Upload
others
View
13
Download
0
Embed Size (px)
Citation preview
Toggweiler (2018) 1
Regression function and explained variance of NEO-PI-R Big Five –
graphology as valid observer-assessment.
Stephan Toggweiler
INTRODUCTION
GraphoPro as a tool for data collection of handwriting signs
GraphoPro (Keel, 2016) by dipl. Psych., ingeneer FH and graphologist Bruno Keel is a software,
which enables the systematic recording of handwriting signs. GraphoPro is an open system, the
integrated handwriting signs have been collected in the relevant literature and can be easily
supplemented. At the moment 234 handwriting signs are programmed in GraphoPro, e. g. conspicuous,
determined, weakly printed, etc. They are coded in four levels (characteristic does not exist,
characteristic occurs, characteristic is clearly present, characteristic dominates).
Figure 1. GraphoPro user interface.
NEO-PI-R from Ostendorf and Angleitner (2004)
In order to be able to assess the impact of the handwriting signs of GraphoPro in the sense of a
construct validation, and in order to optimise GraphoPro in this respect, the five Big Five scales (Table
1) of the NEO-PI-R (Ostendorf & Angleitner, 2004) were included. The aim is to predict these
Toggweiler (2018) 2
characteristics by means of the handwriting signs. If this should succeed, a personality test could be
replaced through graphological analysis. The NEO-PI-R was chosen because it is a well differentiated
and widely used test. The NEO-PI-R is based on the German translation of the Revised NEO
Personality Inventory of Costa and McCrae (1992; McCrae & Costa, 1996,1999). It serves to evaluate
basic personality traits, the so-called Big Five: Neuroticism, extraversion, openness to experience,
agreeableness, and conscientiousness. Each of these dimensions is divided into six subscales with
eight items each. This allows a very differentiated measurement of the five superior personality traits
as well as 30 sub-facets of personality. The advantage of the NEO-PI-R is its wide range of personal
characteritics, which is captured and identifies the test as a screening instrument. To name just a few
arreas of its application the NEO-PI-R is used in counselling, clinical psychology and psychiatry, in
behavioural medicine and health psychology, in career counselling, in occupational, industrial and
organisational psychology and in research (Ostendorf & Angleitner, 2004, p. 4). The NEO-PI-R is a
validated standard instrument (Ostendorf & Angleitner, 2004, p. 140ff), easy to use, well known and it
has been translated into more than 30 languages. There is as well a version available that enables an
observer-assessments, e. g. by family members or close acquaintances. The following Table 1 shows
the scales and sub-scales of the NEO-PI-R including examples of items.
Table 1. The Big Five dimensions of the NEO-PI-R (Ostendorf & Angleitner, 2004)
Scales Subscales (examples)
Neuroticism (p. 33f.) Persons with a high level of neuroticism are more sensitive and tend to
lose their balance more easily under stress. In stressful situations, they tend
to be more often annoyed, sad, embarrassed, anxious, ashamed, distressed
and worried. They also tend to develop more inappropriate forms of
problem solving, tend to have unrealistic ideas and are less able to control
their needs. Persons with high level of these characteristics can be
characterised with the following adjectives: Tensioned, needy, anxious,
worried, alarmed, emotional, sensitive, irritable, helpless, helpless, moody,
nervous, self-pitying, self-doubting, hypersensitive, unbalanced, restless,
insecure, unhappy, unsatisfied, vulnerable and snivelling.
Anxiety (I'm not easily to be worried)
Angry-Hostility (I often get angry about
how other people treat me)
Depression (I rarely feel lonely or sad)
Self-consciousness (I'm often afraid that I
might attract unpleasant attention when
dealing with others)
Impulsivity (I seldom get too involved in
anything)
Vulnerability (I often feel helpless and
wish a person to solve my problems)
Extraversion (p. 40f.) In everyday language, people with a high degree of extraversion can be
described as sociable, talkative, friendly, enterprising and active.
Extraverted people like the company of others, they feel comfortable in
groups, but are also assertive, self-confident, dominant, love exciting
situations and stimuli. They tend to be optimistic, cheerful and entertaining
[...]. According to Costa and McCrae (1992, p. 15), introversion should be
understood as a lack rather than an opposition to extraversion. Descriptive
adjectives for an extroverted person: Adventurous, active, bright,
enthusiastic, dominant, dynamic, fiery, cheerful, friendly, chatty, sociable,
talkative, cordial, likes to be funny, lively, passionate, courageous,
optimistic, person-oriented, talkative, communicative, assertive,
spontaneous and temperate.
Warmth (Most people I meet are really
likeable for me)
Gregariousness (I like to have a lot of
people around me)
Assertiveness (I am dominant, self-
confident and assertive)
Activity (I work and play in a leisurely
way)
Excitement-Seeking (I often long to
experience more exciting things)
Positive emotions (I've never really jumped
up in the air with joy)
Openness to Experience (p. 42f.)
People with high level of openness to experience are interested in new
experiences and impressions. They are interested in the outside world, but
also in their inner world. They claim to have a lively fantasy and to
perceive their own positive and negative feelings very clearly. They
embrace new ideas and are unconventional in their value orientations.
They describe themselves as being interested in a variety of subjects,
interested in knowledge, creative and interested in theories and cultural
events, inclined to critically question existing norms and values and
willing to deal with new ethical, political and social issues and
orientations. […] A high level of openness to experience can be described
with the following adjectives: Imaginative, sentient, inventive, sensitive,
progressive, witty, has many interests, full of ideas, creative, critical,
liberal, musical, curious, non-conformist, open, original, unconventional,
and eager to learn.
