Corps de l’article

1. Introduction

Michif, a critically endangered Metis mixed language still spoken by an estimated few dozen to a few hundred people across the Canadian Prairie provinces and British Columbia (Mazzoli 2019), makes use of two different gender systems: the French sex-based system contrasting masculine and feminine gender, and the Algonquian animacy-based system contrasting animate with inanimate gender. Here I discuss the previous literature on Michif gender, and suggest that by broadening how we view variation in Michif gender assignment, we may be able to learn more about grammatical gender.

2. Introduction to the Michif language

The Michif language developed out of contact between French voyageurs and Indigenous peoples in the Red River valley of Manitoba, Canada, where a new language was created when the new Metis Nation was born. This new language, Michif, was created by the first half of the nineteenth century, and can be generally described as mixing French nouns with Plains Cree verbs. Although this characterization is a simplification, it does a reasonable job as an elementary description of Michif. Michif is most often described as a ‘mixed language’, which is a type of language said to form out of the formation of new ethnicities, often due to mixed marriages. They are said to mark identity rather than fill a communicative need (Bakker 1997: 12; Thomason 2003; Golovko 2003; Meakins 2013: 186). Michif falls into this category, as the Métis people are a new nation resulting from mixed marriages between (primarily) Cree women and French men, and the language was not created to fill a communicative need due to lack of common code. The Métis people were multilingual, and Michif was used as a home language with other Métis people; it was not a language used for communicating with outsiders.

The Michif language was created by the Métis people, the descendants of (primarily) French fur traders and Cree women who married in the late eighteenth and early nineteenth centuries, in what is today the Prairie provinces. The children of these unions emerged as a new identity by the early nineteenth century in the Red River Settlements in Manitoba, with a new set of traditions taken from both parents. The Métis, born bicultural and having grown up (at least) bilingual, were in a position of privilege in the changing Red River region in the mid 1850s. Many had access to a European education, but also knew the local traditional ways, where they could move seamlessly between settler and First Nations communities (Sealey and Lussier 1975; St-Onge and Podruchny 2012), and they had good relations with both. Multilingual, the Métis often served as interpreters and guides. Although most Métis people spoke First Nations languages such as Dene, Cree or Ojibwe, or colonial French and/or English in their interactions with others, they are reported to have spoken a different, in-group language at home, which has been called many things—French Cree, Chippewa Cree, not-the-real Cree, Cree spoken by the Michif —but today is usually called Michif in both scholarly (Bakker 1997; Rosen 2007; Gillon and Rosen 2018; Sammons 2019) and community publications (Rosen and Souter 2015; Gabriel Dumont Institute 2019; Fleury 2013).

3. Gender in Michif

Gender systems are normally motivated by a semantic core, i.e. there is some overlap with nouns in a particular group and some semantic feature (Corbett 2013). This semantic feature is most often related to biological sex. Sex-based systems are common in Indo-European languages such as French, Italian, German, Russian. While sex-based systems are the most common gender systems, there are also classes of languages which organize noun classes based on notions of animacy, for example the family of Algonquian languages, which have animate versus inanimate nouns. Most Algonquian languages mark nominal animacy distinctions in a number of ways, including on the verb. Note that because Michif is descended from both French and Cree, it does have both these distinctions in its grammar.

3.1. Sex-based gender in Michif

Since Michif nouns are historically derived primarily from French, it is perhaps unsurprising that there is sex-based gender marking on Michif nouns. In Michif, this marking surfaces primarily on the singular forms of articles and possessives. Note that as in French, gender is neutralized in plural forms in Michif.

The examples in (1a)–(4a) all show masculine articles and possessives on a masculine noun, while those in (1b)–(4b) show the feminine counterparts. The examples in the plural forms seen in (5)–(6), however, do not vary based on whether they are masculine or feminine nouns; the possessive and the article remain invariable. The articles and possessives are given in Tables 1 and 2. Note that the plural possessors do not show gender, even with a singular possessum (not, vot, loer).

Table 1

Michif articles

Michif articles

-> Voir la liste des tableaux

Table 2

Michif possessive articles[2]

Michif possessive articles2

-> Voir la liste des tableaux

In sum, Michif marks sex-based gender via (in)definite articles and possessive elements, only for singular nouns with singular possessors.

