Jump to ContentJump to Main Navigation
Show Summary Details
More options …

Journal of Historical Sociolinguistics

Ed. by Rutten, Gijsbert / Auer, Anita / del Valle, José / Vosters, Rik / Pickl, Simon

2 Issues per year

See all formats and pricing
More options …

Searching for standard French: The construction and mining of the Recueil historique des grammaires du français

Shana Poplack / Lidia-Gabriela Jarmasz / Nathalie Dion / Nicole Rosen
Published Online: 2015-05-06 | DOI: https://doi.org/10.1515/jhsl-2015-0002


This paper describes a massive project to characterize “Standard French” by constructing and mining the Recueil historique des grammaires du français (RHGF), a corpus of grammars whose prescriptive dictates we interpret as representing the evolution of the standard over five centuries. Its originality lies in the possibility it affords to ascertain the existence of prior variability, date it, and determine the conditions under which grammarians accept or condemn variant uses. Systematic meta-analyses of the RHGF reveal that grammarians rarely acknowledge the existence of alternate ways of expressing the same thing. Instead, they adopt three major strategies to establish form-function symmetry. All involve partitioning competing variants across distinct social, semantic or linguistic contexts, despite pervasive disagreement over which variant to associate with which. This effectively factors out variability. In contrast, systematic analysis of actual language use, as instantiated in the spontaneous speech of 323 speakers of Quebec French over an apparent-time period of a century and a half, reveals robust variability, regularly conditioned by contextual elements which have never been acknowledged by grammarians. This conditioning has remained largely stable since at least the mid-nineteenth century. Taken together, these results indicate that the “rules” for variant selection promulgated by grammarians do not inform the spoken language, nor do grammars take account of the variable rules structuring spontaneous speech. As a result, grammar and usage are evolving independently.

Keywords: standard language; prescriptive grammar; language variation and change; linguistic ideology; prescription versus praxis; French

Ce qui n’est pas clair n’est pas français.

‘What isn’t clear isn’t French.’

Rivarol (1784)

1 Introduction

Like many sociolinguists whose stock-in-trade is ordinary speech, we have been struck by the number of everyday usages that are routinely labeled, by grammarians and speakers alike, as non-standard. In the particular case we deal with here, that of French spoken in Canada, these items are not restricted to the uneducated or the underprivileged, but are actually used by individuals from all walks of life. The sheer extent of these “lapses”, as many casual observers consider them, piqued our curiosity as to how, when and why they developed, whether they were changes from an earlier, “better” stage of the language, as is often implied, and what that stage might have looked like. More generally, we wanted to characterize the constitution and evolution of Standard French, so as to achieve a better understanding of what the non-standard uses are deviating from.

These are diachronic questions, but ones that pose thornier challenges than most reconstructions tackled by historical linguists. Pertinent data is particularly hard to come by. Indeed, given the heavy ideological component of the normative enterprise (e.g. Milroy and Milroy 1999; Armstrong and Mackenzie 2013), it is difficult enough to arrive at an adequate characterization of contemporary standards, let alone those of earlier times. Complicating the matter is the fact that almost all of the non-standard uses targeted are the product of inherent variability: they simply alternate with prescribed counterparts in expressing the same referential meaning or fulfilling the same grammatical function. Thus, any attempt to trace the trajectory of non-standard forms must incorporate a method for detecting the existence of prior variability. In this paper we describe a long-term project that addresses these challenges by constructing and mining a corpus of grammars, the Recueil historique des grammaires du français (‘Diachronic Compendium of French Grammars’; RHGF), whose dictates we interpret as representing the evolution of standard French over a period of five centuries.

In the popular conception, standardization is construed as the natural outcome of the normative enterprise. Such is the general faith in the invariance and immutability of the standard language that its precepts are rarely challenged. Instead, deviations, real or perceived, are often equated with change, an outcome to be avoided. The goal of our project is to problematize this state of affairs by subjecting the prescriptive injunctions constituting the standard, as instantiated by the RHGF, to systematic diachronic analysis. Adopting the working hypothesis (Poplack et al. 2002) that forms salient enough to have incited the opprobrium of grammarians were not only attested but likely widespread, we show that prescriptive injunctions can reveal the historical antecedents of much current variability. Indeed, when properly exploited, the data of the RHGF enable us to 1) infer the existence of competing variants and approximately date them, 2) discern hints of linguistic and social conditioning of variable usage, and 3) assess the trajectory and fate of the variants designated as standard. This information is in turn incorporated into analyses of contemporary usage in ways described below, allowing us to reconstruct the evolution of specific variables and trace the social history of their variants, thereby providing a check on the inference of change.

Results of our meta-analyses, described in ensuing sections, reveal surprising inconsistencies in the treatment of many important linguistic features across time, amongst grammarians, and even within the same period and grammar. Instead of the consensus generally assumed for standard language, we find pervasive indeterminacy and contradiction, to an extent hitherto unreported. Both prescriptions and proscriptions turn out to be arbitrary and labile. Moreover, much of the variability censured as non-standard in contemporary French was recognized as far back as the earliest grammars. Grammarians’ response has been to endeavour to eradicate this variability, through a number of strategies detailed in Section 8. The goal, now firmly entrenched in prescriptive (and linguistic) thought, is to (re-)establish symmetry by positing a one-to-one relationship between form and function, context, or (especially) meaning. This is reflected in the long-standing tradition of matching each variant choice with a dedicated motive. Comparisons with the way the language was actually spoken during some of the periods in which these prescriptive dictates were being issued (Section 9) show that spontaneous speech has remained virtually impervious to them. The regular, if implicit, rules constraining variability in speech are in turn unacknowledged by the prescriptive enterprise. As a result, the spoken language turns out to be far more structured, systematic and stable than the injunctions making up the standard. These findings suggest that there is not now, nor has there been, a coherent enduring entity that can objectively be qualified as Standard French.

2 The legacy of the French grammatical tradition

As one of the most codified of the world’s languages, French has been the subject of a highly developed grammatical tradition. Indeed, the very notion of grammaire is synonymous with “correct”. The concern with regulating the French language originated in officially sanctioned efforts to emancipate it from Latin, culminating in the Edict of Villers-Cotterêts of 1539. To achieve the required legitimacy, the language would need rules; apparently, the more intricate and dogmatic, the better. The objective of the first grammarians was thus to create structure out of the anarchy and variability thought to characterize French in its pre-codified stage. The earliest (sixteenth-century) works were modeled after classical grammars, forcing grammarians to fit French into Greek and Latin categories, even where not relevant (e.g. nominal declension). Conversely, French innovations, such as the article, were not acknowledged. Thus began the enduring tradition of endorsing and promoting as “correct” forms which may never actually have been used by ordinary people, while denigrating many that were. By the seventeenth century, while still attempting to “fix” the language, grammarians had shifted their benchmark from Latin to the language spoken by their aristocratic peers. An increasingly powerful grammatical machine (Denis 1949: 220) was bolstered by the creation, early in the century, of the Académie française, whose mandate, under the patronage of Louis XIII, was

…travailler avec tout le soin et toute la diligence possibles à donner des règles certaines à notre langue et à la rendre pure, éloquente et capable de traiter les arts et les sciences. (Académie française 1635: article 24)

‘to work with all possible care and diligence to endow our language with fixed rules, and render it pure, eloquent and capable of treating the arts and sciences.’

In response, constructions deemed worthy of the “honnêtes gens” ‘polite society’ were privileged, while those associated with the masses were condemned. The foundation of a “grammaire de classe” (Brunot cited in Chevalier 1994), ratifying only the usage of the “plus saine partie de la cour” ‘the finest of the courtiers’ (Vaugelas 1647), and reviling that of “la lie du peuple” ‘the dregs of society’ (Estienne 1582), was thus laid.

As Wagner (1968: 56–57) perspicaciously remarked, it is precisely when speakers have a choice between different ways of expressing the same meaning or function that norms are established. One way, typically that preferred by the governing class, is deemed standard, while others are stigmatized in an effort to obliterate them. Eradicating variability was seen as essential to the ultimate goal of perfecting the French language. By the seventeenth century, such normative efforts were considered to have paid off (e.g. Denis 1949; François 1959; Chevalier 1994). Vaugelas (1647) famously proclaimed that French had achieved a state of perfection. Henceforth, the greatest danger was change, which could only constitute corruption (e.g. Ayres-Bennett 2004), and grammarians undertook to prevent it. As far as usage is concerned, they of course failed, but the prescriptive grammar of French is considered to have remained highly conservative (e.g. Tell 1874; Denis 1949; Cohen 1954).

This deeply entrenched tradition of ratification and concomitant stigmatization, established in the very earliest works written in French and for the French, has been amply illustrated in literally thousands of grammars over the last five centuries (Tell 1874; Chevalier 1994); it continues apace today. The designation as a model for correctness of a form of language spoken at best by a small cohort of the elite set the stage for the contemporary disconnect between grammatical prescription and actual usage; it is also responsible for the great and growing gulf between written and spoken French (Poplack and Dion 2009; Poplack et al. 2013). And the practice of imbuing variant expressions with the presumed characteristics (rural, ignorant, puerile) of those who use them led to the ascription of negative social meanings to many forms. These constitute a precious backdrop to their current distribution across the community, and help elucidate pathways of change in the social arena as well.

3 Building the RHGF

3.1 Sampling

The robust grammatical enterprise sketched above resulted in vast numbers of French grammars. How can such an extensive data pool be circumscribed in a principled way? Exhaustivity is not an option, but we wanted to attain a modicum of systematicity. Our sampling procedures were dictated in the first instance by the goals of our project: to characterize the standard language on the basis of prescribed usage, to detect prior variability in actual usage, and to provide a methodical assessment of how grammarians have treated it at various points. We particularly wanted to determine which variants they judged acceptable (“standard”), and the extent to which their proclamations were consistent across grammars and persistent over time. We also sought to achieve a certain degree of representativity and authoritativeness.

Three monumental compilations of French-language grammars that together span the fourteenth century through the beginning of the twentieth century constituted our points of departure: Stengel’s (1976 [1890]) chronology of French grammars (fourteenth to eighteenth centuries), Chervel’s (1982) repertory (1800–1914) and the Archives de la linguistique française (Quemada 1974), a collection of treatises, dictionaries, and reference works prepared by academic institutions devoted to the study of French. For the twentieth century, a period barely or not at all touched upon by these works, the task was more daunting. We cross-referenced five bibliographical sources: Chervel (1982), Colombat (1998), which includes 29 entries for twentieth-century grammars, Lauwers (2004), covering 1907–1948, the extensive ELICOP (Étude LInguistique de la COmmunication Parlée) bibliography (Debrock et al. 2001) categorizing descriptive and pedagogical grammars, and Huot (1991), which summarizes grammar work done between 1870 and 1960.