Openness to Fantasy (I have a very lively
imagination)
Openness to Aesthetics (Aesthetics and art
mean a lot to me)
Openness to Feelings (Without strong
feelings life would be uninteresting for me)
Openness to Actions (I am quite used to my
courses)
Openness to Ideas (I often have fun playing
with theories or abstract ideas)
Openness to Values (I believe that it is
often confusing and misleading for students
to listen to speakers who take controversial
points of view)
Toggweiler (2018) 3
Agreeableness (p. 44f.)
A person with a high level of agreeableness can be characterised as
helpful, accommodating, trusting and eager to help others. In addition,
such a person is usually convinced that other people as well react with
helpfulness. Persons with a high degree of tolerance are open to other
people with benevolence, tend to be good-natured, are willing to give in in
disputes and may in extreme cases appear to be submissive or dependent.
A pathological extreme variant is the personality disorder of the
"dependent personality". Characteristic adjectives for high level of
agreeableness are as follows: Altruistic, undemanding, innocent, self-
sacrificing, sincere, unambitious, modest, direct, honest, accommodating,
free, outspoken, kind, pleasing, straightforward, generous, trustful, good-
natured, willing, helpful, gullible, credulous, and sympathetic.
Trust (I am rather cynical and sceptical
about the intentions of others)
Straightforwardness (I don't think I'm
cunning)
Altruism (Some people think I'm selfish
and complacent)
Compliance (I would rather work with
others than compete with them)
Modesty (I don't mind stating my skills and
achievements)
Tender-Mindedness (Politicians should
care more about the human side of their
policies)
Conscientiousness (p. 45f.)
People with a high degree of conscientiousness describe themselves as
determined, strong-willed and purposeful. Only a few people become great
musicians or sportsmen without a fairly high expression in
conscientiousness. Digman and Takemoto-Chock (1981) describe this
range of characteristics as "will to achieve". High level corresponds to
academic and professional achievement. On the other hand, an
exaggeratedly high level of demands, almost compulsive neatness and
workaholism can be seen as negative examples of high level of
conscientiousnes. A high level of of conscientiousness can be outlined by
the following adjectives: Hardworking, persistent, prudent, ambitious,
eager, diligent, precise, conscientious, competent, efficient, motivated,
orderly, tidy, disciplined, dutiful, committed, planned, principled,
punctual, righteous, self-disciplined, careful and systemic.
Competence (I am known for my prudence
and my senses)
Order (I prefer to leave myself free from
decisions instead of planning everything in
advance)
Dutifullness (I try to do all the tasks
assigned to me very conscientiously)
Achievement Striving (I am carefree and
indifferent)
Self-Discipline (I can allocate my time
quite well, so that I can finish my business
in time)
Deliberation (I've done some stupid things
in my life)
These scales and subscales are measured on the basis of 240 items scaled at five grades (strong
rejection – rejection – neutral – approval – strong agreement). The completion of the questionnaire
takes about 30 - 40 minutes, but there is no time limit. The target group of the NEO-PI-R is people
over 16 years of age.
Research Questions
This publication answers two questions:
1. Is it possible to construct reliable graphological scales of the Big Five dimensions via
GraphoPro and NEO-PI-R?
2. What is the quality of this solution in terms of convergent validity between the graphological
scales and the Big Five dimensions?
Relevance of the study
The relevance of this study is based on the fact that the author is not aware of any studies which have
attempted to construct valid graphological scales by means of a more or less standardised
psychometrical approach and to predict or rather to diagnose the characteristics of the NEO-PI-R self-
assessment. If in this way it should be possible to construct graphological scales based on an
established psychometrical instrument, a significant step towards validity methodology of
graphological characteristics is likely to be taken.
METHOD
Instruments
For answering the research questions, the GraphoPro software described above and the NEO-PI-R
(Ostendorf & Angleitner, 2004) were used.
Toggweiler (2018) 4
Recruting of the sample
The sample was obtained based on a chance approach. The potential participants were contacted by
Bruno Keel and Etienne Bühler (student ZHAW) and were asked to take part in the survey. Both of
them did not know each other's candidates. Approximately 300 people were asked to participate in the
study, whereas 121 have agreed to submit a handwriting sample and fill out the NEO-PI-R
questionnaire.
Obtaining the NEO-PI-R data
The participants received an instruction on the handwriting sample (by e-mail), together with the
NEO-PI-R questionnaire (by mail). Participants were instructed to write one page about any topic of
their choice on an unlined A4 page with a ballpoint pen. In addition, they were instructed to use five
A4 sheets as a pad and to provide their signature as well. Participants were asked not to use any
writing aids or lined patterns. All handwriting samples were then evaluated using GraphoPro's 234
handwriting signs. The handwriting signs were coded by Bühler or Keel and were cross-checked.
Data cleaning
Missing values in the NEO-PI-R were not replaced. 18 Participants had to be excluded because they
did not fully follow the instructions for the handwriting sample. This resulted in a sample of 103
persons.
Additional sample from the GraphoPro database
At the time of data collection, there were already 98 completely coded handwriting samples in
GraphoPro, but without NEO-PI-R inquiery. Psychologists and graphologists Christian Katz, Bruno
Keel, Roman Krapf and Martin Leisebach coded these handwriting samples. In each case, it was
ensured that at least one graphologist was involved as second coder who did not know the author of
the the handwriting sample.
Statistical procedures
For the statistical analysis the following procedures were used: Descriptive statistics, cluster analyses
(method complete linkage, distance measure as Pearson correlation), correlation analyses, item and
scale analyses, and multiple linear regressions (method enter). The raw values of the graphological
scales were calculated by averaging.
RESULTS
Description of the sample
As shown in Table 1, the gender of the participants was approximately equally distributed within the
whole sample with 51.7 % women and 48.3 % men, and 59.2 % women and 40.8 % men for the NEO-
PI-R sample. 32.3 % of the participants have an academic education, 27.4 % have a high-school or a
special professional education, and 15.9 % were registered for a secondary school or high school at the
time of the research. These relations are also preserved more or less in the NEO-PI-R sample, whereby
the candidates are in average about seven years younger.
