See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/270222238 Quantifying principles of the narrative text formation ARTICLE· DECEMBER 2014 Source: arXiv CITATION 1 READS 71 8 AUTHORS, INCLUDING: Stanislaw Drozdz Institute of Nuclear Physics and Cracow Univ … 134PUBLICATIONS1,832CITATIONSSEE PROFILE Andrzej Kulig Institute of Nuclear Physics 6PUBLICATIONS13CITATIONSSEE PROFILE Katarzyna Bazarnik Jagiellonian University 8PUBLICATIONS1CITATIONSEE PROFILE Jan Rybicki Jagiellonian University 10PUBLICATIONS31CITATIONSSEE PROFILE Available from: Stanislaw Drozdz Retrieved on: 25 January 2016
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Quantifying principles of the narrative text formation
Stanislaw Drozdza,b, Pawel Oswiecimkaa, Andrzej Kuliga,Jaroslaw Kwapiena, Katarzyna Bazarnikc, Iwona Grabska-Gradzinskaa,d,
Jan Rybickic, Marek Stanuszekb
a Complex Systems Theory Department, Institute of Nuclear Physics, Polish Academy of Sciences, ul. Radzikowskiego 152, 31-342 Krak´ ow, Poland
bFaculty of Physics, Mathematics and Computer Science, Cracow University of Technology, ul. Warszawska 24, 31-155 Krak´ ow, Poland
cInstitute of English Studies, Faculty of Philology, Jagiellonian University, ul. prof.S. Lojasiewicza 4, 30-348 Krak´ ow, Poland
d Faculty of Physics, Astronomy and Applied Computer Science, Jagiellonian University,
ul. Reymonta 4, 30-059 Krak´ ow, Poland
Abstract
In natural language using short sentences is considered efficient for commu-nication. However, a text composed exclusively of such sentences looks tech-nical and reads boring. The text composed of long ones, on the other hand,demands significantly more effort for comprehension. Studying characteris-tics of the sentence length variability (SLV) in a large corpus of world-famous
literary texts shows that an appealing and aesthetic optimum appears some-where in between and involves selfsimilar, cascade-like alternation of variouslengths sentences. A related quantitative observation is that the power spec-tra S (f ) of thus characterised SLV universally develop a convincing ‘1/f β’scaling with the average exponent β ≈ 1/2 , close to what has been identi-fied before in musical compositions or in the brain waves. An overwhelmingmajority of the studied texts simply obeys such fractal attributes but espe-cially spectacular in this respect are hypertext-like, ”stream of consciousness”novels. In addition, they appear to develop structures characteristic of irre-ducibly interwoven sets of fractals called multifractals. These observationsand results, beside their obvious interdisciplinary implications, open room
for novel informetrics measures of potentially great applicability.Keywords: Text formation, World literature, Long-range correlations,
Mirroring cultural progress (Agmon and Bloch, 2013) during their evo-lution natural languages - the most advanced and imaginative carriers of in-formation - developed remarkable quantifiable patterns of behaviour such ashierarchical structure in their syntactic organization (Nowak, Komarova andNiyogi, 2002) a corresponding lack of characteristic scale (Newman, 2005) asevidenced by the celebrated Zipf law (Zipf, 1949) and long-range correlationsin the use of words (Montemurro and Pury, 2002; Ausloos, 2012). A majorityof such patterns are common to a large class of natural systems known ascomplex systems (Kwapien and Drozdz, 2012) and they all open a formalframework for extending the informetrics measures (Egghe, 2005; Bar-Ilan,2008). Since the capacity of language is to generate an infinite range of ex-pressions from the finite set of elements (Hauser, Chomsky and Fitch, 2002)the complexity concept suggests to inspect the linguistic constructs longerthan mere words. The most natural of them are sentences - strings of wordsstructured according to syntax and grammar principles (Akmajian, Demers,Farmer and Harnish, 2001). Typically it is within a sentence that wordsacquire a specific meaning. Furthermore, in a text the sentence structureis expected to be correlated with the surrounding sentences as dictated by
the intended information to be encoded, fluency, harmony, intonation andpossibly due to many other factors and feedbacks including the authors’preferences. This may thus introduce even more involved and more essen-tial correlations than those identified so far. The composition of sentencesof varied length dictates the reading rhythm which thus involves sound andperception. This, therefore, opens up a possibility that the celebrated Weber-Fechner law (Coren, Ward and Enns, 2004) - stating that in perception it isthe relative proportions that matter primarily, and not differences in absolutemagnitudes - leaves its imprints also in the sentence arrangement. At thesame time a sentence cannot usually be expanded continuously but by addingclauses, so that syntax and grammar rules are obeyed. The first of these two
potential factors makes some variant of the multiplicative cascade a likelycomponent of the mechanism that amplifies the associative leaps especiallyanticipated in the ”stream of consciousness” (SoC) narrative, while the otherfactor may set some constraints. The multifractal formalism (Halsey et al.,
1985; Stanley and Meakin, 1988) offers a particularly appropriate framework
to quantify such effects.
