Skip to content
BY 4.0 license Open Access Published by De Gruyter Mouton May 20, 2020

Breton a-marking of (internal) verbal arguments: A result of language contact?

Elisabeth Stark and Paul Widmer
From the journal Linguistics


We discuss a potential case of borrowing in this paper: Breton a- ‘of’, ‘from’ marking of (internal) verbal arguments, unique in Insular Celtic languages, and reminiscent of Gallo-Romance de/du- (and en-) arguments. Looking at potential Gallo-Romance parallels of three Middle Breton constructions analyzed in some detail (a with indefinite mass nominals in direct object position, a-marking of internal arguments under the scope of negation, a [allomorphs an(ez)-/ahan-] with personal pronouns for internal arguments, subjects (mainly of predicative constructions) and as expletive subjects of existential constructions), we demonstrate that even if there are some semantic parallels and one strong structural overlap (a and de under the scope of negation), the amount of divergences in morphology, syntax and semantics and the only partially fitting relative chronology of the different constructions do not allow to conclude with certainty that language-contact is an explanation of the Breton facts, which might have come into being also because of internal change (bound to restructuring of the pronominal system in Breton). More research is necessary to complete our knowledge of a-marking in Middle Breton and Modern Breton varieties and on the precise history of French en, in order to decide for one or the other explanation.

1 Introduction

Modern Breton (from 1800 on well attested and still spoken in Brittany, North-Western France, by approximately 500,000 fluent and occasional speakers, cf. Bodlore-Penlaez and Kervella 2011: 56) features a morphological type of argument marking that is unknown in many respects in other Celtic languages and varieties (e. g., Welsh or Irish). Attested since at least Middle Breton (1200–1650: poorly attested, cf. Le Berre 2012, followed by Early Modern Breton; 1650–1800: well attested, but poorly described), argument marking by the element a and its allomorphs (see below) had been grammaticalized by the nineteenth century (cf. Hemon 1975; Ledunois 2002; Favereau 2018). The marker a corresponds to the Middle Breton phrasal marker a ‘from, ofʼ (cf. Widmer 2017 on its morphosyntactic properties) and is of Insular Celtic descent (Middle Welsh o, Cornish a, Middle Irish a). Nowadays, some Breton varieties do not feature a any longer as a phrasal marker (Ledunois 2002; Wmffre 1998). As far as it can be established for Middle Breton, object marking with a optionally occurs with mass nouns in the singular like a bara guen literally ‘of bread white’ in Example (1) and contrasts with the use of the (slowly emerging) indefinite article with count nominals as exemplified in (2), where a is not licensed with assiet ‘plate’.


Middle Breton

prenit … evit un guennec a bara guen
buy.imp.2pl for one penny a white

‘Buy … some white bread for one penny!ʼ

(Qu.; 1632)

NOT: ‘Buy … a (loaf of) white bread for one penny!ʼ


Middle Breton

it da querc’hat un/*a assiet
go.imp.2pl to fetch.inf indef/*a plate

‘Go fetch a plate!ʼ

(Qu.; 1632)

Replacing a with the indefinite article un in Example (1) is possible but induces a count reading ‘a (loaf of)/one piece of white bread’.

This specific construction is lacking in the other Insular Celtic languages as shown in Example (3) from Middle Welsh (with vara ‘bread’ in direct object position without any marker), which suggests that we are dealing with an innovation that was introduced during the development of Breton (cf. already Ernault 1897; Hemon 1975; the Welsh parallel adduced by German 2007: 175 is not pertinent). It is strongly reminiscent of the French so-called “partitive article”, found, e. g., in the literal Modern French translation of Example (1), given in (4):


Middle Welsh

o-r kaff-ant vara ac vn enllyn
since-ptcl get-prs.3pl bread and one condiment

‘since they get bread and one condiment’

(Peniarth 32: 98, 22–23)


Modern French

Achetez pour un sou du pain blanc.
buy.imp.2pl for one sou du bread white

‘Buy … some white bread for one penny!ʼ

The choice of du in the direct objet nominal du pain blanc ‘white bread’ in this example is an indefinite determiner indicating a mass reading of pain blanc, which is conceived of as a substance, the whole nominal denoting sets of sets of portions (cf. Krifka 2013). Note that du can also alternate with the indefinite article un in Example (3), yielding a count interpretation (‘one loaf of white bread’, ‘one sort of white bread’). There is no option of bare arguments, however, in Modern Gallo-Romance, at least not in French varieties (dialectes d’oïl).

In this paper, our aim is to describe the syntactic distribution of Breton a (allomorphs with personal pronouns: anez-/ahan-) and compare it to that of the Gallo-Romance “partitive article” and “partitive pronoun” en, in order to answer our general research question: Are there plausible arguments in favor of (or against) the assumption that the development of Breton a is the result of borrowing from Gallo-Romance?

Two facts speak in favor of this hypothesis: First, from an external perspective, Breton varieties have lived for hundreds of years in close contact with Northern Gallo-Romance ones, both Gallo (cf. Tréhel-Tas 2007) and (North-Western) regional French (see Figure 1 on the next page). Second, from a structural perspective, a derives etymologically from an Insular Celtic preposition meaning ‘from, of’ (cf. Widmer 2017: 226), just like the first part of the “partitive article” du/de la/des in Modern French (Latin: DE, ‘of’; cf. Stark 2008, Stark 2016).

Following Poplack and Levey (2009), however, we will carefully try to avoid a frequent shortcoming of language contact studies, i. e., comparing the structures under scrutiny too superficially and concluding too hastily that their existence can be explained by borrowing. One condition that Poplack and Levey (2009: 298) postulate for justifying language contact as an explanation of innovations is fulfilled, albeit only weakly: Breton does not seem to have developed the a-structures before 1500. By that time, attestations of the “partitive article” in Gallo-Romance are found, but it is not yet very widespread and certainly not yet fully grammaticalized (this only became the case by the seventeenth century or even later, cf. Carlier and Lamiroy 2014). The second substantial condition, which is the object of our study, is even more difficult to meet, i. e., the demonstration that Breton a-structures “parallel in some non-trivial way the behavior of a counterpart feature in the source” (Poplack and Levey 2009: 298 [emphasis ours]). This means that we will have to look for non-trivial parallels between Breton a and Gallo-Romance de/du, on the form and the function side, keeping in mind that it is always possible for the observed innovations to be due to internal change (drift) rather than to language contact (Poplack and Levey 2009: 297–298).