Toggweiler (2018) 5
Table 2. Description of samples
Whole Sample
(1)
NEO-PI-R
Sample (2)
Characteristic n (%) n (%)
Gender
Feminine 104 (51.7) 61 (59.2)
Masculine 97 (48.3) 42 (40.8)
Age (M1 = 50.45, SD1 = 20.91 resp. M2 = 42.00, SD2 = 15.06)
18 – 25 years
16 (8.0)
10 (9.7)
26 – 35 years 45 (22.4) 34 (33.0)
36 – 55 years 66 (32.8) 35 (34.0)
> 55 years 74 (36.8) 24 (23.3)
Educational level
Unfinished lower secondary school - -
Lower secondary school 4 (2.0) 4 (3.9)
Apprenticeship without national vocational qualification - -
National vocational qualification, not yet certificated 6 (3.0) 6 (5.8)
National vocational qualification 36 (17.9) 24 (23.3)
Upper secondary school, not yet certificated 26 (12.9) 11 (10.7)
Upper secondary school 5 (2.5) 5 (4.9)
Upper secondary school and ongoing higher education qualification 14 (7.0) 14 (13.6)
Upper secondary school and higher education qualification 65 (32.3) 39 (37.9)
Missing information 45 (22.4) -
Handedness
Left 11 (5.5) 10 (9.7)
Right 190 (94.5) 93 (90.3)
Question 1: Scale construction
First approach – Scale construction via cluster analysis
Via cluster analysis (method complete linkage, distance measure Pearson correlation) a total of 13
clusters were found at a distance of 15. These clusters had the following reliability in terms of
Cronbach's Alpha: .89 (13 handwriting signs), . 88 (24 handwriting signs), .62 (11 handwriting
signs), .90 (20 handwriting signs), .73 (22 handwriting signs), .56 (6 handwriting signs), .87 (15
handwriting signs), .81 (15 handwriting signs), .90 (22 handwriting signs), .93 (25 handwriting
signs), .81 (26 handwriting signs), .86 (19 handwriting signs) and .81 (15 handwriting signs). Both
Pearson correlations and multiple linear regressions, however, showed no significant connections
(explanation of variance) with the Big Five dimensions of at least a medium effect.
Second approach – Scale construction via correlation analysis
With the next approach, it was examined whether the desired association of personality to the
graphological scales could be achieved by means of correlation analysis. In order to construct these
scales, all handwriting signs with significant and highest positive (indexed with +) and significant and
highest negative (indexed with -) correlations were combined with the corresponding dimensions of
the NEO-PI-R. Table 3 shows these Pearson correlations (the original German names of the
handwriting signs are preserved). It can be seen that they rarely are above .30, which would be
equivalent to a medium effect size. Regarding the distributions of the handwriting signs it can be
mentioned that skewness and/or kurtosis exceed the critical value of 2.00 for 17 out of 68 handwriting
signs. However the optical check shows that these handwriting signs are at least unimodal. On the
Toggweiler (2018) 6
other hand, the discriminative powers are clearly inadequate in 26 cases with values less than .30. The
Cronbach's Alpha of the graphological scales is .24 (agreeableness+), .38 (openness to experience
+)
and .86 (extraversion-). Two graphological scales (agreeableness
- and conscientiousness
+) could only
be formed by just one handwriting sign, so there is no Cronbach’s Alpha available. One handwriting
sign (“fein”) within the scale openness to experience- had to be recoded due to its negative
discriminative power.
Table 3. Descriptives, Pearson correlations and psychometrics of handwriting signs and graphological scales
Descriptives (N = 201) Pearson correlations (N = 103) Scale characteristics
(N = 201)
M SD Skew Curt N E O A C rit α-i α-
Neuroticism+ 0.63 0.40 0.73 -0.11 .71
Längenunterschied klein 0.55 0.81 1.16 0.11 .20* .00 -.04 -.14 .05 .43 .68
unauffällig 0.43 0.73 1.51 1.14 .20* -.11 .11 -.06 .02 .06 .72
linksläufig 0.55 0.62 0.68 -0.50 .21* .03 .00 .11 .02 .34 .69
voll 0.82 0.95 0.72 -0.79 .20* .08 -.06 -.09 -.02 .49 .67
binneneng 0.78 0.95 0.88 -0.43 .28** -.01 -.17 -.03 -.05 .54 .66
Wortabstand klein 0.31 0.60 1.81 2.06 .24* .06 .02 -.09 -.10 .29 .70
langsam 0.45 0.76 1.57 1.53 .20* .02 -.03 -.05 -.11 .37 .69
zögernd 0.22 0.49 2.42 6.59 .23* .04 .04 -.15 -.14 .31 .70
Oberzeichen tief 0.78 0.86 0.92 0.16 .23* -.04 -.13 -.20* -.16 .38 .68
endbetont 0.46 0.67 1.24 0.69 .22* -.01 -.01 .01 -.16 .06 .72
schmale Ränder 1.01 1.05 0.53 -1.06 .31*** .17 .15 -.02 -.17 .37 .69
Linksrand schmal 1.14 1.16 0.41 -1.36 .31** .13 .06 -.18 -.26** .44 .68
