Hostname: page-component-6898cfcb55-w5hgh Total loading time: 0 Render date: 2026-11-11T18:52:19.610Z Has data issue: false hasContentIssue false
  • English
  • Français

More on (distinctive!) vowel length in historical French*

Published online by Cambridge University Press:  01 July 2008

RANDALL GESS
RANDALL GESS*
Affiliation:
Carleton University
*
Address for correspondence: Randall Gess, School of Linguistics and Applied Language Studies, 215 Paterson Hall, Carleton University, 1125 Colonel By Drive, Ottawa, ON K1S 5B6, Canada e-mail: randall_gess@carleton.ca
Rights & Permissions [Opens in a new window]

Abstract

While rejecting a claim for the disappearance of distinctive vowel length in historical French as ‘counterfactual’ (Picard, 2004: 3), Picard's own arguments in support of the existence of vowel length do not rise to the level of fact. Picard fails to differentiate between derived versus underlying (hence distinctive) features. Further, his assumptions regarding vowel length from the Middle French period on are ill founded. Regarding the truly minor vocalic contrasts that do exist in Canadian French mid and low vowels, Picard makes several unmotivated assumptions and unsupported assertions that preclude consideration of other, plausible scenarios for their existence. Minor vocalic contrasts such as these, with little to no functional load, can be modeled in phonological grammars without an unnecessary proliferation of phonological categories.

Information

Type
Articles
Information
Journal of French Language Studies , Volume 18 , Issue 2 , July 2008, pp. 175 - 187
Copyright
Copyright © Cambridge University Press 2008

1 INTRODUCTION

Following an analysis of rhyming pairs in Old French poetry, Gess (Reference Gess and Brinton2001) concludes that vowel length distinctions introduced by compensatory lengthening following Old French /s/-deletion (e.g., /tɛːtə/ < /tɛstə/) were lost in the sixteenth century. Picard (Reference Picard2004) asserts that this conclusion ‘cannot possibly be correct’ [. . .] ‘[g]iven that many of these long vowels are still present in Canadian French’(1). In this paper, I will address some of the problems with Picard's reasoning and address the issue of distinctive vowel length in historical French in more detail.

One problematic aspect of Picard's analysis has to do with his failure to distinguish between vowel length as a derived feature and as an underlying, phonemic contrast. The remaining problems relate to several of his key assumptions that are left unmotivated. Among these are: that length was a truly distinctive feature in the phonological system of Middle French; that any minor, non-distinctive vocalic contrasts that do exist in Canadian French (CF) mid and low vowels involve primarily quantity rather than quality; and that any such contrasts are sufficiently robust to project phoneme categories (for the distinction between contrasts and categories see, for example, Steriade (Reference Steriade, Broe and Pierrehumbert2000) and Faber, Di Paolo and Best (Reference Faber, Di Paolo and Best2007)). These assumptions preclude consideration of alternative scenarios for the existence in CF of minor vowel contrasts.

In the following section, I address the lack of differentiation between derived properties and phonemic contrasts. In Section 3, I question Picard's underlying assumption that there were phonological length distinctions from the Middle French period on (at least to the seventeenth century and the settling of Nouvelle-France). In Section 4, I advance alternative scenarios for the existence of minor, residual vocalic contrasts in present-day CF mid and low vowels, one assuming the more likely possibility that they are really quality contrasts, with length as a secondary phonetic feature, the other assuming the less likely possibility that they really are length contrasts. These scenarios involve well-understood linguistic principles and are perfectly reasonable once we dispense with a priori assumptions unsupported by empirical evidence. In Section 5, I show that whether minor, non-functional contrasts are of quality or of quantity, one need not assume that they are reflected in the inventory of phoneme categories. Section 6 provides some concluding remarks.

2 DERIVED FEATURES VERSUS PHONEMIC CONTRASTS

Picard (Reference Picard2004) neglects the important distinction between derived features and underlying, phonemic contrasts. The latter is clearly the focus of Gess’ (2001) study. Besides appearing in the title, the word ‘distinctive’ (or ‘contrastive’ or ‘phonemic’) appears as a modifier of ‘vowel length’ (or a contextually unambiguous reference to same) at least 12 times in Gess’ (2001) short article, averaging more than once per page. In comparison, Picard (Reference Picard2004) refers to ‘contrastive vowel length’ only once outside of quotes with equivalent expressions from Gess (Reference Gess and Brinton2001) and Morin (Reference Morin2000). Moreover, his use of the term in this context (saying that it ‘has existed in Standard French (SF) almost to this day, e.g., /bɛːl/ bêle vs. /bɛl/ belle’) is problematic at best (one might even say ‘abusive’ (see discussion of Dumas (Reference Dumas, Morin, Picard, Pupier and Santerre1974) below)). The use of this example as the first pair to argue against Gess (Reference Gess and Brinton2001) is ironic, as even Martinet (Reference Martinet1945), an adherent of the view (also problematic) that length had some phonologically distinctive status in French at the time of his study, regrets the use of this pair in his survey as it ‘présente l'inconvénient que bêle est une formation onomatopéique où les sujets peuvent être tentés de conserver une voyelle anormalement longue’ (126).