Our paper is structured as follows: In Section 2, we will describe the syntactic distribution of a-marked constituents in Middle and Modern Breton as far as it is understood at present.[1] Section 3 will try to identify structural parallels in Older and Modern stages of Gallo-Romance varieties in close geographical contact with Breton, and Section 4 will discuss the plausibility of language contact as an explanation for the existence of a-marking in this Celtic language. We will see that even if borrowing cannot be excluded in the early stages of Breton, the Middle and especially Modern Breton facts diverge considerably from the Gallo-Romance ones and do not seem to relate to indefiniteness or “nominal classification” (mass marker) in a broader sense, unlike their Gallo-Romance equivalents, e. g., in Example (4) (see Stark 2008, Stark 2016; Herslund 2008; Cardinaletti and Giusti 2016 for Italo-Romance “partitive articles”). The a-element does not show any behavior of nominal determiners (e. g., gender or number inflection agreeing with lexical elements in noun phrases); moreover, it can combine with personal pronouns, whether free or clitic (cf. Widmer 2017: 222 and 226, Example (8)), a phenomenon that is also found in Welsh, though under differing syntactic conditions. Section 5 will provide a short conclusion and identify some perspectives for future research on the issue of a-marking in Breton and Breton and Gallo-Romance contact.

2 Syntactic distribution of a in Middle and Modern Breton

The Middle Breton corpus, most of which is available to us in electronic form, consists of mostly religious, partly metrical texts in a uniform linguistic shape that does not reflect much of present-day dialectal diversity (cf. Schrijver 2011; Widmer and Jørgensen 2011; Le Berre 2012 for further references). Up to the end of the Middle Breton period (ca. 1650), the use of a-marked arguments in this corpus is clearly limited in terms of token frequency.

Besides locative or ablative arguments (nominal and pronominal), where the use of a [2] ‘from’ is clearly semantically motivated (see examples under (5) below, with a denoting ‘from’, combined with the abstract noun drouc ‘evil’ in (5a)), a may also occur with indefinite mass nominals, like bara ‘bread’, in Example (1) indicating an unspecified amount of a non-given substance. In these contexts, a seems to have lost its original meaning. Furthermore, two allomorphs of a (anez /ahan) are attested in fused constructions with personal pronouns coreferential with a preceding definite nominal, yielding a real partitive meaning (‘of it’; see Example (6), with anez-aff ‘from it’ being coreferential with the prementioned beurage ‘beverage’). Here, a (or its allomorphs) seems to be functionally motivated and preserving its ablative meaning. These two structures concern mainly direct objects and mass nouns or reference to substances (beverages, food). A third attested use of a is particularly striking, as a does not contribute anything semantically to the resulting expression: it appears with indefinite internal arguments (direct objects or the single argument of unaccusatives and presentational constructions, both nominal and pronominal) under the scope of negation (see Examples (7), (8a) and (8b)). In the existential/presentational construction nac eux quet (‘and there is even not’; see Example (7)), which looks very much like Modern French il n’y a pas de X, the use of a-marking is triggered by negative polarity, like in Example (8a) with a personal pronoun and in (8b) with a plural object nominal.

hon diliuret a pep drouc
1pl.obl deliver.imp.2pl a all evil

‘deliver us from all evilʼ

(Heuriou 6; 1576)

na alle enem deliura anez-a
and.not refl free.inf a-3sg.m

‘and was not able to free herself from itʼ

(Beach 58; 1656)

beuraigej dicouraichaff; anez-aff j pan
…beverage awful a-3sg.m when



‘…an awful drink, when one tastes itʼ

(Mirouer 2482; 1575)

nac eux quet a trecte
and.not be.prs emph a mercy

‘and there is no mercyʼ

(Mirouer 1602; 1575)

palamour na ezneuez muy anez-i
since neg know.prs.2sg more a-3sg.f

‘Since you don’t know her any longer.ʼ

(Cathell 34; 1576)

ne desiromp quet a traezou superflu
neg desire.prs.1pl emph a superfluous

‘We don’t desire any superfluous things.’

(Bellarmin 68; 1625)

It is noticeable that in some Modern Breton varieties a ‘of, fromʼ does not occur with nouns anymore, and has been replaced in its locative or ablative function (see Examples (5a) and (5b)) by other prepositions, e. g., døs ‘from, ofʼ in Central Breton (Wmffre 1998: 29). The prepositions that replaced a did not, however, spread to the other “partitive” or indefinite or pronoun-marking functions of a. In Standard Modern Breton (i. e., broadly the literary form of Cornouaillais, Léonais, and Trégorrois Breton; s. Press 1986: 4; concerning the sociolinguistic situation, s. Broudic 1995), which is the main source of our data, a still covers both the locative and ablative as well as the “partitive” or indefinite and pronoun marking functions (Ledunois 2002: 234–235; Favereau 2018: 424–427).

All three non-locative constructions mentioned above, i. e., a-marking of indefinite mass nominals in direct object position, a-marking of personal pronouns with a real partitive reading, and a-marking of any kind of object nominal under the scope of negation, do not abound, but are fairly well attested. In all three constructions, a-marking is available in Middle Breton, but optional. For a-marked nominal internal arguments, the alternative encoding strategy consists of bare nominals, whereas anez/ahan-marked pronouns alternate with object markers that are prefixed to the verb. As for the further development in the period following Middle Breton, the constructions evolved in different directions: a-marking under the scope of negation remains optional (Kervella 1976: 343–344, 396–397; Favereau 2018: 425), whereas a-marked pronominal objects ousted the prefixed object marking strategy in most dialects (see below).

