resilient grammars: on vo/ov in germanic linguistic islands in northern italy

In this work we consider some residual cases of OV order in Cimbrian and show that this is due to the interaction between verb movement, a language specific property, and the syntax of bare quantifiers. This has consequences on a general theory on the change of the basic word order, since it shows that the passage from OV to VO can involve different structures in different languages depending on other properties, hence it is not possible to trace a common path in the diachronic change for all languages that have undergone this mutation.

In this work we consider some residual cases of OV order in Cimbrian and show that this is due to the interaction between verb movement, a language specific property, and the syntax of bare quantifiers. This has consequences on a general theory on the change of the basic word order, since it shows that the passage from OV to VO can involve different structures in different languages depending on other properties, hence it is not possible to trace a common path in the diachronic change for all languages that have undergone this mutation.
[1] introduction In this work we concentrate on Cimbrian, a Germanic variety spoken in Northern Italy. Cimbrian used to be spoken in the whole Asiago Plateau, the area of the so called seven communities and thirteen communities (among them Giazza, Roana) as well as Luserna in the Trento province, where it is still relatively well preserved and spoken by a community of about 300 speakers. The geographical area Cimbrian used to cover extends between Trento, Verona and Vicenza, but nowadays most varieties are at best residual languages or have totally disappeared from the active competence of the inhabitants of the area. Since the only well-preserved variety still in daily use nowadays is the one of Luserna, we will discuss data from this variety (and compare it to the one found in the literature when necessary), where the attempts to preserve it include courses at school, publications festivals etc. When not otherwise indicated, the data stem from our own field work in Luserna extended over the years. 1 When necessary, we will compare Cimbrian with Plodarisch, 2 another linguistic island spoken in Northern Veneto at the border with Friaul, which has a comparable [1] We thank the collaborators of the Cimbrian Institute and particularly Fiorenzo Nicolussi and Andrea Nicolussi Golo for their help and patience in checking data with us. [2] Also for Plodarisch the data stem from our own field work. We thank the family Kratter for their help. [336] poletto & grewendorf OSLa volume 11(2), 2020 sociolinguistic situation. The reason why Cimbrian is particularly interesting is that, although it has maintained the verb second property up to now, it is possibly the only German dialect with a VO structure, which clearly looks like a case of massive languagecontact induced change. The sociolinguistic situation of Luserna, where speakers have native competence of Cimbrian, standard Italian (which is the so-called 'Dachsprache') and the Trentino/Veneto Northern Italian non-standard varieties spoken in the neighbouring villages lends itself to such a hypothesis. Luserna speakers use Cimbrian within the community (in colloquial situations even outside the family), Trentino/Veneto with speakers from other villages (for instance at the weekly market) and standard Italian in formal situations.
Cimbrian has maintained the V2 property typical of all German varieties (see Grewendorf & Poletto 2011), although with peculiar features, which are similar to those found in Old Italian (see Poletto 2014). However, it is striking that the language is a regular VO language with only some interesting residues of OV, which we will discuss. The aim of this work is to show that a) there is actually no direct evidence that the VO property of Cimbrian is borrowed from Italian b) the VO property of Cimbrian displays interesting differences with respect to the neighbouring Italian varieties, which c) will bring us to conclude that language contact applies to syntax in a peculiar way. So, it might be true that Cimbrian is VO due to the pressure of Italian, but it is certainly not true that Cimbrian has borrowed the type of VO typical of Romance (and in particular of Italian) tout court. This means that syntactic change due to language contact does not simply borrow syntactic properties as they are and transfers them from the lending language to the borrowing language, i.e. syntactic borrowing does not work like lexical borrowing, where the lexeme is taken from one language and used in the other as well. Since syntax is a unitary system where "tout se tient", it would be very surprising indeed that languages just take a piece of a foreign grammar and, so to speak, "swallow it whole". Since the system is an integrated derivational machine, the new property must also be integrated.
One general problem when talking about language contact is that one needs to prove that the change has occurred as a consequence of the contact, i.e. one needs criteria. According to Poplack & Levey (2010), it is only possible to prove that the change of property P has occurred in a language A as the consequence of language contact with a language B when the following conditions are met: (ii) The change must have occurred in an appropriate span of time, for instance, it is not plausible that such a basic property as VO/OV has changed in the span of one generation.