In order to show that ‘all the phonemic vowels [with two exceptions] have a long and short version’ (the last word suggests conditioned variant to me), Picard (Reference Picard2004: 4–5) presents 14 pairs of words containing long and short vowels, 13 of which are clear instances of context-induced, allophonic variation:

  • vise /viːz/ vs. vie /vi/

  • jure /ʒyːʀ/ vs. jus /ʒy/

  • bouge /buːʒ/ vs. bout /bu/

  • neige /neːʒ/ vs. né /ne/

  • creuse /kʀøːz/ vs. creux /kʀø/

  • grosse /gʀoːs/ vs. gros /gʀo/

  • épaisse /epɛːs/ vs.épais /epɛ/

  • soeur /sœːʀ/ vs. seul /sœl/

  • grasse /gʀɑːs/ vs. gras /gʀɑ/

  • sainte /sẽːt/ vs. saint /sẽ/

  • défunte /defːt/ vs. défunt /def/

  • blanche /blãːʃ/ vs. blanc /blã/

  • ronde /ʀːd/ vs. rond /ʀ/

What we have here are two types of lengthened vowel that are very well known as contextually determined, phonetic phenomena, even in SF, and one type that is morphologically induced (marking the masculine/feminine distinction). Purely phonological lengthening applies to 1) vowels of any quality induced by the so-called lengthening consonants /v z ʒ ʀ/ and the cluster /vʀ/, and 2) the nasal vowels, which are inherently long in final, closed syllables, along with /ø o ɑ/ (see Walker, Reference Walker2001: 42ff). The morphologically induced lengthening can be seen as resulting from the addition of a floating moraFootnote 1 that is the feminine-marking morpheme (along with the x-slot that allows for the realization of the stem-final /s/), so that the feminine of /epɛ/ is actually as shown below, with bracketing indicating morphological structure:

The derivation in this case involves the attachment of the x-slot to the stem-final consonant (the only available landing site), and the attachment of the mora to the final vowel (the only licit landing site). The surface form of this adjective would then be as shown below:

Lengthening of precisely this type is mentioned (in far different language, of course) by Maupas (Reference Maupas1618), so that in this respect, Picard is correct in saying that ‘CF has remained remarkably conservative in regard to vowel length’ (2004: 4). However, what is crucial in all of the cases cited above is that the length in question is derived – it is not underlying, and hence not phonologically distinctive. That is, the vowel inventory of CF does not consist of short vowels and long vowels. Rather, the latter are the surface results of phonological conditioning and morphological derivation. The absence of long vowels alongside short vowels in the underlying vowel inventory of CF renders vowel length a non-distinctive property of the language.

The other long/short pair mentioned by Picard is passion /pæːsj/ vs. passif /pæsɪf/, illustrating ‘the only length alternation that does not occur in final syllables’ (2004: 5). The length in the first member of the pair might be accounted for by what Martinet refers to as ‘allongement expressif’ (1945: 95, 102, 109). In any case, as for any claim of distinctive length in any modern variety of French, the assertion should be backed up with spectrographic support showing not only the length difference (produced consistently by an adequate sampling of speakers in a methodologically sound elicitation task or, obviously more unlikely, in naturalistic speech), but also the absence of any qualitative difference, since spectral differences are known to be more salient than differences in length. In addition, perception studies (e.g., using an AX discrimination task or a standard identification task) should verify the phonological salience of vowel length distinctions. Without empirical evidence to support them, claims of vowel length in any modern variety of French are little more than empty assertions.