Despite its plausible origin from an expression with partitive semantics (see Example (6)), the further development of the use of a with pronominal arguments diverges heavily already in the Middle Breton period (see Example (8a)). Most importantly, a-marked pronominal internal arguments without any partitive semantics, a completely unattested phenomenon in Romance, increasingly expand and start to replace prefixed person-number-gender markers. In Early Modern Breton, both constructions still co-occur, as is shown in the examples in (9a) and (9b), which both come from the same text. In (9a), the pronominal 1st sg. direct object is expressed with a prefix on the verb (-m-), whereas in (9b) the 3rd pl. direct object is marked with the innovative a(n)-.

neuse e-m-jnstrufed
then ptcl-1sg-instruct.fut.2pl

‘then you will give me instructionsʼ

(EN 1376; late 18th)

e tispinfes ane
ptcl spend.subj.2sg a.3pl

‘you would spend them (sc. all the écus)’

(EN 804; late 18th)

In the late eighteenth century text the examples under (9) are taken from, a theatre play, a-marking for pronominal direct objects (19/98 instances) is still clearly outnumbered by prefixed person-number-gender markers, which occur in 79 out of 98 examples (cf. Dottin 1914: 216–217). During the nineteenth century, prefixed person-number-gender markers are superseded by a-marked pronouns in almost all Breton dialects except for Vannetais (cf. Figure 2; Hemon 1975: 114; Favereau 2018: 109).

Figure 1: 
Breton and Gallo-Romance varieties (Bodlore-Penlaez and Kervella 2011: 58).

Figure 1:

Breton and Gallo-Romance varieties (Bodlore-Penlaez and Kervella 2011: 58).

Figure 2: 
ALBB map 288 showing the dialectal distribution of prefixed person-number-gender markers for direct objects (areas inside the red lines: almost exclusively in Vannetais in the southwest) and a-marking (rest of the Breton speaking area) in the mid twentieth century.

Figure 2:

ALBB map 288 showing the dialectal distribution of prefixed person-number-gender markers for direct objects (areas inside the red lines: almost exclusively in Vannetais in the southwest) and a-marking (rest of the Breton speaking area) in the mid twentieth century.

Finally, a-marked pronouns may be used as subjects of intransitive (unaccusative) verbs and even sometimes as pronominal subjects and direct objects of transitive and subjects of predicative constructions (Ernault 1897; Timm 1985; Ledunois 2002; Schapansky 1996: 105–121). Cf. the postverbal 3rd sg. fem. pronoun anezhi in subject function (of an unaccusative verb) in (10a) and the 3rd Pl. anezo (of a transitive verb) in (10c). Note that a-marked subjects predominantly occur in negative clauses, but not exclusively (Stephens 1993: 372; Schapansky 1996: 110–111; Wmffre 1998: 26; German 2007: 175) and are also widely attested in Vannetais (e. g., Sarzeau, Ernault 1878: 59; Saint Ivy, German 2007: 177). The 3rd sg. masc. pronoun occurs in direct object function in (10b), and the 3rd Sg. masc. subject pronoun anezhañ in a predicative construction in (11). From Middle Breton onwards, the form anezy (3rd sg. fem.) also serves as non-referential expletive subject pronoun with e. g., diurnal verbs, as illustrated in (12) (Ernault 1897; Favereau 2018: 427).

a. Ha e chome anezhi un tammig
and ptcl stay.pst.3sg a. 3sg.f a little_piece

‘and she stayed a little whileʼ

(Modern Breton, Cornouaille; Stephens 1993: 372)

b. an dra-ze a lazo anezañ
the thing-there ptcl kill.fut.3sg a. 3sg.m

‘that thing will kill himʼ

(Modern Breton; Trépos 1980: 248)

c. ne rafent ket an dra-se anezo
neg do.subj.3pl neg the thing-that a. 3pl

‘They wouldn’t do that!ʼ

(Modern Breton, Cornouaille; Trépos 1980: §444)

ur brezhoneger ez eus anezhañ
a speaker.of.Br. ptcl be.prs.ind.3sg a.3sg.m

‘He is a speaker of Breton.ʼ

(Ledunois 2002: 237)

hogos dez eu anezy
close day be. prs.ind.3sg a.3sg.f

‘It will soon be morning.ʼ

(Middle Breton; Pa. 3742)

All these constructions around the grammaticalized a (anez etc.) with personal pronouns, i. e., definite, given arguments, without partitive semantics, are completely unknown in Gallo-Romance, and will not be taken into account any further in what follows.

3 Parallel structures in old and modern Gallo-Romance?

Before discussing potential source structures for the Breton facts described in Section 2, a word is in order on the language contact situation in Brittany. Breton has been in close contact with two different types of Gallo-Romance varieties since the settlement of Brittany by British migrants starting in the fifth century AD (cf. Tréhel-Tas 2007: 29), i. e., non-dialectal French and “le gallo” (< Breton gall: ‘foreign’),[3] a French (oïl) dialect spoken in Haute-Bretagne and highly endangered nowadays (only old speakers, no monolingual speakers). Older stages of Gallo are only rarely attested (some lexical elements in medieval and Renaissance documents). Furthermore, Gallo has never been standardized, and until recently, almost no systematic modern linguistic descriptions of this variety were available (but see Deriano 2005; Auffray 2012; and the projects CREDILIF and SyMiLa, where Nicolas Guillot is in charge of syntactic fieldwork in Gallo varieties). This means that evidence is very sparse or lacking for the old dialectal Gallo-Romance structures which potentially might have been borrowed or imitated by Breton speakers around 1100–1500, and that we have to rely, on the one hand, on diachronic accounts of the French “partitive article”, and, on the other, on modern descriptions of parallel elements in Gallo.

