Microvariation in verbal and nominal agreement: An analysis of two Lombard Alpine dialects

: In Bregagliotto and Mesolcinese, two Lombard Alpine dialects, feminine plural agreement/concord is marked by the formative - n , a reflex of the 3rd person plural verbal ending. In Bregagliotto, plural - n triggers mesoclisis of the feminine subject clitic in contexts of inversion, whereas in the noun phrase - n behaves as a second-position element marking plural feminine concord. Mesolcinese exhibits verbal gender agreement as the formative - n occurs on the inflected verb whenever a feminine plural subject or the feminine plural object clitic occurs; in feminine plural DPs, - n is attached to any element except the definite article. I argue that the Bregagliotto system emerged when - n was reanalysed as an adjunct pluraliser, whereas in Mesolcinese -n has been turned into a marker of morphophonological concord/agreement.


Introduction
This article deals with agreement in two Italo-Romance (groups of) dialects that are spoken in two valleys of southern Switzerland: the Mesolcina and the Bregaglia valley. Although both -N marking is asymmetric and, at first sight, the asymmetry in one variety seems the mirror image of the asymmetry in the other: in Mesolcinese the suffix -n tends to occur on the rightmost elements in the DP, whereas in Bregagliotto -n usually occurs on determiners, see (1)a.
-In both varieties, -n (which is a reflex of the 3rd person plural verbal ending) occurs on finite verbs. In Mesolcinese -n occurs if and only if either the subject, see (1)b, or the object, see (1)c, are feminine plural. In Bregagliotto, conversely, -n occurs with all 3 rd person plural subjects regardless of their gender, see (1)b, and it never marks object agreement.
(1) a. Mesolcinese, which revolves around the hypothesis that the externalization of agreement features is sensitive to syntactic phases (Chomsky 2001).
The above research questions will be addressed in the light of data from Mesolcinese and Bregagliotto. These Lombard Alpine dialects have been extensively debated in the dialectological literature, most of which focused on the synchronic distribution of -n and its possible etymology, see Ascoli 1875: 269-270, 273-274;Schuchardt 1880:153; Meyer-Lübke 1890(II): §32-33; Salvioni 1902: 139;Keller 1932;Sganzini 1933;Jaberg 1951;Tuttle 1982, Loporcaro 2006Pescarini 2020. This article aims to provide a closer comparison between Mesolcinese and Bregagliotto in the light of data from the aforementioned literature, from syntactic atlases (in particular the AIS), and from my own fieldwork notes. I intend to show that Bregagliotto provides crucial evidence to reach a better understanding of Mesolcinese, which exhibits a more innovative and peculiar system.
From a theoretical standpoint, I argue that the agreement phenomena we find in present-day Mesolcinese and Bregagliotto originated from the syntactic reanalysis of a morphophonological irregularity that is still visible in Bregagliotto. In Bregagliotto the plural verbal ending -n undergoes metathesis when it is followed by the feminine plural enclitic la, yielding the order Root > la > -n that is unparalleled in the other Italo-, Gallo-, and Rhaeto-Romance dialects. I argue that this morpho-phonological irregularity triggered a sequence of three cascading changes: First, -n was reanalysed as an adjunct PLURALISER (Wiltschko 2008;Dobrovie-Sorin 2012;Cyrino & Espinal 2020), which is not incorporated by the verb.
ii) The PLURALISER is merged with Person probes such as T and D (Longobardi 2008).
Person probes were thus the pivots triggering the emergence of N-marking in the DP.
iii) Differences between the two varieties result from the status of n: in Bregagliotto n still lexicalises the adjunct PLURALISER, whereas in Mesolcinese it externalises number agreement via morphophonological concord (Cyrino & Espinal 2020). explore a syntactic analysis of the Bregagliotto T and D systems, respectively; §5 argues that in Mesolcinese the formative n is a marker of morphophonological agreement/concord; §6 elaborates on the other properties that distinguish Mesolcinese from Bregagliotto: gender and object agreement on finite verbs; §7 concludes.

