1 Introduction: the many puzzles of double plurality

The study of the mechanisms involved in language mixing¹ below the word level has recently gained prominence in the morpho-syntactic literature, as witnessed by the bulk of work in this area, see e.g., González-Vilbazo & López (2011), Alexiadou (2011b; 2017; 2020), Bandi-Rao & den Dikken (2014), Åfarli (2015), Callies & Stolz (2016), Alexiadou & Lohndal (2018; to appear), Grimstad et al. (2018), Riksem et al. (2019), López (2020), Fisher et al. (2022), Lohndal & Putnam (2024) to mention a few. These authors all argue that this type of mixing may be used as an important tool to identify the basic units that are involved in both monolingual and bilingual word formation. These papers make a convincing case that data from language mixing support a decompositional view of morphology, whereby morphemes are the realizations of abstract syntactic features. Thus, studying word-internal mixing, as Alexiadou & Lohndal (to appear) point out, informs theories of what a word is, how it is built, and how concepts are lexicalized across languages. This work shares with MacSwan (1999) the view that no extra rules are needed to account for code switching, but departs both his lexicalist framework, and also from earlier work that had argued language mixing is conditional upon phonological integration: Mixing can only take place if a morphological ending is integrated into the language of the lexical form, see e.g., Poplack (1980) and Sankoff & Poplack (1981), and Alexiadou & Lohndal (2018), López (2020) for critical overviews.

This paper follows this decompositional tradition and probes the question of how a late insertion model such as Distributed Morphology (Halle & Marantz 1993 and subsequent work) may be employed to account for double marking of plurality in language mixing, illustrated in (1) with an Ewe English example:

Double marking of plurality has long been observed in various language contact situations. For instance, it is a feature found in several English based Creoles. According to Mühlhäusler (1985), in Tok Pisin, there are cases in which the NPs might be doubly marked by two pluralizers, an affix and the plural determiner ol shown in (2), from Blaxter Paliwala (2015: 220), and see also Bobyleva (2013) and references therein for Jamaican Creole:

Doubling of plurality in language contact raises several issues. While roots may be borrowed freely, borrowability of inflectional morphology is controversial, raising the question of whether the borrowed plural form is interpreted as an affix or not, see e.g., Backus (1999), Myers-Scotton (2002), Matras (2009), Muysken (2011), Bakker & Hekking (2012), Gardani (2012), Seifart (2015), Fisher et al. (2022), Gouskova (2023) and references therein for contrasting views. Matras (2009: 212), while admitting that borrowing of inflectional morphology is rare, signals that the plural is one systematic case of borrowing. Gardani (2012: 93) concludes that “the higher borrowability of plurals in and through language contact can be explained based on two facts: First, the category of plural is more similar to derivation than other categories of inflection and, second, in most instances, the plural borrowings investigated are formed via agglutination. These properties seem to favor not only the temporary use but also the permanent integration of plural morphemes in the recipient languages.” More recently, Fisher et al. (2022) discuss allomorphic variation in Pennsylvania Dutch plurality, claiming that -s plurals, e.g., Schnuppi-s ‘hankies’ in the language are structurally identical to English plurals. The authors propose that -s realizes the plural predominantly in combination with English roots. According to Myers-Scotton (2008), the plural maybe be borrowed and as a result doubled in mixing, it being an early system morpheme in her system. As she explicitly states, “this is supported by all published corpora”, Myers-Scotton (2008: 31). As detailed in Hakimov (2021), previous research on double plurals in mixing has attributed their presence to either structural factors, i.e., type of affix (Boumans 1998: 91; Myers-Scotton 2002: 91, Gardani 2012), erroneous access in production (Myers-Scotton 1993: 132–136; Myers-Scotton & Jake 1995: 1000) or high frequency of some plurals in the source language (e.g., Backus 1999: 97–99). In Hakimov’s own study of Russian German mixing, frequency, phonetic shape, and mismatch of forms are also shown to play a role. Backus (1999) proposes that these forms are borrowed as complex chunks and Muysken (2011) suggests that plural forms are pied-piped together with the stem and are not interpreted as plurals.

From a morpho-syntactic point of view, double plurals raise a more general issue concerning form-meaning correspondence and the locus of plurality. Double plurals are a sub-case of multiple exponence, defined as in (3), see Harris (2017) for an overview:

The issue here is that if there is only one semantic feature in a functional projection in the nominal spine associated with plurality, namely Number (Ritter 1992), why should it receive two realizations, thereby violating the principle in (4)?