This means that in order to assume that the VO order of Cimbrian has been influenced by Italian, we would have to verify conditions a) and b), i.e. provide incontrovertible historical evidence for a stage previous to the contact where Cimbrian was OV. This is clearly not possible, since the first attestations of Cimbrian we have are from the beginning of the XVII century, and at this point Cimbrian already displays VO. However, we will see that this is not even necessary, since we can show that Cimbrian has a type of VO which is NOT the one of the neighbouring Italian varieties. Even though Italian might have influenced Cimbrian, it has done so in an indirect way, and not by "lending" a piece of its grammar to Cimbrian. Furthermore, Cimbrian has retained some very resilient properties typical of Germanic, like the position of clitic pronouns, which remain in enclisis, and some OV orders with bare quantifiers that are not attested in the neighbouring Italian varieties.
[2] vo vers us ov The first logical move to make when examining the problem of VO versus OV is to establish criteria which clearly define a language as VO or OV. Several proposals have been made in the typological literature (see for instance Dryer 1997, Dryer 2007). The problem with typological criteria is that even standard German does not always fit into them as an SOV language; for instance the fact that SOV languages generally have postpositions is by and large not met by German (which has only a couple of residual postpositions or specific constructions). 4 The same holds for subordinators, which in SOV languages like Lezgian (described in Dryer 2007, p. 62 ff.) or Japanese can occur in sentence final position. This is never the case in standard German. Also criteria like the structure of comparison (see Dryer 2007, p. 63) with standard comparison [3] Notice that also Heine and Kuteva (2005:8) warn against language contact as the cause of linguistic change in absence of strong sociohistorical and linguistic evidence that this is the case. preceding the comparison marker are not met by German. Therefore, using typological criteria is not fine-grained enough to determine the difference between standard German and Cimbrian. 5 Also in the generative literature there are some proposals for criteria to determine the SVO/SOV status of a language, see among others those proposed on the basis of Germanic by Fuß (2005) and Fuß & Trips (2002) according to whom a VO-language must meet the criteria in (1): (1) (i) It shows verbal particles in postverbal position. (i) It shows light elements in postverbal position (pronouns and light adverbials which cannot undergo right dislocation). (ii) It never shows the sequence finite verb -object -main verb.
These criteria are also problematic because they give us the wrong results for Romance languages, which are nonetheless standardly considered as VO languages both in the generative and typological literature.
The first criterion cannot be applied, since Romance languages do not generally display verbal particles. It is however true that precisely in the North-Eastern area, i.e. Veneto and Trentino a small set of prepositions can be combined with the verb also in semantically non transparent contexts, so that one might analyze them as verbal particles (see Benincá & Poletto 2006 for Veneto, Cordin 2011 for Trentino) and it is true that they generally occur after the past participle: (2) [Venetian] El me ga dito su. He to-me has told up 'He told me off' However, it is not clear what the status of these elements is, whether they are already completely grammaticalized as particles or not. Therefore, we can conclude that the first test is not applicable to all languages. Furthermore, and rather paradoxically, these verb+preposition constructions are often assumed to be due to language contact with Germanic, so we would have a somewhat circular criterion. Therefore, we consider this criterion as non-applicable. The second test is also very problematic in Romance, since clitics and weak pronouns never occur in postverbal position. Notably clitics are in proclisis to the inflected verb, not in enclisis, but also weak pronoun as Italian loro studied [5] There is actually a debate in the typological literature concerning the OV status of German, which we will not enter here.

VO/OV IN GERMANIC LINGUISTIC ISLANDS IN NORTHERN ITALY
[339] OSLa volume 11(2), 2020 by Cardinaletti (1998) occur between the auxiliary and the past participle in what could be viewed as a typical OV position. The third criterion is also problematic for Romance: if you consider a language like French, where bare object quantifiers do occur in front of the past participle, one would reach the rather awkward conclusion that French is not a VO language. (3) [French] a.
Je n'ai rien vu. I not have nothing seen 'I have seen nothing' b.