Despite the impressive number of grammars some of these bibliographies cover, none lay any claim to exhaustivity. Indeed, each invokes different criteria for inclusion.1 We too eliminated on principle, from an initial sample frame comprising thousands of titles, those that could not be expected to shed light on usage, actual or prescribed. These included treatises devoted to specific topics (e.g. past participles [Toutes les difficultés des participes français, résolues en trois petites règles (Ledrut 1820)] or punctuation [Petite grammaire de la ponctuation (Pontis 1881)]), manuals on logic (Traité d’analyse logique et grammaticale [Vanier 1827]), conjugation (e.g. Tableau synoptique, contenant les dix sortes de mots ou parties du discours, le participe, la conjugaison [Leclere 1841]), etc. Grammars based on literary works (e.g. Grammaire française tirée de nos meilleurs auteurs [Boffy 1830]), or written for children (e.g. Grammaire pittoresque, pour l’amusement et l’instruction de l’enfance [Pons-Lambert 1830]), catechisms (e.g. Principes de grammaire française en forme de catéchisme [Rouxel 1812]), schoolbooks (e.g. Grammaire française élémentaire à l’usage des jeunes filles du pensionnat établi à Strasbourg par Mme Delahaye [Garnier 1801]), and miscellaneous others (e.g. Le Rudiment des dames [Galimard 1803], La Grammaire en vaudevilles, ou Lettres à Caroline sur la grammaire française [Simonin 1806]) were also excluded at the outset.

Initially retained were titles advertising a clear prescriptive component, such as L’art de bien parler françois, qui comprend tout ce qui regarde la Grammaire, & les façons de parler douteuses (de la Touche 1973 [1730]), or Manuel de la pureté du langage (Blondin 1823). We also sought some indication of attention to the spoken usage of the time, as could be inferred from titles like Le langage populaire: Grammaire, syntaxe et dictionnaire du français tel qu’on le parle dans le peuple de Paris (Bauche 1929), La grammaire des fautes (Frei 1929), or Grammaire québécoise d’aujourd’hui: comprendre les québécismes (Léard 1995). These were expected to make some acknowledgment, direct or indirect, of variability.

The linguistic focus is on grammatical alternations, since here the ideological component of prescriptivism enjoys free reign – the standard variant can be (and often is) justified as the most rational, if not the only, option dictated by logic or common sense. This led us to privilege works giving priority to morphology and syntax (declensions, tense, aspect, mood, number, gender, etc.; e.g. Grammaire du verbe français des formes aux sens: Modes, aspects, temps, auxiliaires [Leeman-Bouix 1994]), as opposed to lexicon, pronunciation or spelling (e.g. Méthode de langue française: Grammaire, Prononciation et Orthographe, Vocabulaire [Lefrançais 1923]). Requiring that both target language and meta-language of the grammar be French further narrowed the sample frame.2 We also sought to include as many influential grammars as possible.

3.2 The “incontournables”

We appealed to three lines of evidence to ascertain which works could be characterized as influential. The first came from repeated mentions in the secondary literature on the French grammatical tradition (Livet 1967 [1859]; Harnois 1928; Levitt 1968; Wagner 1968; Huot 1991; Trudeau 1992; Chevalier 1994; Colombat 1998). The grammars of Ramus (1587), Meigret (1888 [1550]), and Estienne (1569), for example, were all cited by Trudeau (1992), Chevalier (1994) and Livet (1967 [1859]), qualifying them for inclusion in the subsample of the RHGF we dubbed the incontournables ‘the unavoidables’. The second derived from explicit commentary on the status of a particular work, as in (1).


Une autre grammaire célèbre est celle de Claude Mauger. (Chevalier 1994: 47)

‘Another famous grammar is Claude Mauger’s.’


The Grammaire des grammaires [Girault-Duvivier 1848], to which Flaubert refers, was generally regarded as the most complete and authoritative of the nineteenth-century French grammars… (Levitt 1968: 17)

Number of re-editions was also taken into account. Thus, Noël and Chapsal’s (1845) Nouvelle grammaire française: sur un plan très-méthodique was included on the basis of 80 editions (Colombat 1998) since its original publication date of 1823. Grammars fulfilling one or more of the above conditions and which also correspond to our original inclusion criteria figure in the sub-corpus of incontournables.3

Though the Archives de la linguistique française (Quemada [1974]; containing microfiches of many otherwise inaccessible grammars) was an inestimable boon, the major limiting factor, as already observed decades ago in this same connection by Stengel (1976 [1890]: 3), was necessarily accessibility. Indeed, most titles in the initial sample pool had to be excluded because we simply could not obtain them. Yet another limiting factor stems from the type of systematic analysis to which we wished to subject the sources contained in the RHGF. Given the many difficulties in searchability they posed (detailed in Section 4.1), the database had to be kept manageable.

The resulting corpus differs in both its aims and its constitution from the many other grammar compilations we consulted. Their goals have been largely bibliographical, seeking to amass as exhaustive a list as possible of works published on French grammar.4 The RHGF, in contrast, is conceived as an analytical tool, with the express purposes of tracing how the prescriptive enterprise construes the standard, uncovering evidence of prior grammatical variation, and ascertaining its normative treatment over time. The ultimate goal (illustrated in Section 9) is to confront this information with actual praxis, as instantiated in everyday speech, so as to determine the success of the prescriptive enterprise in moulding ordinary usage.

3.3 Characteristics of the RHGF

The final corpus is comprised of 163 grammars published between 1530 and 1998, as depicted in Table 1.5 Although this is not a representative sample of the grammars written about the French language over the duration, the figures in the table can nonetheless be construed as roughly proportional to the number of works published in each of the five periods we identified as pertinent for linguistic analysis. The fact that three quarters of them are clustered between 1800 and 1999 reflects the estimate that “thousands” of French grammars were published during these two centuries alone (Chevalier 1994). In contrast, relatively few grammars date from the two centuries we refer to as Period I (Stengel 1976 [1890]).6 We further distinguish the latter half of the twentieth century, which coincides with the flowering of modern linguistics, as Period V. This period represents an important bellwether in the development of the doctrine of form-function symmetry and the concomitant treatment of inherent variability (Poplack and Dion 2009; Poplack et al. 2013; Section 8).

Table 1:

Distribution of grammars comprising the RHGF by period (adapted from table 1, Poplack and Dion 2009).

Nor does the RHGF lay claim to homogeneity, which in any event, would be virtually precluded simply by the size of the corpus and the time span it covers. Inevitably, some RHGF grammars adopt a more prescriptive stance, others are more descriptive, and some (Period V) grammars are even somewhat informed by linguistic theorizing. Likewise, some may target foreign learners while others are directed more towards native speakers. A single grammar may incorporate a variety of perspectives; indeed, even the author’s avowed aims for his own work may not correspond to the content (Ayres-Bennett in press). These facts in no way interfere with the pursuit of the specific project goals enunciated above, nor detract from the findings presented in ensuing sections.

Finally, the RHGF is necessarily constituted mainly of grammars originating from France, few French grammars having been published in Canada before 1850. Of the 29 RHGF grammars that were published in Canada, two thirds are reprints or reproductions of European French grammars, suggesting that a specifically Canadian French prescriptive norm distinct from a European counterpart has not been established. The few grammars that did invoke “canadianismes” generally limited their observations to pronunciation or lexis, as opposed to the morphosyntactic features of interest here. These tend to be more stable across varieties.7

4 Exploiting the RHGF

As mentioned above, the RHGF is not a bibliography, but an analytical tool. Perhaps what most distinguishes the way we exploit it is our adherence to the principle of accountability (e.g. Labov 1972): rather than assembling a smattering of selected examples, we take account not only of all pertinent mentions of the competing variants of the linguistic variable under study, but also the absence thereof, i.e. grammars that fail to acknowledge them. We define as “pertinent” those mentions that shed light on the diachronic development of standard French and its relationship to spoken usage by providing one of the following types of information:

  • a)

    attestations of variability, both direct, as in Bouhours’ acknowledgment of widespread vacillation between the auxiliaries avoir and être in compound tenses in (2), and indirect, as in (3), where the injunction against the conditional in protases of hypothetical si-clauses (Laurence 1957) implies that this form co-exists with another, accepted, variant.

  • b)

    diachronic benchmarks for the variability: the phenomena mentioned can be inferred to have been extant, and likely even current, prior to the time of writing.

  • c)

    the normative perspective on the variants involved in the variability, as emerges from Buffier’s condemnation of auxiliary avoir (and implicit ratification of auxiliary être) with the verbs in (4).

  • d)

    the social evaluation of the variants involved in the variability, as in Brunot’s qualification of the conditional in si-clauses as “common” (5).

  • e)

    hints of linguistic conditioning of variability. These may be semantic, as when doubt and desire are invoked as meanings of the subjunctive in (6), or contextual, as with Dauzat’s admission of the futurate present only in temporally disambiguated contexts (7).


Il a passé, il est passé. J’Ay veû des gens bien en peine de sçavoir lequel il faut dire. (Bouhours 1675: 384)8

He passed [conjugated with both auxiliary avoir and être]. I have seen people having a lot of trouble deciding which one to use.’


On ne dit pas et on n’écrit pas: Si tu voudrais(Laurence 1957: 153)

‘Do not say and do not write: If you would wish…’


C’est une faute de dire j’ai allé, j’ai arrivé, j’ai parti, j’ai retourné. (Buffier 1709: 266)

‘It’s a mistake to say I went, I arrived, I left, I returned [with auxiliary avoir].’


Le conditionnel [après si] ne se rencontre qu’en langue populaire. (Brunot 1965: 890)

‘The conditional [after si] is only found in common speech.’


On met le verbe au subjonctif, quand par ce verbe on veut marquer une chose qui tient du doute ou du souhait, sans affirmer absolument qu’elle est, étoit, a été, sera, seroit, ou auroit été. (de Wailly 1768: 277)

‘We put the verb in the subjunctive when by that verb we want to mark a thing that is in doubt or wished for, without absolutely asserting that it is, was, has been, will be, would be or would have been.’


Une précision d’époque, à l’aide d’un adverbe de temps, d’une indication de jour, de mois… est nécessaire pour donner au présent la valeur de futur. (Dauzat 1943: 171)

‘A specification of time, with the aid of a temporal adverb, an indication of date, month… is necessary to endow the present with the value of future.’

Readers whose acquaintance with grammars is limited to those published in recent times may not be aware of the challenges posed by systematic research on earlier materials. The oldest grammars, couched in terms translated from classical philosophy and logic, listing the accidents of the parts of speech (e.g. gender, number, and person) and invoking, among others, the espèces des verbes (‘species of verbs’), are particularly difficult to decode. Terminology was not only often strikingly different from today’s (e.g. aoriste passé for passé simple), but also highly inconsistent (e.g. the modern present conditional is alternately referred to as suppositif, optatif second imparfait or conjonctif 2e imparfait; intransitive verbs are either verbes absolus or verbes neutres). Distinctions invoked are often opaque and/or obsolete (e.g. the passé simple was also referred to as the passé défini, while passé indéfini was used for the periphrastic past form now labeled the passé composé; optatif – from the Latin mood – refers to the subjunctive in independent clauses (as in plaise au ciel ‘if it please God’), while subjunctives in subordinate clauses (often preceded by a conjunction) are termed conjonctifs.