In this paper, I will argue that multiple exponence in the case of mixing plurals is epiphenomenal, see also Caha (2021) for a general such approach to multiple exponence. Specifically, I will argue that exponents of plurality realize two distinct heads present in the morpho-syntax of nouns and will offer an analysis of (1) and similar patterns along the lines of analyses of double marking of plurality in a variety of typologically unrelated languages, e.g., Amharic (Kramer 2016) and Breton (Acquaviva 2008 and references therein), and more recently Robinson (2022). My hypothesis is that one of the plural affixes, mostly the borrowed one, is a word marker, i.e., it is used to create a word out of a root. Thereby I will show that the patterns Kramer and Acquaviva identify for Amharic and Breton also hold for the mixing cases. The gist of the analysis follows Acquaviva’s (2008) and Alexiadou’s (2011b) treatment of split plurality, employing the structure in (5): there are two loci for plural morphology, n, which hosts lexical plurals, and Div, which is the locus of grammatical plurals. The two heads differ in terms of function and semantic contribution. As detailed in Acquaviva (2008: 271 ff.), a decomposition of nouns into n and root makes it possible that nouns that share the same root may surface with distinct plural forms, e.g., braccia – bracci ‘arms (body parts, also measurement)’, ‘arms (any other sense)’. These forms are analyzed as distinct lexemes, and only the former is inherently plural and expresses a non-individual conceptualization. This information is clearly lexical/derivational, not inflectional. Crucially, -a plurals do not contain affixes that realize number. Plurals on n are thus part of the word defining domain, while plurals in Div are outside this domain, see also Alexiadou (2011b), and Wiltschko (2008, 2021).

This analysis predicts that there will be languages where plural markers in Div may co-occur with lexical plurals in n. Acquaviva (2008) discusses Breton, see (6), from Acquaviva (2008: 260), see also Robinson (2022); Kramer (2016) argues that Amharic is another example, see (7).²

In this paper, I will further develop the split analysis for the mixing cases under discussion, cf. Gutiérrez-Bravo & Uth (2020). Specifically, building on Alexiadou & Lohndal (2018; to appear) and López (2020), I will propose that double plurals in mixing may also be captured under (5): in order to signal that a noun has been created, which in turn can be pluralized, speakers recruit a lexical plural affix to signal the presence of the nP layer. This is related to Backus’s (1999) and Muysken’s (2011) idea that these forms are borrowed as complex chunks and that one of the plural forms is not interpreted as grammatical plural. As we will see, the mixing data to be discussed support the view that the plural words are not re-interpreted as roots, i.e., brothers is not re-analyzed as a root, but the employment of the plural is systematic. Thus, both grammatical and lexical plural are realized. This fine-grained decomposition of the nominal spine is similar in spirit with work following Kayne’s 2005 Principle for Decompositionality, see also Putnam (2020), Lohndal & Putnam (2021; 2024), Fisher et al. (2022) and Sauerland & Alexiadou (2020; to appear).

Adopting the proposals on split plurality in Acquaviva (2008), Alexiadou (2011a, 2021), Kramer (2016), Wiltschko (2008, 2021) among others, I will argue that in doubling each plural form lexicalizes a different head in (5) and will discuss the semantic contribution of each head in section 3.1 in detail. This is a consequence of the principle in (8), building on Martin et al. (2023), Guasti, Alexiadou & Sauerland (2023), cf. Hein et al. (to appear), and Fabrégas’s Principle in (9). A split analysis of plurality captures the mixed inflectional-derivational nature of the plural mentioned above: Div plurals are inflectional, e.g., they create pluralities out of words, while n plurals are derivational, i.e., they create words out of roots:

In cases a morpheme may receive zero realization, multilingual speakers, as discussed at length in e.g., Polinsky (2018), avoid zero exponence,³ i.e., bilingual speakers prefer overt elements. This further relates to Scontras’s et al. (2018) insights into the structure of languages in contact. These authors argue that two pressures shape contact grammars: representational economy, which leads to more parsimonious representations, and analyticity. I will suggest that there are tools employed in Distributed Morphology (such as Maximize Exponence) that may be employed to model analyticity (in the spirit of Lohndal & Putnam 2024). By contrast, representational economy can be modeled as a case of Minimize Exponence, Siddiqi (2006), which may be the result of Fusion, see sections 3.1, and 4, and Putnam (2020), and Lohndal & Putnam (2024) for alternative ideas.

