EMPIRICAL STUDIES OF THE ARTS, Vol. 10(2) 157-181, 1992

ONTHE METER AND PROSODY OF FRENCII12-SYLLABLE VERSE

MARC DOMINICY Linguist/que generale Un/verslte Libre de Bruxelles

ABSTRACT This article provides the reader with a survey of the basic metrical and prosodic properties of the French classical and post-classical 12-syllable alexandrine line. It is argued that these features can be captured within a neo-Jakobsonian framework which maintains a clear distinction between linguistic and non-linguistic parallelisms. Metrical equivalence is accounted for by positing abstract "line designs" that are linked to "line instances" by a set of "matching rules." The versification of Racine, Hugo, and late Nineteenth-century poets of the Parnassian and post-Parnassian generations is studied in some details. This makes it possible to evaluate the adequacy of recent theorizing by Benoit de Cornulier and Paul Verluyten.

THE ESSENTIALS OF FRENCH VERSIFICATION: MASCULINE AND FEMININE LINES One of the major insights of Roman Jakobson's poetics [1-2, 4-10] is, no doubt, the assumption that any well-founded theory of verse should be based on a general study of parallelism. By reformulating this basic principle, I show that it is possible to give a description of the fundamentals of French verse. In order to make this discussion less abstract, I will use as an example one of the French poems which has frequently been commented upon by poeticians (see [3]), Baudelaire's "Les Chats": Les amoureux fervents et les savants austeres Aiment egalement, dans leur mare saison, Les chats puissants et doux, orgueil de la maison, Qui comme eux sont frileux et comme eux sedentaires,

157 CJ 1992, Baywood Publishing Co., Inc.

doi: 10.2190/570B-ARK0-6MM7-JP7H http://baywood.com 158 I DOMINICY

Amis de la science et de la volupte, Ils cherchent le silence et l'horreur des tenebres; L'~be les eOtpris pour ses coursiers funebres, S'ils pouvaient au servage incliner leur fierte.

lis prennent en songeant les nobles attitudes Des grands sphinx allonges au fond des solitudes, Qui semblent s'endormir dans un reve sans fin;

Leurs reins fcconds sont pieins d'etincelles magiques, Et des parcelles d'or, ainsi qu'un sable fin, Etollentvaguement leurs prunelles mystiques.

The Schwa It is a well-known fact that, in French, polysyllabic words are stressed on their last syllable, unless it contains the vowel which is usually called "e muet," mute "e" or schwa. Consequently, the posttonic part of a polysyllabic word consists of at most one syllable. Let us assume, for the sake of the argument, that any polysyllabic word can be divided into two and only two parts: the posttonic or "feminine" part, and the remaining "masculine" part. Let us assume, furthermore, that, except for two postverbal enclitics ("-je" and "-ce"), all monosyllabic words are "masculine" (i.e., they only contain a "masculine" part). Given such premises, borrowed in a simplified version from [11-15], we are able to describe the crucial features of the so-called "classical alexandrine" used in "Les Chats."

The Hiatus Suppose we define the notion of "hiatus" in the following way. There exists a hiatus between two vowels VI and V2 occurring in that order if, and only if: (i) VI and V2 are separated by a word boundary-or a stronger (e.g., phrase or sentence) boundary; (ii) VI is not a schwa; (iii) the spelling of the word to which VI belongs does not end with a consonantal letter; (iv) the spelling of the word to which V2 belongs does not begin with the consonantal letter "h." (This definition is not totally adequate [11, 14, 16], but it will be sufficient for my purpose.) So, there is no hiatus in "beant" (condition (i}), "vraie et" (condition (ii)), "maison ou" (condition (iii)), "it hair" (condition (iv)), but there is a hiatus in "tu es." If we can tum back to "Les Chats," we observe that there are two cases of hiatus ("volupte, lis" and "fierte, Us"), and that in both cases the word boundary coincides with a line boundary. In fact, any French poem written according to the conventions of classical versification conforms to a very simple constraint which FRENCH 12·SYLlABLE VERSE I 159 prohibits line-internal hiatus, so that the occurrence of a hiatus becomes one of the marks of line division. Interestingly enough, this property already shows that the commutation of linguistically equivalent segments may lead to metrical ill­ formedness. Indeed, suppose we invert the syntagmatic order of the two preposi­ tional phrases of line 5: Amisde 18 volupteet de 18 science

The resulting formula is linguistically perfect, but it happens to violate the con­ straint which prohibits line-internal hiatus.

Syllable Countand Elision Another feature of French meter can be illustrated by a short comparison between lines 3, 4 and 5 of our poem. Assume we are able to count the syllables of any given (string ot) word(s). Assume, furthermore, that, whatever our own pronunciation, we are familiar with some additional conventions stipulating, for instance, that the word "science" of line 5 contains 3 syllables (sci-en-ce). Then, syllable count applied to lines 3, 4 and 5 will respectively yield 12, 15 and 13 syllables: 3 Les/chats!puis/sants!et/doux/or/gueiVde/la/mai/son 4 Qui/com!me/eux/sont/fri/leux/et/com!me/eux/s6/den!tai/res 5 Nmis/de/la/sci/en!ce/et/de/la/vo/lup/t6 Now, if we look at line 3, we observe that it does not contain any schwa separated from a subsequent vowel by a word boundary only (in "de la," the consonant "I" occurs between "e muet" and "a"). As for lines 4 and 5, they respectively contain 2 and 1 tokens of such a sequence: commeeux; commeeux; scienceet, Let us say that in "comme eux" and similar cases, we have an "e-V sequence." The reader can easily check that the following regularity holds for "Les Chats" (and for any poem written within the same metrical tradition): if a line x contains m syllables, with m-12 ~ 1, then:

(i) if the last syllable of x is the feminine part of a word, x contains m-13 e-V sequences; (ii) if the last syllable of x belongs to the masculine part of a word, x contains m-12 e-V sequences.

For instance, line 4 contains 15 syllables: since 15-12 = 3 and the syllable "-res" is the feminine part of "sedentaires," it follows that there are 15-13 = 2 e-V sequences; line 5 contains 13 syllables: since 13-12 = 1 and the syllable "-t6" belongs to the masculine part of "volupte," it follows that there is 13-12 =1 e-V sequence. An alternate way to capture this regularity consists in formulating a rule of "elision" which deletes any word-final "e muet" in the domain of an e-V sequence. We then get the following syllabic divisions for lines 4 and 5: 160 I DOMINICY

Qui/comlmeux/sont/fri/1eux/et/comlmeux/~/denjtai/res Nmis/de/l8/sci/enjc et/de/la/vo/lup/tC

More generally, syllable count will yield 12 for all "masculine" lines (whose last syllable belongs to the masculine part of a word), and 13 for all "feminine" lines (whose last syllable is the feminine part of a word). Again, this highly restrictive property blocks themutual commutation ofsegments which arelinguis­ tically equivalent. For instance, the interversion of thetwoadjectives of line 4:

Quicomme eux sont ~dentaires et comme eux frileux wouldcreatea masculine lineof 13syllables, wheretheword-final consonant "s" prevents "s~dentaires et" from beingan e-V sequence. The prohibition of hiatus and the principles which underlie elision are not totally unrelated, since classical French versification rules out any (word-final) stringV-e unless it is situated at the endof theline or belongs to an e-V sequence [11, p. 105; 14]. So, in thefollowing twoseriesof examples:

DanscettegrandeplaineOU l'autan froid se joue (a) DanscettegrandeplaineOU se joue l'autan (b)