Neuroticism- 0.85 0.77 0.68 -0.26 .59
Zeile gerade 1.15 1.00 0.21 -1.20 -.24* .04 -.12 .05 .21* .42
Zeile straff 0.56 0.83 1.30 0.66 -.21* -.10 -.09 .15 .20* .42
Extraversion+ 0.70 0.41 0.73 0.13 .66
unübersichtlich 0.38 0.73 1.80 2.15 -.01 .24* .16 -.03 .16 .27 .65
Unterlängen offen 0.82 0.99 0.88 -0.44 -.03 .22* .11 -.03 .07 .29 .65
schlaff 0.19 0.45 2.38 5.13 -.06 .23* .12 -.12 .04 .12 .67
primitiv 0.58 0.82 1.07 -0.11 .11 .35*** .07 -.18 .01 .61 .59
teigig 0.47 0.73 1.52 1.69 .06 .28** .18 -.11 -.01 .15 .67
üppig 0.68 0.86 1.09 0.29 .09 .21* -.07 -.13 -.02 .35 .64
ungeordnet 0.30 0.66 2.14 3.71 .08 .19* .14 .07 -.04 .26 .65
Unterschrift ungleich 1.00 1.15 0.67 -1.09 .08 .28** .04 -.08 -.04 .42 .62
gross 1.04 0.98 0.42 -1.00 .11 .21* .13 -.07 -.05 .29 .65
leserlich 1.71 1.01 -0.36 -0.95 .12 .20* .00 -.01 -.06 .21 .67
undifferenziert 0.54 0.82 1.15 -0.17 .12 .29** .01 -.10 -.07 .50 .61
Extraversion- 0.58 0.42 1.04 0.49 .86
sachlich 0.78 0.83 0.71 -0.45 -.16 -.20* -.02 -.02 .16 .31 .86
Endfaden 0.23 0.57 2.67 7.03 -.08 -.25* .03 .14 .14 .27 .86
gespannt 0.67 0.80 0.91 -0.12 .03 -.25* -.09 .05 .13 .19 .86
Faden 0.34 0.67 1.84 2.17 -.06 -.31*** .09 .08 .11 .48 .85
nüchtern 0.67 0.90 1.05 -0.11 -.10 -.24* .09 -.06 .10 .65 .85
geöffnet 0.63 0.80 0.95 -0.19 -.11 -.20* .03 .05 .07 .58 .85
reserviert 0.35 0.65 1.76 2.17 -.06 -.21* -.01 -.07 .06 .01 .87
leicht 0.53 0.76 1.16 0.19 -.17 -.26** -.05 -.07 .02 .49 .85
schwer leserlich 0.56 0.80 1.21 0.36 -.09 -.22* -.08 -.02 .01 .44 .86
vereinfacht 1.05 0.95 0.49 -0.74 -.15 -.25* -.03 -.03 .01 .68 .84
Doppelbogen 0.36 0.65 1.82 2.80 -.02 -.28** .18 .03 .01 .35 .86
Toggweiler (2018) 7
Descriptives (N = 201) Pearson correlations (N = 103) Scale characteristics
(N = 201)
M SD Skew Curt N E O A C rit α-i α-
kühl 0.54 0.77 1.13 0.06 -.03 -.22* .00 -.05 -.01 .46 .85
Unterlängen mager 0.62 0.88 1.15 0.12 -.03 -.26** -.16 -.20* -.02 .35 .86
rechtsläufig 0.76 0.87 0.72 -0.73 -.09 -.26** -.02 -.08 -.02 .54 .85
offen 0.58 0.88 1.33 0.64 -.02 -.23* .06 .04 -.03 .45 .86
binnenweit 0.51 0.79 1.43 1.07 -.03 -.23* -.06 .05 -.05 .51 .85
karg 0.56 0.90 1.36 0.53 -.03 -.31*** -.01 -.08 -.05 .75 .84
geistig 0.63 0.80 1.12 0.60 -.10 -.26** -.04 -.04 -.06 .54 .85
endunterbetont 0.59 0.76 0.98 -0.12 .06 -.22* -.07 .00 -.09 .27 .86
formschwach 0.59 0.90 1.36 0.77 -.05 -.23* .04 -.14 -.15 .56 .85
Openness to Experience+ 0.49 0.42 0.96 0.82 .38
Unterlängen voll 0.69 0.85 0.90 -0.36 .01 .07 .21* -.06 .05 .10 .41
Druck gestaut 0.29 0.59 2.03 3.60 .12 .17 .23* .12 -.04 .14 .37
schmaler oberer Rand 0.64 1.00 1.30 0.30 .21* .12 .24* .18 -.17 .16 .37
ungleichmässig 0.54 0.76 1.42 1.58 .04 .14 .20* -.03 -.20* .26 .28
unsicher 0.28 0.60 2.26 5.03 .14 .21* .28** -.05 -.25* .37 .22
Openness to Experience- 1.38 0.45 -0.12 -0.39 .57
fein (recodiert) 2.56 0.77 -1.67 1.94 .03 -.07 -.22* -.06 -.15 .04 .60
sicher eingeteilt 1.42 0.97 -0.23 -1.05 -.20* -.15 -.25* -.04 .17 .35 .51
prägnant 1.06 0.92 0.26 -1.05 -.13 -.16 -.27** .06 .14 .44 .48
geordnet 1.50 0.99 -0.15 -1.02 -.10 -.22* -.25* .04 .12 .23 .55
bestimmt 1.79 0.90 -0.45 -0.47 .03 -.01 -.21* -.05 .07 .49 .46
gestaltet 0.98 0.92 0.66 -0.41 .08 .07 -.20* .08 .04 .27 .54
drängend 0.83 0.83 0.44 -1.11 -.06 -.02 -.23* .05 .00 .10 .59
hart 0.92 0.82 0.43 -0.72 .09 -.16 -.30** .06 -.07 .29 .53
Agreeableness+ 0.41 0.42 1.26 2.28 .24
Unterlängen eckig 0.32 0.69 2.23 4.29 -.13 .05 -.02 .21* .14 .15 .12
gelötet 0.24 0.50 2.17 5.51 -.13 -.07 .08 .24* .12 .14 .17
reich 0.67 0.79 0.91 -0.02 .02 -.03 .05 .21* .05 .11 .23
Agreeableness- 0.65 0.93 1.26 0.45 -
breiter oberer Rand 0.65 0.93 1.26 0.45 -.07 -.05 -.18 -.26** .06 - -
Conscientiousness+ 1.40 1.02 -0.11 -1.18 -
gleichmässig 1.40 1.02 -0.10 -1.18 -.05 -.02 -.05 .06 .20* - -
Conscientiousness- 0.46 0.47 1.83 4.22 .71
Zeile schwankend 0.74 0.70 0.51 -0.53 .17 .03 .12 -.16 -.20* .34 .72
unrhythmisch 0.46 0.79 1.52 1.11 .14 .12 -.01 .10 -.23* .55 .64
ungeschickt 0.24 0.58 2.78 8.21 .11 .18 .22* -.02 -.25* .61 .62
gebremst 0.55 0.70 0.89 -0.46 .21* .05 -.07 .11 -.25** .39 .70
haltlos 0.29 0.65 2.21 3.96 .00 .02 .00 -.07 -.32*** .51 .65
Note. * p ** p *** p Potential range of items is 0 to 3. M = Mean, SD = Standard deviation, Skew = Skewness, Curt =
Curtosis, rit = Discriminative power, α = Cronbach’s Alpha, α-i = Cronbach’s Alpha without item. Bold are convergent correlations of
handwriting signs and NEO-PI-R.
