Partial interrogatives in Quebec French: a diachronic look at a spreading variant

This study offers a diachronic look (1840-today) into the direct partial interrogation system in Quebec French. The main goals are to provide empirical evidence of the rise of the in situ variant within the Quebec French system, and to understand how the system evolved over centuries, independently from Hexagonal French. The data (based on more than 1600 tokens) show that the emergence of the in situ variant seems to depend on the wh word itself, but its usage seems to be more important in Montreal. There are a few examples of in situ found before the 1970s, but this variant was never publicly condemned nor associated to “ bad usage ” . This article, inspired by a historical sociolinguistics approach, provides much needed original data from five different corpora, both (written) historical and (oral) contemporary, and shows that the partial interrogative system in Quebec is not stable over time. The Montreal data suggests a system evolving differently in this city than in other regions of the province, and the data from each wh word show that interrogatives with comment still only marginally allow the in situ variant.


INTRODUCTION
The French language in Quebec benefits from decades of interests by sociolinguists that aimed to describe the specificities of the oral variety, through the use of specially constructed corpora of sociolinguistic interviews (for a complete account, see Thibault, 2001).Despite this relative profusion of data, most existing corpora have not been searched for interrogatives (notable exceptions are described in Barbarie, 1982;Elsig, 2009); as a consequence, we lack a systematic, quantitative portrait of usages in Quebec.
Yet, we know that the Quebec French interrogative system, both total and partial, has some interesting properties in terms of variation, which makes it a perfect paradigm for theoretical and empirical case studies.This study proposes two objectives in order to partially fill the gap of knowledge concerning the most recent innovation of the Quebec French system, the in situ variant: i. provide a global diachronic portrait of the wh interrogative system in Quebec French, from 1840 to 2010s; ii.offer empirical evidence for the development of the in situ variant within the Quebec French partial interrogative system, which has been said to be not identical to the France one (Elsig, 2009).
We know that the in situ variant is attested in today's spoken Quebec French, as found in this corpus recorded in the 1980s in the western part of the province of Quebec (Ottawa-Hull Corpus, Poplack, 1989): (1) Que c'est que j'allais dire, tu pourrais repasser quand?(OH.115.302)(Elsig, 2009: 146) However, we do not know much about its specificities, nor its origins.In France, the wh in situ variant was extremely rare (in writing) before the 19 th century (Coveney, 2011;Farmer 2015;Guryev and Larrivée, 2021).Based on these chronological facts, this research tests the hypothesis that the in situ variant was not, in fact, part of the wh interrogative system of the first settlers coming to New France (French Canada, which will later become the Province of Quebec, where Quebec French is spoken) during the 17 th and 18 th centuries, and that its evolution is parallel and independent from what is described in France.This article aims to verify this hypothesis through the study of different types of documents throughout the 20 th century to track the appearance of the variant in the Quebec variety with the five following wh words: quand; comment; où, qui and combien.
Examining the partial interrogative system in Quebec French to confront the findings to what we know about the France partial interrogative system will not only bring extraand differentarguments to the table about the characteristics of each variant, but will also contribute to reinforcing the idea that this particular paradigm is unique in the language in so many ways.It is the perfect language element to explore language change and the spread of innovations within a specific language paradigm, across language varieties.We will show that traces of in situ can be found early in the 20 th century, but that it is only after the 1970s that we see its usage grow.
This article is divided into the following parts: section 2 briefly exposes relevant pieces of knowledge about the emergence and usage of the in situ variant in France.Section 3 is dedicated to the partial interrogative system in Quebec French; what do we know about it and how is it different from what is found in France?This section presents the results of the diachronic search done from the 19 th century on to try to understand the evolution of the system for Quebec French speakers.I will show that, as is put forward by Lauersdorf (2018), a historical sociolinguistics approach benefits from considering as many different types of data as possible, to make up for the famous 'bad data' problem (Labov, 1994: 11).The discussion section will be dedicated to exposing hypotheses about the different uses of an innovative variant in distinct French language communities.