3.2. Animacy-based gender in Michif

Nouns in Michif are marked for animacy, though the details of how grammatical animacy would be assigned to French-origin words remains unclear. This animacy marking shows up in Michif on demonstratives and on verb forms/agreement.

In Michif, animacy is marked on the demonstratives. Table 3 shows this paradigm.

Table 3

Michif demonstratives

Michif demonstratives

-> Voir la liste des tableaux

Both the demonstrative and the article in Michif mark number, while the demonstrative marks animacy and the (singular) article marks masculine or feminine gender (7)–(9).

The examples in (7) use the inanimate demonstratives while the examples in (8)–(9) use the animate demonstratives. Note that in (10), while fraa’nbwaaz ‘raspberry’ takes animate marking, it is not animate in the natural world, discussed in section 3.0 for the Plains Cree ayôskan ‘raspberry’.

Transitive and intransitive verb stems in Michif are found in pairs differentiated by animacy. Intransitive verbs distinguish between an animate and inanimate subject, while transitive verbs distinguish between an animate and inanimate object. Consider the following examples, showing that the animacy of the object determines which verb is used, as both subjects are identical, and only the objects change. The verbs also change. In the following examples, (10a) shows that the form with /ht/ denotes an inanimate object, while in (10b) we see that the form with /m/ denotes an animate object.

Now compare the data in (11) with that in (12). The relevant morphemes are bolded.

Note that the root in the examples in (11) is different from the root in (12). Since the subject, verb and person and number of the object are all identical, it is clear that the crucial difference is in the object’s animacy; the roots denote an animate object in (11) and an inanimate object in (12). A verb paradigm of a single root in Michif being inflected to denote changes in valency and animacy can be seen in Table 4. Note that in some cases the verb root is identical or nearly identical (examples in Table 4; (10a,b)), and in other cases the two verbal roots are different (11–12).

Table 4

Verbal animacy marking in Michif: paash- ‘to dry’

Verbal animacy marking in Michif: paash- ‘to dry’

-> Voir la liste des tableaux

We have seen that animacy is marked both on the demonstratives and on the verbal forms, on affixes and sometimes on roots, in Michif, and therefore has a central role in the Michif grammar. We have also seen examples where animacy is not only semantic, as in (7)–(8) ‘horse’ vs ‘chair’, but also arbitrary, as in the ‘raspberry’ example in (9).

4. Previous research on gender in Michif

In this section I outline the main research that has been published on Michif gender, in the order in which it was published. Papen (2003a) was the first to do a gender-specific study on Michif, while Gillon and Rosen (2018) and Sammons (2019) took the question up more recently.

Although gender assignment was mentioned in work on Michif previously, by Bakker (1997) and Rhodes (1977), the first published research specifically investigating details of Michif gender appears to have been Papen (2003a). Papen explains his research question with the following:

La question qui nous intéresse ici est de savoir comment le locuteur du mitchif, qui, faut-il le rappeler, n’est pas nécessairement un locuteur du français et presque jamais un locuteur du cri, parvient à attribuer les genres appropriés aux substantifs afin de pouvoir effectuer les accords appropriés…

Papen 2003a: 131

In essence, Papen is asking how Cree animacy gets attributed ‘correctly’ to historically French nominals, when there is no grammatical animacy in French. Note that under the dual grammar hypothesis in Michif (Bakker 1997; Papen 2003b; Bakker & Papen 1997), that is to say, where the historically French items in Michif follow French grammar, and historically Cree items follow Cree grammar, this is indeed an important question (but see Rosen 2007; Rosen 2006; Gillon and Rosen 2018; Rosen et al. 2020 for an alternate approach which does not propose dual grammars). The dual-grammar hypothesis posits that nouns—the historically French items—will follow the gender assigned in French, therefore a masculine noun mur ‘wall’ in French will also be masculine in Michif. However, since these nouns must also be assigned grammatical animacy, and there is no grammatical animacy in French, how does a Michif speaker attribute this grammatical gender in Michif, without speaking Cree? Papen investigates whether Michif assigns the same gender that would be assigned to the ‘equivalent’ historically French item. In other words, one must find the Cree–French translation equivalent and see whether the same grammatical animacy is assigned to both.

