational approaches (Cronbach, 1957): As pointed out by
Simonton (2003), it is difficult to conceive a basic creative process which is neither
enhanced nor inhibited by at least one individual-difference variable that has been
empirically associated with actual creative behavior. One of these variables surely is
psychometric intelligence.
But contrary to some early believes of the other extreme, a high IQ (‘giftedness’;
Jensen, 1996) has proven as insufficient for creativity ever since Terman’s (1925;
Burks, Jensen & Terman, 1930; Terman & Oden, 1947, 1959) famous longitudinal study
of 1528 highly gifted children, which had a mean IQ of 151. While most of these children
achieved remarkable occupational success in later life, none of them showed a
noteworthy sign of creativity (see Eysenck (1995a) for an extensive review of the
Terman study). Indeed, most of the numerous empirical studies concerning the
association between psychometric intelligence and creativity yielded only a weak
relationship (Barron & Harrington, 1981). For example, Torrance (1975) reported that
the median of 178 correlation coefficients between IQ and the Torrance Test of Creative
Thinking (TTCT; Torrance, 1974) was only .20, and Hocevar (1980) reported a
correlation of .26 between IQ and mean self-reported creative activities and
accomplishments across various domains in a sample of 94 students. Not surprisingly,
common factor analyses of IQ and creativity tests yielded separate factors (see
Introduction 4

Amelang & Bartussek, 1997, p. 274). When Carroll (1993) re-analyzed 42 data sets that
included creativity tests for his Three Striatum Taxonomy of Intelligence, he found a
factor loaded highly by these test that “represents a distinct dimension of individual
differences that is linearly independent of other such dimension” (p. 427) and “can be
measured with considerable reliability” (p. 431). Consequently, not even the strongest
proponents of a g factor of general intelligence, from Spearman and the early British
School (Hargreaves, 1927) to Eysenck (1995a) and Jensen (1996), doubt the
differences between intelligence and creativity.
Yet a creative person’s IQ is practically inevitably at least a standard deviation above
the mean, often more (Cox, 1926; Haensly & Reynolds, 1989; Simonton, 1984, 1999a).
These apparently contradicting findings were integrated by Guilford (1967a, p. 168,
1967b, 1981) to a threshold hypothesis of creativity. The hypothesis stated that a
minimal level of IQ, often rather arbitrary set to 120, should be necessary, but not
sufficient for creativity. Recognized creative achievement was thought to be impossible
below this threshold (Simonton, 1994). Guilford proposed that the correlation between
IQ and creativity should be insignificant above an IQ of 120 - a rather week claim
because of the inevitable range restriction effect and the fact that all cognitive activities
show weaker correlations in high-IQ samples (Detterman & Daniel, 1989). Second, he
proposed that scatter plots of IQ and creativity should show a triangular pattern (which
gave Guilford’s claim sometimes the name ‘triangularity hypothesis’), with no data
points in the low-IQ / high-creativity quadrant. This threshold view of creativity is so
plausible that it is widely accepted, though empirical test are scarce and more likely to
show a disconfirming tendency (Mednick & Andrews, 1967; Runco & Albert, 1986;
Lubinski et al., 2001). A theoretical extension of Guilford’s hypothesis, suggested by
Peterson and colleagues (2002; Carson et al., 2003), will be discussed in section 1.4.
From a knowledge-centered cognitive perspective, Hayes (1989) proposed an
alternative ‘certification hypothesis’, which doubted intrinsic links between creativity and
intelligence. Instead, it stated that most possibilities to display a recognizable level of
creativity, like occupations in architecture or science, simply require a high level of
formal education. Since academic performance is correlated with IQ, society simply
denies creative individuals of low IQ the chance to express their talent adequately. It
must be noted that exactly the same idea was already discussed by Guilford (1967a,
pp. 168-169), who puzzles over cases of triangular IQ-creativity-relationships in
children, for whom the ‘certification hypothesis’ would not hold.
Introduction 5

Related to these issues is the question if creativity is stronger associated with fluid (gf)
or crystallized intelligence (gc or ag; Cattell, 1963, 1971) While the former is more
representative of reasoning ability and mental speed, the latter one is more influenced
by knowledge and learned response habits to complex cues and is often allocated in
tests of verbal ability. Cattell (1971) himself saw creativity, especially real-life creative
performance, as mainly determined by fluid intelligence, with the contribution of gc being
small compared it and to personality traits facilitating concentration and impulse
restriction in favor of inner activity (p. 443). He argued that “by its very definition and
nature, ag deals with things that are already known, and judgmental skills that have
already been applied before” (p. 435), leaving it in the mere role of “a requisite
foundation and contribution to the creative steps taken by virtue of gf” (p. 443)
According to him, it also bears the risk of hindering gf’s creative insight through negative
transfer (i.e. false application of existing knowledge in a new situation; p. 441). He
suggested any relation between creativity and gc being attributable to its substantial
empirical confoundation with gf. This view is challenged by studies of Crawford (1974;
Crawford & Nirmal, 1976), who found stronger relations of test creativity (measured by
the TTCT) to gc than to gf in multivariate analyses of data from 172 and 163 elementary
school children. Differences between gf and gc were, however, small, and divergent test
such as the TTCT are questionable creativity criteria (see below).
Beside intelligence, a second and rather independent cognitive component of creativity
is widely acknowledged (e.g. Guilford, 1967a, b; Amabile, 1983a, b; Sternberg & Lubart,
1995; Eysenck, 1993, 1995a; Jensen, 1996; Simonton, 1999a, b): A cognitive style of
divergent (Guilford, 1950) or associational thinking (Mednick, 1962), sometimes called
originality (Eysenck, 1993). Guilford’s approach to this cognitive style got much attention
after he included it as an own operation dimension (‘divergent production’) in his famous
Structure of Intelligence (SOI) model and operationalised it in form of divergent
(production) tests (Guilford, 1967a). These divergent test, of whom beside Guilford’s
(1967a) those of Torrance (1974) and Wallach & Kogan (1965) are well-known
examples, deviate from standard ‘convergent’ cognitive abilities tests in demanding not
one right answer, but as many different responses as possible, like various uses for an
object or titles for a short story. The responses are subsequently scored on objective
scales measuring assumingly different dimensions of creative ability, traditionally
fluency (number of responses), flexibility (number of different categories covered by the
Introduction 6

responses), originality (statistical infrequency of the responses) and elaboration

(amount of details given).
Tests of divergent production have often been criticized. While they do show a
considerable amount of convergent validity (e. g. Barron, 1968) and do often relate to
non-test indices of creativity (Barron & Harrington, 1981), their content validity has often
been questioned (e. g. Wallach, 1971; Cattell, 1971; Barron & Harrington, 1981; Brown,
1989). After presenting an extensive taxonomy and critique of creativity measures,
Hocevar & Bachelor (1989) suggested using divergent tests only as a measure of a
subcomponent of creativity, but warn against using them as the only criterion.
Additionally, the scoring dimensions of divergent tests have been shown to lack
discriminant validity and to depend exclusively on fluency (that is, all other dimensions
lack internal consistency and convergent validity when fluency is statistically controlled;
Hocevar, 1979a, b). Even though Hocevar proved this detrimental effect for objectively
determined originality (i.e. statistical infrequency) as well as for subjectively rated
originality, he (Hocevar, 1979a) and also Eysenck (1995a, p. 93) suggested subjective
originality ratings by several raters with controlling for number of responses as the
scoring method of choice. According to Hocevar (1979a), number of responses can be
controlled either by using item means as total scores, by requiring only a certain number
of responses per item in the instructions, or by scoring only a certain number of
responses per item. Note that this approach is very much in line with Amabile's (1982)
consensual definition of creativity (discussed in section 1.1).
In the literature, individual differences in a disposition towards an original cognitive style
have been closer linked to temperamental factors and personality traits than to pure
cognitive abilities. Therefore, in the next two sections, I will shortly review empirical
findings on and theoretical integrations of creativity and personality.

1.3 Creativity and personality

The first-ever meta-analysis in the field of creativity research was conducted by Feist
(1998) and dealt with the personality of creative people. More precisely, Feist analyzed
the results from 26 studies comparing scientists vs. non-scientists (total N = 4852), 28
studies comparing more creative vs. less creative scientists (total N = 3918) and 29
studies comparing artists vs. non-artists (total N = 4397) according personality traits that
Introduction 7

predicted creative achievement (all studies were published ones). As a reference model
for all the various traits differentiating between the groups, Feist used the Five-Factor
Model of personality (FFM; see John & Srivastava, 1999), which represents the
essence of a growing consensus on the taxonomic structure of personality. He
classified all traits according to their highest correlation with a FFM factor reported in the
literature to either the positive or the negative pole of Neuroticism (N), Extraversion (E),
Openness to Experience (O), Agreeableness (A) or Conscientiousness (C), as long as
the correlation exceeded .25. The factors of the FFM are constituted as broad
dimensions of personality and therefore inevitably blur more fine-grained personality
dimensions. Since the relationship between creativity and Extraversion is known to be
rather complex (and has even been called paradox (Martindale, 1993; see section 1.4)),
Feist used two interrelated subdimensions of Extraversion – confidence/dominance and
sociability - additionally to the global dimension of E. Whenever sufficient data was
accessible, Feist also used the scales of the California Personality Inventory (CPI;
Gough, 1957), the Sixteen Personality Factors Questionnaire (16PF; Cattell, Eber &
Tatsuoka, 1970) and the Eysenck Personality Questionnaire (EPQ; Eysenck & Eysenck,
1975) to compare the three pairs of creativity groups. Only the FFM and EPQ results
are relevant for the purpose of this thesis, and therefore the CPI and 16PF results will
not be discussed. The EPQ measures the three broad superfactors of Eysenck’s P-E-N
model, Psychoticism (P), Extraversion (E), and Neuroticism (N) (Eysenck & Eysenck,
1985).
The meta-analytic results of the FFM dimensions and somewhat associated traits
showed that what differentiated creative from less creative scientists and artists from
non-artists is high Openness to Experience (Cohen’s d = .31 and .47, respectively;
Cohen, 1977), low Conscientiousness (d = .30 and .75), high Extraversion (d =.39 and
.15), which is completely attributable to confidence/dominance (d = .40 and .21), but not
sociability (d = .00 and .02), and, to a less extend, low Agreeableness (d = .19 and .21).
Additionally, artists, but not creative scientists, showed a lower Emotional Stability (N-)
compared to their comparison group (d = -.24 and .09, respectively). On the other hand,
scientist in general were distinguished from non-scientist by a higher Introversion (E-; d
= .26) combined with a slightly higher confidence/dominance (d = .17), the latter
showing a decline with age, a lower Openness to Experience (d = .30) and a much
higher Conscientiousness (d = .51) – results that differed markedly from the other two
Introduction 8

comparisons. It seems like a scientific occupation imposes much less demand on

creativity than an artistic one.
The EPQ results, however, rendered the Extraversion result just reported inconsistent.
The congruence of FFM and P-E-N Extraversion has often been poven (see Angleitner
& Ostendorf, 2003). It is therefore surprising that EPQ results indicated scientist being
more extraverted than non-scientist (d = .33). Feist explained this result solely with the
importance to distinguish the confidence/dominance and sociability subdimensions of E.
Extraversion showed no noteworthy difference between artists and non-artists, neither
did Neuroticism for scientists or artists. Data for comparisons of highly vs. less creative
scientists were not available for the EPQ. Scientists and artists, however, markedly
differed from the normal population in Psychoticism (d = .45 and .66, respectively).
For all analyses together, Feist concluded: “Creative people are more autonomous,
introverted, open to new experiences, norm-doubting, self-confident, self-accepting,
driven, ambitious, dominant, hostile, and impulsive. Out of these, the largest effect sizes
are on openness, conscientiousness, self-acceptance, hostility, and impulsivity. Yet,
creative people in art and science do not completely share the same unique personality
profiles: Artists are distinguished more by their emotional instability, coldness, and their
rejecting group norms than are scientists. […] Finally, less creative scientists, compared
with the effect sizes of their more creative peers in science and in art, are more
conscientious, conventional, and closed-minded […].” (pp. 299-300). These conclusions
were very much in line with those of an earlier qualitative review by Barron and
Harrington (1981, p. 453): “In general, a fairly stable set of core characteristics (e.g.,
high valuation of esthetic qualities in experience, broad interests, attraction to
complexity, high energy, independence of judgment, autonomy, intuition, self-
confidence, ability to resolve antinomies or to accommodate apparently opposite of
conflicting traits in one’s self-concept, and finally a firm sense of self as “creative”)
continued to emerge as correlates of creative achievement and activity in many
domains.”
Feist (1998) also reviewed longitudinal studies informative about the temporal order of
personality and creativity, and concluded that there is no hint for any influence of
creative achievement on subsequent personality: “Every longitudinal study has found
that the same traits that distinguish creative people later in life also distinguish them
earlier in life. […] Taken in total, longitudinal studies of creative personality over time
Introduction 9

suggest that the personality structure of highly creative people tends to remain relatively
stable.” (Feist, 1998, p. 299; see also Feist & Barron, 2003).
At this point, a cautious remark has to be made. Most reviews of creativity and
personality, like those of Feist (1998) or Barron and Harrington (1981), focus on creative
achievement. It surely is convincing to see eminence or recognized creative products as
the purest criteria of creativity (Hocevar & Bachelor, 1989). But creative achievement in
a field, let alone eminence or genius, is without doubt synergistically determined by
more than just a disposition towards creativity (e.g. (Amabile, 1983a, b; Woodman &
Schoenfeldt, 1989; Eysenck, 1993, 1995a, b; Feist, 1998; Csikszentmihalyi, 1999;
Simonton, 1999b; Jensen, 1996; Sternberg & Lubart, 1995). Beside multiple external
factors facilitating or inhibiting a creative career (from education, socioeconomic status
and home environment to cultural, religious and historical factors to oddities such as
season of birth and activity of sunspots; Eysenck, 1995a), some internal factors may be
mere catalysts for bringing a creative product to public recognition. The hypothesized
function of IQ as a certification for creative occupations has already been discussed
(section 1.2). Other possible candidates are self-confidence and dominance, autonomy
and independence, ambition and intrinsic motivation, and even a certain degree of
hostility and anti-social tendency to defend ones ideas against premature criticism.
Some further internal factors may be specific for certain fields, such as emotionality in
the arts (Feist, 1998). As long as one conceptualizes the general disposition towards a
creative cognitive style as a dimensional trait owned by everyone to a greater or lesser
extend, studying everyday creativity in general population samples should lead to less
confounded, yet valid estimates (Richards, Kinney, Benet & Merzel, 1988).
Of those studies using less professional subjects, McCrae's (1987) is maybe the most
recognized one. In a sample of over 200 adult men, he studied relations between
creativity and the Five-Factor Model, intensively measured by different versions of NEO
questionnaires (Costa & McCrae, 1992) and an adjective list, all in self-, peer-, and
spouse-report form, as well as a Q-Sort. Further personality measures included EPQ
Psychoticism. The creativity criteria were five of Guilford’s divergent tests, four of which
were only scored for fluency, while the fifth (remote consequences) was additionally
rated for originality. Since all tests loaded on the same factor, a total sum score was
calculated. As a second criterion, Gough's (1979) Creative Personality Scale (CPS), a
well-validated, empirically keyed, 30-item adjective scale, was administered. All six
measures of Openness to Experience correlated highly significant (between .29 and
Introduction 10

.41) with the divergent tests total score, and four remained significant when verbal
intelligence, age and years of education (correlated .53, -.19 and .52 with total divergent
tests score, respectively) were partialled out. The CPS did also correlate significantly
with all Openness measures (between .26 and .61), as well as .26 with the divergent
tests total. No other of the FFM factors, neither Psychoticism, showed a consistent
relation to the divergent tests. Results of these dimensions were also inconsistent for
the CPS, except for Extraversion, which correlated between .16 and .58 with the CPS
(five of six coefficients being significant). Note, however, that according to Hocevar and
Bachelor (1989), both creativity criteria did not fully qualify as such, since both
measured only an aspect of creativity.
While Rawlings and colleagues (1998), also using divergent tests as criteria, were able
to replicate McCrae’s result of a positive correlation with Openness, the results of
Woody and Claridge (1977) contradict those of McCrae in finding a strong association
of five Wallach and Kogan (1965) divergent tests with EPQ Psychoticism self-reports
(.32 to .45 for fluency, .61 to .68 for originality, both dimensions objectively scored) in
100 British university students. However, replications of Stayte (1977) and Rawlings
(1985) yielded much lower and only partly significant positive correlations between
divergent tests and P. In 54 Australian adults, Wuthrich and Bates (2001) found no
relation between two divergent tests (Wallach and Kogan’s (1965) ‘pattern meaning’
and Torrance’s (1974) ‘unusual uses’) and the revised P scale (Eysenck, Eysenck &
Barrett, 1985), but also failed to find relations with NEO-PI-R E, O or A, though the
‘unusal uses’ test correlated positively in the .30ies with N and C (all self-reports). An
unpublished study by Anton, Griepenstroh, Poggenpohl & Rothenpieler (2003) surveyed
78 German adults and found different results for NEO-PI-R self- and peer-reports in
relation to the twelve divergent tests of the BIS-4 (Jäger, Süß & Beauducel, 1997; see
section 2.1.2): Only N correlated low with a total score, but with negative sign.
Other studies did not or not solely rely on divergent tests when exploring the relations
between creativity and these personality dimensions in non-professional samples.
Dollinger and Clancy (1993) gave 257 American psychology undergraduates the task to
create within the next semester a booklet of twelve photos that should describe their
identity. These booklets were rated by four independent raters on a 5-point overall
“richness” scale, which was explicitly anchored to include creativity, artistic sensitivity
and aesthetic value. Of all self-report NEO-PI scales, only O was a significant predictor
of the composite rating (β = .31). Wolfradt and Pretz (2001) found in a sample of 204
Introduction 11

German students that, out of the FFM dimensions (measured by NEO-FFI (Costa &
McCrae, 1992) self-reports), E and O predicted CPS scores, but only O predicted
creativity ratings (sensu Amabile's (1982) Consensual Assessment Technique) of a list
of hobbies and a short story written to a picture. In a study by King, McKee Walker and
Broyles (1996), 75 American psychology students provided a list of creative
accomplishments over a two-year period and completed the verbal scale of the TTCT.
The accomplishments were counted and rated for creativity by two raters. The FFM was
measured by Big Five Inventory (BFI; John, Donahue & Kentle, 1991) self-reports.
While E was correlated .26 with TTCT verbal creativity and A was negatively correlated
(-.23) with the reported creative accomplishments, O was the only scale related to both
creativity measures (.38 and .47, respectively). McCrae (1993/94) reported a correlation
of .53 between Openness to Experience and artistic interests. A study of Martindale and
Dailey (1996), however, failed to find an association between O and creativity ratings of
a story written to a given topic, the Alternate Uses Test (scored for fluency) and the
remoteness of associations given to a list of 100 words in a small sample of 37 male
American psychology students. Of the remaining EPQ and NEO-PI scales, only E was
significantly related to a composite creativity score. Aguilar-Alonso (1996) let 400 adults
fill out the Spanish version of the EPQ, construct a crossword puzzle in a 5 x 5 scheme
(scored with one point for every letter belonging to two words) and complete four
incomplete drawings (objectively scored for fluency, flexibility and originality). When
participants were divided into extreme groups for each of the three EPQ scale (using
mean splits), no groups differed significantly on crossword puzzle constructing
performance or drawing completion flexibility and originality, but both the group high on
E and the group low on P were each significantly more fluent on the drawing completion
task. Finally, Soldz and Vaillant (1999) showed that, when professional ratings of 132
male Harvard sophomores on 25 traits were classified to fit the dimensions of the FFM,
the O factor extracted from these ratings was the only one that predicted a rating of total
creative accomplishments (based on a detailed record of achievements, interests and
hobbies) in the subsequent 45-year period (r = .40). Additionally, only NEO-PI
Openness, assessed via self-reports in late adulthood, related significantly (.27) to this
creativity rating.
Taken together, the results reviewed in this chapter strongly suggest Openness to
Experience as the most likely correlate of a disposition towards creativity. Other
candidates, though with less clear empirical support, include Extraversion and
Introduction 12

Psychoticism. The other dimension of the Five Factor Model and Eysenck’s P-E-N
model show no consistent relation. It is striking that the clear majority of these studies
measured personality solely with self-report questionnaires. Additionally, all studies
reported here were rather atheoretical and only correlational, therefore not able to
provide causal explanations. In the next chapter, I will summarize some attempts to
integrate what is known about creativity and personality into models and theories.

1.4 Models and theories of a disposition towards creativity

Arguably the most ambitious attempt to integrate diverse findings into a theory of
creativity has been conducted by Eysenck (1993, 1995a, b). The central statement of
Eysenck’s theory is that the genetic, biological and cognitive underpinnings of a
normally distributed disposition towards a creative cognitive style (‘originality’) are the
same that underlie Psychoticism. More precisely, Eysenck argues that genes increasing
dopamine level and/or sensitivity in the frontal lobe as well as in the mesolimbic
dopaminergic system, and genes that decrease serotonin level and/or sensitivity in the
serotonergic mesolimbic system will, partly mediated by hippocampal formation
activities, lower latent inhibition (Weiner, 1990). Latent inhibition is a preconscious
gating mechanism that keeps stimuli previously experienced as irrelevant from entering
the focus of attention (Lubow, 1989). A similar phenomenon from cognitive research is
negative priming, resulting from experimental paradigms that show higher thresholds for
associations with stimuli which previously have been used as distractors. Negative
priming is one measure of cognitive inhibition, a state of high cortical activity
(sometimes called arousal or drive). Cortical activity is known to be primarily inhibitory.
In line with Hull’s (1943) “behavioral law”, such a state makes dominant stimulus-
response-connections even more dominant, leading to reduced behavioral flexibility.
Eysenck proposes that a condition of low expression in these three factors (assumingly
habitual in individuals with corresponding genetic make-up) leads to an overinclusive
cognitive style, marked by a broad, defocused attention, primary process thinking, and
flat association hierarchies. Such a cognitive style is not only common in schizophrenic
disorders (one extreme pole of the Psychoticism dimension), but shall also be
supportive for creative thinking processes.
Introduction 13

To understand why the latter shall be the case, we have to take an excursion to the
creative process theory on which Eysenck’s approach is build. It is Donald Campbell's
(1960) theory of blind variation and selected retention, later extended by Simonton
(1988) to the Chance-Configuration Theory, and it is on the way to become the
predominant theory of the creative process (Simonton, 1999b, c; Cziko, 1995).
Campbell’s theory, which is also the fundament of evolutionary epistemology in
philosophy (Campbell, 1974), is a secondary Darwinian theory, i.e. a theory using
Darwin’s (1859) theory of biological evolution by natural selection metaphorically.
Simply stated, it claims that any creative insight stems from a nonteleological, blind
mental combinatory process, which generates chance configurations of mental
elements, and whose results are selectively kept and elaborated. Plenty of empirical
support from experimental, psychometric and historiometric sources exists
(accompanied by a long list of introspective reports from creative genii), most of which
are not easily explained by any other theory of creativity. For example, cognitive
computer simulations of human creativity as well as artificial intelligence systems that
lack a random element are unable to do better than reproducing already known
inventions from data limited to the relevant elements. The only objective psychometric
measures that relate to creative processes rely on such factors as ideational fluency
and remoteness of associations, i.e. manifold and blindness of ideas. And in any field,
the creative quality of products is a mere function of quantity, with no sign of a greater
probability of success for more experienced producers. For far more detailed reviews
and discussions, see Simonton (1998, 1999b, c, d), Cziko (1998), and Eysenck (1995a).
However, while the ‘retention’ part of the theory is rather uncontroversial – such an
evaluative step of verification or elaboration was already included in early models of the
creative process (Helmholtz, 1896; Wallas, 1926) – the central statement of random
variation and mental trial-and-error being fundamental kindled much discussion (e.g.
Jensen, 1996; Sternberg, 1998). Most often, criticism roots in two sources, namely (1) a
deep preoccupation of the critics against moving creativity out of the scope of intention
and volition, and (2) fundamental misunderstandings what is meant by ‘blind’ or
‘chance’. According the first point, Simonton (1999c, d) convincingly argues that having
the goal to be creative does not imply having the ability. Even the greatest genii were
unable to will producing creative products, a fact that has many historical examples. The
second criticism has caused several changes in the naming of the central process (in
his most recent publication, Simonton (2003) talks about “constrained stochastic
Introduction 14

behavior”). The theory should not be caricaturized as stating that the creative process
resembles a monkey writing a Shakespeare play by means of randomly typing on a
typewriter (as done by Jensen, 1996). Just like biological evolution, secondary
Darwinian processes do not create “hopeful monsters” (resulting from radical and
coordinated functional variation), but, step by step, build incrementally on what is
already there (Dawkins, 1986), i.e. biologically evolved information-processing
adaptations and achieved expertise in the creativity case. Darwinian creativity should
also not be interpreted as demanding precise equiprobability for all possible events,
which is also not true for biological evolution, where e.g. genetic linkage constraints
genetic recombination (Simonton, 1999c). At this point, it is important to emphasize that
a Darwinian conceptualization of the creative process does by no means deny individual
differences in creativity, nor does it reduce them to differences in expertise and
motivation: Any individual difference that alters the amount of quasi-random variation in
mental combinatory processes is likely to alter the probability of creative production.
This is best described by Austin (1978), who distinguishes four types of chance: Only
the first represents “blind luck”, while the other three favor those who are “in motion”
(motivation), “prepared” (intelligence and expertise) and “act distinctively” (cognitive
style).
Here is the point where Eysenck introduced overinclusive thinking. First, the mere
amount of internal and external information entering the attentional focus of an
individual with low latent inhibition and negative priming provides masses of rather
random input for the creative process. Second, the defocused, associative cognitive
state resulting from low cognitive inhibition allows for a large number of simultaneously
activated mental representations (Martindale, 1989, 1999), and therefore increases the
number of potential combinations. It is noteworthy, however, that creative people do not
show a constant pattern of cognitive arousal (which would locate them on Eysenck’s
Extraversion dimension). Instead, they show cortical disinhibition only in creative
achievement situations, while they appear more inhibited and rather oversensitive in
other situations (Martindale, 1989). Eysenck saw such a variable pattern of arousal as a
feature of Psychoticism, not Extraversion (a view that was gratefully accepted by
Martindale (1993), who noted that he was always unsatisfied with the “paradoxical”
relations of creativity and Extraversion in his own studies).
Distilling a creative idea from a blind combinatory process is of course facilitated by
processing capacity, mental speed and reasoning ability, all attributable to general
Introduction 15