The partial interrogative system in Quebec French
There is not a great variety of studies to know what is precisely going on with Quebec French speakers' interrogative grammar.Most studies about Quebec French interrogatives have focused on the total interrogative system, since Quebec French is quite special in having kept in a productive way the particle -tu (formerly -ti).
A few recent studies have also looked at the indirect interrogatives system, which also displays a great amount of variation.Blondeau and Ledegen (2021) show that in oral French, the in situ variant in subordinate contexts is attested in a variety of francophone areas, including in Quebec French, as early as the 1970s (Blondeau and Ledegen, 2021: 175).
(3) personne sait c'est qui (Les Parent, saison 2, épisode 3) (Blondeau and Ledegen, 2021: 173) As for direct partial interrogatives, the studies available are even scarcer.For the various forms possible in today's Quebec French, I rely mainly on Elsig's published doctoral thesis (2009), which adopts a variationist approach.Using two corpora of ethnographic and variationist interviews, one recorded in the 1940s and 1950s and one in the 1980s, Elsig described the language-internal and (socially) external factors that influenced the choice of one variant over the other.His results are still the most detailed portrait of the usage of the partial interrogative system in two Quebec French speech communities.His study is used throughout the present article as a point of comparison.
As is found in oral France French today (but contrary to what was found in a beginning of the 20 th century written corpus, see section 3.3), the standard form with inversion of the subject and the verb (QVS word order) is almost absent of Elsig's two corpora: "As the results show, [:::].Inversion does not play any role in wh-questions."(Elsig, 2009: 53, emphasis added).The majority variant in partial interrogatives in both corpora is the one with a reinforcer of the est-ce que kind, in the qu'est-ce que form (4), of course, but not only (5).
(RFQ.044.1204,Elsig, 2009: 146) This variant represents more than 70% of all variants in both corpora (Elsig, 2009: 147;149) and one form or the other of the est-ce que reinforcer is represented with all wh words (Elsig, 2009: 152).The second most used variant is the fronted wh word variant without inversion of the subject and the verb, representing around 15% of all occurrences: (6) L'union c'est supposé de t'aider, comment ça se fait qu'il y en a qui travaillent jusqu'à neuf, dix heures puis toi tu es t icitte, assis?(OH.088.924,Elsig, 2009: 52) Finally, and this is the most interesting piece of evidence for our purposes, Elsig showed that in Quebec French, just like in France (Coveney, 2002;Coveney, 2011), the in situ variant spread in usage during the 20 th century.The Quebec French results are different from the Coveney's ones though; in Elsig's corpora, the in situ variant is not playing a role nearly as important to what it plays in the France French wh interrogative system.In the 1950s corpus (only speakers born before 1900), the in situ variant represents less than 1% (7 occurrences out of 813; Elsig, 2009: 149), and in the 1980s, it represents 7.6% of the total number of occurrences (73 occurrences out of 959; Elsig, 2009: 147), making it the third variant most used.This is the only variant for which Elsig found a significant rate of use difference between his two corpora, which allows him to conclude that this variant is on the rise during the 20 th century.
The predominance of est-ce que variants in the Quebec French partial interrogation system is not surprisingit seems to be a continuation of what was found in 17 th century French, in which the est-ce que variants were very strongly associated to the partial interrogative system, despite not being majority variants at that time (Ayres-Bennett, 2004: 54-55), which will be discussed in the next subsection.As for the in situ variant, Elsig's study gives empirical evidence for its presence in the Quebec French system.I will show however that it might be underrepresented in his corpora compared to other oral corpora examined for the purpose of the present study.

The evolution of the partial interrogative system and the appearance of in situ
As mentioned in the previous section, in addition to the standard QVS variants, wh-est-ce que variants were found in 17 th century French (Ayres-Bennett, 2004: 54), but "[the] wh-interrogatives in the late Middle and Classical French data shows that there was hardly any variation" (Elsig, 2009: 162).Inversion variants (pronominal and complex) were found almost categorically in the corpus of plays and literature (15 th to 17 th centuries) that he examined (N=1309, Elsig, 2009: 159).Mathieu (2009) gives an example of in situ from Diderot during the 18 th century, and suggests a spread of the variant at that time, but he does not base his hypothesis on a systematic corpus search.
In contrast, Guryev and Larrivée (2021) manually searched for the wh in situ variant in the Frantext corpus, in texts dating from 1100 to 1840.They extracted 6750 occurrences of sentence-final interrogative sequences, and out of those, only 32 were analyzed as "true" in situ (see Guryev and Larrivée, 2021: 78-79, for details about the way they sorted the data).It is therefore a very rare structure in written texts.Guryev and Larrivée convincingly argue that the discursive context that contributed to the emergence of the in situ variant is very restricted from the beginning, appearing only in questions referring to already known information ("questions à valeur d'information explicitement ancienne" Guryev and Larrivée, 2021: 78).
Given the diachronic development of the variant, we can therefore expect to find instances of SVQ word order in Quebec French even before it becomes "an option of French grammar" (Guryev and Larrivée, 2021: 78), but it should appear in discursive contexts where a regular wh question could not be uttered.