Although animacy is indeed grammatical in Cree, in many cases, semantic animacy still often applies: humans and animals are given animate gender, and the animacy of objects can be shifted either due to different circumstances or for stylistic purposes. For instance, in Plains Cree the word for ‘stick’ and ‘tree’ is the same word, with different animacy, as in (13) below. Note that the plural markers are different for animate and inanimate nouns in Plains Cree.

Similarly, nouns can be shifted between animate and inanimate in a story as a literary device. For example, Bloomfield (1930) includes a telling of the ‘Rolling Head’ story which contrasts inanimate and animate values for a head which has been cut off but continues to be inhabited by the woman’s mind. Muehlbauer (2008) outlines the different uses back and forth of animacy and inanimacy, showing that semantic animacy does apply, with some exceptions, at which point grammatical animacy takes over.

Since semantic or real-life gender[3] is applied to lexical items, it would be unsurprising for two languages with semantic animacy to share the same animacy assignment for most items, other than perhaps items where different world views might lead to different natural animacy. For example, plants may be designated as animate in one language but not in another. However, in Plains Cree, despite some earlier literature trying to argue for semantic animacy based on a semantic feature (often referred to as ‘power’) that triggers animate agreement (Hallowell 1960; Darnell and Vanek 1976, among others), there is some consensus that while semantic animacy is often followed, there are also cases of animacy assignment which cannot be explained away by semantics, and that this ‘power’ feature has not been able to exhaustively characterize this set of nouns. Ratt (2016: 27), for example, gives a table of examples of semantically inanimate nouns which take animate agreement and some inanimate nouns which might be seen to be animate given arguments made for those in the animate category on the left.

Table 5

Plains Cree animacy (adapted from Ratt (2016: 27))

Plains Cree animacy (adapted from Ratt (2016: 27))

-> Voir la liste des tableaux

For example, ayôskan ‘raspberry’and kaskitêmin ‘blackberry’ are animate and inanimate, respectively. It is difficult to argue that these two similar berries should be of different semantic animacies, especially when mitêhimin ‘strawberry’, iyinimin ‘blueberry’ and misâskwatômin ‘saskatoon berry’ are also all inanimate in Plains Cree. What properties a blackberry possesses that these other berries do not possess is not straightforward, leading to an analysis that while semantic animacy does play a role in Cree, there is also grammatical animacy at work in the language, underivable by natural meaning, no matter the worldview.

Because semantic animacy applies to many areas of the lexicon, and perhaps also because it is not always easy to find exact translations for terms, Papen (2003a) focuses on two areas where there is obvious grammatical animacy in Cree: clothing, and fruit and vegetables. He examines animacy of these terms in Plains Cree and Michif (and East Cree, which is less relevant to our question here) in published dictionaries to see whether the animacy of the Michif terms are the same or different as those in Plains Cree. Most of the terms assigned the same animacy in both Michif and Plains Cree, with very few exceptions, but what also becomes apparent in his analysis is that there is some variability in terms of usage between animate and inanimate gender assignment in Michif. Four of the fourteen fruit and vegetable terms were found to be acceptable as either animate or inanimate in Michif: strawberry, raspberry, potato and radish. While Papen’s goals are to see how Michif speakers attribute gender as compared to the source languages, and he finds that they overwhelmingly assign gender along the same lines as the source languages, it is the variability in this gender assignment in Michif that I would like to focus on in this paper. This variability is mentioned in Papen (2003a) as well:

le genre (animé / inanimé) des substantifs cris est systématiquement attribué à leur équivalent français. Il existe, cependant, quelques exceptions qui peuvent être dues à la variation dialectale ou à la variation inhérente.

Papen 2003a: 138–139

This ‘inherent variation’ is not discussed in Papen, but it is of interest to the discussion of gender here.