intelligence. Eysenck worships this in conceptualizing creative achievement as a

synergistic interaction of several variables, including P and g. Jensen's (1996) review of
genius and giftedness, owing much to Eysenck’s work, simplifies matters in stating
genius being a product of high ability times high productivity times high creativity, and
proposing g as the underpinning of ability, P as the underpinning of creativity, and
mental energy, a still unclear construct borrowed from Galton (1869), as the
underpinning of productivity.
Compelling in its theoretical richness and integration of diverse findings, Eysenck’s
theory is less convincing on empirical grounds. Especially, studies have found a
stronger connection of Psychoticism with creative achievement than with a disposition
towards creativity, the latter being unclear at best (see section 1.3). Even worse, while
some studies confirmed the association of high Psychoticism and low latent inhibition
(Baruch, Hemsley & Gray, 1988; Lubow et al., 1992; De la Casa, Ruiz & Lubow, 1993),
Peterson and colleagues (Peterson, Smith & Carson, 2000; Peterson & Carson, 2002)
found a much stronger association of latent inhibition with Openness to Experience and
(less consistently) Extraversion than with Psychoticism (based on NEO-FFI and EPQ
self-reports from student samples). Additionally, they found relations of O, E, and latent
inhibition with the CPS. Interestingly, a composite score of O and E was the best
predictor of the CPS and latent inhibition. The authors propose an underlying,
dopamine-driven exploratory tendency (mediated by a low latent inhibition), with E as its
behavioral and O as its abstract-intellectual manifestation (Peterson & Carson, 2002). A
study by Wuthrich and Bates (2001), however, contradicted these results by failing to
find any relation between latent inhibition and priming on the one hand and self-reports
on the NEO-PI-R and the revised Psychoticism scales as well as two divergent tests on
the other hand.
Even though the relation between Psychoticism and creativity remains unclear, there is
support for the basic mechanism predicted by Eysenck: Carson, Peterson & Higgins
(2003) recently found that Harvard students low in latent inhibition had higher originality
scores in four of Torrance’s divergent tests (d = .65), higher CPS scores (d = 1.06), and
reported more creative accomplishments (effect size in a meta-analysis of two studies: r
= .31). Comparing 25 distinguished creative achievers with 23 students low in creative
achievement showed latent inhibition being able to discriminate between them (d = .93).
In this sample, a multiple regression of latent inhibition and IQ on creative achievements
yielded 30% explained variance and a highly significant interaction of both. The
Introduction 16

interaction indicated high IQ combined with low latent inhibition being most favorable for
creative achievement. According to the authors, these results give a rational to
Guilford’s threshold hypothesis. Peterson and colleagues (Peterson & Carson, 2002, p.
1145) concluded: “This would make the individual predisposed to schizophrenia
suffering, in principle, from the pathological and possibly synergistic combination of
excess experiential, ideational or associational variability, and a decrement in methods
of selecting from that excess, while the healthy, open and creative individual would be
characterized by a broader gate and careful post-experience selection and culling. So,
we have an operationalized quasi-Darwinian approach to the problem of psychosis and
creativity (as originally suggested by Campbell (1960) and Simonton (1999[c])).”
Peterson and colleagues’ results are highly supportive for major parts of Eysenck’s
theory, but they strongly suggest a modification in its most central claim: An exchange
of P with the common core of O and E. Indeed, out of the lexicographical Big Five
personality factors, which are the emerging consensus on a taxonomy of personality
descriptive terms across languages (Goldberg, 1990; Ostendorf, 1994), the factor
corresponding to O, factor V, is most often interpreted as including creativity as an
important aspect (Goldberg, 1994; Ostendorf & Angleitner, 1994), and is sometimes
even labeled “Creativity” (Johnson, 1994). Very much in line with the empirical evidence
reviewed so far, Ostendorf & Angleitner (1994) suggested from results of an Abridged
Big Five Circumplex (AB5C) analysis based on German disposition adjectives that the
factor with high positive loadings of factor V and I (Extraversion) markers (V+I+) should
be labeled “Creativity” (or “Surgent Mentality”). While this underlines the allocation of
creativity in descriptive personality taxonomies, such a phenotypical description lacks
explanatory value, since it would be a tautological explanation (Martindale, 1989).
Factor V is, however, not identical with Openness to Experience in the NEO-Five Factor
Model (Goldberg, 1994). Unlike the descriptive Big Five taxonomy, the FFM is a
comprehensive taxonomy of personality traits, which can very well be assumed as
causal factors. The Five Factor Theory proposed by Costa and McCrae conceptualizes
the FFM dimensions as representing fundamental, temperamental, and highly genetic
sources of individual differences in personality (McCrae et al., 2000). Such a
perspective, while controversial, is partly supported by empirical evidence on a
biological link between O/E and latent inhibition, which itself is likely to be influenced by
genes moderating serotonin and dopamine level.
Introduction 17

Overall, theoretical approaches converge with empirical findings in suggesting general

intelligence, Openness to Experience, and, less clearly, Extraversion and Psychoticism
as determinants of a disposition towards creativity. All these predictors show moderate
to substantial heritabilities (Plomin, DeFries, McClearn & Rutter, 1997; Heath, Eaves &
Martin, 1985), so the reasonable next step, undertaken in the next section, will be to ask
the same question for creativity itself.

1.5 The etiology of creativity

Inquiries into the heritability of creativity in its highest form – eminence and genius – are
even older than the most common behavioral genetic method, the study of twins. In fact,
Sir Francis Galton published his famous “Hereditary Genius” in 1869, 13 years before
he conducted what is acknowledged as the first twin study. Galton emphasized that
genius runs in families. A re-examination of Galton’s data conducted by Bramwell
(1948), however, came to the conclusion that this was only the case for judges, but not
for any profession normally viewed as creative. Bullough, Boulough and Mauro (1981)
concluded in their literature review that creative achievement is rarely inherited for more
than one generation. Overall, there is no evidence indicating that creative achievement
aggregates within families.
So far, modern behavioral genetic studies of creativity have focused almost exclusively
on divergent tests. Additionally, I was unable to identify a single adoption study. Instead,
all genetically informative studies on creativity, including the one reported in this thesis,
are twin studies. The behavioral genetic twin method decomposes the variance of
measured characteristics into shares influenced by genetic (A), shared environmental
(C) and unshared environmental (E) sources. This is done on the rational that identical
or monozygotic (MZ) twins share an identical genome, while fraternal or dizygotic (DZ)
twins share only 50% of their genes on average. Under the assumptions that (1)
differences in environmental similarity do not affect differences in pair resemblance of
MZ and DZ twins and (2) all three variance sources contribute additively to the
observable, phenotypic variance, both of which are well supported (Plomin et al., 1997),
the proportion of phenotypic variance attributable to genetic effects (heritability) can be
calculated from twin pair intraclass correlations (ICCs) as: h² = a² = 2 * (ICCMZ – ICCDZ)
(Falconer formula; Falconer & Mackay, 1996). Additionally, environmental effects can
Introduction 18

be estimated as: c² = 2 * ICCDZ – ICCMZ and: e² = 1 – ICCMZ, the latter including error of
measurement as long as reliabilities of measures are unavailable (Plomin et al., 1997).
Nichols (1978) reported mean ICCs, weighted for sample size, of .61 for MZ and .50 for
DZ twins in a summary of ten studies on divergent test performance published before
1971. They indicate modest genetic influences of 22%, 39% shared environmental
influences and 39% non-shared environmental and error influences on the observed
variance. Reznikoff, Domino, Bridges and Honeyman (1973) administered a battery of
ten creative ability tests to a sample of 63 MZ and 54 DZ adolescent twin pairs. The
battery was rather diverse, consisting of five of Guilford’s classic divergent tests, a test
requiring remote associations to word triplets, a figure preference test, and three test in
which responses were rated for originality, including a drawing completion test (the
Franck Drawing Completion Test; Anastasi & Schaefer, 1971) similar to the one used in
the studies reported in this thesis. Tests were all scored by the same single person.
Variance-analytically derived heritability indices yielded a mean heritability index of .14
for the Guilford tests and of .56 for the remote associations. Two of the three tests rated
for originality, including the drawing completion test, showed heritability indices below
zero (the third, ‘Similies’, of .39), as did the figure preference test. Calculating variance
components from the reported ICCs, the five Guilford tests showed a mean heritability
of 41%, a mean shared environmental influence of 38%, and 37% non-shared
environmental and error variance. Similarly, estimates were a² = .70, c² = .08 and e² =
.22 for the remote associations tests and a² = .68, c² = .00 and e² = .41 for the ‘Similies’
test. Estimates were not meaningfully calculable for the two tests where ICCDZ were
greater than ICCMZ. Grigorenko, LaBude and Carter (1992) tested 60 MZ and 63 DZ
adolescent twin pairs from the former Soviet Union with the Russian version of the
verbal TTCT scale. ICCs were .86 for MZ and .64 for DZ pairs, yielding estimates of
44% genetic effects, 42% shared environmental effects and 14% non-shared
environmental effects plus measurement error. The authors speculated about a
relationship between the somewhat higher heritability in their studies compared with
Nichols review and expectations of conformity in the communistic system of the Soviet
Union, since conformity reduces environmental variances and therefore highlights
genetic differences.
In sum, the heritability of performance in classic divergent tests seems to be moderate
at best. When the results from the two more recent studies were added with unit
weights to the ten studies summarized by Nichols (1978), a rough estimate of 25%
Introduction 19

genetic, 38% shared environmental, and 37% non-shared environmental and error
influences on classic divergent test variance emerges. It must be noted, however, that
divergent test (TTCT) performance within 56 married couples was correlated .33 in a
study by Wallinga & Crase (1983). Creativity and verbal fluency are also discussed as
characteristics preferred in mate choice (Miller, 2000). Heritability estimates for such
tests might therefore be conservative, since they are attenuated not only by unreliability
of measurement, but also by assortative mating (which increases DZ, but not MZ twin
pair similarity; Plomin et al., 1997). For the same reason, shared environmental effects
might be inflated. Still, heritabilities for divergent tests remain lower than those normally
found for personality traits, and markedly lower than those found for cognitive abilities.
Shared environmental influences, on the other hand, seem to be strong for a cognitive
ability measure (Plomin et al., 1997). Results from a twin study by Canter (1973)
suggest that the genetic influences on divergent thinking are completely attributable to
those of general intelligence: Within-pair resemblance in divergent test performance
was very similar for MZ and DZ twins after general intelligence was statistically
controlled.
To my knowledge, the only twin study of creativity not based on creative ability tests
was published by Waller, Bouchard, Lykken, Tellegen and Blacker (1993). They
extracted a general factor from the CPS and calculated ICCs for factor scores of 45 MZ
pairs, a MZ triplet, and 32 DZ pairs, all reared apart and unified in adulthood. While the
MZ twins showed a moderate resemblance of .54, indicating a heritability of equal
amount, the DZ pair correlation was practically zero (-.06). Waller and colleagues
concluded from this pattern and the contradictory findings of evidence for a low
heritability of divergent tests on the one hand and no evidence for family aggregation of
creative achievement on the other that creativity is an emergenic phenomenon.
Emergenesis (Lykken 1982; Lykken, McGue, Tellegen & Bouchard, 1992) describes the
inheritance of complex higher-order traits, which are synergistically determined by an
interaction of multiple, more fundamental, eventually partly heritable traits. Emergenic
traits will not run in families, since their components are independently inherited and will
be torn apart by sexual recombination. Emergenesis differs from epistasis in that it
encompasses the configuration of molar, partly heritable composite traits, not just
interactions of alleles at different genetic loci in polygenetic traits. Thus, emergenic traits
are heritable (influenced by genes), but it is unlikely that relatives who do not share the
complete genetic make-up (i.e. all but MZ twin siblings) show much resemblance. This
Introduction 20

is exactly the pattern Lykken and colleagues suggest as indicative of emergenesis in

twin studies: Non-negligible MZ and negligible DZ intra-pair resemblance. “What we
suggest now is that perhaps many complex human psychological traits (e.g., many of
the idiosyncrasies of personal style that we have observed in twins), as well as traits like
extraversion; the ability to stop reacting to a meaningless repeating stimulus
(habituation); more familiar traits like leadership, artistic ability, selling ability, teaching
ability, creativity, parenting ability – and also many examples of human genius – are
emergenic traits. Because these attributes do not tend to run in families, the possibility
of their having a genetic basis has been overlooked. As we have seen, we will not
discover the emergenic character of such traits unless we study twins. Until this idea is
more broadly accepted, we will not believe what we discover unless we study twins
reared apart.” (Lykken et al., 1992, p. 1569; italics original, underlinings added).
An emergenic view of the etiology of creativity seems to be the emerging consensus in
the literature (e.g. Martindale, 1999; Simonton, 1999b). While surely appealing for
creative achievement and genius, where synergistic interactions of determinants are
already included in theories, such a perspective is rather odd for a more fundamental
disposition towards creativity. This is especially the case because, as stated at the end
of the last section, the theoretically and empirically identified determinants of such a
disposition are moderately to highly heritable. Thus, an emergenic inheritance pattern
would only emerge if indeed several of these determinants are necessary conditions for
a disposition towards creativity. This conclusion would be premature given the empirical
evidence available to date, since only components of creativity have been studied with
behavioral genetic methods so far. There exists no such study of molar creativity.
Multivariate behavioral genetic analyses of the links between dispositional creativity and
its determinants would also be helpful, but are completely absent from the literature.
Though a need for more research is expressed in many areas, it seems truly urgent
here.
Introduction 21

1.6 Aims and hypotheses of the present study

1) Consensus on a common person’s dispositional creativity

Following the consensual definition of creativity (Amabile, 1982), it is hypothesized that

independent raters will agree on a person’s creativity. Assuming a general, dimensional,
and relatively stable disposition towards creativity, this should be the case for people
from the normal population, who are not necessarily distinguished creative achievers in
any field. It should also hold across measurement times, kinds and amount of
information about the person, and degrees of acquaintanceship.

2) Dispositional creativity and intelligence

Based on previous studies, but contrary to the certification hypothesis, it is assumed

that a disposition towards creativity will show a low to moderate, but significant relation
to general intelligence. Following Cattell (1971), a similar relation is expected to the
facets of fluid and crystallized intelligence, with that to fluid intelligence being stronger
than that to crystallized intelligence, since fluid intelligence should explain most of the
link between crystallized intelligence and creativity. Finally, contrary to the weakly
proven threshold hypothesis, the relation of general intelligence and dispositional
creativity is expected to be linear, with similar correlations of both variables across the
whole intelligence spectrum.

3) Dispositional creativity and personality

Out of the broad personality dimensions of the Five Factor Model and the P-E-N Model,
the literature suggests strong relations with Openness to Experience and weaker
relations with Extraversion for a disposition towards creativity. Research on latent
inhibition suggests a common core of both as especially important, a possibility that will
be explored. An alternative hypothesis would be a strong relation of dispositional
creativity with Psychoticism. Both will be compared in the main study. In any case,
personality dimensions should be incremental to general intelligence in the prediction of
dispositional creativity. To test configurational theories of creativity, the predictive value
of personality x general intelligence interactions will be tested for Openness to
Experience, Extraversion, an Openness-Extraversion composite, and Psychoticism1.
1
: Eysenck (1995b) did only predict synergistic effects of Psychoticism and intelligence on creative
achievement. The conceptualisation of dispositional creativity in this thesis comes of course closer to
what Eysenck called trait creativity or originality. Therefore, only a direct effect of Psychoticism would
be expected. However, if (high) Psychoticism is taken as a possible proxy for (low) latent inhibition,
then the results of Carson et al. (2003) would suggest the possibility of a synergistic effect.
Introduction 22

4) Necessity of personality and intelligence factors for dispositional creativity

Since the personality and ability dimension that will be related to dispositional creativity
in this study are assumed to be more biologically basic and are partly suggested to be
causal in theoretical approaches, it is hypothesized that statistically controlling the
variance of more fundamental personality dimensions and general intelligence will
markedly decrease the relationship between different measures of dispositional
creativity and dissolve their structure.

5) Genetic and environmental influences on dispositional creativity

The amount genetic, shared environmental and non-shared environmental effects on a

disposition towards creativity will be estimated. Previous behavioral genetic studies
focused on single components of creativity and yielded different results for different
components. This is the first behavioral genetic study of a molar creativity measure,
perceived dispositional creativity.

6) Genetic and environmental links between dispositional creativity and its

predictors

To date, not a single study has applied multivariate behavioral genetic designs to
explore the overlap of genetic and environmental effects on creativity and its assumed
determinants or the mediation of observed phenotypical correlations between creativity
and related constructs by genetic and environmental links. This gap will also be filled
with the present thesis.

Before all of these questions and aims will be tackled in the main study, a pilot study is
reported, which was conducted to explore and validate one of the measures used in the
main study.
Pilot study: The T-88 as a measure of creativity 23

2. Pilot study: The T-88 as a measure of creativity

Aim of the pilot study was to test the reliability of the T-88 (Cattell & Warburton, 1967), a
drawing completion test which is also used in the main study, as a measure of creativity,
and to compare it with classic divergent tests. This was important, since (a) the T-88
was not constructed as a measure of divergent thinking or creativity, and (b) it was not
administered in a standardized situation, but self-administered by the participants at
home in both studies. It is hypothesized that (1) the T-88, when administered this way,
can still be reliably scored by different raters for originality and elaboration, even when
rated rather subjectively, (2) both scores are related, but not identical (i.e. raters are
able to discriminate the dimensions), and (3) both scores will show modest, but
significant relations to classic divergent tests, administered in a standard test situation
and objectively scored for fluency and flexibility. It is important to remark that this pilot
study can only be a very weak attempt to validate the T-88 as a measure of true
creative ability, since the use of divergent tests as such a criterion has been strongly
criticized (Brown, 1989; Hocevar & Bachelor, 1989).

2.1 Method

2.1.1 Sample

As part of an empirical practical, four students² of the University of Bielefeld recruited 78

subjects (48 females, 28 males, two missing data) as voluntary participants for a more
comprehensive study (see Anton, Griepenstroh, Poggenpohl & Rothenpieler, 2003),
which included this pilot study. Eight subjects where later excluded because of missing
data, leading to a final sample of 70 subjects for the pilot study (44 females, 25 males,
one missing data). Age of subjects ranged from 18 to 58 years (M = 27.1, SD = 9.4,
median = 25.0, one missing data). The majority of subjects (65.4 %) were students from
various faculties (mostly psychology), the others were friends or relatives of the
students and came from various backgrounds. None of the subjects were paid for
participation, but all were offered a personality profile based on a measure not further
discussed in this pilot study. Additionally, students of psychology received certificates of
their participation, which they required for their Diploma.
²: I would like to thank Friederike Anton, Julia Griepenstroh, Henrike Poggenpohl and Sandra
Rothenpieler for collecting the data of this pilot study.
Pilot study: The T-88 as a measure of creativity 24

2.1.2 Materials

T-88: The T-88, here administered in its German version (Häcker, Schmidt,
Schwenkmezger & Utz, 1975), is a subtest of the Objective Test Battery (Cattell &
Warburton, 1967) and was originally constructed as a measure of Cattell’s exvia factor.
It consists of eighteen incomplete line drawings (see figure 1 for examples of items and
responses and the appendix (section 6.1) for the whole test). The subjects are
instructed to complete and name the drawings in any way they like within a four minutes
time limit.

Figure 1:
Exemplary Items and Responses of the T-88

The T-88 shows remarkable resemblance to the Franck Drawing Completion Test
(FDCT; Franck & Rosen, 1949), a test consisting of 36 line drawings very similar to
those of the T-88. It was originally constructed as a projective measure of a person’s
sex adjustment or concern over his or her sex role. Barron (1958), however, used
ratings of FDCT responses as a measure of “creative originality” and Yamamoto (1964)
used six FDCT drawings (and their titles) as a test of divergent production sensu
Guilford (1967a). Torrance later developed a subtest of the TTCT (Torrance, 1966)
based on this work. The first twelve FDCT drawings were administered to 800 high
school students by Anastasi and Schaefer (1971; Schaefer, 1970). They showed that
rating the responses for originality, elaboration and asymmetry, but not for abstraction
or scoring them objectively for the number of different content categories used
(flexibility), differentiated creative from non-creative students (qualifying as such in both
Pilot study: The T-88 as a measure of creativity 25

teacher nominations and Guilford’s Alternate Uses and Consequences tests) in both
boys and girls and in both arts and science students.
The present study adapted Anastasi and Schaefer’s scale format and scoring
instructions (see also Schaefer, 1969). The T-88 responses were rated itemwise (with
item order randomized) by three independent raters (two female, one male) according
originality (5-point Likert-type scale ranging from 1 (very poor) to 5 (very clever or
unusual idea)) and elaboration (3-point Likert-type scale ranging from 0 (not identifiable
or arbitrary) to 2 (identifiable and detailed or decorated)). Rater instructions included
definitions of originality and elaboration, detailed anchors for each scale point of both
scales, sensibilisations to rater biases, and emphasizes to evaluate form as well as
content and discriminate elaboration from drawing ability (see the appendix (section 6.2)
for the original German instructions). Additionally, raters were provided with a list of the
most common responses for each item, including frequencies in percentages (see
section 6.3). These lists were based on frequency counts of response content for each
item in 278 randomly selected test sheets from the main study sample and included all
response contents mentioned in more than five percent of the cases (following Vernon,
1971, p. 252). The scale definitions and response baselines were thought solely as an
orientation for the raters, who were explicitly advised to use them that way, i.e. to
maintain their subjective impression in their ratings. Note that this scoring procedures
differs from those most widely used for divergent tests (e.g. by Guilford (1967a) or
Torrance (1974)), who emphasize reduction of subjectivity by means of objectively
counting response frequencies (originality) or included features (elaboration).
The present scoring of the T-88 further deviates from that of most classical divergent
tests by excluding ideational fluency (number of responses given). While Yamamoto
(1964) and Torrance (1966) scored their FDCT adaptations for fluency, Torrance
already mentions the limited usefulness of such a score in this test (Torrance, 1966, pp.
14-15), leading Anastasi and Schaefer (1971) to drop fluency. But since they still
calculated total scores for the scoring dimensions used in their study by summing over
item scores, they included an indirect fluency effect, because a greater number of
responses easily inflates such sum scores. The present study went further and, in line
with Hocevar’s (1979a) suggestion (discussed in section 1.2) controlled for fluency by
averaging item scores for originality and elaboration. Subjects were not excluded for
failing to complete a critical amount of items, since (a) such a procedure is likely to be
Pilot study: The T-88 as a measure of creativity 26

an overcontrolling of fluency and might induce range restriction, and (b) no subject
completed less than one third and only one completed less than half of the items.

BIS-4: The ‘Berliner Intelligenzstruktur-Test’ (Berlin Intelligence Structure Test, BIS-4;

Jäger, Süß & Beauducel, 1997) is a multifactorial German intelligence test based on the
well-acknowledged Berlin Intelligence Structure Model of Jäger (1982, 1984). This
model is hierarchical, because it includes the g factor of intelligence (Spearman, 1904)
on its highest level as well as a subordinated bimodal classification of abilities into three
content (figural, verbal, numerical) and four operation (processing speed, remembering
ability, imaginativeness, processing capacity) aspects. The aspects are assumed to
underlie all intellectual abilities, but with markedly different weightings. The BIS-4
assesses all twelve ability aspects in this 3 x 4 matrix with a total of 45 different
subtests (three to five per cell) and allows calculating a g value, too. In this pilot study,
only the twelve subtests of the imaginativeness scale (abbreviated E for German
“Einfallsreichtum”) in all three content aspects are used. These subtests resemble
classic divergent tests, but were chosen by the test authors to require problem-oriented
ideational flexibility (i.e. imaginativeness) and to be easily and objectively scoreable.
They are therefore scored for number of mentioned categories (given in the manual)
where possible (flexibility, five tests), and for number of correct responses (fluency)
where this has not been objectively possible so far (seven tests). Because more
complex and more subjective to score, originality and elaboration measures were not
included in the E-scale. This was the rational to call it a measure of imaginativeness
instead of creativity (Jäger et al., 1997, p. 32). To provide a more differentiated picture
for analyses on subtest level, all subtests scored for flexibility were additionally scored
for fluency in this study. However, following the manual, only the flexibility scores were
included in calculations of the E scores in these cases. Table 1 lists the E-scale
subtests with their scoring in the BIS-4, as well as a short description of them.
 Pilot study: The T-88 as a measure of creativity 27

Table 1
Subtests of the BIS-4 E Scale

Subtest Name Scoring Description M SD

figural:
ZF continue signs flexibility Completing identical line drawings to different real objects. 102.74 10.72
LO layout fluency Creating graphical emblems for a shop advertisement. 91.20 10.75
ZK combine signs fluency Combining four geometric figures to figural composites. 90.50 10.03
OJ object design flexibility Combining four geometric figures to various real objects. 98.20 11.43
verbal:
MA Masselon fluency Inventing sentences including three given nouns. 100.72 10.14
AM alternate uses flexibility Listing different possible uses for a given object. 96.59 9.29
EF traits-abilities flexibility Listing traits unfavorable for a given occupation. 100.76 8.83
IT insight test flexibility Listing reasons for a given social behavior. 102.73 10.24
numerical:
TN telephone numbers fluency Constructing easy-to-remember telephone numbers. 98.70 11.78
DR divergent calculations fluency Producing arrays of numbers that fit a given equation. 91.21 8.45
ZR number riddle fluency Filling number patterns into a geometric scheme. 92.32 8.87
ZG number equations fluency Producing equations from given numbers and operations. 93.84 9.34

2.1.3 Procedure

First, the E-scale subtests of the BIS-4 were administered to the subjects either in
individual (25%) or group testings of 2-5 participants (75%) on scheduled occasions.
After the BIS-4 testing, subjects received an envelope including the T-88 and further
personality measures not discussed in this thesis, and were asked to return them within
the next weeks after completing them at home. Subjects received a certification of their
participation on return of the materials and a personality profile a few weeks later.
Note that administering the T-88 in this manner deviated markedly from the standard
procedure for ability test. Though, ever since Wallach and Kogan (1965), more relaxed
and less evaluative test conditions have been suggested as favorable for divergent
tests. Ferris, Feldhusen and van Mondfrans (1971) compared the predictive validity of
divergent tests administered under four different conditions for academic achievement
in 5th, 8th and 11th grade pupils, and found that those tests filled out at home during
spare time and without time limit did best. Vernon (1971), in a review and study of
administration effects on divergent tests, concluded: “Our major finding is, then, that
divergent test scores obtained under relaxed conditions have generally richer
psychological meaning than those obtained under more formal, test-like conditions.”
The T-88 instructions, on the other hand, insist minding a 4 minutes time limit for the 18
items, and visual inspections of the test sheets (especially of the reported time used)
indicated that most subjects indeed paid attention to it. While there is no reason to
Pilot study: The T-88 as a measure of creativity 28

believe that such a de-relaxation of test conditions should have detrimental effects on
test scores, uncontrollable individual differences in keeping the time limit might
introduce error variance, at least in the number of items completed. This gives a second
rational for using item means instead of item sum scores for the analyses.

