Modeling Syntactic Change under Contact: The Case of Italiot Greek

: In this paper, we investigate patterns of persistence and change affecting the syntax of nominal structures in Italiot Greek in comparison to Modern (and Ancient) Greek, and we explore the role of Southern Italo-Romance as a potential source of interference. Our aim is to highlight the dynamics that favor syntactic contact in this domain: we provide an overview of the social context where these dynamics have taken place and of the linguistic structures involved.


Introduction
The relation between Greek and Romance in Southern Italy represents an ideal setting to explore syntactic microvariation and to investigate the impact of horizontal transmission 1 on syntactic change: in fact, these communities are a "natural laboratory" (Katsoyannou 1999) to observe the mechanisms of language change under contact.
Our research in this area focuses on nominal structures, with two main purposes: 1. a. Explore, describe and represent syntactic microvariation in this domain. b.
Define the role of horizontal transmission in triggering language change.
To examine these points, in this paper we summarize the findings of our previous work 2 , we combine them with novel evidence, and we tentatively identify the structural factors which favor or hamper horizontal change.
With respect to 1a, we raise the following questions: 2. a. Does microvariation manifest itself in nominal structures? b.
If yes, is it possible to single out specific (sub)domains exhibiting a higher degree of variation (or, vice versa, stronger resistance)?
In previous works, we suggested that: (a) the patterns of divergence between Italiot Greek and other (ancient and contemporary) varieties of Greek are mostly due to innovations introduced in Italiot Greek under the pressure of contact with Romance; (b) in certain domains, these innovations have been more pervasive than in others, and (c) Calabria Greek has been more impermeable to changes than Salento Greek. These aspects are discussed in Section 2, where we present the sociolinguistic settings of the two communities, in Section 3, where we summarize the relevant data, and in Section 4, where we propose an explanation for such unbalanced effects.
As far as 1b is concerned, we focus on the following questions:

a.
What is the impact of horizontal transmission on syntactic diversity? b.
Is syntactic borrowing sensitive to structural similarity? 3 In the domains we focused on so far, a condition that seems to trigger the processes of reanalysis inducing structural change is the availability of overlapping linear strings 4 (even if they emerged from different structural sources) between the source (Romance) and the target (Greek) language. The phenomena we investigate in this paper suggest that this condition is not sufficient: more precisely, it is their combination with specific aspects of the internal configuration of each given domain that triggers (or blocks) structural change.

Background: Syntactic Contact and Structural Borrowing
Our work is launched in the framework of the generative approaches to dialectal variation 5 . Dialectal varieties are often "in a constant interaction with one or more standard languages and with other dialects" (Barbiers and Cornips 2000, p. 3); hence, they offer a helpful testing ground for exploring the role of contact in language change 6 . Furthermore, the study of microvariation is a powerful tool for understanding the mechanisms that underlie diversity of the language faculty (Kayne 1996(Kayne , 2005Savoia 2005, 2019, a.o.). Under this view, a major goal is to devise a model able to detect, describe and explain microscopic diversity accounting for both the "external" impulses acting on it (i.e., the dynamics of interaction between speakers and/or speaking communities and their consequences on the E-languages 7 available to the speakers) and the "internal" structures that determine it (i.e., the speakers' grammatical competence).
The role played by contact in the dynamics of language change was first systematically acknowledged by Weinreich (1953) and subsequently modeled by Thomason and Kaufman (1988). Since then, contact has been regarded as a "bridge" between variation and change: "the implications of contact are pervasive and fundamental to language change, whether the contact be between speakers of different languages or between those of different varieties of the same language" (Bowern 2009, p. 187) 8 .
An important issue associated with contact is borrowing, defined as "the incorporation of foreign elements into the speakers' native language" (Thomason and Kaufman 1988, p. 21). Specifically, syntactic borrowing has been defined as a transfer of syntactic rules (Thomason 2004) not induced by lexical or other kinds of transfer, and has been shown to be less conscious, massive and dependent on the physical context as compared to other contact-induced changes (Thomason 2001).
As stressed by Thomason and Kaufman (1988), two classes of potential stimuli operate on the mechanisms of structural borrowing: social factors 9 and the internal structure of 1 In the phylogenetic literature, the term horizontal transmission is used to refer to changes/innovations transmitted from one language to another when these languages are in geographical contact (though not necessarily genetically related). 2 Guardiano (2014a); Guardiano and Stavrou (2014, 2019a, 2019b; Guardiano et al. (2016. 3 The conjectures about structural contact that we discuss here have been inspired by Weinreich (1953) and by some aspects of the models proposed in Thomason and Kaufman (1988) and Heine and Kuteva (2005; with particular reference to the notion of "equivalence": chp. 1, 6). 4 By "overlapping linear strings" we mean sequences that are (superficially) identical in the source and in the target language. 5 Black and Motapayane (1996); Cornips (1998);Auer et al. (2005); Barbiers and Cornips (2000); Adger and Trousdale (2007), a.o. 6 Thomason and Kaufman (1988); Bowern (2009); Thomason (2001); Heine and Kuteva (2005); Hickey (2010), a.o. The understanding of the impact of contact in determining structural changes is fruitful for both historical reconstruction (Noonan 2010) and the analysis of language transmission (Corrigan 2010). As far as the reconstruction of historical relatedness is concerned, "linguists [ . . . ] need to engage with the central question of whether linguistic features which owe their existence to descent from an ancestral variety or protolanguage within a family can be distinguished from those which have been borrowed or remodeled on the basis of another language" (McMahn 2010, p. 128; see also, among many others, Corrigan 2010). 7 Chomsky (1986). 8 Thomason (2001); Heine and Kuteva (2005); see also the papers collected in Hickey (2010), among many others. 9 "Long-term contact with widespread bilingualism among borrowing-language speakers is a prerequisite for extensive structural borrowing" (Thomason and Kaufman 1988, p. 67). See also Smith and Veenstra (2001); Matras and Bakker (2003); Pountain (2006); Aikhenvald and Dixon (2007) the source and the target language 10 . As a matter of fact, in their model, social factors are assigned a primary role 11 : "it is the social context [ . . . ] that determines the direction and the degree of interference" (Thomason and Kaufman 1988, p. 19) 12 . They identify three major social factors as triggers for structural borrowing: 13 4. a. "long-term cultural pressure from source-language speakers on the borrowinglanguage speaker group"; b.
"a history of several hundred years of intimate contact"; c.
Concerning structural similarity, it must be remarked that we are dealing with two language groups (Greek and Romance) genealogically very close and typologically quite similar to one another 15 ; hence, in order to explore instances of convergence/divergence, we must first identify a set of domains exhibiting variation between the two groups. This is done in Section 3.
To investigate the role of structural similarity in triggering syntactic change under contact, we adopt as a starting point a hypothesis first proposed by Guardiano et al. (2016) and further developed in , labeled "Resistance Principle". The Resistance Principle submits that syntactic borrowing "is likely to be the result of some intrinsic resistance by language's most internally structured systems (such as syntax) to accept changes even when they are motivated by external pressures (like contact)" (Guardiano et al. 2016, pp. 147-48) 16 . Therefore, in order for syntactic change to occur, it must be triggered by interference data already available ("familiar", Sitaridou 2014) in the interfered language.