Even later works depart from what we now conceive of as the traditional format of grammars: from the late seventeenth to the nineteenth century, entretiens or interviews (e.g. de Fenne’s [1690] Entretiens familiers pour les amateurs de la langue françoise) in the form of questions and answers were a popular way of dealing with grammatical points. Dictionnaires (Chevalier 1994), such as the Nouveau manuel de la pureté du langage, ou, Dictionnaire des difficultés de la langue française (Biscarrat 1835), listing common solécismes, barbarismes and other “mistakes” were characteristic of the late eighteenth, and especially, the nineteenth centuries. Like standard dictionaries, these works were organized by lexical item, although the criterion for inclusion was frequent “misuse” (see Ayres-Bennett and Seijido 2011 for a useful discussion). For example, a verb such as tomber, whose compound tenses are often formed with avoir, might feature its own entry condemning this non-standard usage. To ascertain the class of verbs displaying this auxiliary alternation, the entire work must be searched.

Many grammars lack clear subsections, making the location of pertinent mentions especially onerous. Even in modern grammars, information on a given phenomenon tends to be scattered over several sections rather than contained within a single one. The subjunctive, for example, is often treated not only in a dedicated section in the chapter on mood, but also under different types of subordinate clause, and in passages on different ways of conveying semantic nuances. Pertinent material on the expression of future temporal reference may be found in the section on tense (in the subparagraphs on the future and the present), as well as in the section on semi-auxiliaries (where the multiple functions of aller ‘go’ are dealt with). Sometimes no (sub)section is devoted solely to the construction of interest: only a relatively small portion of all the excerpts on the conditional deals with its non-standard use in protases of hypothetical si-clauses, for instance.

The structure of grammars, in terms of organization of information, treatment of different variables and variants, and the terms they employ to refer to them, precludes relying on a single search procedure, even within a single period. Often the same grammar had to be searched repeatedly using several methods. Wherever possible, both an index search (for keywords such as futur/futur proche, aller, présent, subjonctif, que, other subordinators, different subordinate clauses, avoir, être, conditionnel, si, etc.) and a table of contents search (for general sections on mood/tense, auxiliaries, clause types) were performed. A full scan had to be carried out on works that had neither.

Such disparate patterns of rules, exceptions and observations make the principle of accountability particularly difficult to implement. An unexpected advantage of the multiple searches they required us to perform, however, is the discovery of numerous contradictions and inconsistencies, even within the same grammar (Poplack and Dion 2009; Poplack et al. 2013; see also Section 6 below).

5 Method: coding data from grammars

Marshalling the procedures described above, each RHGF grammar is systematically examined, and all pertinent mentions are extracted. From these materials, we can categorize the preoccupations of grammarians with respect to specific variables and variants. This may require extensive content analysis, since their assessments are often proffered only in passing: see Leeman-Bouix’s aside (8), conflating the juxtaposition of si and the conditional with “ignorance”, or its equation with child language by Charaudeau (9).


La raison la plus souvent avancée pour expliquer ces emplois “déviants” [si + conditionnel] est, en dehors de l’ignorance attribuée à la personne qui parle (liée à son âge, à son milieu social), l’analogie… (Leeman-Bouix 1994: 108)

‘The reason most often advanced to explain these “deviant” uses [si + conditional] is, beyond the speaker’s ignorance (linked to his age, social class), analogy…’


Le fameux: Si j’aurais su, j’aurais pas venu, du langage enfantin, est d’une parfaite logique… (Charaudeau 1992: 474)

‘The famous: If I would have known, I wouldn’t have come, of child language, is perfectly logical…’

Categories we established include the “qualité” of the form, as in (10), the meaning(s) ascribed to it, as in (11), and the linguistic rules governing its use, whether absolute (12a), or conditioned (12b).


Cette tournure barbare [le conditionnel dans la protase] est à proscrire énergiquement. (Dauzat 1943: 279)

‘This barbaric turn of phrase [conditional in the protasis] is to be energetically proscribed.’


Il a monté, il est monté. Ces deux expressions sont bonnes; mais elles ne peuvent pas s’employer indifféremment. Le verbe avoir marque l’action, et le verbe être, l’existence ou le repos. (Molard 1810: 181)

‘These two expressions [auxiliary avoir vs. être] are good, but they cannot be used interchangeably. The verb avoir marks action, and the verb être, existence or rest.’

… il faut en François que le Verbe qui la suit [la conjonction si], soit à l’Indicatif. (Restaut 1730: 90)

‘… in French the verb that follows [si] must be in the indicative.’

Le présent … se met au lieu du futur de l’indicatif; mais alors il est toujours accompagné de quelque nom ou adverbe de temps qui marque le futur… (Vallart 1744: 237)

‘The present... is used instead of the indicative future, but then it is always accompanied by some noun or temporal adverb marking the future...’

Each pertinent mention is then coded according to the categories emerging from the content analysis. Coupled with the principle of accountability, this procedure enables a quantitative characterization of 1) the overall salience of a variable, as expressed by the proportion of grammars mentioning it, 2) the type of treatment accorded to its variant expressions, 3) the number of semantic readings and/or linguistic functions or contexts associated with each variant, and 4) their social evaluation, including degree of stigma or approval. At the same time, we can assess the consistency of variant treatment across grammars and the degree to which such treatment persists over time.

6 Grammar according to grammarians

6.1 Salience of variability

Consider first the extent to which specific grammatical alternations have preoccupied grammarians, based on the number of pertinent mentions accrued. Analysis reveals not only that different variables are differentially salient, but that the level of attention they attract may fluctuate considerably over time. Figure 1 shows, for example, that while only about half of all grammars constituting the RHGF make pertinent mentions of variability in expressing future temporal reference and tense selection in protases of si-clauses, over three quarters cite alternations involving the subjunctive and auxiliary avoir/être. Which alternations attract most normative attention? We hypothesized that variables not subject to strong opprobrium would receive less coverage in works devoted to the bon usage. But results to date indicate that salience does not necessarily correlate with stigma: choice of the conditional variant in si-clauses has provoked far more virulent condemnation than auxiliary avoir in compound tenses, for instance, yet the latter has accrued far more normative attention (Willis 2000; Poplack 2015).

Salience to RHGF grammarians of four morphosyntactic variables.
Figure 1:

Salience to RHGF grammarians of four morphosyntactic variables.

The overall percentages in Figure 1 tell only part of the story. A breakdown of pertinent mentions over time shows an abrupt increase in references to the apparently less conspicuous constructions, both of which achieve, in Periods IV and V, the same level of visibility as their more salient counterparts (Figure 2).

Pertinent mentions over time: future temporal reference and tense selection in si-clauses.
Figure 2:

Pertinent mentions over time: future temporal reference and tense selection in si-clauses.

Examination of actual usage patterns in spontaneous speech representing these periods (Poplack and Dion 2009; Poplack 2015; Section 9) reveals that these variables are sites of vigorous changes in progress, with the (colloquial) periphrastic and (non-standard) conditional variants making substantial gains at the expense of their ratified counterparts. It stands to reason that ongoing changes should be particularly salient to grammarians, yet the number that apparently escapes them is striking. For example, ongoing research shows that grammarians barely mention the choice of 3rd p. singular indefinite pronoun on ‘one’ over 1st p. plural nous ‘we’ until the twentieth century (Figure 3), although on was already virtually categorical in this context in Quebec French in the nineteenth century (Blondeau 2007), and remains so through the present, at rates ranging between 99.7% and 100%. Yet only as of 1922 do grammarians really begin to weigh in on it. Evaluations, predictably, are almost all negative (on is vulgaire, négligé, populaire; see also King et al. 2011). Likewise, alternation between indicative pluperfect (si j’avais su) and subjunctive pluperfect (si j’eusse su) in protases of past conditional complexes did not begin to accrue mentions until after 1950, a period far postdating the extinction of the latter in the spoken language (Section 8.1).

Acknowledgment of nous/on variability in the RHGF (subsample).
Figure 3:

Acknowledgment of nous/on variability in the RHGF (subsample).

Another illustration of the relationship between salience of variability to grammarians and their efforts to stamp it out comes from auxiliary avoir/être alternation in compound tenses. Ongoing analysis shows that the constitution of the “class” prescriptively designated as être verbs has varied wildly over time. Today conventionally circumscribed to about 16,9 it has historically included, at one point or another, 172 verbs. The problem (for grammarians as well as some linguists) is that the few “être” verbs that have persisted through the present are at least occasionally conjugated with avoir; some (e.g. descendre), almost uniquely so. Indeed, Willis (2000) showed that proportions of avoir selection with the 21 verbs she found to co-occur with both auxiliaries range from 8% to 98% in spoken Ottawa-Hull French (echoing Sankoff and Thibault’s [1980] findings for Montreal). This means that avoir is the usual auxiliary for two thirds of the “être” verbs. Further calculations showed for each that the higher the proportion of avoir in actual usage, the greater the number of semantic explanations (Section 8.3) offered for the alternation in the RHGF.

These results suggest that salience of variability to grammarians is labile, and likely kicks in once the disfavored variant reaches usage levels too high to ignore. We take this as a first indication of the arbitrary nature of the prescriptive enterprise in determining which of the competing variants to designate as standard.

6.2 Persistence of prescribed uses over time

Our methodology also enables us to measure the persistence of prescriptive dictates over time. The grammatical tradition, as instantiated in the RHGF, has a poor track record in this regard. Nowhere is this more clearly illustrated than through the treatment of elements prescribed to govern a subjunctive in an embedded clause. 785 such elements (including both those explicitly cited and those simply appearing in grammarians’ illustrations) have been invoked over the duration, an improbably large number. More unexpected, a breakdown of citations by period shows the maximum number of prescriptions per grammar to have risen steadily over time (Figure 4), reaching a zenith of 131 in the modern linguistics period (V).

Maximum number of governors prescribed per grammar over time.
Figure 4:

Maximum number of governors prescribed per grammar over time.

The full significance of this finding emerges when we consider that almost all subjunctive morphology actually used during this period was concentrated under only four governors (Poplack 1992; Poplack et al. 2013; Section 9.4). We also note that surprisingly few of these prescriptions have persisted over the duration (Figure 5).

Persistence of prescriptions involving subjunctive governors (adapted from figure 1, Poplack et al. 2013).
Figure 5:

Persistence of prescriptions involving subjunctive governors (adapted from figure 1, Poplack et al. 2013).

On the contrary, whether they involve verbal governors, non-verbal governors or entire semantic classes of governors, well over half of all prescriptions are idiosyncratic: prescribed in one period, never to recur. Here again, approximately two thirds of these prescriptions date from Period V. A similar result was found with respect to future temporal reference (Poplack and Dion 2009).

6.3 Consistency of prescribed uses across grammars

Another recurrent trend among the variables studied to date is the remarkable lack of consistency, both across and within grammars, over which meaning, function or context to associate with which variant expression. As is already evident from many of the preceding citations, these tend to be presented contrastively, implying that the functions are isomorphic with the forms (13).