Nevertheless, satisfied at ‘[h]aving established that [+long] is a pervasive feature of the present-day CF vowel system’ (Picard, Reference Picard2004: 5), Picard then goes on to ‘seek to determine whether the vowel length that specifically emanates from CL, i.e., the one that Gess maintains had all but disappeared by the sixteenth century, is still productive in that dialect’ (5). In this context, he makes much of the fact that non-high vowels from ‘original CL-induced long vowels’ are ‘phonetically diphthongized’. This fact supposedly ‘fully reflects the older state of affairs’ (5). He quotes in this context Dumas’ (1981) statement that ‘le français québécois . . . utilise la diphtongaison comme mode de réalisation privilégié de la durée dans certains contextes’ (1981: 1). However, it is clear from Dumas (Reference Dumas, Morin, Picard, Pupier and Santerre1974, Reference Dumas1981) that he does not consider vowel length as a distinctive feature of CF. Besides excluding long vowels from the phonemic inventory in Dumas (Reference Dumas, Morin, Picard, Pupier and Santerre1974) (and including /ɛː/ and /aː/ only as parentheticals, along with // in Dumas (Reference Dumas1981)), Dumas (Reference Dumas, Morin, Picard, Pupier and Santerre1974) has the following to say on the subject:

[. . .] c'est seulement en diachronie profonde qu'il trouve une explication, en ce sens qu'il est clairement hérité d'un modèle de durée compensatoire, très ancien en français, qui n'est plus productif depuis longtemps mais qui s'est transmis et conservé de façon étonnamment stable en français québécois, alors qu'il est à peu près disparu du dialecte de la région parisienne, par exemple. Cette grande stabilité n'en est pas moins strictement résiduelle, et ne saurait être la source de distinctions véritables dans le lexique. Il serait pour le moins abusif de parler de valeur distinctive, même potentielle, au hasard de quelques malheureuses paires lexicales [emphasis added – RG] dont les termes, la plupart du temps, n'appartiennent pas à la même catégorie grammaticale et risquent peu de se trouver dans le même contexte syntaxique, pour ne rien dire des restrictions sémantiques contextuelles. Ainsi, ce n'est que de façon très accessoire et déjà à demi métaphorique qu'on peut parler de valeur distinctive dans le cas de paires ‘minimales’ comme les suivantes [emphasis added – RG]:

  1. (2)

même si les termes en sont toujours phonétiquement différents. S'ils ne l'étaient pas, en l'absence de cette durée acquise pour l'un des deux termes, on se retrouverait simplement en face d'un cas d'homonymie, ce qui ne serait pas très grave. (Dumas Reference Dumas, Morin, Picard, Pupier and Santerre1974: 16)

Dumas (Reference Dumas1981) likewise concludes that these two contrasts ‘n'ont pas dans la langue le statut général et largement productif des grandes oppositions vocaliques’ (4).

Dumas does suggest that length ‘doit être posé à priori dans la représentation phonologique sous-jacente des morphèmes qu'il affecte’ (Dumas, Reference Dumas, Morin, Picard, Pupier and Santerre1974: 16). However, this does not render it a phonological category-defining feature if we assume, with adherents of current exemplar-theoretic models (discussed further in Section 5), that ‘long-term representations of words include more phonetic detail than previously imagined’ (Pierrehumbert, Reference Pierrehumbert, Gussenhoven and Warner2002: 101), without that detail involving an unnecessary proliferation of phonological categories. Pierrehumbert (Reference Pierrehumbert, Gussenhoven and Warner2002) provides a nice summary of evidence in support of this assumption. In any case, in CF, we are more likely dealing with a difference in vowel quality, for which phonetic length is a secondary feature. This will be discussed more in the following section, followed by a discussion of the question of lexical representation of whichever is primary, length or quality, in Section 5.

Picard (Reference Picard2004) also points to the statement made by Walker (Reference Walker2001: 42), that ‘[h]istorically and dialectically, long vowels have played and continue to play a significant role in French’ although ‘[i]n SF [. . .] their role is significantly reduced’ (2). However, Walker clearly emphasises, throughout that section, the phonetic nature of vowel length, stating in the following sentence that ‘[l]ong vowels are largely contextually determined, and the remnants of an earlier long-short opposition are minimal’ (42). No long vowels are included in Walker's phonemic inventory (ix, 41), and the only opposition discussed is the ‘limited’ one between /ɛ/ and /ɛː/. In this regard, he states that ‘[t]he list of words with /ɛː/ varies from author to author, but normally contains some 250 forms (Walter Reference Walter1976: 126–133)’ (Walker Reference Walker2001: 45). He goes on to list what he considers to be the most notable inclusions, including the everyday favorites bêle ‘bleats’ (see my note on this onomatopoeic word above), blême ‘pallid’, and brème ‘bream’ (a sort of fish).