2.2 Results

2.2.1 Reliability analyses

Table 2 shows the means and standard deviations of the elaboration ratings for each of
the three raters, as well as their intercorrelations (ranging from .70 to .80). These values
were evaluated as high enough to sum all three raters to a composite score. The
composite yielded a satisfactory interjudge agreement (ICC 3, 3 = .72, indexed
according to the taxonomy of intraclass correlations suggested by Shrout and Fleiss,
1979; Cronbach’s α = .89).

Table 2
Pearson-Correlations of the Elaboration-Ratings

Elaboration Rater 1 Rater 2 M SD

Rater 1 - - 1.08 .17

Rater 2 .76** - 1.05 .23

Rater 3 .70** .80** .94 .26

Note: * p < .05, ** p < .01 (both two-tailed), N = 70.

Means and standard deviations as well as rater intercorrelations of the originality ratings
are listed in table 3. Because they were even better than those of the elaboration ratings
(.82 to .87), they were also summed to a composite, which had good reliability (ICC 3, 3
= .83, Cronbach’s α = .94).

Table 3
Pearson-Correlations of the Originality-Ratings

Originality Rater 1 Rater 2 M SD

Rater 1 - - 2.93 .36

Rater 2 .84** - 1.84 .38

Rater 3 .82** .87** 2.38 .44

Note: * p < .05 **, p < .01 (both two-tailed), N = 70.

Pilot study: The T-88 as a measure of creativity 29

Pearson product-moment correlations between elaboration and originality ratings were

.61 (p < .01), .44 (p < .01), .27 (p < .05) and .47 (p < .01) for raters 1, 2, 3, and rater
composite, respectively.
Table 4 gives psychometric properties for the BIS-4 total imaginativeness (E) scale, as
well as for the figural, verbal, and numerical E subscales. Reliability (Cronbach’s α) was
satisfactory for the total scale (.79), but was only modest (.50 to .62) for the subscales.
However, total scale internal consistency equals the value reported in the manual (.77),
and the test authors argue that Cronbach’s α is a coefficient of very limited usefulness
for the heterogeneous BIS-4 (Jäger et al., 1997, p. 37). Mean inter-item and part-whole
corrected item-total-correlations were good. Scale mean of total imaginativeness was
slightly lower than in the normative sample reported in the manual (1200; standard
deviations for all scales and means for subscales were not reported). The last two
columns of table 1 (p. 27) show means and standard deviations for the subtests of the E
scale. None of them differed markedly from the normative values (M = 100, SD = 10).

Table 4
Psychometric Properties of the BIS-4 E-Scale and Subscales

Number of
Scale Number Cronbach’s rii items with M SD
of items α corrected
rit < .20
E 12 .79 .24 0 1160.87 65.58
E (figural) 4 .57 .25 0 382.21 28.47
E (verbal) 4 .50 .20 0 404.41 23.59
E (numerical) 4 .62 .29 0 376.12 25.98

Note: N = 70, rii = mean inter-item correlation, rit = item-total correlation.

Visual inspections of histograms indicated that neither the T-88 nor the BIS-4 scores
deviated from a normal distribution.

2.2.2 Correlational analyses

Biserial correlations of T-88 and BIS-4 E scales with subjects’ sex and Pearson
correlations with age are shown in table 5. None of the correlations with age reached
significance, but elaboration correlated positively (.29) with sex, indicating women being
more elaborated in their T-88 drawings, and verbal imaginativeness correlated negative
(-.24) with sex, indicating higher values for men.
Pilot study: The T-88 as a measure of creativity 30

Table 5
Correlations of the T-88 and the BIS-4 with Sex and Age

T-88 T-88 BIS-4 E BIS-4 E BIS-4 E BIS-4 E

elaboration originality (figural) (verbal) (numerical)
Sex .29* .01 -.20 -.14 -.24* -.13
Age -.09 .07 .09 .02 .12 .09

Note: * p < .05, ** p < .01 (both two-tailed), N = 69.

Males are coded as 1, females as 2.

In the upper part of table 6, correlations of elaboration and originality with the total E
scale as well as the content-specific subscales are shown. Elaboration correlated
positively with all scales except verbal imaginativeness. However, none of these
correlations reaches significance. Originality shows low but significant associations with
all but numerical imaginativeness, which is still positive.

Table 6
Correlations of the T-88 with the BIS-4 E-Scale

T-88 T-88
elaboration originality
Scales:

BIS-4 E .14 .25*

BIS-4 E (figural) .19 .26*
BIS-4 E (verbal) -.01 .26*
BIS-4 E (numerical) .15 .11

Subtests:

figural:
ZF (‘continue signs’; flexibility) .23 .34**
ZF (‘continue signs’; fluency) # .15 .41**
LO (‘layout’; fluency) .15 .25*
ZK (‘combine signs’; fluency) .08 .08
OJ (‘object design’; flexibility) .06 .03
OJ (‘object design’; fluency) # .14 .13

verbal:
MA (‘Masselon’; fluency) .09 .18
AM (‘alternate uses’; flexibility) .09 .12
AM (‘alternate uses’; fluency) # .02 .14
EF (‘traits-abilities’; flexibility) -.04 .27*
EF (‘traits-abilities’; fluency) # -.07 .21
IT (‘insight test’; flexibility) -.16 .10
IT (‘insight test’; fluency) # -.29* -.02

numerical:
TN (‘telephone numbers’; fluency) .12 .06
DR (‘divergent calculations’; fluency) -.01 -.10
ZR (‘number riddle’; fluency) .08 .27*
ZG (‘number equations’; fluency) .19 .15

Note: # not included in the E-scales, * p < .05, ** p < .01 (both two-tailed), N = 70.
Pilot study: The T-88 as a measure of creativity 31

The lower part of table 6 gives correlations between T-88 scores and E scale subtests.
Of these 34 coefficients, only two reached significance on the one percent α level, both
between subtest ZF (flexibility and fluency score) and originality. Because of multiple
testing, results significant on a less conservative α level were ambiguous to interpret. Of
the remaining 15 originality correlations, 13 were in the expected direction, but only
three were significant on five percent α level. Five of the 17 correlations between
subtests and elaboration were not in the expected direction, one of them (IT fluency)
even significant on the five percent α level. No other correlation with elaboration
reached significance.

2.3 Discussion

Consistent with hypothesis one, completed T-88 drawings proved to contain enough
useful information to allow raters to make intersubjectively reliable judgments about
their originality and elaboration. In line with hypothesis two, both dimensions were
moderately related, but discriminable. Partially disproving hypothesis three, however,
only the originality ratings shared a small, but significant amount of variance with the
imaginativeness scale of the BIS-4. IT stem from significant relations with the figural and
verbal, but not the numerical aspects of imaginativeness - a reasonable result, since
originality of drawings (figural aspect) as well as titles (verbal aspect) affected T-88
scores. Associations are especially strong with the subtests most similar in content (ZF,
LO). These results indicate some support for convergence despite different test
situations, but question the generality across content aspects of the creative abilities
measured by the T-88. All in all, validity results seem disappointing at first glance,
especially for elaboration.
We have to keep in mind, however, that the T-88 and the BIS-4 E subtests are scored
very differently. Fluency is known to be the determining factor in classic divergent tests
(Hocevar, 1979a, b; see section 1.2) and serves this function also in the E scale (mean
correlation between flexibility and fluency scores of subtests for which both were
available was .74 in this sample). These kinds of tests do only measure some aspect of
real creative ability, and, as mentioned above, do not qualify as a creativity criterion.
Indeed, the criterion validities of the E scale reported in the manual are not very
convincing (Jäger et al., 1997, pp. 45-46). Since the T-88 scores used here were
Pilot study: The T-88 as a measure of creativity 32

controlled for fluency, the results are still remarkable, tentatively indicating that what is
shared between these operationalizations of originality and imaginativeness might be
true creative ability.
What can be concluded, then, is that elaboration and originality of T-88 responses can
be reliably rated, are distinguishable, but related dimension, and do not measure what
classical divergent tests measure, with only originality showing some association to that
aspect. The results Anastasi & Schaefer (1971) received with a similar measure, on the
other hand, turn optimistic that the T-88 is a valid measure of creativity. The clearest
limitation of the T-88 usage in this study is its uncontrolled time limit, reducing test
objectivity and therefore reliability and validity. However, the pilot study showed no hint
to markedly detrimental effect of the administration procedure used, suggesting some
robustness of the T-88 to it.
Main study: Creativity in BiLSAT and GOSAT 33

3. Main study: Creativity in BiLSAT and GOSAT

3.1 Method

3.1.1 Sample

The main study of this thesis is based on the Bielefeld Longitudinal Study of Adult Twins
(BiLSAT) and the German Observational Study of Adult Twins (GOSAT). The BiLSAT
register was started in 1993 by Alois Angleitner and Jan Strelau with over 1,100 pairs of
adult twins who voluntarily reacted on announcements in the media and twin clubs. The
register is population-based and very heterogeneous. By now, these twins were
surveyed three times by mail (BiLSAT), and a subsample of 300 twin pairs was invited
to the University of Bielefeld, one pair per day, for intensive assessment (GOSAT)
(Spinath, Angleitner, Borkenau, Riemann & Wolf, 2002).
The present study draws on data from BiLSAT waves II. and III. as well as GOSAT. Not
all twins completed all measure used in it, since (a) the BiLSAT sample size changed
over time due to both attrition and inclusions of twins being added to the register at a
later stage, and (b) only a subsample participated in GOSAT. To increase statistical
power, each analysis was conducted with the largest possible sample size, yielding
different subsamples in different analyses. I will therefore report sample descriptions
and measure reliabilities for the maximal total sample used as well as for the minimal
core sample that completed all relevant measures (on which analyses on highest
aggregation level were based).

3.1.1.1 Total sample

In total, data from 2,608 individuals (622 male (23.8%), 1842 female (70.6%)) was used.
Age at the last wave ranged from 21 to 74 (M = 39.0, SD = 12.9, median = 36.1). These
subjects belonged to 742 monozygotic (MZ), 289 same-sex dizygotic (DZss), and 226
opposite-sex dizygotic (DZos) pairs of twins. Subject’s sex was not available for 144
cases, age was even missing for 1452 cases and zygosity could not be reliably
determined for 320 individual twins. However, except for some basic analyses at the
very beginning, all further steps were based on data corrected for age and sex effects,
Main study: Creativity in BiLSAT and GOSAT 34

therefore only conducted with those subjects for whom these data were available. This
lead to a sample of 1,155 individuals (246 male (21.3%), 909 female (78.7%)), at the
last wave in the age of 21 to 74 (M = 39.0, SD = 12.9, median = 36.1), who belong to
352 MZ pairs, 179 DZss pairs and 42 DZos pairs. The zygosity of 9 individual twins was
unclear.

3.1.1.2 Core sample

Complete data for all measures used in this study was available for 267 subjects (55
male (20.6%), 212 female (79.4%)) who were at the age of 21 to 70 (M = 38.4, SD =
12.8, median = 34.1) at last wave of assessment. They combined to 53 MZ and 43
DZss twin pairs, leaving a residual of 38 individual MZ and 37 individual DZ twins,
whose co-twin failed to complete all relevant measures.

Comparisons of total and core sample descriptives indicated that age and sex structure
was similar.

3.1.2 Materials and procedure

The present study uses only a small part of the measures administered to the twins.
See Spinath et al. (2002) for a detailed list of all measures.

3.1.2.1 Measurement point 1: II. BiLSAT wave

In 1995, a battery of self-report questionnaires and two shorter batteries of peer-report

questionnaires were mailed to the twins of the register. They were asked to fill out the
self-report questionnaires and let themselves be rated on the peer-report questionnaires
by two peers who favorably knew one, but not the other, twin of the pair very well.
Postage for returning completed questionnaires was paid by the researchers. Three
measures administered in this wave are of interest for the present study:
Main study: Creativity in BiLSAT and GOSAT 35

EPQ-RK: The short version of the revised Eysenck Personality Questionnaire (EPQ-
RS; Eysenck & Eysenck, 1991) in its German adaptation (EPQ-RK; Ruch, 1999), a
measure of Eysenck’s P-E-N model of personality, was administered to the twin and in
a third-person version to both peers. Only the Psychoticism (P) scale of the EPQ-RK
was used herein. It asks the subject to answer 14 questions about his or her personality
with ‘yes’ or ‘no’. Exemplary items are “Are good behavior and tidiness important for
you?” or “Are you strongly affected by seeing children or animals suffer?”.

UNIPOL and BIPOL adjective lists: Both the UNIPOL and BIPOL adjective lists stem
from the German taxonomy of personality-descriptive terms (Angleitner & Ostendorf,
1994; Ostendorf, 1994). The UNIPOL list includes 100 unipolar adjectives and asks the
subject to indicate how well each of the adjectives describes him or her on a 5-point
Likert scale (ranging from 1, ‘not at all fitting’, to 5, ‘very fitting’). The BIPOL list consists
of 119 bipolar adjective pairs with 6-point Likert scales (ranging from -3 to +3), which
should be used by the subjects to indicate where on the dimension marked by the
adjectives he or she locates him- or herself. While the UNIPOL list was only given to the
twins, the BIPOL list was also filled out by both peers. Only the adjective ‘creative’ of the
UNIPOL list and the pair ‘uncreative – creative’ of the BIPOL list are of interest for the
present study.

3.1.2.1 Measurement point 2: GOSAT

Over a two-year period between 1995 and 1997, 300 BiLSAT twin pairs were studied,
pair by pair, for a whole day at the University of Bielefeld (see Spinath et al. (1999) for
details). Co-twins were separated during the whole GOSAT assessment day. Out of the
multitude of measures applied in GOSAT, two intelligence tests and video-based
personality ratings are of relevance for this thesis:

APM-20: The German version of Raven’s Advanced Progressive Matrices (Raven,

1958; Kratzmeier & Horn, 1974) was administered with a 20-minutes time limit. As
shown by Frearson and Eysenck (1986), the 20-minutes version is highly correlated
with the standard 40-minutes version.
Main study: Creativity in BiLSAT and GOSAT 36

LPS-K: The Leistungsprüfsystem (Horn, 1962) is a highly reliable German intelligence

test developed to measure Thurstone’s (1938) primary mental abilities. In GOSAT, the
short form of the test (Sturm & Willmes, 1983) was applied, consisting of seven subtests
(number 1, 2, 4, 5, 9, 10, and 12; 40 items each) measuring verbal comprehension,
reasoning, word fluency, space, and closure.

Neubauer, Spinath, Riemann, Angleitner and Borkenau (2000) calculated two oblique
factors resembling Cattell's (1963, 1971) fluid (gf) and crystallized (gc) intelligence from
the LPS-K data of the GOSAT sample, as well as a factor of general intelligence (g)
sensu Spearman (1904) from both the LPS-K and APM-20 results. The present study
will only use these three factors based on the factor values of the Neubauer et al. study.

Video ratings: During their GOSAT day, each twin was, separately from his co-twin,
videotaped in 15 assessment-center-like situations. These were (in timely order): (1)
introducing oneself, (2) arranging three photographs in a meaningful order and telling an
interesting story to it, (3) telling a dramatic story to each of three pictures from the
Thematic Apperception Test (TAT; Murray, 1943), (4) telling a joke, (5) persuading an
‘obstinate neighbor’ to turn down her music after 11 PM in a telephone role-play, (6)
refusing a request for help by a friend who just had a car accident in a second
telephone role-play, (7) introducing oneself to a stranger (actually a confederate) after
the confederate introduced herself, (8) recalling objects one has just seen in a waiting
room, (9) solving a complex logical problem, while the confederate from setting 7 solves
the same problem in enormous speed, (10) introducing a different confederate to the
experimenter, (11) inventing a definition for a neologism and provide arguments for why
that definition would be appropriate, (12) rigging up a high and stable paper tower within
5 minutes, using only scissors, paper, and glue, (13) reading 14 newspaper headlines
and their subtitles aloud, (14) describing multiple uses of a brick pantomimicly, and (15)
singing a song of one’s choice. Videotaped sequences of the situations had an average
duration of between one and twelve minutes, and summed up to about 60 minutes per
twin. Each twin’s personality was rated on base of the videos by four independent
judges per situation (i.e. 15 x 4 = 60 independent judges per twin) on 35 bipolar
adjective pairs, using computerized 5-point Likert scales. Only one of these adjective
pairs, ‘creative – uncreative’, will be dealt with in the following. Borkenau, Riemann,
Angleitner and Spinath (2001), whose article also gives further details on GOSAT and
Main study: Creativity in BiLSAT and GOSAT 37

the video ratings, calculated an aggregate score of the 60 ‘creative – uncreative’ ratings
per twin, which had an excellent interjudge reliability (ICC 1, 60 = .90). This aggregated
score was used in analyses reported below.

3.1.2.3 Measurement point 3: III. BiLSAT wave

Paralleling the procedure of the second wave described above, all twins of the register
were mailed further sets of questionnaires for self- and two peer-reports between 2000
and 2001. Two of these measures are of interest here:

NEO-PI-R: The revised NEO Personality Inventory (Costa & McCrae, 1992) in its
German adaptation (Angleitner & Ostendorf, 2003) is a measure of the NEO-Five Factor
Model trait domains of Neuroticism (N), Extraversion (E), Openness to Experience (O),
Agreeableness (A) and Conscientiousness (C). The five global scales consist of 48
items each, divided into six eight-item-long facet scales, totaling to 240 items. For each
item, the value of a statement like “I often feel tense and nervous” or “I like parties with
many people” as a description of one’s personality shall be indicated on a 5-point Likert
scale (ranging from 1, ‘not at all fitting’, to 5, ‘very fitting’). Only global scales will be
analyzed in this study. Apart from the twins themselves, two peers per twin rated them
using a version of the NEO-PI-R formulated in the third person.

T-88: Administration, scoring and raters of the T-88 were identical to the pilot study
(chapter 2). Since this time over 800 response sheets had to be rated, they were
randomly assigned to six booklets, which the raters worked through in randomized
order. Only 1.45% of the subjects completed less than one third of the T-88 items, and
only 5.19% completed less than half of the items. Therefore, again no subjects were
excluded for having completed too few items.
Main study: Creativity in BiLSAT and GOSAT 38

3.1.3 Zygosity determination

For 283 GOSAT twin pairs, zygosity diagnosis was conducted using either blood or
salvia samples. The procedure applied for 248 of these twin pairs was based on five
highly polymorphic dinucleotide repeat marker loci (Becker et al., 1997), with the
probability of misclassifying DZ as MZ twins being less than 0.9%. For further 35 of
these twin pairs, zygosity determination relied on semiautomated genome mapping on
ten highly polymorphic dinucleotide repeat marker loci, with an according error rate of
less than 0.1%. The classification of the remaining GOSAT and BiLSAT pairs was
based on questionnaire measures (Oniszczenko, Angleitner, Strelau & Angert, 1993) or
physical similarity data, with an estimated error rate of 7.0%.

3.2 Results

3.2.1 Reliability analyses

Tables 7 and 8 give means, standard deviations and intercorrelations for the three T-88
raters on the two rating dimensions (elaboration and originality). Average rater
intercorrelation was .80 for elaboration and .81 for originality, both evaluated as high
enough to sum raters to composites. For both composites, interrater reliability was good
(elaboration: ICC 3, 3 = .76, Cronbach’s α = .90; originality: ICC 3, 3 = .80, Cronbach’s
α = .92). Correlations between elaboration and originality were .70, .53, .47 and .61 for
rater 1, 2, 3 and rater composite, respectively (all p < .01).

Table 7
Pearson-Correlations of the Elaboration Ratings

Elaboration Rater 1 Rater 2 M SD

Rater 1 - - 1.03 .20

Rater 2 .82** - .97 .28

Rater 3 .73** .83** .81 .32

Note: * p < .05, ** p < .01 (both two-tailed), N = 828.

Main study: Creativity in BiLSAT and GOSAT 39

Table 8
Pearson-Correlations of the Originality Ratings

Originality Rater 1 Rater 2 M SD

Rater 1 - - 2.79 .37

Rater 2 .78** - 1.80 .41

Rater 3 .77** .86** 2.20 .47

Note: * p < .05 **, p < .01 (both two-tailed), N = 828.

Tables 9 shows results from reliability analyses of the NEO-PI-R and EPQ-RK, and
means and standard deviations of the LPS-K and the APM-20, for the total study
sample. Table 10 contains the same results for the core sample of 267 individuals.
Overall, reliabilities were good for all NEO-PI-R global scales. Internal consistencies
ranged between .85/.87 (A) and .93 (E) for self-reports and between .84 (O) and .93 (E)
for peer-reports. These values, as well as all NEO-PI-R means and standard deviations,
were very similar to those Angleitner and Ostendorf (2003) reported for the German
normative sample. Mean inter-item correlations seemed rather low and number of items
with critical part-whole corrected item-total correlations rather high, but such values are
expectable for long, heterogeneous scales measuring broad constructs.
For the EPQ-RK P scale, internal consistencies were rather low, especially for the self-
reports (.59/.59). Indeed, self-report values were markedly lower than those in a
normative sample reported by Ruch (1999), which ranged from .72 to .76. These values
were critical, even though the P scale is conceptualized as heterogeneous and strongly
varying in item difficulties, rendering Cronbach’s α an inadequate index for it (Ruch,
1999). Means and standard deviations of self-reports were also somewhat lower than
those reported by Ruch (M = 3.37 and SD = 2.54 for the total normative sample). No
such normative values were available for peer-reports. Corresponding to these results,
mean inter-item correlations were rather low and number of items with low corrected
item-total correlations rather high for a 14-item scale, especially for self-reports.
A comparison of tables 9 and 10 indicates very similar psychometric properties, means
and standard deviations in both samples for all measures, except peer-reported
Psychoticism, for which internal consistencies varied about .05 (.68/.66 vs. .74/.61).
However, since peer-reports were more reliable for one and less reliable for the other
peer in the core sample compared with the total sample, and both peer scores will be
aggregated later on, these differences will be averaged out. Overall, these comparisons
Main study: Creativity in BiLSAT and GOSAT 40

underpin optimism that later analyses using different subsamples will be based on
comparably reliable data.

Table 9
Psychometric Properties of Predictor Variables: Total Sample

No. of
Scale N No. of Cronbach’s rii items with Item number M SD
items α corrected
rit < .20
NEO-PI-R
Self
N 844 48 .93 .22 3 21, 141, 226 84.23 23.56
E 844 48 .89 .15 5 22, 112, 157, 167, 197 109.57 19.28
O 844 48 .87 .13 8 58, 88, 143, 148, 153, 116.31 17.55
208, 218, 238
A 844 48 .85 .11 5 29, 54, 129, 194, 219 119.19 15.45
C 844 48 .89 .16 6 10, 35, 105, 140, 150, 123.43 17.56
240
Peer 1
N 839 48 .93 .21 5 21, 81, 111, 141, 226 79.53 22.14
E 839 48 .89 .15 4 22, 112, 167, 212, 111.91 18.96
O 839 48 .87 .12 10 58, 78, 88, 143, 148, 108.76 16.84
153, 183, 208, 218, 238
A 839 48 .88 .14 7 29, 54, 84, 129, 194, 118.74 17.42
219, 239
C 839 48 .92 .21 2 150, 240 129.81 19.75

Peer 2
N 839 48 .93 .22 3 21, 141, 226 81.08 21.89
E 839 48 .89 .15 4 22, 112, 167, 212 112.04 18.29
O 839 48 .86 .11 10 58, 78, 88, 143, 148, 110.18 15.93
153, 183, 198, 208, 238
A 839 48 .89 .15 5 84, 129, 194, 219, 239 118.88 17.67
C 839 48 .92 .20 0 - 129.47 19.11

EPQ-RK
Self
P 1760 14 .59 .09 6 6, 26, 31, 37, 47, 50 2.44 1.84

Peer 1
P 1762 14 .68 .13 3 6, 31, 47 2.49 2.22

Peer 2
P 1753 14 .66 .12 5 6, 26, 31, 37, 47 2.45 2.16

LPS-K 590 7 x 40 - - - - 184.26 31.12

RAPM-20 578 36 - - - - 16.47 5.88

Note: rii = mean inter-item correlation, rit = item-total correlation.

Main study: Creativity in BiLSAT and GOSAT 41

Table 10
Psychometric Properties of Predictor Variables: Subsample with Complete Data
No. of items
Scale No. of Cronbach’s rii with Item number M SD
items α corrected
rit < .20
NEO-PI-R
Self
N 48 .93 .22 4 21, 126, 141, 226 83.70 23.52
E 48 .89 .16 6 22, 52, 112, 157, 167, 197 111.77 19.20
O 48 .88 .14 7 58, 88, 143, 148, 208, 218, 236 118.60 18.11
A 48 .87 .12 8 29, 54, 84, 164, 194, 219, 239, 144 118.30 16.13
C 48 .90 .17 5 35, 20, 105, 140, 150 122.24 18.34

Peer 1
N 48 .93 .21 5 21, 81, 111, 141, 226 81.46 22.81
E 48 .88 .14 7 2, 22, 42, 112, 157, 167, 212 111.54 18.48
O 48 .86 .11 11 58, 78, 88, 138, 143, 148, 153, 109.17 16.39
198, 208, 228, 238
A 48 .88 .13 7 29, 84, 89, 129, 194, 219, 239, 117.84 17.24
C 48 .92 .21 3 140, 150, 240 129.09 20.05

Peer 2
N 48 .93 .22 3 21, 141, 226 82.23 22.28
E 48 .89 .14 6 22, 67, 112, 132, 197, 212 112.28 18.24
O 48 .84 .10 15 28, 58, 78, 83, 88, 113, 153, 183, 111.13 15.27
193, 198, 118, 143, 148, 208, 238
A 48 .90 .16 5 84, 129, 194, 219, 239 118.32 18.05
C 48 .92 .20 1 20 128.61 19.18

EPQ-RK
Self
P 14 .58 .10 6 3, 6, 16, 26, 31, 39 2.67 1.85

Peer 1
P 14 .74 .17 0 - 2.57 2.41

Peer 2
P 14 .61 .10 6 6, 26, 31, 37, 44, 47 2.62 2.05

LPS-K 7 x 40 - - - - 184.90 30.78

RAPM-20 36 - - - - 17.01 5.93

Note: N = 267, rii = mean inter-item correlation, rit = item-total correlation.

