Beruflich Dokumente
Kultur Dokumente
ORIGINAL PAPER
Graphology for the diagnosis of suicide attempts: a blind
proof of principle controlled study
S. Mouly,1 I. Mahé,1 K. Champion,1 C. Bertin,1 P. Popper,2 D. De Noblet,2 J. F. Bergmann1
Table 1 Number of correct graphological diagnoses (%) of parasuicide by two graphologists and two internists after reading letters from patients
following attempted suicide (n = 40) and healthy controls (n = 40)
Graphologists Internists
ITT analysis
Suicide (n = 40) 31 (77%) 31 (77%) 32 (80%) 26 (65%) 27 (68%) 27 (68%)
Control (n = 40) 33 (83%) 31 (77%) 33 (83%) 34 (85%) 31 (77%) 34 (85%)
Per-protocol analysis
Suicide (n = 30) 21 (70%) 22 (73%) 22 (73%) 16 (53%) 16 (53%) 16 (53%)
Control (n = 38) 33 (87%) 31 (82%) 33 (87%) 34 (89%) 31 (82%) 34 (89%)
was no statistical difference regarding the accuracy written by subjects who had survived such an
of the diagnosis between the graphologists and int- attempt. These subjects were known to have
ernists (p = 0.45). attempted suicide and there was no risk of diag-
For the per-protocol (PP) analysis, 12 letters nostic error amongst the suicide population. We
expressing sadness were excluded: 10 from the sui- therefore preferred not to evaluate their degree of
cide group and two from the control group. Ele- depression by asking them to complete a MADRS
ven of these were classified as ‘suicide’ by both questionnaire, so that they did not feel the study
teams of evaluators in the ITT analysis. In the PP to be a psychometric test, which might have influ-
analysis, the characteristics of the graphological enced their writing. We performed the study in
test for graphologists and internists after confron- patients who had recovered a normal level of
tation were, respectively, sensitivity 73 and 53%, alertness, on the day of discharge from the hospi-
specificity 88 and 89%, positive predictive value tal. To keep the study blinded, patients were asked
81 and 80%, negative predictive value 82 and to write a letter not related to their parasuicide or
71%. Both the graphologists and the internists their mental health status or history. The control
remained significantly different to chance in the subjects did not show any depressive or suicidal
PP analysis (p < 0.001 for both evaluator groups). tendencies, and we believe that the populations
After initial evaluation of the 80 letters, the two studied were correctly chosen for measurement of
graphologists disagreed on the diagnosis of 12 let- the metrological properties of graphology. Our
ters; the consensus session led to a final correct findings show that there truly are detectable differ-
diagnosis in eight letters. For the internists, diver- ences between the writing of subjects who have
ging opinions were observed in 22 letters and a attempted suicide and healthy subjects even if both
consensus was obtained with a final correct diag- graphologists missed the suicidal intent in 23% of
nosis in 14. cases. The fact that this difference was also detec-
ted by practising internists who were not graphol-
ogists could represent an argument against the
Discussion
specific and rigorous nature of this technique. But
This pilot study is the first controlled trial to have one may also consider that the principles of
demonstrated the ability of graphologists to detect a graphological diagnosis are based on good sense
particular psychiatric status, i.e. parasuicide with a and a logic which is perceptible by nonspecialists.
sensitivity and specificity around 80%. However, the graphologists were slightly though
Graphology is based on a theory according to not significantly better than the internists in this
which the shape, rhythm, size and position of the study with a sensitivity of 73 and 53%, respect-
handwritten letters and words reflect the character ively.
of the writer (5). But the correlation between the Our study had some limitations. The study eval-
writing and the psychological profile is essentially uated the accuracy of only two graphologists in
based on case reports in which graphological char- only one medical situation without follow-up of
acteristics are described with full knowledge of the the patients. Our results could not be extrapolated
psychological characteristics of the subject (9). to all graphologists and to other psychological dis-
However, it would be very useful to be able to orders. The grounds which led the graphologists to
precisely determine the psychological status of an consider the letters to have been written by sub-
individual by studying his ⁄ her handwriting. If jects who had attempted suicide are not unequivo-
handwriting could define a dangerous situation cal. There is no global or composite score, but
such as, for example, a suicide risk, and if graph- factors such as falling lines, a loss of the link
ological analysis could predict this risk with a between the letters, an heterogeneity of the graph-
good level of predictive value, it could become a ics, a weak holding of the lines, a change in the
tool in the decision-making process (10). The disposition of the words, an unusual possession of
decision to hospitalise or discharge a depressive the space of the page are taken into account. The
patient is always difficult to take due to the risk subjective dichotomous global decision was built
of suicide which, if incorrectly assessed, can have on the size, shape, pressure, speed and movement
serious consequences (11,12). We therefore chose of the letters and words and on the signature.
this parasuicide model in order to try to validate However, no graphic element in itself was purely
the graphological diagnosis. As graphologists significant of one definite quality leading to a pre-
believe that suicidal tendencies continue to be diction of suicide.