3.1 Indefinite de

Let us start with the most frequent occurrence of French “partitive articles”, i. e., its obligatory use in direct object position with indefinite mass nominals, as in Example (4). Semantically, there is no given set that the du-marked constituent is a subset of. Even if de in du (= de + le) means literally ‘of’, it does not perform any partition – du in these structures is a simple indefinite determiner and the du-constituent is not a PP (see Ihsane 2008, Ihsane 2013 for syntactic evidence, e. g., the impossibility of extractions from real PPs vs. possible extractions from indefinite du-constituents). In contrast to its Breton equivalent in (1), possibly also indefinite as to its semantics, only the composition or fusion of the original preposition de ‘of’, ‘from’ and the definite article (le/la/les), yielding the inflected (gender/number) indefinite determiner du/de la /des, is grammatical in this structure in Modern French:

a. *Achetez de pain!
buy.imp.2pl of bread
b. Achetez du pain!
buy.imp.2pl part bread

‘Buy (some) bread!’

In Modern Romance, we find the structure in (13a) in some varieties of Francoprovençal, spoken nowadays in South-Eastern France, some parts of French-speaking Switzerland and the Aosta Valley in Northern Italy (Kristol 2014), and in Gascon and Provençal, i. e., Occitan varieties from Southern France (cf. Bossong 2016: 69), where any kind of (geographical) contact with Breton is, however, highly implausible. See Example (14) from Provençal varieties, where the plural direct object nominal oulivas ‘olives’ only combines with the uninflected element de:



Vòli d’ oulivas
want.prs.1sg of olives

‘I want (some) olives.’

In Late Latin we do find superficially similar structures with de – but almost exclusively in real partitive structures, i. e., denoting an unspecified quantity of a given, specified and often prementioned set (cf. Carlier and Lamiroy 2014: 489–494). A look back at older stages of French leaves structures like the one in (13a) almost unattested (see Carlier 2007: 8); what is found in that period is bare nouns with mass interpretation and also bare plurals (cf. Foulet 1958 [1919]: 61–62, 72, but see below, Foulet 1958 [1919]: 75). The so-called “partitive article” du is, according to, e. g., Brunot (1899: 377), “très rare dans les anciens textes” [‘very rare in the old texts’], and we find structures comparable to Modern French only from the end of the fourteenth century onwards (cf. Carlier 2007: 35, for a quantitative analysis of Old and Middle French translations of a falconry treatise, where the “partitive article” is very sporadically attested by the end of the fourteenth century, cf. also Jensen 1990: 127–129). Its grammaticalization started first in the plural, in the thirteenth century, only later in the singular (cf. Carlier and Lamiroy 2014: 493–495) and was not completed until classical French (by the seventeenth or eighteenth century). Finally, Modern Gallo varieties do not possess structures like the one in (13a), but they do have a “partitive article”, morphosyntactically and functionally completely parallel to Modern French (cf. e. g., Auffray 2012: 53).

In sum, there is no morphological similarity between Breton a in Example (1) and the Gallo-Romance structures (except the almost synonymous preposition), and as we do not know much about its semantics, we have to refrain from postulating a structural influence of Gallo-Romance on these Breton structures. The available Middle Breton data show however a in postverbal position with mass nouns, with indefinite internal arguments, which alternates freely (?) with bare nominals, and that is pretty much the picture we have for du in Old French.

3.2 “Partitive” en

Examples like (6) or rather their French translation evoke another “partitive” element of French grammar, the clitic en (cf. Ihsane 2013). In Example (6), we find a “real partitive” (Koptjevskaja-Tamm 2001), i. e., the initial set (‘an awful beverage’) is given, and Breton anez-aff or French en denote an (unspecified) amount of this set, i. e., a partition operation takes place conceptually. But here is where the parallels end: French en, deriving from Latin inde ‘from there’ is not inflected for gender or number (like the Latin adverb); it is a morphological simplex and not prepositional in nature at all. Anez-aff, in contrast, means literally ‘from-it’; it is the transparent combination of the element a (or its bound allomorph anez-) plus the respective anaphoric personal pronoun, coindexed and agreeing in gender and number with the preceding nominal. Contrary to en, this combination is always referentially definite, maybe the most important difference.

In order to evaluate a possible functional influence on the anez-aff structure by Gallo-Romance en, let us quickly recapitulate the functional range of en in Modern French. First, en replaces real locative de-PPs, meaning ‘from’:

Il rentre du travail .
he return.prs.3sg from.the work
Il en rentre.
he therefrom return.prs.3sg

‘He returns from work.’ ‘He returns therefrom.’

Second, en replaces de-complements of verbs and adjectives:

a. Il est fier de ses enfants . –
he be.prs.3sg proud of his children.

‘He is proud of his children.’

b. Il en est fier.
he en be.prs.3sg proud

‘He is proud of them.’

Il parle de sa femme.
he speak.prs.3sg of his wife
– Il en parle
he en speak.prs.3sg

‘He speaks of his wife.’ – ‘He speaks of her.’

These two functions have nothing to do semantically with partitivity (but feed classical analyses which see en as replacing de-PPs in general, cf. e. g., Kayne 1975) and are certainly not related to the anez-aff-structure in (Middle) Breton like the ones illustrated in Example (6). They rather remind us of the prepositional objects in (5). En is named “partitive” because it can substitute for prepositional and nominal elements in partitive, pseudo-partitive and indefinite object nominals (or internal arguments of unaccusative verbs). In real partitive constructions, en replaces definite de-complements of nouns:

J’ai bu un verre
I-have.prs.1sg drink.ptcp a glass
de ce vin
of this wine.

‘I drank a glass of this wine.’

J’en ai bu un verre _.
I-en have.prs.1sg drink.ptcp a glass

‘I drank a glass of it.’

This function comes close to that of anez-aff in Example (6), without, again, any structural parallel; note that – aff is a resumptive pronoun of the preceding constituent, i. e., (6) presents the most common and unmarked literal translation of ‘of it’ in (Middle) Breton.

Even more important for evaluating a potential borrowing phenomenon from en to anez-aff, however, are cases where en replaces (parts of) indefinite nominal constituents, as in the following examples: either NPs without their determiners (verre de vin ‘glass of wine’, or vin très bon ‘very good wine’) or du-nominals, with the indefinite mass determiner discussed in Section 3.1. In neither of the following examples does en denote (or replace) an (unspecified) subset of a given set; “partitive clitic” or “partitive pronoun” is thus a misnomer:

J’en ai bu un _ (= verre de vin)
I-en have.prs.1sg drink.ptcp one (glass of wine)

‘I drank one (glass of wine).’