A well-known problem of the Psychoticism scale is its difficulty and the resulting positive
skew of the score distribution. While its revision (Eysenck, Eysenck & Barrett, 1985;
Ruch, 1999) tackled this problem, it is not unlikely that it will still appear in the revised
version. Indeed, inspections of histograms and highly significant Kolmogorov-Smirnov-
z-tests (all p < .001) indicated that this was true in the present study for self- as well as
both peer-reports. Taking the square root of the raw scores (increased by 1) was an
adequate transformation to correct the skewness. All subsequent analyses were based
on these transformed P scores. The other personality, intelligence and creativity
measures were inconspicuous with respect to deviation from normal distribution.
Main study: Creativity in BiLSAT and GOSAT 42

3.2.2 Aggregation of measures

The left side of table 11 depicts the agreement of the two peers on personality
measures for the total sample. Pearson product-moment as well as intraclass
correlations were all between .40 (C and P) and .51 (E). These values indicate rather
high consensus (Funder, 1987). Both peer-reports were therefore aggregated for each
scale. The left side of the table reveals even higher accuracy (i.e., self-other agreement)
of personality judgments, lowest for A (r = .47, ICC 1, 2 = .47) and P (r = .50, ICC 1, 2 =
.44) and highest for E (r = .61, ICC 1, 2 = .60). In order to receive maximally pure
estimates of true personality dimensions, self- and mean peer-reports were also
averaged to one highly aggregated score per dimension.

Table 11
Correlations and Reliabilities of Personality Reports: Peer X Peer and Self X Mean Peer

Peer 1 X peer 2 Self X mean peer

r ICC 1, 2 r ICC 1, 2
NEO-PI-R
(N = 837-840)
N .44** .44 .54** .52
E .51** .51 .61** .60
O .45** .45 .53** .52
A .46** .46 .47** .47
C .40** .40 .53** .52

EPQ-RK
(N = 1746-1751)
P .40** .40 .50** .44
Note: * p < .05, ** p < .01 (both two-tailed).

Self ratings on ‘creative’ and ‘uncreative – creative’ adjective scales correlated .67 in the
total sample (p < .01, N = 1759) and were averaged after z-standardization. Consensus
of peers on the BIPOL adjective ‘uncreative – creative’ was reasonably high for a single
item (r = .31, p < .01, ICC 1, 2 = .31, N = 1738) and thus satisfactory for aggregation.
The aggregate of the 60 ‘creative – uncreative’ video ratings from GOSAT was recoded
for subsequent analyses, with high values now indicating high creativity ratings.
So far, theses aggregation steps leave us with five indicators of a disposition towards
creativity: aggregated originality and elaboration ratings from the T-88, and mean self,
peer and video-based stranger ratings. Table 12 lists the intercorrelations of these five
indicators, below diagonal for the total sample and above diagonal for those cases with
complete data on all indicators, yielding an N slightly higher than that of the core
Main study: Creativity in BiLSAT and GOSAT 43

sample. Differences between corresponding coefficients of both samples were small.

Correlations were highest for both T-88 scores (.61/.55) and between self- and mean
peer-reports on the adjective scales (.47/.45).

Table 12
Aggregated Creativity Indicators Intercorrelations

T-88 T-88 Mean self Mean peer Mean video

elaboration originality ‘creative’ ’creative’ ’creative’
T-88 elaboration - .55** .19** .20** .25**

T-88 originality .61** - .19** .12* .23**

(828)
Mean self ‘creative’ .20** .18** - .45** .27**
(806) (806)
Mean peer ‘creative’ .24** .16** .47** - .27**
(799) (799) (1732)
Mean video ’creative’ .24** .24** .25** .23** -
(294) (294) (523) (523)

Note: * p < .05, ** p < .01, (both two-tailed).

Below diagonal: Missing cases pairwise excluded, N in braces.
Above diagonal: Missing cases listwise excluded, N = 287.

A striking feature of this table is that all these various measures of a creative disposition
correlate positively and significantly (i.e., the matrix exhibits a positive manifold sensu
Spearman, 1904), indicating noteworthy nomological validity. It illustrated in a
nomological network (figure 2), using core sample results from above the diagonal of
table 12.

Figure 2
Nomological Network of Aggregated Creativity Indicators

T-88
elaboration
.55** .19**
.25** .20**
T-88 Self
originality .19** ratings
.12* .27**

.23** .45**

Video Peer
ratings .27** ratings

Note: * p < .05, ** p < .01 (both two-tailed), missing cases listwise excluded, N = 287.
Main study: Creativity in BiLSAT and GOSAT 44

For further investigation of creativity structure, the five indicators were entered into an
exploratory principal component factor analysis. Data adequacy was already suggested
by the correlation matrix and confirmed by a highly significant Bartlett test of sphericity
(p < .01) and an acceptable measure of sample adequacy (MSA ≥ .5; Kaiser-Meyer-
Olkin criterion). Initial eigenvalues were 2.08, 1.16, .75, .58, and .43. A parallel analysis
of 100 random correlation matrices with identical frame conditions yielded eigenvalues
of 1.20, 1.10, 1.03, .97 and .90, which converged with Kaiser and scree test criteria in
suggesting a two factor solution. Since table 12 indicated a meaningful interrelation
between all variables, direct oblimin rotation was applied, with a delta of .119 derived
from iterative hyperplane counts (based on factor pattern matrices). The final two factor
solution explained 64.72% of the variance, with the correlation of the two factors being
.31. Table 13 reports the final factor structure matrix. As can be seen, the first factor
was marked by high loadings of the self and peer ratings (both .81), as well as by a
substantial loading of the video ratings (.59). Loadings on the second factor were
highest for the two T-88 scores (.87 and .88). The video ratings showed also a
substantial secondary loading of .40 on this factor. Communalities indicate that the
video ratings were least well represented by the two factors.

Table 13
Factor Structure Matrix of the Aggregated Indicators after Oblimin Rotation

1. 2. h²
T-88 elaboration .28 .87 .75
T-88 originality .22 .88 .77
Mean self ‘creative’ .81 .20 .65
Mean peer ‘creative’ .81 .16 .67
Mean video ‘creative’ .59 .40 .40
Eigenvalues after rotation / Explained variance 1.78 1.74 64.72 %

Note: N = 287

The two oblique factors were re-entered into a principal component factor analysis to
receive a hierarchical creativity factor structure with a higher-order general factor on top
(Bartlett test and MSA were again acceptable). The upper half of table 14 reports factor
loadings of the two oblique factors on and their communalities with this higher-order
factor, which explained 65.33% of their variance. In the lower half of the table, values for
the first unrotated principal component of the five creativity indicators are given. This
general factor explained 45.55% of their variance, was nearly equally well represented
by all five variables (with loadings ranging between .61 and .70) and was perfectly
(1.00) correlated with the higher-order general factor.
Main study: Creativity in BiLSAT and GOSAT 45

Table 14
Higher-order Factor of the Aggregated Indicators

1. h²
Factor 1 .81 .65
Factor 2 .81 .65
Explained variance (%) 65.33 65.33

T-88 elaboration .70 .49

T-88 originality .67 .44
Mean self ‘creative’ .63 .40
Mean peer ‘creative’ .61 .38
Mean video ‘creative’ .61 .38
Explained variance (%) 41.55 41.55

Note: N = 287

Figure 3 summarizes the aggregation of indicators (lower part) and factor structure of
the disposition towards creativity (upper part) in this study.

Figure 3
Indicators and factor structure of the disposition towards creativity

General
factor

.81 .81

Factor 1 Factor 2

.81 .81 .59 .40 .87 .88

Self- Peer- Video T-88 T-88

reports reports ratings elaboration originality
BIPOL peer 1

BIPOL peer 2

15 x 4 raters
UNIPOL self

T-88 rater 1

T-88 rater 2

T-88 rater 3

T-88 rater 1

T-88 rater 2

T-88 rater 3
BIPOL self

Note: Factor loadings below .40 were omitted.

Main study: Creativity in BiLSAT and GOSAT 46

3.2.3 Age and sex effects

Before the relation between the indicators and factors of a creative disposition and other
psychometric measures were explored, their dependence on subjects’ age and sex had
to be controlled. This was important since (a) developmental and sex difference
questions were out of scope of this study, so age and sex effects could bias conclusions
about general relations between variables in adult age, and (b) uncorrected age and sex
effects inflate twin correlations, therefore biasing estimates of genetic and
environmental influences on variables in behavior genetic analyses (McGue &
Bouchard, 1984). Age and sex effects on creativity indicators and factors are shown in
table 15. Positive correlations with sex indicate higher values for women. Video ratings
were omitted in this table, since the aggregate taken from Borkenau and colleagues
(2001) was already corrected for age and sex effects. Biserial correlations of the others
with sex and Pearson correlations with age were generally very low (ranging between
±.14). Only peer ratings were significantly higher for women. Both T-88 scores and the
second creativity factor were significantly higher for younger subjects, while self ratings
were significantly higher for older subjects.

Table 15
Correlations of Creativity Indicators with Sex and Age

T-88 T-88 Mean self Mean peer Creativity Creativity Creativity

elaboration originality ‘creative’ ’creative’ factor 1 factor 2 g-factor
Sex .03 .00 .00 .14** -.03 -.06 -.06
(828) (828) (1759) (1738) (287) (287) (287)
Age -.14** -.07* .10** .05 -.02 -.14* -.10
(790) (790) (1074) (1064) (287) (287) (287)

Note: * p < .05, ** p < .01 (both two-tailed), N in braces.

Males were coded as 1, females as 2.

Table 16 shows the same correlations for personality (aggregated over self- and mean
peer-reports) and intelligence measures. Age and sex effects were a bit stronger
(between -.23 and .24 for sex and between -.46 and .20 for age) and more common
here. In this sample, men were more emotionally stable, less agreeable, more
conscientious, more psychotic and more intelligent than women, and younger subjects
were more neurotic, extraverted and open to experience, less agreeable and
conscientious, more psychotic, and had a remarkably higher general and fluid, but
slightly lower crystallized intelligence. All these effects resemble those normally
reported in the literature in size and direction (e. g. Angleitner & Ostendorf, 2003; Ruch,
1999; Horn & Cattell, 1967).
Main study: Creativity in BiLSAT and GOSAT 47

Table 16
Correlations of Personality and Intelligence Measures with Sex and Age

NEO-PI-R NEO-PI-R NEO-PI-R NEO-PI-R NEO-PI-R EPQ-RK g gf gc

N E O A C P
Sex .24** .01 .06 .16** -.10** -.08** -.23** -.21** -.16**
(840) (840) (840) (840) (840) (1746) (568) (590) (590)
Age -.24** -.09* -.13** .20** .15** -.29** -.37** -.46** .09*
(801) (801) (801) (801) (801) (1069) (568) (590) (590)

Note: * p < .05, ** p < .01 (both two-tailed), N in braces.

Males were coded as 1, females as 2.
All personality measures were aggregated over self- and two peer reports.

Subjects’ age and sex were regressed on all variables (including non-aggregated
personality self- and mean peer-reports, which were omitted here), and all subsequent
phenotypic and behavior genetic analyses were solely based on residual scores
corrected for sex and age effects.

3.2.4 Phenotypic relations of creativity with intelligence and personality

3.2.4.1 Relations of creativity with intelligence

First, relations between creativity indicators and general, fluid and crystallized
intelligence were explored. Correlations of these variables are listed in table 17. A first
inspection reveals that all but one correlation, which was close to zero (-.01), were
positive. However, none was greater than .35. Interpreting false-positive results
stemming from multiple testing will be avoided by considering only results significant on
one percent α level. To test whether the gf and gc differ significantly in their relation to
creativity variables, two-tailed difference t-tests were conducted (Diehl & Staufenbiel,
2001, p. 693; see Bortz, 1999, formula 6.97, for a similar z-test). The row in the middle
of table 17 gives significance levels for these tests.
While all correlations between T-88 elaboration and intelligence factors were highly
significant and of similar magnitude, only crystallized intelligence exhibited a strong
association with originality, which was significantly higher than that of gf. No intelligence
factor correlation with self- or peer-reported creativity was highly significant. On the
other hand, the GOSAT video ratings showed the strongest relation of all creativity
indicators with all intelligence factors to a similar extend. Only the grossly overlapping
 Main study: Creativity in BiLSAT and GOSAT 48

Table 17
Correlations of Creativity with Intelligence

T-88 T-88 Mean self Mean peer Mean video Creativity Creativity Creativity
elaboration originality ‘creative’ ’creative’ ‘creative’ factor 1 factor 2 g-factor
g .19** .07 .03 .09 .35** .23** .20** .28**
(286) (286) (511) (511) (555) (270) (270) (270)
gf .20** .02 .07 .11* .29** .27** .16** .29**
(300) (300) (532) (532) (574) (282) (282) (282)
gc .22** .22** -.01 .02 .32** .07 .30** .21**
(300) (300) (532) (532) (574) (282) (282) (282)

Difference n. s. ** n. s. * n. s. ** * n. s.
test (gf-gc)

gf.gc .10 -.12* .09* .12** .16** .28** .01 .18**

(297) (297) (529) (529) (571) (279) (279) (279)
gc.gf .14* .25** -.05 -.05 .20** -.09 .25** .10
(297) (297) (529) (529) (571) (279) (279) (279)

Note: * p < .05, ** p < .01, n. s.: non-significant (all two-tailed), N in braces.

fluid and general intelligence correlated significantly with the first creativity factor, the
difference between gf and gc relations being highly significant in favor of gf. All three
intelligence factors, but especially gc (.30, difference to the .16 relation of gf significant
on the five percent α level), correlated with the second factor. Finally, all three
intelligence factors were significantly correlated with the creativity general factor
(general and fluid intelligence slightly, but insignificantly higher than crystallized).
To further test whether the gf or the gc facet of intelligence was a more important
predictor of a disposition towards creativity, partial correlations were calculated, with
one facet controlled in correlations of the other facet with creativity variables (last two
rows of table 17). It has to be noted that gf and gc were strongly overlapping (r = .52, p <
.01, N = 590), with a meaningful core of general intelligence. These partial correlations
therefore removed valid variance from both variables. They are informative, however, in
that they indicate the relative importance of genuine fluid and crystallized intelligence
aspects.
Partialling out the other facet decreased the correlations of both gf and gc with T-88
elaboration score, but slightly increased the advantage of gc. The pure aspects of gf and
gc seem to have antagonistic effects on originality, with controlling the other facet
yielding a significantly negative relationship with gf (-.12) and a slightly increased one
with gc. A similar effect in the opposite direction, but of neglectable magnitude, can be
suspected in the pattern observable in both correlations with self- and peer-reported
creativity. The correlations with the video ratings, in contrast, were similarly decreased
Main study: Creativity in BiLSAT and GOSAT 49

for gf and gc residuals, with both staying highly significant. All three aspects, genuine
fluid and crystallized intelligence as well as their common core, seem to be influential
here. The relations with creativity factor scores reflected what was already observable
on aggregated measures level: Only genuine gf correlated with factor 1 and only
genuine gc correlated with factor 2. Controlling the other facet had reductive effects on
relations with the general creativity factor for both. But while the non-significant
difference between the correlations of gf and gc with this factor stayed the same, only
the one with the gf residual remained significantly different from zero.
Guilford’s threshold hypothesis was tested using the two methods he suggested
(Guilford, 1967a, b, 1981). First, general intelligence was plotted against the creativity g
factor (figure 4a). While this scatter plot might remind one of the triangular pattern
Guilford hypothesized, it has to be remarked that 3 of the 270 data points depicted here
lie outside 3 standard deviations (two > 3 SD on the creativity g factor, one < 3 SD on
intelligence g). When these three outliers were removed (figure 4b), the triangular
pattern diminished completely.

Figure 4
Scatter Plots of Intelligence and Creativity g Factors

a) b)

Guilford’s second suggested test of the threshold hypothesis demands correlating

creativity and intelligence measures in a high-IQ-sample. It hypothesizes a non-
significant correlation for such a sample. To avoid false interpretations of such a
decrease in correlations as a confirmation of the threshold hypothesis when it was just
Main study: Creativity in BiLSAT and GOSAT 50

based on range restriction, two extreme groups were build: The first contained all
subjects with general intelligence factor values at least one standard deviation above
sample mean (N = 44), while the second contained all subjects at least one standard
deviation below mean (N = 44). Neubauer and colleagues (2000) calculated an IQ
mean of 111 for this sample’s LPS-K results. Since the LPS norms stem from 1962,
Neubauer and colleagues argued for an effective IQ mean of around 100 in this sample,
based on the Flynn effect (describing an increase of mean population IQ of about 3
points per decade). IQ of the high-g extreme group subjects should therefore be at least
around 115, qualifying for a test of Guilford’s hypothesis.
Results for the high-g group indeed indicated a non-significant correlation of general
intelligence with the first and second creativity factor as well as with the higher-order
general factor (.10, -.09, and -.04, respectively, all n. s. in two-tailed tests). However,
results were similar for the low-g group (.06, .08, and .09, respectively, all n. s. in two-
tailed tests), even though all results were in positive direction here.

3.4.2.2 Relations of creativity with personality

Correlation analyses to clarify the relation of dispositional creativity and personality were
separately conducted for personality self-reports, mean peer-reports, and aggregated
self- and mean peer-reports, yielding a total of 144 correlation coefficients (first 8
columns of table 18). Since danger of false-positive significant correlations was
especially high here, only correlations that were significant on one percent α level and
meaningfully replicable within this dataset will be interpreted. Openness to Experiences
easily passed these criteria: All 24 correlations between O and creativity indicators and
factors were highly significant, even reaching .52 on highest level of aggregation.
Second, 18 of 24 correlations between Extraversion and creativity indicators reached
significance on one percent α level, with the non-significant ones being exclusively all
those with T-88 scores. Interestingly, E was also significantly associated with the
second creativity factor, on which both T-88 dimensions showed high loadings.
Openness and Extraversion were the only two personality dimensions which showed
noteworthy relations to the highest-level creativity aggregate, the creativity general
factor. In contrast, Neuroticism was only significantly related (in negative direction) with
creativity self- and peer-reports. But while all six correlations were significant, they were
 Table 18
Correlations of Creativity with Personality

T-88 T-88 Mean self Mean peer Mean video Creativity Creativity Creativity
elaboration originality ‘creative’ ’creative’ ‘creative’ factor 1 factor 2 g-factor

Self-report
NEO-PI-R (N = 789) (N = 789) (N = 789) (N = 781) (N = 300) (N = 286) (N = 286) (N = 286)
N .03 .04 -.13** -.09* -.01 -.04 .13* .06
E -.01 .05 .25** .17** .26** .33** .17** .31**
O .19** .26** .35** .26** .34** .42** .32** .46**
A -.07* -.11** .02 .05 .13* .00 -.09 -.05
C -.05 -.05 .09** .02 -.10 .02 -.21** -.12*

EPQ-RK (N = 776) (N = 776) (N = 1073) (N = 1063) (N = 525) (N = 287) (N = 287) (N = 287)

P .02 .05 .00 .04 .11* .00 .05 .03
Main study: Creativity in BiLSAT and GOSAT

Mean peer-report
NEO-PI-R (N = 785) (N = 785) (N = 786) (N = 778) (N = 299) (N = 285) (N = 285) (N = 285)
N -.04 .00 -.07* -.12** -.03 -.02 .03 .00
E .03 .04 .16** .17** .19** .25** .16** .26**
O .23** .23** .22** .27** .40** .38** .31** .43**
A .03 .02 .02 .10** .11 .05 -.02 .02
C -.03 -.05 .01 .04 .02 .00 -.14* -.08

EPQ-RK (N = 774) (N = 774) (N = 1067) (N = 1060) (N = 524) (N = 286) (N = 286) (N = 286)

P .09** .04 .04 .02 .07 -.05 .08 .01

Aggregate
NEO-PI-R (N = 785) (N = 785) (N = 786) (N = 778) (N = 299) (N = 285) (N = 285) (N = 285)
N .01 .03 -.12** -.12** -.02 -.04 .10 .04
E .01 .05 .23** .19** .25** .33** .18** .32**
O .24** .28** .33** .31** .42** .47** .37** .52**
A -.02 -.06 .02 .09* .14* .03 -.07 -.03
C -.05 -.06 .06 .04 -.05 .01 -.20** -.11

EPQ-RK (N = 773) (N = 773) (N = 1066) (N = 1059) (N = 524) (N = 286) (N = 286) (N = 286)

P .06 .06 .02 .03 .10* -.04 .07 .02
51

Note: * p .05, ** p < .01 (all two-tailed), N in braces.

Main study: Creativity in BiLSAT and GOSAT 52

low and two only reached the five percent α level. Additionally, the association with N
was not found for the first creativity factor, which was highly loaded by the self- and peer
ratings. Conscientiousness showed only one spurious and not replicated correlation
with creativity indicators, but surprisingly consistent and quite substantial negative
correlations (up to -.21, one only significant on five percent α level) with the second
creativity factor. Agreeableness and Psychoticism showed only few and not replicable
significant correlations with creativity indicators, and none of those was meaningful
enough to reappear on factor level.

3.4.2.3 Regression analyses

Next, relations between variables were closer examined by several stepwise regression
analyses of all personality dimensions (on highest aggregation level) and fluid and
crystallized intelligence on the three creativity factors. General intelligence was
excluded, because it was highly correlated with both fluid (.92) and crystallized
intelligence (.77), which promised a more differentiated picture. F-test p levels were .05
for inclusion and .10 for exclusion of variables in all analyses. In none of the three
regressions did visually inspected scatter plots of standardized predicted values by
standardized residuals indicate signs of non-linearity or heteroskedasticity, nor did any
Durbin-Watson test of autocorrelation reach significance (d = 1.87, 1.85, and 1.94, for
factor 1, 2 and the general factor regressions, respectively, all n. s.; Backhaus,
Erichson, Plinke & Weiber, 2000).
The first stepwise regression analysis of personality and intelligence scores on creativity
factor 1 yielded an explanation of 28% variance by Openness to Experience,
Extraversion, and fluid and crystallized intelligence (table 19). O alone explained 21%.
The entering of gc in the fourth step had an interesting effect: While showing a positive
(but insignificant) zero-order correlation with the first creativity factor, gc had a
significantly negative beta weight in this multiple regression. On the other hand, the
introduction of gc increased the beta weights of O (with rgc, O = .28, p < .01, N = 305) and
especially gf, tentatively indicating a suppression effect of gc. This replicates the
antagonistic effect of gf and gc in predicting self- and peer-reported creativity, which
were the marker variables of factor 1. It is noteworthy, though expectable, that the
Main study: Creativity in BiLSAT and GOSAT 53

entering of gc in the regression had an reductive effect on the tolerances (from all
greater .81 to greater .66).

Table 19
Stepwise Regression of Personality and Intelligence on Creativity Factor 1

Predictors Beta weights R² change Total adjusted R²

Step 1 O .46*** .21*** .21***

Step 2 O .43***
gf .17** .03** .24***

Step 3 O .35***
gf .19***
E .19*** .03** .26***

Step 4 O .39***
gf .27***
E .16**
gc -.17** .02** .28***
Note: Predictors: N, E, O, A, C, P, gf, & gc.
All personality predictors were aggregated over self- and two peer-reports.
* p < .05, ** p < .01, *** p < .001, N = 279.

The second creativity factor was best predicted by Openness to Experience, crystallized
intelligence, Conscientiousness (-), and Extraversion (table 20). These predictors
explained only 21% of its variance, though. O alone explained 14%. C, which added
about 3% incremental predictive validity, did not reach significance in the zero-order
correlations reported above (but showed a similar negative trend). The entrance of E in
step 4 reduced tolerances (from all greater .90 to greater .70), supposingly because of
its overlap with O (rE, O = .39, p < .01, N = 801).

Table 20
Stepwise Regression of Personality and Intelligence on Creativity Factor 2

Predictors Beta weights R² change Total adjusted R²

Step 1 O .37*** .14*** .13***

Step 2 O .30***
gc .21*** .04*** .17***

Step 3 O .29***
gc .21***
C -.16** .03** .19***

Step 4 O .22**
gc .23***
C -.19**
E .14* .02* .21***
Note: Predictors: N, E, O, A, C, P, gf, & gc.
All personality predictors were aggregated over self- and two peer-reports.
* p < .05, ** p < .01, *** p < .001, N = 279
Main study: Creativity in BiLSAT and GOSAT 54

The variance of the general creativity factor was predictable to 30% by Openness to
Experience, fluid intelligence, and Extraversion (table 21). O alone explained 26%.
Again, predictors were somewhat interrelated, but all tolerances were still greater than
.78 when all three variables were entered.

Table 21
Stepwise Regression of Personality and Intelligence on the Creativity g Factor

Predictors Beta weights R² change Total adjusted R²

Step 1 O .51*** .26*** .26***

Step 2 O .48***
gf .16** .02** .28***

Step 3 O .42***
gf .18**
E .15** .02** .30***

Note: Predictors: N, E, O, A, C, P, gf, & gc.