5.
Resistance principle: "Resetting of parameter α from value X to Y in language A as triggered by interference of language B only takes place if a subset of the strings that contribute to constituting a trigger or value Y of parameter α in language B already exists in language A". (Guardiano et al. 2016, p. 148) 17 .

Sociolinguistic Factors: Greek and Romance in Southern Italy
The relation between Greek and Romance in Southern Italy displays all the conditions that are assumed to favor structural borrowing, in terms of both social factors and structural similarity. As far as social factors are concerned, the sociolinguistic settings of the Greekspeaking areas are compatible with the components listed in 4:

6.
a. Romance is the dominant group (intense pressure); b.
The two groups have been in contact for centuries (intimate contact); c.
Currently, no Greek speaker is monolingual (all the speakers of Greek also speak at least a regional variety of Italian, and often a Romance dialect) (bilingualism).
As is well known 18 , there are two Greek-speaking communities in Southern Italy, one in Salento (Grecia salentina) and one in Calabria (Bovesia). In terms of social prestige, Romance has been the dominant language in both areas: especially in recent history, the Greek communities have had very low social prestige in Southern Italy. Particularly in Calabria, they are traditionally associated to poor and rural populations. Additionally, until very recently, speakers have not shown any positive attitude towards their own language and have made no effort towards preserving their cultural identity and language. Thus, despite attempts of "revitalizing" the language in both areas, Greek is currently in a state of regression/obsolescence: native speakers have almost disappeared (these varieties are no longer acquired as first languages) and Romance varieties are massively adopted in everyday use.
Concerning bilingualism, there is an intricate debate about whether and how the languages that have been spoken in the area since ancient times have interacted to one another and how this interaction has affected the current structure of Greek and Romance dialects (see Fanciullo 2001 for a summary). As far as more recent times are concerned, there are differences between Salento and Calabria 19 . In Salento, bilingualism has been the rule for much longer than in Calabria, where Greek-speaking villages were generally isolated and had little contact with the Romance-speaking communities; as a consequence, until recent times, Greek speakers in Calabria were mostly monolingual. In previous work, we suggested that the different pace at which Salento and Calabria have integrated innovations induced by contact with Romance is connected to this condition. We will return to this issue in Section 4.
To sum up, the sociolinguistic conditions of Greek and Romance in Southern Italy are compatible with Thomason and Kaufman (1988, pp. 74-75) definition of "(very) strong cultural pressure", typically associated to "moderate-to-heavy" structural borrowing. Romance acts as the source language, with Greek acting as the target one. According to Thomason and Kaufman (1988, p. 63) "the identification of a source language requires the establishment of present or past contact of sufficient intensity between the proposed source language and the recipient language"; in turn, the existence of some degree of "intensity of contact" between a source language and a recipient language is deducible from the presence, in the recipient language, of "innovations that may reasonably be attributed to that source language". As we see in Section 3, this is precisely the situation we observe in Southern Italy. Our previous work on Italiot Greek has revealed that the differences detected between Italiot Greek and standard Greek 20 can be attributed to innovations induced by contact with the Romance dialects of the area 21 . In addition, as already mentioned, the differences 18 See Guardiano and Stavrou (2014) for a partial overview of the literature concerning the history and structure of the Greek speaking communities in Southern Italy, and Silvestri and Schifano (2017); Squillaci (2017); Remberger (2018) for recent surveys. 19 Chilà (2021); Chilà and De Angelis (2021) and the literature therein. 20 The term "standard Greek" (often comprising also the adjective Modern-"Standard Modern Greek") is uniformly employed to denote the standardized contemporary Greek language (both spoken and written/literary) that is used in the big urban centers and is distinguished from the local (geographical) varieties, both inside and outside Greece, traits of which it has absorbed. In this sense standard Greek is an idealization/abstraction reflecting forms of the Greek language all of which derive from the Koine. In this paper we chose to take standard Greek as a reference language for our study because this is somehow the default Greek language of our times. Moreoever, standard Greek in its idealization (and/or abstraction) remains (more) immune to contact with other languages Our choice does not imply any evaluation or bias in favor or against standard Greek (or any of its dialects between Salento and Calabria in terms of intensity of contact with Romance explain why such innovations were introduced in the two communities at a different pace. However, horizontal transmission has not obscured the genealogical connection between Italiot Greek and the rest of Greek: in all the taxonomic experiments performed so far using as an input syntactic (parametric) data from the nominal domain (Guardiano et al. 2016;Ceolin et al. 2020), the two Italiot Greek varieties are systematically identified as members of the Greek group.