2. Material and Methods
In order to study the long-range correlations among sentences, particu-larly those that refer to fractals and cascading effects, we select a rich corpusof 113 world-famous English, French, German, Italian, Polish, Russian, andSpanish literary texts of considerable size and for each individually form aseries l( j) from the lengths of the consecutive sentences j expressed in termsof the number of words. Thus, a sentence is defined in purely orthographicterms, as a sequence of words starting with a capital letter and ending in a
full stop. Since the present study has a statistical character, an additionalcriterion we impose specifies that each text contains no fewer than 5000 sen-tences. A complete list of the titles included in this corpus is given in theAppendix.
The simplest, second-order linear characteristics are measured in terms of the power spectra S (f ) of such series. Such spectra are calculated as FourierTransform modulus squared
S (f ) = | jmax j=1
l( j)e−2πifj|2 (1)
of the series l( j) representing lengths of the consecutive sentences j.A complementary approach towards higher order correlations consists in
the wavelet decomposition of l( j). The corresponding ‘mathematical micro-scope’ wavelet coefficient maps T ψ(s, k) are obtained as
T ψ(s, k) = (1/√
s)
jmax j=1
l( j)ψ(( j − k)/s) (2)
where k represents the wavelet position in a text while s the wavelet resolutionscale. The wavelet ψ used in the present study is a Gaussian third derivative.
The wavelet decomposition, is optimal for visualisation and, in princi-ple, it is well suited to extract the multifractal characteristics (Muzy, Bacryand Arneodo, 1994). However, the newer method, termed Multifractal De-trended Fluctuation Analysis (MFDFA) (Kantelhardt et al. 2002) is numer-ically more stable (Oswiecimka, Kwapien and Drozdz, 2006), though even
here the convergence to a correct result is a subtle issue (Drozdz, Kwapien,
Oswiecimka and Rak, 2009). Accordingly, for a series l( j) of sentence lengthsone evaluates its signal profile L( j) ≡ jk=1 [l(k)− < l >], where < · > de-
notes the series average and j = 1,...,jmax with jmax standing for the numberof sentences in a series. This profile is then divided into 2M s disjoint seg-ments ν of length s starting from both end points of the series. Next, thedetrended variance
F 2(ν, s) = 1
s
sk=1
{L((ν − 1)s + k) − P (m)ν (k)} (3)
is determined, where a polynomial P (m)ν of order m serves detrending. Finally,
a q -th order fluctuation function
F q(s) = 1
2M s
2M sν =1
[F 2(ν, s)]q/21/q
, (4)
is calculated and its scale s dependence inspected. Scale invariance in a formF q(s) ∼ sh(q) indicates the most general multifractal structure if the gener-alised Hurst exponent h(q ) is explicitly q -dependent, while it is reduced tomonofractal when h(q ) becomes q -independent. h(q ) determines the Holderexponents α = h(q ) + qh′(q ) and the singularity spectrum
f(α) = q [α − h(q )] + 1 (5)
the latter being the fractal dimension of the set of points with this particularα . For a model multifractal series (like a binomial cascade), f(α) typicallyassumes a shape resembling an inverted parabola whose widths ∆α = αmax−αmin is considered a measure of the degree of multifractality and thus oftenalso of complexity.