All personality predictors were aggregated over self- and two peer-reports.
* p < .05, ** p < .01, *** p < .001, N = 279

A more differentiated perspective on the relative influences of each of these partly

overlapping variables on dispositional creativity would be useful, especially with respect
to the gf-gc question and Peterson et al.'s (2002) result of a composite of E and O best
predicting creativity. To get such a perspective, communality analyses were conducted
for the relationships of all three creativity factors to their significant predictors from the
stepwise regressions. In a communality analysis, the explained criterion variance of a
multiple regression is partitioned into unique influences of the predictors and influences
common to several predictors, i.e. criterion variance predicted by the overlap of several
predictors (Cooley & Lohnes, 1976, pp. 218-223). Table 22 lists the results for all three
factors. When multiplied with 100, values in this table can be interpreted as percentages
of explained criterion variance. They add up to the determination coefficient (R²) of the
corresponding multiple regression (last row), which is not adjusted for number of
predictors here, therefore slightly larger than in the tables 19 to 21.
Main study: Creativity in BiLSAT and GOSAT 55

Table 22
Communality Analyses of Creativity Factors Predictions

Creativity Factor 1 Creativity Factor 2 Creativity g Factor

U(O): .115 U(O): .038 U(O): .139
U(gf): .054 U(gc): .048 U(gf): .030
U(E): .019 U(C): .035 U(E): .018
U(gc): .022 U(E): .014 C(O, gf): .040
C(O, gf): .011 C(O, gc): .041 C(O, E): .080
C(O, E): .078 C(O, C): .016 C(gf, E): -.006
C(O, gc): -.019 C(O, E): .042 C(O, gf, E): .008
C(gf, E): -.001 C(gc, C): -.002
C(gf, gc): -.020 C(gc, E): -.008
C(E, gc): .008 C(C, E): -.008
C(O, gf, E): .013 C(O, gc, C): .009
C(O, gf, gc): .024 C(O, gc, E): .002
C(O, E, gc): -.003 C(O, C, E): -.008
C(gf, E, gc) -.005 C(gc, C, E) .004
C(O, gf, E, gc): -.003 C(O, gc, C, E): -.005
R²: .293 R²: .218 R²: .309

Note: U: Uniqueness, C: Communality.

N = 279.
Fractions ≥ |.020| (= 2% explained variance) are printed in bold face.

Results for creativity factor 1 are shown in the first column. More than one third of the
variance explained in this factor (11.5%) was solely attributable to Openness to
Experience, and it increased to about two thirds, when the variance O shared with E
was added. Unique E variance, on the other hand, contributed only a very small amount
to it. The uniqueness of fluid intelligence was responsible for the next bigger share
(5.4%), surprisingly followed by 2.4% common variance of O, gf and gc plus 2.2%
unique gc variance. This positive influence of crystallized intelligence was hidden in the
stepwise regression by its suppression effects on gf and O (identifiable by negative
contributions to R²).
The communality analysis of the regression on the second creativity factor revealed that
the most important unique influence on it was not O (3.8%), but gc (4.8%). The variance
shared by O and E (4.2%) was about as predictive as the unique variance of O, with the
unique E variance again only contributing a very small amount (1.4%). Other noteworthy
contributions came from the overlap of O and gc (4.1%) and unique C (3.5%).
Influences on the general creativity factor (last column of table 22, also illustrated in
figure 5) were rather straightforward: The single major predictor was O, uniquely
contributing 13.9%, plus 8% via shared variance with E and 4% shared with gf.
Extraversion’s unique contribution was again small (1.8%), while gf added unique 3% of
predicted variance.
Main study: Creativity in BiLSAT and GOSAT 56

Figure 5
Results for the Communality Analysis of the General Creativity Factor

Explained
variance (R²)
31%

Unexplained
variance
69%
C(gf, E)
2% C(O, gf, E)
U(E) 2%
U(gf) 6%
9% U(O)
44%

C(O, gf)
12%

C(O, E)
25%
Note: U: Uniqueness, C: Communality. N = 279.

Some authors suggested synergistic effects of personality dimensions and intelligence

on creativity: Eysenck (1995a, b) and Jensen (1996) for P, Peterson et al. (2002) for O
and E. A serial of additional regression analyses was conducted to test these interaction
effects. Four dummy variables containing the product of g times either E, O, a
composite (sum) score of E and O, or P (each aggregate of self- and mean peer-
reports) were calculated. They were entered as predictors into a hierarchical (blockwise)
multiple regression in the second step, after g and the corresponding personality
dimension(s) had been entered as predictors in the first step. All three creativity factors
were, one after the other, entered as dependent variables in each of the four
regressions, yielding a total of twelve analyses. In none of these twelve cases did the
interaction term contribute incrementally to the prediction, with none of the t- and F-tests
being even marginally significant (all p > .10).
Main study: Creativity in BiLSAT and GOSAT 57

3.4.2.4 The structure of creativity after controlling for intelligence and personality

To test the necessity of broad personality dimensions and general mental abilities for a
disposition towards creativity, all dimensions of the FFM, Psychoticism (all aggregated
over self- and mean peer-reports), and general, fluid and crystallized intelligence were
regressed on each of the five creativity indicators (T-88 elaboration and originality, self-,
peer- and video-based stranger rated creativity). The five residuals were intercorrelated
(table 23). Comparisons with the zero-order intercorrelations of the same variables
(table 12 and figure 2, p. 41) revealed that controlling for broad personality and
intelligence dimensions (as well as sex and age) affected the convergence of the
creativity indicators: All correlations were reduced, with three of the ten former
significant correlations now insignificant. While nine of the ten zero-order correlations
were significant on one percent α level, only two of the residual correlation reached it
(T-88 elaboration with T-88 originality and mean peer ratings of creativity). All values
remained, however, positive.

Table 23
Intercorrelations of Creativity Indicators after Controlling for all Predictors
T-88 T-88 Mean self Mean peer
elaboration originality ‘creative’ ’creative’
T-88 elaboration - - - -
T-88 originality .48** - - -
Mean self ‘creative’ .11 .14* - -
Mean peer ‘creative’ .17** .11 .37* -
Mean video ’creative’ .09 .15* .14* .16*

Note: * p < .05, ** p < .01 (both two-tailed). N = 267 to 283.

All variables were corrected for sex, age, N, E, O, A, C, g, gf, & gc.

A highly significant Bartlett test (p < .01) and an acceptable MSA indicated this matrix
as adequate for entering into a principle component factor analysis. Initial eigenvalues
were 2.08, 1.16, .75, .58, .43, and a parallel analysis of 100 random matrices with
identical frame conditions (mean eigenvalues: 1.17, 1.07, 1.00, .92, .84), as well as
scree and Kaiser criteria, suggested the extraction of two factors, which explained
59.17%. This was 5.55% less than the two factors extracted from zero-order
correlations explained (see table 13). Similar to the factor analysis of uncorrected
creativity indicators, the two factors were rotated using direct oblimin rotation with a
delta of .119. An iterative hyperplane count based on factor pattern matrices confirmed
that this was a tenable solution. The correlation of both factors was .26, not significantly
Main study: Creativity in BiLSAT and GOSAT 58

different from the correlation of the oblique factors derived from uncorrected
correlations. Table 24 shows that the first factor was again marked by high loadings of
the self and peer creativity ratings, as well as a moderate loading of the video-based
stranger ratings, while the second factor was marked by the two T-88 scoring
dimensions. The secondary loading of the video ratings on factor two was missing in the
analysis of corrected indicators. The video ratings were also not very well represented
by these two factors, the communality being only .27. However, factor structure of both
factors was nearly identical with that of the corresponding factor extracted from
uncorrected creativity indicators (Tucker’s phi was .998 for factor 1 and .986 for factor 2;
Bortz, 1999, formula 15.72a), and Pearson correlations between corresponding factors
were high (.83 and .88 for factors 1 and 2, respectively, both p < .01, N = 267).

Table 24
Factor Structure Matrix of the Creativity Indicators, Corrected for all Predictors, after Oblimin
Rotation

1. 2. h²
T-88 elaboration .21 .86 .74
T-88 originality .20 .86 .75
Mean self ‘creative’ .78 .15 .61
Mean peer ‘creative’ .77 .17 .59
Mean video ‘creative’ .52 .18 .27
Eigenvalues after rotation / Explained variance 1.55 1.57 59.17%
Note: N = 267. All variables were corrected for sex, age, N, E, O, A, C, g, gf, & gc.

The two oblique factors were again re-factorized in a principle component analysis to
extract a corrected higher-order general factor (with Bartlett test being highly significant
(p < .01) and MSA acceptable). The general factor explained 62.91% of the two oblique
factors’ variance (2.42% less than in the first general factor did, see table 14), with both
exhibiting loadings of .79 on it (upper part of table 25). The first unrotated factor
extracted from the five indicators, a general factor which was perfectly correlated with
the higher-order factor, explained 35.82% of the variance (5.73% less than the
corresponding uncorrected factor, see table 14). The lower part of table 25 shows the
loadings of the corrected indicators, ranging between .44 for the video ratings and .68
for both T-88 elaboration and originality. This factor structure was again nearly identical
with the corresponding uncorrected factor (Tucker’s phi = .994), and the two correlated
.82 (p < .01, N = 267).
Main study: Creativity in BiLSAT and GOSAT 59

Table 25
Higher-Order Factor of the Creativity Indicators, Corrected for all Predictors
1. h²
Factor 1 .79 .63
Factor 2 .79 .63
Explained variance (%) 62.91 62.91

T-88 elaboration .68 .46

T-88 originality .68 .46
Mean self ‘creative’ .58 .33
Mean peer ‘creative’ .59 .34
Mean video ‘creative’ .44 .19
Explained variance (%) 35.82 35.82

Note: N = 267. All variables were corrected for sex, age, N, E, O, A, C, g, gf, & gc.

Figure 6 summarizes, following figure 3, the hierarchical factor structure of the

disposition towards creativity after statistical control of personality dimensions and
intelligence factors.

Figure 6
Factor Structure of the Disposition towards Creativity, Corrected for all Predictors

General
factor

.79 .79

Factor 1 Factor 2

.78 .77 .52 .86 .86

Self- Peer- Video T-88 T-88

reports reports ratings elaboration originality

Note: Factor loadings below .40 were omitted.

Main study: Creativity in BiLSAT and GOSAT 60

3.2.5 Genetic and environmental influences on creativity and its correlations with
personality and intelligence

The etiology of dispositional creativity, i.e. the genetic and environmental influences on
individual differences in it, was explored in two steps: First, within-pair intraclass
correlations for MZ and DZ twins were calculated and univariate behavioral genetic
structure equation models were fitted to the twin intra-pair variances and covariances in
order to estimate the relative effects of genetic, shared environmental and non-shared
environmental influences on creativity indicators, factors, and significant predictors. To
further test the necessity of personality and intelligence for dispositional creativity, these
models were also fitted to creativity indicators and factors corrected for all personality
and intelligence variables assessed in this study. Second, multivariate structure
equation models were fitted to twin intra-pair cross-trait cross-twin covariances (i.e. the
covariance between trait 1 in twin 1 and trait 2 in twin 2 of a pair) to estimate genetic,
shared environmental and non-shared environmental influences on the correlations
between creativity variables and their predictors, and to estimate the amount of overlap
between the sources influencing the variance of both variables.
For all subsequent analyses, all variables were corrected for effects of sex and age, and
all personality dimensions were aggregates of self- and mean peer-reports. Out of the
DZ twin pairs, only same-sexed pairs were included in the analyses. Behavioral genetic
analyses for the personality and intelligence predictors have already been conducted
with data from this sample. See Neubauer et al. (2000) for the three intelligence factors
and Angleitner (2002) for the NEO-PI-R-based FFM dimensions.

3.2.5.1 Univariate behavioral genetic analyses

First hints on the etiology of individual differences in a diposition towards creativity can
be deduced from a comparison of MZ and DZ twin within-pair intraclass correlations
(ICC 1, 1), listed in table 26 (upper part). All ICCs were greater zero, indicating that not
only environmental influences unshared by siblings contributed to the variances.
However, since no ICC was perfect (or close to the reliability of the measure), all
variables can still be expected to be partly influenced by the non-shared environment.
Main study: Creativity in BiLSAT and GOSAT 61

Table 26
Intraclass Correlations (ICC 1, 1) between MZ and DZ Twins

MZ DZ DZ/MZ ratio
Creativity indicators and factors:

T-88 elaboration .46 .48 1.04

(222) (78)
T-88 originality .30 .32 1.07
(222) (78)
Mean self report ‘creative’ .48 .19 .40
(318) (161)
Mean peer report ‘creative’ .28 .19 .68
(316) (155)
Mean video ratings ‘creative’ .57 .38 .67
(158) (126)
Creativity factor 1 .53 .30 .57
(63) (44)
Creativity factor 2 .48 .44 .92
(63) (44)
Creativity g factor .57 .35 .61
(63) (44)

Predictors:

E (Self- & mean peer-reports aggregated) .51 .29 .57

(223) (84)
O (Self- & mean peer-reports aggregated) .56 .25 .45
(223) (84)
C (Self- & mean peer-reports aggregated) .52 .20 .38
(223) (84)
g .80 .46 .58
(154) (127)
gf .73 .41 .56
(164) (129)
gc .76 .46 .61
(164) (129)

Corrected creativity indicators and factors:

T-88 elaboration .39 .46 1.18

(60) (47)
T-88 originality .50 .38 .76
(60) (47)
Mean self-report ‘creative’ .33 .28 .85
(61) (49)
Mean peer-report ‘creative’ .12 .31 2.58
(62) (49)
Mean video ratings ‘creative’ .40 .03 .08
(59) (45)
Creativity factor 1 .33 .25 .76
(53) (43)
Creativity factor 2 .52 .28 .54
(53) (43)
Creativity g factor .53 .14 .26
(53) (43)

Note: MZ = monozygotic, DZ = dizygotic, N in braces.

All analyses were based on values corrected for age and sex effects.
Corrected creativity indicators and factors were additionally corrected for N, E, O, A, C, g, gf, & gc.
Main study: Creativity in BiLSAT and GOSAT 62

Effects of genetic differences can be concluded from higher ICCs in MZ than in DZ

pairs, that is a DZ/MZ ratio below 1 (last column). This was the case for most creativity
variables, except the two T-88 dimensions. DZ/MZ ratios below .5, here only true for
self-reported creativity, indicated non-additive genetic effects (like dominance, epistasis
and emergenesis). Since all other creativity indicators and factors had DZ/MZ ratios
between .5 and 1, genetic as well as shared and non-shared environmental influences
on their variance (to various degrees) can be expected.
On the same rational, twin similarities on all personality dimensions and intelligence
factors indicated genetic and non-shared environmental effects acting on their variance.
Comparisons of MZ-DZ resemblance for Extraversion and the three intelligence factors
also suggested small shared environmental effects; those for Openness to Experience
and Conscientiousness were more in line with an assumption of non-additive genetic
sources of variance. Interesting results emerged for twin correlations of creativity
indicators and factors corrected for all personality dimensions and intelligence factors
assessed in this study: Despite the proposed necessity of some of the controlled
variables for dispositional creativity (e.g. O, P or g), twin pair resemblance for the most
residuals remained greater than zero, some of them even rose. The new pattern of MZ-
DZ ICCs also held some surprises: While T-88 elaboration remained unaffected by
genetic sources, T-88 originality suddenly showed substantial heritability. The contrary
effect was observable for self- and peer-reported creativity as well as creativity factor 1,
whose heritability decreased markedly. It even disappeared completely in peer-reports.
The heritability of creativity factor 2, but especially of the video-based stranger ratings
and the creativity general factor, on the other hand, showed a substantial increase. For
the latter two, this led to an ICC pattern suggestive of non-additive genetic effects.
According to Lykken (1982), at least the DZ/MZ ratio of the video ratings could be
interpreted as indicative of emergenesis.
For a more precise estimate of genetic and environmental influences on individual
differences in these variables, univariate structure equation models were fitted to the
data, using Mx (version 1.3.51; Neale, Boker, Xie & Maes, 2002). The model specified
is depicted in figure 7. Genetic (A), shared environmental (C) and non-shared
environmental (E) effects on the variance of each variable were modeled as latent
constructs, with As being correlated 1.0 for MZ and .5 for DZ twin pairs, while Cs were
set as perfectly correlated and Es as uncorrelated. Squared estimates of the path
coefficients from these constructs on the observed variables (a², c² and e²) can be
Main study: Creativity in BiLSAT and GOSAT 63

interpreted as relative influences of the corresponding latent construct on the variance

of the observed variable in percent.

Figure 7
Univariate Genetic ACE Model

MZ = 1.0, DZ = 0.5 MZ = DZ = 1.0

1.0 1.0 1.0 1.0 1.0 1.0

A C E A C E

a c e a c e

Twin Twin
1 2

For some variables (self-reported creativity, O, C, and both video-based creativity

ratings and the creativity general factor after controlling for all predictors), ICCs were
more indicative of non-additive genetic instead of shared environmental influences.
Since an ACE model as that in figure 7 would have been inappropriate in these cases,
an ADE model was fitted instead, with a latent D construct for dominance effects in
exchange for C. Genetic dominance effects (i.e. intra-loci interactions of alleles) are
completely shared by MZ twin siblings, while DZ twins share them to 25%. The
corresponding path was fixed accordingly (i.e. 1.0 for MZ and .25 for DZ).
Table 27 and 28 show the fit indices for ACE and ADE models and their nested
submodels, respectively. For most variables, highly insignificant p values for the chi-
squared tests and negative AIC fit indices (Aikake’s Information Criterion; χ² - 2df)
indicate that the assigned full model had a good fit to the data. Exceptions were T-88
originality, where the ACE model deviated significantly from the data, aggregated
Extraversion, where the AIC was positive and the chi-squared test approached
significance (p = .08), and aggregated Conscientiousness, where the ADE model
deviated significantly.
Table 27
Fit of Univariate ACE Models to Creativity Indicators, Factors, and Predictors

N (pairs) ACE AE CE E

MZ DZ χ2 p AIC χ2 p LRT p χ2 p LRT p χ2 p LRT p

(df = 3) (df = 3) (df = 3) (df = 4) (df = 1) (df = 5) (df = 2)

Creativity indicators:
T-88 elaboration 222 78 2.94 .40 -3.06 8.70 .07 5.76 .02* 2.94 .57 .00 .99 75.83 .00** 72.89 .00**
T-88 originality 222 78 11.01 .01* 5.01 11.97 .02* .96 .33 11.60 .02* .60 .44 41.65 .00** 30.65 .00**
Peer-report ‘creative’ 316 155 1.78 .62 -4.22 1.91 .75 .13 .72 3.57 .47 1.79 .18 32.34 .00** 30.56 .00**
Video rating ‘creative’ 158 126 3.06 .38 -2.94 4.47 .35 1.41 .24 8.21 .08 5.15 .02* 83.35 .00** 80.29 .00**
Creativity factor 1 63 44 2.77 .43 -3.23 2.77 .60 .00 .99 6.43 .17 3.66 .06 28.61 .00** 25.84 .00**
Creativity factor 2 63 44 1.93 .59 -4.08 3.20 .53 1.28 .26 2.53 .64 .61 .44 26.77 .00** 24.84 .00**
Creativity g factor 63 44 3.31 .35 -2.69 3.32 .51 .01 .94 8.12 .09 4.81 .03* 34.23 .00** 30.92 .00**
Main study: Creativity in BiLSAT and GOSAT

Predictors:
E (aggregated) 223 84 6.77 .08 .77 6.77 .15 .00 # 17.20 .00** 10.43 .00** 81.07 .00** 74.30 .00**
g 154 127 .82 .84 -5.18 1.58 .81 .76 .38 37.37 .00** 36.54 .00** 190.19 .00** 189.36 .00**
gf 164 129 4.58 .21 -1.42 5.33 .26 .75 .39 24.27 .00** 19.69 .00** 153.43 .00** 148.86 .00**
gc 164 129 4.47 .22 -1.53 5.85 .21 1.38 .24 28.35 .00** 23.87 .00** 172.99 .00** 168.51 .00**

Indicators corrected
for all predictors:
T-88 elaboration 60 47 3.27 .35 -2.73 5.48 .24 2.22 .14 3.33 .51 .06 .81 24.07 .00** 20.80 .00**
T-88 originality 60 47 1.12 .77 -4.88 1.52 .82 .40 .53 2.88 .58 1.76 .19 25.39 .00** 24.28 .00**
Self-report ‘creative’ 61 49 2.58 .46 -3.42 3.11 .54 .53 .47 2.75 .60 .17 .68 13.51 .02* 10.94 .00**
Peer-report ‘creative’ 62 49 1.71 .64 -4.29 3.32 .51 1.61 .20 1.71 .79 .00 # 6.62 .25 4.91 .09
Creativity factor 1 53 43 1.14 .77 -4.86 1.40 .84 .26 .61 1.42 .84 .28 .60 10.54 .06 9.40 .01*
Creativity factor 2 53 43 .13 .99 -5.87 .14 1.00 .00 .96 2.59 .63 2.45 .12 19.40 .00** 19.27 .00**

Note: * p < .05, ** p < .01, #: incalculable.

MZ = monozygotic, DZ = dizygotic.
Best fitting model is printed in bold face.
All analyses were based on values corrected for age and sex effects.
Variables corrected for all predictors were additionally corrected for N, E, O, A, C, P, g, gf, & gc.
64
Table 28
Fit of Univariate ADE Models to Creativity Indicators, Factors, and Predictors

N (pairs) ADE AE DE E

MZ DZ χ2 p AIC χ2 p LRT p χ2 p LRT p χ2 p LRT p

(df = 3) (df = 3) (df = 3) (df = 4) (df = 1) (df = 5) (df = 2)

Creativity indicators:
Self-report ‘creative’ 318 161 1.54 .67 -4.46 2.33 .68 .80 .37 2.13 .71 .59 .44 88.59 .00** 87.05 .00**

Predictors:
O (aggregated) 223 84 .66 .88 -5.34 .85 .93 .19 .66 1.65 .80 .99 .32 90.40 .00** 89.74 .00**
C (aggregated) 223 84 7.98 .05* 1.98 8.79 .07 .80 .37 8.25 .08 .26 .61 81.90 .00** 73.91 .00**

Indicators corrected
Main study: Creativity in BiLSAT and GOSAT

for all predictors:

Video ratings ‘creative’ 59 45 1.40 .71 -4.61 2.19 .70 .80 .37 1.40 .85 .00 # 10.96 .05 9.57 .01*
Creativity g factor 53 43 .68 .88 -5.33 1.40 .84 .73 .39 .68 .95 .00 .99 18.97 .00** 18.29 .00**

Note: * p < .05, ** p < .01, #: incalculable.

MZ = monozygotic, DZ = dizygotic.
Best fitting model is printed in bold face.
All analyses were based on values corrected for age and sex effects.
Variables corrected for all predictors were additionally corrected for N, E, O, A, C, P, g, gf, & gc.
65
Main study: Creativity in BiLSAT and GOSAT 66

Fits of more parsimonious nested submodels were tested by fixing one or two paths
from latent constructs to zero. Results yielded AE and CE or DE models as tenable for
most variables, except for E, gf and gc, where AC models were unlikely, and T-88
originality, where all models deviated significantly from the data. The latter resulted from
higher intra-pair resemblance in DZ than in MZ twin pairs, a pattern for which no
reasonable model is known. Therefore, it seems likely that this pattern must be
attributed to sampling error. A model assuming only a non-shared environmental
influence (E model) was only acceptable for corrected peer and video ratings as well as
the corrected first creativity factor.
The best-fitting model was found by testing nested submodels against the full model,
using likelihood ratio tests (chi-squared difference tests; LRT) as well as comparing the
AIC of the models, which is a fit index that takes the parsimony of the model into
account. The final decision (based on the AIC) is set off in bold face in tables 27 and 28.
For each variable, the relative size of the latent constructs’ influence were estimated for
the full and the best-fitting reduced model. These estimates are reported with 95%
confidence intervals in table 29. Unlike in the reduced models, confidence intervals for
the estimates of A and C or D effects in the full models almost exclusively included zero.
Obviously, statistical power was too low to simultaneously support both kinds of effects.
Individual differences in both T-88 scoring dimensions and the resulting creativity factor
2 were best explained by shared and non-shared environmental influences only, with
both being of equal strength for elaboration and the second creativity factor, and one
third shared to two third non-shared environmental influences for originality. On the
other hand, results indicated genetic and non-shared, but no shared, environmental
influences on the remaining creativity indicators and factors. Genetic influences
appeared to be additive for these variables (suggesting an inheritance by quantitative
trait loci; Plomin et al., 1997), with the exception of self-reported creativity, where a
model assuming 49% non-additive dominance effects and 51% non-shared
environmental effects fitted best. Additive genetic effects were strongest on the general
creativity factor (61%) and weakest on peer-reported creativity (29%).
Genetic and solely non-shared environmental effects were also the most likely and
parsimonious explanation for E, O, C and the three intelligence factors, with all genetic
effects but those on C being additive. Parameter estimates indicated roughly half
genetic and half non-shared environmental influences on personality dimensions, and
72% to 80% additive genetic effects on g, gf and gc.
 Main study: Creativity in BiLSAT and GOSAT 67

Table 29
Parameter Estimates from Univariate Analyses

Full model Best fitting nested model

A² c² d² e² a² c² d² e²
Creativity indicators:

T-88 elaboration .00 .46 - .53 - .47 - .53

(.00 - .40) (.09 - .55) (.44 - .63) (.37 - .55) (.45 - .63)
T-88 originality .16 .18 - .66 - .31 - .69
(.00 - .46) (.00 - .40) (.54 - .79) (.20 - .41) (.59 - .80)
Self-report ‘creative’ .23 - .26 .51 - - .49 .51
(.00 - .55) (.00 - .56) (.44 - .60) (.40 - .57) (.43 - .60)
Peer-report ‘creative’ .23 .05 - .71 .29 - - .70
(.00 - .29) (.00 - .31) (.61 - .82) (.19 - .39) (.61 - .81)
Video ratings ‘creative’ .38 .19 - .43 .58 - - .42
(.05 - .65) (.00 - .46) (.34 - .54) (.48 - .66) (.34 - .52)
Creativity factor 1 .56 .01 - .44 .56 - - .44
(.00 - .70) (.00 - .48) (.30 - .64) (.38 - .70) (.30 - .62)
Creativity factor 2 .22 .29 - .49 - .45 - .55
(.00 - .65) (.00 - .58) (.34 - .70) (.29 - .59) (.41 - .71)
Creativity g factor .59 .02 - .39 .61 - - .39
(.06 - .74) (.00 - .46) (.26 - .57) (.44 - .74) (.26 - .56)

Predictors:

E (aggregated) .54 .00 - .46 .54 - - .46

(.22 - .63) (.00 - .28) (.37 - .56) (.44 - .63) (.37 - .56)
O (aggregated) .40 - .17 .43 .57 - - .43
(.00 - .64) (.00 - .64) (.35 - .53) (.47 - .65) (.35 - .53)
C (aggregated) .20 - .34 .46 - - .54 .46
(.00 - .61) (.00 - .62) (.38 - .56) (.44 - .62) (.37 - .56)
g .68 .12 - .20 .81 - .19
(.44 – .84) (.00 - .35) (.15 - .25) (.75 - .85) (.15 - .25)
gf .59 .13 - .28 .72 - - .28
(.32 - .77) (.00 - .38) (.22 - .35) (.65 - .78) (.22 - .35)
gc .59 .17 - .24 .76 - - .24
(.34 - .80) (.00 - .40) (.19 - .31) (.69 - .81) (.19 - .31)

Indicators corrected
for all predictors:

T-88 elaboration .07 .37 - .56 - .42 - .58

(.00 - .60) (.00 - .57) (.38 - .74) (.25 - .57) (.43 - .75)
T-88 originality .38 .16 - .46 .55 - - .45
(.00 - .69) (.00 - .55) (.31 - .68) (.36 - .69) (.31 - .64)
Self-report ‘creative’ .14 .20 - .66 - .31 - .69
(.00 - .54) (.00 - .46) (.46 - .87) (.13 - .47) (.53 - .87)
Peer-report ‘creative’ .00 .21 - .79 - .21 - .79
(.00 - .37) (.00 - .38) (.61 - .98) (.02 - .38) (.62 - .98)
Video ratings ‘creative’ .00 - .38 .62 - - .38 .62
(.00 - .52) (.00 - .56) (.44 - .85) (.15 - .56) (.44 - .85)
Creativity factor 1 .19 .16 - .65 .37 - - .63
(.00 - .55) (.00 - .47 (.45 - .88 (.14 - .56) (.44 - .86)
Creativity factor 2 .50 .02 - .48 .52 - - .48
(.00 – 68) (.00 - .32) (.32 - .71) (.31 - .68) (.32 - .69)
Creativity g factor .01 - .52 .47 - - .53 .47
(.00 - .66) (.00 - .68) (.32 - .69) (.32 - .68) (.32 - .68)
Note: 95% Confidence intervals in braces.
All analyses based were on values corrected for age and sex effects.
Variables corrected for all predictors were additionally corrected for N, E, O, A, C, P, g, gf, & gc.
Main study: Creativity in BiLSAT and GOSAT 68

The surprising effects of statistically controlling personality dimensions and

intelligence factors on the etiology of dispositional creativity reappeared in the
structure equation models: While T-88 elaboration remained best fitted by a CE
model, originality and creativity factor 2 suddenly showed about 50% additive genetic
instead of shared environmental influences. The residuals of creativity self- and peer-
reports were no longer heritable, but influenced by the shared environment (31% and
21%, respectively). Video ratings and the creativity g factor, both formerly determined
to about 60% by additive genetic sources, were determined by non-additive genetic
sources (to 38% and 53%) after all predictors were partialled out. Creativity factor 1
remained additively heritable, though to a slightly lower degree.
Taken altogether, the following picture emerged: Individual differences in T-88 scores
and the resulting creativity factor 1 appeared only determined by environmental
effects, while all indicators based on molar creativity ratings, their resulting factor,
and the common core they share with the T-88 dimensions (the creativity general
factor) were weakly to moderately heritable, but lacked signs of shared environmental
influences. Their assumingly causal predictors were all moderately (personality
dimensions) to highly (intelligence factors) heritable, but also unaffected by the
shared environment. When these predictors were controlled, only creativity self- and
peer-reports lost their genetic foundation, while originality and consequently factor 2
lost environmental influences and became heritable. Genetic effects on the video-
based stranger ratings as well as on the first and the general creativity factor were
slightly reduced, with the video ratings and the g factor no longer suggesting an
additive mode of inheritance.