'%Je n'ai vu rien.' 6 Therefore, the situation is rather complex and it is not really possible to use the criteria that have been [3] cimbrian as a vo language: analogies with italian The very first naïve observation that can be made looking at Cimbrian, is that it is clearly a VO language like Romance and unlike German, because the direct object is located on the right of both the auxiliary and the past participle: Haüte die Mome hat gekoaft die Öala al Mercà/kam bir/in die Buteghe. today the mother has bought the eggs at the market/in the bar/in the shop 'Today mum bought eggs at the market/in the bar/ in the shop' b.
Haüte die Mome hat gebäscht di Piattn. today the mother has washed the dishes 'Today mum washed the dishes' c.
*Haüte di Mome hat di Piattn gebäscht. today the mother has the dishes washed Notice that even if Cimbrian is a V2 language which allows for V3 structures with topics, the contrast with German represented in (5) emerges in the fact that the direct object, as well as PPs, occurs after the past participle and not before (see the reverse grammaticality pattern in 4b,c and 5,b,c).
Today has the mother the eggs at.the market/at.the shop bought 'Today mum bought eggs at the market/in the bar/ in the shop' b. *Heute hat die Mutter gespült die Teller.
Today has the mother washed.up the dishes 'Today mum washed the dishes' c. Heute hat die Mutte die Teller gespült.
Today has the mother the dishes washed.up Differently from German, Cimbrian of Luserna also shows the sequence finite verb -main verb -object in embedded clauses introduced by a complementizer, as can be seen in (6). In this sense Cimbrian is also analogous to Romance in having the past participle before direct and indirect objects. Notice that this cannot be analysed as a case of embedded V2 because in German varieties the complementizer bans verb movement (as we have also shown in Grewendorf and Poletto 2011) [Cimbrian] a. I boas ke di Mome hat gekoaft di Öala al Mercà. I know that the mother has bought the eggs at the market 'I know that today mum bought eggs at the marker' b. Dar issese darzürnt obrom gestarn honne i get an Libar in Peatar.
he is-refl angry since yesterday have I given a book to Peter 'He got angry since yesterday I gave a book to Peter' c. S Beibe bo da-r-en hat geet an Liber.
the woman who there-he-her has given a book 'The woman to whom he gave a book' d. Sa homar gevorst zega bas-to tüast hoint abas.
they have-me asked to-say what-you do tonight 'They asked me what you are going to do tonight' If we consider more complex verbal forms, like those occurring with modals, auxiliaries and past participles, we can also observe that Cimbrian still has the same VO order typical of Romance. Furthermore, Cimbrian does not display the phenomenon "infinitive pro participle" typical of German (and Dutch) where under a modal both the main verb and the auxiliary have an infinitival form: OSLa volume 11 (2), 2020 (7) [Cimbrian] a. Sa hom khött ke dar Gianni hat geböllt gian pit se.
'they have said that the John has wanted go with them' 'They said that Gianni wanted to go with them' b. *Sa hon khött ke dar Gianni hat gian geböllt.
'they have said that the John has go wanted' (8) [Italian] a. Mi hanno detto che Gianni ha voluto andare con lei.
To.me have said that John has wanted to.go with her 'They said that Gianni wanted to go with them' [German] b. Sie haben mir gesagt, dass Hans mit ihr hat gehen wollen.
they have to.me said that John has with her has go want 'They said that Gianni wanted to go with them' We can conclude that the order of the verbal cluster is of the Romance type and not of the German type. Furthermore, Cimbrian has another striking property, which is not common to all Romance, but typical of Italian, i.e. Cimbrian has postverbal subjects. (9) [Cimbrian] a. Haüte geat vort dar Giani.
Today goes out the Gianni 'Today Gianni goes out' b. Haüte geata vort niamat.
'Today goes out noone' 'Today nobody goes out' c. I khoch-mar-en i in vormas. ("i" is focused) I cook-me-it I the lunch 'I am cooking myself lunch' (10) [Italian] a. Oggi va via Gianni.
Today goes away Gianni 'Today Gianni goes out' b. Oggi non va via nessuno.