3. Results and discussion
A highly significant result is obtained already by evaluating the power
spectra S (f ) of the series representing the sentence length variability (SLV)of all the text considered. As documented in Fig.1, the overall trend of allsample texts, and especially its average, shows a clear 1/f β scaling withβ ≈ 1/2 over the entire range of more than two orders of magnitude in fre-quencies f spanned. For the individual texts β is seen to range between
Figure 1: Power spectra S (f ) of the sentence length variability for 113 worldfamous literary works. They are calculated as Fourier Transform modulus squared(Eq. (1)) of the series l( j) representing lengths of the consecutive sentences j expressedin terms of the number of words. S (f ) is seen to display 1/f β scaling. Middle solidline (green) denotes average over the individual power spectra, properly normalised, of all the corpus elements and it fits well by β =1/2. Boundaries of the dispersion in β areindicated by taking average over 10 corpus elements, with the largest β -values, whichresults in β =3/4 and over 10 its elements with the smallest β -values, which results inβ =1/4. The two extremes in the corpus, explicitly indicated, are Henry James’s The Ambassadors (upper) and Artamene ou le Grand Cyrus , the 17th century novel sequence(lower), considered the longest novel ever published which resembles S (f ) for white noise.
1/4 and 3/4. This kind of scaling points to the existence of the power-law
long-range temporal correlations in SLV - thus to its fractal organization- and indicates that it balances randomness and orderliness into a uniqueattractive whole, just as it does for music, speech (Voss and Clark, 1975),heart rate (Kobayashi and Musha, 1982), cognition (Gilden, Thornton andMallon, 1995), spontaneous brain activity (Kwapien, Drozdz, Liu and Ioan-nides, 1998), and for other ‘sounds of Nature’ (Bak, 1996; Theunissen andElie, 2014). From this perspective human writing appears to correlate withthem. Even the range of the corresponding β -values from about 1/4 to 3/4overlaps significantly (more on the Mozart than Beethoven’s side) with those(1/2 to 1) found (Levitin, Chordia and Menon, 2012) for musical composi-tions, which may provide a quantitative argument for our tendency to refer
to writing as ‘being composed’ when we care about all its aspects includingaesthetics and rhythm to be experienced in reading.
The two extremes in the corpus, explicitly indicated in Fig. 1, are HenryJames’s The Ambassadors (upper) and Artamene ou le Grand Cyrus , the17th century French novel sequence (lower), considered the longest novel everpublished. The latter appears largely consistent with the white noise whoseS (f ) is flat. It is also appropriate to notice that at the largest frequencies,which corresponds to the smallest scales, the power spectra of all the textshave some tendency to flatten. This may suggest that the long-range coher-ence in their 1/f organization is locally somewhat coarsened by grammatical
constraints.Our central result relates, however, to the nonlinear characteristics thatmay manifest themselves in heterogeneous, self-similarly convoluted struc-tures, undetectable by S (f ). Such structures may demand using the wholespectrum of the scaling exponents and are then termed multifractals (Stanleyand Meakin, 1988). That such structures in SLV may be present within thecorpus analysed here can be inferred from Fig. 2, which shows four, some-what distinct, categories of behaviour. A majority of the texts in our studyresembles the case displayed in (I). SLV is here seen to be rather homoge-nously ‘erratic’ and, consequently, the distribution of cascades seen throughthe wavelet decomposition is largely uniform. The three other cases, (II),
(III) and (IV), commonly considered representatives of the SoC literarystyle, are visibly inhomogeneous in this respect, as SLV displays clusters of intermittent bursts of much longer sentences. Such structures are character-istic of multifractals and thus an appropriate subject of the analysis withinthe above formalism.