3.2.5.2 Bivariate behavioral genetic analyses

These rather odd relations between dispositional creativity and its predictors might be
illuminated by fitting bivariate models to the data. The model applied in this study, a
correlated factors model (Neale & Cardon, 1992, p. 252), is shown in figure 8. Its
basic logic is to compare the prediction of trait 1 in one twin from trait 2 in its co-twin
(and vice versa) for MZ and DZ twins. Statistically, it is equivalent to the classic
Cholesky decomposition (Loehlin, 1996), but it allows the estimation of two different
Main study: Creativity in BiLSAT and GOSAT 69

Bivariate Genetic ACE Model

Figure 8
Main study: Creativity in BiLSAT and GOSAT 70

kinds of informative coefficients: First, the genetic, shared environmental and non-
shared environmental correlations (rA, rC and rE), indicative of an overlap of the
corresponding sources of variance, i.e. the degree to which individual differences in
two characteristics are influenced by identical sources. Second, the bivariate
heritabilities and shared and non-shared environmental mediations (bivh², bivc² and
bive²), resulting from the decomposition of phenotypical (observable) covariances of
two characteristics, and indicative of the corresponding latent source’s mediating
effect on the phenotypic relation. Both kinds of coefficients are independent of each
other. For example, two correlated traits might be influenced by exactly the same
genes (perfect rA), but when both traits have very low heritabilities (of whom rA is
independent), their association will not be mediated by this shared genetic foundation
(low bivh², which is ax * ra * ay in figure 8), but by overlapping environmental
influences.
A bivariate genetic model has two minimal requirements: Both traits must be (1)
phenotypicaly correlated (rP), and (2) show influence by the same latent sources. The
first requirement was met by the following relations of interest: Creativity factor 1 with
factor 2, both before and after controlling for all predictors, creativity factor 1 with
each E, O, g and gf (gc was excluded since it did not show a significant zero-order
correlation), creativity factor 2 with E, O, C, g and gc, and the creativity general factor
with E, O, g and gf. C was recoded for this analysis (high values now indicate low
Conscientiousness) to ease the interpretability of results. E, O and C did not show
shared environmental influences in the univariate analyses, therefore partially failing
the second requirement. Thus, only bivariate AE models were fitted here. For all
other analyses, full bivariate ACE models were justifiable. Again, full models were
tested against nested submodels to receive the best-fitting model.
Fit statistics are listed in table 30. Where applicable, the ACE model showed a good
fit to the data, as did the AE model for the remaining cases. However, LRTs
suggested reduced models being superior over the full ACE model, and AICs
indicated an AE model fitting the data best in all cases.
Table 31 gives parameter estimates for the full and the best-fitting reduced model.
Due to limited statistical power, confidence intervals for the full model were again
huge. Consequently, these estimates lacked meaningful interpretability, though they
were already in line with a greater importance of additive genetic effects. In the
following, only results from the better fitting AE model will be discussed.
Table 30
Fit of Bivariate ACE Models to Creativity Indicators, Factors, and Predictors

N (pairs) ACE AE CE E

MZ DZ χ2 p AIC χ2 p LRT p χ2 p LRT p χ2 p LRT p

(df = 11) (df = 14) (df = 3) (df = 14) (df = 3) (df = 17) (df = 6 / 3)

Creativity factor 1 X 2 63 44 15.22 .17 -6.78 17.67 .22 2.44 .49 20.57 .11 5.35 .15 65.19 .00** 49.97 .00**

Creativity factor 1 X E 61 44 - - - 9.11 .82 - - - - - - 45.85 .00** 36.74 .00**

Creativity factor 1 X O 61 44 - - - 14.33 .43 - - - - - - 52.97 .00** 38.64 .00**
Creativity factor 1 X g 55 43 5.70 .89 -16.30 7.40 .92 1.70 .64 19.91 .13 14.21 .00** 99.01 .00** 93.31 .00**
Creativity factor 1 X gf 61 43 11.37 .41 -10.63 13.54 .49 2.17 .54 24.25 .04* 12.88 .00** 96.85 .00** 85.48 .00**

Creativity factor 2 X E 61 44 - - - 9.55 .79 - - - - - - 47.37 .00** 37.82 .00**

Main study: Creativity in BiLSAT and GOSAT

Creativity factor 2 X O 61 44 - - - 14.56 .41 - - - - - - 57.89 .00** 43.33 .00**

Creativity factor 2 X C- 61 44 - - - 10.27 .74 - - - - - - 50.18 .00** 39.91 .00**
Creativity factor 2 X g 55 43 12.50 .33 -9.50 14.92 .38 2.42 .49 22.17 .08 9.66 .02* 101.68 .00** 89.18 .00**
Creativity factor 2 X gc 61 43 9.29 .60 -12.72 11.06 .68 1.77 .62 22.18 .08 12.90 .00** 96.22 .00** 86.94 .00**

Creativity g factor X E 61 44 - - - 9.75 .78 - - - - - - 56.00 .00** 46.25 .00**

Creativity g factor X O 61 44 - - - 20.03 .13 - - - - - - 67.18 .00** 47.15 .00**
Creativity g factor X g 55 43 10.25 .51 -11.75 11.75 .63 1.51 .68 24.85 .04* 14.61 .00** 106.06 .00** 95.81 .00**
Creativity g factor X gf 61 43 13.42 .27 -8.58 14.65 .40 1.23 .75 26.70 .02* 13.28 .00** 102.87 .00** 89.46 .00**

Creativity factor 1 X 2, 53 43 8.51 .67 -13.50 11.63 .64 3.12 .37 13.71 .47 5.21 .16 39.47 .00** 30.97 .00**
all predictors contr.

Note: * p < .05, ** p < .01, #: incalculable.

MZ = monozygotic, DZ = dizygotic.
Best fitting model is printed in bold face.
All analyses based on values corrected for age and sex effects.
Variables corrected for all predictors were additionally corrected for N, E, O, A, C, P, g, gf, & gc.
71
Table 31
Parameter Estimates from Bivariate Analyses

rP Full ACE model Best fitting nested model (AE)

rA rC rE bivh² bivc² bive² rA rE bivh² bive²

Creativity Factor 1 X 2 .29 1.00 -1.00 .09 .34 -.09 .04 .43 .12 .24 .05
(-1.0 -1.0) (-1.0 - 1.0) (-.13 - .33) (117%) (-31%) (14%) (.15 - .67) (-.12 - .34) (83%) (17%)
Creativity Factor 1 X E .28 - - - - - - .15 .40 .07 .21
(-.26 - .47) (.18 - .58) (25%) (75%)
Creativity Factor 1 X O .45 - - - - - - .48 .41 .24 .21
(.18 - .72) (.20 - .58) (53%) (47%)
Creativity Factor 1 X g .27 .55 1.00 -.10 .30 .00 -.03 .44 -.09 .30 -.03
(-.05 - 1.0) (-1.0 - 1.0) (-.34 - .17) (111%) (0%) (-11%) (.20 - .65) (-.33 - .17) (111%) (-11%)
Creativity Factor 1 X gf .33 .90 -1.00 -.09 .44 -.08 -.03 .54 -.07 .36 -.03
Main study: Creativity in BiLSAT and GOSAT

(.24 - 1.0) (-1.0 - 1.0) (-.31 - .15) (133%) (-24%) (-9%) (.32 - .73) (-.30 - .17) (106%) (-9%)
Creativity Factor 2 X E .22 - - - - - - .13 .31 .06 .16
(-.26 - .45) (.08 - .50) (27%) (73%)
Creativity Factor 2 X O .38 - - - - - - .36 .40 .18 .20
(.04 - .61) (.19 - .57) (47%) (53%)
Creativity Factor 2 X C- .12 - - - - - - .18 .06 .09 .03
(-.16 - .50) (-.17 - .29) (75%) (25%)
Creativity Factor 2 X g .25 .21 .62 .07 .08 .15 .02 .36 .05 .23 .02
(-1.0 - 1.0) (-1.0 - 1.0) (-.19 - .33) (32%) (60%) (8%) (.12 - .58) (-.20 - .30) (92%) (8%)
Creativity Factor 2 X gc .31 .12 1.00 .21 .05 .20 .06 .38 .18 .25 .06
(-1.0 - 1.0) (-1.0 - 1.0) (-.04 - .44) (16%) (65%) (19%) (.14 - .59) (-.06 - .41) (81%) (19%)
Creativity g Factor X E .32 - - - - - - .16 .48 .08 .24
(-.23 - .45) (.28 - .64) (25%) (75%)
Creativity g Factor X O .52 - - - - - - .50 .54 .26 .26
(.23 - .71) (.35 - .68) (50%) (50%)
Creativity g Factor X g .32 .44 1.00 -.03 .25 .08 -.01 .47 -.03 .33 -.01
(-.09 - 1.0) (-.71 - 1.0) (-.28 - .23) (78%) (25%) (3%) (.25 - .66) (-.28 - .22) (103%) (-3%)
Creativity g Factor X gf .34 .67 1.00 -.08 .36 .00 -.02 .55 -.07 .36 -.02
(.13 - 1.0) (-1.0 - 1.0) (-.31 - .17) (106%) (0%) (-6%) (.34 - .73) (-.30 - .17) (106%) (-6%)
Creativity Factor 1 X 2, .21 1.00 -1.00 .06 .30 -.13 .04 .37 .07 .17 .04
all Predictors contr. (.13 - 1.0) (-1.0 - 1.0) (-.17 - .29) (143%) (-62%) (19%) (-.03 - .76) (-.18 - .31) (81%) (19%)
72

Note: 95% Confidence intervals in braces.

All analyses were based on values corrected for age and sex effects.
Variables corrected for all predictors were additionally corrected for N, E, O, A, C, P, g, gf, & gc.
Main study: Creativity in BiLSAT and GOSAT 73

Note that phenotypic correlations (rP) in table 31 differ somewhat in size from those
reported in tables 17 and 18 because of small sample variations and a different
calculation procedure.
The additive genetic sources acting on creativity factor 1 and 2 correlated moderately
(.43). This considerable overlap resulted in 83% of the phenotypic correlation being
mediated genetically. After controlling all predictors in both factors, the latent pattern
remained unaltered, only the phenotypic correlation was lower. The etiology of the
relation between E and O on the one side and the three creativity factors on the other
was very similar: Genetic correlations between E and all three factors were low (.13
to .16, while genetic correlations between O and all factors, as well as non-shared
environmental correlations between both E and O and all factors were moderate (.31
to .54). The phenotypic correlation of E with the creativity factors (.22 to .32) was
always mediated to one quarter by genetic and to three quarters by non-shared
environmental effects. The higher observed correlation between O and the factors
(.38 to .52), on the other hand, was mediated by equally by genetic and non-shared
environmental effects in each case. C shows only weak genetic and environmental
correlations with the second creativity factor, but the phentotypic correlation was also
small (.12). It decomposed into three quarters genetic and one quarter non-shared
environmental mediation. The genetic correlations of g and its facet gc were .36 and
.38, respectively, while the non-shared environmental ones were low (.18) for gc and
effectively zero for g. Therefore, genetic effects were responsible for 92% of the
correlation with g and 81% of those with gc. Similarly, the genetic correlations of g
and gf with the first and the general creativity factor were moderate (.44 to .55), while
the shared environmental ones were even slightly negative. This resulted in genetic
effects being the only link between these factors, i.e. a genetic mediation of 100%.
It has to be remarked that, because of the small sample size for a multivariate
analysis, most confidence intervals included zero. Taking them into account, what
can be safely concluded is that the genes influencing Openness to Experience as
well as general intelligence and its facets, and the non-shared environmental effects
acting on Openness and Extraversion, overlapped with those influencing
dispositional creativity. On the same rational, partly identical genes influenced both
creativity factor 1 and 2, at least as long as personality and intelligence predictors
were not controlled.
Main study: Creativity in BiLSAT and GOSAT 74

3.3 Discussion

Supporting hypothesis 1, all indicators of a disposition towards creativity used in this

study showed significant intercorrelations. Even though the T-88 was a derivate of an
divergent test, all indicators were based on subjective ratings of creativity or some of
its components, therefore following Amabile’s (1982) consensual definition. The
indicators were very different in the kind, amount and standardization of information
accessible to the judges, as well as the length of acquaintanceship between judge
and target subject: They ranged from self ratings to ratings of well acquainted peers,
which had plenty of opportunities to form themselves a picture of the subject over
many years, to strangers who saw the subject for the first time in a few minutes long,
standardized situation on video tape, to raters who had nothing more than eighteen
entitled line drawings of a subject they never met. Additionally, different indicators
were assessed at different times, with about five years lying between the first (self-
and peer-reports) and the last (T-88) assessment of creativity. The indicators
themselves were aggregates of one to sixty independent judges, all of which showed
considerable consensus. That creativity judgments of normal adult persons under all
these diverse conditions had enough in common to allow the extraction of a general
factor which explained over 40% of the indicators’ variance is indeed impressive.
Because of the variety of measurement methods and times contributing to it, diverse
sources of measurement error have likely been attenuated. This result leaves no
doubt that something like a relatively stable disposition towards creativity does exist
in normal people. That this general disposition splitted into two factors has several
possible explanations: Compared to the self and peer ratings marking creativity factor
1, the T-88 marking factor 2 was later assessed, more divergent test-like (though not
identical, see chapter 2) and based on fewer and more standardized information. The
video-based stranger ratings, which showed substantial loadings on both factors, laid
in between.
In line with hypothesis 2, all three creativity factors correlated low to moderately with
general intelligence. This result underlines that creativity and intelligence are far from
being identical, but also contradicts their fundamental independence aside from
professional requirements, as suggested in the ‘certification hypothesis’ (Hayes,
1989). The differentiation of general intelligence into its fluid and crystallized facets
shed a bit light into the distinction of the two creativity subfactors: Only fluid
Main study: Creativity in BiLSAT and GOSAT 75

intelligence was supportive for the creativity aspects represented by factor 1 (with
genuine crystallized intelligence even having a slight detrimental effect), and only
crystallized intelligence affected factor 2. Paralleling this, creativity self-concept and
the peer perceptions, which were the markers of the first factor, were only (weakly)
affected by genuine fluid intelligence, while the two T-88 scores, marking factor 2,
showed a stronger influence of crystallized intelligence aspects. The video-based
stranger ratings, which loaded on both factors, were equally influenced by all facets
of general intelligence. Since, as the regression analyses revealed, personality
influences were rather similar for both factors (except for low C acting on factor 2,
see below), different demands on fluid relative to crystallized intelligence might have
been the prime difference between the two creativity factors found in this study. Mere
effects of measurement time, on the other hand, seem unlikely. The demands of the
creativity indicators might be better candidates: Maybe more standardized, test-like
conditions, especially when responses are judged by a stranger (to whom your old
ideas and jokes are unfamiliar), allow for creative solutions to appear out of old
knowledge (gc), while observed real-life behavior in spontaneous tasks and
interactions does not, but instead requires quick combination of various information
(gf). The antagonistic effects of gf and gc found in both factors is in line with this
interpretation, also confirming the negative transfer effect of gc on creativity proposed
by Cattell (1971). The general creativity factor was subsequently affected by both
intelligence facets, with a slight superiority of genuine fluid over genuine crystallized
aspects. Interestingly, these results support Cattell’s (1971) suggestion of fluid
intelligence as the fundament of everyday creativity, but also explain the findings of
Crawford (1974) and Crawford and Nirmal (1976), which showed divergent tests
being stronger related to crystallized intelligence. The popular ‘threshold hypothesis’,
on the other hand, was again not supported in this study: The relationship between
general intelligence and all three intelligence factors was perfectly linear.
A very clear picture emerged for the strongest relation of a personality dimensions to
dispositional creativity: As predicted, Openness to Experience showed consistently
the highest correlations with all creativity factors, making it the best single predictor of
a disposition towards creativity. The second best, incremental personality predictor
was Extraversion, except for creativity factor 2, where low Conscientiousness did
even better. It is interesting that the supportive effect of low C seems to go hand in
hand with that of gc. A possible interpretation would be that learned knowledge is
Main study: Creativity in BiLSAT and GOSAT 76

only able to bear new, creative outcomes when handled in a playful, unorganized, not
very goal-directed or even mildly chaotic manner. Extraversion, on the other hand,
seemed to be especially important for creativity indicators based on directly observed
behavior. Its unique contribution might therefore be due to a facilitation of displaying
one’s ideas to an audience. This would mirror Feist’s (1998) finding of the
confidence/dominance aspect of Extraversion being more important than the
sociability aspect for professional creative achievement. But since the results of this
study were not sufficiently differentiated, such an interpretation remains speculative.
On the other hand, communality analyses did show that not the common core of
Extraversion and Openness to Experience, but the Openness variance independent
of Extraversion was most predictive for all three creativity factors. If indeed a
dopaminergic neuronal system, mediated by latent inhibition, underlies the common
core of Openness and Extraversion (Peterson et al., 2002), than it can only be a
partial explanation of why Openness relates to dispositional creativity. Since the other
personality dimension proposed to relate to both dopamine/latent inhibition and
dispositional creativity, Psychoticism (Eysenck 1993, 1995a, b), did even worse as a
predictor of dispositional creativity, as did synergistic effects of these dimensions with
general intelligence, we are left with the latent inhibition mechanism being only an
incomplete account for dispositional creativity, and we are left with only an partial
explanation of why Openness to Experience does relate to dispositional creativity. I
will return to the latter point in the general discussion (chapter 4). Anyway, these
conclusions are only indirect, since latent inhibition was not measured in this study. A
further comment has to be made about Psychoticism, since two more technical
explanations for its failure to relate to creativity in this study are possible: First,
reliabilities (internal consistencies) of both self- and peer-reports were at the lower
edge of what is acceptable for a scientific study. But since correlations with creativity
were virtually zero, disattenuating them cannot be expected to make much
difference. Second, a more critical point can be made about the EPQ P scale in
general: As noted by Harrington (1993), its unclear relation to psychotic symptom
susceptibilities (like latent inhibition), diverse item content, multifactorial nature and,
maybe most important, substantial revision in 1985 (Eysenck, Eysenck & Barrett,
1985), make the evidence about its relations to creativity hard to interpret. To my
knowledge, only Wuthrich and Bates (2001) and the present study related the revised
version of the P scale to a non-achievement measure of creativity – both with
Main study: Creativity in BiLSAT and GOSAT 77

negative results. The meta-analytic association of creative achievement and P found

by Feist (1998), assumingly incorporating results from both versions of the P scale,
might therefore indicate that (at least) the revised version measures something that
facilitates a creative career, not creativity per se.
While in this study certain personality and ability dimensions explained up to one third
of the variance in the dispositional creativity factors, at least two thirds of these
relatively unbiased estimates (see above) remained unexplained by them. Indeed,
controlling for all the comprehensive dimensions of individual differences assessed in
this study did somewhat reduce the intercorrelations of the creativity indicators (which
can be taken as pseudo-criteria for each other), but did not alter their factor structure.
Since most residuals did not drop in heritability (this was only true for self- and peer-
reported creativity and consequently factor 1), but even gained genetic influence (T-
88 originality and factor 2) or suddenly exhibited signs of non-additive inheritance
(video ratings and the general factor), the controlled personality and ability
dimensions cannot be regarded as necessary determinants of a disposition towards
creativity.
Behavioral genetic analyses revealed a complex etiology for dispositional creativity.
In one of its aspect (factor 1), it was very similar to what is often found for personality
traits – roughly 50% heritability, no effects of the shared, but substantial ones of the
non-shared environment. The other aspect (factor 2) showed – in line with previous
results on divergent tests – remarkable environmental effects of both the shared and
non-shared kind (especially true for the elaboration component measured by the T-
88) and only weak genetic influences, which failed to reach significance because of a
lack of statistical power. But since bivariate analyses revealed a mainly additive
genetic link between both creativity subfactors, their existence is likely. Additional
evidence stem from the sudden appearance of genetic influences in substantial
magnitude (combined with a disappearance of shared environmental effects) after
the statistical control of personality and intelligence dimensions. However, this latter
result is hard to interpret, especially since all controlled influences showed mainly
additive genetic, but not shared environmental, associations with factor 2 in bivariate
analyses. It surely awaits replication before a definite interpretation can be given. The
consistent shared environmental effect on T-88 elaboration is special, since such
effects are relatively pure estimates (heritabilities also include gene x shared
environmental interactions, while non-shared environmental effects also include gene
Main study: Creativity in BiLSAT and GOSAT 78

x non-shared environmental interactions and measurement error; Purcell, 2002),

which hold the promise of leading to isolatable critical environmental factors. Traits
showing strong shared environmental are seldom found, but these results make a
disposition to develop and enrich one’s ideas a likely candidate.
That creativity factor 1 and its indicators showed an etiological pattern typical for
personality traits is not very surprising, since they were measured with methods most
typical for personality research. Factor 1 simply reflected one’s self-concept and
social appearance as creative. It must be remarked that the measurement of its
indicators with single, direct items is no disadvantage, because people have a valid
implicit concept of who is a creative person, and creativity research has so far failed
to provide something better (see section 1.1). Influences of fluid intelligence on factor
1 were completely genetic, and also Openness to Experience and Extraversion
contributed to it via shared genes, though common influences of the non-shared
environment were also important here (especially for Extraversion). Controlling them
reduced the heritability of factor 1 and revealed some weak shared environmental
effects on the creativity self-concept and peer evaluation.
Arguably interpretable as the very core disposition towards creativity, the general
factor showed a considerable heritability of 61%. Such a strong genetic influence
differed markedly from what has been reported for creativity so far (section 1.5) and
underlines that divergent test results should not simply be taken as representative for
creativity. Influences of its predictors fluid and general intelligence, Openness to
Experience and Extraversion took the same pathways in a similar pattern as they did
in the case of the first creativity factor, especially emphasizing shared genes with
general intelligence and Openness. However, controlling these influences had
substantial reductive effects on the pair resemblance of DZ, but not MZ twins, and
thus brought forth further non-additive genetic effects. The video ratings exhibited a
similar pattern. Since assortative mating effects can be expected for dispositional
creativity, it is likely that all genetic effects reported here were attenuated, while
shared environmental ones were inflated (confirm section 1.5). Taking this into
account would suggest that the genetic effects on general dispositional creativity
beyond those attributable to general intelligence, Openness and Extraversion were
unshared by relatives except identical twins, i.e. they were emergenic. Note that this
result differs markedly from those of Canter (1973), who found very similar
correlations for MZ and DZ twin pairs in divergent test performance after controlling
Main study: Creativity in BiLSAT and GOSAT 79

for general intelligence (separate analyses not reported so far indicated that the
general creativity factor remained heritable to 55% when only general intelligence
was controlled).
To summarize, what made the subjects of this study creative was largely general
intelligence, Openness to Experience, Extraversion, a bit low Conscientiousness and
some unspecified idiosyncrasies stemming from synergistically interacting genes. But
most importantly, dispositional creativity was substantially heritable. Since this result
clearly contradicts the current view, I hope it encourages further inquiries into this
underdeveloped area.
Several limitations of this study should also be mentioned. First of all, while this study
already used multiple indicators of dispositional (everyday) creativity, further ones,
like assessments of creative hobbies, activities and non-professional
accomplishments (see Hocevar, 1980; Carson et al., 2003; King et al., 1996; Wolfradt
& Pretz, 2001) or retrospective reports of one’s lifetime creativity (Richards et al.,
1988) would be helpful for getting a less biased and more valid criterion. Such an
approach would be especially urgent for behavioral genetic studies, where something
like this has never been done. I doubt, however, that such methods can be valid
without relying on subjective judgments of creativity, an aspect excessively
considered in the present study. On the other hand, unlike most studies, the present
one assessed personality via two peer-reports additionally to self-reports. Method
effects were therefore much better controlled than in the majority of previous
attempts to relate creativity to personality. Behavioral genetic studies of adult
personality have also been criticized for relying solely on self-reports (Brody, 1993),
making the present etiological inquiry into the relation of creativity to adult personality
even more valuable. Second, like in most twin studies, a larger number of twin pairs
would have allowed to receive more reliable parameter estimates in the structure
equation models. Third, measurement points for creativity indicators and predictors
were spread over several years in an rather arbitrary manner. Since no
developmental perspective was taken in here, a concurrent assessment of all
variables would have been more appropriate for the aims pursued in this study. That
the results came out as clearly as they did is therefore a positive surprise. Fourth, a
removal of the T-88’s uncontrollable time limit would have been a definite
improvement. Studies like those of Ferris and colleagues (1971) have shown that
both controlled, test-like and relaxed, untimed administrations of tests comparable to
Main study: Creativity in BiLSAT and GOSAT 80

the T-88 yielded valid results, but the same is unclear for a mix of administration
conditions as the one that has happened here. Since test sheets suggested some
individual differences in dealing with it, the T-88 results should interpreted with some
care (though no hint to any form of directional error resulting from it was found in the
analyses). Finally, all predictors of a disposition towards creativity were of a very
broad kind. While this allowed a rather comprehensive overview of cognitive and
personality effects acting on dispositional creativity, more fine-grained influences
were blurred. These include a lack of insight into the relative effect of
confidence/dominance and sociability aspects of Extraversion and the roles of
classical divergent thinking, traits like sensation seeking or tolerance for ambiguity,
and motivational dispositions. Facet-level analyses of the relevant NEO-PI-R scales
would surely have been an informative first step in this direction.
General discussion and conclusion 81