Question 2: Explained Variance
The following Table 4 shows the Pearson correlations of the graphological scales with the Big Five
dimensions of the NEO-PI-R. The significant convergent correlations range from .20 to .49, which
corresponds to an explained variance of 4.00 respectively 24.01 %. Five out of ten correlations achieve
medium convergent validity, i.e. the correlations reach values between .40 and .59.
Toggweiler (2018) 8
Table 4. Pearson correlations and explained variance between graphological scales and NEO-PI-R.
N E O A C
Neuroticism+ .44*** (r2 = .19) .07 -.01 -.15 -.18
Neuroticism- -.27** (r2 = .07) -.03 -.13 .12 .25*
Extraversion+ .13 .45*** (r2 = .20) .14 -.13 -.01
Extraversion- -.11 -.40*** (r2 =.16) -.01 -.05 .02
Openness to
Experience+ .19 .24* .42*** (r2 = .18) .06 -.23*
Openness to
Experience- -.08 -.19 -.49*** (r2 = .24) .04 .10
Agreeableness+ -.09 -.04 .06 .35*** (r2 = .12) .15
Agreeableness- -.07 -.05 -.18 -.26** (r2 = .07) .06
Conscientiousness+ -.05 -.02 -.05 .06 .20* (r2 = .04)
Conscientiousness- .17 .11 .06 .00 -.35*** (r2 = .12)
Note. N = 103. * p ** p *** p ld are medium validity coefficients (.40 r < .60).
Results of linear multiple regressions indicated that all five prediction models of the NEO-PI-R
dimensions were statistically significant, whereas for neuroticism F(10, 92) = 4.07, p <.001, for
extraversion F(10, 92) = 6.14, p < .001, for openness to experience F(10, 92) = 7.02, p < .001, for
agreeableness F(10, 92) = 3.28, p = .001 and for conscientiousness F(10, 92) = 2.33, p < .05.
Toggweiler (2018) 9
N
euro
tici
sm
Extr
aver
sion
O
pen
nes
s to
exp
erie
nce
A
gre
eab
len
ess
Con
scie
nti
ou
snes
s
B
S
E(B
) 𝛽
t
B
SE
(B)
𝛽
t B
S
E(B
) 𝛽
t
B
SE
(B)
𝛽
t B
S
E(B
) 𝛽
t
Con
stan
t 6
5.6
1
14.4
6
4
.54
***
129
.96
11.6
5
1
1.1
5***
145
.33
9.1
9
1
5.8
2***
136
.59
11.2
0
1
2.1
9***
123
.75
14.3
9
8
.60
***
Neu
roti
cism
+
31.4
5
6.2
3
.64
5.0
5***
-21
.93
5.0
2
-.5
2
-4.3
7***
-8.7
3
3.9
6
-.2
5
-2.2
0*
-8.6
6
4.8
3
-.2
3
-1.7
9
-11
.55
6.2
0
-.2
5
-1.8
6
Neu
roti
cism
- -6
.02
2.9
9
-.2
0
-2.0
1*
0.1
7
2.4
1
.01
0.0
7
-3.5
8
1.9
0
-.1
7
-1.8
8
2.1
0
2.3
2
.09
0.9
1
2.9
9
2.9
8
.11
1.0
0
Extr
aver
sion
+
-8.3
6
6.8
4
-.1
8
-1.2
2
16.9
7
5.5
1
.43
3.0
8**
2.2
1
4.3
4
.07
0.5
1
-9.0
1
5.3
0
-.2
6
-1.7
0
11.9
3
6.8
1
.28
1.7
5
Extr
aver
sion
- 9
.30
6.5
0
.20
1.4
3
-17
.16
5.2
4
-.4
4
-3.2
8***
-3.4
2
4.1
3
-.1
1
-0.8
3
-13
.48
5.0
4
-.3
9
-2.6
8**
-1.6
0
6.4
7
-.0
4
-0.2
5
Op
enn
ess
to
exp
erie
nce
+
3.6
2
6.0
6
.07
0.6
0
7.1
7
4.8
8
.17
1.4
7
14.8
4
3.8
5
.43
3.8
6***
-1.2
3
4.6
9
-.0
3
-0.2
6
-4.5
8
6.0
3
-.1
0
-0.7
6
Op
enn
ess
to
exp
erie
nce
- 0
.15
5.3
4
.00
0.0
3
-8.4
8
4.3
0
-.2
0
-1.9
7
-17
.08
3.3
9
-.4
9
-5.0
3***
-2.0
8
4.1
4
-.0
6
-0.5
0
-2.2
5
5.3
2
-.0
5
-0.4
2
Ag
reea
ble
nes
s+
0.7
4
6.1
8
.01
0.1
2
-1.5
6
4.9
8
-.0
3
-0.3
1
-0.2
4
3.9
3
-.0
1
-0.0
6
12.1
9
4.7
9
.27
2.5
4*
6.0
2
6.1
5
.11
0.9
8
Ag
reea
ble
nes
s- -0
.37
2.1
7
-.0
2
-0.1
7
-0.7
6
1.7
5
-.0
4
-0.4
3
-0.8
7
1.3
8
-.0
5
-0.6
3
-4.8
2
1.6
8
-.2
8
-2.8
6**
1.3
7
2.1
6
.06
0.6
3
Con
scie
nti
ou
snes
s+
-0.2
2
2.8
4
-.0
1
-0.0
8
1.1
2
2.2
9
.06
0.4
9
4.3
5
1.8
0
.29
2.4
1*
1.6
4
2.2
0
.10
0.7
5
-0.6
3
2.8
2
-.0
3
-0.2
2
Con
scie
nti
ou
snes
s- 0
.59
5.3
2
.01
0.1
1
-4.1
6
4.2
8
-.1
2
-0.9
7
-6.5
6
3.3
8
-.2
3
-1.9
4
6.7
6
4.1
2
.22
1.6
4
-11
.41
5.2
9
-.3
0
-2.1
6*
R
2 =
.31
R2 =
.40
R2 =
.43
R2 =
.26
R2 =
.20
R
2 a
dj
= .23
R2 a
dj
= .3
4
R2 a
dj
= .3
7
R2 a
dj
= .1
8
R2 a
dj
= .1
2
R
ad
j =
.48
R a
dj
= .58
R a
dj
= .61
R a
dj
= .43
R a
dj
= .34
K
2 =
.3
0 (
med
ium
) K
2 =
.5
0 (
larg
e)
K2 =
.5
9 (
larg
e)
K2 =
.2
2 (
med
ium
) K
2 =
.1
3 (