Figure 2: (Previous page) Four examples illustrating different fractal/multifractal char-
acteristics identified within the corpus of the canonical literary texts: I, War and Peace by Lev Tolstoy; II, Rayuela (Hopscotch) by Julio Cortazar; III, The Waves by VirginiaWoolf and IV, Finnegans Wake (FW) by James Joyce. The panels inside each containcorrespondingly: (a) The series l( j) of the consecutive sentence lengths throughout thewhole text. Insets illustrate the corresponding probability distributions P (l) of l( j); (b)Wavelet coefficient maps (T ψ(s, k)) obtained for l( j). The wavelet ψ used is a Gaussianthird derivative. Horizontal axis represents the sentence position in a text while verticalaxis - the wavelet resolution scale s. Colour codes denote magnitude of the coefficient fromthe smallest (dark blue) to the largest (red); (c) q -th order fluctuation functions calculated
according to Eq. (2) using the detrending polynomial P (m)ν of second order (m=2) and for
q ∈ [−4, 4]; (d) The resulting singularity spectra f(α) for (i) the series l( j) representingoriginal texts (black), (ii) for their Fourier-phase randomised counterparts (blue); here f(α)is seen shrunk essentially to a point as is characteristic of a pure monofractal, and (iii) fortheir randomly shuffled counterparts (gray). V. Chronological progress of James Joyce’s“engineering work” on writing FW, which he described as “boring a mountain from twosides” (Ellmann, 1982) . This chart may be also taken as a visualisation of Joyce’s dreamabout a Turk picking threads from heaps on his left and right sides, and weaving a fabricin the colours of the rainbow, which the writer interpreted as a symbolic picture of BooksI and III of FW .
The fluctuation functions F q(s) obtained according to Eq.(3) display (Fig.2) a convincing scaling with different degree of q -dependence, however. Thisis corroborated by the corresponding singularity spectra f(α), which rangefrom very narrow in War and Peace (I), indicating essentially monofrac-
tal structure, through significantly broader - thus already multifractal - butasymmetric like the strongly left sided Rayuela (II) or right sided The Waves (III) up to an exceptionally broad and simultaneously almost symmetric case(IV) of Finnegans Wake (FW).
The left side of f(α) is determined by the positive q -values, which filter outlarger events (here longer sentences), and its right side reflects behaviour forsmaller events as filtered out by the negative q -values. Hence, asymmetry inf(α) signals non-uniformity of the underlying hypothesized cascade. Rayuela is thus seen to be more multifractal in the composition of long sentences andalmost monofractal on the level of small ones. To some extent the oppositeapplies to The Waves . In fact, these effects can be inferred already from thenon-uniformities of the corresponding SLV wavelet decompositions (Fig. 2).In this respect FW appears impressively consistent; being one of the mostintriguing literary ‘compositions’ ever, mastered imaginatively in the SoCtechnique, freely exploring the mental labyrinth of dreams and thus often
breaking conventional rules of syntax and of linguistic rigour. However, from
the perspective of our formal quantitative approach, its architecture looks - orperhaps just is - a result of these factors - to be governed consistently by thesame ‘generators’ on all scales of sentence length. An extra intellectual factorshaping FW is very likely to be also related to its top-bottom development- much like model mathematical cascades - as evidenced by its chronologyof writing graphically sketched in the lowest panel V of Fig. 2. Bearing inmind that Joyce himself considered FW an “engineering work” and expressedthe wish (Ellmann, 1982) that this work should be studied by exact sciencemethods, the present results provide further arguments for considering himan ingenious and visionary linguistic “engineer” (Bazarnik, 2011), possiblyopening some new horizons for language, enabling it to explore better the
brain capacity, and echoing the sounds of nature more profoundly. Sufficeus to say in this context that it was FW that inspired Murray Gell-Mann topropose spelling for quarks - the most fundamental constituents of matter.