Theoretical assumptions and methodology
In this section, I give an extensive summary of what we know about the in situ variant in Canadian French, and what this study brings in terms of empirical evidence.As was mentioned in preceding sections, we know that the in situ variant is present in today's spoken Quebec French (Barbarie, 1982;Elsig, 2009), but not to the extent that it is in France French (Adli, 2015;Guryev, 2017).This section offers details about the state of knowledge about partial interrogatives for three distinct periods: 1) before 1840 (section 3.2); 2) from 1840 to 1970 (section 3.3); 3) from 1970 to today (section 3.4).
The division between the first and second periods corresponds to the rise of metalinguistic documents during the 1840s, and the division between the second and third periods corresponds to the gathering of the first sociolinguistic oral corpora during the 1970s.To get a complete picture, various types of linguistic evidence had to be examined, since the variant under study is both new and associated to informal registers, at least in France (Coveney, 2011: 124).In historical sociolinguistics, "Given that the available data is 'imperfect' (i.e., limited, fragmentary, or incomplete), it is imperative to gather as much of it as possible for a given investigation, from all interrelated sources, linguistic and socio-historicalin other words, it is imperative to use all the data!" (Lauersdorf, 2018: 111-112).This is why we present in this section the results of our investigation that includes both 'real' usage data from written and oral corpora, but also data from metalinguistic documents and commentaries.The adopted approach is global: despite the sociolinguistic angle, the aim of the present study is to get a global picture of the interrogation usage in Quebec French, diachronically.The exact value of each variant as well as specific sociohistorical factors may be overlooked, but we hope to make up for the lack of details by giving an overview that will be as complete as possible in terms of types of sources and timespan (two centuries are covered).

Before 1840
Historical data for Quebec French before the 19 th century is scarce and varied, and it is not an easy task to distinguish, amongst the documents that are left from the New France period (17 th and 18 th centuries), which writer is Canadian and which is French (but see Martineau, 2009).The society in New France at that time was less divided and stratified than France society at the same time; in fact, a good proportion of settlers had a fairly educated and urban background (Choquette, 2001).Most of the traces that we have from the language of the time came from travelers' notes, remarks, and diaries (Caron-Leclerc, 1998), but mostly concern pronunciation and lexicon.There is therefore no systematic study of wh interrogatives for this period, and it remains to be done.For the purpose of this study, I searched four documents for any direct partial interrogatives: three private texts and one grammatical 'testimony'. 1 Two texts didn't contain any interrogatives, and the few that were found in the other two were inversion variants.In fact, very few direct questions are found in letters, memoires and diaries from the period, and as far as our research went, there are no traces of the in situ variant in New France.

1840-1970
If the in situ variant followed a similar pathway in Quebec/Canadian French as it did in France French, this period should correspond to the appearance of the variant in various texts, and its spread during the 20 th century.I will show that in Canada, the evidence for the spread of the variant is not as strong as what is found in France (see for instance Rossi-Gensane and al., 2021 on the spread of the variant in plays).Elsig (2009) mentions that it did exist, since he found seven occurrences in the 1 Two texts (from the LFFA Corpus, Martineau (dir.), 2005-) written by women, the nun Marie Morin, who wrote from 1697 to 1725 and Elizabeth Bégon, a higher-class writer who wrote letters to her son-in-law between 1748 and 1753.Our warmest gratitude to France Martineau from the University of Ottawa, who took the time to look into these two texts for us (see Martineau, 2009 for details about these texts and the period).The two other texts are the writings of a miller from the end of the 17 th century (Juneau and Poirier, 1973), and the remarks about the French language in Canada made by Père Potier (1743-1758) (Halford, 1994).
corpus that was recorded in 1940-1955 (ethnographic interviews with older rural people).Although it represents less than 1% of all partial interrogatives, it proves that the SVQ word order existed.The first direct written evidence of the existence of the variant in Quebec was found in 1922 in a newspaper language chronicle.