Si, parunenuitbleueet froide de decembre (a) Si, parunenuit froide et bleuede decembre (b)

the (a)-line, quoted from Baudelaire's works, is well-formed, while the (b)-one is defmitely out. The deletion of "e muet" in "bleue et" does not create ill­ formedness because condition (ii)in thedefinition of hiatusappliesbeforeelision. Oncewe are able to assign the syllabic number 12or 13 to a line,we can label eachof itssyllables by using the integers 1, 2, ... , 12(13) in theirusualincreas­ ingorder; so wecan speakof,say,syllable 4 of line9. In theclassical alexandrine, syllables 6, 7 and12haveto meeta verysimple constraint: theymustbelongto the masculine part of a word. For instance, the following three lines are metrically ill-formed, even if they conform to the requisites concerning hiatus, elision, V-e sequences, andsyllabic number:

Lessavants aus~res, les amoureux fervents Lesamoureux austeres, les savantsfervents Lesamoureux fervents, les savants austeres

Moreover, syllables 6 and 7 cannot belong to the same word; this rules out a formula like:

Lesausteres amoureux, les savants fervents

in spiteof theperfectlinguistic coupling of its two nounphrases. FRENCH 12-SYLLA.BLE VERSE I 161

LINE DESIGN AND LINE INSTANCES In order to understand such a complex set of constraints, we have to tum back to the neo-Jakobsonian [4-8] insight that verse is a matter of non-linguistic parallelism. More precisely, we have to describe the autonomous grid which creates metrical equivalences between certain segments of an independently organized linguistic material. In the case at hand, we will posit what Jakobson [1, p. 34] would have called "the classical alexandrine line design," i.e., a pattern or model of the following kind: XXXXXXXXXXXX(X) In this abstract representation, each syllabic position (labelled "X") is either obligatory or optional: in fact, there is only one optional position (flanked by round parentheses) which corresponds to syllable 13 of feminine lines. According to a plausible universal of verse theory, an optional position which is not followed by any obligatory position is "extrametrical," i.e., does not play any role in the identification of the syllabic number assigned to the line design (which will then be 12 for all alexandrines). Moreover, we can single out those line designs where all optional positions are extrametrical by saying that they are "syllabic" [1, p. 29] or, more explicitly, by saying that the number of their "metrical" positions does not vary. So, the "syllabic" design of the French classical alexandrine contains 12 metrical positions plus one extrametrical position. Our line design can be conceived of as a sort of ruler or filter which selects those segments of French discourse that will qualify as "line instances" [1, p. 34] of a certain class. More specifically, the metrical equivalence holding between any two of the 14 alexandrines in "Les Chats" stems from the fact that all these lines "instantiate" the same abstract line design. In order to be able to establish, in every possible case, whether a given segment of French discourse is, or is not, linked to a given line design by this relation of "instantiation," we have to manipulate a set of "matching rules." In the case of the classical alexandrine, we first have to verify that our putative line instance conforms to the constraints regarding hiatus, elision, and V-e sequences; then we will label its 12 or 13 syllables and check whether syllables 6, 7 and 12-or 6, 7, 12 and 13-really meet the requirements presented above. What I have done so far amounts to summarizing the basic recipes whose knowledge is imperative in order to "scan" a classical alexandrine. The theoretical interest of such rules lies in the fact that they illustrate, or even follow from more general principles. Cornulier [11, 12, 13, 14] has shown that, in French (and probably in many other languages), any line design containing more than 8 metrical positions is "complex," i.e., divides into two or more parts. For instance, the classical alexandrine line design has two constituents of 6 metrical positions: XXXXXX-XXXXXX(X) 162 I DOMINICY

In this slightly modified pattern, the hyphen represents what I will call a "(metri­ cal) break" [1, p. 35], each constituent being a "subline" of the whole design. Classical French metries does not possess any line design with an extrametrical position at the end of a non-final subline. This is the reason why a line such as: Les amoureux austeres mais les savants fervents

(with an extrametrical syllable "-res") would be ruled out as an instantiation of our line design. However, the same example would have been a well-formed "pre­ classical alexandrine" [16, pp. 63-66], since it conforms to the following design: XXXXXX(X)-XXXXXX(X)

Similarly, the design of the Spanish "alejandrino" has two extrametrical posi­ tions at the end ofeach subline [17]:

XXXXXX(XXX)-XXXXXX(XXX) Example from Antonio Machado:

i,Soyclasico 0 roman(ti)(co)?-No set Dejar quisie(ra) What this means is that the complex line design of the classical alexandrine results from the integration of two sublines which progressively lost part of their mutual autonomy. Such a process was surely favored by the fact that certain formulas like, e.g.:

Un amoureux austere et un savant fervent can instantiate either the pre-classical design (with the extrametrical syllable "-re") or the classical design (with the elision rule applying to the e-V sequence "austere et"). On a more technical level, I would be inclined to put forth the following two principles as tentative universals of verse theory:

(i) any optional position that is not followed, in its subline, by an obligatory position is extrametrical; (ii) if a non-final subline has m extrametrical positions, then any other subline (includingthe final subline) has m extrametrical positions,but not vice-versa. From (ii) we deduce that no language can develop a line design which allows non-final extrametricity while prohibiting final extrametricity:

·XXXXXX(X)·XXXXXX

Notice that this constraint holds for line designs, but not for a particular line instance; there is nothing wrong in a pre-classical alexandrine like (repeated):

Les amoureux austeres mais les savants fervents

where the only extrametrical syllable does not match with a line-final position. FRENCH 12·SYLLABLE VERSE I 163

BREAKS AND CAESURAS The hyphen occurring in the schematic representation of a complex design does not, in itself, give any information about the interface between the metrical break and the linguistic material of the line instance. In classical French metrics, the syllables which are separated by the break cannot belong to the same word. This creates a strong parallelism between line-internal breaks and line boundaries. For example, the fact that in a classical alexandrine, the last syllable (6) of the first subline and the initial syllable (7) of the second one never belong to the same word clearly corresponds to the fact that, except for some metrical tricks [11, pp. 120-122], no line begins with a syllable belonging to the last word of the preceding line. Recall, however, that the parallelism between breaks and line boundaries is not complete, since line-internal extrametricity is no longer allowed. Notice also that we do not find any comparable constraint in the English or Italian lO-syllable line, where the last syllable (4 or 6) of the first subline and the initial syllable (5 or 7) of the second subline may belong to the same word, provided it has a paroxytonic stress pattern [18]. Examples: Th' expence of Spi-rit in a waste ofshame (Shakespeare) al gran tiran-no del cui toseo sugge (Dante) In view of such typological differences, it is convenient to distinguish, when we deal with line instances, between breaks (which separate syllables) and caesuras (which separate words). More technically, given a break B, between sublines x and y, the "caesura" Ci associated with Br will be the first word boundary which follows x (even if it occurs in y). In the classical alexandrine, the interval bounded by Br and Ci can only be filled with consonants or elided schwas. Here are some examples, where the hyphen indicates the break, and the slash indicates the corresponding caesura: Aiment egalement-« / dans leur mOresaison Amis de la scien--<:(e) / et de la volupte Et des parcelles d'()-r / ainsi qu'un sable fin In order to have an explicit terminology, I propose to call "hemistichs" the two strings of words which are delimited by the caesura. Obviously, breaks (or sub­ lines) and caesuras (or hemistichs) do not necessarily coincide, but this possible discrepancy has no metrical relevance in classical French lines [19, 20]. In the English or Italian 10-syllable line, however, the interval bounded by, say, a break after syllable 4 and the associated caesura may be filled with (at most) one vowel belonging to syllable 5, so that it is useful to classify caesuras in two categories. Borrowing Cornulier's terminology with some significant adaptations [11, 12, 13, 14], I will say that a caesura is "synthetical" if, and only if, it is not separated from its break by a non-elided vowel; otherwise, it is "analytical." Here too, I would advance two universal principles. The first one stipulates that a line design may impose (but not prohibit) the use ofsynthetical caesuras. When 164 I DOMINICY synthetical caesura is compulsory, as in the classical alexandrine, this will be represented by substituting the "+" sign for the hyphen: XXXXXX+XXXXXX(X) In the unmarked situation (i,e., in the absence of any constraint on the syntheticity or analyticity of the caesura), as is the case for English or Italian lO-syllable line with a break after syllable 4, I will simply maintain the hyphen:

XXXX·XXXXXX(X)(X) The second universal relating to breaks and caesuras states that no line design may allow internal extrametricity without requiring synthetical caesuras-or, equivalently, that extrametrical syllables cannot belong to the same word as a subsequent metrical syllable. Accordingly, pre-classical alexandrines or Spanish "alejandrinos" only have synthetical caesuras (this need not be marked by a "+" sign, since it is deductible from the extrametrical positions situated at the end of the first subline).