… le futur catégorique [simple] engage un processus de certitude […] qui reste toutefois moins certain que le procès présenté par le futur périphrastique… (Leeman-Bouix 1994: 162)

‘… the categorical [synthetic] future engages a process of certainty […] which, however, remains less certain than the process presented by the periphrastic future…’

Yet Poplack and Dion (2009) found that the same function was sometimes assigned to all of the future temporal reference variants, while elsewhere the same variant was assigned contradictory functions. This was true even within a single time period and even (occasionally) the same grammar. Thus the synthetic variant has been variously characterized as both progressive and non-progressive (Dubois 1967: 117), determined and undetermined (Laveaux 1846: 684) and definite and indefinite (Silvestre de Sacy 1799: 125–126). In this context, it is perhaps not surprising that rates of inter-grammar consistency with respect to variant treatment are so low (Figure 6).

Inter-grammar agreement on non-idiosyncratic associations with future temporal reference variants (adapted from figure 2, Poplack and Dion 2009).
Figure 6:

Inter-grammar agreement on non-idiosyncratic associations with future temporal reference variants (adapted from figure 2, Poplack and Dion 2009).

Even after setting aside the idiosyncratic prescriptions (which together account for two thirds of all prescriptions), Figure 6 shows that agreement rates for the most consistent associations with the synthetic variant (neutrality) and the futurate present (proximity) reach only 13% and 32% respectively.

Results of the meta-analyses detailed in this section reveal the difficulty in predicting which variable linguistic phenomena will attract normative attention, and how volatile and inconsistent the prescriptive response is once they do.

7 The RHGF as a window on prior variability

The avowed goal of the normative enterprise is to “fix” the language, i.e. to render it invariant by reducing it to a set of categorical rules. In this context, appealing to prescriptive grammars to document variability may seem like a contradiction in terms. But prescriptivists have targeted variability since the earliest times (see also Ayres-Bennett in press; Ayres-Bennett and Seijido 2011), if only in their quest to eradicate it. Analysis of the RHGF turned up a variety of avenues. While some grammarians opt to ignore variability altogether, as does Girard, in citing only the inflected variant of future temporal reference (14), many more acknowledge it, albeit rarely directly or unconditionally, as in (2). Still, the existence of variability can almost always be inferred, if only from injunctions to eschew certain forms, as in (15) or (3). Even the simple issuing of hard-and-fast rules, as in (16), or more “reasoned” explanations for why a particular variant is unacceptable, such as semantic redundancy in (17), are tacit admissions that some contexts sometimes host more than one variant.


Lorsqu’on représente l’évenement comme devant positivement arriver dans la suite, cela fait le temps avenir; qu’on nomme FUTUR, tel qu’on le voit dans cette frase: je me donnerai de la peine; mais j’en viendrai à bout. (Girard 1982 [1747]: 20)

‘When the event is presented as definitively taking place at a time to come, that calls for the future tense, which we call [synthetic] future, as illustrated in this sentence: I will work hard; but I will prevail.’


C’est une faute d’employer le conditionnel après la conjonction si. (Carpentier 1860: 81)

‘It’s a mistake to use the conditional after the conjunction si.’


… malgré son nom, le conditionnel est exclu des propositions de condition introduites par si. (Grevisse and Goosse 1989: 260, fn 2)

‘… in spite of its name, the conditional is excluded from conditional clauses headed by si.’


Parcequ’exprimant elle même [si] la condition, ce seroit un pléonasme que de lui faire encore régir les tems conditionnels. (de Mauvillon 1754: 526)

‘Since si itself expresses a condition, it would be redundant to make it govern the conditional tenses as well.’

We found, however, that such mentions are nearly always further qualified: variants may not be used by the same types of people (18a), in the same linguistic contexts (18b), or with the same meaning (18c). In this way, the so-called “vacuous” variability that is the hallmark of spontaneous speech is, for all intents and purposes, factored out.

Noter d’ailleurs le tour populaire incorrect, mais très répandu: si je serais riche… (Gaiffe et al. 1936: 319)

‘Note the widespread but incorrect popular expression: if I would be rich…’

Le conditionnel ne s’emploie guère que dans les propositions indépendantes ou principales; il ne s’emploie jamais dans les propositions subordonnées de condition. (Cayrou et al. 1948: 291)

‘The conditional is rarely employed outside of independent or main clauses; it is never used in subordinate conditional clauses.’

D. Peut-on mettre le conditionnel présent après la conjonction si? R. Non; à moins que le premier verbe ne marque du doute. (Anonyme 1810: 110)

‘Q: Can you put the present conditional after the conjunction si? A: No, unless the first verb marks doubt.’

Indeed, the achievement of form-function symmetry requires either that one of the forms be designated “standard” and the others rejected, or each form be assigned to a distinct meaning or context. Our analyses reveal three principal strategies to achieve these goals.

First, different variants may be associated with different social groups. According to Cayrou et al., the conditional in si-protases is used only by foreigners (19a); for Martinon, it is characteristic of “popular” speech (19b); and Charaudeau relegates it to child language (9). Different registers or speech styles may also be invoked (19c). Alternatively, a variant may be restricted to a particular linguistic context. Thus, according to Dauzat (7) or Vallart (12b), when the present tense refers to future time, it must always occur with a temporal modifier. But the most pervasive strategy is to assign a distinct meaning to each variant. Thus, by labelling the synthetic form “futur absolu défini éloigné” ‘absolute definite remote future’ and the periphrastic variant “futur absolu défini prochain” ‘absolute definite proximate future’, Silvestre de Sacy (1799: 125) denies that they are just two different ways of saying that something will occur in the future, since the former would denote temporal distance while the latter would refer to proximate events.

Mais seuls les étrangers ignorant le génie de notre langue peuvent proférer si vous viendriez. (Dauzat 1943: 178)

‘But only foreigners unfamiliar with the spirit of our language could utter if you would come.’

D’ailleurs le peuple ne se prive jamais de dire si j’aurais su ou si j’aurais eu… (Martinon 1950: 362, fn 1)

‘In fact, the masses never miss a chance to say, if I would have known or if I would have had…’

Dans l’usage familier […], le conditionnel s’emploie aussi dans la proposition introduite par si. (Riegel et al. 1998: 318)

‘In informal usage […], the conditional is also used in clauses introduced by si.’

The strategies employed by grammarians are not mutually exclusive; on the contrary, all three have been invoked with respect to the same variable.10 Still, quantitative analysis shows that in the aggregate, different variables enjoy prototypical, though to all appearances, arbitrary, treatments. As illustrated in the next section, prescriptive dictates with respect to the subjunctive, for example, focus mainly on the semantic nuances ascribed to the choice between it and the indicative, while variability in the expression of future temporal reference (via synthetic, periphrastic or futurate present options) is frequently “explained” by invoking the linguistic contexts in which the variants are admitted. No RHGF injunction appeals to either kind of reasoning with respect to choice of the conditional (as opposed to the indicative) in protases of hypothetical si-complexes; instead, as has already been amply illustrated, grammarians mostly weigh in on its social connotations.

8 Prescriptive strategies for factoring out variability

8.1 Factoring out variability through social and stylistic attributions: the conditional in protases of hypothetical si-complexes

At least since 1632, when Maupas observed that the conditional “went well” with si (20), grammarians have steadfastly proscribed this form in this context (21; 15).


Quand. Lorsque. Si. Si tost que. Si est-ce-que. Vont bien avec le second imparfait [conditional], & second plusqueparfait & futur. (Maupas 1632: 186)

Si [and other conjunctions]. Go well with the conditional and the past conditional and the [synthetic] future.’

Cette conjonction [si] ne régit jamais que l’Indicatif. (de Mauvillon 1754: 525)

‘This conjunction [si] never governs anything but the indicative.’

Dans les propositions de condition introduites par si, le bon usage n’admet pas le conditionnel... (Grevisse 1998: 353)

‘In conditional clauses headed by si, good usage does not tolerate the conditional...’

Proportion of grammars providing a rule (and qualification) categorically excluding the conditional from the protasis, by period.
Figure 7:

Proportion of grammars providing a rule (and qualification) categorically excluding the conditional from the protasis, by period.

Figure 7 shows that this variable has become increasingly salient to grammarians, especially post-1950, when close to a third add further qualification (dotted line) to the prescriptive injunction (solid line) against the conditional. In contrast to so many other variable phenomena, no attempt is made to “explain” the variation on linguistic grounds.11 Instead, grammarians qualify the rule itself as “paradoxical” (Gaiffe et al. 1936: 122), “disconcerting” (Bruneau 1940: 319), “curious” (Martinon 1950: 362), and especially, as “illogical” (Bergeron 1972: 80; Grevisse and Goosse 1989: 260) and arbitrary, fruit of the famous “servitude grammaticale” ‘grammatical bondage’ (Gougenheim 1963: 190; Brunot 1965: 890; LeBidois and LeBidois 1971: 208). Rather than attempt to factor out the variability by appealing to meaning or context, the properties of the speaker are invoked. In the case of the conditional, the grammaire de classe reigns: the proscribed variant is not the province of educated adult native speakers, but only of foreigners, children, the uneducated and the masses (19). By showing how these appreciations have evolved over time, Figure 8 lays the foundation for a social history of this variant.

Social connotations of the conditional in si-protases over time.
Figure 8:

Social connotations of the conditional in si-protases over time.

Through the eighteenth century, the conditional is qualified as an error, and attributed largely to foreigners. Not until the twentieth century do grammarians begin to acknowledge that it is also used by native francophones, albeit only those of the lowest classes. During this period attributions like “erroneous” and “foreign” diminish (though they do not disappear), in favour of negative native connotations like puerile.

By way of contrast, consider the surprising amount of continuing (positive) coverage of the subjunctive pluperfect (si j’eusse su) in this context. Though this variant has long been absent from protases of past conditional complexes in speech, the proportion of grammars commenting on its alternation with the indicative counterpart paradoxically increases after 1950, as shown in Figure 9. Granted, it is increasingly qualified as “archaic” or “affected”, but more grammars persist in portraying it (rather flatteringly) as “literary”.

Social connotations of the subjunctive pluperfect in si-protases.
Figure 9:

Social connotations of the subjunctive pluperfect in si-protases.

Thus, while the choice of conditional in protases of present conditional complexes is almost universally reviled, the extinct subjunctive pluperfect is deemed to be elegant and literary in the past temporal context, right up to the latest period covered by the RHGF (when it is finally branded as “pedantic” and “archaic”). This is also one of the very few variants we have discovered that is permitted to alternate indiscriminately (i.e. with no “explanation”) with the standard indicative pluperfect form.12 The opposing treatment of these tenses is another telling illustration of the arbitrary nature of the standard.

8.2 Factoring out variability through contextual conditioning: the case of the future

The three major exponents of future temporal reference have all been attested since the earliest grammars making up the RHGF. But here again, beyond acknowledging the variant forms, only rarely do grammarians identify them as alternate expressions of the same referential meaning. A number of their attempts to factor out variability match variants with specific linguistic conditions or contexts. The most frequently invoked association is between variant and temporal distance; different variants are reserved for different degrees of proximity of the future predication. But there is little consensus as to which variant expresses which, since each has been associated with both proximate and distal futures, as illustrated in (22–24).