The paper is structured as follows: I will first introduce double marking of plurality in language contact situations focusing on pairs for which a detailed description exists that facilitates their comparison to e.g., Amharic and Breton. In section 3, I will then turn to the split analysis of plurality as introduced for a variety of languages and then see how it applies to the other cases discussed here. In section 4, I will then turn to a discussion of why the specific contexts favor doubling and why the plural is prone to doubling and no other numbers. In section 5, I conclude and highlight some issues that merit further investigation.

3 Towards an analysis

4 Why the plural? And why mixing?

In this paper, I focused on double plurals in language mixing considering what we know about double plurals in several typologically unrelated languages. I argued that all instances of double plurality can be treated as cases of split plurality, suggesting that multiple exponence is in fact an epiphenomenon. This shares with Backus (1999) and Muysken (2011) the intuition that some plural forms are word markers. As stated in Backus (1999: 98), “plurals are established lexical units for the speakers who use them. This means they are chunks […]: they are lexical units that consist of more than one morpheme.” I argued that this follows from some version of a general principle that requires a one-to-one form-meaning mapping, along the lines of (8), and cf. earlier work by Slobin (1977, 1985) and van Hout (2008):

I now turn to two questions: why (8) applies to the plural, and why language mixing is where we see it at play. Beginning with the latter, language contact situations have been argued to trigger overexplicitness in one way or another, see e.g., the by now extensive work on heritage languages, and Polinsky (2018), Lohndal & Putnam (2021) for recent summaries. For instance, there are claims that “creole languages lose opaque features and become more transparent”, Seguin (2020: 218). In his discussion of vernacular universals, Chambers (2004: 128) notes that explicitness “features occur not only in working-class and rural vernaculars but also in child language, pidgins, creoles and interlanguage varieties. Therefore, they appear to be natural outgrowths of the language faculty.”

Thus, the more general question is what it is about the language faculty that it allows in some domains to Maximize Exponence and to strictly respect Bi-uniqueness by avoiding zero (and thus fusion of functional nodes). Some reasons have been suggested in the literature, for instance, it can be argued that it is due to communication pressure (see Leufkens 2015) for lengthy discussion. I will develop my ideas about that considering the discussion in Scontras et al. (2018) with respect to heritage languages. These authors argue that there are two pressures that shape heritage grammars: representational economy, which leads to more parsimonious representations, and analyticity, which leads to fully articulated structures. In their terms, analyticity favors one-to-one correspondence between form and meaning. Let us generalize this and claim that these two pressures shape grammars in general and lead to speaker variation. Specifically, analyticity leads to a representation where underlying semantic relations should be marked overtly and clearly and there is a preference not to mark a semantic category by ∅. Perhaps zero marking is avoided as it leads to processing difficulties, see Truong (2024). For analyticity to be preferred, there must be an alternation that involves a zero form and/or a form that has been built based on Fusion of two terminals and a form that contains overt realization of each feature/terminal, see Guasti, Alexiadou & Sauerland (2023), and also Alexiadou (2017), and Alexiadou & Rizou (2023) for discussion of such alternations in the verbal domain. Analytic vs. synthetic alternations may be recruited for register purposes, showing sensitivity to certain situations/contexts, hence allowing speakers to signal register shifts by changing the form of linguistic expressions, along the lines of Adger’s (2006) combinatorial variability, see also Alexiadou et al. (2022) for some discussion. This may yield formal vs. informal and/or spoken vs. written language contrasts of the type discussed in Bokamba (2012). By contrast, representational economy is in line with Minimize Exponence and Fusion; note that this is unlike the proposal in Scontras et al. (2018) that structural heads are eliminated as the result of bundling. Rather, here the functional sequence is always present, and Fusion is one way to post-syntactically model representational economy. Note furthermore that an explanation along these lines does not predict that Maximize Exponence will always be at work, it simply predicts that speakers may sometimes opt for an analytic representation, explaining perhaps the rarity of doubling across domains and its non-obligatoriness. This type of explanation departs from views, according to which double forms are a symptom of erroneous access in production (Myers-Scotton 1993: 132–136).