STRESS, SYNTAX ANDVERSE Thus, it turns out that the constraints which apply to syllable 7 of classical alexandrines follow from the fact that the underlying line design rules out analyt­ ical caesuras. Yet this pattern does not account for the requirements bearing on syllables 6 and 12, which must belong to the masculine part of a word. The problem can easily be solved by adding a condition ensuring that syllables 6 and 12 are stressed [19, 20]:

XXXXXS+XXXXXS(X)

This, together with the supplementary constraint that syllables 12 and 13 should belong to the same word, rightly predicts that the extrametrical syllable is neces­ sarily feminine (posttonic). Again, a universal tendency seems to be at work here: except for some enclitic words (like French"-je" and"-ce") and for very artificial tricks [11, pp. 116-117; 16, pp. 80-81], extrametrical syllables always belong to the same word as the last metrical syllable which precedes them. It should be noticed that the stress constraints we have just introduced do not support the widespread, but nevertheless ill-founded, hypothesis which claims that French verse is "accentual" (see, e.g., [21, 22]). In an accentual verse design (e.g., the pattern of the English iambic pentameter), each syllabic position should be labelled "8" ("strong," "+ stress") or "w" ("weak," "-stress"). Here, two syllables are labeled "8," but the remaining ones receive no specification at all. According to another mistaken conception (which still influences recent authors), any alexandrine should contain two concatenated "prosodic" or "intonational" FRENCH 12-SYLLABLE VERSE / 165 groups, so that the stresses falling on syllables 6 and 12 each occupy a rhyth­ mically prominent position. Yet in a line such as: J'y suis encor malgre-s-tes infidelites (Racine) the prepositional phrase "rnalgre-s-tes infidelites" constitutes one and only one prosodic group, where the main stress falls on "-tes." More generally, there does not exist any conclusive argument in favor of a strictly prosodic analysis of the classical alexandrine. It is true that in most examples. line boundaries do coincide with prosodic boundaries (exceptions to this regularity are traditionally called "enjambements"); it is also true that the following line could not appear in Racine's works [19, 20]: *J'y demeure encor par-tes infidelites But the principles which underlie such regularities are extremely different. The coincidence between line boundaries and prosodic boundaries is easy to obtain for long lines: this explains why the use of classical alexandrines does not exert any significant influence on the expressive power of ordinary language. In other words, the regular distribution of prominent stresses in line-final position stems from the necessity of organizing, simultaneously, the discursive network of the whole message and the metrical parallelisms between equivalent lines. On the contrary, the constraint bearing on syllable 6 in Racine's corpus does not follow from such an overall tendency. What distinguishes Racine's original line from its ill-formed modified version is simply the fact that, in Racine's meter. no alexandrine may exhibit the following syntactic pattern: * ... (;... 5J;# L6- ... 1... where the word boundary "#" would correspond to a major syntactical division. I.e., if syllable 6 coincides with the beginning of a constituent that extends to a syllable of the second subline, then this constituent is embedded in another line-internal constituent which necessarily extends to syllable 5. This constraint, which is abundantly illustrated in [19, 20, 23], seems to distinguish Racine from earlier poets, like Agrippa d' Aubigne [24]. Furthermore, it bears some superficial resemblance to the prohibition of analytical caesuras. since both prevent linguistic units of a certain type (words or constituents) from "spreading" over syllables 6 and 7. However, the prohibition of analytical caesuras does not put any condition on the left boundary ofthe relevant word. Verluyten [19, pp. 337·343; 20] has also noticed that in most alexandrines containing two major constituents, the main syntactic boundary occurs in the first hemistich; i.e., lines like: II n'y travaillera I que trop bien, l'infidele! are clearly marginal in Racine's works. This statistical divergence could result from two independent facts. First, such a syntactic bracketing often creates 166 I DOMINICY enjambement, i.e., a marked discrepancy between prosodic and metrical boundaries. Secondly, lines such as the example quoted above, in which no enjambement is initiated, conflict with the general tendency to place shorter prosodic groups before larger ones [19, pp. 111,201; 20, p. 72). The syntactical constraint bearing on syllable 6 is exceptionally violated by Andre Chtnier (whose works were only published in 1819):

Ne pouvait songer ... Mais I que nous font ses ennuis La terre aux dauphins, I'onde I aux taureaux est fenn6e and it is no longer valid for Hugo's versification [25; 19, pp. 311, 327-329, 357; 20; 23]:

(a) Assieds-toi, Cromwell. Mets I ton chapeau sur ta tete Cette muraille, bloc I d'obscurite funebre L'abime d'en bas, plein I de I'ombre inferieure

(b) Ah! pas d 'injures!-Carr / n'a sauve qu'Israel lis acceptent tout, vie I et tombeau, tlamme et cendre De peur des soupcona-c-Mals I tout cela me confond

(c) Et que Ie Pinde, grac(e) I au nom de Republique Hors du village et pres I de la borne du mille Que disent-ils lA?-SaufI opinion meilleure Entre les cates, par lIe poumon, jusqu'au foie Ala grand messe et chez I Gabrielle d'Estrees

This sample of the Hugolian corpus is organized along the following lines. In the lines under (a), the monosyllabic word occurring before the caesura functions as the head of an endocentric construction. In the (b)-lines, it belongs to an exocentric ("multiple-head") construction (compare with Chenier's lines). Category (c) shows that, in such circumstances, Hugo did not hesitate to fill syllable 6 with a monosyllabic preposition which is clearly dependent on a head occurring in the second hemistich. In addition to the syntactical constraint on syllable 6, classical poets systemati­ cally avoided the use of a monosyllabic preposition or determiner before the caesura. Exceptional lines like [16, p. 67]:

Adieu; je m' en vais ~ I Paris pour mes affaires (Voltaire)

(with, additionally, a major syntactic boundary after syllable 2) aimed at a marked prosaic or colloquial effect. Again, Victor Hugo sometimes violated this supple­ mentary requirement [12, p. 275; 21, p. 104; 25, pp. 124-125, 135]:

(c') Prepositions: Et quand i1s seront pres I des degres de lumiere Je t'approuve.-II faut pour I ne rien faire ademi FRENCH 12-SYLLABLE VERSE I 167