The genesis of N marking
The table in (2) compares the distribution of -n in the clause (DPs will be examined in §5): in Mesolcinese, -n is suffixed to the verb whenever a feminine plural subject or object clitic occurs.
Conversely, in Bregagliotto -n is the regular ending of all 3 rd person plural verbs, regardless of the gender of subjects and objects.
(2) Distribution of n in the clause: ii) in Bregagliotto, finite verbs never agree in gender with object clitics (as in the vast majority of Italo-Romance dialects); iii) Bregagliotto exhibits no gender verbal agreement as verbal morphology is regular (n is the 3rd person plural ending of the verb, regardless of the gender of the subject).
The only irregularity displayed by Bregagliotto is that the subject clitic la 'her/they.F' precedes the formative -n when it is enclitic, yielding a pattern of mesoclisis, cf. (2)b. As shown  First of all, mesoclisis of subject clitics is unparalleled and, from a syntactic standpoint, it is completely unexpected: it is a widely held view that enclisis results from movement of the inflected verb to the left periphery (say, to Rizzi's 1997 Fin), crossing the position harbouring the subject clitic: Since mesoclisis should be impossible under (6), to account for Bregagliotto we must invoke a mechanism that "overrides" (6) when the subject is feminine. The most probable explanation is that mesoclisis was initially triggered at PF by metathesis, a phonological change preventing sequences of sonorants. As shown by numerous examples (e.g. dormir > dromir 'to sleep' in (3)), Lombard (Alpine) dialects tend to avoid clusters of sonorants by reversing the order of segments or syllables. Since the feminine plural enclitic la begins with a sonorant, enclisis yielded a marked cluster of consonants (*nl) that was avoided by means of metathesis, see (7)a.
In the masculine, conversely, metathesis did not take place as the sequence formed by the plural ending -n and the masculine plural clitic i is phonologically licit. Consequently, the masculine clitic of Bregagliotto has kept occurring after the agreement ending, see (7)b, as subject enclitics normally do. In the next section, I will therefore argue that metathesis triggered reanalysis (i.e. syntactic rebracketing) of -n as an adjunct PLURALISER. This in turn triggered a series of further changes that made Bregagliotto and Mesolcinese diverge from the other Lombard dialects and from each other. The hypothesis that a marginal irregularity reverberated throughout the agreement

systems of Bregagliotto and Mesolcinese is consistent with the tenets of Longobardi's 2001
Inertial Theory of syntactic change, the gist of which is summarised in the following quote: "syntactic change should not arise, unless it can be shown to be caused-that is, to be a well-motivated consequence of other types of change (phonological changes and semantic changes, including the appearance/disappearance of whole lexical items) or, recursively, of other syntactic changes." (Longobardi 2001: 278) According to this view, which is endorsed by the present work, syntactic change cannot result from the invariable principles of Narrow Syntax, which are by definition inert to change. Innovationsand, consequently, variationis first triggered by reanalysis of interface effects, such as metathesis, which end up modifying the featural endowment of functional elements (where syntactic variation is encoded, according to the so-called Borer-Chomsky Conjecture).
In this respect, it is worth noting that in Lombard Alpine dialects inversion (and, consequently, mesoclisis) was originally much more pervasive than nowadays. In present-day dialects, inversion is found in interrogative clauses, but for centuries northern Italian dialects displayed traces of verb-second syntax of the kind that is still productive in the nearby Rhaeto-Romance varieties spoken in the Grisons (see Fig. 1). This amounts to saying that, in previous diachronic stages, metathesis was more frequent than nowadays and, given its higher incidence, it could have prompted the kind of syntactic reanalysis reconstructed here.