The second question is: why the plural feature, see also Hakimov (2021) for discussion and references. As stated in Gardani (2012 93): “the category of plural is more similar to derivation than other categories of inflection.” This dual nature is what the analysis of split plurality straightforwardly captures. Specifically, plurality is manifold (Acquaviva 2008): it creates at least (i) collection of singularities, and (ii) lexical meanings (collectives, groups). (ii) is the derivational nature of plurality, while (i) comes close to the inflectional nature of plurality. I assume, following Wiltschko (2014), that the primitives of the functional spine are universal and category neutral, see (40). In the nominal domain, (41), DP realizes discourse linking, while nP, the locus of lexical number, is associated with classification, and DivP, the locus of grammatical number, is associated with point of view (dividing in Wiltschko’s 2021 terms, presumably #P associates with anchoring):

Importantly then, plural morphology may associate with both point of view and classification, allowing double realization that is associated with different layers of the spine, i.e., double marking may be expected as the function of the form differs.¹⁵

A further question, raised by an anonymous reviewer, concerns doubling affecting other numbers, e.g., singular, dual and paucal in languages that have these distinctions. With respect to the singular, the following remarks are in order. The plural is considered morphologically marked (Greenberg 1963, Jakobson 1968), as in several languages the singular is expressed via a zero form. If the singular is expressed via a zero form, it is unlikely that doubling of singulars will be found. In languages with singulative morphology, we do have examples of plural of the singulative, see Acquaviva (2008), Mathieu (2014) and Kramer (2015). If the singulative is treated as being on n, see Acquaviva (2008), Kramer (2015), but see Mathieu (2014), it may of course co-occur with plural on Div or combine with other n heads bearing class information. With respect to duals, Acquaviva (2008: 253ff.) argues that in e.g., Breton duals may be analyzed as lexical plurals, i.e., plurals on n. In Breton, dual morphology may co-occur with plural morphology, see (42), from Acquaviva (2008: 249).

Paucals and other minor numbers may also be treated as instances of lexical number, Acquaviva (2008: 71), thus, again predicting co-occurrence with plurals in Div. Acquaviva cites Koasati, Acquaviva (2008: 67), see (43), as an example of a language with the paucal affix may co-occur with the plural one:

The prediction for mixing is that if duals and paucals are lexical numbers, they should co-occur with grammatical plural in Div. This awaits further research.

5 Conclusion and outlook

In this paper, I discussed double plural marking found in language mixing pairs. I proposed that this double marking, a case of multiple exponence, may be treated along the lines of better studied cases of double plurality in languages such as Breton and Amharic as an instance of split plurality and offered an analysis thereof within the framework of Distributed Morphology. I further discussed why the plural and why language mixing favor doubling. I argued that double marking is favored in the context of the plural as plurality is manifold, being associated with two semantic primitives in a universal functional spine. Language mixing favors double plurality, as multilingual speakers avoid zero realization of semantic features and tend to lexicalize all the functional heads in the spine.

While several mixing pairs can be analyzed as involving split plurality, there is one pair that I did not discuss in detail, namely Quechua Spanish. Bakker & Hekking (2012) argue that Spanish plurals in mixing are not notational variants of their singular forms, i.e., they are not borrowed as such. In all the other cases of plurality doubling discussed here, the word order of the markers is fixed. However, in Quechua there is variation among different varieties, (44a, from Muysken 2012: 485) and (44b), from Lorenzino (2000: 117):

According to Muysken (2012), in Bolivian Quechua, -s is most frequent after a vowel, while -kuna appears after a consonant. This seems to be the rule also in Santiago Quechua, according to Lorenzino (2000). However, while -s is unrestricted, -kuna is restricted, as it appears only on animate nouns and does not co-occur with numerals in traditional Quechua. Moreover, in Ecuadorian Quechua, Spanish plural –s is limited to Spanish nouns. Often these are count nouns reinterpreted as Quechua mass nouns (45), Muysken (2012: 484). This suggests an analysis of Spanish -s as a lexical plural, also supported by the fact that in some cases, (45), -s simply marks the Spanish word as a borrowing:

However, if -kuna is indeed restricted to animate nouns, it seems to function as an exponent of n, see Kramer (2015) on gender-number interactions. Recently, Sánchez & Vengoa (2023) discussed the distribution of -kuna and concluded that it is not a marker of plurality, but rather acts as a sorting mechanism. This conclusion is in line with an analysis of -kuna as a realization of n, meaning that both -s and -kuna act as realizations of n, not contributing the meaning of grammatical plurality. This constellation may differ from variety to variety. This further raises the question of whether the Yucatec Maya Spanish data discussed in section 2 could be re-analyzed in this light. Both issues await further research.

A further question is whether there are mixing pairs where doubling does not occur and how the present analysis may explain them. Doubling of plurality is supposedly absent from French based Creoles: for instance, in Haitian Creole, it is always D that bears plural marking, and never n, data from Cyrino & Espinal (2020: 171), Déprez (2005):

This has been attributed to the fact that French number cliticizes to D, rather than to the noun, see e.g., Bouchard (2002). According to Pomino (2016), French nouns no longer bear a plural marking (see also Carlier 2016, and Gerards & Stark 2021).