Metalinguistic documents and commentaries
Quebec has a strong tradition of language chronicles and commentaries, ever since the mid 19 th century (Remysen 2012).This has to do with a redefinition of the normative locus as well as a constant quest by French-Canadians for identity definition; stuck between the anglophone population who held the economic and social power and a mère-patrie with whom they had less and less in common, French-Canadian bourgeois, in particular, were trying to find their place (Martineau and Remysen, 2019).The date of 1840 was selected as a somewhat arbitrary cutoff between significant periods because the 1840s saw many descriptive and prescriptive documents published about the French language in Quebec.
Our investigation shows that during the 19 th and 20 th centuries in Quebec, the focus of descriptive and prescriptive documents when it came to nonstandard interrogatives was unequivocally the presence of the complementizer que.Out of the 16 metalinguistic documents that were surveyed, eight mentioned direct partial interrogation.These documents all focus on nonstandard constructions, and the mentions of partial interrogation are no exceptions to this: those mentions were to insist on forms that did not follow the standard wh-verb-subject (QVS) pattern.
All eight of the documents mentioning partial interrogatives focus on one or multiple form(s) of the reinforcer est-ce que.The following example was found in a glossary published in 1930: (Société du parler français au Canada, 1968Canada, [1930]]: 485).
This entry is typical of what our investigation found, although some documents also include stylistic comments about various formsin general, the wh question with reinforcer, excluding qu'est-ce que, is considered informal but typical of oral Quebec French at the time.The in situ variant is not mentioned explicitly, but we found, in a document published in 1939, the following example: (8) Pour conjuguer le verbe interrogativement on laisse le pronom devant le verbe.Ex.Je vous dois combien?Il s'appelle comment?Il va où? (Barbeau, 1939: 184 2 )   This same example is reproduced in a later book published by the same author: (9) Pour conjuguer le verbe interrogativement, on laisse le pronom devant le verbe.Ex.Il s'appelle comment lui? -Il va où ton frère?(Barbeau, 1963) Barbeau used these examples in the section about "The verb", and not in the section about interrogativeshe insists on the fact that there is no subject-verb inversion, but does not comment on the fact that the wh word is not fronted.As for the interrogation section in these two books, it only contains examples with the est-ce que reinforcer.We'd like to bring attention to the fact that in 1963 (example 9), Barbeau chose to repeat the subject after the wh element.This yields a nonstandard word order, but unmarked for oral Quebec French.It is not clear what the connection is between subject doubling and SVQ, but it might be the case that Barbeau found it more natural to use in situ with either a non-standard word order, or with a word order most typical of oral Quebec French (Auger 1994).
3.3.1.2.Language chronicles (ChroQué).In addition to metalinguistic (mostly descriptive) documents, in order to get a better idea of the stylistic value of the wh in situ if it appears, a corpus much more prescriptive in nature was surveyed: ChroQué (Verreault, Mercier and Remysen, 1998Remysen, -2011)).This corpus contains 7936 language chronicles published in newspapers from around the province of Quebec between 1865 and 1996 (see https://fdlq.recherche.usherbrooke.ca).Given the extensive size of this corpus, only the wh words où and combien were extracted.These two wh words were selected since Coveney (2002: 220) mentions that combien favours the in situ variant, and Druetta (2009) mentions that both où and combien favour the SVQ word order (see also Rossi-Gensane and al., 2021: 183).This search yielded a total of 169 occurrences, out of which six mentioned the in situ variant; five with combien (in 1922, 1929, 1960, 1967 and 1977)  The most interesting element to retain from these results is that although the aim of these chronicles was to condemn certain usages, none of the six mentions of the in situ variant were part of the condemned constructions.Of course, this investigation is not complete since we only looked at two wh words, but it still shows that despite interrogation being a topic of choice for many language chronicles during the 19 th and 20 th centuries (our search did yield a total of 169 occurrences only with où and combien), the in situ variant is very rarely found, and never condemned as "bad usage".