3 LENGTH FROM MIDDLE FRENCH ON?

An important questioned raised by Gess (Reference Gess and Brinton2001) is left unanswered by Picard (Reference Picard2004): ‘if there were truly a phonological vowel length contrast in the language in the 16th and 17th centuries, what is the explanation for the fact that it was a matter of debate among grammarians? This is totally unexpected for a truly distinctive feature of a language’ (Gess, Reference Gess and Brinton2001: 152). In fact, there are highly contradictory views on vowel length in grammars from Estienne (Reference Estienne1549) on. Several, including Matthieu (Reference Matthieu1559–Reference Matthieu1560), Lartigaut (Reference Lartigaut1669), Dupuis (Reference Dupuis1836) and Ricard (Reference Ricard1887), deny the existence of vowel length distinctions. Others simply make no mention of vowel length (Lanoue (Reference Lanoue1596) mentions six from his century who make no reference to it). Again, this is a surprising omission if length were truly a distinctive feature.

In addition to contradictory views on vowel length, those who do mention it give wildly inconsistent reports as to its nature and distribution. Some specific claims are espoused by this or that grammarian and rejected or not mentioned at all by others. This is summarised in Biedermann-Pasques (Reference Biedermann-Pasques1992: 195ff), and I will limit myself here to the works cited therein, although the riches of inconsistency would surely grow with the number of grammarians included.

Of course, the most commonly cited rule of length is the one related to etymological s or the circumflex that replaced it. According to Biedermann-Pasques (Reference Biedermann-Pasques1992: 195), Estienne (Reference Estienne1549) ties the perception of long vowels particularly to this marker, whereas eight years later, he mentions a conditioned (therefore not underlying, as discussed in the preceding section) lengthening triggered by a final ‘e caduc’:

‘Quelque fois E ne se prononce qu'à demi son & en refermant la bouche, & la syllable de deuant est longue, comme Sapíence, Iustíce, Châmbre’. (Etienne (1557: 6), as cited in Biedermann-Paques (1992: 195)

Estienne (Reference Estienne1557) also mentions a morphological length distinction marking the indicative versus subjunctive contrast (orthographically signaled by s). Nicot (Reference Nicot1606) mentions etymological length marked by s and, according to Biedermann-Pasques (Reference Biedermann-Pasques1992: 195), ‘de durées proprement françaises’. The latter is the same wording used earlier to refer to the allophonic lengthening triggered by the final ‘e caduc’. Nicot also mentions length marked by orthographic double vowels (e.g., aage, aiséement – N.B. that the latter can be considered a morphological instance) and the existence of duration, as Biedermann-Pasques (Reference Biedermann-Pasques1992: 196) puts it, ‘permettant la distinction de couples d'homonymes’. One wonders why, if they are truly distinguishable phonologically, they are referred to as homonyms in the first place.

Maupas (Reference Maupas1618) adds to the list of length distinctions the morphological length distinctions between masculine and feminine forms, and between singular and plural forms of nouns and adjectives as well as of verb endings (presumably the third person singular versus third person plural). Another novel addition by Maupas (Reference Maupas1618) is the length signaled by mute consonants after nasal vowels (Biedermann-Pasques Reference Biedermann-Pasques1992: 196). Oudin ([1632] Reference Oudin1640) adds even more:

[. . .] la règle d'allongement de formes du singulier ayant une finale caractéristique du pluriel, du type tas, refus, paix; la fonction allongeante de r suivi de e caduc final, type meure (adjectif feminine, fr. mod. mûre), de ʒ noté g(e), type liège, de z (noté s) après voyelle et devant e caduc final, type extase; l'allongement dû à la présence des consonnes doubles rr, ss, type beurre, caisse; de l mouillé suivi de e caduc final, type bataille; la fonction allongeante de yod suivi de e caduc dans les finales –aye, -oye, déjà signalée par Nicot et Poisson, type playe, essayent, voye, monnoye. (Biedermann-Pasques, Reference Biedermann-Pasques1992: 196)

It is difficult to avoid the conclusion that, at some point, there was some unspoken, competitive imperative to discover more sources of length in the language than did the previous grammarian. This conforms to the view of many that the grammarians were desperately trying to find something in French to fit the Latin and Greek models. As Citton and Wyss (Reference Citton and Wyss1989) put it, ‘[s]ouvent les linguistes de la Renaissance ont voulu trouver en français des voyelles longues et des brèves, sur le modèle des langues antiques’ (125). These authors go on to show four inconsistent views on vowel length from sixteenth-century grammarians.