reflected in handwriting after a suicide attempt, it As we did not include a formal depression rat-
was methodologically justified to study the letters ing scale to evaluate subjects who had attempted
suicide, it was impossible to search for a correla- positives and 13% of false negatives, this study
tion between handwriting analysis and level of yielded positive results and should lead to further
depression. We preferred to keep a more powerful studies. Prospective studies with a follow-up of
binary primary end-point in accordance with our patients with various graphological diagnoses of
sample size calculation. A previous study compar- psychiatric diseases would allow us to ascertain
ing the assessment of 10 graphologists in the diag- whether graphology provided additional informa-
nosis of extraversion showed good agreement tion to the usual psychological diagnosis. In this
between the graphologists, but in this study, only case, graphology could act as a complement to
six subjects were evaluated without any healthy standard methods and become an additional tool
control group (6). In another study (8), graphol- in therapeutic decision-making (14). Indeed it
ogy was compared with results of psychological should be noted that in psychiatry, many therapeu-
tests for vocational guidance. tic decisions are taken on the basis of impressions
The experimental conditions of our pilot study the predictive value of which has been even less
certainly do not allow definitive conclusions to be well studied than the graphology in the present
drawn as to the validity of graphology. Our design study (15).
with a dichotomic answer of suicide yes–no greatly We can conclude that graphology is able to differ-
facilitates graphological analysis. The content of the entiate letters written by patients who attempt sui-
letters themselves may influence analysis of the cide from those written by healthy subjects with an
handwriting; indeed, 11 of the 12 letters expressing acceptable degree of accuracy. This difference is not
sadness were classified in the suicide group. How- due to the ideas expressed in the content of the let-
ever, even after the exclusion of these letters, ters.
graphological analysis still had a predictive value of
about 80%. Our study should be considered as a
preliminary report, a kind of go–no go study. Had
References
our findings been negative, we would have demon-
strated, using a solid methodology, that graphology 1 Hawton K, Arensman E, Townsend E et al. Deliberate self
harm: systematic review of efficacy of psychosocial and pharma-
was not reliable in the diagnosis of psychiatric ill- cological treatments in preventing repetition. BMJ 1998; 317:
nesses such as suicide. Conversely, our positive 441–7.
results are not sufficient to provide overall valid- 2 Haste F, Charlton J, Jenkins R. Potential for suicide prevention in
primary care ? An analysis of factors associated with suicide. Br J
ation of this method. A future study using letters
Gen Pract 1998; 48: 1759–63.
written by foreign patients in a language unknown 3 Modestin J, Kamm A. Parasuicide in psychiatric inpatients: results
by the graphologists and ⁄ or mixing several differ- of a controlled investigation. Acta Psychiatr Scand 1990; 81: 225–
ent types of clearly differentiated psychiatric illnes- 30.
4 Leon AC, Friedman RA, Sweeney JA, Brown RP, Mann JJ. Statisti-
ses and letters written by healthy volunteers would
cal issues in the identification of risk factors for suicide behavior:
allow our findings to be clarified. Of course, the the application of survival analysis. Psychiatry Res 1990; 31:
prediction of future suicide risk is clinically more 99–108.
important than establishing a previous suicide 5 Brewer JF. Graphology. Complement Ther Nurs Midwifery 1999; 5:
6–14.
attempt but needs longer and larger prospective 6 van Rooij JJ, Hazelzet AM. Graphologists’ assessment of extraver-
studies and was not the goal of our proof of prin- sion compared with assessment by means of a psychological test.
ciple trial. Percept Mot Skills 1997; 85: 919–28.
In conclusion, we have shown that graphological 7 Ludewig R, Dettweiler C, Lewinson TS. Possibilities and limits of
medical graphology. Determination current status and perspectives
analysis is different to chance in evaluating the (I). Z Gesamte Inn Med 1992; 47: 549–57.
presence of a suicide risk in handwriting, but fur- 8 Super DE. A comparison of the diagnoses of a graphologist with
ther studies are needed before the true role of this results of psychological tests. J Consult Clin Psychol 1992; 60: 323–
6.
technique in the diagnostic process can be estab-
9 Crumbaugh JC, Stockholm E. Validation of graphoanalysis
lished. by ‘global’ or ‘holistic’ method. Percept Mot Skills 1977; 44:
An association between words used to describe 403–10.
handwriting features and personality traits might 10 Zalsman G, Netanel R, Fischel T et al. Human figure drawings in
the evaluation of severe adolescent suicidal behavior. J Am Acad
be a reason for our positive results (13). A high
Child Adolesc Psychiatry 2000; 39: 1024–31.
risk for suicide might modify the form or the 11 Power K, Davies C, Swanson V, Gordon D, Carter H. Case-
position of the letters, the words or the lines in a control study of GP attendance rates by suicide cases with or
letter and ⁄ or a signature, but the goal of our study without a psychiatric history. Br J Gen Pract 1997; 47: 211–
5.
was not to attempt to explain the theory of graph-
12 Hawton K, Zahl D, Weatherall R. Suicide following deliberate self-
ology but simply to evaluate the value of graphol- harm: long-term follow-up of patients who presented to a general
ogy as a diagnostic test. Even with 27% of false hospital. Br J Psychiatry 2003; 182: 537–42.
13 King RN, Koehler DJ. Illusory correlations in graphological infer- 15 Modestin J, Wurmle O. Role of modelling in in-patient suicide:
ence. J Exp Psychol Appl 2000; 6: 336–48. a lack of supporting evidence. Br J Psychiatry 1989; 155:
14 Knox KL, Litts DA, Talcott GW, Feig JC, Caine ED. Risk of suicide 511–4.
and related adverse outcomes after exposure to a suicide preven-
tion programme in the US Air Force: cohort study. BMJ 2003;
327: 1376–8. Paper received August 2005, accepted March 2006