J’en ai bu _ (= du vin) (mass)
I-en have.prs.1sg drink.ptcp (wine)

‘I drank some (wine).’

J’en ai bu un _ (= vin très bon)
I-en have.prs.1sg drink.ptcp one (very good wine)

‘I drank one (very good wine).’

Note that (20) is in fact ambiguous: without any prementioned wine, it can mean ‘I drank wine’, en being indefinite in this case; but in a different context, it can also mean ‘I drank of this wine’, du and en being in fact homonymous: indefinite mass determiner/pronoun or the fused morpheme of the preposition de plus the definite article le, also synchronically. Translated literally, we get for (20): ‘I some have drunk’; with a prementioned wine, we get: ‘I en drank’, en being functionally equivalent here to ‘of-it’.

These data lead immediately to reject analyses which see en generally as a de-PP replacing pronoun (see Ihsane 2013), because it replaces constituents of different category and size (articleless NP as a complement of un in [19] and [21]: verre de vin/vin très bon; complete indefinite mass object nominal in [20]: du vin). For more details on the ongoing debate about the actual constituency of en-replaced elements in French syntax see Shlonsky (2014), Gerards and Stark (forthcoming).

As for the historical stages of French, this large array of functions of en seems to have been deployed after the Old French period – at least, that is the state-of-the-art, as nothing can be found in the literature on the precise development of en replacing de-PPs in real partitive constructions (see Examples (18), (6)) or (different parts of) indefinite nominal internal arguments (see Examples (19)–(21)). In contrast to this, en replacing prepositional, often locative arguments (with de; already in the thirteenth century also other de-complements: parler de – en parler, see Examples (15], (16) and (17)) or causal adjuncts with de, is attested in Old French. This seems to have been its only function until at least the thirteenth century (cf. e. g., Kaminska 1965, but also Foulet 1958 [1919]; Jensen 1990).

Modern Gallo varieties, finally, behave exactly like Modern French (cf. e. g., Auffray 2012: 75). Thus, again, very few functional parallels (in fact only one semantic similarity, replacing a definite set from which an unspecified amount is taken) can be stated between (Middle) Breton and (Modern) French or Gallo: the anez-aff construction and en are completely different as to their etymology, syntax and morphology. In fact, any language possessing a preposition meaning ‘of’ and combining it with a personal pronoun would translate en in examples like (18) by the combination of both: ‘of it’.

3.3 De under the scope of negation

The Modern French equivalent of Example (8b) (see also (7)) looks completely parallel, for the first time in our discussion, to the (Middle) Breton structures:

Et nous ne désirons pas de choses
and 1pl neg desire.prs.1pl neg de thing.f-pl



‘And we do not desire superfluous things.’

In fact, de is obligatory in Modern French under the scope of negation with postverbal indefinite internal arguments following negative adverbs such as pas/point/plus ‘not’, ‘not at all’, ‘no more’ etc. (cf. Rowlett 1998), irrespective of their semantic class (mass, count, abstract, animate) or morphological features (gender/number). The only prerequisite to be fulfilled is reference type: the internal argument has to be indefinite; definite internal arguments take the determiner they also have in affirmative contexts (Nous désirons les choses superfluesNous ne désirons pas les choses superflues ‘We (do not) desire the superfluous things’).

Historically, this de, quite unique in Romance (it does not exist, e. g., in Northern Italian varieties and standard Italian, which both feature a “partitive article” comparable to Modern French), starts to be found occasionally under the scope of negation in the thirteenth century, but only with point (Foulet 1958 [1919]: 74–76, 268–271 with internal arguments of so-called “fragmentative predicates” like EAT and DRINK). There are no attestations with pas until the fourteenth century, only with the preverbal negator ne, which was still able to express sentential negation alone in Old French:

Ne mengiez de char ne
neg eat.imp.2pl de meat neg.comp
ne bevez de vin
neg drink.imp.2pl de wine

‘Don’t eat meat or drink wine.’

(La Queste de Saint Graal, 1225–123, 129.14; cited in Jensen 1990: 129)

Structures like these could indeed have influenced Middle Breton varieties and given rise to a case of pattern borrowing. This hypothesis is weakened, however, by the facts in Modern Gallo varieties, which like Northern Italian varieties or standard Italian keep the respective indefinite determiners of the internal argument in affirmative sentences under the scope of negation, even the so-called “partitive article” (cf. e. g., Auffray 2012: 53), with an unaltered indefinite mass meaning in Gallo:

I manje pâs de la che
3sg.m eat.prs.3sg neg part meat

‘He does not eat meat.’

Table 1 gives a synthesis for a- or de-marking of internal arguments under the scope of negation.

Table 1:

A- or de-marking under the scope of negation.

Obligatory under scope of negation Inflected for gender/number like PA Preference for semantic class Postverbal position in non-dislocated structures
Mod. French + +
Old French +
Gallo + + +
Mid. Breton +
Mod. Breton +

In sum, the most parallel structure between (Middle) Breton and French (but not Gallo!) leaves us with some doubts about similarity in morphosyntactic regularity and degree of obligatorification, let alone the low probability for (pattern) borrowing in argument marking in marked contexts, i. e., under the scope of negation.

4 Discussion: Pattern similaritiesand divergencies

In order to assess the plausibility of borrowing as an explanation of the a/anez-structures in Breton, we start from the similarities discovered in Sections 2 and 3 and will then discuss differences in function and morphosyntactic distribution.