Usage study based on written corpora
As for actual usage during this period, I chose to investigate a corpus of correspondence (1893( -1939( , Tailleur, 2014-)-), 3 from which I extracted all wh words, in all contextsmatrix and embedded, with or without a form of the est-ce que reinforcer.The results were sorted to keep only the direct questions, the results of which are presented in this section.
The Dubuc-Palardy family corpus was chosen because it represents direct evidence from the turn of the 20 th century.The data is written and the five writers are highly educated, but the register is very intimate; members from a single family write to each other while they are away on trips, at school or at work.Previous studies done on a subset of this corpus (Tailleur and Rouillard, 2020;Tailleur and Saint-Gelais, 2018) showed that the results for the expression of the future matched results done on contemporary oral corpora, and that communication strategies typically associated to oral registers such as accommodation can be found in this corpus.This is why, in addition to the fact that it contains 1481 manuscript pages approximating to a total of 139 000 words written by five different writers from two generations, this corpus is a good source of data to examine usage of partial interrogation.
Despite what one might think about such a corpus that is not akin to a conversation, questions are reasonably common in our data.However, all question words are not equally represented, since we find an overwhelming number of comment "how", given that writers almost systematically ask about "how" their 3 The corpus Tailleur (2014-) comes from the archive center BAnQ Saguenay, Fonds Famille Dubuc (P1).The original manuscripts were numerised and diplomatically transcribed to respect not only orthography and word boundaries, but also line breaks, page breaks, etc. addressee has been, about "how" life is going, etc.The detailed results of our extraction are presented in Table 1.
This corpus does not present any variation in the use of partial questions.All members of the family used the standard variant with inversion, and the wh element used does not influence the choice of variant.
The only two variants found that did not fit this pattern are presented in ( 12) and ( 13 Example (12) is clearly an example of fronted wh without inversion, despite the fact that the writer didn't use a question mark.According to the context it might be a rhetorical question, but it definitely is a question nonetheless.The second example (13) was classified in the table as an in situ variant, but the discourse situation is definitely marked.In fact, Guryev (2021: 98) considers these kind of examples elliptical wh-constructions, in which the sequence is used to project a subsequent answer or expectation.This example is quite interesting in a letter from the beginning of the 20 th century, in which the "dialogue" is not immediate, and the writer is using a discourse strategy that clearly creates an interaction with the reader of the letter.However, even if we cannot posit that this is a true case of an interrogative in situ, we see that the SVQ order is possible for this writer, but in a very marked discourse situation.
These 9 occurrences are actually QV, since qui is the subject: « Qui choisira pour elle dans la vie [:::]?» (AMP, 1927).There are no occurrences of qui as an object, except for the 'in situ' occurrence presented in (13).
Interestingly, while the descriptive grammars and manuals contemporary to the writings of the Dubuc-Palardy family almost all described the various ways to ask a partial question with reinforcers, as seen in section 3.3.1 above, these variants are completely absent from this written corpus for direct interrogation.It therefore seems that for these (highly educated) French Canadian writers, forming interrogatives with inversion was the only acceptable interrogative form in writing (while Tailleur and Rouillard (2020) show that they all use English words and expressions in their French letters, despite the intimate but highly formal register).
These usage results do not inform us about the possible presence or spread of the in situ variant at the turn of the 20 th century in Quebec French, but the empirical implications of the absence of variation in partial interrogative usage will be briefly addressed in the Discussion section below.