Le futur […] exige une date objective ou une distance avec le présent… (Léard 1995: 197)

‘The [synthetic] future […] requires an objective date or distance from the present…’

Le futur simple […] s’agit d’un moment futur, mais très proche. (Grevisse 1993: 1257)

‘The synthetic future […] refers to future time, but very proximate.’


… [aller + infinitif] marque un futur vu du présent, souvent un futur proche, parfois un futur plus lointain… (Grevisse 1993: 1192)

‘[The periphrastic variant] marks a future viewed from the present, often a proximate future, sometimes a more remote future…’


Mais il [le présent] peut aussi traduire le futur immédiat […] le présent peut traduire un futur qui peut être assez éloigné…’ (Chevalier et al. 1971: 338)

‘But it [the futurate present] can also express immediate future […] the present can express a rather distant future…’

This having been said, when associations between temporal context and variant are tallied, as in Figure 6, we find the one linking the periphrastic variant with proximity to enjoy the greatest amount of inter-grammar consensus (though still far from impressive, at 59%). This is also the only association that persists over all five periods. A variety of other elements relating to the discourse context have also been suggested to influence variant selection, albeit at far lower agreement levels. These are depicted in Figure 10.

Inter-grammar agreement on contexts of use associated with future temporal reference variants.
Figure 10:

Inter-grammar agreement on contexts of use associated with future temporal reference variants.

For the periphrastic variant, the privileged context is speech, especially informal (25), although one grammarian does acknowledge that it can also be used in writing (26). The futurate present is identified as a stylistic device to render the predication more vivid (27); many grammarians further require that it be temporally disambiguated by an adverbial ([7] and [12b]). In general, the preponderance of contexts associated to the periphrastic and present variants relative to the synthetic future leads us to further infer that the latter is the default marker of futurity, while the former are reserved for more specialized uses.


Dans la langue familière, la périphrase aller + infinitif tend à prendre la place du futur. (Baylon and Fabre 1973: 126)

‘In colloquial language, the periphrasis aller + infinitive tends to replace the [synthetic] future.’


Ce futur périphrastique […] sert aussi dans la langue écrite la plus élaborée. (Grevisse 1993: 1192)

‘This periphrastic future […] can also be used in the most elaborate written language.’


Souvent, pour rendre l’expression plus vive, plus animée, on emploie figurément le présent à la place du futur. (Bescherelle 1877: 619)

‘Often, to make the expression more lively, more animated, we use the present figuratively instead of the future.’

8.3 Factoring out variability through semantic assignments: the case of the subjunctive

Of all of the strategies for factoring out variability, the grammarian’s ace-in-the-hole resides in the meanings s/he attributes to competing variants. Nowhere is this more evident than in the treatment of the subjunctive. Despite the gargantuan effort invested in delimiting the class of lexical governors (Section 6.2 and Poplack et al. 2013), when subjunctives fail to appear in “subjunctive-selecting” contexts, or show up in presumed indicative contexts, the “explanations” proffered are invariably semantic. Regardless of other considerations, insofar as the speaker is the ultimate arbiter of what s/he wants to express, the mood s/he selected must always be appropriate. Once the grammarian has furnished an exegesis of what that was, s/he can proceed to pronounce on its acceptability. In keeping with this program, a remarkable total of 76 distinct readings have been assigned to the subjunctive variant. Predictably, they include the prototypical meanings conventionally attributed to the (Latin) subjunctive, such as doubt, desire and volition. But these are far exceeded by more elusive concepts like “psychic energy” (LeBidois and LeBidois 1971: 501), “doubtful modesty” (Fischer and Hacquard 1959: 328) and “non-judgment” (Damourette and Pichon 1930: 479), among many others. The affect, emotions and even the soul of the speaker are also frequently invoked, as in (28). Such meanings are particularly difficult to decode, let alone implement, rendering them invulnerable to challenge.

Le subjonctif exprime les dispositions de l’âme relativement à des faits… (Crouzet et al. 1912, Section 400)

‘The subjunctive expresses the dispositions of the soul relative to facts…’

Toutes les fois que la parole est comme chargée de sentiment, dans toutes les phrases qui supposent une tension et un élan de l’âme, le subjonctif a ses raisons suffisantes en soi-même. (LeBidois and LeBidois 1971: 510)

‘Whenever speech is charged with feeling, in all phrases that imply a tension and an impulse of the soul, the subjunctive is justified.’

Inexplicably (though by now predictably), however, not only the number of readings assigned, but also the proportion of grammars assigning them (Figure 11a, b) has nearly doubled in the twentieth century.

(a) Number of meanings assigned to the subjunctive by period (adapted from figure 5, Poplack et al. 2013); (b) Proportion of grammars assigning semantic readings to the subjunctive by period.
Figure 11:

(a) Number of meanings assigned to the subjunctive by period (adapted from figure 5, Poplack et al. 2013); (b) Proportion of grammars assigning semantic readings to the subjunctive by period.

And here again, only a very small minority (5%) has persisted across the entire duration (Figure 12a). Instead, nearly half of the readings are idiosyncratic, and again, most (66%) of these stem from Period V (Figure 12b).

(a) Persistence of semantic readings assigned; (b) Concentration of idiosyncratic readings by period (figure 6, Poplack et al. 2013).
Figure 12:

(a) Persistence of semantic readings assigned; (b) Concentration of idiosyncratic readings by period (figure 6, Poplack et al. 2013).

Even the most persistent of readings, displayed in Figure 13, receive inconsistent treatment: only 11 of the 76 we documented achieved so much as a 10% inter-grammar agreement rate. The frontrunner, doubt, is cited by only 1/3 of the RHGF grammars. Most of the others are mentioned by far fewer.

Semantic readings achieving an inter-grammar agreement rate of 10% or better (adapted from figure 7, Poplack et al. 2013).
Figure 13:

Semantic readings achieving an inter-grammar agreement rate of 10% or better (adapted from figure 7, Poplack et al. 2013).

Why have so many meanings been invoked for this form, why have so few recurred over time, and why have they been marked by such inconsistency? The very volatility and idiosyncrasy of the normative treatments we have documented militate against the idea that the disparities result from actual linguistic change. Rather, they derive from the goal of eradicating vacuous variability and establishing form-function symmetry. In the case of the French subjunctive, as with so many other variables, the preferred method to achieve this is by imbuing each variant with a privative meaning, whatever it may be, so long as it is distinct from that of its counterpart(s). The sheer number of readings, their instability, and their vague, contradictory, convoluted or unverifiable nature all bolster the conclusion that here, the aim of grammarians is not so much to prescribe the correct use of the subjunctive as to bring order to the perceived chaos of variable subjunctive use. The multiplication of such efforts in the modern linguistics period (Figures 5 and 11a) suggests that variability is becoming more salient and less palatable to prescriptivists. Negation of variation through semantic assignments is one way to counter it.

8.4 Summary

Summarizing, the intent of these strategies, whatever form they take, is to establish that competing expressions of a single construction do not constitute variant expressions in the technical sense of the term, either because they do not mean the same thing, because they do not occur in the same (linguistic or stylistic) context, or because they are not used by the same kinds of individuals. Assigning distinct roles to each variant restores the isomorphic relation between function and form, while implicitly rejecting the possibility of bona fide grammatical variation. The enduring legacy of this position is encapsulated in the doctrine of form-function symmetry, which continues to shape prescriptive (and many linguistic) treatments of variability through the present day.

9 Confronting prescription with praxis

A key (and ongoing) mission of this project is to assess whether, how and to what extent the model of the standard promulgated by the normative tradition affects the way French is actually used in ordinary spoken discourse. This, after all, is the express purpose of the more overtly prescriptive grammars, and an implicit goal of others. A number of avenues of investigation suggest themselves. On a global level, if prescriptive dictates affect the course of linguistic change, it seems reasonable to hypothesize that variants with a long tradition of stigma should eventually wane over time. Alternatively, we might at least expect them to be relegated to informal or vernacular registers, while those which have been endorsed as “standard” (or even enjoyed benign neglect) might be concentrated in more formal styles. Another possibility is that forms that have attracted the most prescriptive attention will display least variability in usage. At the level of specific variables, conditions cited by grammarians as permitting, requiring or prohibiting certain variants can be operationalized and tested against usage data. In the next section, we appeal to these considerations in testing the three normative strategies for factoring out variability outlined in Section 8 in speech.

9.1 Usage data

The data come from corpora representing three stages of spoken Quebec French. The Récits du français québécois d’autrefois (Poplack and St-Amand 2007), a compendium of folk tales, legends and interviews, were collected in the 1940s and 1950s (Lacourcière 1946; Roy 1981) from Quebec francophones born between 1846 and 1895. These materials can be taken to represent Quebec French as it was spoken in the second half of the nineteenth century. A twentieth-century counterpart, recorded in the early 80s, is typified by the Ottawa-Hull French Corpus (Poplack 1989). Speakers, born between 1893 and 1964, are stratified according to standard sociolinguistic factors. Twenty-five years later, part of the community sampled for the Ottawa-Hull French Corpus was revisited to build the Corpus du français en contexte: milieux scolaire et social (Poplack and Bourdages 2005). The focus was on the speech of teenagers born between 1989 and 1991, and that of their French teachers. Taken together, these three corpora afford a unique opportunity to track the way French has been spoken by a large number (N = 323) of individuals in a variety of contexts over an apparent-time (Labov 1994) span of nearly a century and a half. Though gathered at different periods and by different fieldworkers, all the data include spontaneous everyday speech, i.e. the vernaculars of their respective times. As the style least affected by self-monitoring or, crucially for these purposes, a superimposed standard, the vernacular offers the “most systematic data for linguistic analysis” (Labov 1984: 29). This makes it ideally suited to studying the way prescription affects praxis. Recall that the “Canadian” grammars in the RHGF adopt European morphosyntactic conventions wholesale (Section 3); such a comparison is thus fully warranted.

9.2 Illustration #1: the conditional in protases of si-clauses

The paradigm illustration of the gulf between prescriptive efforts and actual usage comes from the trajectory of the conditional in protases of si-clauses. We observed above that this is the most consistently decried construction of any we have studied, and remains subject to overt correction today (Poplack 2015). This variable is also subject to vigorous change in progress (Figure 14). But instead of diminishing over time in favor of the normatively endorsed imperfect, as might be expected in response to the prescriptive condemnation it has garnered, the conditional has been increasing rapidly, from an overall rate of 13% in the nineteenth century to a full 78% in the twenty-first (based on 1248 tokens of conditional complexes; Figure 14). Nor is it restricted to certain disfavored sectors of the population, as per the normative depiction (Figure 8) – on the contrary, conditional users include native francophones of all stripes, from “ignorant” to intellectual, child to adult.13

Rate of conditional selection in protases of si-clauses over time.
Figure 14:

Rate of conditional selection in protases of si-clauses over time.