A syntactic analysis of the Bregagliotto T system
In this section, I entertain the hypothesis that Bregagliotto speakers rebracketed sentences including feminine plural clitics in order to accommodate metathesis. As a consequence, the -n co-occurring with the feminine la was reanalysed as an adjunct PLURALISER (Wiltschko 2008).
In origin, the ending -n spelled out T's person and number features that are valued via agree(ment) with the subject. As shown in (9)b, -n is merged in T when T probes a plural subjects such as i 'they.M' (let us assume, for the sake of simplicity, that subject clitics are first merged in v). The formative -n forms a complex head with the verb and, when the latter moves to C, as in (9)e, -n moves along with the verbal root above the subject clitic i, as in the other Romance languages and dialects exhibiting inversion.
(9) a. b. c. d. e. [ Sentences featuring feminine plural subjects hadin originthe same syntactic derivation, until metathesis reversed the order of -n and the feminine enclitic la at PF, yielding a misalignment between syntax and morphology, in (10)f: d. e. f. [ PF Across generations, this irregularity triggered rebracketing: as shown in (11)b-c, n does no longer lexicalise T, but it is adjoined to T. From now on I adopt the symbol ^ (taken from Hornstein & Nunes 2008) to distinguish the inflectional marker n, lexicalising T, from the freestanding PLURALISER ^n that is adjoined to T. The latter does not belong to the constituent that is subject to movement (Hornstein & Nunes 2008: 68), possibly because adjuncts are merged "on a separate plane" (Chomsky 2004: 118; see also Chomsky 2020: 47), and is not c. d. e. f.
[ There are alternative ways to account for the excorporation of -n. Calabrese and Pescarini 2014 explored a Distributed Morphology analysis in which metathesis triggers fission of feature bundles. In a previous version of this work I argued for an alternative analysis based on D'Alessandro's 2016, 2017 analysis of central Italo-Romance dialects that display agreement phenomena analogous to those of Bregagliotto and Mesolcinesei.e. verbal gender agreement and omnivorous agreement (Nevins 2011). D'Alessandro argues that dialects showing gender and/or omnivorous agreement exhibit a complex probe formed by T and an extra probe π. However, the analysis proposed in the present work, where I assume that -n becomes an adjunct, seems to me more elegant.
in nature, i.e. the verbal ending spells out T's agreement features, whereas, with feminine subjects, plural marking is modificational (Wiltschko 2008).

A syntactic analysis of the Bregagliotto D system
Wiltschko 2008: §2 argues that in several languages number marking is not inflectional, but modificational: the plural obtains by adjoining a PLURALISER to the functional spine of the DP. Port. os livros / os livro / *o livros 'the books', In this section, I argue that the hypothesis of an adjoined PLURALISER accounts for the syntax of n in Bregagliotto, which is exemplified in (12) With respect to the data in (12), two main questions need elaboration: i) why ^n behaves as a second position element in the DP; ii) how ^n became a marker of nominal concord.
The hypothesis that number is encoded by a PLURALISER accounts for the second-position syntax of ^n: since the PLURALISER is merged in D, ^n is correctly predicted to follow all kinds of determiners: Adjectives and nouns that precede ^n in (12)d-f are not real counterexamples to the hypothesis that ^n is merged in D. In fact, head movement to D is a rather common phenomenon in determiner-less DPs (Longobardi 1994). In this respect, it is worth noting that many instances of Adj^n or N^n sequences, which nowadays are quite marginal, are found in vocatives, which Espinal 2013 argues to be derived by moving Ns or As to a dedicated position (Voc) above D.
Hence, adjectives and nouns precede -n in (12) (16). Given the constrained distribution of ^n in (12) one can eventually hypothesise that the PLURALISER n is subcategorised for person (I will resume this discussion in § §5-6). (

16) a. [T{Person}^n [ … [nP Root]]] b. [D{Person}^n [ … [vP Root]]]
The idea that D, like T, probes Person is supported by evidence of N-to-D movement of pronouns: unlike nouns, personal pronouns bearing Person features move above prenominal modifiers such as numerals in (17) and adjectives in (18)  To conclude, in this section I argued that the person/number ending -n has been reanalysed as a PLURALISER ^n that is adjoined to T (first) and then to both person probes: T and D. The hypothesis accounts for the extension of ^n to the DP, once we assume that i) D probes Person

4
The displacement of personal pronouns in D takes place to value D's Person feature in the same way in which subject move to spec T.