smal
l)
Tab
le 5
. M
ult
iple
reg
ress
ion
an
alysi
s p
red
icti
ng t
he
NE
O-P
I-R
dim
ensi
on
s.
No
te. *
p
** p
***
p
Bo
ld -
sig
nif
ican
t B
eta-
wei
ghts
. Th
e ef
fect
po
wer
fo
r m
ult
iple
lin
ear
regr
essi
on
is c
alcu
late
d a
s fo
llow
ers
(Bo
rtz
& D
öri
ng,
20
06
, p. 6
06
): K
2 =
(R2 a
dj /
(1
- R
2 a
dj)
, wh
ere
K2 ≥
0.0
2 (
smal
l
effe
ct),
K2 ≥
0.1
5 (
med
ium
eff
ect)
, K2 ≥
0.3
5 (
larg
e ef
fect
).
Toggweiler (2018) 10
The raw and standardized regression coefficients of the predictors together with the standard errors
are shown in Table 5. The predictors relate, with the exception of NEO-PI-R-conscientiousness, as
expected. That is these NEO-PI-R dimensions have significant Beta-weights in both corresponding
graphological scales. In the case of extraversion, openness to experience and conscientiousness, other
scales also significantly account for explained variance: Neuroticism+ in the case of extraversion,
neuroticism+ and conscientiousness
+ in the case of openness to experience and extraversion
- in the case
of agreeableness.
The predictors accounted for about 23% of the variance in neuroticism (R2 = .31, R
2 adj = .23, R
adj = .48, K2 = medium), for about 34 % in the case of extraversion (R
2 = .40, R
2 adj = .34, R adj = .58,
K2 = large), for 37 % in the case of openness to experience (R
2 = .43, R
2 adj = .37, R adj = .61, K
2 =
large), for 18 % in the case of agreeableness (R2 = .26, R
2 adj = .18, R adj = .43, K
2 = medium) and for
12 % in the case of conscientiousness (R2 = .20, R
2 adj = .12, R adj = .34, K
2 = small). The stepwise
exclusion of handwriting signs with insufficient discriminative power (in the sense of scale
optimisation) always led to a lower extent of explained variances.
DISCUSSION
Sample
The two samples (overall sample and NEO-PI-R sample) are not representative of a general
population; they include consistently high educational qualifications (national vocational qualification,
upper secondary school, upper secondary school and higher education). However, representativeness
is not required for the current questions, since the primary interest was the method (the construction of
handwriting sign scales by means of correlation analysis) and its potential (explained variance of the
NEO-PI-R). From this point of view the homogeneous setting of the samples can even be seen as an
advantage, as it allows statements about a relatively well-defined target group of high educational
persons. Of course, these high educational level corresponds with an increased age of the sample. The
approximatly equal distribution of gender is ideal, because it leads to a balanced gender-specific
influence on the results.
Question 1: Construction of graphological scales
A first attempt (Bühler, 2015) to construct graphological scales by means of cluster analysis was
successful, since 13 scales could be constructed with high reliability in terms of Cronbach's Alpha.
Unfortunately, these scales neither correlatively, nor by means of multiple linear regressions are
significantly associated with the Big Five dimensions of the NEO-PI-R in a considerable extent. In this
respect, question 1 can be answered in such a way that it is easily possible to construct reliable
graphological scales in terms of Cronbach’s Alpha. However, the internal structure revealed by means
of cluster analysis does not relate to the Big Five personality traits. It is unclear what this structure
correlates with. Unfortunately, the sample was too small for factor analysis because that would have
been a worthwhile attempt: Scaling by means of factor analysis.