Moreover, in contrast to a number of other variables, whose patterns of style shifting belie at least some sensitivity, however slight, to prescriptive norms, rates of conditional remain resolutely high, regardless of context or situation (Poplack 2015).

9.3 Illustration #2: future temporal reference

As regards linguistic conditioning, several of the contexts normatively associated with the variant expressions of the future were operationalized by Poplack and Dion (2009). Making use of multivariate analysis, they tested the effects of those and other factors in nearly 7,700 utterances making reference to future time (Table 2).

Table 2:

Multivariate analysis of the contribution of linguistic factors to the selection of the synthetic over the periphrastic variant in spoken Quebec French over time (adapted from table 8, Poplack and Dion 2009).

Two contexts prescribed by grammarians turned out to be statistically significant to variant choice, but in a direction opposed to normative trends.14 Some have associated the periphrastic future with informal (colloquial) speech, but with a probability of .49, it is actually as likely to occur in this context as not. There is a stylistic effect, but it involves the increasing association of the synthetic variant with formal speech, an effect that has more than doubled since the nineteenth century. Proximity in the future, which emerged as the single most consistent and persistent condition for variant choice in the RHGF, does not distinguish among variants in actual usage. In fact, the effect of temporal distance, barely discernible in the nineteenth century, has all but neutralized in contemporary French. Finally, the small and waning tendency for the synthetic variant to co-occur with a temporal adverbial was invoked by only one RHGF grammarian. In usage, the strongest predictor of variant choice by far is polarity of the future eventuality. The synthetic variant is almost entirely restricted to negative contexts, an effect that was already firmly in place in the nineteenth century. This too was mentioned only once (Léard 1995) in the five centuries of grammatical tradition we studied, and even this author was citing corpus-based analyses of variant use. Sociolinguistic studies were the first to detect this polarity condition, and continue to be virtually the only sources to acknowledge its primordial effect. Contrary to the implications of prescriptive commentary, it is the periphrastic variant that represents the default expression of future, especially in affirmative contexts. Here again, prescription bears little relation to praxis.

9.4 Illustration #3: subjunctive selection

We noted above that in the prescriptive view, the major determinant of subjunctive selection is semantic, but few of the 76 readings attributed to it could be operationalized and tested empirically. This is because so many of them reside in speaker intent or hearer interpretation, elements to which the analyst does not have independent access. Poplack et al. (2013) nonetheless attempted to capture some of them through intervening variables (e.g. presence of indicators of non-factual modality, reality of the predication), a number of which emerged from RHGF treatments. Large-scale quantitative analysis of 4,440 contexts in which the subjunctive could have been selected showed that any apparent semantic effect was in fact an epiphenomenon of the overriding effect of another factor: lexical identity.

Figure 15 shows that since the nineteenth century, four governors (falloir, vouloir, aimer and pour que) together have represented 66–76% of all the governors actually in use. Rates of subjunctive selection with these governors are high and rising, such that together they account for the vast majority of all subjunctive morphology.

Relative “importance” of subjunctive governors over time.
Figure 15:

Relative “importance” of subjunctive governors over time.

When these are paired with a small cohort of four highly frequent and irregular embedded verbs (aller, avoir, être and faire), those numbers skyrocket. The only other factors affecting variant choice in usage pertain to the construal of the context as a “canonical” subjunctive construction. When favorable conditions obtain (embedded verb is morphologically suppletive, que is overt and clauses are adjacent), choice of the subjunctive is favoured. Everywhere else, it is rare and receding (Poplack et al. 2013). These quantitative patterns are invisible to any but systematic quantitative analysis, so it is perhaps not surprising that nowhere in the normative tradition have they ever been acknowledged.

9.5 Summary

The brief illustrations provided in this section expose robust and regular, if untaught, community norms for variant selection. These have clearly developed independently of normative injunction. This is most evident in the decline of the prescriptively endorsed variant, observed across the board. More subtle but even more revealing is the conditioning of variant selection, which, in stark contrast to prescription, has remained stable since the nineteenth century (at least). Taken together, these results indicate that the “rules” for variant selection promulgated by grammarians do not inform the spoken language. Nor do grammars acknowledge the variable rules structuring spontaneous speech. Grammar and usage are evolving independently.15

10 Discussion

In this paper we have described a massive and ongoing project to identify, through grammarians’ injunctions, the constitution of “Standard French”. The Recueil historique des grammaires du français is the first large-scale, diachronically comprehensive corpus of French grammars which is readily exploitable for the purposes of systematic analysis. Its originality lies in the possibility it affords to ascertain the existence of prior variability, date it, and determine the conditions under which grammarians accept or condemn variant uses. The number, variety and scope of sources included, and the quantitative analyses to which we subject them, offer a broad view of the pronouncements construed to constitute the standard. Meta-analyses of prescriptions culled from 163 grammars published over four and a half centuries test the popular assumption that this variety represents both an earlier and superior stage of the language, and allow us to confront the results with actual speech over an apparent-time period of a century and a half.16 Results revealed, first, extreme volatility in grammarians’ injunctions regarding which form is “correct”, damaging to the widespread belief that the standard is a definable entity that can be readily acquired by ordinary speakers. Second, all of the non-standard variants singled out for condemnation that we have studied to date had been attested in the RHGF since the earliest times, showing that they could not be the product of recent innovation, as is often alleged. This invalidates the standard as a benchmark against which to measure change. Third, prescriptive injunctions have had little if any effect on usage that we have been able to detect. Far from stemming the tide of linguistic change, they have been powerless to stop it. Indeed, the most highly stigmatized variant we have studied is the one that is gaining most ground. Other prescribed variants (e.g. negative ne, 1st p. pl. pronoun nous) are simply eschewed by speakers, often even in their most formal registers (Poplack 2015). The contextual conditioning grammarians invoke to explain some of the observed variability was also shown to be inoperative in speech. Finally, systematic analysis of actual language use, as instantiated in the spontaneous speech of the 323 speakers we studied, showed that far from random, as the normative tradition has repeatedly implied, variability is highly structured, and current usage patterns, as instantiated by the conditioning of variability, have remained basically stable in spoken Quebec French at least since the mid-nineteenth century. This, despite huge gains in compulsory education, migration to urban centres and other social upheavals. Prescriptive injunctions about “standard” French, on the other hand, were shown to be not only arbitrary, but also inconsistent and ephemeral. We have shown that this volatility stems in large part from unwarranted associations of variant forms with meanings or contexts irrelevant to the way they are actually used – when variability is acknowledged at all. Thus is the consecration of the standard converted into exclusionary ideology encouraging the denigration of vernacular varieties. Lamentably, this trend is only increasing, to judge by the (otherwise inexplicable) proliferation of prescriptions post-1950.17 As a result, the gulf between (what is perceived as) standard usage and the actual usage of real people is profound and growing.


We are grateful to Jennifer Anderson, Carmen Leblanc and Lauren Willis for their participation in building the RHGF, and to Martine Leroux and Kimberley Miller for assistance in exploiting it. Suzanne Robillard aided in the preparation of the manuscript. We thank Wendy Ayres-Bennett, Annette Gerstenberg, Lars Hinrichs, David Sankoff, Benedikt Szmrecsanyi, Rena Torres Cacoullos, Gerard Van Herk and two anonymous readers for perspicacious commentary.


  • Académie française. 1635. Statuts et règlements. Article 24. http://www.academie-francaise.fr/sites/academie-francaise.fr/files/statuts_af.pdf (accessed 11 November 2014). 

  • Anonyme. 1810. Élémens de la langue française. Saint-Omer: Loy. Google Scholar

  • Armstrong, Nigel & Ian E. Mackenzie. 2013. Standardization, ideology and linguistics. Basingstoke: Palgrave Macmillan. Google Scholar

  • Ayres-Bennett, Wendy. 1987. Vaugelas and the development of the French language. London: Modern Humanities Research Association. Google Scholar

  • Ayres-Bennett, Wendy. 2004. Sociolinguistic variation in seventeenth-century France. Cambridge: Cambridge University Press. Google Scholar

  • Ayres-Bennett, Wendy. In press. Codification and prescription in linguistic standardisation: Myths and models. In Francesc Feliu (ed.), Constructing languages: Norms, myths and emotions. Amsterdam & Philadelphia: John Benjamins. Google Scholar

  • Ayres-Bennett, Wendy & Magali Seijido. 2011. Remarques et observations sur la langue française: histoire et évolution d’un genre. Paris: Classiques Garnier. Google Scholar

  • Bauche, Henri. 1929. Le langage populaire: grammaire, syntaxe et dictionnaire du français tel qu’on le parle dans le peuple de Paris. Paris: Payot. Google Scholar

  • Baylon, Christian & Paul Fabre. 1973. Grammaire systématique de la langue française avec des travaux pratiques d’application et leurs corrigés. Paris: Fernand. Google Scholar

  • Bergeron, Richard. 1972. Code essentiel du français parlé et écrit. Montréal: Beauchemin. Google Scholar

  • Bescherelle. 1991. L’art de conjuguer: dictionnaire de 12 000 verbes. La Salle, QC: Éditions Hurtubise HMH. Google Scholar

  • Bescherelle, Louis-Nicolas. 1877. Grammaire nationale. Paris: Garnier Frères. Google Scholar

  • Biscarrat, Félix. 1835. Nouveau manuel de la pureté du langage ou Dictionnaire des difficultés de la langue française, relativement à la prononciation, au genre des substantifs, à l’orthographe, à la syntaxe et à l’emploi des mots; et où sont signalées et corrigées les expressions et les locutions vicieuses usitées dans la conversation. Paris: Roret. Google Scholar

  • Blondeau, Hélène. 2007. L’épreuve du temps réel et la variation pronominale à la première personne du pluriel en français québécois du XIXe et XXe siècles. Verbum: Actes du colloque Phénomènes de changement en français 29(1–2). 53–64. Google Scholar

  • Blondin, Jean Noël. 1823. Manuel de la pureté du langage ou Recueil alphabétique du corrigé des barbarismes, des néologismes, des locutions vicieuses et des expressions impropres employés journellement à la Tribune, au Théâtre et dans la Société, par les personnes même qui se piquent de parler et d’écrire avec le plus de perfection la Langue française. Paris: Published by author. Google Scholar

  • Boffy, A. J. 1830. Grammaire française, tirée de nos meilleurs auteurs. Dôle: Imprimerie de Joly. Google Scholar

  • Bouhours, Dominique. 1675. Remarques nouvelles sur la langue françoise. Paris: S. Mabre. Google Scholar

  • Bruneau, Charles. 1940. Grammaire et linguistique: causeries prononcées aux postes du réseau français de la Société Radio-Canada. Montréal: Éditions Bernard Valiquette. Google Scholar

  • Brunot, Ferdinand. 1965. La pensée et la langue: méthode, principes et plan d’une théorie nouvelle du langage appliquée au français. Paris: Masson et Cie. Google Scholar

  • Buffier, Claude. 1709. Grammaire françoise sur un plan nouveau. Paris: N. Leclerc. Google Scholar