4. General discussion and conclusion

In front of the results, we can now reconsider the central question from the very
beginning: What makes a person creative? As already said in the last section, the
superficial answer for everyday creativity in normal people that has emerged is: To
over 60% his or her genes and to less than 40% his or her non-shared environment.
Mainly a high Openness to Experience and general intelligence, plus a bit
Extraversion and a pinch of low Conscientiousness. But why? A high Openness to
Experience, which without doubt holds the lion’s share, might be supportive in
several ways: First, it might be an indicator of increased dopaminergic activity in the
frontal lobe and mesolimbic system, resulting in a lowered latent inhibition that allows
more information to enter the focus of attention. This additional information can,
following Campbell (1960) and Simonton (1988, 1999b), be beneficial for more and
more unusual ideational variation as a result of blind mental combinatory processes,
which can be selectively retained as creative sparks. But as mentioned before, if this
is only true for the common core of Openness and Extraversion, it explains only a
part of the Openness influence. Other important elements of Openness to Experience
include intellectual curiosity, active imagination, unconventional attitudes,
preferences for novelties and independence in opinion-formation (McCrae, 1994;
Angleitner & Ostendorf, 2003), all of which are supportive for entering more and more
unusual information into the blind mental combinatory process, though in a more
active manner than simply perceiving more information from the environment.
Additionally, Openness to Experience is related to divergent thinking, which is mainly
ideational fluency (Hocevar, 1979a, b). Fluency can be seen as the speed of the
variation generator. Here, we slowly reach common ground with general intelligence.
The fluid aspect of general intelligence is beneficial because it allows the extraction
of more complex information from the environment, faster learning of new information
in new environments, the faster combination (mental speed) of more information units
(working memory capacity), and, last but not least, a better evaluation of what is a
reasonable result in the selective-retention process. The crystallized facet, on the
other hand, is responsible for the amount of older information that has been learned
earlier and can now be entered into the blind combination process. However, as
remarked by Cattell (1971), high crystallized intelligence is not solely beneficial, since
it might lead to knowing an existing solution or behavior for most situations, therefore
General discussion and conclusion 82

hindering that a creative one results from fluid intelligence. Low Conscientiousness
can introduce informational variation through its unsteady, chaotic and unsystematic
nature, though a lack of goal-directedness and volition might be detrimental even for
small creative acts in everyday life. Therefore, its net effect is only weak.
Extraversion’s influence might either function via low latent inhibition, confidence and
dominance to display one’s creative ideas, or it might facilitate the generation of
ideational variation through positive affectivity (Isen, Daubman & Nowicki, 1987): A
state of positive affect is marked by low cortical arousal, yielding an attenuation of
dominant response patterns (Hull, 1943) and therefore allowing for creative ones.
While these personality and intelligence dimensions surely are the most important
(and supposingly most common) single determinants of what makes a person
creative, the results this study suggest that meaningful individual differences in
everyday creativity exist independent of these factors. The behavioral genetic
analyses have revealed that about half of these remaining differences stem from non-
shared environmental influences, which could include pure chance events of
creativity or be due to other hardly reconstructable causes. The other half, in
contrast, appeared to be under the influence of genetic effects only shared by
individuals with identical genome. What follows is that a substantial amount of
creative behavior observable in everyday life results from behavioral idiosyncrasies of
people, which in turn stem from interactions of heritable traits and individual genes.
Despite the lack of an association between Eysenck’s Psychoticism dimension and
creativity in this and several other studies (confirm section 1.3), and contrary to
humanistic perspectives on creativity (Woodman, 1981), prime candidates remain
various psychopathological disorders. Severe forms of disorders do without doubt
lead at best to ‘novel’ ideas that lack usefulness or appropriateness – and therefore
creativity. But there exist plenty of evidence from diverse sources that suggest the
link between genius and madness (which goes far back to Aristotle) being more than
just a myth (Simonton, 1999b; Prentky, 1989; Eysenck, 1995a). Both Eysenck and
Simonton propose that mild forms of disorders have to pair up with high levels of ego-
strength to allow for creativity to emerge. According to them, this is because the
random ideational variation provided by the pathology, when controllable and
selectively retained, can act as ideational raw material for creative products. Such
interactions might therefore be possible pathways to idiosyncratic creativity. But they
are surely not the only ones. As the Minnesota study of twins reared apart has shown
General discussion and conclusion 83

(Lykken et al., 1992), complex patterns of polymorphic behavior-influencing genes

open up a universe of unique (except for one’s MZ co-twin) behaviors and (via gene x
environment interactions and correlations) experiences (Plomin et al., 1997;
Bouchard, Lykken, Tellegen & McGue, 1996), all of which have the potential to
facilitate the generation of novel and useful combinations of information.
This brings us back to a question I have raised on the very first page, but haven’t
reconsidered since: The heterogeneity of creativity. Contrary to the popular picture of
creativity as a single, homogeneous trait in naïve - and often enough scientific -
psychology, it now appears to be a heterogeneous disposition with multiple sufficient,
but no necessary (though some fairly common) determinants (confirm Asendorpf,
1999, p. 178). Thus, what we are left with is a secondary Darwinian process at the
very core of creativity, and various individual differences adjusting the amount and
diversity of elements it can randomly combines as well as the efficiency with which it
selects results that fit the task or situation. The heterogeneity of personality factors
potentially contributing to a disposition towards a stochastic phenomenon called
creativity (Simonton, 2003) might be the reason why creative genius has never lost
the veil of mysticism which cloaked it ever since the term ‘genius’ was coined in
ancient Greek (where it labeled the ‘Daimon’ or ‘guardian spirit’ that possessed all
those with unexplainable creative abilities; Albert & Runco, 1999).
A final comment has to be added about the double role of general intelligence in the
creative process: Unlike all the other potential determinants of dispositional creativity
discussed in this thesis, general and especially fluid intelligence do not only facilitate
the generation of ideational variation, but are of prime importance for the selective
retention process. The level of reasoning ability a creator posses determines the
complexity and amount of ideational combinations he can evaluate and the degree of
fit he can achieve to often complex demands of the environment, task, fashion or
existing information structure. A (theoretical) complete lack of general intelligence
would result in arbitrary productions that completely lack fit – useless, inappropriate
and therefore uncreative. This might indeed be the case in schizophrenia, where the
evaluative mechanism is overwhelmed by the amount of generated variation (see
e.g. Carson et al., 2003). What it implies is that, unlike all the other determinants, a
certain amount of general intelligence cannot be substituted. The necessary amount
is given by the field of creative endeavor: Completing a line drawing is fairly
unconstrained, while making up an excuse why one is unable to help a friend (as in
General discussion and conclusion 84

one of the GOSAT video situations) is already constrained by rules of social behavior
and perhaps by the cleverness of other person. Spontaneous jokes or witty
comments normally require the quick integration of a new ideational combination into
the actual situation in a surprising, yet meaningful manner. Pieces of art have to fit
aesthetic preferences and often defining features of a style or fashion - acquired
knowledge that has to be kept in mind as an evaluative standard for every stroke of
the brush. And having a creative scientific insight requires picking the meaningful
complex of information that fits a huge net of empirical evidence out of random
combinations of diverse knowledge that float in the scientist’s stream of
consciousness.
Surely most of the creative tasks the subjects in the main study had to complete (and
also the vast majority of divergent tests) do not require a high level of reasoning
ability to manage the evaluative part of the creative process. It is unlikely that this
part imposed high cognitive demands on the subjects. Thus the beneficial effect of
general and fluid intelligence on dispositional creativity found in the main study might
nearly exclusively stem from its facilitating effects on the variation generation
process, making it replaceable. More complex forms of creative activities –
professional as well as recreational ones – should be more dependent on a minimal
level of general (fluid) intelligence. This should not be understood as an attempt to
reintroduce the ‘threshold’ or the ‘certification hypothesis’ through the backdoor: In
any case, more general intelligence is more beneficial in one or both ways it
influences the creative process, without an upper limit. But what it implies is, given a
minimal complexity level of the creative activity (i.e. demands of the activity to fit
one’s ideas to existing information) and an unrestricted range of both creativity and
intelligence, the only thing that can be reliably inferred about a person from his
creative performance is his level of general intelligence. This makes most creative
performances, when compared to the whole population, valid cues of general
intelligence. I therefore agree with the conclusion of Haensly and Reynolds (1989, p.
130): “Yet each creative act may be the ultimate expression of intelligence, in which
all of the cognition and comprehension that individuals have developed at that point
in their time (age) and situation (context) with their degree of training (experience)
have been brought to bear upon a particular idea or problem. We propose that
creativity is not another “breed” of mental processing, but is the ultimate expression
of that finely honed system of thinking we know as intelligence.”
References 85

5. References

Aguilar-Alonso, A. (1996). Personality and creativity. Personality and Individual

Differences, 21 (6), 959-969.
Albert, R. S., & Runco, M. A. (1999). A history of research on creativity. In R. J.
Sternberg (Ed.), Handbook of creativity (pp. 16-31). Cambridge: University Press.
Amabile, T. M. (1982). Social psychology of creativity: A consensual assessment
technique. Journal of Personality and Social Psychology, 43 (5), 997-1013.
Amabile, T. M. (1983a). The social psychology of creativity. New York: Springer.
Amabile, T. M. (1983b). The social psychology of creativity: A componential
conceptualization. Journal of Personality and Social Psychology, 45 (2), 357-376.
Amelang, M., & Bartussek, D. (1997). Differentielle Psychologie und Persönlichkeits-
forschung (4. Auflage). [Psychology of individual differences and personality
research (4th edition).] Stuttgart: Kohlhammer.
Anastasi, A., & Schaefer, C. E. (1971). The Franck Drawing Completion Test as a
measure of creativity. The Journal of Genetic Psychology, 119, 3-12.
Angleitner, A., & Ostendorf, F. (1994). Von aalglatt bis zynisch: Merkmale
persönlichkeitsbeschreibender Begriffe. [From slippery as an eel to cynical:
Characteristics of personality-descriptive terms.] In W. Hager & M. Hasselhorn
(Eds.), Handbuch deutschsprachiger Wortnormen (pp. 340-381). [Handbook of
German word norms.] Göttingen: Hogrefe.
Angleitner, A. (2002). Personality in Adulthood: First findings from the Bielefeld
Longitudinal Study of Adult Twins (BiLSAT). Talk given at the 11th European
Conference of Personality of the EAPP, Jena, Germany, 21. - 25. 07. 2002.
Angleitner, A., & Ostendorf, F. (2003). NEO Persönlichkeitsinventar (revidierte Form,
NEO-PI-R) nach Costa und McCrae. [NEO Personality Inventory (revised
version, NEO-PI-R) after Costa and McCrae.] Göttingen: Hogrefe.
Anton, F., Griepenstroh, J., Poggenpohl, H., & Rothenpieler, S. (2003). Zum
Zusammenhang von Persönlichkeit und Einfallsreichtum. [Towards the relation of
personality and imaginativeness.] Unpublished practical report, University of
Bielefeld.
Asendorpf, J. B. (1999). Psychologie der Persönlichkeit (2. Auflage). [Psychology of
personality (2nd edition).] Berlin: Springer.
References 86

Austin, J. H. (1978). Chase, chance, and creativity. New York: Columbia University
Press.
Backhaus, K., Erichson, B., Plinke, W., & Weiber, R. (2000). Multivariate
Analysemethoden. 9. Auflage. [Multivariate analyse methods. 9th edition.] Berlin:
Springer.
Barron, F. X. (1955). The disposition towards creativity. Journal of Abnormal and
Social Psychology, 51, 478-485.
Barron, F. X. (1958). The psychology of imagination. Science Monthly, 199 (3), 150-
166.
Barron, F. X. (1968). Creativity and personal freedom. Princton, NJ: Van Nostrand.
Barron, F. X., & Harrington, D. M. (1981). Creativity, intelligence, and personality.
Annual Review of Psychology, 32, 439-476.
Baruch, I., Hemsley, D. R., & Gray, J. A. (1988). Latent inhibition and “psychotic
proneness” in normal subjects. Personality and Individual Differences, 9, 777-
783.
Becker, A., Busjahn, A., Faulhaber, H.-D., Baehring, S., Robertson, J., Schuster, H.,
& Luft, F. C. (1997). Twin zygosity: Automated determination with microsatellites.
Journal of Reproductive Medicine, 42, 260-266.
Borkenau, P., Riemann, R., Angleitner, A., & Spinath, F. M. (2001). Genetic and
environmental influences on observed personality: Evidence from the German
Observational Study of Adult Twins. Journal of Personality and Social
Psychology, 80 (4), 655-668.
Bortz, J. (1999). Statistik für Sozialwissenschaftler. [Statistics for social scientists.]
Berlin: Springer.
Bouchard, T. J., jr., Lykken, D. T., Tellegen, A., & McGue, M. (1996). Genes, drives,
environment, and experience: EPD theory revised. In C. Benbow & D. J. Lubinski
(Eds.), Intellectual talent: Psychometric and social issues (pp. 5-43). Baltimore,
MD: John Hopkins University Press.
Bramwell, B. S. (1948). Galton’s ‘Hereditary Genius’ and three following generations
since 1869. Eugenics Review, 39, 146-153.
Brody, N. (1993). Intelligence and the behavioral genetics of personality. In R. Plomin
& G. E. McClearn (Eds.), Nature, nurture, and psychology (pp. 161-178).
Washington, DC: American Psychological Association.
References 87

Brown, R. T. (1989). Creativity: What are we to measure? In J. A. Glover, R. R.

Ronning & C. R. Reynolds (Eds.), Handbook of creativity (pp. 3-32). New York:
Plenum Press.
Bullough, V., Bullough, B., & Mauro, M. (1981). History and creativity: Research
problems and some possible solutions. Journal of Creative Behavior, 15, 102-
116.
Burks, B., Jensen, D., & Terman, L. M. (1930). The promised youth. Stanford:
University Press.
Campbell, D. T. (1960). Blind variation and selective retention in creative thought as
in other knowledge processes. Psychological Review, 67 (6), 380-400.
Campbell, D. T. (1974). Unjustified variation and selective retention in scientific
discovery. In F. J. Ayala & T. Dobzhansky (Eds.). Studies in the philosophy of
biology (pp. 131-149). London: Macmillan.
Canter, S. (1973). Personality traits in twins. In G. Claridge, S. Canter & W. I. Hume
(Eds.), Personality differences and biological variations (pp. 21-51). New York:
Pergamon Press.
Carroll, J. B. (1993). Human cognitive abilities. Cambridge: University Press.
Carson, S., Peterson, J. B., & Higgins, D. M. (2003). Decreased latent inhibition is
associated with increased creative achievement in high-functioning individuals.
Journal of Personality and Social Psychology, 85 (3), 499-506.
Cattell, R. B. (1963). Theory of fluid and crystallized intelligence: A critical
experiment. Journal of Educational Psychology, 54, 1-22.
Cattell, R. B. (1971). Abilities: Their structure, growth, and action. Boston: Houghton
Mifflin.
Cattell, R. B., Eber, H. W., & Tatsuoka, M. M. (1970). The handbook for the Sixteen
Personality Factor (16PF) Questionnaire. Champaign, IL: Institute for Personality
and Ability Testing.
Cattell, R. B., & Warburton, F. W. (1967). Objective Personality & Motivation Tests. A
Theoretical Introduction and Practical Compendium. Urbana: University of Illinois
Press.
Cohen, J. (1977). Statistical power analysis for the behavioral sciences (revised
edition). New York: Academic Press.
References 88

Costa, P. T., & McCrae, R. R. (1992). Revised NEO Personality Inventory (NEO-PI-
R) and NEO Five-Factor Inventory (NEO-FFI) professional manual. Odessa, FL:
Psychological Assessment Resources.
Cooley, W. W., & Lohnes, P. R. (1976). Evaluation Research in Education. New
York: Irvington.
Cox, C. (1926). The early mental traits of three hundred geniuses. Stanford:
University Press.
Crawford, C. B. (1974). Fluid and crystallized intelligence, creativity and achievement
in elementary school children. Journal of Behavioral Science, 2 (2), 43-48.
Crawford, C. B., & Nirmal, B. (1976). A multivariate study of measures of creativity,
achievement, motivation, and intelligence in secondary school students.
Canadian Journal of Behavioral Science, 8 (2), 189-201.
Cronbach, L. J. (1957). The two disciplines of scientific psychology. American
Psychologist, 12, 671-684.
Csikszentmihalyi, M. (1999). Implications of a systems perspective for the study of
creativity. In R. J. Sternberg (Ed.), Handbook of creativity (pp. 313-335).
Cambridge: University Press.
Cziko, G. A. (1995). Without miracles. Universal selection theory and the second
Darwinian revolution. Cambridge, MA: MIT Press.
Cziko, G. A. (1998). From blind to creative: In defense of Donald Campbell's
selectionist theory of human creativity. Journal of Creative Behavior, 32 (3), 192-
209.
Darwin, C. (1859). On the origin of species by means of natural selection. London:
John Murray.
Dawkins, R. (1986). The blind watchmaker: Why the evidence of evolution reveals a
universe without design. New York: Norton.
De la Casa, L. G., Ruiz, G., & Lubow, R. E. (1993). Latent inhibition and
recall/recognition of irrelevant stimuli as a function of preexposure duration in
high and low psychiotic-prone normal subjects. British Journal of Psychology, 84,
119-132.
Detterman, D. K., & Daniel, M. H. (1989). Correlations of mental tests with each other
and with cognitive variables are highest for low IQ groups. Intelligence, 13, 349-
359.
References 89

Diehl, J. M., & Staufenbiel, T. (2001). Statistik mit SPSS Version 10.0. [Statistics with
SPSS version 10.0.] Eschborn: Klotz.
Dollinger, S. J., & Clancy, S. M. (1993). Identity, self, and personality: II. Glimpses
through the autophotographic eye. Journal of Personality and Social Psychology,
64 (6), 1064-1071.
Eysenck, H. J. (1993). Creativity and personality: Suggestions for a theory.
Psychological Inquiry, 4 (3), 147-178.
Eysenck, H. J. (1995a). Genius. The natural history of creativity. Cambridge:
University Press.
Eysenck, H. J. (1995b). Creativity as a product of intelligence and personality. In D.
H. Saklofske & M. Zeidner (Eds.), International handbook of personality and
intelligence (pp. 231-247). New York: Plenum Press.
Eysenck, H. J., & Eysenck, M. W. (1985). Personality and individual differences: A
natural science approach. New York: Plenum Press.
Eysenck, H. J., & Eysenck, S. B. G. (1975). Manual of the Eysenck Personality
Questionnaire (junior & adult). London: Hodder & Stoughton.
Eysenck, H. J., & Eysenck, S. B. G. (1991). Manual of the Eysenck Personality
Scales (EPS Adult). London: Hodder & Stoughton.
Eysenck, S. B. G., Eysenck, H. J., & Barrett, P. (1985). A revised version of the
Psychoticism scale. Personality and Individual Differences, 6, 21-29.
Falconer, D. S., & Mackay, T. F. C. (1996). Introduction to Quantitative Genetics.
Essex: Longman.
Feist, G. J. (1998). A meta-analysis of personality in scientific and artistic creativity.
Personality and Social Psychology Review, 2 (4), 290-309.
Feist, G. J., & Barron, F. X. (2003). Predicting creativity from early to late adulthood:
Intellect, potential, and personality. Journal of Research in Personality, 37, 62-88.
Ferris, P. R., Feldhusen, J. F., & van Mondfrans, A. P. (1971). The relationship
between academic grades and creativity test scores derived from four different
methods of testing. Journal of Experimental Education, 40 (1), 35-40.
Franck, K., & Rosen, E. (1949). A projective test of masculinity-femininity. Journal of
Consulting Psychology, 13, 247-256.
Frearson, W. & Eysenck, H. J. (1986). Intelligence, reaction time (RT) and a new
odd-man-out RT paradigm. Personality and Individual Differences, 7, 807-818.
References 90

Funder, D. C. (1987). Errors and mistakes: Evaluating the accuracy of social

judgment. Psychological Bulletin, 101 (1), 75-90.
Galton, F. (1869). Hereditary genius: An inquiry into its laws and consequences.
London: Macmillan.
Gigerenzer, G. (1998). Surrogates for theories. Theory and Psychology, 8 (2), 195-
204.
Goldberg, L. R. (1990). An alternative “description of personality”: the Big-Five factor
structure. Journal of Personality and Social Psychology, 59, 1216-1229.
Goldberg, L. R. (1994). Resolving a scientific embarrassment: A comment on the
articles in this special issue. European Journal of Personality, 8, 351-356.
Gough, H. G. (1957). Manual of the Californian Psychological Inventory. Palo Alto:
Consulting Psychologists Press.
Gough, H. G. (1979). A creative personality scale for the Adjective Check List.
Journal of Personality and Social Psychology, 37 (8), 1398-1405.
Grigorenko, E. L., LaBude, M. C., & Carter, A. S. (1992). Similarity in general
cognitive ability, creativity, and cognitive style in a sample of adolescent Russian
twins. Acta Geneticae Medicae et Gemellologicae, 41, 65-72.
Guilford, J. P. (1950). Creativity. American Psychologist, 5, 444-454.
Guilford, J. P. (1967a). The nature of human intelligence. New York: McGraw Hill.
Guilford, J. P. (1967b). Creativity: Yesterday, today, and tomorrow. Journal of
Creative Behavior, 1, 3-14.
Guilford, J. P. (1981). Potentiality for creativity. In J. C. Gowan, J. Khatena, & E. P.
Torrance (Eds.), Creativity: Its educational implications (2nd edition, pp. 1-5).
Dubuque, IA: Kendall/Hunt.
Häcker, H., Schmidt, L. R., Schwenkmezger, P., Utz, H. E. (1975). Objektive
Testbatterie, OA-TB 75. [Objective test battery, OA-TB 75.] Weinheim: Beltz
Test.
Haensly, P. A., & Reynolds, C. R. (1989). Creativity and intelligence. In J. A. Glover,
R. R. Ronning & C. R. Reynolds (Eds.), Handbook of creativity (pp. 111-132).
New York: Plenum Press.
Hargreaves, H. L. (1927). The ‘faculty’ of imagination. British Journal of Psychology
(Monograph Supplement), 10, 74.
Harrington, D. M. (1993). The problematic elevation of concepts to positions of
preeminence. Psychological Inquiry, 4 (3), 205-209.
References 91

Hayes, J. R. (1989). Cognitive processes in creativity. In J. A. Glover, R. R. Ronning

& C. R. Reynolds (Eds.), Handbook of creativity (pp. 135-145). New York:
Plenum Press.
Heath, A. C., Eaves, L. J., & Martin, N. G. (1985). The genetic structure of
personality: III. Multivariate genetic item analysis of the EPQ scales. Personality
and Individual Differences, 10, 877-888.
Helmholtz, H. (1896). Vorträge und Reden. [Lectures and talks.] Braunschweig:
Vieweg.
Hocevar, D. (1979a). Ideational fluency as a confounding factor in the measurement
of originality. Journal of Educational Psychology, 71 (2), 191-196.
Hocevar, D. (1979b). A comparison of statistical infrequency and subjective judgment
as criteria in the measurement of originality. Journal of Personality Assessment,
43 (3), 297-299.
Hocevar, D. (1980). Intelligence, divergent thinking, and creativity. Intelligence, 4, 25-
40.
Hocevar, D., & Bachelor, P. (1989). A taxonomy and critique of measurements used
in the study of creativity. In J. A. Glover, R. R. Ronning & C. R. Reynolds (Eds.),
Handbook of creativity (pp. 53-75). New York: Plenum Press.
Horn, J. L., & Cattell, R. B. (1967). Age differences in fluid and crystallized
intelligence. Acta Psychologica, 26, 107-129.
Horn, W. (1962). Leistungsprüfsystem. L-P-S (2. Edition). [Achievement Check
System. L-P-S (2nd edition).] Göttingen: Hogrefe.
Hull, C. L. (1943). Principles of behavior: Its relation to human abilities. New York:
Wiley.
Isen, A. M., Daubman, K. A., & Nowicki, G. P. (1987). Positive affect facilitates
creative problem solving. Journal of Personality and Social Psychology, 52 (6),
1122-1131.
Jäger, A. O. (1982). Mehrmodale Klassifikation von Intelligenzleistungen.
Experimentell kontrollierte Weiterentwicklung eines deskriptiven
Intelligenzstrukturmodells. [Multimodal classification of intelligence achievements.
Experimentally controlled extension of a descriptive intelligence structure model.]
Diagnostica, 28, 195-226.
Jäger, A. O. (1984). Intelligenzstrukturforschung: Konkurrierende Modelle, neue
Entwicklungen, Perspektiven. [Intelligence structure research: Concurring
References 92

models, new developments, perspectives.] Psychologische Rundschau, 35, 272-

289.
Jäger, A. O., Süß, H.-M., & Beauducel, A. (1997). Berliner Intelligenzstruktur-Test.
Handanweisung. Göttingen: Hogrefe.
Jensen, A. R. (1996). Giftedness and genius: Crucial differences. In C. Benbow & D.
J. Lubinski (Eds.), Intellectual talent: Psychometric and social issues (pp. 393-
411). Baltimore, MD: John Hopkins University Press.
John, O. P., Donahue, E. M., & Kentle, R. L. (1991). The Big Five Inventory –
versions 4a and 4b. Technical report, Institute of Personality and Social
Research, University of California, Berkley.
John, O. P., & Srivastava, S. (1999). The Big Five trait taxonomy. History,
measurement, and theoretical perspectives. In L. A. Pervin & O. P. John (Eds.),
Handbook of personality: Theory and research (pp. 102-138). New York: Guilford
Press.
Johnson, J. A. (1994). Clarification of Factor Five with the help of the AB5C model.
European Journal of Personality, 8, 311-334.
King, L. A., McKee Walker, L., &Broyles, S. J. (1996). Creativity and the Five-Factor
Model. Journal of Research in Personality, 30, 189-203.
Klahr, D., & Simon, H. A. (1999). Studies of scientific creativity: Complementary
approaches and convergent findings. Psychological Bulletin, 125, 524-543.
Kratzmeier, H., & Horn, R. (1974). Raven-Matrizen-Test - Advanced Progessive
Matrices. [Raven-Matrices-Test - Advanced Progressive Matrices.] Weinheim:
Beltz Test.
Loehlin, J. C. (1996). The Cholesky approach: A cautionary note. Behavior Genetics,
26 (1), 65-69.
Lubinski, D., Webb, R. M., Morelock, M. J., & Benbow, C. P. (2001). Top 1 in 10,000:
A 10-year follow-up of profoundly gifted. Journal of Applied Psychology, 86 (4),
718-729.
Lubow, R. E. (1989). Latent inhibition and conditioned attention theory. New York:
Cambridge University Press.
Lubow, R. E., Ingberg-Sachs, Y., Zalstein-Orda, N., Gewirtz, J. C. (1992). Latent
inhibition in low and high “psychotic-prone” normal subjects. Personality and
Individual Differences, 6, 587-597.
References 93

Lykken, D. T. (1982). Research with twins: The concept of emergenesis.