The second attempt to construct graphological scales using correlation analysis was not successful
immediately. The prerequisites for parametric analyses were not ideal, because one quarter of the
handwriting signs were not normally distributed, but nevertheless were unimodal. In these
distributions nothing can be optimized, since the handwriting signs, unlike verbal items, cannot simply
be formulated more appropriately. The method for selecting the proper handwriting signs from the
total of 234 handwriting signs consisted of constructing scales with the most significant positively and
negatively correlated characteristics with the NEO-PI-R dimensions. The discriminative power of the
handwriting signs was finally inadequate at about one third, which will of course be detrimental to the
results, however, good results under unfavourable conditions are a strong fact. The internal
consistence in terms of Cronbach’s Alpha of the graphological scales finally turned out to be not
conclusive. They range from insufficient .24 to pleasant .86. Two graphological scales (agreeableness-
and conscientiousness+) contain only one single item, therefore they are not actual scales but rather
Toggweiler (2018) 11
single characteristics. Nevertheless, the interrater reliability of the included 68 handwriting signs must
be quite good, otherwise not one satisfying results would have been revealed by the analyses.
As answer to question 2 can be noted: The first attempt (scaling by means of cluster analysis) has
shown that it is absolutely possible to construct reliable graphological scales in terms of Cronbach’s
Alpha – we simply do not know what they correlate with; not with personality – and certainly not
directly with the dimensions of the Big Five. The second attempt (scaling by means of correlation
analysis) must also be accounted for as failed: With this method, only three reliable graphological
scales could be constructed – one third of which are inadequate in their discriminative power. Ten
reliable graphological scales would have been necessary. The value of this method is recognized if this
solution is used in the attempt to answer question 2.
Question 2: Quality of the solution
The convergent correlations of the graphological scales with the Big Five of the NEO-PI-R (Table 4)
already showed a medium validity in five out of ten cases (neuroticism+, extraversion
+, extraversion
-,
openness to experience+, and openness to experience
-), the explained variance is at least 16 %, which
corresponds to a correlation of at least .40. This finding is already a positive result and justifies the
diagnostical use of five out of ten graphological scales for the Big Five dimensions neuroticism,
extraversion and openness to experience. The result is even better with the help of linear combinations
of the graphological scales: With the primary goal of a maximizing the explained variance, all
graphological scales were in into five linear multiple regressions onto the Big Five of the NEO-PI-R.
The result was one small (conscientiousness), two medium (neuroticism, agreeableness) and two large
(extraversion, openness to experience) effects, i.e. a total of four considerable effects – agreeableness
is also present now. As a side note: Although the discriminatory power of approximately one third of
the included handwriting signs was insufficient (as far as calculable), an exclusion of these items did
not lead to an improvement of the results in any case, i.e. to an improvement of the validity of the
graphological scales respectively to an improvement of the explained variance of the NEO-PI-R. In
this respect, the numbers, apart from the explained variance of conscientiousness, express a good and
usefull quality of the solution – one may predict the Big Five with a quiet conscience in four out of
five dimensions by means of a linear combination of the defined graphological scales.
The performance of these graphological scales compared to paper-pencil tests
How considerable these reported effects are can be illustrated by the correlations between self- and
observer-assessment of the Big Five with various verbal instruments. The data in Table 6 is taken from
the manual of the NEO-PI-R (Ostendorf & Angleitner, 2004, p. 142) as well as from a recent study
from Lee and Ashton (2016, p. 8). We assume that graphological assessment is a kind of observer-
assessment.
Toggweiler (2018) 12
Table 6. Explained variances among different self- and observer-assessments.
Big Five dimension
Neu
ro
ticis
m
Ex
tra
ver
sio
n
Op
enn
ess
to
Ex
per
ien
ce
Ag
reea
ble
nes
s
Co
nsc
ien
tio
usn
ess
NEO-PI-R self-assessment
Graphological scales1 (cf. Tab. 5) R2adj (effectsize K2) .23 (M) .34 (L) .37 (L) .18 (M) .12 (S)
NEO-PI-R observer-assessment2 r2(effectsize r) .38 (M) .40 (M) .31 (M) .30 (M) .41 (M)
BARS 179 observer-assessment2 r2(effectsize r) .32 (M) .22 (S) .14 (S) .17 (S) .27 (S)
BARS1794 self-assessment
BARS 179 observer-assessment2 r2(effectsize r) .41 (M) .30 (M) .18 (S) .14 (S) .35 (M)
NEO-PI-R observer-assessment2 r2(effectsize r) .35 (M) .20 (S) .18 (S) .16 (S) .27 (S)
HEXACO-PI-R-100 self-assessment
HEXACO-PI-R-100 self-assessment3 r2(effectsize r) .37 (M) .31 (M) .31 (M) .22 (S) .27 (S)
Note. S = Small, M = Medium, L = Large. All measures are significant at p BARS179 = Bipolar adjective-rating-scales. 1) N = 103. 2) The
correlations are based on different sample sizes between N = 83 and N = 573. 3) N = 2863. 4) Ostendorf (1990). In bold are the convergent
correlations of the graphological scales with the NEO-PI-R.
As seen from Table 5 and taken up again in Table 6, the graphological scales reach explained
variances in percentages of small (conscientiousness), medium (neuroticism, agreeableness) and large
effect sizes (extraversion, openness to experience). If one compares these numerical results with the
other values in the respective columns (i.e. the explained variance that were achieved by verbal
observer-assessment), one can see that the explained variances in the columns extraversion and
openness to experience only once respectively not at all were higher. Although neuroticism explained
by graphological scales is numerically outperformed by the other explained variances in the column, it
still maintains a medium effect size. The same applies to conscientiousness, which is also surpassed by
all other explained variances. In contrast, the graphologically explained variance of agreeableness is
approximately in the midfield of the values in the column. The balance for the graphological scales is
an approximatly middle-ranking position: Extraversion and openness to experience perform very well
and are, with both large effect sizes, winners over verbal observer-assessments. Agreeableness is in
the midfield whereas neuroticism and conscientiousness are better explained by the verbal observer-
assessment instruments. It has to be said that the explained variance calculated by means of the
adjusted R2, represents a very strict measure. Its upper limit is the (unadjusted) R
2, which reaches at
minimum the same or even much higher values than the adjusted R2. The true explained variance is
somewhere between the adjusted and the unadjusted R2 (see Table 5).