Collection of Data
The Italiot Greek data discussed in this paper originate from native speakers' judgments and from a selection of written records. By "written records" we refer to collections of texts originally transmitted orally, which were gathered and put down in written form by local experts around the middle of the 20th century. Since these texts offer, apparently, a (partial) window to the language as it was when it was more productive and used for everyday needs, we included them in our research data, in combination with the speakers' judgments (and the grammars), in order to obtain a more complete view of the language. The comparison between these two types of sources reveals important differences that allowed us to identify different steps of changes in progress. 22 As far as Romance is concerned, we focus on the dialects spoken in the same areas as Italiot Greek, Salentino and Southern Calabrese in particular, both belonging to the "extreme Southern" group (Pellegrini 1977). The data emerged from on-purpose interviews with native speakers, and were subsequently combined, when needed, with the evidence provided in the literature. The extreme Southern Romance dialects of Italy do not significantly differ from one another in the domains under investigation in this paper.
All the examples, where not otherwise specified, have been tested in argument (i.e., subject/object) position.

Results
We start with the domains that are known to display differences between Greek and Romance outside of the geographic area under investigation; we observe whether, in these domains, Italiot Greek converges or diverges from Greek and, in the latter case, whether it converges or not with Romance. They are listed in 7.
Concerning 7a and b (adjectives and demonstratives), we explore two major aspects: Linearization with respect to other constituents; b.
Co-occurrence with the definite article.
As far as 8a (linearization) is concerned, in standard Greek (and also Ancient Greek), demonstratives have by and large the same distribution as (postnominal) adjectives 27 . In contrast, in Italiot Greek, the distribution of demonstratives is incompatible with that of 22 We refer to Guardiano and Stavrou (2019a, 2019b for an extensive list of these sources. 23 Stavrou (2012Stavrou ( , 2013, Guardiano and Stavrou (2014, 2019a, 2019b. 24 Horrocks and Stavrou (1987); Stavrou and Horrocks (1989); Guardiano (2012Guardiano ( , 2014b; Guardiano and Michelioudakis (2019); . 25 Guardiano (2011); Guardiano and Longobardi (2018); ; . 26 Guardiano et al. (2016; Mertyris (2014). 27 See also Horrocks and Stavrou (1987); Stavrou and Horrocks (1989); Alexiadou et al. (2007); Manolessou and Panagiotidis (1999) adjectives, which display the same linearization patterns as the Romance dialects of the area. A first difference between Italiot Greek and standard Greek is that, in Italiot Greek, the unmarked (even the only possible) position for most adjectives (as in the Romance dialects of the area) is the postnominal one (as shown in 9a). Like the neighboring Romance dialects 28 , only a very restricted group of (speaker-oriented) adjectives is found after determiners/numerals 29 and before the noun (see 9b); by contrast, in standard Greek almost all types of adjectives can be prenominal. Additionally, in Italiot Greek, adjectives are never found to the left of numerals (see the ungrammaticality of 9c), again unlike standard Greek 30 . As far as demonstratives are concerned, they are linked to a DP-initial position (see 10a; like in the Romance dialects of the area): 31 all other linear orders are ungrammatical (see 10b,c).    29 Reference to numerals in the present context is due to the fact that they are the leftmost modifiers in the DP. This holds both when there is no article present and when D is occupied by the definite article. In the former case it may be assumed that numerals act as determiners ; in the latter they are on a par with (weak) quantifiers. Therefore, numerals provide a solid reference point for determining the position of adjectives (and other modifiers). 30 In standard Greek, only demonstratives and the quantificational adverb olos ("all") may appear to the left of numerals, more correctly before the definite article, which is itself the leftmost element of the DP. Notice however that, in standard Greek, both demonstratives and (articulated) adjectives can occur to the left of determiners: afto to vivlio (lit. "this the book"), to oreo to vivlio (lit. "the nice the book"). 31 Salentino: šti/ddi tri krištjani (lit. "these/those three men"); * tri šti/ddi krištjani; * i (tri) krištjani šti/ddi.     To round up the discussion so far, concerning linearization of adjectives and demonstratives, we make the following observations: 11. a.
Italiot Greek systematically goes with the Romance dialects of Southern Italy and against standard Greek. b.
Concerning adjectives, there are overlapping strings between Greek and Romance: [Num N A] and [Num A N] (Tables 1 and 3 respectively). c.
Concerning demonstratives, DP-initial ones are found in all the three groups.
Turning to 8b (co-occurrence with the definite article), in standard Greek, demonstratives systematically co-occur with the definite article, no matter of their position, as shown in 12. Similarly, adjectives originally merged in postnominal predicative structures (Stavrou 2012(Stavrou , 2013(Stavrou , 2019 systematically take a copy of the definite article in definite DPs, a phenomenon known as "polydefiniteness", shown in 13.
Languages 2021, 6, x FOR PEER REVIEW 8 of 23 Table 3. Num X N. To round up the discussion so far, concerning linearization of adjectives and demonstratives, we make the following observations: 11. a. Italiot Greek systematically goes with the Romance dialects of Southern Italy and against standard Greek. b. Concerning adjectives, there are overlapping strings between Greek and Romance:

Italiot Greek
[Num N A] and [Num A N] (Tables 1 and 3 respectively). c. Concerning demonstratives, DP-initial ones are found in all the three groups.
Turning to 8b (co-occurrence with the definite article), in standard Greek, demonstratives systematically co-occur with the definite article, no matter of their position, as shown in 12. Similarly, adjectives originally merged in postnominal predicative structures (Stavrou 2012(Stavrou , 2013(Stavrou , 2019 systematically take a copy of the definite article in definite DPs, a phenomenon known as "polydefiniteness", shown in 13.