Dans ton prodige et dans II'horreur demesuree

(d) Determiners: Des profondeurs que nul / prophete n'a sondees Venait rire entre deux / ecroulements d'empire C'est acause de tous / ces songes formidables It should be pointed out, however, that he never resorted, in syllable 6 of (c-c'j-lines or (d)-lines, to the preposition "de" or to the determiners "le" and "ce," in spite oftheir very high frequency in ordinary French texts. This means that his versification system prohibits the use of a schwa in such cases, even if it is "masculine" (i.e., belongs to a masculine monosyllable, see [11, 12, 13, 14, 15]). Notice, incidentally, that masculine "e muets" can occur in classical lines like [12, p. 163; IS, pp. 10-11; 22, p. 87]: Le sang coule, bois-le / qu'il suffise ata rage (Rotrou)

S'ecrie: epargnez-le / nous n'avons plus que lui (Florian) where the (stressed) enclitic "le" forms a phonological word with the preceding verb. (However, Cornulier [12, p. 159] suggests that no such lines appear in Hugo's works.) Among the 25 (c-c') lines found in [21] and [25], 16 contain a complex preposition (grace d, loin de, pres de, hors de) whose second element coincides with syllable 7. This may be one of the reasons why Hugo replaced the following unpublished occurrence:

Qui fus designe des / Ie ventre de ma mere with a more traditional formula [25, p. 54]. Given the extremely small number of (d)-lines, it is hard to imagine that the difference between simple determiners (e.g., "nul," "deux") and complex determiners (e.g., tous ces) may have played a significant role. Yet, if we are to believe Richard Lesclide, Hugo was gripped by a kind of "horror" ("une sorte d'epouvante [ ... ] s'empara de lui") when he realized he had written an (unpublished) alexandrine with the definite article "Ies" in position 6 [25, p. 50]:

Dans les palais, dans les / chateaux, dans les chaumieres In this respect, later poets of the so-called "Parnassian" generation (Baudelaire, Banville, Leconte de Lisle) were less conservative (12, pp. 200-201; 21, pp. 108-109; 26]). For instance, we find 27 occurrences ofthe (c-c', d) pattern in Leconte de Lisle's Poemes barbares: Us s'en venaient de la / montagne et de la plaine Car il connut, dans son / esprit, que c'etait la Ni les aigles, ni les / vautours ne mangeront Le jour tombe. Que mon / Seigneur se leve et mange 168 I DOMINICY

Je me leve dans la I fureurqui me consume Voicivenir, pour la I curee,a Roi sanglant, Commedes merlesdans Il'epaisseur des buissons Par coupsde foudreet parI rafalesemporte La queue en eercle sousI leurs ventrespalpitants De son ombre, de son I repos, de sa colere Quandj'embrassais dans un(e)I irresistible envie Pour I'absorber dans tonI impassiblebeaute J'aurais passe par la Ilumiere et les t6nebres Parmiees cris et ees I angoisseset ces fievres En ees beaux siecles,saufI votre permission Et son ombre dans sa I chaleuret sa poussjere Arnanature,sans I colereet sans exces Mais de ceci, pour mon/malheur, ne sachant rien Vivez,Rabbi,dans la I priereet Ie silence Que Dieu la juge en son I infaillible Cquit6 Cachela tete sous Ila nappe,a monenfant Qu'on ne lui doit, en tout(e)I equite, qu'une corde L'autre s'abat, alaI renverse, 6claboussant Avecla Reine,et dans I son lit dormirton somme Seul, immobile, et sur Ila dalle agenouille Et triomphantdans sa I hideusederaison Serontfruits murs de votr(e)I impeccabilite As pointed out by Verluyten [19, 20], Leconte de Lisle and other Pamassian poets also avoided the use of "de," "le" and "ce" in position 6. Interestingly enough, Leconte de Lisle's Poemes barbares contain two occurrences of the masculine enclitic "-Ie" before the caesura: Allez, lapidez-IeI car il parletres mal Enoutre, sachez-Ie I j'en ai mangebeaucoup Moreover, all of Leconte de Lisle's lines listed above have a word stress on syllable 4 or (inclusively) on syllable 8. It thus turns out that Leconte de Lisle and, presumably, other poets introduced two very different constraints on lines which, anyway, were felt as extremely deviant with respect to the classical-and even romantie--models (see Anatole France's comments, quoted in [26, p. xv], and [27]). Notice, however, that Baudelaire and the younger Catulle Mendes did not always conform to the second constraint [12, p. 200; 19, pp. 358-360; 20]): Lui,se tournevers son I foumisseurordinaire(Baudelaire)

Oil la Lune comme unI froidIinceulsemblesuivre La cigaleavec ses I grelescris obsedants(Mendes) Furthermore, Comulier has shown, in his detailed study of Mallarme and Rimbaud [12, pp. 192-200; 28], that these authors did not correlate the occurrence of a monosyllabic preposition or determiner in syllable 6 with any constraint FRENCH 12·SYLLABLE VERSE / 169 regarding schwa or the stress properties of syllables 4 and 8. This is immediately illustrated by lines such as: Dont Ie vol selon Ie / reverbere decouche (Mallarme)

Noire et froide 011 vers Ie / crepuscule ernbaume (Rimbaud)

From this point of view, Verlaine's versification seems a little more conserva­ tive. In the Poemes saturniens (1866), all alexandrines that exhibit a monosyllabic preposition or determiner in position 6 meet the two requirements of the Parnas­ sian system (see the list in [19, p. 364; 20, p. 62]). E.g.,:

De la douceur, de la / douceur, de la douceur

(which bears a striking resemblance to Hugo's suppressed line). After the Poemes satumiens and before Amour (1888), all alexandrines where syllable 6 is filled with "de," "Ie" or "ce" have a stress on syllable 4 or (inclusively) on syllable 8. Examples [12, p. 299]:

De par la grace, Ie / sourire et la bonte Un parfum.e-et c'est Ie / deluge qui consomme Ce chef horrible et Ie / vide de la hantise

However, Verlaine's works also contain a line like:

Se recueillent dans un / egotsme de marbre which was published before the Poemes saturniens [12, p. 200].

THE DISLOCATION OF THE CLASSICAL ALEXANDRINE

When dealing with Mallarme, Rimbaud, Verlaine, and many later poets, we cannot dissociate the problems we have just discussed from deeper violations of classical rules. Relying on Comulier's typology, I will distinguish between three categories ofdeviant 12-syllable lines. "F7-lines" have an analytical caesura while respecting the stress constraint on syllable 6:

Font les saules, d'oll sau-tent / les oiseaux sans bride (Rimbaud) Dans votre sein, sur vo--tre / coeur qui fut Ie notre (Verlaine)

"F6-lines" have a synthetical caesura but violate the stress constraint on syllable 6:

Nubiles plis I'astre / milri des lendemains (Mallarme) Forets, soleils, rives / savanes!-lI s'aidait (Rimbaud) Une candeur d'une / fraicheur delicieuse (Verlaine)

"M6-lines" both have an analytical caesura and violate the stress constraint on syllable 6: 170 I DOMINICY