Overall, Papen (2003a) discusses Michif gender, asking how speakers assign gender to nouns that are not native to the language assigning the gender. Because his approach assumes that two grammars are operating within the language (Bakker and Papen 1997; Papen 2003b), he asks how Michif assigns sex-based gender to Cree and English nouns, and how it assigns animacy-based gender to French and English nouns, and whether gender assignment from French or Cree to Michif is straightforward. Indeed, most French-based nouns in Michif do assign the same sex-based gender as in French.

Gillon and Rosen (2018) analyze Michif gender as an amalgam of two systems: the French system marking sex-based gender, and the Cree system marking animacy-based gender. They do not address the question of how French- or English-source items are assigned animacy-based gender, but rather address how to analyze such a dual system in a generative framework. Adopting a framework developed by Kramer (2015), they argue that the Michif system, while complex, is easily analyzable, focusing on the idea that while Michif may have been formed in a fairly unusual contact situation, it behaves as a language like any other, and does not require extra arbitrary structure to be explained, such as designating lexical items for their historical source in the grammar. This was in reaction to claims posited elsewhere in the literature (i.e., Bakker 1997; Bakker and Papen 1997), that lexical items in Michif are marked for their historical source in order to follow the grammar of that language. Gillon and Rosen (2018) adopt a generative framework to analyze of the complex synchronic Michif system while not attempting to answer the question as to how Michif speakers assigned gender.

Kramer’s (2015) framework argues that gender features are syntactically located on the n head, and distinguish between natural gender and grammatical gender. Natural gender features, i.e. gender that is assigned on the basis of some semantic property in the real world, are interpretable, while arbitrary (or grammatical) gender is the realization of uninterpretable gender features. For instance, animacy would be interpretable on nouns such as the above example ‘horse’ in Michif, but uninterpretable on ‘raspberry’. Likewise, sex-based gender would be interpretable on items such as la fille (f) ‘the girl’ or le garcon (m) ‘the boy’ in French, but uninterpretable on items such as la table (f) ‘the table’ or le mur (m) ‘the wall’. Gender assignment, then, for Kramer, operates along two dimensions: according to some natural or semantic property in the real world (interpretable features) or arbitrarily (uninterpretable features). The Michif merged system results in six possible gender feature combinations on nouns, when we combine the sex-based and animacy-based features. These are given in (14), with the semantically impossible combinations shaded out.

In the Gillon and Rosen (2018) analysis, it is posited that given the complexity of the Michif gender system, one might expect some simplification over time. Specifically, based on Kramer’s framework, they posit that one might expect arbitrary feminine gender (that is; items which are marked for feminine grammatical gender but which do not correspond to feminine in the real-world, i.e. table ‘table’ vs fille ‘girl’) to erode first. While a few Michif examples pointing to this erosion in process are given, it was left as an open question. However, Sammons’ (2019) work allows this prediction to be examined more closely.

Sammons (2019) investigates Michif nominal classification in a natural language dataset. The dataset consists of a multimodal corpus with 25 Michif speakers, representing a rich array of language speech types, bolstering Papen’s (2003a) initial dictionary study of gender in Michif. Sammons (2019) corroborated Papen’s dictionary findings, finding strong association between Michif and Cree animacy, and with French and Michif gender. Further, delving into Papen’s question as to other lexical innovations, she found that English-origin items were primarily associated with masculine gender. That Michif gender (both animacy-based and sex-based) is strongly correlated with Plains Cree and French gender is certainly unsurprising, and seems to be well-established and agreed upon.

Sammons’ corpus is a rich source of naturally-occurring speech, especially for an endangered language such as Michif. Because her methodology included only extracting nouns with overt sex-based and animacy-based gender, she was able to analyze a total of 1261 tokens representing 261 unique nouns (called lemmas). The repetition of nouns allows for the study of variability of the gender of these items in a way that is impossible in elicitation. Both Gillon and Rosen (2018) and Papen (2003a) cite speaker intuitions allowing variation between masculine and feminine gender for a given lexical item, but Sammons’ corpus gives us a way to study this variability in natural speech.