a. to vivlio afto
Standard Greek the book this 'this book' b. afto to vivlio c. * afto vivlio d. * vivlio afto 13. a. to vivlio to kokino the book the red 'the red book' b. to kokino to vivlio c. * kokino to vivlio d. * to vivlio kokino In the currently spoken varieties of Italiot Greek, polydefinite DPs are unattested (see 14 and 15, 33 from Guardiano and Stavrou 2020, p. 126). The written sources provide evidence that the phenomenon was active in the past and disappeared only recently. In particular, Calabria Greek apparently retained it until very recent times: in most texts (e.g., those collected in Caracausi and Rossi Taibbi 1959), postnominal adjectives are regularly articulated in definite DPs and demonstratives often co-occur with articles. By contrast, we found only residual instances of polydefinite DPs in the written sources of Salento Languages 2021, 6, x FOR PEER REVIEW 8 of 23 Table 3. Num X N. To round up the discussion so far, concerning linearization of adjectives and demonstratives, we make the following observations: 11. a. Italiot Greek systematically goes with the Romance dialects of Southern Italy and against standard Greek. b. Concerning adjectives, there are overlapping strings between Greek and Romance:

Italiot Greek
[Num N A] and [Num A N] (Tables 1 and 3 respectively). c. Concerning demonstratives, DP-initial ones are found in all the three groups.
Turning to 8b (co-occurrence with the definite article), in standard Greek, demonstratives systematically co-occur with the definite article, no matter of their position, as shown in 12. Similarly, adjectives originally merged in postnominal predicative structures (Stavrou 2012(Stavrou , 2013(Stavrou , 2019 systematically take a copy of the definite article in definite DPs, a phenomenon known as "polydefiniteness", shown in 13.