A first common point is the distribution of our elements: Breton a + NP,[4] a similar to French de under the scope of negation and en are all found only with internal arguments (i. e., direct objects or P arguments in an alternative terminology, see footnote 4, S arguments only of unaccusatives, and the argument of presentational constructions).[5] A second one is a partial overlap in function between a with mass nouns and du (not de!) with mass nouns in direct object position, marking both indefiniteness and being optional in Old French and Middle Breton. The anez-personal pronoun combination and en overlap, thirdly, in one specific function: anez + personal pronoun denotes the specific substance (PP, ‘of it’) in real partitive constructions in Middle Breton (end of sixteenth century; Example (6)), one of the functions of Modern French en (Example (17)), though we lack information on its precise history and first appearance. The fourth parallel is the behavior of a and de under the scope of negation (see Section 3.3).

As for the differences, out of three potential Gallo-Romance sources for the Breton a/anez-element, the last one (see Section 3.3) is the most convincing, i. e., de under the scope of negation, which is attested since the fourteenth century. Note, however, that this de does not convey any partitive meaning at all; in fact, its semantic contribution to the sentence is null (and it is incompatible with pronouns). Moreover, we know very little about the morphosyntax of a in negative contexts for Middle and Modern Breton (where it is most probably always optional, see Table 1). The main reason discussed in the literature for the grammaticalization of de with indefinite internal arguments even under the scope of negation is nominal morphology, more precisely the complete loss of overt number marking in French (as opposed to other Gallo-Romance varieties, see Carlier and Lamiroy 2014; Stark 2016; and Gerards and Stark forthcoming for more detailed information). It is a kind of “default noun marker” and not linked at all to partitivity. The function of Breton a in these contexts remains enigmatic to date and will have to be studied in fine-grained corpus analyses.

The first construction, de with indefinite internal arguments (Section 3.1), has been attested with a certain frequency only since the fifteenth/sixteenth century (cf. Carlier and Lamiroy 2014: 493497), where it is already combined with the definite article to result in the so-called “partitive article”, an indefinite mass determiner. Both chronologically and morphologically, there is little overlap with the Breton data exemplified by Example (1): no gender-number marking on a, unclear distribution of this particle, and unclear grammaticalization status.

Finally, en with real partitive constructions (see Section 3.2), functionally, but not morphosyntactically, parallel to Example (6), is not mentioned in grammars of Old French, which only describe the locative/adverbial/indirect complement function by the time relevant for Gallo-Romance–Breton contact – and at the time of writing, there is no research available on the “partitive” or indefinite en in older stages of French or Gallo.

A look at the diachronic spreading of the initial locative prepositions a (and its allomorphs) and de (plus definite article) to additional contexts where a/anez and de-(and en)-elements are found today reveals heavily diverging developments: Gallo-Romance extends de + definite article and en from initially real partitive constructions (syntactically still prepositional/adverbial constituents) to pseudo-partitivity and finally indefiniteness marking (plus some sort of “nominal classification”), yielding reanalyzed nominal constituents (cf. Carlier and Lamiroy 2014), whereas a seems to never have fused with (definite) determiners and to have stuck to bare nominal internal arguments, also under the scope of negation (cf. Widmer 2017: 227, Table 5 therein). Its grammaticalization as the first part of complex non-partitive pronominal arguments, always definite ones, is not (Gallo-)Romance in nature and is most certainly due to a rearrangement of the Breton pronominal system, which loses affix-like person-number-gender markers in most varieties and sets up a new paradigm of non-clitic pronominal elements.

5 Conclusion

In order to answer our overall research question: Do we find plausible arguments in favor of (or against) the assumption that the development of Breton a is the result of borrowing from Gallo-Romance?, we first examined the distribution of a, originally meaning ‘of’, ‘from’, in Middle and Modern Breton, before analyzing seemingly parallel structures with de in Gallo-Romance. The reasons why we started from assuming language-contact as a reason for the a-structures coming into being are, on the one hand, the absence of parallel structures in other Celtic languages, and, on the other hand, the close geographical contact of Breton with Gallo-Romance varieties. We then tried to identify parallels in Gallo-Romance varieties in contact with Breton, looking for structural and functional similarities of Gallo-Romance constructions with de (and the pronominal en, deriving from the Latin adverb inde). Apart from the fact that in Gallo-Romance, two different etyma, one for nominals, and one for pronominals, correspond to a in different contexts, our analysis and discussion of the facts showed relatively little overlap and historically stable differences in morphology (no inflection marking on a and its allomorphs in nominal arguments; person-number-gender information in a-fusions with personal pronouns vs. uninflected, adverbial en). Semantically, only the indefinite structure exemplified in (1) is equivalent to the French one in Example (4). Functional parallels in real partitive constructions such as (6) to Gallo-Romance en are trivial in the sense of Poplack and Levey (2009), as the latter is normally to be translated by ‘of it’ (= the literal equivalent of Breton anez-aff) in any language possessing a preposition ‘of’ and personal pronouns.

The closest parallels are found under the scope of negation, where Breton and French use the former prepositions a/de with any type of internal postverbal nominal arguments, the only condition being indefiniteness. This fact needs further examination and a detailed syntactic and semantic account of the Breton structures. Later developments in the Breton pronominal system (grammaticalization of the a-allomorphs with personal pronouns in most varieties, cf. Widmer 2017; Figure 2) are foreign to (Gallo)-Romance and probably linked to internal changes in the Breton pronominal system.

Though the singularity of the Breton facts in the concert of Celtic languages and the close contact to Gallo-Romance are striking and make the pattern borrowing hypothesis quite plausible, we have to keep in mind the following facts: i. We did not manage to identify many non-trivial parallels between Breton a and Gallo-Romance de/en; ii. The use of a seems subject to variability (which is not the same as contact-induced language change) with nominal arguments; iii. The development of a with pronouns diverges completely from Gallo-Romance. What is noticeable is the specific and isolated development of a in Breton as opposed to other Insular Celtic languages.