Sammons (2019: 183–184) employs three criteria to her dataset in which to retain ‘truly’ variable tokens: 1) any item with fewer than four available tokens was excluded from study, 2) if the token count was four or more and the proportion of use of the two genders was 80: 20, then the item was assigned the majority gender value with the ‘exceptions’ being deemed nonce speech errors, and 3) any instances not fitting into the first two criteria, where variability affected more than 20% of the tokens was treated as variable animacy, i.e. not attributed to nonce speech errors. With these criteria in place, she found eleven ‘truly’ variable items in all; two items were variable for (Cree-like) animacy-based gender, and nine for (French-like) sex-based gender. Otherwise, items generally strongly correlated to the genders in their respective source language, with 92.7% of Michif items corresponding to Cree animacy (Sammons 2019: 208), and 88.4% of Michif items corresponding to French sex-based gender (Sammons 2019: 232). Sammons also found that 91% of English-source items were assigned masculine gender. This work shows that there is some, but not extensive variability in gender assignment in Michif, and that sex-based gender is more somewhat variable than animacy-based gender. Sammons (2018: 229) points out that this sex-based variability also tends to apply to arbitrary-feminine nouns, rather than the default-masculine nouns, and so Gillon and Rosen’s (2018) prediction that sex-based gender is more likely to erode than animacy-based gender is at least partially supported.[4] She does point out however that there is variability in both dimensions of gender in Michif.

5. Discussion

Sammons’ research questions mainly centered around finding the underlying gender assigned to a given lexical item in Michif, and as part of that question, investigating whether Michif had cases of lexical items with variable gender. However, underpinning these questions is the assumption that there is an underlying gender assigned to each item, with deviations from this being considered nonce speech errors. For examples, because the goal was to find cases of variable gender, an item which used, for instance, masculine gender 80% of the time, was assumed to be a masculine noun, and not ‘variable’. Any instances where one of these items was used with feminine gender would be considered a nonce speech error. This assumes, however, that variability in gender assignment is not part of the grammar underpinning the linguistic system, and that there is a ‘true’ gender assigned to a given lexical item. Sammons makes an explicit methodological choice which is successfully argued for, but it is equally possible to make a different choice. Rather than exclude these items a priori, we might want to consider the full range of examples in a study of gender variation more broadly in Michif, with all the variability considered as part of the Michif variable grammar.

While gender is often (and perhaps always, prescriptively) considered to be an immutable feature of a lexical item, with items being assigned a single gender, that is to say, a given lexical item may only be feminine or masculine, examples abound where this is not the case. Often this variability is dialectal; for example, French vidéo is masculine in Canada and feminine in France, while job tends to be feminine in Canada but masculine in France. It is clear, however, that even within the Quebec dialect of French, there is more variability in gender than in is allowed prescriptively, and that this is sometimes, but not always based on register of speech, as argued by Remysen:

il est fréquent au Québec qu’on parle d’une autobus, d’un affaire et d’un heure… Il faut noter toutefois que ces traits ne sont pas toujours généralisés dans la variété québécoise: ainsi, parler de la bus est typique des situations de communication informelles et du parler populaire. D’autres termes, par contre, prennent un autre genre que celui qui est d’usage en Europe, même s’il est utilisé en contexte formel : une trampoline.

Remysen (2003: 33)

Relatedly, Klapka (2002) gives examples in a corpus of storytelling interviews with Quebecois speakers born in the nineteenth century (Récits du français québécois d’autrefois, RFQ) (Poplack and St-Amand 2007, where gender is used variably by a single speaker, even within a single sentence.

In example (15), we see the two different genders being used for the same (prescriptively feminine) noun cheminée. In examples (16)–(17), the masculine adjective (in bold) is used with a feminine noun. In each of these cases, the nouns are very common, familiar lexical items, so it cannot be that the gender is unknown due to the lexical item being unknown. Klapka found that 5–6% of lexical items displayed variable gender (Klapka 2002: 26).

Gillon & Rosen (2018) argue that this confusion has been partially systematized in Quebec French, where many vowel-initial words have developed arbitrary feminine features from their historical default masculine (Chamberlain 1895, Barbaud et al. 1982, Klapka 2002). Barbaud et al. show that while 38% of the changing vowel-initial masculine nouns became feminine, only 1.6% of feminine vowel-initial words became masculine, showing a particular asymmetry from masculine to feminine. Klapka also finds similarly (Klapka 2002: 28) This can be attributed to phonological processes such as analogical extension, as the demonstratives and prenominal adjectives in these cases are all pronounced identically to the feminine forms. The following are examples of nouns which prescriptively are masculine, but often surface with feminine articles and adjectives in Quebec.