a. to vivlio afto
Standard Greek the book this 'this book' b. afto to vivlio c. * afto vivlio d. * vivlio afto 13. a. to vivlio to kokino the book the red 'the red book' b. to kokino to vivlio c. * kokino to vivlio d. * to vivlio kokino In the currently spoken varieties of Italiot Greek, polydefinite DPs are unattested (see 14 and 15, 33 from Guardiano and Stavrou 2020, p. 126). The written sources provide evidence that the phenomenon was active in the past and disappeared only recently. In particular, Calabria Greek apparently retained it until very recent times: in most texts (e.g., those collected in Caracausi and Rossi Taibbi 1959), postnominal adjectives are regularly articulated in definite DPs and demonstratives often co-occur with articles. By contrast, we found only residual instances of polydefinite DPs in the written sources of Salento In the currently spoken varieties of Italiot Greek, polydefinite DPs are unattested (see 14 and 15, 33 from Guardiano and Stavrou 2020, p. 126). The written sources provide evidence that the phenomenon was active in the past and disappeared only recently. In particular, Calabria Greek apparently retained it until very recent times: in most texts (e.g., those collected in Caracausi and Rossi-Taibi 1959), postnominal adjectives are regularly articulated in definite DPs and demonstratives often co-occur with articles. By contrast, we found only residual instances of polydefinite DPs in the written sources of Salento Greek we checked. This suggests that Calabria Greek, until very recently, had remained much more impervious to changes than Salento Greek. We will come back to this issue in Section 4. 'this daughter' c. ecini ti γineka tutese e δio monakese that the woman these the two nuns 'that woman' 'these two nuns' There are no instances of polydefinite DPs, or of co-occurrence of demonstratives and articles, in the Romance dialects of Southern Italy: hence, concerning 8b, the currently spoken varieties of Italiot Greek go, again, with Romance rather than with standard Greek, as shown in Table 4. The third domain that we observe is the realization of nominal arguments, primarily genitive DPs (7c). In Greek, since the earliest stages of the language, genitive DPs are morphologically marked, because the noun denoting the possessor, or an argument (agent or theme) of the head noun, bears genitive case.
The loss of morphological case distinctions is described as a major diachronic change in the history of Greek: it has variously affected all the areas where the language has been spoken and almost all its dialects across space and time. We refer to Mertyris (2014) for an extensive discussion of the relevant facts and literature.
Several authors 34 discuss cases of paradigm gaps due to the weakening/loss of inflectional (case) morphology in Italiot Greek. Yet, despite diachronic weakening/recession, "the genitive is largely maintained in the varieties of Southern Italy, according to the most prominent grammatical description of the dialect which was conducted by the German linguist Gerhard Rohlfs (1977, pp. 182-83), who disputes earlier claims that it had been lost (cf. Morosi 1870). […] The higher distributional potential of the genitive in Italiot helps the preservation of its morphological productivity […]. Most grammatical descriptions (cf. Rohlfs 1977;Karanastasis 1997) provide full paradigms without genitive gaps" (Mertyris 2014, pp. 269-70). In short, in spite of evidence of inflectional reduction, the genitive system of Italiot Greek displays "high functionality" and "relative paradigmatic productivity" (Mertyris 2014, p. 271 There are no instances of polydefinite DPs, or of co-occurrence of demonstratives and articles, in the Romance dialects of Southern Italy: hence, concerning 8b, the currently spoken varieties of Italiot Greek go, again, with Romance rather than with standard Greek, as shown in Table 4. The third domain that we observe is the realization of nominal arguments, primarily genitive DPs (7c). In Greek, since the earliest stages of the language, genitive DPs are morphologically marked, because the noun denoting the possessor, or an argument (agent or theme) of the head noun, bears genitive case.
The loss of morphological case distinctions is described as a major diachronic change in the history of Greek: it has variously affected all the areas where the language has been spoken and almost all its dialects across space and time. We refer to Mertyris (2014) for an extensive discussion of the relevant facts and literature.
Several authors 34 discuss cases of paradigm gaps due to the weakening/loss of inflectional (case) morphology in Italiot Greek. Yet, despite diachronic weakening/recession, "the genitive is largely maintained in the varieties of Southern Italy, according to the most prominent grammatical description of the dialect which was conducted by the German linguist Gerhard Rohlfs (1977, pp. 182-83), who disputes earlier claims that it had been lost (cf. Morosi 1870). […] The higher distributional potential of the genitive in Italiot helps the preservation of its morphological productivity […]. Most grammatical descriptions (cf. Rohlfs 1977;Karanastasis 1997) provide full paradigms without genitive gaps" (Mertyris 2014, pp. 269-70). In short, in spite of evidence of inflectional reduction, the genitive system of Italiot Greek displays "high functionality" and "relative paradigmatic productivity" (Mertyris 2014, p. 271). Morphologically, genitive case distinctions are visible on the definite article, while they appear weakened on other categories (e.g., noun and adjectives), depending on various constraints (cf. the There are no instances of polydefinite DPs, or of co-occurrence of demonstratives and articles, in the Romance dialects of Southern Italy: hence, concerning 8b, the currently spoken varieties of Italiot Greek go, again, with Romance rather than with standard Greek, as shown in Table 4. The third domain that we observe is the realization of nominal arguments, primarily genitive DPs (7c). In Greek, since the earliest stages of the language, genitive DPs are morphologically marked, because the noun denoting the possessor, or an argument (agent or theme) of the head noun, bears genitive case.
The loss of morphological case distinctions is described as a major diachronic change in the history of Greek: it has variously affected all the areas where the language has been spoken and almost all its dialects across space and time. We refer to Mertyris (2014) for an extensive discussion of the relevant facts and literature.
Several authors 34 discuss cases of paradigm gaps due to the weakening/loss of inflectional (case) morphology in Italiot Greek. Yet, despite diachronic weakening/recession, "the genitive is largely maintained in the varieties of Southern Italy, according to the most prominent grammatical description of the dialect which was conducted by the German 33 As far as demonstratives are concerned, sometimes speakers use contracted forms obtained from morphophonological fusion with the definite article. Such forms seem to freely alternate with non-contracted (non-articulated) ones. 34 Including, e.g., Katsoyannou (1995); Nucera (1993); Minuto et al. (1988); and also Rohlfs (1977), a.o.
As far as the syntactic realization of genitive DPs is concerned (Longobardi 2001), in Italiot Greek, nominal arguments are realized as non-prepositional DPs, morphologically marked with genitive case (where available), as in standard Greek. They further display two properties (see the examples in 17): first, they are postnominal and, second, they cannot be iterated (a head noun bears only one genitive argument). This strategy for realizing adnominal genitives has been typical of Greek since ancient times (Guardiano 2011;Guardiano and Longobardi 2018): some examples from standard Greek are given in 18. Structurally, this strategy shares the major properties of a specific structural configuration called "GenO" by Longobardi and Silvestri (2013;"GenL" in Crisma et al. 2020) listed in 16. Note that, in Greek, nominal genitive arguments always appear postnominally.
Non iterable (i.e., a DP cannot contain multiple nominal genitives modifying one and the same head noun); c.
Languages 2021, 6, x FOR PEER REVIEW 10 of 23 As far as the syntactic realization of genitive DPs is concerned (Longobardi 2001), in Italiot Greek, nominal arguments are realized as non-prepositional DPs, morphologically marked with genitive case (where available), as in standard Greek. They further display two properties (see the examples in 17): first, they are postnominal and, second, they cannot be iterated (a head noun bears only one genitive argument). This strategy for realizing nominal genitives has been typical of Greek since ancient times (Guardiano 2011, Guardiano andLongobardi 2018): some examples from standard Greek are given in 18. Structurally, this strategy shares the major properties of a specific structural configuration called "GenO" by Longobardi and Silvestri (2013;"GenL" in Crisma et al. 2020 ii. to vivlio to kokino tu agoriu iii. to vivlio tu agoriu to kokino iv. * to (kokino) tu agoriu (kokino) vivlio In contrast, the Romance dialects of Southern Italy (with limited exceptions, Silvestri 2013, Massaro 2019) use a different type of genitive realization, namely prepositional genitives. Crosslinguistically, prepositional genitives are postnominal and can be freely ordered with respect to other postnominal modifiers. Additionally, DPs containing multiple prepositional genitives modifying one and the same head noun are possible. These properties are typical of a further strategy crosslinguistically adopted to realize nominal arguments, labeled "Free genitive" by Longobardi and Silvestri (2013). An exemplification of prepositional genitives in the Romance dialects of Southern Italy is given in 19 and 20.
Languages 2021, 6, x FOR PEER REVIEW 10 of 23 As far as the syntactic realization of genitive DPs is concerned (Longobardi 2001), in Italiot Greek, nominal arguments are realized as non-prepositional DPs, morphologically marked with genitive case (where available), as in standard Greek. They further display two properties (see the examples in 17): first, they are postnominal and, second, they cannot be iterated (a head noun bears only one genitive argument). This strategy for realizing nominal genitives has been typical of Greek since ancient times (Guardiano 2011, Guardiano andLongobardi 2018): some examples from standard Greek are given in 18. Structurally, this strategy shares the major properties of a specific structural configuration called "GenO" by Longobardi and Silvestri (2013;"GenL" in Crisma et al. 2020 ii. to vivlio to kokino tu agoriu iii. to vivlio tu agoriu to kokino iv. * to (kokino) tu agoriu (kokino) vivlio In contrast, the Romance dialects of Southern Italy (with limited exceptions, Silvestri 2013, Massaro 2019) use a different type of genitive realization, namely prepositional genitives. Crosslinguistically, prepositional genitives are postnominal and can be freely ordered with respect to other postnominal modifiers. Additionally, DPs containing multiple prepositional genitives modifying one and the same head noun are possible. These properties are typical of a further strategy crosslinguistically adopted to realize nominal arguments, labeled "Free genitive" by Longobardi and Silvestri (2013). An exemplification of prepositional genitives in the Romance dialects of Southern Italy is given in 19 and 20.
In contrast, the Romance dialects of Southern Italy (with limited exceptions, Silvestri 2003; Massaro 2019) use a different type of genitive realization, namely prepositional genitives. Crosslinguistically, prepositional genitives are postnominal and can be freely ordered with respect to other postnominal modifiers. Additionally, DPs containing multiple prepositional genitives modifying one and the same head noun are possible. These properties are typical of a further strategy crosslinguistically adopted to realize nominal arguments, labeled "Free genitive" by Longobardi and Silvestri (2013). An exemplification of prepositional genitives in the Romance dialects of Southern Italy is given in 19 and 20.

