Two methods of Author Identification: the Gary/Ajar case. author style statistics

1. The Romain Gary/Émile Ajar
controversy
In 1974 the well-established French author Romain Gary, a recipient of the Prix Goncourt (France’s highest literary award) wanting to escape
from “le parisianisme” and the context in which
critics and readers alike had pegged him, published Gros-Câlin under the pseudonym of Émile
Ajar. By adopting this new name, he wanted to
start over and have his work judged on its own
merits and not on his established reputation. His
stratagem payed off: Gros-Câlin immediately attracted the attention of critics and readers alike,
became a best-seller while new novels published
under the Gary name were not as successful. When
a few astute critics noticed similarities between
Gary and Ajar, Gary vehemently denied having
any connection with Ajar. But when these rumours
persisted, Gary fearing that he might be found out,
coaxed Paul Pavlowich, his nephew, to impersonate Ajar. To quash any further rumour that he was
Ajar, he even accused Ajar of plagiarising him.
With increasing paranoia, he wrote a second Ajar
novel, La Vie devant soi which became an immediate success and was awarded the Prix Goncourt.
Romain Gary thus became the first author (and
presumably the last one) to receive this award
twice, which is strictly forbidden by the Goncourt
academy.
It was after his suicide in 1980, that two books Vie
et Mort d’Émile Ajar by Romain Gary (published
posthumously in 1981) and L’Homme que l’on
croyait (1981) by Paul Pavlowich enabled readers
to demystify the double disguise: Ajar was the
pseudonym of Gary and not of Paul Pavlowich.
Critics immediately saw similarities between the
Gary and the Ajar novels in terms of ideas, characters, images, recurring motifs and phrasings.
But so far, no one has undertaken a comparative
analysis of the Gary and Ajar style using statistical
methods. If Buffon’s assertion that “le style c’est
l’homme même” is true, the Émile Ajar corpus
should prove to be statistically similar to the Romain Gary corpus. Romain Gary thus provides an
excellent example for authorship attribution study, or stylometric study in it’s broad sense. Authorship attribution study is the analysis of stylistic
idiosyncracies of an author as an index of authenticity, or an attempt to capture quantitatively the
essence of an author’s use of language. However,
the purpose of this paper is not to attribute unknown works to Romain Gary (we now know that
Gary and Ajar are the same author) but rather to
compare his style in two novels written under the
pseudonym of Émile Ajar to two novels written
under the Romain Gary name.
2. Hypothesis
Nearly all experts from Buffon to Barthes postulate that style, which is dictated by the subconscious, forms the genetic fingerprint of a writer’s
work. It is therefore impossible to disguise one’s
style. It would then follow that works written
under a pseudonym should contain the genetic
fingerprint of the original writer. A stylistic analysis of the pseudonymous corpus would reveal
that it is statistically similar to the work of the
original writer. Therefore, there should be no significant difference between the style of Romain
Gary and that of Émile Ajar.
The problem with this assumption, however, is
that there is no hard evidence to support the idea
that authors have an unconscious as well as a
conscious aspect to their style. The two applications of author recognition, namely (1) authorship
attribution and (2) chronological studies, have
made contradictory claims. The attribution method claims that the unconscious aspect of a writer’s style remains constant throughout his/her
life, in other words, an author will leave his/her
stylistic fingerprint on each and every one of
his/her works. However the chronological studies
claim that the unconscious features change
throughout the author’s life and develop rectilinearly which then allow a work to be dated.
3. Methodology
Since the findings of Mosteller and Wallace on the
Federalist papers in 1964, linguists, statisticians
and literary critics alike use the method of stylistic
fingerprinting to attribute authorship to disputed
works. Like many experts, they assert that style
which comes from the unconscious mind, forms
the genetic fingerprint of any writing and helps
distinguish one author from the next. While the
methods used are not the same, statistical models
play an important role for their findings.
The various methods used for the determination of
authors and the measurement of style have been:
word-lengths (Mendenhall, 1887), (Brinegar,
1963), (Mosteller and Wallace, 1964); number of
syllables per word (W. Fucks, 1952); sentencelength (Yule, 1938), (Williams, 1940) and Kjetsså
(1979) etc. But this paper focuses on vocabulary
distribution as a general style discriminant. The
analysis deals mostly with (a) high frequency
words and (b) synonyms as discriminants of style.
In their analysis of The Federalist Papers Mosteller and Wallace focus their research mainly on the
use of synonyms such as “while” and “whilst” as
style discriminants to make their conclusions,
while J.F. Burrows in Computation into criticism:
A Study of Jane Austen’s novels and an experiment
in Method concentrates on the use of high frequency words as an essential element of an author’s
style. Burrow’s assertion (and the one shared in
this paper) rests on the premise that the essential
element of an author’s style is not confined in the
rare lexical words likely to evoke love, hate or war,
but in the forty or fifty unambiguous and most
common word types in the entire corpus. Although
this method has been put into question by F.J.
Damereau in 1975 it is still widely used and is the
one used in this paper.