Historically, nouns have switched their prescriptive gender even in hexagonal French: légume ‘vegetable’ was feminine until the eighteenth century, and pamplemousse ‘grapefruit’ is listed in the Larousse as being either feminine or masculine. The English borrowing sandwich used to be feminine and is sometimes still employed as a feminine in Quebec. Gillon & Rosen (2018) argue based on these examples, that there is a certain degree of instability of the sex-based gender marking in French, and most notably, in the French input that would have been present in the development of Michif.

There seem to be at least three types of nouns in French which are prone to gender variability: a) vowel-initial words, where the elision of the determiner and the appearance of the normally-silent feminine-marking consonant in liaison leads to a reanalysis (un/e autobus); b) those who have switched by analogy with other phonologically similar forms (un/e heure, un/e trampoline); and c) borrowed words (un/e job). Note that in each of these types of items, changes in gender follow natural language change processes commonly found in languages around the world: in the first two, the changes are due to internal change, especially reanalysis. In the third, external language contact triggers the variability. There is indication that this variability is more present in Canadian French, both spoken and written, according to Remysen (2003). Arguably, Canadian French, even in addition to being more in contact with English, has been less historically influenced by prescriptive norms than hexagonal French, given the strength and tradition behind bodies such as the Académie française in France. It seems likely that varieties of French that are less likely to endure conservative pressure from prescriptive or written codes may display more gender variability than those that have a long tradition of highly prescriptive rules which are ingrained throughout the school system. This seems to be borne out when looking at orally-dominant varieties of French, such as child language or French-based creoles. Both of these tend to reanalyze words commonly found with a liaison consonant such as un ours as un nours, where the initial n has been lexicalized into French lenounours ‘the teddy bear.’ This reanalysis is also present in Michif, which has led to the well-documented allomorphy of many items that were vowel-initial in French, including nours and lours ‘bear’, or zwezo, lwezo and nwezo ‘bird’ (Bakker 1997, Papen 2005, Rosen 2007). If reanalysis can trigger gender variability in vowel-initial French nouns, as we have seen above, we could expect that this in turn might act as a catalyst for gender variability elsewhere in the language. In addition, a variety which combines this reanalysis with a general lack of prescriptive pressure might be more susceptible to this sort of change. If this is the case, it would be unsurprising that Michif, a language that has never traditionally been taught in schools, and which even today does not have a strongly encoded written form, would be less pressured to conform to gender assignment rules than hexagonal French.

French is not the only language where gender is more absolute in prescriptive norms than in spoken norms. There is also known gender variability between dialects in other languages, for example in German and in Arabic. For example, Butter is masculine (der Butter) in Swabian German, while feminine (die Butter) in most other German varieties. Other examples which are region-dependent in German are Joghurt (das/der), Virus (das/der) and Radio (das/der)[5].

Similarly, Arabic displays gender variability between dialects (Hamdi 2017). The following examples are from Makkah, in Saudi Arabia, where rural tribal varieties treat ‘tummy’ as masculine while urban ones treat it as feminine, as is seen in the examples below.

In the above examples, the gender marking on the verb is different in the different varieties of Arabic. This is yet another example where gender-marking norms are found to be variable.

Considering that there does seem to be some variability in gender systems cross-linguistically, we could reframe the question of Michif gender assignment in particular to look at it assuming conditioned variability in the system. While it seems clear from both Papen and Sammons’ work that gender in Michif follows gender in French and Cree, the question is what we can learn by not setting aside the examples which are variable only less than 20% of the time. Sammons’ criteria make sense for the goals of her study, which is primarily to document the gender assigned to Michif lexical items and its correlation with French and Cree gender assignment. With the backdrop of variability which may be inherent even in gender assignment, however, it may also be interesting to include all cases of variation in the Michif dataset, to answer another research question, namely, to assess the role of gender and of variability in the language.