Psychophysiology, 19 (4), 361-373.
Lykken, D. T., McGue, M., Tellegen, A., & Bouchard, T. J., jr. (1992). Emergenesis.
Genetic traits that may not run in families. American Psychologist, 47 (12), 1565-
1577.
MacKinnon, D. W. (1962). The nature and nurture of creative talent. American
Psychologist, 17, 484-495.
Martindale, C. (1989). Personality, situation, and creativity. In J. A. Glover, R. R.
Ronning & C. R. Reynolds (Eds.), Handbook of creativity (pp. 211-232). New
York: Plenum Press.
Martindale, C. (1993). Creativity, causality, and the separation of personality and
cognition. Psychological Inquiry, 4 (3), 209-211.
Martindale, C. (1999). Biological bases of creativity. In R. J. Sternberg (Ed.),
Handbook of creativity (pp. 137-152). Cambridge: University Press.
Martindale, C., & Dailey, A. (1996). Creativity, primary process cognition and
personality. Personality and Individual Differences, 20 (4), 409-414.
Mayer, R. E. (1999). Fifty years of creativity research. In R. J. Sternberg (Ed.),
Handbook of creativity (pp. 449-460). Cambridge: University Press.
McCrae, R. R. (1987). Creativity, divergent thinking, and openness to experience.
Journal of Personality and Social Psychology, 52 (6), 1258-1265.
McCrae, R. R. (1993/1994). Openness to experience as a basic dimension of
personality. Imagination, Cognition and Personality, 13, 39-55.
McCrae, R. R. (1994). Openness to Experience: Expanding the boundaries of factor
V. European Journal of Personality, 8, 251-272.
McCrae, R. R., Costa, P. T., jr., Ostendorf, F., Angleitner, A., Hrebícková, M., Avia,
M. D., Sanz, J., & Sánchez-Bernardos, M. L. (2000). Nature over nurture:
Temperament, personality, and life span development. Journal of Personality and
Social Psychology, 78 (1), 173-186.
McGue, M., & Bouchard, T. J., jr. (1984). Adjustment of twin data for the effects of
age and sex. Behavior Genetics, 14 (4), 325-343.
Mednick, S. A. (1962). The associative basis of the creative process. Psychological
Review, 69, 220-232.
Mednick, S. A. & Andrews, P. M. (1967). Creative thinking and level of intelligence.
Journal of Creative Behavior, 4, 428-431.
References 94

Miller, G. F. (2000). The mating mind: How sexual choice shaped the evolution of
human nature. New York: Anchor Books.
Murray, H. A. (1943). Thematic Apperception Test manual. Cambridge, MA: Harvard
University Press.
Neale, M. C., & Cardon, L. R. (1992). Methodology for genetic studies of twins and
families. Dordrecht, NL: Kluwer.
Neale, M. C., Boker, S. M., Xie, G., & Maes, H. H. (2002). Mx: Statistical modeling.
Richmond, Department of Psychiatry.
Neubauer, A. C., Spinath, F. M., Riemann, R., Angleitner, A., & Borkenau, P. (2000).
Genetic and environmental influences on two measures of speed of information
processing and their relation to psychometric intelligence: Evidence from the
German Observational Study of Adult Twins. Intelligence, 28 (4), 267-289.
Nichols, R. C. (1978). Twin studies of ability, personality and interests. Homo, 29,
158-173.
Oniszczenko, W., Angleitner, A., Strelau, J., & Angert, T. (1993). The questionnaire
of twins’ physical resemblance. Unpublished manuscript, University of Warsaw,
Poland.
Ostendorf, F. (1994). Zur Taxonomie deutscher Dispositionsbegriffe. [Towards the
taxonomy of German dispositional terms.] In W. Hager & M. Hasselhorn (Eds.),
Handbuch deutschsprachiger Wortnormen (pp. 382-441). [Handbook of German
word norms.] Göttingen: Hogrefe.
Ostendorf, F., & Angleitner, A. (1994). Reflections on different labels for factor V.
European Journal of Personality, 8, 341-349.
Peterson, J. B., & Carson, S. (2000). Latent inhibition and openness to experience in
a high-achieving student population. Personality and Individual Differences, 28,
323-332.
Peterson, J. B., Smith, K. W., & Carson, S. (2002). Openness and extraversion are
associated with reduced latent inhibition: Replication and commentary.
Personality and Individual Differences, 33, 1137-1147.
Plomin, R., DeFries, J. C., McClearn, G. E., & Rutter, M. (1997). Behavioral genetics
(3rd edition). New York: W. H. Freeman and Company.
Prentky, R. (1989). Creativity and psychopathology: Gamboling at the seat of
madness. In J. A. Glover, R. R. Ronning & C. R. Reynolds (Eds.), Handbook of
creativity (pp. 243-269). New York: Plenum Press.
References 95

Purcell, S. (2002). Variance components models for gene-environment interaction in

twin analysis. Twin Research, 5 (6), 554-571.
Raven, J. C. (1958). Advanced Progressive Matrices. London: Lewis.
Rawlings, D. (1985). Psychoticism, creativity, and dichotic shadowing. Personality
and Individual Differences, 6, 737-742.
Rawlings, D., Twomey, F., Burns, E., & Morris, S. (1998). Personality, creativity, and
aesthetic preference: Comparing psychoticism, sensation seeking, schizotypy,
and openness to experience. Empirical Studies of the Arts, 16, 153-178.
Reznikoff, M., Domino, G., Bridges, C., & Honeyman, M. (1973). Creative abilities in
identical and fraternal twins. Behavior Genetics, 3 (4), 365-377.
Richards, R., Kinney, D. K., Benet, M., & Merzel, A. P. C. (1988). Assessing
everyday creativity: Characteristics of the Lifetime Creativity Scales and
validation with three large samples. Journal of Personality and Social
Psychology, 54 (3), 476-485.
Ruch, W. (1999). Die revidierte Fassung des Eysenck Personality Questionnaire und
die Konstruktion des deutschen EPQ-R bzw. EPQ-RK. [The revised version of
the Eysenck Personality Questionnaire and the construction of the German EPQ-
R and EPQ-RK.] Zeitschrift für Differentielle und Diagnostische Psychologie, 20
(1), 1-24.
Runco, M. A. (1993). Creativity, causality, and the separation of personality and
cognition. Psychological Inquiry, 4 (3), 221-225.
Runco, M. A., & Albert, R. S. (1986). The threshold theory regarding creativity and
intelligence: An empirical test with gifted and non-gifted children. Creative Child
and Adult Quarterly, 11, 217-218.
Runco, M. A., & Charles, R. E. (1993). Judgments of originality and appropriateness
as predictors of creativity. Personality and Individual Differences, 15 (5), 537-546
Schaefer, C. E. (1969). Originality scale - Franck Drawing Completion Test:
Preliminary manual. Unpublished manuscript, Psychological Department,
Fordham University, New York.
Schaefer, C. E. (1970). Development of an originality scale for the Franck Drawing
Completion Test. Perceptual and Motor Skills, 31, 402.
Shrout, P. E., & Fleiss, J. L. (1979). Intraclass correlations: Uses in assessing rater
reliability. Psychological Bulletin, 86 (2), 420-428.
References 96

Simonton, D. K. (1984). Genius, creativity, and leadership: Historiometric inquiries.

Cambridge: Harvard University Press.
Simonton, D. K. (1988). Scientific genius: A psychology of science. Cambridge, UK:
University Press.
Simonton, D. K. (1994). Greatness: Who makes history and why. New York: Guilford
Press.
Simonton, D. K. (1998). Donald Campbell's model of the creative process: Creativity
as blind variation and selective retention. Journal of Creative Behavior, 32 (3),
153-159.
Simonton, D. K. (1999a). Creativity and genius. In L. A. Pervin, & O. P. John (Eds.),
Handbook of personality: Theory and research (pp. 629-652). New York: Guilford
Press.
Simonton, D. K. (1999b). Origins of genius. Darwinian perspectives on creativity.
Oxford: University Press.
Simonton, D. K. (1999c). Creativity as blind variation and selective retention: Is the
creative process Darwinian? Psychological Inquiry, 10 (4), 309-328.
Simonton, D. K. (1999d). The continued evolution of creative Darwinism.
Psychological Inquiry, 10 (4), 362-367.
Simonton, D. K. (2003). Scientific creativity as constrained stochastic behavior: The
integration of product, person and process perspectives. Psychological Bulletin,
129 (4), 475-494.
Soronkin, P. A. (1937-1941). Social and cultural dynamics (4 volumes). New York:
American Book Press.
Soldz, S., & Vaillant, G. E. (1999). The Big Five personality traits and the life course:
A 45-year longitudinal study. Journal of Research in Personality, 33, 208-232.
Spearman, C. (1904). “General intelligence”, objectively determined and measured.
American Journal of Psychology, 15, 201-293.
Spinath, F. M., Riemann, R., Hempel, S., Schlangen, B., Weiß, R., Borkenau, P., &
Angleitner, A. (1999). A day in the life: Description of the German Observational
Study on Adult Twins (GOSAT) assessing twin similarity in controlled laboratory
settings. In I. Mervielde, I. J. Deary, F. DeFruyt & F. Ostendorf (Eds.), Personality
psychology in Europe (Vol. 7, pp. 311-328). Tilburg: University Press.
References 97

Spinath, F. M., Angleitner, A., Borkenau, P., Riemann, R., & Wolf, H. (2002). German
Observational Study of Adult Twins: A multimodal investigation of personality,
temperament and cognitive ability. Twin Research, 5, 372-375.
Stayte, S. (1977). Incidential learning, creativity, and schizophrenia. Unpublished
dissertation, Institute of Psychiatry, London.
Stein, M. I. (1953). Creativity and culture. Journal of Psychology, 36, 311-322.
Sternberg, R. J. (1998). Cognitive mechanisms in human creativity: Is variation blind
or sighted? Journal of Creative Behavior, 32 (3), 159-177.
Sternberg, R. J., & Lubart, T. I. (1995). Defying the crowd: Cultivating creativity in a
culture of conformity. New York: Free Press.
Sternberg, R. J., & O'Hara, L. (1999). Creativity and intelligence. In R. J. Sternberg
(Ed.), Handbook of creativity (pp. 251-272). Cambridge: University Press.
Sternberg, R. J. (Ed.) (1999). Handbook of creativity. Cambridge: University Press.
Sturm, W., & Willmes, K. (1983). LPS-K – eine LPS-Kurzform für hirngeschädigte
Patienten; mit Anleitung zur psychometrischen Einzelfalldiagnostik [LPS-K – a
short form of the LPS for patients suffering from brain damage; with instructions
for single case diagnostics.]. Diagnostica, 29, 346-358.
Terman, L. M. (1925). Genetic studies of genius: Vol. I. Mental and physical traits of a
thousand gifted children. Stanford: University Press.
Terman, L. M., & Oden, M. H. (1947). Genetic studies of genius: Vol. IV. The gifted
child grows up. Stanford: University Press.
Terman, L. M., & Oden, M. H. (1959). The gifted child at mid-life. Stanford: University
Press.
Thornhill, R. (2003). Darwinian aesthetics informs traditional aesthetics. In E. Voland
& K. Grammer (Eds.), Evolutionary aesthetics (pp. 13-35). Berlin: Springer.
Thurstone, L. L. (1938). Primary mental abilities. Chicago: University of Chicago
Press.
Torrance, E. P. (1966). Torrance Test of Creative Thinking: Norms-technical manual.
Princeton: Personnel Press.
Torrance, E. P. (1974). Torrance Test of Creative Thinking: Norms-technical manual.
Princeton: Personnel Press/Ginn.
Torrance, E. P. (1975). Creativity research in education: Still alive. In I. A. Taylor, & J.
W. Getzels (Eds.), Perspectives in creativity (pp. 278-296). Chicago: Aldine.
References 98

Vernon, P. E. (1971). Effects of administration and scoring on divergent tests. British

Journal of Educational Psychology, 41 (3), 245-257.
Voland, E., & Grammer, K. (Eds.) (2003). Evolutionary aesthetics. Berlin: Springer.
Wallach, M. A. (1971). The intelligence/creativity distinction. New York: General
Learning Press.
Wallach, M. A., & Kogan, N. (1965). Modes of thinking in young children. New York:
Holt, Rinehart & Winston.
Wallas, G. (1926). The art of thought. New York: Harcourt, Brace.
Waller, N. G., Bouchard, T. J., jr., Lykken, D. T., Tellegen, A., & Blacker, D. M.
(1993). Creativity, heritability, familiality: Which word does not belong?
Psychological Inquiry, 4 (3), 235-237.
Wallinga, C. R., & Crase, S. J. (1983). Creativity in husbands and wives. Journal of
Creative Behavior, 16 (4), 277-283.
Weiner, I. (1990). Neural substrates of latent inhibition: The switching model.
Psychological Bulletin, 108, 442-461.
Weisberg, R. W. (1992). Creativity: Beyond the myth of genius. New York: Freeman.
Weisberg, R. W. (1999). Creativity and knowledge: A challenge to theories. In R. J.
Sternberg (Ed.), Handbook of creativity (pp. 226-250). Cambridge: University
Press.
White, R. K. (1931). The versatility of genius. Journal of Social Psychology, 2, 460-
489.
Wolfradt, U., & Pretz, J. E. (2001). Individual differences in creativity: Personality,
story writing, and hobbies. European Journal of Personality, 15, 297-310.
Woodman, R. W. (1981). Creativity as a construct in personality theory. Journal of
Creative Behavior, 15 (1), 43-66.
Woodman, R. W., & Schoenfeldt, L. F. (1989). Individual differences in creativity: An
interactionist perspective. In J. A. Glover, R. R. Ronning & C. R. Reynolds (Eds.),
Handbook of creativity (pp. 77-91). New York: Plenum Press.
Woody, E., & Caldridge, G. (1977). Psychoticism and thinking. British Journal of
Social and Clinical Psychology, 16, 241-248.
Wuthrich, V., & Bates, T. C. (2001). Schizotypy and latent inhibition: Non-linear
linkage between psychometric and cognitive markers. Personality and Individual
Differences, 30, 783-798.
References 99

Yamamoto, K. (1964). Experimental scoring manual for Minnesota tests of creative

thinking and writing. Kent: Bureau of Educational Research, Kent State
University.
Appendix 100

6. Appendix

6.1 The T-88

Appendix 101
Appendix 102
Appendix 103

6.2 Rater instructions for the T-88

Rater-Instruktionen

- Auf die richtige Kennummer beim Eintragen in den Ratingbogen achten

- Auf die Nummerierung der Items achten!
- Immer pro Heft zunächst ein Item für alle Versuchspersonen in diesem Heft auf
beiden Dimensionen bewerten, dann das nächste Item
- Zeichnungen dürfen auch so ergänzt sein, dass das Bild auf der Seite liegt oder
auf dem Kopf steht
- Sei möglichst durchgehend konsistent in Deiner Beurteilung
- Nutze die ganze Breite der Skala, sei also nicht zu mild oder streng.
- Vermeide auch zu häufige Extremurteile oder eine starke Tendenz zu mittleren
Einschätzungen
- Es hilft, die Zeichnungen immer wieder mit den vorgegebenen Linien zu
vergleichen.
- Achte darauf, wie gut Du Dich noch konzentrieren kannst, damit Du nicht in
unterschiedlichen Zeichnungen immer die Standardantworten siehst!

Originalität ist die Fähigkeit, etwas zu produzieren, das einerseits neu (d.h.
unkonventionell, unerwartet, ungewöhnlich, statistisch selten, einfallsreich) ist,
andererseits aber auch einen Realitätsbezug hat (letzterer kann hier aber großzügig
bewertet werden). Auch abstrakte Formen und Muster zählen, sollten aber eine
gewisse ästhetische Qualität besitzen und nicht zufällig oder lustlos sein.

In ein Originalitätsrating einfließen sollte:

- Neuheit, Seltenheit, Originalität, Überraschung, ferne Assoziation (siehe
Baselines)
- Die Idee, also der Inhalt (aus Zeichnung und evtl. Titel) und die Form

Nicht in ein Originalitätsrating einfließen sollten:

- Die Originalität der anderen Items einer Versuchsperson (also jedes Item einzeln,
kein Gesamteindruck!)
- eigene Vorlieben oder Abneigungen
- die Qualität der Ausführung, im Sinne von Zeichenfähigkeit
- wie Ordentlich oder Unbeholfen die Ausführung ist
- fehlender Titel (ist egal)
- Handschrift, Rechtschreibfehler im Titel o.ä.

Elaboriertheit bezieht sich rein auf die Qualität der Ausführung, d.h. in der Regel, wie
gut die Zeichnung erkennbar und wie detailliert und/oder künstlerisch sie ist. Bei
Mustern und abstrakten Zeichnungen ist die Elaboriertheit subjektiv. Ob ein Titel
gegeben wurde, die Handschrift etc. sind auch hier egal. Auch Zeichentechnik oder
-übung sollte hier nicht einfließen.

Bei gleichen Titel können die Zeichnungen dennoch unterschiedlich elaboriert

und/oder originell sein!
Appendix 104

Die Skalen:

Elaboriertheit: 0-1-2

/ (9): Missing: Nicht bearbeitet oder nur durchgestrichen

0: Bearbeitet, aber nicht erkennbar (oder Zufallsproduktion)
1: Bearbeitet, erkennbar
2: Bearbeitet, erkennbar und verziert, verfeinert, mit Details, Schatteneffekten o.
ä.

Originalität: 1-2-3-4-5

/ (9): Missing: Nicht bearbeitet oder nur durchgestrichen

1: Sehr arm. Plan- oder lustlos bearbeitet, Zufallsproduktionen oder sehr nah an
den vorgegebenen Linien orientiert (extrem stimulusgebunden)
2: Arm. Nah an den vorgegebenen Linien (stimulusgebunden), banal und
abgedroschen
3: Durchschnittlich. Gewöhnliche Zeichnung, die mittelmäßig von der Vorgabe
entfernt ist
4: Überdurchschnittlich. Clever, gewitzt, ästhetisch
5: Sehr clevere oder ungewöhnliche Idee. Hat einen fesselnden Effekt
Appendix 105

6.3 Response baselines for the T-88

Baselines: Top-Antworten Cut-Off: 5%

Item 1 % (korr.) Item 5 % (korr.)

Wolke 9,36 (Sand-)Burg / Festung 37,5
Zahn 9,36 Skyline / Wolkenkratzer /Stadt 34,09
(Enten-)Teich, See 8,51 Haus / Fabrik 17,42
Apfel 6,81 Summe 89,01
Gesicht 6,38 % (unkorr.)
Kartoffel 5,11 nichts o. unkenntlich 5,71
Summe 45,53
% (unkorr.)
nichts o. unkenntlich 15,47

Item 2 % (korr.) Item 6 % (korr.)

Herz 73,53 Blume, Blüte, Tulpe 17,7
Po (von Frau o. Kind) 5,51 Krone 16,05
(Klee-)Blatt 5,15 Mütze, Hut 7,82
Summe 84,19 Schmetterling 7,82
% (unkorr.) Summe 49,39
nichts o. unkenntlich 2,16 % (unkorr.)
nichts o. unkenntlich 12,59

Item 3 % (korr.) Item 7 % (korr.)

Blatt (angefressen etc.) 14,04 Kaktus 15,75
Fleck, Klecks 9,36 Hand 10,96
Seepferdchen 7,02 Glied / Penis 9,59
Drache 5,85 Gebirge 8,9
Landkarte 5,85 Summe 45,2
Summe 42,12 % (unkorr.)
% (unkorr.) nichts o. unkenntlich 47,65
nichts o. unkenntlich 38,49

Item 4 % (korr.) Item 8 % (korr.)

Nackte von hinten, weibl. Akt 34,57 Fuß 37,33
Birne 14,4 Hand 6,45
Nase (eines Gesichtes) 11,52 Bumerang 6,45
Hund 7,82 Summe 50,23
Summe 68,31
% (unkorr.) % (unkorr.)
nichts o. unkenntlich 12,9 nichts o. unkenntlich 21,94

% (korr.): Prozentuale Anteile an den erkennbaren Antworten von bearbeiteten Items

Appendix 106

Item 9 % (korr.) Item 14 % (korr.)

Luftballon 43,64 Turm / Burg / Mauer (Zinnen) 40,68
Raupe / Wurm / Larve 7,73 Schiff 16,95
Summe 51,37 Eisenbahn / Lok 9,75
% (unkorr.) (Hoch-)Haus (mit Kamin) 8,9
nichts o. unkenntlich 20,86 Skyline / Stadt 6,36
Summe 82,64
% (unkorr.)
nichts o. unkenntlich 15,11

Item 10 % (korr.) Item 15 % (korr.)

Stern 93,45 See / Pfütze / Teich 23,16
Komet / Sternschnuppe 4 Gesicht / Kopf / Fratze 7,89
Summe 97,45 Spiegelei 6,32
% (unkorr.) Fahne 6,32
nichts o. unkenntlich 1,08 Wolke 6,32
Fleck / Klecks 5,79
Summe 55,8
% (unkorr.)
nichts o. unkenntlich 31,65

Item 11 % (korr.) Item 16 % (korr.)

(Vogel-)Fuss(abdruck) 60,19 Flugzeug 17,07
Gespenst 5,83 Fuss 17,07
See / Pfütze 5,83 UFO 16,67
Tier auf dem Kopf 5,34 Auto 12,6
Summe 77,19 Hut 6,1
% (unkorr.) Summe 69,51
nichts o. unkenntlich 25,9 % (unkorr.)
nichts o. unkenntlich 11,51

Item 12 % (korr.) Item 17 % (korr.)

Vogel / Ente 38,02 Zahl '8' 32,22
Flasche / Gefäß / Vase 24,48 Wurm / Schlange 17,15
Hut / Mütze / Haube 5,05 Unendlich-Zeichen 5,44
Summe 67,55 Summe 54,81
% (unkorr.) % (unkorr.)
nichts o. unkenntlich 30,94 nichts o. unkenntlich 14,03

Item 13 % (korr.) Item 18 % (korr.)

Vogel (im Flug) 31,25 Kopf ( Kasper, Comic,...) 52,58
Narrenkappe / Kasper 15,38 Baum 12,89
Muster/Ornament/Wappen 11,54 Kaktus 7,22
Palme 5,29 Summe 72,69
Summe 32,21 % (unkorr.)
% (unkorr.) nichts o. unkenntlich 30,22
nichts o. unkenntlich 25,18

Durchschnittlich nicht bearbeitet o. unkenntlich: 20,19

Durchschnittlicher Anteil von Antworten, die häufiger als 5% sind: 63,07
% (korr.): Prozentuale Anteile an den erkennbaren Antworten von bearbeiteten Items
Eidesstattliche Erklärung (Affidavit) 107

7. Eidesstattliche Erklärung (Affidavit)

Hiermit versichere ich, dass ich diese Arbeit selbständig verfasst und ausschließlich
die angegebenen Quellen und Hilfsmitteln benutzt sowie Zitate kenntlich gemacht
habe.

Lars Penke
Bielefeld, den 14. November 2003