i)
Multiple agreement: While in Bregagliotto ^n always follows Person probes (and, consequently, occurs once in the clause and once in the DP), in Mesolcinese the formative n occurs on all inflected elements in the clause and in the DP (demonstratives, nouns, quantifiers, past participles, etc.), except the clitic/article la.
The distribution in the DP is shown in table (19) (23) Plural-marking-on-D constraint If X (that is, a pluralized D) c-commands Y (that is, N or A), which in its turn c-commands Z (N or A), plural marking may be overt on X alone, on X-Y, on X-Y-Z, but not on X-Z.
The above constraint, however, accounts neither for Bregagliotto nor for Mesolcinese. In the former, n behaves as a second position element, in the latter n occurs on any inflected element except definite articles.
To conclude, both Bregagliotto and Mesolcinese exhibit systems of partial N-marking that are isomorphic in the clause and in the DP. However, while in Bregagliotto n lexicalises the otherwise abstract PLURALISER (see § §3-4), in Mesolcinese number is externalised via morphophonological concord within a Spell-Out domain. In the latter, n has the same distribution as other inflectional markers such as F.SG -a and is constrained by a parameter along the line of (27). The difference between the two dialects is shown in the following trees (recall that the symbols ^, + signals syntactic adjunction and insertion at PF, respectively): Building on this hypothesis, the next section seeks to account for the participle-like agreement of Mesolcinese finite verbs, which undergo gender and object-clitic agreement. The gist of the next section is that Mesolcinese finite verbs are subject to morpho-phonological agreement whenever they are in the c-command domain of feminine plural object and/or subject clitics.

Mesolcinese: gender and object agreement on T
Verbal gender agreement is very rare in Romance; it is attested in few dialects scattered along the Alps and in central Italy (D'Alessandro 2016(D'Alessandro , 2017. In Mesolcinese, verbal gender agreement arguably emerged when the formative -n marking 3 rd person plural agreement was lost, a phenomenon that is widespread in the nearby Lombard varieties spoken in the basin of the river Ticino (to which the Mesolcina valley belongs, see the map in Fig. 1). In the dialects of Ticino, 3 rd person singular and plural verbal forms became syncretic because the ending -n, marking 3 rd person plural agreement, was lost (Salvioni 1902), e.g. kanta-n > kanta 'they sing' = 'he/she sings'. person ending -n, but spared the homophonous +n marking morphophonological concord/agreement ( §5). Once the full-fledged Person ending -n was lost, 3 rd person plural verbs became syncretic with 3 rd person singular verbs in the masculine, whereas in the feminine the marker +n kept occurring, yielding the present-day pattern of verbal gender agreement: (29) proto-Mes./Breg. Mes.
a. kanta-n > kanta(*-n) 'they.M sing' b. kanta-n > kanta+n 'they.F sing' The change in (29)   The fact that the Mesolcinese +n became a morphophonological marker of number/gender agreement after the person/number element -n had been lost supports the view that morphophonological agreement and person agreement are somehow in complementary distribution. However, the agreement pattern illustrated in (33), in which a 1 st person singular verb agrees with a feminine plural object clitic, does not corroborate Baker's idea that verbs can agree in number and gender if they lose the ability to agree in person.

Conclusions
This article examined the agreement systems of two Lombard Alpine dialects. In both varieties, feminine plural agreement/concord is marked by the formative n, which resulted from the reanalysis of the 3 rd person plural verbal ending.
The verbal morphology of Bregagliotto is quite regular (save for metathesis that occurs when the inverted subject clitic follows the plural formative -n). Concord within the noun phrase is marked by -n, which always occurs after the first element of the phrase.
Mesolcinese is more complicated as it exhibits verbal gender agreement: the formative -n is attached to the inflected verb when either the subject or the object (clitic) is feminine plural. In the noun phrase, n is attached to any element except the definite article.
As for Bregagliotto, I argued that metathesis was eventually reanalysed, yielding an alternation between two homophonous items: the agreement head -n and the adjunct ^n. The latter carries feminine plural features and, unlike -n, it is not incorporated by the inflected verb.
The formative ^n is merged with person probes such as T and D.
In Mesolcinese, conversely, +n does not lexicalise the adjunct pluraliser, but it is a feminine plural marker involved in morpho-phonological agreement/concord. In Mesolcinese +n combines with all inflected categories within a Spell-Out domain, both at the clausal and DP level (Manzini & Savoia 2019). The inflected verb, which, like in other Lombard dialects, shows no distinction between the 3 rd and 6 th person and can no longer undergo inversion, is subject to morpho-phonological, participle-like agreement: as a result, inflected verbs in Mesolcinese are combined with +n when they are in the c-command domain of feminine plural subject and/or object clitics.