Mais eux, pourquoi n'endos-ser I pas, cesbaladins (Mallarme) Cependant que, silen-cieux I sous lespilastres (Rimbaud) Etlatigresse epou-c-vantable I d'Hyrcanie (Verlaine) The existence of such lines has given rise to a rather technical and sometimes confusingdebatebetween Verluyten[19, 20] and Cornulier[12]. In the following, I will try to summarizethe main lines of this discussion, and provide some clues for evaluatingthe explanatoryadequacyof each approach. Let us first examine the case of Mallarme, Cornulier has shown that all his 12-syllable lines conform to a simple implicational rule: Ifa line does not instantiate the classicaldesign: XXXXXS+XXXXXS(X) [dominant pattern] then it can be scanned with one break after syllable 4 and another one after syllable8, withoutany constrainton the corresponding caesuras: XXXS·XXXS-XXXS(X) [marginal pattern] the dominant pattern being significantly morefrequent than the marginal pattern. The same regularity can be observed in Rimbaud's works, except for one line of the Album zutique, and for two long poems belonging to the so-called Derniers vers,to which I will come back below. So that we obtain scansions like [15, p. 16]: Nubiles plis--I'astre mQri-deslendemains Forets, soleils-e-rives, sava-nes! IIs'aidait Mais eux, pourquoi-n'endosser pas-ces baladins Cependant que-silencieux-sous lespilastres ConcerningVerlaine,the situation is again more complex. The whole set of his production seems to be structured by a series of embedded implicational rules. Indeed,we can classify his 12-syllable lines in 4 categories: (i) those lines which instantiatethe classicaldesign (pattern 1); (ii) those lines which instantiatethe Mallarmean design (pattern 2); (iii) those lines which do not instantiate one of the two precedingdesigns, but can be scannedwitha break aftersyllable8, withoutany constrainton the corresponding caesura: XXXXXXXS-XXXS(X) (pattern3); (iv) thoselineswhichdo not instantiate one of thethreeprecedingdesigns,but can be scannedwith a break aftersyllable4, withoutany constrainton the corresponding caesura: XXXS-XXXXXXXS(X) (pattern4); with the similar statistical result that, for any i<4, pattern i is significantly more frequent that pattern i+1. Here are some examples for categories.(ii), (iii) and (iv): FRENCH 12-SYlLABlE VERSE I 171

(ii) Dans votre sein-sur votre cceur-i-quifut Ie notre (F7) Une candeur-d'une fratcheur-i-delicieuse(F6) Et la tigres-s(e) epouvanta-s-bled'Hyrcanie (M6)

(iii) Dans les cages plus d'une clo---ch(e) encore bruit (F7) Une plaine lourde de sang-s-bleme de nuit (F6) Et que sonnent les Angelu~roses et noirs (M6)

(iv) Saint-Ouen, Montrou-ge, d'autres peut-etre ou les vagues (F7) Ace qu'assu-r(e) une dure philosophie (F6) En robe d'or-il adore, gloire et symbole (M6)

Although Rimbaud's exceptional line from theA/bum zutique clearly belongs to category (iii):

Eclatent, tncolorement-s-enrubannes most problematic 12-syllable lines of the Derniers vers seem to be simply devoid of any internal division [12,21,28,29, 3D}. In Verlaine's own works, some rare examples do not even conform to the extremely liberal system I have just described, e.g.:

De cette Science intruse dans la Maison unless one posits a 6+6 break which induces a morphological remotivation of "in-truse" [12, p, 236}. However, the general relevance of categories (i) to (iv) cannot be doubled. For instance, the apparently chaotic versification of Jules Laforgue's 12-syllable lines exhibits a series of metrical properties that only make sense with respect to a basic 8-4 or 4-8 break [31}. Moreover, some more recent poets obviously rely on a "neo-classical" diction which systematizes Verlaine's practice. One clear illustration is provided by Jacques Audiberti's Race des hommes (1937). The 1853 12-syllable lines of this collection which were examined [32] divide up as follows:

(a) 1738 lines belong to category (i), (b) 77 lines belong to category (ii). Puis il bondit-s-commodebouc-et comme d'or (F7) 11 traversait-une foret-Contre les baies (F6) A quel creuset-me taxiez-vous-s-a queUestiges (M6)

(c) 27 lines belong to category (iii). Celul-la therapeute fut-et bohemien (F7) Les barons consultent 1a bie-r(e). Un juif dehors (F6) S'incorpore viridite-lucidite (M6)

(d) 11 lines belong to category (iv). Se rompt, soudain-d'une membraneuse fenetre (F6) 172 / DOMINICY

Related results are reported by Cornulier in his study on YvesBonnefoy [33]. The emergence of the Mallarmean ternary design (4-4-4) could result from a complex interaction of meter and prosody. We have seen above that Racine's corpusdoesnotcontainanyalexandrine whereposition 6 coincides withtheinitial syllable of a major polysyllabic constituent. This prohibition, together with the stressconstraint on syllable 6 and thegeneral avoidance of prominent stressclash [19,pp. 84-107], entailsthatno rhythmically salientstresswill fall on syllable5. Indeed, sucha patterncouldonlybe triggered by bracketings whichviolate either the stressconstraint or thesyntactic constraint:

s F • [i 5 6 ]i ("F6-line") s • [i 5 ]i # L6- ... 1...

Notice that, contrary to what is sometimes claimed (e.g., [21, p. 89], see [19,20)), no similarcondition applies to position 7. For instance, in:

Seigneur,si j'ai trouv6-grAce devantvos yeux (Racine)

theprominent stressof theconstituent "trouvegrace"fallson syllable7. Besides, the classical conventions automatically favor the occurrence of two prosodic peaks,onesituated before syllable 5, for instance on syllable 4, and the otherone near to, or on syllable 8:

MauditchAteau! mauditamour! mauditvoyage (La Fontaine) Toujoursaimer, toujourssouffrir,toujoursmourir(Corneille) Et Mardoch6eest-il ausside ce festln(Racine) Et pres de vous ce sont des sots que tous les hommes (Moliere)

Indeed, recent research [34, 35] has shown that rhythmical groups tend to counterbalance each other and that they have, in French, a mean syllabic length which oscillates between 2 and 5 and cannotexceed 8 (recall Cornulier's "8-syllable rule"). This possibility of coupling a 6+6 design and a ternary 4-4-4 rhythm was intensively exploited by romantic poets. Let me only recall Hugo's famous slogan(for moreexamples, see [25,pp. 108-111 and36)):

J'ai disloquece grand niaisd'alexandrin

Accordingly, readers of romantic poetry (including all poetsof the Parnassian generation) mayhavebegun to perceive theternary rhythm as "meaning" the6+6 design in almost the sameway as spots"mean" measles [37]. This wouldbe the reason whytheuse of a monosyllabic preposition or determiner in position 6 is so frequently "compensated" by an easilyidentified 4-4-4prosodic pattern(see [12, pp. 201-204)) and theexamples of Leconte de Lislequoted above). FRENCH 12·SYLLABLE VERSE I 173

CORNULIER VERSUS VERLUYTEN

Up to now, I have followed Cornulier's main line of argument, while relying on Verluyten's findings about the syntactic constraint on syllable 6. In fact, the debate I have alluded to before concerns the theoretical status of what Cornulier calls "the 4-4-4 measure," and the role played by this controversial entity in the emergence of the innovative line designs. Let us consider F7., F6-, and M6-lines, which violate either the prohibition of analytical caesuras or (inclusively) the stress constraint. Cornulier's explanation runs as follows. Experienced readers of romantic and Parnassian poetry, for whom a 4-4-4 prosodic pattern "meant" the abstract 6+6 design, could extend this recognition strategy to F7-, F6- and M6­ lines, provided these had the required rhythmical features. Once this new percep­ tual habit was established, it became possible for a handful of "happy few" (the poets themselves and maybe some other readers) to keep perceiving the original design when either the first or the second prosodic peak was absent, so that two derivative patterns (4-8, 8-4) progressively emerged. Now, the question arises whether such patterns were purely rhythmical (i.e., continued to function as "signs" or "symptoms" of the underlying 6+6 design) or turned into autonomous metrical models. In my presentation of the data, I have opted for the second interpretation, which increases the empirical content of Cornulier's theory, and makes the comparison with Verluyten's approach easier. Cornulier does not make a clear choice in this respect, so that one might believe, at first sight, that his disagreement with Verluyten is mainly terminological. But in my view, Verluyten's originality lies precisely in the fact that he denies any metrical relevance to the ternary pattern and its derivatives, and prefers to modify the "matching rules" which relate the classical 6+6 design to its possible instances. Verluyten uses the technical notion of "mismatch" (see [38]) in order to describe the possible conflicts between the underlying design and the line instances. We have a "bracketing mismatch" when the caesura is analytical (to use my own terminology), i.e., when the 12-syllable line is either F7 or M6. We have a "vowel quality mismatch" when the nucleus of syllable 6 is a schwa (Verluyten does not distinguish between masculine and feminine schwa). Finally, we have a "prominence labelling mismatch" when syllable 6 is not stressed, i.e., is pre- or posttonic. Furthermore, Verluyten formulates an "alternate stress rule" which puts a "secondary stress" on each syllable which is separated from the syllable bearing the main stress by an odd number of intermediate syllables. So he assumes that the stress pattern of French words conforms to the following schemes:

s(W) roi, reine WS(W) amour, voiture SWS(W) amoureux, habitude wsws(W) assurement, mathematique swswsrwi immoralite,anthropocentrisme 174 I DOMINICY