The significance of the above results for the singularity spectra f(α) of theseries l( j) representing the original texts has also been tested against the twocorresponding surrogates. One standard surrogate in this kind of analysis isobtained by generating the Fourier-phase randomised counterparts of l( j).This destroys nonlinear correlations and makes probability distribution of fluctuations Gaussian-like, but preserves the linear correlations and, as it isclearly seen in Fig. 2, shrinks f(α) essentially to a point as is characteristic of
a pure monofractal. Another surrogate is obtained by randomly shuffling theoriginal series l( j). Consequently, any temporal correlations get destroyedbut the probability distributions of fluctuations remain unchanged. The cor-responding singularity spectra calculated according to the same MFDFA al-gorithm are also shown in Fig. 2 (gray). Consistently with the lack of anytemporal correlations they all get shifted down to α ≈ 0.5 but some nonzerowidth of f(α) still remains to be observed. However, at least a large partof this remaining multifractality in this last case may be apparent due toa relatively small size of the samples. As shown by Drozdz et al. (2010),for the uncorrelated series the result of calculating the multifractal spectraends up in either mono-fractal for the series whose fluctuation probability
distributions are Levy-unstable, or in bi-fractal for those whose distributionsare Levy-stable. Contrary to the correlated series, the convergence to theultimate correct results in this case is very slow. We also wish to note at thispoint that in spite of the Menzerath-Altmann law, all the relevant resultsshown here remain essentially unchanged if the sentence length is measured
Figure 3: Special case of Ulysses. The same convention as in Figure 2 is used here. Thetwo additional insets in the panels a and c display results for Ulysses after bisecting it into
halves. Ulysses-I corresponds to the text from the beginning to the end of Chapter 10 andUlysses-II to the remaining text (without its last two disproportionately long sentences).
in terms of the number of characters instead of the number of words.Another, even better known SoC novel by Joyce - Ulysses , which played
a central role in formulating the scale - free word rank-frequency distribu-tion law by Zipf - also deserves here an extended attention, however, for adifferent reason. As illustrated in Fig. 3, no unique multifractal scaling canbe attributed, and thus no f(α) assigned, for this novel. The SLV inspectedboth in terms of the sentence length distribution and through its wavelettransform indicate clearly that Ulysses splits into two parts such that eachof them may independently have well defined scaling properties. Indeed,by bisecting it approximately into halves (between Chapters 10 and 11) al-lows us again to comprise Ulysses within the present formalism. The firstpart appears essentially monofractal, while the other is clearly multifractal,
though asymmetrically left-sided, just as Rayuela . In fact, this result pro-
vides a quantitative argument in favour of some earlier literary criticism onthe ”doubleness” of Ulysses (McHale, 1992).The results, represented in terms of the width ∆α of f(α) and of α ≡
α(q = 0) , which stands for the most frequent H older exponent (maximumof f(α)) and thus can be considered a measure of the degree of persistence inSLV , for the whole studied corpus are collected in Fig. 4.
This ‘scatter plot’ opens up room for many further interesting informetricsrelated observations and hypotheses or even definite conclusions of generalinterest. Some of them that can be straightforwardly listed as follows:
(i) All the studied texts that are seen in the multifractality region are com-
monly classified as SoC literature. The only exception found here, the Old Testament , has not been considered before in this context.
(ii) ∆α for all the texts that do not belong to SoC is located below theborder of definite multifractality. Their complexity is thus poorer.
(iii) Also, several texts, by some considered as SoC, appear to be located sig-nificantly below this border. An important example of this is A la recherche du temps perdu by Marcel Proust (no. 76 in the list given in Appendix)),which is clearly monofractal. The present methodology may thus also serve
as a quantitative criterion in resolving the related disputes.
(iv) Artamene ou le Grand Cyrus is seen to have characteristics just op-posite to FW . Here ∆α equals nearly zero and α gets shifted down to almost1/2, which complements its flat power spectrum seen in Fig. 1, to mean thatthe corresponding SLV is of the white noise type.
4. Summary
The present analysis, based on a large corpus of world famous literarytexts, uncovers the long-range correlations in their sentence arrangement.
The linear component of these correlations universally reveals the scale-free‘1/f β’ form as characteristic to many other ‘sounds of Nature’ and thus thisobservation may serve as an indicator of those factors that shape humanlanguage. The corresponding β -value ranges from about 1/4 to 3/4 andmay thus serve also as a very useful and inspiring bibliographic measure.