It is worth pointing out here, that if all (or even most) of these sources of conditioned vowel length did exist, it renders the case for truly distinctive, underlying length all the more implausible. If a learner is faced with a multitude of surface long vowels, from a variety of sources, the viability of survival of an underlying distinction would be tenuous. The principal reason for this is that the more sources there are for derived length, the more overlap there will be in the environments in which derived and underlying length occur, so that it would be more difficult for the learner to reconstruct which instances are underlying rather than derived. This would lead to a smaller pool of clearly underlying distinctions, rendering the distinction less robust in terms of functional load. Distinctions could be maintained by other means, of course; namely, for example, by adding qualitative aspects to the vowel which might then take over the contrastive role of quantity. We turn to the issue of quality versus quantity in the following section.

4 ALTERNATIVE PLAUSIBLE SCENARIOS FOR MINOR CONTRASTS

4.1 Quality, not quantity

Let us assume, based on the findings of Gess (Reference Gess and Brinton2001), that ‘the lengthening that accompanied the deletion of syllable-final /S/ was distinctive through at least the 14th century’, that there was, ‘in the latter half of the 15th century [. . .] an instability in the perception of length’, and that there was ‘an obvious, cataclysmic loss of distinctive vowel length in the following century’ (151). Now it is obviously not the case that the loss of distinctive vowel length happened overnight (Gess (Reference Gess and Brinton2001) is referring here to 100-year units), or that it necessarily applied to all vowel qualities equally or at the same time. It is well known, for example, that there is a near universal correlation between vowel height and intrinsic vowel duration, with lower vowels being longer for physiological reasons (see Keating, Reference Keating1984: 37–39). This could have a retarding effect on the loss of vowel length on the lower vowels, perhaps even resulting in minor, residual length distinctions being maintained in those.

However, the most plausible scenario is that the collapse of distinctive vowel length resulted in the reinterpretation of length differences in mid and low vowels as quality differences. This is hardly a novel idea. For example, Martinet (Reference Martinet1959) postulates that the distinction between [a] and [aː], [ø] and [øː], and [ɔ] and [ɔː] eventually became a distinction between [a] and [ɑ], [ø] and [œ], and [ɔ] and [o]. While I agree with this, I disagree sharply with Martinet's assumption that the length distinctions were still intact at the end of the seventeenth century. This assumption, developed fully in Martinet (Reference Martinet1959: 155–167), rests entirely on the highly problematic work of Vaudelin (Reference Vaudelin1713). Its description of the speech of women who frequented the court and the salons (Cohen, Reference Cohen1946: 3) was ignored by grammarians of the time, including Thurot and Brunot (Cohen, (Reference Cohen1946: 7), Martinet (Reference Martinet1959: 156)), and was carried out by a man with dubious credentials in matters linguistic (Cohen, Reference Cohen1946: 2–12). Nevertheless, a generous reading of Vaudelin (Reference Vaudelin1713) shows at most morphologically derived length and phonologically derived length (both of them variable (‘arbitraire’ according to Vaudelin himself (Cohen, Reference Cohen1946: 4))), not underlying, phonologically distinctive length of the type we are talking about here (phonogically distinctive features are never arbitrarily realised).

I also disagree with Martinet's (1959) assumption that the distinction between [ɛ] and [ɛː] continued on despite the collapse of the other distinctions of this type. Rather, it is much more plausible that this distinction, like all of the others, was reinterpreted as one of quality rather than quantity, so that the new distinction (limited as it was) was between [ɛ] and [] or even [æ], the latter two likely having length as a secondary correlate. In other words, given the near universal correlation between vowel height and intrinsic vowel duration, it is likely the case that any erstwhile [ɛː] came to fall somewhere in the vowel space between [ɛ] and [æ], with length then being relegated to a feature dependent on height rather than vice versa (this would be reinforced by the absence of length on the high vowels). This provides a straightforward explanation for the low quality of the first member of [] in today's CF tête [tt] (Picard, Reference Picard2004: 5) and poêle [pwail] (Dumas, Reference Dumas, Morin, Picard, Pupier and Santerre1974: 16).

The correlation between vowel height and length can be formalized in terms of a constraint, >low ⇒ >long (‘the lower the vowel, the longer it will be’). Flemming (Reference Flemming2001: 26) addresses the question of whether functional constraints of this type are actually a part of speakers’ grammars, or whether instead they are simply external forces that impinge on it. I agree with his two principal conclusions: that even if such constraints derive ultimately from external forces ‘it does not necessarily mean that these constraints are not also represented in mental grammars’; and that ‘[m]ore definitive statements will have to await further empirical work in psycholinguistics’.