It is obvious that such a hypothesis provides a very elegant explanation for the fact that most M6-lines have a stress on syllable 8. What happens, according to Verluyten, is that a secondary stress realizes the strong "S" position of the classical design: Les scrupules des Zlbertins et des begueules Car ce myst~re,Z'Incarnation est tel Pense! Et quel beau cas batrachomyomachique (Verlaine) This can also be seen as a strategy which aims at avoiding the cooccurrence of a bracketing mismatch and a prominence labelling mismatch. By the same token, Verluyten rightly predicts that most instances where positions 6 and 7 coincide with the penultimate and final syllables of a word are F7-lines. In order to account for F6-lines, Verluyten simply admits that vowel quality mismatches became possible at some point in the history of French versification. Yet he claims that the cooccurrence of a vowel quality mismatch and a bracketing mismatch is systematically avoided. Indeed, such lines are not common:

Accueillent d 'escroque-rie Aprele poete (Verlaine) Notre marche venge-resse a tout occupe (Rimbaud)

As with lines which combine a bracketing mismatch with a prominence labelling mismatch, the two preceding lines should be considered as metrically ill-formed. Because of its restrictiveness, Verluyten's theory is very easy to falsify. He does not account for the fact that F7- and F6-lines strongly favor the occurrence of stress on syllable 4 or (inclusively) on syllable 8; nor does he explain why the same regularity often holds for lines where syllable 6 is filled with a monosyllabic preposition or determiner (Which he considers to be "prosodically W") [12, pp. 201-204; 20, p. 58]. More generally, his alternate stress rule comes up against serious objections [3D, p. 303). In French, secondary stresses- are distributed according to different principles, which do not always produce Verluyten's binary patterns [34,35,39,40]. First, polymorphemic words may have a secondary stress at the end of a non-final morpheme; e.g.:

[anti]constitutionnel, [m~Zo]dramatisation. [hyper]sensibilise Otherwise, the secondary stress may fall: (i) on the antepenultimate masculine syllable of the word (this is a subcase of Verluyten's rule); or (ii) on the first or second syllable of the words e.g.: char-mant, forma lite, intoxication When the secondary stress falls on the initial syllable, we have what Milner and Regnault call a "counter-accent" [22, p. 82]. In certain cases, especially when the word begins with a vowel, the "counter-accent" may shift to the second syllable, unless this syllable bears the main stress or is immediately contiguous to the syllable bearing the main stress. (The same process is at work with the so-called FRENCH 12·SYLLABLE VERSE I 175

"expressive accent" in a word like "~pouvantable" [19, p. 374; 20, p. 66; 34, 35, 39]; see Cornulier's comments [12, p. 242] on Verlaine's "tigresse epou-s-vant­ able.") Either secondary stress or main stress may be deleted by further prosodic rules; this deletion has to apply when there is a stress clash. E.g. [34, 35, 39]: --charmant, un charmant homme, un channant garfon, un charmant garfon -Ies formalites administratives (binary secondary stress, no stress deletion) -Ies formalites administratives (secondary stress on the first or second syllable, no stress deletion) -Iesformalite-sadministratives (secondary stress on first syllable with an initial consonant resulting from "liaison," deletion of the first main stress).

Such data illustrate the complex nature of French secondary stress, and raise the question of whether this phenomenon could exert any significant influence on metrical well- or ill-formedness (compare with the situation in Spanish, where meter is "accentual" [41]). Consider, for instance, the following two lines of Verlaine:

Avec, rares, des bouquets d'arbres et de l'eau Paix pour de bon, paix definitive et sans treve The prosodic pattems which are indicated conform to the general principles of French accentuation. (In the first line, we have a clear example of "stress retrac­ tion," i.e., deletion of the main stress in "bou-quets"; see [19, pp. 374-375; 20, pp. 66-67].) However, Verluyten himself acknowledges that prosodic rules only account for a quite limited number of cases, and sometimes lead us to mix matters of meter with judgments about possible phonetic deliveries. In Comulier's approach, the two lines I have just discussed simply belong to category (iii) or to category (iv), Yet, the late emergence of the 4-8 design remains a mystery. Given the general propensity to have shorter intonational groups before larger ones (see above), one would have expected a different chronology [19, p, 381; 20, p. 72]. This is all the more intriguing since few experienced readers manage to identify 4-8 lines. (personally, I perceive most 8-4 lines at first or second sight, but I have to count syllables in order to verify that a line actually instantiates the 4-8 design.) Moreover, if we endorse the radical version of Comulier's theory-i.e., if we assume that we have to distinguish between four underlying designs, then we are committed to the claim that complex designs may derive from other complex designs by a kind of reanalysis mechanism, and not only by the concatenation of two autonomous designs. I do not know of any independent evidence which would support this hypothesis.

AMBIVALENT LINES Whatever their disagreements, both Cornulier and Verluyten maintain that many 12-syllable lines are "ambivalent," i.e., instantiate the underlying 6+6 176 I DOMINICY design while having at least one property which makes them look deviant with respect to this classical model. For instance, Victor Hugo's works contain some M6-lines, where the break always occurs in a compound or morphemically com­ plex word [12, p. 276; 25, p. 120; 36]: Mon bien-aime, mon bien-aime, mon bien-aime Ne croyez pas qu'a Bray-sur-Marne, ()citadins Et j'ai fait mettre au For-l'Eveque la Duthe Gallus: Inaccessible, inex ... Gunich: ... pugnable (Souriant et saluant) Et vulnerable.

All these lines may receive a clear 4-4-4 pattern (even the last one, although it is spoken by two characters). Yet, the morphological status of the problematic words induces us to divide each instance into equal segments of 6 syllables. Notice also that "inexpugnable" normally has a secondary stress on its second syllable, since it begins with a vowel and exhibits a strong morphological boundary after "inex-" (see above). We find many similar cases in Verlaine, with a 4-4-4 pattern or one of its derivatives [12, pp. 166-167, 239-246; 19, pp. 365-366; 20, pp. 62-63]. For example:

(a) secondary stress in a compound word: Au premier sang pour Cou-de-Marbre ... -I.e bruit court La Colombe, Ie Saint-Esprit, Ie saint Delire Cent fois.-Et eet Ivry-commune d'~ present!