What has to be done now is to explore the syntax and semantics of Breton a, both diachronically (corpus studies, potential correlations with other language internal changes) and for Modern Breton (fieldwork) in order to evaluate its function in the grammar of Breton. This function might have been created via language contact, imitating Gallo-Romance partitive structures, but has subsequently taken its very own directions. Diachronic work on French de/du has shown that language internal changes inside nominal morphology (Carlier 2007; Stark 2008; Carlier and Lamiroy 2014; Stark 2016) led to the obligatorification of an otherwise freely available structure meaning ‘of’, ‘from’ (possible in almost any language) – corpus based diachronic analyses of the development of en are still missing to date. It is possible that the introduction of a-marking with certain types of arguments could have happened in Breton without any influence of Gallo-Romance, and given the fact that prepositions are regularly turned into argument markers, it is not necessary to assume the influence of Gallo-Romance on the grammaticalization of a – but it remains a thought-provoking idea worth pursuing.


This research was possible through the financial support of the URPP Language and Space from the University of Zürich. We are grateful for many helpful comments from the audience of the international workshop Partitivity and language contact, especially Ilja Seržant and two anonymous reviewers. All shortcomings are, of course, of our responsibility.

Data Sources

ALBB=Le Roux, Pierre. 1924. Atlas linguistique de la Basse-Bretagne. Rennes: Plihon et Hommay.Search in Google Scholar

Beach=Ar Speret Santel, Bernard. 1656. Ar veac’h devot hac agreabl evs a Perc’herinet santes Anna e Gvenet. Montrovlez: Nicolas dv Brayet.Search in Google Scholar

Bellarmin=Bellarmin, Robert. 1625. Declaration abvndant eves an catechism, hac an Doctrin Christen. Montroulles: Allienne.Search in Google Scholar

Cathell=Ernault, Émile. 1887. La vie de sainte Catherine. Revue Celtique 8. 76–95.Search in Google Scholar

CREDILIF. (accessed 9 February 2019).Search in Google Scholar

Dottin, Georges. 1914. Louis Eunius ou le Purgatoire de saint Patrice (suite). Annales de Bretagne et des pays de l’Ouest 30. 214–233.10.3406/abpo.1914.1416Search in Google Scholar

EN=Dottin, Georges. 1911. Louis Eunius ou le Purgatoire de Saint Patrice. Paris: Champion.10.3406/abpo.1911.1379Search in Google Scholar

Heuriou=Stokes, Whitley. 1846. Middle-Breton hours. Calcutta.Search in Google Scholar

Mirouer=Ernault, Émile. 1914. Le Mirouer de la mort. Poème breton du XVIe siècle. Paris: Champion.Search in Google Scholar

Pa.=Le Berre, Yves. 2011. La passion et la résurrection bretonnes en 1530. Suivies de 3 poèmes, Tremenuan an ytron guerches Maria, Pemzec leuenez Maria, Buhez mab den. Brest: C.R.B.C.Search in Google Scholar

Qu.=Quiquier de Roscoff, Guillaume. 1632. Dictionnaire et colloques françois et Breton, Vols. 3. Morlaix: G. Allienne.Search in Google Scholar

SyMiLA. (accessed 9 February 2019).Search in Google Scholar


Auffray, Régis. 2012. Chapè chapiao. Grammaire de Gallo. Rennes: Rue des Scribes.Search in Google Scholar

Bodlore-Penlaez, Mikael & Divi Kervella. 2011. Atlas de Bretagne/Atlas Breizh. Spézet: Coop Breizh.Search in Google Scholar

Bossong, Georg. 2016. Classifications. In Adam Ledgeway & Martin Maiden (eds.), The Oxford guide to the Romance languages, 63–72. Oxford: Oxford University Press.10.1093/acprof:oso/9780199677108.003.0006Search in Google Scholar

Broudic, Fañch. 1995. La pratique du breton de l’Ancien Régime à nos jours. Rennes: Presses Universitaires. de Rennes.Search in Google Scholar

Brunot, Ferdinand. 1899. Précis de grammaire historique de la langue française. Paris: Masson et Cie.Search in Google Scholar

Cardinaletti, Anna & Giuliana Giusti. 2016. The syntax of the Italian determiner dei. Lingua 181. 58–80.10.1016/j.lingua.2016.05.001Search in Google Scholar

Carlier, Anne. 2007. From preposition to article. The grammaticalization of the French partitive. Studies in Language 31(1). 1–49.10.1075/sl.31.1.02carSearch in Google Scholar

Carlier, Anne & Béatrice Lamiroy. 2014. The grammaticalization of the prepositional partitive in Romance. In Silvia Luraghi & Tuomas Huumo (eds.), Partitive case and related categories, 477–519. Berlin & Boston: De Gruyter Mouton.10.1515/9783110346060.477Search in Google Scholar

Deriano, Patrik. 2005. Grammaire du gallo. Ploudalmézeau: Editions Label LN.Search in Google Scholar

Ernault, Emile. 1878. Le dialecte Vannetais de Sarzeau. Revue Celtique 3. 47–59; 232–239.Search in Google Scholar

Ernault, Emile. 1897. Etudes bretonnes X: Sur les pronoms. Revue Celtique 18. 199–211.Search in Google Scholar

Favereau, Francis. 2018. Grammaire du breton contemporain. Yezhadur ar brezhoneg a-vremañ, 2nd edn. Morlaix: Skol Vreizh.Search in Google Scholar

Foulet, Lucien. 1958 [1919]. Petite syntaxe de l’ancien français, 3rd edn. Paris: Champion.Search in Google Scholar

Gerards, David & Elisabeth Stark. forthcoming. Why ‘partitive articles’ do not exist in (Old) Spanish. In Tabea Ihsane (ed.), Bare Nouns vs. ‘Partitive articles’: Disentangling functions (Syntax and semantics). Leiden & Boston: BrillSearch in Google Scholar

German, Gary. 2007. Language shift, diglossia and dialectal variation in western Brittany: The case of southern Cornouaille. In Hildegard Tristram (ed.), The Celtic languages in contact, 147–192. Potsdam: Potsdam University Press.Search in Google Scholar

Hemon, Roparz. 1975. A historical morphology and syntax of Breton. Dublin: Dublin Institute for Advanced Studies.Search in Google Scholar