a. lu/nnu ritrattu te lu Ggiuanni
Salentino the/a picture of the Ggiuanni 'John's picture/a picture of John' b. lu ritrattu ngrazziatu te lu Ggiuanni the picture nice of the Ggiuanni 'John's nice picture' c. lu ritrattu te lu Ggiuanni te la Maria 35 the picture of the Ggiuanni of the Maria 'John's picture of Mary' 20. a. la kasa di lu sinniku Southern Calabria the house of the major 'the major's house' b. i. la bbella kasa di lu sinniku the nice house of the major 'the nice house of the major' ii. la kasa bbella di lu sìnniku iii. la kasa di lu sinniku bella A phenomenon that has drawn the linguists' attention concerning Italiot Greek genitives 36 is the (adnominal) possessive use of a prepositional construction headed by "the ablative preposition atse/aʃe 'from' (<AG ἐξ "out of")" (Mertyris 2014, p. 275). 37 This construction is sometimes used to realize nominal genitives; yet, when used in this function, it is subject to constraints. As shown by the contrast between 21a and b, it is usually accepted when the head noun is indefinite, while being disliked by most speakers in DPs headed by a definite article. Also, it is dispreferred when the genitive is a definite DP (21c) or a proper name (21d). 38 Finally, in terms of usage, the prepositional construction with atse is quite marginal according to the speakers (especially in Salento) and is not frequent in the written sources. 21. a. mia (megali) ikoni/fotografia (megali) (a)tse ena athropo/gineka (megali) a big portrait/picture big ATSE a man/woman big 'a big picture of a man/woman' b. ? i (megali) ikoni/fotografia (megali) (a)tse ena athropo/gineka (megali) the big portrait/picture big ATSE a man/woman big 'the big portrait/picture of a man/woman' c. * mia ikoni/fotografia atse to(n) athropo d. * mia ikoni/fotografia atse to Ianni With the exception of Karanastasis (1997, p. 53), 39 the literature uniformly agrees that these structures are a consequence of contact with Romance. There is also agreement that 35 Although speakers tend to avoid two genitives, the possibility of more than one genitive argument of the noun is not excluded. 36 E.g., Karanastasis (1997); Rohlfs (1977); Katsoyannou (1995); Profili (1985), a.o. 37 The prepositional construction with atse is also found in sequences resembling partitive constructions with di in Romance (Alexiadou and Stavrou 2019) and ablative constructions with da, as shown for instance in Profili (1985) and reported in Mertyris (2014, p. 276). 38 The judgments of the speakers are variable; at the present stage we cannot provide any more detailed data. 39 From Mertyris (2014, p. 276): " Karanastasis (1997, p. 53) does not accept the effect of Italian influence, as opposed to Rohlfs (1977, p. 69), and claims that the possessive use of this preposition was an internal development in the dialect". It is also worthwhile pointing out that the replacement of inflected genitive complements with prepositional constructions is a phenomenon attested in several varieties of Greek, including standard Greek, and has been described as a language-internal one, i.e., not (necessarily) induced by contact with non-Greek languages. 20. a. la kasa di lu sinniku Southern Calabria the house of the major 'the major's house' b. i. la bbella kasa di lu sinniku the nice house of the major 'the nice house of the major' ii. la kasa bbella di lu sìnniku iii. la kasa di lu sinniku bella A phenomenon that has drawn the linguists' attention concerning Italiot Greek genitives 36 is the (adnominal) possessive use of a prepositional construction headed by "the ablative preposition atse/aʃe 'from' (<AG ἐξ "out of")" (Mertyris 2014, p. 275). 37 This construction is sometimes used to realize nominal genitives; yet, when used in this function, it is subject to constraints. As shown by the contrast between 21a and b, it is usually accepted when the head noun is indefinite, while being disliked by most speakers in DPs headed by a definite article. Also, it is dispreferred when the genitive is a definite DP (21c) or a proper name (21d). 38 Finally, in terms of usage, the prepositional construction with atse is quite marginal according to the speakers (especially in Salento) and is not frequent in the written sources.

a. mia (megali) ikoni/fotografia (megali) (a)tse ena athropo/gineka (megali)
a big portrait/picture big ATSE a man/woman big 'a big picture of a man/woman' b. ? i (megali) ikoni/fotografia (megali) (a)tse ena athropo/gineka (megali) the big portrait/picture big ATSE a man/woman big 'the big portrait/picture of a man/woman' c. * mia ikoni/fotografia atse to(n) athropo d. * mia ikoni/fotografia atse to Ianni With the exception of Karanastasis (1997, p. 53), 39 the literature uniformly agrees that these structures are a consequence of contact with Romance. There is also agreement that 35 Although speakers tend to avoid two genitives, the possibility of more than one genitive argument of the noun is not excluded. 36 E.g., Karanastasis (1997); Rohlfs (1977); Katsoyannou (1995); Profili (1985), a.o. 37 The prepositional construction with atse is also found in sequences resembling partitive constructions with di in Romance (Alexiadou and Stavrou 2019) and ablative constructions with da, as shown for instance in Profili (1985) and reported in Mertyris (2014, p. 276). 38 The judgments of the speakers are variable; at the present stage we cannot provide any more detailed data. 39 From Mertyris (2014, p. 276): " Karanastasis (1997, p. 53) does not accept the effect of Italian influence, as opposed to Rohlfs (1977, p. 69), and claims that the possessive use of this preposition was an internal development in the dialect". It is also worthwhile pointing out that the replacement of inflected genitive complements with prepositional constructions is a phenomenon attested in several varieties of Greek, including standard Greek, and has been described as a language-internal one, i.e., not (necessarily) induced by contact with non-Greek languages.
A phenomenon that has drawn the linguists' attention concerning Italiot Greek genitives 36 is the (adnominal) possessive use of a prepositional construction headed by "the ablative preposition atse/a e 'from' (<AG ἐξ "out of")" (Mertyris 2014, p. 275). 37 This construction is sometimes used to realize adnominal genitives; yet, when used in this function, it is subject to constraints. As shown by the contrast between 21a and b, it is usually accepted when the head noun is indefinite, while being disliked by most speakers in DPs headed by a definite article. Also, it is dispreferred when the genitive is a definite DP (21c) or a proper name (21d). 38 Finally, in terms of usage, the prepositional construction with atse is quite marginal according to the speakers (especially in Salento) and is not frequent in the written sources. 20. a. la kasa di lu sinniku Southern Calabria the house of the major 'the major's house' b. i. la bbella kasa di lu sinniku the nice house of the major 'the nice house of the major' ii. la kasa bbella di lu sìnniku iii. la kasa di lu sinniku bella A phenomenon that has drawn the linguists' attention concerning Italiot Greek genitives 36 is the (adnominal) possessive use of a prepositional construction headed by "the ablative preposition atse/aʃe 'from' (<AG ἐξ "out of")" (Mertyris 2014, p. 275). 37 This construction is sometimes used to realize nominal genitives; yet, when used in this function, it is subject to constraints. As shown by the contrast between 21a and b, it is usually accepted when the head noun is indefinite, while being disliked by most speakers in DPs headed by a definite article. Also, it is dispreferred when the genitive is a definite DP (21c) or a proper name (21d). 38 Finally, in terms of usage, the prepositional construction with atse is quite marginal according to the speakers (especially in Salento) and is not frequent in the written sources.