In addition, the qualifications of the authors of the NEO-PI-R (Ostendorf & Angleitner, 2004) and
NEO-FFI (Borkenau & Ostendorf, 2008) are to be taken into account, such as reported in their
manuals, concerning validity of their instruments: "The convergence of [...] the main scales can be [...]
considered as satisfactory. On average, the self-observer-agreement for the main scales [of the NEO-
PI-R] is .54 [r2 = .29] and for the facets is .47 [r
2 = .22]" (Ostendorf & Angleitner, 2004, p. 140).
Exactly the same average self-observer-agreement of R2 adj = .29, calculated by Fisher's z
transformation, is obtained with the graphological scales. Borkenau and Ostendorf (2008, p. 26) had a
similar conclusion about their NEO-FFI: "These agreements [.24 < r2 < .37] are comparatively high
Toggweiler (2018) 13
and show that there exists [between observer- and self-assessments with the NEO-FFI] a high
agreement regarding the expression of these five personality traits [...]. The self-descriptions done by
the test persons are therefore highly realistic in terms of inter-personel agreement”. This qualification
(we achieve .12 R2 adj .37) at least partially certifies the graphological scales a "comparatively
high" agreement between observer- and self-assessment.
From all these findings, it can be concluded that the regression equations defined in Table 5 for the
Big Five dimensions neuroticism, extraversion, openness to experience and agreeableness allow a
valid prognosis of the NEO-PI-R personality traits. In other words, the validity is high enough to feel
disclosed within the Big Five dimensions. The similarities are as considerable as it is to be expected
between self- and observer-assessments - not more and not less. The two graphological scales for the
dimension conscientiousness (especially conscientiousness+) should be further optimized. Considering
that these calculations were carried out in a straightforward manner, there is certainly potential for
even better coefficients of validity.
FUTURE DIRECTIONS
The available results are very encouraging and justify the implementation of new algorithms in
GraphoPro. Nevertheless, an urgent methodological difficulty must be addressed here: Due to the high
number of correlations that were necessary to construct the graphological scales (5 * 234 = 1170
correlations), it has to be considered, that various significant correlations between handwriting signs
and NEO-PI-R may have occurred just accidentally. At a significance level of p = .05 one should
expect 59 just by chance significant relations at a total of 68 handwriting signs, which were found to
be relevant (Table 3). This uncertainty, i.e. the question of the role of chance, can be resolved in two
ways:
1. One increases the sample size from currently 103 to about 600 subjects (Rho (H1) = .20, Rho
(H0) = 0, p < .000043, Power = .80) and lowers the significance level of the correlations to p =
.05/1170 = .000043 by means of the Bonferroni correction.
2. A replication study is carried out with the 68 handwriting signs and the ten graphological
scales. On the basis of the correlations and regression values, it will be possible to evaluate the
influence of handwriting signs that were found to be relevant just by chance.
Unfortunately, the concern just described cannot be solved in this study. From these two options,
replication is certainly the more economic one. The existence of a larger sample would also allow the
use of much more elaborated methods (factor analysis or the use of structural equation modelling). As
well the unfavourable differences between the explained variance R2 and the adjusted R
2 would be
reduced due to the larger sample size. Furthermore, the inclusion of other observer ratings would be
interesting in order to test them against the graphological scales – it is a matter of competitive validity
or what graphology achieves independently of verbal diagnostic instruments.
Toggweiler (2018) 14
REFERENCES
Borkenau, P. & Ostendorf, F. (2008). NEO-FFI. NEO-Fünf-Faktoren-Inventar nach Costa und
McCrae. Bern: Hogrefe.
Bortz, J. & Döring, N. (2006). Forschungsmethoden und Evaluation für Human- und
Sozialwissenschaftler (4. Aufl.). Heidelberg: Springer Medizin Verlag.
Bühler, E. (2015). Die Konstruktvalidität der Schriftpsychologie – Eine Validitätsuntersuchung von
Schriftmerkmalen (unpubl. Bachelor-Thesis). Zürich: Zürcher Hochschule für Angewandte
Wissenschaften ZHAW, Psychologisches Institut.
Costa, P. T. & McCrae, R. R. (1992). Revised NEO Personality Inventory (NEO PI-R) and NEO Five
Factor Inventory. Professional manual. Odessa, FL: Psychological Assessment Resources.
Digman, J. M. & Takemoto Chock, N. K. (1981). Factors in the natural language of personality: Re-
analysis, comparison, and interpretation of six major studies. Multivariate behavioral research,
16, 149-170.
Keel, B. (2016). Computeruntersützte Graphologie mit GraphoPro. Graphologie News, Juli/August
(http://graphologie-
news.net/cms/upload/archiv/Computerunterstuetzte_Graphologie_mit__GraphoPro.pdf)
Lee, K. & Ashton, M. C. (2016). Psychometric Properties of the HEXACO-100. Assessment, 0(0).
doi: 10.1177/1073191116659134.
McCrae, R. R. & Costa, P. T. (1996). Towards a new generation of personality theories: Theoretical
contexts for the five-factor model. In J. S. Wiggins (Ed.), The five-factor model of personality
(pp. 51-87). New York: Guilford Press.
McCrae, R. R. & Costa, P. T. (1999). A five-factory theory of personality. In L. A. Pervin & O. P.
John (Eds.), Handbook of personality. Theory and research. New York: Guilford Press.
Ostendorf, F. (1990). Sprache und Persönlichkeitsstruktur. Zur Validität des Fünf-Faktoren-Modells
der Persönlichkeit. Regensburg: Roderer.