(b) secondary stress before a morphological boundary: Universel, cosmopolite et polyglotte O'ailleurs, toe, galvanoplastie et similor! A cause de l'emmerdement de la mitraille

However, contrary to Cornulier [12, pp. 235-246], I would distinguish such cases from those lines where Verlaine strives to create a "metrical pun" by putting an obscene syllable in position 6 [19, pp. 378-379; 20, p. 69]. Adverbs in "-(e)ment" are extremely well-adapted to an ambivalent context, given their high degree of semantic motivation with respect to adjectives [41, p. 49]. Recall Dante's famous "endecasillabo"-"Con tre gole caninamente latra"-where the 6-4 break coincides almost exactly with a morphological boundary. Similarly, in Verlaine's «Lc Rossignol» (Poemes saturniensy; a poem written in 5+5 decasyllables, a morphologically justifiable break couples the adverb "melancoliquement" with the related adjective: FRENCH 12-SYLLABLE VERSE / 177

Qui melancolj--quement coule aupres [ ... ] Nuit melancoli-i-qufe) et lourde d'ete The same pattern is frequently used in alexandrines (see [12; 25, p. 59; 36, p. 57]:

Elle filait pensi-vement la blanche laine (Banville) Contre Ie marbre vai-nement de Baudelaire (Mallarme) Puis franchement et sim-plement viens il rna Table (Verlaine) Belatent, tricolo-rement enrubannes (Rimbaud) Nous avancions tranquil-Iement sous les etoiles (Verhaeren) Observe that, again, each of these lines conforms either to the 4-4-4 or to the 8-4 design. If we resorted to morphological or pseudo-morphological (re)motivation without the minimal care, we would certainly be led to adopt rather eccentric scansions. I have already mentioned that some desperately deviant verses, such as:

De cette Science intruse dans la maison (Verlaine) can be made regular by an "ad hoc" morphemic analysis. I do not want to deny the existence of such stylistic devices, but I think it must be corroborated, in each particular case, by independent textual or philological evidence. Here are two examples of this procedure. Let us consider the two M6-lines of Rimbaud's Bateauivre [12, pp. 166, 175, 208;28]: Je courus! Et les Peninsules demarrees N'ont pas subi tohu-bohus plus triomphants Given the clear 4-4-4 pattern of the second line, and the stress on syllable 8 in the first one, one could reasonably assume that they respectively belong to category (ii) and to category (iii). However, "tohu-bohus" is a compound word whose reduplicate structure strongly favors the occurrence of a secondary stress on syllable 6. As for "Peninsules," its initial upper case letter and its etymology ("paene-insula") suggest the presence of a parallel secondary stress, which would precede a 6-6 break. (The idea that capitalization may signal the regular stress required by an abstract verse design was put forward by Jakobsen and Jones [I, p. 294].) This interpretation would make us understand why, later in the poem: Presque ile, balloUant sur mes bords les quereUes Rimbaud did not use the standard orthography for "presqu'tle." In both contexts, he wanted to indicate that the "bateau ivre" cannot be compared to an ordinary peninsula (a typically static object), but rather to something which is on the verge of becoming an island [15, pp. 39-40]. 178 I OOMINICY

Apollinaire's «Palais» (in Alcools), a poem written in a variety of the post­ Verlainian neo-classical diction, begins with an M6-line which has an easily perceivable 4-4-4 prosody:

Versle palaisde Rosemonde au fond du Reve

Again, a 6-6 break would remotivate an etymological division of "Rosemonde"; and again, we have independent evidence for this hypothesis [42]: Je 18 sumommaiRosemonde Vaulantpouvoirme rappeler Sa bouchefleurieen Hollande Puis lentement je m'en allai Pourqueter la Rose du Monde(<

EPILOGUE Any versification system consists of line designs-which underlie metrical equivalences-and of matching rules-which constrain the linguistic realization of these abstract models. In spite of the rapid evolution of modem poetic diction, the classical 6+6 design and its derivatives keep functioning as a reference point for many poets, who generally relax the matching rules regarding syllable count, hiatus, V-e sequences, and elision. This stability of the underlying designs is perhaps the most characteristic feature of meter in general [43]. Yet one would surely miss it by sticking to an intuitive approach where all relevant levels (meter, syntax, prosody, ... ) are continuously mixed. In particular, no true understanding of poetry can be attained by relying on "delivery designs" (as do Milner and Regnault [22]) or, even worse, on "delivery instances" that frequently select one specific interpretation without any further justification (see [I, pp. 36-38]). In several recent articles [4, 5, 44, 45, 46, 47], I have tried to develop a nee-Aristotelian poetics which would integrate the insights gained by the (post-) Jakobsonian research program. In this perspective, I assume that verse tends to acquire the formal and cognitive properties which are typical of words or "ready­ made" expressions like idioms and proverbs. On the formal level, we can point to striking similarities between the rules regarding hiatus or elision and some phonological principles which mainly apply within the domain of the word (and sometimes extend to the immediately higher domain of the so-called "phonological word," see [22, 48]). Consider, for instance, the contrast between "maison oli"t'Castillon amoureux" (no hiatus) and "alle oli"t'Ledru amourcux" (hiatus); and compare with the fol­ lowing morphological data: "creation" - "creationnisme," "Petain" ­ "petainisme" as compared to "dada" - "dadaisme," and "Bouddha - "boud­ dhisme" [49]. When the "-isme" suffix is added to a word ending with a nasal vowel (i.e., to a word whose orthographical notation ends with a nasal consonantal FRENCH 12-SYLLABLE VERSE I 179 letter), a fully pronounced nasal consonant surfaces, which avoids the creation of a V-V sequence. (Recall that French phonology prohibits word-internal VI-V2 sequences where VI is a nasal vowel, see [50].) On the contrary, the conflicting treatments illustrated by "dadai"sme" (no loss of the final a) and "bouddhisme" (loss of the final a) show that no automatic rule operates when a V-V sequence would normally appear. In other words, derivational morphology is sensitive to one of the parameters which discriminates hiatus from non-hiatus contexts. Similarly, we may draw a parallel between the basic metrical rule of elision and the fact that no acceptable French word contains a phonological e-V sequence [50, pp. 202-204]; or between the blocking of elision by any orthographical consonant (even if it is never pronounced) and the surfacing of purely orthographical con­ sonants before a suffix: see, e.g., "Flaubert" (where the t is never pronounced) ­ "flaubertien" as compared to "Wagner" - "wagnerien." Part of the formal and semantic density of poetic discourse certainly originates from this almost "agglutinating" nature of verse. As Mallarme magnificently put it,

Le vers qui de plusieurs vocables refait un mot total, neuf, etranger It la langue et comme incantatoire, acheve cet isolement de la parole: niant, d'un trait souverain, le hasard demeure aux termes malgreI'artifice de leur retrempe altemee en Ie sens et la sonorite, et vous cause cette surprise de n'avoir oUI jamais tel fragment ordinaire d'elocution, en meme temps que la reminiscence de I'objet nomme baigne dans une neuve atmosphere.