1970-today
The previous sections showed that the in situ variant is somewhat attested before the arrival of major sociolinguistic oral corpora, but nothing can be said of its actual importance within the system.For the period post-1970, it was possible to focus solely on usage data, since this is the only period for which major oral corpora exist.This section provides the results of systematic extractions of wh interrogative variants from the spoken French of different regions of the province of Quebec.
Four major corpora were investigated, one recorded during the 1970s in the Estrie region of Quebec (which is about 100 km south-east of Montreal) (Beauchemin, Martel andThéorêt, 1971-1974), two corpora from Montreal (Thibault and Vincent, 1984;Vincent, Laforest and Martel (eds), 1995) (these three latter corpora are available fully transcribed through the project Fonds de données linguistiques du Québec, whose main investigator is Wim Remysen [https://fdlq.recherche.usherbrooke.ca])as well as one corpus recorded in 2012 and 2014 in a working-class neighborhood of Montreal, Hochelaga-Maisonneuve (Corpus FRAN, Martineau et al., 2011-) 5 .With this 30 years timespan between the two corpora (and Elsig's (2009) most recent corpus being in between), I am hoping to find out if the in situ variant 1) is present in the interrogative system of the speakers; and 2) is taking on more importance within the system as we get closer to today's.
All of these corpora consist of sociolinguistic interviews with speakers of various ages, males and females, and are hundreds of thousands of words in size.The Hochelaga-Maisonneuve corpus also contains one ecological interview (four speakers between the ages of 21 and 23 years old).We extracted all occurrences involving direct interrogatives involving these five wh-words: quand (example 14), comment (example 15, excluding comment ça, which means pourquoi), qui (example 16), où (example 17) and combien (example 18).( 14) Tu me rembourses quand?
(QSV) (Yannick C. 126'85, Montreal, Thibault and Vincent, 1984) We searched by wh word (and not by the interrogative punctuation, since these corpora are oral) and we kept a 30-words context before and after the occurrence.We rejected all nonverbal occurrences and for each occurrence kept, we made sure that an actual answer could be given to eliminate rhetorical questions or préannonces (Coveney, 2011).
For the Estrie corpus of the 1970s (Beauchemin andal., 1971-1974), 127 occurrences of direct questions were found, and the distribution of each variant (in raw numbers) is presented in Table 2.
We can see that the fronted wh without inversion is the majority variant, and that the standard fronted wh with inversion is virtually absent.Despite the fact that the in situ variant (SVQ) only represents 9% of all occurrences (11 out of 127), it is not equally represented through each wh words: combien alone represents 9 out of 11 occurrences of in situ.This corroborates Coveney's findings (2002: 220) that the wh word combien seems to favour the SVQ word order.The wh words où and comment favour the variant fronted without inversion (QSV), but in variation with fronted with reinforcer.As for quand and qui, the low number of tokens makes them not significant, but they both display variation.This portrait is slightly different from what Elsig found in his 1950s corpus, in which variants with est-ce que were overwhelmingly used, around 70% of the time (Elsig, 2009: 149).However, with a closer look, we notice that the wh word combien, even in Elsig's corpora, is the only one that does not favour the est-ce que variant: it is used 39% of the time as an SVQ, and only 30% of the time with est-ce que (Elsig, 2009: 157).
For the Montreal corpus recorded a decade later (Thibault and Vincent, 1984), the amount of data is more substantial: we extracted 464 occurrences of direct questions, distributed through four variants (Table 3).
Similar to the Estrie corpus, it is the fronted variant without inversion (QSV) that is the overall majority variant, but there is much variation according to the wh word.The word combien favours once again the appearance of the in situ variant, but so does où (to a slightly lesser extent).Quand is found most often used with the reinforcer (QeskSV).Finally, we notice that in this corpus, the in situ variant is used with all wh words, and at an overall rate of more than 20%.If around 1 occurrence of direct question out of five uses the in situ variant, we can no longer consider it a marginal usage (in Estrie a decade earlier, it was only 1 out of 10).
The third corpus consulted is the Montreal 1995 corpus (Corpus Montréal 1995), from which we extracted 166 direct questions of the five wh words under study.The results are presented in Table 4.
The portrait is somewhat similar to the Montreal 1984 corpus: the preferred overall variant is the QSV one, but each wh word has a different distribution.Again, the in situ variant is used with all wh words, and it is the preferred variant for où and quand (it is the second most preferred variant for both qui and combien).
The overall rate of usage of the in situ is almost exactly the same as the 1984 rate (21 % compared to 22 %), but what is significant is that it is for the first time more used than the variant with reinforcer (which is something that Elsig, 2009 didn't find in his 1989 corpus; the QeskSV was the majority variant).
Finally, for the last corpus recorded in Montreal (Corpus FRAN, Martineau et al., 2011-), 261 occurrences of direct questions were found, and the distribution of each variant is presented in Table 5.
The portrait is quite different from what was found 40 years before: the in situ variant is now the second most used variant with an overall use of 42%, not far behind the fronted variant without inversion (46%).The variant with reinforcer lost a lot of importance in this corpus, since it is only used with 17% of all occurrences.