In the French-Algonquian mixed language Michif, French derived nouns pluralize via D ((47), Gillon & Rosen 2018: 62):

In Michif, only the Algonquian derived nouns can bear both markers of plural, the affixal one from Algonquian and the French determiner lii, (48a), which according to Gillon & Rosen (2018) is in NumberP, our DivP. Algonquian plurals can be argued to be on n: support comes from two observations: first they are conditioned by animacy, and second, these plurals cannot co-occur with diminutives, see (48c), from Gillon & Rosen (2018: 67). This follows from an analysis of diminutives as being on n, see Kramer (2015) on this point for Ojibwe plurals, which also cannot co-occur with diminutives.

Amuzu (2013) points out that Akan and Ga plurals may not combine with English plurals: the English noun either carries the English plural or the Akan/Ga plural, foɔ or -nom and -fo^!i respectively. Akan/Ga nouns never take the English plural. Interestingly, Amuzu (2013: 96) suggests that the plural forms in Akan and Ga are derivational morphemes and that the English plural and the Akan/Ga forms are in competition. This analysis comes very close to viewing the Akan/Ga forms as plurals on n, (see Ahenkorah 2024 for a recent such proposal for Akan), and thus in competition with the English plural, which is an exponent of n in the context of mixing. A further pair that does not show double plurality is Hausa English, according to Amuzu (2013). Hausa is a language in which there are no gender distinctions in the plural, but the choice of a particular plural from correlates with gender. In addition, certain feminine derivational affixes are dropped in the presence of plural morphology (Newman 2022). It seems reasonable to suggest that such plurals are n plurals, in view of certain similarities with Somali, Lecarme (2008), and Romanian, Kramer (2015). If this is the correct analysis for the Hausa plural, the fact that it does not co-occur with the English plural on n is expected.

Finally, Hakimov (2021) reports that in Russian German mixing, German nouns either appear with Russian plural forms or with German plural forms, but no instances of double plural marking are found in his corpus. Hakimov suggests that this may be attributed to the fact that both languages have declension classes, and a conflict arises, thus only one determination of class is retained. This suggests that lack of doubling may be a general characteristic of languages with declension classes. In the case of Pontic, a Greek dialect in contact with Russian, I was not able to detect double plural forms on Russian loan words, see Berikashvili & Lobzhanidze (2017). Russian roots are borrowed and assigned into Pontic Greek declension classes, similarly to what has been discussed in Alexiadou et al. (2015) for English and German roots and Greek declensions. From a Distributed Morphology perspective, declension class information is on n. Typically, in Greek plural surfaces on n, as a result of fusion between Div and nP, Alexiadou & Stavrou (1997), but cf. Markopoulos (2019). If double marking of plurality is indeed prohibited in languages with declension classes of the Greek/Russian type, we may conclude that this is the result of a conflict for class assignment on n. This issue certainly requires further empirical research.

Abbreviations

ADJ = adjective, ANIM = animate, CL= class, COMP = complementizer, DET = determiner, DIM = diminutive, FEM = feminine, FV = final vowel, HAB = habitual, IN = inanimate, INF = infinitive, NOM = nominative, PL = plural, PROG = progressive, REP = repetitive, SG = singular, SM = subject marker

Acknowledgements

Materials discussed in this paper were presented at the Humboldt Universität zu Berlin in November 2021, the 46th Austrian Linguistics Conference in December 2021, the syntax-semantics seminar at the University of Potsdam in February 2022, the workshop Flexible and multiple plural marking in language contact and creolization: Social and situation correlates organized by the Collaborative Research Center 1412 Register in November 2022 in Berlin, the workshop The syntax of small objects hosted at the University of Hamburg in December 2022, the Exo-words Workshop hosted at Penn State University in March 2023 and the 10th International Contrastive Linguistics Conference in Mannheim in July 2023. I am grateful to the participants in these events for their questions and comments. I am indebted to three anonymous reviewers for their constructive feedback as well as to the editors of this special volume, and Terje Lohndal, Fabienne Martin and Miriam Meyerhoff for comments, which have greatly improved this contribution.

Funding information

This paper is part of a project that has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant agreement No. 856421). The contribution also benefited from a project funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – SFB 1412 Register, 416591334.

Competing interests

The author has no competing interests to declare.