  • Carpentier, Nicolas Joseph. 1860. Dictionnaire du bon langage: contenant les difficultés de la langue française, les règles et les fautes de prononciation, les locutions vicieuses, les wallonnismes, les flandricismes, etc. Liège: L. Grandmont-Donders. Google Scholar

  • Cayrou, Gaston, Pierre Laurent & Jeanne Lods. 1948. Le français d’aujourd’hui. Paris: Colin. Google Scholar

  • Charaudeau, Patrick. 1992. Grammaire du sens et de l’expression. Paris: Hachette. Google Scholar

  • Chassang, Alexis. 1886. Nouvelle grammaire française: cours supérieur avec des notions sur l’histoire de la langue et en particulier sur les variations de la syntaxe du XVIe au XIXe siècle. Paris: Garnier. Google Scholar

  • Chervel, André. 1982. Les grammaires françaises, 1800–1914: Répertoire chronologique. Paris: Institut national de recherche pédagogique. Google Scholar

  • Chevalier, Jean-Claude. 1994. Histoire de la grammaire française. Paris: Presses Universitaires de France. Google Scholar

  • Chevalier, Jean-Claude, Claire Blanche-Benveniste, Michel Arrivé & Jean Peytard. 1971. Grammaire Larousse du français contemporain. Paris: Librairie Larousse. Google Scholar

  • Cohen, Marcel. 1954. Grammaire et style 1450–1950: Cinq cents ans de phrase française. Paris: Éditions Sociales. Google Scholar

  • Colombat, Bernard (ed.). 1998. Corpus représentatif des grammaires et des traditions linguistiques [Special issue]. Histoire, épistémologie, langage 2(1). Paris: Société d’histoire et d’épistémologie des sciences du langage. Google Scholar

  • Colombat, Bernard, Jean-Marie Fournier & Wendy Ayres-Bennett (eds.). 2011. Grand corpus des grammaires françaises, des remarques et des traités sur la langue (XIVe-XVIIe s.). Paris: Classiques Garnier Numérique. Google Scholar

  • Crouzet, Paul, Georges Berthet & Marcel Galliot. 1912. Grammaire Française Simple et Complète pour toutes les classes. Toulouse: Privat. Google Scholar

  • Damourette, Jacques & Édouard Pichon. 1930. Des mots à la pensée: essai de grammaire de la langue française. Paris: Collection des Linguistes Contemporains. Google Scholar

  • Dauzat, Albert. 1943. Le génie de la langue française. Paris: Payot. Google Scholar

  • Debrock, Mark, Piet Mertens, Fred Truyen & Veerle Brosens. 2001. Étude LInguistique de la COmmunication Parlée (ELICOP). Louvain-la-Neuve: Linguistics Department, University of Louvain. Google Scholar

  • Denis, Roland. 1949. Les vingt siècles du français. Montréal: Fides. Google Scholar

  • Dubois, Jean. 1967. Grammaire structurale du français, vol. 2: le verbe. Paris: Larousse. Google Scholar

  • Elsig, Martin & Shana Poplack. 2006. Transplanted dialects and language change: Question formation in Québec. University of Pennsylvania Working Papers in Linguistics 12(2). 77–90. Google Scholar

  • Elsig, Martin & Shana Poplack. 2009. Synchronic variation in diachronic perspective: Question formation in Québec French. In Andreas Dufter, Jürg Fleischer & Guido Seiler (eds.), Describing and modeling variation in grammar, trends in linguistics: Studies and monographs, 255–270. Berlin & New York: Mouton de Gruyter. Google Scholar

  • Estienne, Henri. 1582. Hypomneses. Paris: Honoré Champion. Google Scholar

  • Estienne, Robert. 1569. Traicté de la gramaire françoise. Paris: Robert Estienne Imprimeur du Roy. Google Scholar

  • Fenne, François de. 1690. Entretiens familiers pour les amateurs de la langue françoise. Leyde: Corneille Boutesteyn. Google Scholar

  • Fischer, Maurice & Georges Hacquard. 1959. À la découverte de la grammaire française. Paris: Librairie Hachette. Google Scholar

  • François, Alexis. 1959. Histoire de la langue française cultivée des origines à nos jours, vol. I. Genève: Alexandre Jullien. Google Scholar

  • Frei, Henri. 1929. La Grammaire des fautes. Paris: Paul Geuthner. Google Scholar

  • Gaiffe, Félix, Ernest Maille, Ernest Breuil, Simone Jahan, Léon Wagner & Madeleine Marijon. 1936. Grammaire Larousse du 20e siècle. Paris: Larousse. Google Scholar

  • Galimard, P. G. 1803. Le Rudiment des dames. Paris: Published by author. Google Scholar

  • Garnier, Jean. 1801. Grammaire française élémentaire à l’usage des jeunes filles du pensionnat établi à Strasbourg par Mme Delahaye. Strasbourg: Levrault. Google Scholar

  • Girard, abbé Gabriel. 1982 [1747]. Les vrais principes de la langue françoise. Genève: Droz. Google Scholar

  • Girault-Duvivier, Charles-Pierre. 1848. Grammaire des grammaires ou, Analyse raisonnée des meilleurs traités sur la langue française. Paris: A. Cotelle. Google Scholar

  • Gougenheim, Georges. 1963. Système grammatical de la langue française. Paris: Éditions d’Artrey. Google Scholar

  • Grevisse, Maurice. 1993. Le bon usage: grammaire française, 13th edn. (André Goosse). Louvain-la-Neuve: Duculot. Google Scholar

  • Grevisse, Maurice. 1998. Le français correct: guide pratique, 5th edn. (Michèle Lenoble-Pinson). Bruxelles: De Boeck & Larcier. Google Scholar

  • Grevisse, Maurice & André Goosse. 1989. Nouvelle grammaire française. Paris: Duculot. Google Scholar

  • Haase, Alfons. 1898. Syntaxe française du XVIIe siècle. Paris: A. Picard. Google Scholar

  • Harnois, Guy. 1928. Les Théories du Langage en France de 1660 à 1821. Paris: Les Belles Lettres. Google Scholar

  • Hinrichs, Lars, Benedikt Szmrecsanyi & Axel Bohmann. In press. Which-hunting and the standard English relative clause. LanguageGoogle Scholar

  • Huot, Hélène. 1991. La Grammaire française entre comparatisme et structuralisme 1870–1960. Paris: Armand Colin. Google Scholar

  • Imbs, Paul. 1968. L’emploi des temps verbaux en français moderne: essai de grammaire descriptive. Paris: Klincksieck. Google Scholar

  • King, Ruth, France Martineau & Raymond Mougeon. 2011. The interplay of internal and external factors in grammatical change: First-person plural pronouns in French. Language 87(3). 470–509. CrossrefGoogle Scholar

  • Labov, William. 1972. Sociolinguistic patterns. Philadephia: University of Pennsylvania Press. Google Scholar

  • Labov, William. 1984. Field methods of the project on linguistic change and variation. In John Baugh & Joel Sherzer (eds.), Language in use: Readings in sociolinguistics, 28–54. Englewood Cliffs, NJ: Prentice-Hall. Google Scholar

  • Labov, William. 1994. Principles of linguistic change, vol. I: Internal factors. Oxford: Blackwell. Google Scholar

  • Lacourcière, Luc. 1946. La langue et le folklore. Canada français (Québec) 33(7). 489–500. Google Scholar

  • Laurence, Jean-Marie. 1957. Grammaire française: grammaire raisonnée et code grammatical. Montréal: Centre de psychologie et de pédagogie. Google Scholar

  • Lauwers, Peter. 2004. La description du français entre la tradition grammaticale et la modernité linguistique: Étude historiographique et épistémologique de la grammaire française entre 1907 et 1948. Leuven: Peeters. Google Scholar

  • Laveaux, Jean-Charles. 1846. Dictionnaire raisonné des difficultés de la langue française. Paris: A. Ledentu. Google Scholar

  • Léard, Jean-Marcel. 1995. Grammaire québécoise d‘aujourd’hui: comprendre les québécismes. Montréal: Guérin universitaire. Google Scholar

  • LeBidois, Georges & Robert LeBidois. 1971. Syntaxe du français moderne: ses fondements historiques et psychologiques. Paris: Éditions A. et J. Picard. Google Scholar

  • Leclere J. A. 1841. Tableau synoptique, contenant les dix sortes de mots ou parties du discours, le participe, la conjugaison. Pithiviers: de Chenu. Google Scholar

  • Ledrut, Auguste-Louis. 1820. Toutes les difficultés des participes français, résolues en trois petites règles. Paris: Le Normant. Google Scholar

  • Leeman-Bouix, Danielle. 1994. Grammaire du verbe français des formes aux sens: Modes, aspects, temps, auxiliaires. Paris: Nathan. Google Scholar

  • Lefrançais, P. 1923. Méthode de langue française: Grammaire, Prononciation et Orthographe, Vocabulaire, 4th edn. Paris: Nathan. Google Scholar

  • Lemay, Anne. 2009. Le rôle de l’école et de la communauté dans l’acquisition de la compétence sociolinguistique par rapport à l’utilisation des pronoms forts à Gatineau: une perspective diachronique. MA mémoire: University of Ottawa. Google Scholar

  • Leroux, Martine. 2007. Something old, something new, something borrowed, something true: What are null subjects in French? Paper presented at NWAV 36, University of Pennsylvania, 11–14 October. Google Scholar

  • Levitt, Jesse. 1968. The Grammaire des Grammaires of Girault-Duvivier. The Hague & Paris: Mouton. Google Scholar

  • Livet, Charles-Louis. 1967 [1859]. La Grammaire française et les grammairiens du XVIe siècle. Genève: Slatkine Reprints. Google Scholar

  • Martinon, Philippe. 1950. Comment on parle en français: La langue parlée correcte comparée avec la langue littéraire et la langue familière. Paris: Librairie Larousse. Google Scholar

  • Maupas, Charles. 1632. Grammaire et syntaxe françoise, 3rd edn. Rouen: Jacques Cailloué. Google Scholar

  • Mauvillon, Eléazar de. 1754. Cours complet de la langue françoise distribué par exercices: à l’usage des personnes pour qui cette langue est étrangère. Dresde: G. C. Walther. Google Scholar

  • Meigret, Louis. 1888 [1550]. Le tretté de la grammere françoeze. Paris: Chréstien Wechel. Google Scholar

  • Miller, Kimberly. 2007. La possession dans le français de l’Outaouais: le rôle de l’école et de la communauté. MA mémoire: University of Ottawa. Google Scholar

  • Miller, Kimberley & Nathalie Dion. 2009. Measuring the effectiveness of proscription: Possessive prepositions in French. Paper presented at NWAV 38, University of Ottawa, 22–25 October. Google Scholar

  • Milroy, James & Lesley Milroy. 1999. Authority in language: Investigating Standard English. London: Routledge. Google Scholar

  • Molard, Etienne. 1810. Le mauvais langage corrigé ou Recueil par ordre alphabétique, d’expressions et de phrases vicieuses usitées en France, et notamment à Lyon, 4th edn. Lyon: Yvernault et Cabin. Google Scholar