Today not goes aways nobody 'Today nobody goes away' c. Me lo cucino io, il pranzo. me it cook-I, the lunch 'I am cooking myself lunch' [342] poletto & grewendorf OSLa volume 11 (2), 2020 (9) shows that Cimbrian has the possibility of having postverbal subjects of different types, definite DPs, quantifiers and full pronouns, just like Italian. This is rather striking, since Cimbrian is not a pro drop language like Italian and generally the possibility to have postverbal subjects is considered a consequence of the possibility to have pro in the preverbal subject position (see Lenerz 1977, Grewendorf 1989, Haider 1993). [4] resilie nt propert ie s: diff erences between cimbrian and italian [4.1] Enclitic pronouns So far, we have seen that the position of DP objects is on the right of the past participle, a typical VO property. Other types of objects behave differently from regular DPs. This is true both in Germanic languages and in Romance, where clitics maintain the OV order of Latin and occur in proclisis to the inflected verb.
One of the most stable properties of minority German varieties in Northern Italy is the position of pronouns. Although Cimbrian has VO with DP objects, it lacks another typical property which is characteristic of the Romance VO type: it does not have proclitic object pronouns in finite clauses. 7 In Cimbrian, clitics can only follow the finite verb or the complementizer (see (11c) and (11d) as it is the case in standard German: in embedded clauses the pronoun is enclitic onto the complementizer, in main clauses to the inflected verb in C (11) [Cimbrian] a.
Der Tatta [7] There is only one modern Romance variety known to have regular enclisis with inflected verbs, and this is the Borgomanerese (Piedmontese) dialect described by Tortora (2014). The same is true of infinitival clauses, but this is a topic we will not investigate here. It thus looks like Cimbrian displays the most basic characteristics of a VOlanguage but there are some interesting resilient patterns which have apparently not surrendered to the pressure of the Romance grammar.

[4.2] Preparticipial quantifiers
In addition to pronouns, another type of objects that have maintained a conservative pattern are bare quantifiers. In (13)  The option in (14) is not limited to the expression niamat, it represents a more general phenomenon that has to do with the specific nature of the object phrase. This can be seen from the examples (15)  I bil eparummas seng/seng eparummas. I want somebody see/see somebody These observations suggest the generalization that objects represented by quantifiers permit the order finite verb -object -participle, which is otherwise illicit in a VO-language. It looks like this generalization does not concern a specific phenomenon of the VO-language Cimbrian. Another Germanic VO-language which allows preposing of quantified objects is Icelandic. While Old-Icelandic until the 18. century predominantly shows OV-order with various preverbal complements (nominal objects, PP-objects, adverbials, particles etc.), the OV order declines from the 18. century on. In the 19. century, only a few remainders of the old OVorder can be observed (cf. Hróarsdóttir 2000, p. 155 f.). Modern Icelandic has fully developed VO-order in the VP and only allows the sequence [VP auxiliarymain verb -object]. However, there is an important exception to this constraint: While nominal phrases cannot occur in front of the past participle, the preparticipial position is an option for quantified phrases and is obligatory for negated phrases, as illustrated by the examples in (19) and (20): Strákarnir höfðu hent miklu grjóti í bílana. 'the-boys had thrown many stones on the-cars' 'The boys had thrown many stones to the cars' b.
Strákarnir höfðu miklu grjóti hent í bílana. 'the-boys had many stones thrown on the-cars' (Svenonius 2000, p (Trips 2002, p. 203) In early Middle English, which is in the transition from OV to VO, OV-order with a quantifier located in front of the non finite form of the verb, already occurs more often with quantified phrases than with non-quantified NPs. It has been observed (cf. Rögnvaldsson 1987; Svenonius 2000, among others) that several Scandinavian languages, including Icelandic, show a phenomenon known as "Quantifier Movement" e.g. scope, moves a quantified expression leftward across VP-internal material to a VP-external position. As can be seen from the Icelandic example (22), the landing site of Quantifier Movement must be a VP-external position (which according to Fox and Pesetsky 2005, p. 31, is slightly lower than the landing site of Object Shift): Hann mun mikið hafa vilja [lesa __]. he will much have wanted read 'He has wanted to read much' (Trips 2002, p. 203) In Romance languages, a similar phenomenon can be found. Unlike earlier stages of Romance, the modern Romance languages generally show VO-order. However, French also allows preposing of quantifiers, but notice that the expected VO-order is ungrammatical with bare quantifiers: (24) [French] a.