Summary: a longitudinal glance
The wide survey of different corpora from the second half of the 20 th century shows that the in situ variant did indeed enter usage and started competing with the other variants during this period.In terms of language change within the partial interrogative system, it is difficult to explain the mechanisms behind this innovation.We knew, thanks to Elsig's very detailed study (2009) and through the data from metalinguistic documents since the 19 th century (section 3.3), that in vernacular Quebec French, it had been the QeskSV variants that took over the majority of usages (and comments) within the partial interrogative system.The multiplicity of est-ce que variants and the specialization of each of them for specific stylistic purposes (see Tailleur, 2013) made them almost sufficient, and Elsig's data led us to believe that other variants played a negligeable role in today's system.However, a closer look at other oral corpora and at different types of data complicates this picture: Figure 1 shows that it is only in Elsig's both corpora, 1940Elsig's both corpora, -1955 and 1980s and 1980s, that the QeskSV variant is indeed the majority one.Note that Figure 1 only includes the five wh-words under study (we excluded quoi, pourquoi and quelNP from Elsig's data).Figure 1 summarizes all usage data for 20 th century Quebec French.It is based on a total of 1,666 tokens of direct interrogatives, for five different wh-words, and it shows to what extent variant distribution differs greatly among the six corpora.Based on these data, we can no longer say that the QeskSV variant is the majority one in Quebec, since the QSV variant is the most common in all corpora, except in Elsig's Récits and Ottawa-Hull.It is hard to explain these facts, since Elsig's corpora are different from each other.To make the comparison easier, we divided the variants the same way as Elsig did, so we cannot explain the differences this way; for instance, we considered examples like (19) as in situ just like he did, despite the fact that diachronically, they would be better explained as variants of est-ce que (see Tailleur 2013).
(19) Si il se produit pas là, c'est quand qu'il va se produire?(Roland M._83'84, Mtl, Thibault and Vincent, 1984) We therefore suggest that the high rates of est-ce que variants in Elsig's 1940-55 corpus comes from the fact that it is a mid-20 th century corpus, and that if we believe the content of metalinguistic documents from the same period, this variant did indeed have much importance in the grammar of Quebec speakers of that generation.As for the Ottawa-Hull corpus of 1989, the explanation is not clear, and we are not sure why the rate of wh est-ce que is higher than for the Estrie corpus.Finally, it is important to point out that globally the in situ variant is more common in Montreal than in rural regions (1940-55;Elsig, 2009), in Estrie (1971-1974) or in Ottawa-Hull (1989;Elsig 2009).The three Montreal corpora also have fewer QeskSV variants than the three other corpora.
It is also worth mentioning that the correspondence data presented in section 3.3.2was excluded from Figure 1 (and from Figure 2 below), since for the partial interrogation system, the type of data is crucial and the difference in results between this written corpus and the others is too important to have a meaningful comparison the writers from the Dubuc-Palardy family used the inversion variant 98% of the time (N = 76).These results are interesting for methodological reasons, since they show that a corpus of written but intimate correspondence is truly a hybrid form of language when it comes to registers; sometimes it aligns with oral data (see Tailleur and St-Gelais, 2018 about the expression of future in the Dubuc-Palardy corpus), sometimes it aligns with standard language, such as is the case with the partial interrogative system (see Martineau, 2014 for a discussion about the use of such documents in usage studies).In contrast, the oral data presented here has a very clear pattern: the inversion variant is clearly disfavoured and marginal in all corpora (between 0 and 6% of all data).
As mentioned in the previous section, the actual wh word used seems to influence the choice of the variant.Figure 2 illustrates this in a diachronic way.
The number of tokens is not equivalent for each wh word, and of course we have more occurrences of comment than the others, given the nature of the corpus: sociolinguistic interviews are likely to contain many comment questions (Comment ça va? Comment ça a été?Comment ça s'est passé?Etc.).It is likely that the resistance of the comment wh word to the in situ variant comes from these more or less fixed expressions found in conversation (this is also noteworthy since Larrivée (2019)'s study on the pragmatic value of in situ focuses solely on comment).Figure 2 shows very clearly that the HoMa (Montreal) corpus has a higher rate of in situ, since all wh words except comment are used with the in situ variant more than 40% of the time.It also shows that each wh word has its own trajectory, and that the rate of in situ usage for all is going up over time.However, the curbs are not clean (the varied number of tokens is partially to blamefor instance, quand in the Estrie corpus has a rate of 50% usage of in situ, but it is only based on two tokens), and the tendency could be reversed, since each corpus yields its own results, despite having a single way of being collected (with the exception of Elsig's Récits, which are ethnographic interviews).These empirical and methodological considerations are discussed in the following section.