REFERENCES

1. R. Jakobsen, Selected Writings Ill. Poetry of Grammar and Grammar of Poetry, S. Rudy (ed.), Mouton Publishers, The Hague, 1981. 2. R. Jakobson, Selected Writings V. On Verse, Its Masters and Explorers, S. Rudy and M. Taylor (eds.), Mouton Publishers, The Hague, 1979. 3. M. Delcroix and W. Geerts (eds.), «LesChats» de Baudelaire. Une confrontation de methodes, Presses Universitaires de Namur-Presses Universitaires de France, Namur-Paris, 1981. 4. M. Dominicy, Y a-toil une rhetorique de la poesie?, LangueFrancoise, 79, pp. 51-63, 1988. 5. M. Dominicy, Sur I'epistemologie de la poetique, Epistemologie de la linguistique, Histoire Epistemologie Langage, 13, pp. 151.174, 1991. 6. N. Ruwet, Parallelismes et deviations en poesie, J. Kristeva et alii. (eds.), Langue. Discours. Societe. Pour Emile Benveniste, Editions du Seuil, Paris, 1975. 7. N. Ruwet, Unguistique et poetique, Une breve introduction, Le Francais Moderne, 49, pp. 1·19, 1981. 8. B. de Cornulier, Versifier: Ie code et sa regie, Poetique, 66, pp, 191-197, 1986. 9. S. R. Levin, Unguistic Structures in Poetry, Mouton, The Hague, 1962. 10. N. Ruwet, Langage, musique, poesle, Editions du Seuil, Paris, 1972. 11. B. de Cornulier, Elements de versification fran'iaise, A. Kibedi-Varga (ed.), Theorie de la litterature, Picard, Paris, 1981. 180 / DOMINICY

12. B. de Comulier, Theone du verso Rimbaud, Verlaine, Mallarme, Editions du Seuil, Paris, 1982. 13. B. de Comulier, M~trique, Encyclopedia universalis, Paris, 1989. 14. B. de Comulier, Art poetlque, Centre d'~tudes m~triques, Universi~ de Nantes, Nantes, 1991. 15. M. Dominicy, Sur la notion d'e masculin et f~minin en metrique et en phonologie, Recherches Linguistiques, 12, pp. 7-45, 1984. 16. W. T. Elwert, Trai~ de versification fran~ise des origines il nos jours, Klincksieck, Paris,I965. 17. T. Navarro TomAs. M~trica espaftola, Labor, Barcelona, (6th Edition), 1983. 18. W. T. Elwert, Versificazione italiana dalle origini ai nostri glomi, Le Monnier, Firenze, 1973. 19. S. P. Verluyten, Recherches sur la prosodie et la metrique du fran~ais, doctoral dissertation, Universitaire Instelling Antwerpen [University Microfilms International, Ann Arbor], 1982. 20. S. P. Verluyten, L'analyse de I'alexandrin. M~tre ou rythme?, M. Dominicy (ed.), Le souci des apparences. Neuf etudes de poetlque et de m6trique, Editions de I'Unlversite de Bruxelles, Bruxelles, 1989. 21. J. Roubaud, La vieillesse d'Alexandre. Essai sur quelques ~tats recents du vers fran~ais, Fran~is Maspero, Paris, 1978. 22. J.·C. Milner and F. Regnault, Dire Ie vers. Court trai~ ill'intention des acteurs et des amateurs d'a1exandrins, Editions du Seuil, Paris, 1987. 23. S. P. Verluyten, Contraintes syntaxiques il la cesure, in Linguistics in Belgium 4, M. Dominicy and C. Peeters (eds.), Didier Hatier, Bruxelles, 1981. 24. H. Chataign6(Centre d'etudes ~triques, Universit6de Nantes), personal communication. 25. A. Rochette, L'alexandrin chez Victor Hugo, Hachette, Paris, 1911. 26. J.·P. Bobillot, Recherches sur la crise d'identit6 du vcrs dans la poesie fran~ise 1873·1913, These de doctorat d'etat es lettres, Universit6 de Paris III, Paris, 1991. 27. J.·M. Guyau, Les problemes de I'esthetique contemporaine, F61ix Alcan, Paris, (13th Edition),1935. 28. B. de Cornulier, M6trique du vers de 12 syllabes chez Rimbaud, Le Francais Moderne, 48, pp. 140·174, 1980. 29. B. de Cornulier, Metre "impair," m~trique "insaisissable"? Sur les "derniers vers" de Rimbaud, M. Dominicy (ed.), Le souci des apparences. Neuf etudes de poetique et de m~trique, Editions de l'Universit6 de Bruxelles, Bruxelles, 1989. 30. D. Billy, Quelques apports r~nts ilia metrique fran~ise, Bulletinde la SocMt~ de Linguistique deParis, 84, pp. 283·319,1989. 31. J.·M. Gouvard, M6trique et metrico-metrie des poemes de Jules Laforgue, Diplorne d'etudes approfondies, Universit6 de Nantes, Nantes, 1991. 32. D. Huybrecht, Jacques Audiberti: "Race des hommes." Etude metrique, syntaxique et poetique, memoire de licence, Universit6 Libre de Bruxelles, Faculte de Philosophie et Lettres, Bruxelles, 1990. 33. B. de Comulier, M~trique de I'alexandrin d'Yves Bonnefoy: essai d'analyse m~thodique,LangueFrancaise, 49, pp. 30·48,1981. 34. V. Pasdeloup, Modele de regles rythmiques du fran~is applique il la synthese de la parole, these de doctorat nouveau regime, Universit6 de Provence, Aix-Marsellle, 1990. FRENCH 12·SYLLABLE VERSE I 181

35. V. Pasdeloup, Essai de modelisation du rythme en francais, Journal d'Acoustique, 2, pp. 47·51,1989. 36. M. Grimaud, Trimetre et role poetique de la cesure chez Victor Hugo, Romanic Review, 70, pp.56-68, 1979. 37. H. P. Grice, Meaning, D. D. Steinberg, and L. A. Jakobovits (eds.), Semantics. An Interdisciplinary Reader in Philosophy, Linguistics and Psychology, Cambridge University Press, Cambridge, 1971. 38. P. Kiparsky, The Rhythmic Structure of English Verse, Linguistic Inquiry, 8, pp. 189-247, 1971. 39. V. Pasdeloup, Essai d'analyse du systeme aceentuel du francais: distribution de I'accent secondaire, Actes des l7emes Journees d'Etudes sur la Parole, pp. 65-70, 1988. 40. P. Mertens, L'intonation du fran~ais. De la description linguistique ala reconnaissance automatique, doctoral dissertation, Katholieke Universiteit Leuven, Leuven, 1987. 41. M. Dominicy, Accent et rythme en espagnol, M. Dominicy and M. Wilmet (eds.), Linguistique romane et linguistique fran~aise. Hommages it Jacques Pohl, Editions de l'Universite de Bruxelles, Bruxelles, 1980. 42. A. Fongaro,Apollinaire poete,Presses Universitaires du Mirail, Toulouse, 1988. 43. P. Kiparsky, The Role of Linguistics in a Theory of Poetry, in Language as a Human Problem, E. Haugen and M. Bloomfield (eds.), W. W. Norton, New York,1974. 44. M. Dominicy, De la pluralite semantique du langage. Rherorique et poetique, Poetique, 80, pp. 499-514, 1989. 45. M. Dominicy, Prolegomenes aune theorie generale de l'evocation, M. Vanhelleputte (ed.), Semantique textuelle et evocation, Peeters (Brussels Publications in Artistic and Literary Studies I), Louvain,1990. 46. M. Dominicy, Pour une theone de I'enonciation poetique, Acres du colloque "Enoncia­ tion et parti-pris," Anvers, in press. 47. M. Dorninicy, Du "style" en poesie,Actes du colloque "Qu'est-ce que Ie style?", Paris, in press. 48. l-C. Milner, Reflexions sur Ie fonctionnement du vers francais, Ordres et raisons de langue, Editions du Seuil, Paris, 1982. 49. C. Plantin, Les mots, les arguments, Ie texte, these de doctorat, Universite Libre de Bruxelles, Bruxelles, 1988. 50. F. Dell, Les regleset les sons. Introduction ala phonologie generative, Hermann, Paris, 1973.

Direct reprint requests to:

Marc Dominicy Linguistique generate CP 175 Universite Libre de Bruxelles Avenue Roosevelt, 50 B-I050 Brussels Belgium