Herslund, Michael. 2008. Articles, definite and indefinite. In Henrik Müller & Alex Klinge (eds.), Essays on nominal determination: From morphology to discourse management, 27–43. Amsterdam & Philadelphia: John Benjamins.10.1075/slcs.99.04herSearch in Google Scholar

Ihsane, Tabea. 2008. The layered DP in French: Form and meaning of French indefinites. Amsterdam & Philadelphia: Benjamins.10.1075/la.124Search in Google Scholar

Ihsane, Tabea. 2013. En pronominalization in French and the structure of nominal expressions. Syntax 16(3). 217–249.10.1111/synt.12003Search in Google Scholar

Jensen, Frede. 1990. Old French and comparative Gallo-Romance syntax. Tübingen: Niemeyer.10.1515/9783110938166Search in Google Scholar

Kaminska, Alexandra. 1965. La valeur des pronoms personnels ‘en’ et ‘y’ dans Aucassin et Nicolette, cantefable du Moyen Âge. Revue de Linguistique Romane 29(113/114). 98–104.Search in Google Scholar

Kayne, Richard. 1975. French syntax: The transformational cycle. Cambridge, MA: The MIT Press.Search in Google Scholar

Kervella, Frañses. 1976. Yezhadur bras ar brezhoneg. Repr. Brest: Al Liamm.Search in Google Scholar

Koptjevskaja-Tamm, Maria. 2001. ‘A piece of the cake’ and ‘a cup of tea’: Partitive and pseudopartitive nominal constructions in the Circum-Baltic languages. In Östen Dahl & Maria Koptjevskaja-Tamm (eds.), The circum-baltic languages: typology and contact, Vol. 2. 523–568. Amsterdam & Philadelphia: Benjamins.10.1075/slcs.55.11kopSearch in Google Scholar

Krifka, Manfred. 2013. Cognitive foundations of measuring and counting and their reflection in determiner systems. Paper presented at the international workshop “Mass and count in Romance and Germanic languages”, University of Zurich, 16–17 December.Search in Google Scholar

Kristol, Andres. 2014. Les grammaires du francoprovençal: L’expression de la partitivité. Quelques leçons du projet ALAVAL. La géolinguistique dans les Alpes au XXe siècle: Méthodes, défis et perspectives. Actes de la Conférence annuelle sur l’activité scientifique du Centre d’études francoprovençales ”René Willien”, Aoste, Bureau régional pour l’ethnologie et la linguistique, 29–44.Search in Google Scholar

Le Berre, Yves. 2012. Entre le riche et le pauvre. La littérature du breton entre 1450 et 1650. Brest: Emgleo Breiz.Search in Google Scholar

Ledunois, Jean-Pierre. 2002. La préposition conjuguée en breton. Lille: Atelier national de reproduction des thèses.Search in Google Scholar

Poplack, Shana & Stephen Levey. 2009. Contact-induced grammatical change: a cautionary tale. In Peter Auer & Jürgen E. Schmidt (eds.), Language and space: An international handbook of linguistic variation, vol. 1. 391–418. Berlin & New York: De Gruyter Mouton.10.1515/9783110220278.391Search in Google Scholar

Press, Ian. 1986. A grammar of modern Breton. Berlin & New York: Mouton de Gruyter.10.1515/9783110884975Search in Google Scholar

Rowlett, Paul. 1998. Sentential negation in French. New York & Oxford: Oxford University Press.Search in Google Scholar

Schapansky, Nathalie S. M. 1996. Negation, referentiality and boundedness in Breton: A case study in markedness and asymmetry. Burnaby: Simon Fraser University dissertation.Search in Google Scholar

Schrijver, Peter. 2011. Middle and early modern Breton. In Elmar Ternes (ed.), Brythonic Celtic – Britannisches Keltisch, 359–429. Bremen: Hempen.Search in Google Scholar

Shlonsky, Ur. 2014. One note on quantitative ne-cliticization. Paper presented at the International Exploratory Workshop “Partitivity in Romance and beyond”, Zürich, 11–13. December (accessed 30 June 2017).Search in Google Scholar

Stark, Elisabeth. 2008. The role of the plural system in Romance. In Ulrich Detges & Richard Waltereit (eds.), The paradox of grammatical change. Perspectives from Romance, 57–84. Amsterdam & Philadelphia: Benjamins.10.1075/cilt.293.04staSearch in Google Scholar

Stark, Elisabeth. 2016. Nominal morphology and semantics – Where’s gender (and “partitive articles”) in Gallo-romance? In Susann Fischer & Mario Navarro (eds.), Proceedings of the VII Nereus international workshop “Clitic Doubling and other issues of the syntax/semantic interface in Romance DPs”, 131‒149. Konstanz: Fachbereich Sprachwissenschaft der Universität Konstanz.Search in Google Scholar

Stephens, Janig. 1993. Breton. In Martin J. Ball (ed.), The Celtic languages, 349–409. London & New York: Routledge.Search in Google Scholar

Timm, Lenora. 1985. Pronominal a-forms in Breton: A discourse-based analysis. Journal of Celtic Linguistics 4. 1–34.Search in Google Scholar

Tréhel-Tas, Nathalie. 2007. Parlons Gallo. Langue et culture. Paris: L’Harmattan.Search in Google Scholar

Trépos, Pierre. 1980. Grammaire bretonne. Rennes: Ouest France.Search in Google Scholar

Widmer, Paul. 2017. Cases, paradigms, affixes and indices: Selecting grammatical relations in middle Breton. In Erich Poppe, Karin Stüber & Paul Widmer (eds.), Referential properties and their impact on the syntax of insular Celtic languages, 217–246. Münster: Nodus.Search in Google Scholar

Widmer, Paul & Anders R. Jørgensen. 2011. An Buhez Sant Gwenôlé. Vienna: Praesens.Search in Google Scholar

Wmffre, Iwan. 1998. Central Breton. Munich: Lincom.Search in Google Scholar

Published Online: 2020-05-20
Published in Print: 2020-05-26

© 2020 Elisabeth Stark and Paul Widmer, published by De Gruyter, Berlin/Boston

This work is licensed under the Creative Commons Attribution 4.0 International License.