Linguistic innovations and methodological considerations
The facts outlined in the previous section show that we cannot conclude that the in situ variant came to Quebec during the New France period (17 th -18 th centuries)this is what was proposed for the particle -tu (or -ti) variant in total interrogatives, which survived in Quebec but has virtually disappeared from Europe (see Morin, 2017; among others)nor can we conclude that the variant is completely independent from the one that developed in France.The metalinguistic documents as well as usage data from the 19 th century to the end of the 20 th century point towards a certain presence of wh in situ (or at least of a construction with the SVQ word order), but no direct evidence was found that it was actually in competition with other variants of the partial interrogative system, which is quite similar to what was found in France (Rossi-Gensane 2021).The discourse situations in which the few pre-1980s examples were found seemed to be all marked, therefore not completely part of the unmarked interrogation system.This corroborates Guryev and Larrivée (2021) and Guryev (2021), who show that diachronically, this variant comes from discursively and pragmatically marked situations; it was still the case in 19 th and 20 th century Quebec French.In terms of usage, it is interesting to note that the data from this research softens the claim that "[l']interrogation est un des domaines où les variétés canadiennes et américaines se distinguent le plus du français de référence" (Coveney, 2011: [en ligne]).Indeed, for oral spontaneous speech, Coveney (2011, based on Coveney, 1996) gives between 31 and 46% rates of usage of QSV (our 3 corpora from the 1970s to the 1990s show rates between 50% and 56%), and he shows rates between 12 and 33% of in situ (our three corpora show rates between 9 and 22%).These rates are different, yes, but the corpora are not perfectly comparable, and the relative importance of each variant is more similar than we would have thought, given only Elsig's (2009) study, which showed that the QeskSV variant occupied a much greater place in the system.Similarly, Larrivée (2019: 119-121) shows how much of an increase of usage in situ underwent between the 1969-1974 corpus and the 2014 corpus of sociolinguistic interviews from Orléans that he examined (from less than 1% to 6.8 %, but these low overall rates could be explained by the fact that he only extracted comment).
Since it was shown, through a detailed review of metalinguistic documents over more than a century (section 3.3), that in situ occurrences are not very salient for users of the language (or at least are not overtly condemned) in Quebec, we can consider this variant a non-stylistically marked linguistic innovation.Coveney (2011) gives it the tag "familiar but not stigmatized", which would seem to be fitting for our data (Barbarie (1982), based on the Estrie corpus, deemed it "stylish").In the absence of a detailed study of a contemporary written corpus, however, we cannot conclude, like Coveney (2011), that this variant is associated to an oral register.
Despite not being able to test it, we believe it to be the case, since based on work presented in Tailleur (2013), the SVQ variant would be only compatible with a system lacking inversion, which is the case of oral Quebec French (data from Figure 1), but not of spontaneous written French (data from Table 1although this is not contemporary data).
In terms of methodological and empirical considerations, the question remains if sociolinguistic interviews constitute ideal data for the study of direct partial interrogation.The number of tokens is quite low, despite the considerable size of each corpus, and the representation of each wh word is not equal.However, given its status as non-stylistically marked, the in situ variant would be expected to appear in such corpora of spontaneous speech, and indeed it does.The high degree of variability exhibited by the direct partial interrogative system requires a variety of linguistic genres and registers to get the complete picture.We believe that this complete picture is still lacking for Quebec French, but that this contribution represents another step closer to it.
The evidence presented in this article allows for a proposal of two main observations, which are linked to the two objectives presented at the beginning of this article: i. the Quebec French partial interrogative might not be as different to the France one as was previously thought, since the variant with est-ce que and the in situ variant respectively had less and more importance within the global portrait presented here (contra Elsig 2009), and that the distribution of variants differs for each wh word (as was already observed for France in situ in Druetta (2009) and Coveney (2002)); ii. the in situ variant, or, more accurately, the SVQ word order was attested in Quebec at the beginning of the 20 th century, but only in marked discourse situations.The spread of the variant seems to be as recent as the end of the 20 th century and not equal in all regions, as shown in the higher rates of usage in the Montreal corpora (1984; 1995 and 2012-14).
Of course, more data (i.e.all wh words and more corpora, especially contemporary written ones) should be included in a future study, but for now data suggest that the spread of the in situ variant might be coming from Montreal.Is it an urban feature, or is it an influence of the important French population living in Montreal?This would bring us back to the contact hypothesis, but without very recent data from regions outside of Montreal, this question cannot be answered.An anonymous referee mentions the possible influence of media: this hypothesis could not be completely disregarded, of course.However, it would be surprising that media would have an influence on this feature but not on the rest of the Quebec French system; Quebec French speakers have historically had much contact with the European French variety through media, and it hasn't had much impact beyond an enhanced interdialectal intelligibility.

CONCLUSION
The wh interrogative system represents a unique case study for so many important theoretical, pragmatic and variationist questions about French.In this contribution, I showed how various types of data over a long period of time are necessary to get but a glimpse of the spread of a 'new' variant within a system.The findings from both usage and metalinguistic corpora show that the in situ variant in Quebec French seemed to start spreading only during the 1980s.Before then, the only occurrences found were in restricted and marked discursive patterns (Guryev and Larrivée, 2021).This points to a diachronic pathway almost identical to the one described for Hexagonal French (Coveney, 2002;Rossi-Gensane and al., 2021;Larrivée 2019), which goes against what was previously assumed for the Quebec French system, mainly based on Elsig (2009)'s work.Such a similarity in the two systems, from France and from Quebec, suggests that the variable status of the French partial interrogative system has in fact internal causesthere is something about the system which makes it susceptible to variation.This "something" still needs to be understood, just like the fact that different factors seem to influence each wh word's usage, resulting in a considerable difference between each wh word's variant distribution.Finally, more data is still necessary to understand the distribution of variants from region to region within Quebec, and to confirm if the spread of the variant would be a linguistic change starting from the urban area, maybe influenced by the higher rates of direct contact with Hexagonal French speaking communities (see Auger and Fournier, 2022 for work related to this new Montreal reality).

Figure 1 .
Figure 1.Global longitudinal portrait for each variant (in percentages).

Figure 2 .
Figure 2. Usage of the in situ variant for each wh word over time.

Table 1 .
Usage data: correspondence by the Dubuc-Palardy family