Of course there is an asymmetry in the earlier behaviour of the back vowels, so that length does not correlate with vowel height, but rather with degree of rounding. That length would correlate (as a secondary feature) with degree of rounding should be of no surprise. Such a correlation is motivated on articulatory and perceptual grounds, as a higher degree of rounding requires more articulatory effort if more time is not allowed for its production (see Kirchner (Reference Kirchner2001) and Boersma (Reference Boersma1998) for the correlation of effort and duration), and an increased duration enhances the perceptual salience of the lower second formant (F2) of more rounded vowels. The salience of degree of roundedness is reflected in the off-glide of the CF forms côte [ko͡ ut], pâte [pɑ͡ ot] (Picard, Reference Picard2004: 5), and tâche [tɑuʃ] (Dumas, Reference Dumas, Morin, Picard, Pupier and Santerre1974: 16). The lengthening of the more rounded vowels can be captured by a constraint >round ⇒ >long (‘the rounder the vowel, the longer it will be’), and the ranking of >round ⇒ >long above >low ⇒ >long.

4.2 Retention of length via prescriptivism

In the unlikely scenario in which the relevant distinctions were of length, and not of quality, there is still a plausible scenario for their maintenance. Such distinctions could have been reinforced by prescriptivist admonitions and hyper-corrections based on the orthographic marking of erstwhile length via the circumflex accent and perhaps other means (see Biedermann-Pasques, Reference Biedermann-Pasques1992: 194–202 for the various interpretations of length in the seventeenth century and their orthographic representations). Martinet recognised that orthography could have an impact on the judgments of his survey respondents (1945: 86, 90, 113, 114, 123), and it is not unreasonable to assume that such effects have been widespread and recurrent in the history of the language, at least in the literate population.Footnote 2

In principle, minor length distinctions could even reemerge later, or grow more robust, in the French of Nouvelle-France (independently of their status in European French), on the basis of the same types of pressure (prescriptivism and hypercorrection). This is theoretically possible even if the European variety had no residual length distinctions itself. Again, this process could interact with natural markedness considerations, being blocked in the high vowels, which are intrinsically shorter. In terms of phonological constraints, one could say that prescriptivism and hypercorrection would favour a constraint ‘^’ ⇒ +μ (‘a circumflex accent implies an additional mora on the associated vowel’), and that physiological considerations would favor a constraint *μ/+high (‘a mora must not dominate a high vowel’), with the latter taking precedence over the former, resulting in the following ranking:

This ranking effectively blocks hypercorrection from applying in forms with high vowels while allowing it to apply elsewhere, as can be seen in the following table:

Again, this scenario is based on well-known linguistic concepts, hypercorrection and a physiologically grounded markedness constraint, and is easily formalised in terms of constraints and constraint ranking. This scenario, for either the maintenance or reemergence of minor, residual vowel length, is plausible as applicable either to some stage of SF, with a later transmission to CF, or directly to CF.

5 WORD-SPECIFIC PHONETICS

In either of the cases discussed in the preceding section, when we are talking about minor residual distinctions of this type, whether of length or of quality, the theoretical interest is in how these minor distinctions are maintained in the grammar. Pierrehumbert (Reference Pierrehumbert, Gussenhoven and Warner2002) argues that while subphonemic detail of this type (non-contrastive length, height or both) is maintained in the long-term representation of words, only categorical information is fed into the phonological buffer that feeds the phonetic implementation module. According to the assumptions of classical exemplar theory (which was limited to speech perception, and not production), the actual production goal for a given phoneme would then be determined by ‘aggregate or average properties’ of actually produced instances (exemplars, stored in memory) of that phoneme category (Pierrehumbert Reference Pierrehumbert, Gussenhoven and Warner2002: 114). If this were correct, there would be no variability in the production of a given phoneme, and subphonemic detail could never systematically differentiate words.

Pierrehumbert therefore argues against this classical view, suggesting instead that in the general case, production goals are based on a random selection from the exemplars of a given phoneme category (bear in mind that exemplars are far more numerous towards the center of the exemplar space, around the location of what could be considered the canonical realization of the phoneme). This allows for variation in speech production. However, Pierrehumbert also assumes, based on a ‘number of cases [. . .] in which allophonic details are systematically associated with words’ (134), that there are non-general cases in which some individual words are linked systematically to more specific phonetic outcomes via a biased weighting of exemplars from among those associated with a particular phoneme category.