*Je n'ai aucun garçon/pas tout ça vu. *OV I not=have any boy/not all that seen 'I have not seen any boy/all this' b.
Je n'ai vu aucun garçon/pas vu tout ça. VO 'I not=have seen any boy/not seen all that' Cimbrian displays the same restriction found in French, i.e. only bare quantifiers can be found in a pre-participial position: (25) [Cimbrian] a.
I hon gesek kumman Pua. I have seen no boy 'I have seen no boy' b.
*I hon kumman Pua gesek. 'I have no boy seen' We can conclude that a) several languages display a different pattern when the object contains a quantifier. Quantifiers are generally more easily found in OV position than definite objects both in Germanic and Romance. There is a distinction between Icelandic and Middle English on the one hand and French and Cimbrian on the other, since the latter only allow for bare quantifiers to occur in preparticipial position. In standard Italian as well as in [10] In general, it is possible to have a postparticipial bare quantifier only when it is focused, otherwise the normal position is the preparticipial one. Notice that also standard Italian and Veneto display a higher position of bare quantifiers (at least of the universal one tutto/tutti) with respect to normal objects, although this does not result in reversing the word order with respect to the past participle. As already discussed by Cinque (1999) the bare quantifier tutto 'everything' is located on the left of manner adverbs, while definite object DPs and quantified objects are located to its right: This means that the Italian varieties in contact with Cimbrian have indeed a higher position of bare quantifiers, though this is not immediately visible. We conclude that the pre-or postparticipial position of quantifiers depends on the raising point of the past participle. This means that Cimbrian and French have low past participles, while standard Italian and Veneto have higher past participles, but all of them move bare quantifiers to the left.
There is a more general problem here: as we have seen, the Germanic and the Romance phenomenon of quantifier preposing is different: Germanic languages are more tolerant and also quantified expressions can pied-pipe their lexical restriction to the higher position, while this is not the case in French and Cimbrian.
More generally, our observations can be summarized as follows. In addition to the lack of proclitic pronouns in finite clauses, the VO-language Cimbrian shows a further property which looks prima facie atypical in a VO-language: the pre-participial position of quantified objects.
On the other hand, we have seen that the Germanic languages have a slightly different property, since OV is allowed also with quantified expressions and not only with bare quantifiers. This means that the property of Cimbrian is neither identical to the contact varieties, which have VO, nor to the one of the Germanic languages, which prepose also quantified expressions.

[4.3] Separable prefixes
A third property of Cimbrian which represents a phenomenon typical of OVlanguages concerns separable prefixes. In main clauses of the OV languages like German, separable prefixes are located in the right sentence bracket.
In German embedded clauses and with composed tenses, the prefix precedes the verb: Der Vater hat kritisiert, dass die Mutter die Teller abspült. the father has criticized that the mother the dishes washes-pref 'Dad criticized, that mum washes the dishes' b.
Der Vater hat die Teller abgespült. the father has the dishes washed-pref 'Dad has washed the dishes' This is different from the VO-language English, where with so-called "phrasal verbs", which is the VO-counterpart of separable prefixes in German, the nominal elements can occur on the right or on the left of the particle, while the pronouns generally occur on the left of the particle (at least with a specific class of verbs): [English] a.
I gave up the job. b.
I gave it up. c.
What did you give up?