W. Shakespeare ( d e g r e e o f c o m p l e x i t y )
(the most frequent Hölder exponent)
~
~
À la recherche du temps perdu
Figure 4: Complexity characteristics of the world literature. ‘Scatter plot’, whichfor a collection of 113 most representative literary works indicated by their numbers onour list (Appendix) displays the width ∆α (schematically defined in the inset) and themost frequent Holder exponent α. Shaded area marks the transition (uncertainty) regionbetween fully developed multifractality and definite monofractality. We find it reasonableto assume that the shuffled series are mono-fractal (or at most bi-fractal) and that anytrace of multifractality in this case is an artefact of the finiteness of a series. Therefore,the lower bound of the shaded area is determined as an average of ∆α’s for all the series(texts) shuffled. Due to the thickest tails in the probability distributions P (l) of l( j) inFW (seen in the inset to panel IV of Figure 2), which after shuffling the correspondingseries may yield the strongest apparent multifractality signal, the upper bound of theuncertainty region is taken as ∆α of the shuffled FW .
As far as correlations in the sentence length variability are concerned, some
texts - within the present corpus exclusively belonging to the stream of con-sciousness narrative - develop even more complex scale-free patterns of thenonlinear character. In quantitative terms this results in a whole spectrumof the scaling exponents as compactly grasped by the multifractal spectrumf(α) , whose width reflects the degree of nonlinearity involved. A greatercomplexity of such hypertext-like narrative finds an intriguing parallel inthe biological dynamical system as documented (Ivanov et al., 1999) for thehealthy human heartbeat, which develops broader multifractal spectra ascompared to the heart failure. That the SoC kind of narrative simultane-ously activates greater variety of brain areas seems quite natural. Whetherthis indicates route towards more efficient and thus ‘healthier’ communica-
tion also emerges as an exciting perspective to study. A further argument infavour of such a likely correspondence is that hypertext parallels the under-lying architecture of World Wide Web which proves easy-to-use and flexiblein sharing information over the Internet, indeed.
Acknowledgement: We thank Krzysztof Bartnicki (who translated FW into Polish) for constructive exchanges at the early stage of this Project.
References
Agmon, N., & Bloch, Y. (2013) Statistics of language morphology change:from biconsonantal hunters to triconsonantal farmers. PLoS ONE 8(12),e83780.
Akmajian, A., Demers, R.A., Farmer, A.K. & Harnish, R.M. (2001) Linguis-tics: An Introduction to Language and Communication. MIT Press (Cam-bridge)
Ausloos, M. (2012). Generalized Hurst exponent and multifractal function of original and translated texts mapped into frequency and length time series.Physical Review E 86, 031108
Bar-Ilan, J. (2008) Informetrics at the beginning of the 21st century-A re-view, Journal of Informetrics 2, 1-52.
Bazarnik, K. (2011) Joyce and Liberature, Litteraria Pragensia (Prague)
Coren, S., Ward, L.M. & Enns, J.T. (2004) Sensation and Perception, John
Wiley & Sons, sixth ed. (Hoboken NJ)
Drozdz, S., Kwapien, J., Oswiecimka, P., & Rak, R. (2009) Quantitativefeatures of multifractal subtleties in time series. Europhys. Lett. (EPL) 88,60003.
Egghe, L. (2005) Expansion of the field of informetrics: Origins and con-sequences. Information Processing and Managament 41, 1311-1316.
Ellmann, R. (1982) James Joyce, Oxford University Press (Oxford, 1982)
Gilden, D. L., Thornton. T., & Mallon M. W. (1995) 1/f noise in humancognition, Science 267, 1837-1839 (1995)
Halsey, T.C., Jensen, M.H., Kadanoff, L.P., Procaccia, I., & Shraiman,B.I. (1986) Fractal measures and their singularities: The characterizationof strange sets, Physical Review A 33, 1141-1151
Hauser, M.D., Chomsky N. & Fitch W.T. (2002) The Faculty of Language:What is it, Who has it, and Did it evolve? Science 398, 1569-1579.