Let us take as a concrete example the case of the word tête in CF, associated for the sake of this argument with the percept (phoneme category) /ɛ/, assuming no separate, independent percepts /æ/ or /ɛː/. Let us assume that a speaker stores the word tête in memory, with subphonemic detail based on the way the word is actually pronounced, as ‖tæːt‖ (abstracting away, for ease of exposition, from any diphthongization, which would entail additional transition information). According to Pierrehumbert's model, what will actually be fed into the phonological buffer, which itself then feeds the phonetic implementation module, is the purely categorical string /tɛt/.Footnote 3 While the vast majority of words containing the percept /ɛ/ would randomly select from all exemplars associated with it (heavily weighted in the area of ‘pure’ [ɛ], at the centre of the exemplar space), this word tête, along with a handful of others, will instead bias the weighting of exemplars in favour of those that are lower (towards or at [ɛ̝] or [æ]) and longer (towards or at [ɛː] or [æː]).

This model has the advantages of allowing for variation in production (even words which bias a certain area of the exemplar space allow for variation within that area), and of allowing for cases of word-specific phonetics whereby certain words bias production in favour of targets that fall on the outer edge of a phoneme's exemplar space.

6 CONCLUSION

Gess (Reference Gess and Brinton2001) has sparked more controversy than I would ever have imagined. However, the debate is a healthy one for French phonology, as well as for phonology in general, and brings up some very interesting theoretical problems, the most important of which seems to center on the very notion of a phoneme or a phonologically distinctive feature. In the particular case of vowel length in French (of any current variety), the question must eventually be resolved by empirical evidence (acoustic and perceptual). Given the subtlety of the question, and the sophisticated methods and techniques available to us, reliance on a problematic historicity (revealed here and in Gess (Reference Gess, Gess and Arteaga2006)), or on studies using flawed survey methods (Gess (Reference Gess, Gess and Arteaga2006) discusses of a couple of these), simply will not suffice. Nor will studies that provide no quantification of the functional load borne by a putative length contrast.Footnote 4 If empirical evidence reliably demonstrates the phonological relevance of vowel length in current varieties of French, then claims to the contrary will stand corrected and the field will have truly established ‘facts’ to work with, as well as a fresh lens with which to interpret historical data. However, in the absence of such evidence, length contrasts in French, synchronic or diachronic varieties, cannot be considered as representing ‘fact’.

Footnotes

*

For even more on this issue the reader is referred to Gess (2006) and Morin (2006), who have strongly divergent views on the historicity of vowel length in French and, in particular, its phonological status (distinctiveness, contrastiveness, relevance and/or salience, as well as functional load). Despite the publication date, Gess (2006) was written subsequently to the current article, which was submitted to this journal in February, 2005.

1 I will follow the outcome of the last two decades of nonlinear phonology and represent phonological (underlying or derived) length by a mora, rather than by a feature [±long].

2 Since the original submission date of February, 2005, I have had the occasion to do fieldwork for the international Phonologie du Français Contemporain project, in Coquitlam, British Columbia, and I witnessed one speaker, in reading from the standard word list, stop after a word with a circumflex, correct her pronunciation, and make a circumflex-like gesture with her finger.

3 For the advantages that this assumption entails, based on the large-scale success of so-called ‘modular feedforward’ models of phonetic implementation, the reader is referred to Pierrehumbert (Reference Pierrehumbert, Gussenhoven and Warner2002: 103–106).

4 Of course the functional load criterion does not always work, as in the well-known example of /ʒ/ in English. However, this sound stands in a relation to the phoneme /ʃ/ that is symmetrical on the basis of the feature [± voice], a feature that distinguishes between many phoneme pairs in the language. The same appeal cannot be made to the feature [±long] in French.