In Cimbrian main clauses, separable prefixes show the postparticipial position typical of VO-languages (35a). However, preparticipial order, which is typical of OV-languages, is also possible (35b). However, when we consider other types of complementizers and not only ke, the impossible order becomes possible. Relative clauses (among others) are introduced by a complementizer of the former type, and in relative clauses, separable prefixes can occur in front of the finite verb: Our observations about separable prefixes show a further OV-residue in the syntax of Cimbrian: the option of placing these prefixes in a position preceding (40) Dar Mann bo da o hat geheft a naüga Arbat. the man that has-part-pref started a new work 'The man who started a new job' the participle indicates that Cimbrian of Luserna has not yet completed its change to a VO-language. This hypothesis receives independent support from the fact that Roana, a variety of Cimbrian which has almost finished its development into a VO-language, only allows separable prefixes to occur in the postparticipial position typical of a VO-language. (41) [Cimbrian of Roana] a. Haüte dar Gianni is gont-vort.
today the John is gone-away b. *Haüte dar Gianni is vort-gon.
today the John is away-gone 'Today Gianni went out' It emerges from these observations that the change from an OV-to a VOlanguage takes the position of finite auxiliaries as its starting point (for an explanation of this phenomenon, see Fuß &Trips 2002), while the position of clitic pronouns, quantified objects, and separable prefixes appear to be the most resistent OV-phenomena, which are the latest to surrender to the inevitable typological change. [

4.4] Verb movement and IP adverbs
The order of prefixes and the position of bare quantifiers shows that the Cimbrian inflected verb does not raise in the TP area. In order to be more precise, let us use the standard test to determine the exact position of the inflected verb, i.e. the position of adverbs. Since it would be extremely time consuming to test all the sequence of adverbs that Cinque (1999) argues for, we will concentrate here on those that are relevant to show the height of verb movement in Cimbrian. We start from Cinque's (1999)  'always' and za 'already', and the modal adverb furse 11 'perhaps', which is located higher than the highest AspP. Starting from the lowest adverb we have tested, i.e. the adverb garecht, corresponding to the Specifier of Voice in (42), we notice that it always occurs on the right of the inflected verb, like in Italian: (43) I bill asto åhefst garecht di arbat. I want that you start well the work 'I want you to start your job well' As for the herta, main verbs can either occur to the left or to the right of the adverb. On the assumption that adverbs do not move, originally based on Pollock (1989) and Cinque (1999), we can conclude that the inflected verbs can but need not move higher than the aspectual adverb: (44) a. Dar libar bo da ze lesan herta. The book who there they read always 'The book that they read always' b. Dar libar boma herta lesst worma geat in pett.
The book that.we always read wenn we go in bed 'The book that we always read before we go to bed' (45) a. … az ar za årüaft. … that he already phoned b. … *az ar årüaft za.
... that he phoned already '… that he already phoned' The aspectual adverb za 'already' always occurs to the left of the inflected verb: The same is observed with the modal adverb furse 'perhaps' which always occurs on the left of the inflected verb: shown in (42), and considering that za always occurs to the left of the verb, it is expected that furse does as well. This means that the inflected verb raises much less than the Romance verb: essentially the position of the inflected verb in Cimbrian corresponds to the raising span of the past participle in the Northern Italian varieties (see Cinque 1999 chapter 2), which occurs after za but can occur before or after the adverb corresponding to 'always'. Mapping the adverbs we have considered onto Cinque's hierarchy, we can conclude that tenses have to raise higher than VoiceP, can raise higher than PerfectaspectP containing adverbs like 'always' but cannot raise higher than TanteriorP whose specifier contains the adverb 'already'. We will come back to this conclusion after having considered the movement span of complex tenses.
The VO system of Cimbrian is not the one of the Romance languages, where the inflected verb must raise higher than the adverb already and can raise higher than the adverb perhaps, as shown in (47): (47) a. … che Gianni telefona giá alle cinque.
... that Gianni calls maybe at five '… that maybe Gianni will call at five' [5] concluding remarks In this work we have examined the peculiar system of Cimbrian, the only German dialect which has become a VO language. In Cimbrian, we see that the last residues of the OV grammar are concentrated in the position of bare quantifiers, which remain in OV order also in an otherwise VO grammar. However, Cimbrian has maintained a typical Germanic property, which is most probably also related to the V2 effect, i.e. they still have enclitic and not proclitic pronouns, like Romance. We conclude that some properties are more resilient than others and this raises the question of the reason why this is so, a question we will answer in future work.

dedication
We dedicate this work to Janne, whose vital energy, passion for our work and engagement in creating new methods has always been an inspiration. Her courage in finding new ways to systematize data and means to investigate the language faculty will remain as a model to us all. We will sorely miss her.