Levitin, D.J., Chordia, P., & Menon, V. (2012) Musical rhythm spectra fromBach to Joplin obey a 1/f power law. Proc. Natl. Acad. Sci. 109(10),3716-3720.
McHale, B. (1992) Constructing Postmodernism, Routledge (London)
Newman, M.E.J. (2005) Power laws, Pareto distributions and Zipf’s law,Contemporary Physics 46, 323-351.
Nowak, M.A., Komarova, N.L. & Niyogi, P. (2002) Computational and evo-lutionary aspects of language. Nature 417, 611-617.
Oswiecimka, P., Kwapien, J., & Drozdz, S. (2006) Wavelet versus detrendedfluctuation analysis of multifractal structures. Physical Review E 74, 016103.
Stanley, H. E., & Meakin, P. (1988). Multifractal phenomena in physicsand chemistry. Nature 335, 405-409.
Theunissen, F. E., & Elie, J. E. (2014). Neural processing of natural sounds,Nature Rev. Neuroscience 15, 355-366.
Voss, R.F., & Clark, J. (1975). 1/f noise in music and speech. Nature 258,317-318.
Zipf, G.K. (1949) Human behavior and the principle of least effort. Addison-Wesley (Cambridge)
58 Artamène ou le Grand Cyrus Madeleine & Georges de Scudéry(Artamène, or
Cyrus the Great,1649-53)
0,5387 0,0281
59 Les Liaisons dangereuses Pierre Choderlos de Laclos(The DangerousLiaisons, 1782)
0,6876 0,0873
60 Le Rouge et le Noir Stendhal (Henri Beyle)(The Red and the
Black, 1830)0,6086 0,0327
61 La Comédie humaine Honoré de Balzac(The Human
Comedy, 1830-39)0,7675 0,1075
62 Les Mystères de Paris Eugène Sue(The Mysteries of
Paris, 1842-43)0,8229 0,1825
63 Les Trois Mousquetaires Alexandre Dumas(The Three
Musketeers, 1844)0,818 0,2245
64 La Reine Margot Alexandre Dumas(Queen Margot,
1844-45)0,7659 0,07631
65 Le Comte de Monte-Cristo Alexandre Dumas(The Count ofMonte Cristo,
1844-45)0,7519 0,2156
66 Vingt ans après Alexandre Dumas(Twenty Years
After, 1845)0,6656 0,1678
67 Le Vi comt e de B ragelonne ou Di x ans pl us tard Alexandre Dumas
(The Vicomte ofBragelonne: Ten
Years Later, 1847-50)
0,8056 0,2262
68 Le Collier de la reine Alexandre Dumas(The Queen's
Necklace, 1849-50)0,7909 0,1723
69 Madame Bovary Gustave Flaubert (1857) 0,7154 0,07570 Les Misérables Victor Hugo (1862) 0,82 0,1896
71 Le Petit Chose Alphonse Daudet(Little Good-For-
Nothing1868)0,6716 0,0966
72 Les Rougon-Macquart Émile Zola (1871-93) 0,7093 0,1893
73 Bel Ami Guy de Maupassant (1885) 0,7436 0,191374 A vent ures ext raordi naires d'un s avant rus se Georges Le Faure & Henri de Graffigny (1888) 0,7545 0,0896
75 Le Roman de Tristan et Iseut Joseph Bédier(Romance of
Tristan and Iseult,1900)
0,7678 0,1036
76 À la recherche du temps perdu Marcel Proust(In Search of Lost
Time, 1913-27)0,7545 0,1033
77 Voyage au bout de la nuit Louis-Ferdinand Céline(Journey to the
End of the Night,1932)
0,7776 0,0644
78 Mort à crédit Louis-Ferdinand Céline(Death on Credit,
1936)0,7317 0,4069
79 La Condition Humaine André Malraux (M an's Fat e, 1933) 0, 7133 0, 2569
80 Molloy, Malone Meurt, L'innommable Samuel Beckett