References

REFERENCES

Biedermann-Pasques, L. (1992). Les grands courants orthographiques au XVIIe siècle et la formation de l'orthographe moderne. Tübingen: Niemeyer.CrossRefGoogle Scholar
Boersma, P. (1998). Functional Phonology. Formalizing in interactions between articulatory and perceptual drives. The Hague: Holland Academic Graphics.Google Scholar
Citton, Y. and Wyss, A. (1989). Les doctrines orthographiques du XVIe siècle en France. Geneva: Droz.Google Scholar
Cohen, M. (1946). Le français en 1700 d'après le témoignage de Gile Vaudelin. Paris: Librairie Ancienne Honoré Champion.Google Scholar
Dumas, D. (1974). Durée vocalique et diphtongaison en français québécois. In: Morin, Y.-C., Picard, M., Pupier, P., and Santerre, L. (eds.), Le français de la région de Montréal. Montréal: Les Presses de l'Université du Québec, pp. 13–55.Google Scholar
Dumas, D. (1981). Structure de la diphtongaison québécoise. Revue canadienne de linguistique, 26: 1–61.CrossRefGoogle Scholar
Dupuis, S. (1836). Traité de prononciation ou nouvelle prosodie française. Paris: Hachette.Google Scholar
Estienne, R. (1549). Dictionaire francoislatin, avtrement dict les mots francois, auec les manieres dvser diceulx, tournez en latin. Paris: R. Estienne. Reprinted 1972. Geneva: Slatkine.Google Scholar
Estienne, R. (1557). Traicte de la grammaire francoise. Paris: R. Estienne. Reprinted 1972. Geneva: Slatkine.Google Scholar
Faber, A., Di Paolo, M. and Best, C. (2007). Perceiving the unperceivable: The acquisition of near-merged forms. Unpublished manuscript.Google Scholar
Flemming, E. (2001). Scalar and categorical phenomena in a unified model of phonetics and phonology. Phonology, 18: 7–44.Google Scholar
Gess, R. (2001). Distinctive vowel length in Old French: evidence and implications. In: Brinton, L. J. (ed.), Historical Linguistics 1999. Amsterdam: Benjamins, pp. 145–156.CrossRefGoogle Scholar
Gess, R. (2006). The myth of phonologically distinctive vowel length in Renaissance French. In: Gess, R. S. and Arteaga, D. (eds.), Historical Romance Linguistics: Retrospective and Perspectives. Amsterdam: Benjamins, pp. 53–76.Google Scholar
Keating, P. A. (1984). Universal phonetics and the organization of grammars. UCLA Working Papers in Phonetics, 59: 35–49.Google Scholar
Kirchner, R. (2001). An Effort Based Approach to Lenition. New York: Routledge.Google Scholar
Lanoue, O. de (1596). Dictionnaire des rimes françoises. Geneva: Les héritiers d'Eustache Vignon.Google Scholar
Lartigaut, A. (1669). Les progrès de la véritable ortografe ou l'ortografe francêze fondée sur ses principes. Paris: Laurant Ravenau.Google Scholar
Martinet, A. (1945). La prononciation du français contemporain. Paris: Droz.Google Scholar
Martinet, A. (1959). L'évolution contemporaine du système phonologique contemporain. Free University Quarterly, 7: 1–16. Amsterdam: Free University [also in Le français sans fard (Paris: PUF, 1969), pp. 168–190].Google Scholar
Matthieu, A., sieur des Moystardières. (1559–60). Devis de la langue française, suivi du Second devis et principal propos de la langue française. Paris: Breton.Google Scholar
Maupas, C. (1618). Grammaire et syntaxe françoise contenant reigles bien exactes & certaines de la prononciation, orthographe, construction & usage de nostre langue, en faveur des estrangers qui en sont desireux. Orleans: O. Boynard et I. Nyon. Reprinted 1973. Geneva: Slatkine.Google Scholar
Morin, Y. C. (2000). La prononciation et la prosodie du français du XVIe siècle selon le témoignage de Jean-Antoine de Baïf. Langue Française, 126: 9–28.Google Scholar
Morin, Y. C. (2006). On the phonetics of rhymes in Classical and Pre-Classical French. In: Gess, R. S. and Arteaga, D. (eds.), Historical Romance Linguistics: Retrospective and Perspectives. Amsterdam: Benjamins, pp. 131–162.Google Scholar
Nicot, I. (1606). Thresor de la langue françoyse, tant ancienne que moderne. Paris: Denys Duval. Reprinted 1960. Paris: Picard.Google Scholar
Oudin, A. (1640). Grammaire françoise (2nd ed.). Paris: A. de Sommaville. Reprinted 1972. Geneva: Slatkine.Google Scholar
Picard, M. (2004). /s/-deletion in Old French and the aftermath of compensatory lengthening. Journal of French Language Studies, 14: 1–7.Google Scholar
Pierrehumbert, J. B. (2002). Word-specific phonetics. In: Gussenhoven, C. and Warner, N. (eds.), Laboratory Phonology 7. Berlin: Mouton de Gruyter, pp. 101–139.Google Scholar
Ricard, A. (1887). Système de la quantité syllabique et de l'articulation des sons graves et des aigus. Prague: Gustave Neugebauer.Google Scholar
Steriade, D. (2000). Paradigm uniformity and the phonetics-phonology boundary. In: Broe, M. B. and Pierrehumbert, J. B. (eds.), Papers in Laboratory Phonology V: Acquisition and the Lexicon. Cambridge: Cambridge University Press, pp. 313–334.Google Scholar
Vaudelin, G. (1713). Nouvelle manière d'écrire comme on parle en France. Paris.Google Scholar
Walker, D. C. (2001). French Sound Structure. Calgary: University of Calgary Press.Google Scholar
Walter, H. (1976). La dynamique des phonèmes dans le lexique français contemporain. Paris: France Expansion.Google Scholar