Note that after Sammons’ criteria are applied to her data, 30 lexical items are excluded, with 31 items retained with variable animacy/gender values. However, if we consider that variability is underlying throughout the system, and not that only certain lexical items are variable, with the others being treated as nonce speech errors, we may be missing generalizations or insights of a Michif variable grammar. Expanding the criteria for inclusion enriches the dataset considerably, and allows us more insight into the language’s treatment of gender.

If we expand the criteria for inclusion, we find fourteen cases (instead of two) of variable gender for animacy, and sixteen (instead of nine) cases of sex-based gender. The full list of items is given in the tables below:

Table 7

Michif items with variable sex-based gender (from Sammons 2019: 190)

Michif items with variable sex-based gender (from Sammons 2019: 190)

-> Voir la liste des tableaux

Table 8

Michif items with variable animacy-based gender (from Sammons 2019: 185) [7]

Michif items with variable animacy-based gender (from Sammons 2019: 185) 7

-> Voir la liste des tableaux

When we look at this list, we see that there are many more examples of variability in gender assignment across both sex-based and animacy-based dimensions. The generalizations that Sammons (2019) is able to make based on even just her small subsection of her rich dataset suggests that investigating the additional regularities in gender assignment has the potential to yield even more interesting insights. Considering gender to be a locus of a conditioned variability could tell us what features of gender might be stronger or weaker, and yield insight into how such features change diachronically, along the lines of work such as Burnett et al. (2018).

In Sammons’ data, we can see clearly that nouns with feminine gender in French are most often variable in Michif; if we consider all the variable examples, 12/16 of them are feminine. Sammons finds 7/12 of these to be variable according to her criteria, which she already takes to be partial support for Gillon and Rosen’s predictions. However, if we look at the whole range of examples, then 12/16, or 75%, of the examples, are feminine in French, and surface with either masculine or feminine in free speech. These extra examples do appear to follow the same tendencies as the rest of the data, and also follow the predictions made in Gillon and Rosen (2018). Given that they seem to be following the same pattern, it seems reasonable to include all the examples in our analysis, i.e. including even cases which might be ‘production errors’, into the analysis as part of a larger pattern.

Lastly, an argument that seems to be made when confronted with variable gender is that speakers are simply making ‘production’ mistakes; that they are not fluent, or perhaps no longer fluent in the language, and this is the reason why they make ‘mistakes.’ This is an argument that gets used in particular for Michif speakers generally when they do not follow the rules as expected by linguists, due to the status of Michif as an endangered language. Not only should this be considered a misguided colonial practice to be avoided, assuming native speakers are making mistakes, but all the speakers of Michif studied in these publications are fluent speakers who grew up speaking the language, even if they do not all speak it as their principal language now. These are not second-language learners with imperfect acquisition, whom we might expect would have ‘problems’ with gender. In any case, the larger goal of this paper is to show that gender can be variable for a myriad of reasons (phonological, dialectal, contact-based), across different languages, and that perhaps its discussion could be reframed as having conditioned variability in the same way that variationists investigate variability of the use of the subjunctive in French (Poplack et al. 2013), or variable ways of expressing negation in English (Childs et al. 2020) rather than deciding that this variation a priori should be considered as production mistakes. In this sense, Michif gender would sit alongside other variable aspects of the grammar, such as choice of possessive adjective and lexical choices between Cree and French, with likely many other aspects which are as of yet unstudied.

6. Conclusions

In this paper I have tried to consolidate the literature on Michif gender. I have also used Michif gender as an example of where a normally-prescriptive feature of a related language, French, might benefit from being framed in a different way. Specifically, I suggest that we may want to consider treating gender to be a locus of variable grammar, rather than an absolute rule of grammar, as it is normally portrayed in prescriptive grammars. More specifically, languages such as Michif that are sheltered from the pressure of prescriptive norms should be of interest to test predictions of theories of grammar, a perspective learned from the work of Yves Roberge, the scholar this collection honours. Linguists could benefit from stepping back and re-evaluating language evidence based on what we see, including all examples, and all variation, rather than treating them as “production errors”. There is value in studying where these “errors” appear, as we may find that they follow regular conditioning which leads to a deeper understanding of the languages(s) we are studying.