a. mia (megali) ikoni/fotografia (megali) (a)tse ena athropo/gineka (megali)
a big portrait/picture big ATSE a man/woman big 'a big picture of a man/woman' b. ? i (megali) ikoni/fotografia (megali) (a)tse ena athropo/gineka (megali) the big portrait/picture big ATSE a man/woman big 'the big portrait/picture of a man/woman' c. * mia ikoni/fotografia atse to(n) athropo d. * mia ikoni/fotografia atse to Ianni With the exception of Karanastasis (1997, p. 53), 39 the literature uniformly agrees that these structures are a consequence of contact with Romance. There is also agreement that 35 Although speakers tend to avoid two genitives, the possibility of more than one genitive argument of the noun is not excluded. 36 E.g., Karanastasis (1997); Rohlfs (1977); Katsoyannou (1995); Profili (1985), a.o. 37 The prepositional construction with atse is also found in sequences resembling partitive constructions with di in Romance (Alexiadou and Stavrou 2019) and ablative constructions with da, as shown for instance in Profili (1985) and reported in Mertyris (2014, p. 276). 38 The judgments of the speakers are variable; at the present stage we cannot provide any more detailed data. 39 From Mertyris (2014, p. 276): " Karanastasis (1997, p. 53) does not accept the effect of Italian influence, as opposed to Rohlfs (1977, p. 69), and claims that the possessive use of this preposition was an internal development in the dialect". It is also worthwhile pointing out that the replacement of inflected genitive complements with prepositional constructions is a phenomenon attested in several varieties of Greek, including standard Greek, and has been described as a language-internal one, i.e., not (necessarily) induced by contact with non-Greek languages.
With the exception of Karanastasis (1997, p. 53), 39 the literature uniformly agrees that these structures are a consequence of contact with Romance. There is also agreement that they are recent: "their starting point should be placed at a time when pressure from 35 Although speakers tend to avoid two genitives, the possibility of more than one genitive argument of the noun is not excluded. 36 E.g., Karanastasis (1997); Rohlfs (1977); Katsoyannou (1995); Profili (1985), a.o. 37 The prepositional construction with atse is also found in sequences resembling partitive constructions with di in Romance (Alexiadou and Stavrou 2019) and ablative constructions with da, as shown for instance in Profili (1985) and reported in Mertyris (2014, p. 276). 38 The judgments of the speakers are variable; at the present stage we cannot provide any more detailed data. 39 From Mertyris (2014, p. 276): " Karanastasis (1997, p. 53) does not accept the effect of Italian influence, as opposed to Rohlfs (1977, p. 69), and claims that the possessive use of this preposition was an internal development in the dialect". It is also worthwhile pointing out that the replacement of inflected genitive complements with prepositional constructions is a phenomenon attested in several varieties of Greek, including standard Greek, and has been described as a language-internal one, i.e., not (necessarily) induced by contact with non-Greek languages.
Romance heavily increased. As the Greek-speaking areas decreased in number of speakers and size during the last three centuries, it would be logical to claim that these analytic phenomena are related to such sociolinguistic factors and that these phenomena are not older than that" (Mertyris 2014, p. 278;cf. also Alexiadou 2017;De Angelis 2021). What is relevant for the purposes of the present discussion is that, in Italiot Greek, prepositional constructions are very marginal as a strategy of adnominal genitive realization: according to the speakers, and also as attested in the written sources, the actually productive strategy is GenO, like in standard Greek (cf. also the discussion in Mertyris 2014, pp. 275-77). Table 5 sums up the main points made so far: concerning the realization of adnominal genitives, there is convergence between Italiot Greek and standard Greek, both being different from Romance. As far as pronominal arguments of the noun (henceforth "possessives", 7d) are concerned, Italiot Greek features the same strategies as standard Greek 40 . Possessives are realized as phonologically enclitic items. Such items do not exhibit feature agreement with the head noun (unlike Romance) and co-occur with articles (see example 22 from Salento Greek). Etymologically, they emerged from the genitive of the weak form of personal pronouns 41 .
Languages 2021, 6, x FOR PEER REVIEW 12 of 23 they are recent: "their starting point should be placed at a time when pressure from Romance heavily increased. As the Greek-speaking areas decreased in number of speakers and size during the last three centuries, it would be logical to claim that these analytic phenomena are related to such sociolinguistic factors and that these phenomena are not older than that" (Mertyris 2014, p. 278;cf. also Alexiadou 2017;De Angelis 2021). What is relevant for the purposes of the present discussion is that, in Italiot Greek, prepositional constructions are very marginal as a strategy of nominal genitive realization: according to the speakers, and also as attested in the written sources, the actually productive strategy is GenO, like in standard Greek (cf. also the discussion in Mertyris 2014, pp. 275-77). Table 5 sums up the main points made so far: concerning the realization of nominal genitives, there is convergence between Italiot Greek and standard Greek, both being different from Romance. As far as pronominal arguments of the noun (henceforth "possessives", 7d) are concerned, Italiot Greek features the same strategies as standard Greek 40 . Possessives are realized as phonologically enclitic items. Such items do not exhibit feature agreement with the head noun (unlike Romance) and co-occur with articles (see example 22, from Salento Greek). Etymologically, they emerged from the genitive of the weak form of personal pronouns 41 . Another type of possessive in Italiot Greek is dikommu/dikossu/dikottu, which corresponds to standard Greek [dikos+enclitic genitive pronoun]. In standard Greek, this cluster agrees with the head noun as regards its linearly first item and has the same distribution as adjectives; it is semantically equivalent to the simplex possessive mentioned hitherto but it is more emphatic-even contrastive-than that (to diko mu vivlio = "the book that is mine and not anybody's else"). In Italiot Greek, it is mostly found postnominally (see example 23, from Salento Greek), although some speakers marginally accept it in prenominal position. In the written sources of Calabria Greek, where postnominal adjectives are articulated in definite DPs, it is usually articulated, too. According to traditional descriptions of Italiot Greek, the item diko(s) displays agreement with the head noun (see for instance the paradigm given in Condemi 1995, pp. 156-58), like in standard Greek. Finally, again like in standard Greek, diko-is only used in combination with the enclitic possessive.  Stavrou (2000, 2019), Giusti and Stavrou (2008), Horrocks and Stavrou (1989), Kiparsky (1985), Kolliakou (1997), a.o. 41 The full paradigm of (tonic and clitic) personal pronouns of Italiot Greek is given in Appendix A.