In order to test if indeed the Gary corpus is statistically similar to the Ajar corpus, four books were
scanned, two by Romain Gary: Au-delà de cette
limite votre ticket n’est plus valable (1975) and
Clair de femme (1977), plus two Émile Ajar
books: Gros-Câlin (1974) and La Vie devant soi
(1975). As Romain Gary’s literary career spanned
nearly thirty years, these four books, all written
within a four year period, were chosen in order to
avoid the problem of chronology. After the scanning process, alphabetical concordances and word
counts (using the O.C.P.) were established. From
these another programme sorted out the words in
descending order yielding a list of highest frequency words in the Ajar and Gary novels. But as
testing Ajar against Gary would not have been
conclusive, other French twentieth century novels
were included in the tests: Camus’ L’Étranger
(1942); Gide’s L’Immoraliste (1902) and La Porte
Étroite (1909); Mauriac’s Le Noeud de vipères
(1932). For these novels, a series of texts held in
the ARTFL database were used. These five books
were chosen because they are of similar lengths to
the Gary and Ajar books, they belong to the same
genre (the novel) and in all of them, as in the Gary
and Ajar novels, the narrator is the first person
singular. Furthermore a list of high frequency
words compiled by Engwall in 1974 and made up
of the most common words found in twentyfive
best sellers in France from 1962 to 1968 was
included in this study. All work was done on
keywords, not lemmas.
4. Results
After a graph on occurrences of words per thousand in each text was plotted, a definite pattern was
seen to emerge: the second Ajar novel La Vie
devant soi deviated considerably from the other
Gary novels as well as all the other novels. In order
to determine the significance of this difference,
three statistical tests were done: the t-test, the
Pearson correlation and the chi-squared test. All
these tests showed that La Vie devant soi was
statistically different from the Gary novels as well
as all the other novels. After a confidence interval
of 99% for the t-test was constructed, it was observed that among the high frequency words La
Vie devant soi had 55.5 percent of occurrences that
fell outside the expected range, while the other
novels ranged between 11.6 and 28.3 percent of
the expected value. The Pearson correlation yielded similar results. While correlation between all
the books were high (as expected) they were highest among nearly all the other texts ranging between 0.94 and 0.96. However La Vie devant soi
showed a lower correlation with the other books,
ranging between 0.66 with Engwall (lowest correlation) and 0.82 with Gros-Câlin (highest correlation). It’s correlations with the two Gary novels
were 0.78 and 0.80, while the correlations between
the two Gary novels were 0.95.
The Chi-squared test also showed that among the
high frequency words La Vie devant soi had eleven
observations that fell above 3.84 (a significant
chi-squared value at one degree of freedom) while
the other novels had between one and six observations falling above 3.84. When the high frequency
list was condensed to a context-free list, La Vie
devant soi still gave the highest chi-squared value
of 43.36, while chi-squared values for the other
Gary novels ranged between 9.08 and 18.23. A
further chi-squared test on pairs of synonyms was
done. As expected it showed very high chi-squared values, but the highest value was again found
in La Vie devant soi at 6,198 followed by GrosCâlin at 1,746.
5. Conclusions
The statistical results found in this paper prove
beyond any doubt that high frequency words and
pairs of synonyms, which are considered, by
many, to be the unconscious elements of an author’s style, can indeed be consciously manipulated by the author. The notion that function words
(and synonyms) constitute a genetic fingerprint of
an author’s style has therefore been disputed in the
Romain Gary – Émile Ajar case. While Gros-Câlin, the first Ajar novel closely resembles the two
Gary novels, La Vie devant soi is so significantly
different from the two Gary novels that it could
have been written by another author. It would
appear that Gary did not feel the need to drastically
260
change his style in Gros-Câlin his first Ajar novel,
feeling confident that nobody would make the
connection between him and Ajar. But when critics saw similarities between Gros-Câlin and the
Gary novels, he became increasingly paranoid, set
out to prove that he was not Ajar and wrote La Vie
devant soi. In so doing, he consciously or unconsciously changed the genetic fingerprint of the Gary
style in that novel.
References
Brinegar, C.S. “Mark Twain and the Quintus Curtius Snodgrass Letters: A Statistical Test of
Authorship.” Journal of the American Statistical Association, 58 (1963), 85-96.
Burrows, J.F. Computation into Criticism: A Study of Jane Austen’s Novels and an experiment
in Method. Oxford: Clarendon Press, 1987.
Damerau, F.J. “The Use of Function Word Frequencies as Indicators of Style.” Computers
and the Humanities, 9 (1975): 271-280.
Ellegard, A. A Statistical Method for determining
Authorship: The Junius Letters, 1769-1772
.
Gothenburg: University of Gothenburg, 1962.
Fucks, W. “On the Mathematical Analysis of Style.” Biometrika, 39 (1952): 122-129.
Gary, R. Vie et mort d’Émile Ajar. Paris: Gallimard, 1981?
Kjetsså, G. “And Quiet Flows the Don Through
the Computer.” Association for Literary and
Linguistic Computing Bulletin, 7 (1979): 248-
256.
Mendenhall, T.C. The Characteristic Curves of
Composition.” Science, IX (1887): 237-249.
Mosteller, F. and D.L. Wallace. “Inference and
Disputed Authorship: The Federalist.” Reading, M.A: Addison-Wesley, 1964.
Pavlowich, P. L’Homme que l’on croyait. Paris:
Fayard 1981.
Williams, C.B. “A Note on the Statistical Analysis
of Sentence-Length as a Criterion of Literary
Style.” Biometrika, 31 (1940), 356-361.
Yule, G.U. “On Sentence-Length as a Statistical
Characteristic of Style in Prose, with Application to Two Cases of Disputed Authorship.”
Biometrika, 30 (1938): 363-390.