  • Noël, François & Charles Pierre Chapsal. 1845. Nouvelle grammaire française: sur un plan très-méthodique. Paris: Maire-Nyon. Google Scholar

  • Palsgrave, John. 2003 [1530]. L’éclaircissement de la langue française. Paris: Champion. Google Scholar

  • Pons-Lambert, Pierre. 1830. Grammaire pittoresque, pour l’amusement et l’instruction de l’enfance. Paris: Werdet. Google Scholar

  • Pontis, P. C. 1881. Petite grammaire de la ponctuation. Paris: Hetzel. Google Scholar

  • Poplack, Shana. 1989. The care and handling of a mega-corpus. In Ralph Fasold & Deborah Schiffrin (eds.), Language change and variation, 411–451. Amsterdam: John Benjamins. Google Scholar

  • Poplack, Shana. 1992. The inherent variability of the French subjunctive. In Christiane Laeufer & Terrell A. Morgan (eds.), Theoretical analyses in Romance linguistics, 235–263. Amsterdam: John Benjamins. Google Scholar

  • Poplack, Shana. 2015. Norme prescriptive, norme communautaire et variation diaphasique. In Kirsten Kragh & Jan Lindschouw (eds.), Les variations diasystématiques dans les langues romanes et leurs interdépendances. Série TraLiRo: Société de linguistique romane. Google Scholar

  • Poplack, Shana & Johanne S. Bourdages. 2005. Corpus du français en contexte: milieux scolaire et social. Ottawa: University of Ottawa. Google Scholar

  • Poplack, Shana & Nathalie Dion. 2009. Prescription vs. praxis: The evolution of future temporal reference in French. Language 85(3). 557–587. CrossrefGoogle Scholar

  • Poplack, Shana, Allison Lealess & Nathalie Dion. 2013. The evolving grammar of the French subjunctive. [Special issue]. Probus 25(1). 139–195. Google Scholar

  • Poplack, Shana & Anne St-Amand. 2007. A real-time window on nineteenth-century vernacular French: The Récits du français québécois d’autrefois. Language in Society 36. 707–734. CrossrefGoogle Scholar

  • Poplack, Shana, Gerard Van Herk & Dawn Harvie. 2002. Deformed in the dialects: An alternative history of non-standard English. In Peter Trudgill and Richard Watts (eds.), Alternative histories of English, 87–110. London: Routledge. Google Scholar

  • Quemada, Bernard. 1974. Archives de la linguistique française: Collection de documents relatifs à la langue française publiés entre 1500 et 1900. Paris: France-Expansion. Google Scholar

  • Ramus, Petrus (Pierre La Ramée). 1587. Grammaire. Paris: Denys DuVal. Google Scholar

  • Restaut, Pierre. 1730. Principes généraux et raisonnés de la grammaire françoise. Paris: Jean Desaint. Google Scholar

  • Riegel, Martin, Jean-Christophe Pellat, René Rioul. 1998. Grammaire méthodique du français. Paris: Presses Universitaires de France. Google Scholar

  • Rivarol, Antoine de. 1929 [1784]. Discours sur l’universalité de la langue française. Publisher unknown. Google Scholar

  • Rouxel. 1812. Principes de grammaire française en forme de catéchisme par demandes et par réponses à l’usage des jeunes personnes. Milan: Silvestri. Google Scholar

  • Roy, Carmen. 1981. Littérature orale en Gaspésie, 2nd edn. Ottawa: Léméac. Google Scholar

  • Sankoff, Gillian & Pierrette Thibault. 1980. The alternation between the auxiliaries avoir and être in Montréal French. In Gillian Sankoff (ed.), The social life of language, 311–354. Philadelphia: University of Pennsylvania Press. Google Scholar

  • Silvestre de Sacy, baron Antoine Isaac. 1799. Principes de grammaire générale, mis à la portée des enfans, et propres à servir d’introduction à l’étude de toutes les langues. Paris: A. A. Lottin. Google Scholar

  • Simonin, Antoine J.B. 1806. La grammaire en vaudevilles, ou Lettres à Caroline sur la grammaire française. Paris: Barba. Google Scholar

  • Stengel, Edmund. 1976 [1890]. Chronologisches Verzeichnis französischer Grammatiken vom Ende des 14. bis zum Ausgange des 18. Jahrhunderts [Chronological index of French grammar from the late 14th to the end of the 18th centuries]. Oppeln: Georg Maske. Google Scholar

  • Tell, Julien. 1874. Les Grammairiens français depuis l’origine de la grammaire en France jusqu’aux dernières œuvres connues. Paris: Firmin Didot. Google Scholar

  • Touche, Nicolas de la. 1973 [1730]. L’art de bien parler françois, qui comprend tout ce qui regarde la grammaire & les façons de parler douteuses, 4th edn. Genève: Slatkine Reprints. Google Scholar

  • Trudeau, D. 1992. Les inventeurs du bon usage (15291647). Paris: Les Éditions de Minuit. Google Scholar

  • Vallart, abbé J. 1744. Grammaire françoise. Paris: Desaint & Saillant. Google Scholar

  • Vanier, Victor-Augustin. 1827. Traité d’analyse logique et grammaticale. Paris: Garnier. Google Scholar

  • Vaugelas, Claude-Favre de. 1647. Remarques sur la langue françoise utile à ceux qui veulent bien parler et bien escrire. Paris: Augustin Courbe. Google Scholar

  • Wagner, Robert Léon. 1968. La grammaire française. Paris: Société d’édition d’enseignement supérieur. Google Scholar

  • Wagner, Robert Léon & Jacqueline Pinchon. 1965. Grammaire du français classique et moderne. Paris: Hachette. Google Scholar

  • Wailly, Noël François de. 1768. Principes généraux et particuliers de la langue françoise, confirmés par des exemples choisis, instructifs, agréables & tirés des bons auteurs. Paris: J. Barnou. Google Scholar

  • Willis, Lauren. 2000. Être ou ne plus être: Auxiliary alternation in Ottawa-Hull French. Ottawa: University of Ottawa Masters thesis. Google Scholar


The research on which this article is based is part of a larger project entitled “Confronting prescription and praxis in the evolution of grammar”, generously supported by grants to Poplack from the Social Science and Humanities Research Council of Canada and the Killam Foundation. Poplack holds the Canada Research Chair (I) in Linguistics and Rosen, the Canada Research Chair (II) in Language Interactions.


  • 1

    Chervel (1982), for example, excludes works published outside France, Belgium, Switzerland or Quebec, works written in languages other than French and grammars of French as a foreign language, among others. 

  • 2

    We did retain Palsgrave (2003 [1530]; reputed to be the first grammar of French), originally written in English and translated into French. 

  • 3

    Some works that would qualify as influential by any measure, e.g. Le Nouveau Bescherelle: L’art de conjuguer (e.g. 1991), were not retained unless they fulfilled the inclusion criteria laid out above. 

  • 4

    Colombat (1998) marshalls the grammars in his corpus to trace the evolution of linguistic terminology, while Lauwers (2004) documents the treatment of syntax in the 50 years prior to the structuralist and generative periods. Ayres-Bennett’s concerns (e.g. Ayres-Bennett 1987; Ayres-Bennett and Seijido 2011; Colombat et al. 2011) in mining grammars for evidence of variation and change come closest to our own. 

  • 5

    All but seven grammars published prior to the twentieth century are currently available in digital format, as are approximately a third of Period IV grammars. Period V grammars must be consulted in print. 

  • 6

    As with any archival work of this sort, we cannot ascertain how many grammars may have been published and subsequently lost, particularly in the earliest stages. This is only problematic for a corpus that lays claim to exhaustivity, which the RHGF does not. 

  • 7

    Nonetheless, our corpus exploitation methods, detailed in Section 4, enable us to detect national differences in the prescriptive treatment of grammatical variables, should any exist. Of the nine variables investigated in this respect to date, we have turned up none. They would also permit us to distinguish variant treatment according to type of grammar, assuming they can be categorized objectively. 

  • 8

    Examples reproduced from the RHGF are faithful to the original orthography. Typographical conventions follow those of the journal. 

  • 9

    The “Dr. & Mrs. Vandertramp” verbs: devenir, revenir, monter, rester, sortir, venir, arriver, naître, descendre, entrer, rentrer, tomber, retourner, aller, mourir, partir. 

  • 10

    As a consequence of the contradictions detailed in Section 6. 

  • 11

    The one notable exception incarnates another major trend in the grammatical tradition: the wholesale reproduction (some might say plagiarism) of one grammarian’s exegesis by his successors. Thus Chassang (1886) “explained” Molière’s (prescriptively proscribed) use of the conditional in si vous auriez de la répugnance à me voir votre belle-mère ‘if you would hate to see me as your mother-in-law’ in L’Avare, by observing that si here actually means s’il est vrai que ‘if it is true that’, thereby (according to him) licensing a conditional. His reasoning is repeated quasi-verbatim in the grammars of Haase (1898), Wagner and Pinchon (1965), Imbs (1968) and LeBidois and LeBidois (1971). 

  • 12

    The fact that both variants are deemed standard does not explain this tolerance of variability, since in other such cases, variability is factored out through contextual or semantic conditioning (Sections 8.2 and 8.3). 

  • 13

    The notable exception is French language-arts teachers, who model their own ban on the conditional by maintaining low (6%) rates of this variant in this context (Poplack 2015). 

  • 14

    Recall that many prescribed uses are themselves contradictory; accordingly, here we invoke prescriptive trends, such as they are. 

  • 15

    We refer of course to spoken usage. Prescriptivists have been demonstrably more successful in regulating written French, as evidenced by the very presence in it (however occasional) of variants that are rare if not altogether absent in speech. 

  • 16

    Aside from those referenced here, other variables examined in this connection include direct question formation (Elsig and Poplack 2006, 2009), choice of possessive preposition (Miller 2007; Miller and Dion 2009), (null) subjects (Leroux 2007), and clitic vs. strong pronoun alternation (Lemay 2009). 

  • 17

    Interestingly, Hinrichs et al. (in press) also pinpoint the post-1950 period in assessing the effects of prescriptive discourse on English. In their interpretation, the effects of centuries of prescription begin manifesting themselves during this period, whereas here, it hosts a proliferation of prescriptions that have no palpable effect on the rise of non-standard forms. It would be interesting to explore the sociolinguistic developments contributing to these divergent outcomes. 

About the article

Published Online: 2015-05-06

Published in Print: 2015-05-01

Citation Information: Journal of Historical Sociolinguistics, Volume 1, Issue 1, Pages 13–55, ISSN (Online) 2199-2908, ISSN (Print) 2199-2894, DOI: https://doi.org/10.1515/jhsl-2015-0002.

Export Citation

©2015 by De Gruyter Mouton. Copyright Clearance Center

Citing Articles

Here you can find all Crossref-listed publications in which this article is cited. If you would like to receive automatic email messages as soon as this article is cited in other publications, simply activate the “Citation Alert” on the top of this page.

Gijsbert Rutten
Journal of Sociolinguistics, 2016, Volume 20, Number 1, Page 6

Comments (0)

Please log in or register to comment.
Log in