o (orrio) libbro dikommu
Another type of possessive in Italiot Greek is dikommu/dikossu/dikottu, which corresponds to standard Greek [dikos+enclitic genitive pronoun]. In standard Greek, this cluster agrees with the head noun as regards its linearly first item and has the same distribution as adjectives; it is semantically equivalent to the simplex possessive mentioned hitherto but it is more emphatic-even contrastive-than that (to diko mu vivlio = "the book that is mine and not anybody's else"). In Italiot Greek, it is mostly found postnominally (see example 23 from Salento Greek), although some speakers marginally accept it in prenominal position. In the written sources of Calabria Greek, where postnominal adjectives are articulated in definite DPs, it is usually articulated, too. According to traditional descriptions of Italiot Greek, the item diko(s) displays agreement with the head noun (see for instance the paradigm given in Condemi 1995, pp. 156-58), like in standard Greek. Finally, again like in standard Greek, dikois only used in combination with the enclitic possessive.
Languages 2021, 6, x FOR PEER REVIEW 12 of 23 they are recent: "their starting point should be placed at a time when pressure from Romance heavily increased. As the Greek-speaking areas decreased in number of speakers and size during the last three centuries, it would be logical to claim that these analytic phenomena are related to such sociolinguistic factors and that these phenomena are not older than that" (Mertyris 2014, p. 278;cf. also Alexiadou 2017;De Angelis 2021). What is relevant for the purposes of the present discussion is that, in Italiot Greek, prepositional constructions are very marginal as a strategy of nominal genitive realization: according to the speakers, and also as attested in the written sources, the actually productive strategy is GenO, like in standard Greek (cf. also the discussion in Mertyris 2014, pp. 275-77). Table 5 sums up the main points made so far: concerning the realization of nominal genitives, there is convergence between Italiot Greek and standard Greek, both being different from Romance. As far as pronominal arguments of the noun (henceforth "possessives", 7d) are concerned, Italiot Greek features the same strategies as standard Greek 40 . Possessives are realized as phonologically enclitic items. Such items do not exhibit feature agreement with the head noun (unlike Romance) and co-occur with articles (see example 22, from Salento Greek). Etymologically, they emerged from the genitive of the weak form of personal pronouns 41 . Another type of possessive in Italiot Greek is dikommu/dikossu/dikottu, which corresponds to standard Greek [dikos+enclitic genitive pronoun]. In standard Greek, this cluster agrees with the head noun as regards its linearly first item and has the same distribution as adjectives; it is semantically equivalent to the simplex possessive mentioned hitherto but it is more emphatic-even contrastive-than that (to diko mu vivlio = "the book that is mine and not anybody's else"). In Italiot Greek, it is mostly found postnominally (see example 23, from Salento Greek), although some speakers marginally accept it in prenominal position. In the written sources of Calabria Greek, where postnominal adjectives are articulated in definite DPs, it is usually articulated, too. According to traditional descriptions of Italiot Greek, the item diko(s) displays agreement with the head noun (see for instance the paradigm given in Condemi 1995, pp. 156-58), like in standard Greek. Finally, again like in standard Greek, diko-is only used in combination with the enclitic possessive.  Stavrou (2000, 2019), Giusti and Stavrou (2008), Horrocks and Stavrou (1989), Kiparsky (1985), Kolliakou (1997), a.o. 41 The full paradigm of (tonic and clitic) personal pronouns of Italiot Greek is given in Appendix A. 40 Stavrou (2000, 2019), Giusti and Stavrou (2008), Horrocks and Stavrou (1989), Kiparsky (1985), Kolliakou (1997), a.o. 41 The full paradigm of (tonic and clitic) personal pronouns of Italiot Greek is given in Appendix A.
Finally, it is worthwhile observing that there seems to be no trace, in Italiot Greek, of "adjectival" possessives, which by contrast are available in Ancient Greek. 42 These items agree with the head noun in gender, number and case and have the same distribution as adjectives. They have not been preserved in standard Greek either. In contrast, possessive items with analogous properties are found in Asia Minor Greek (e.g., in Romeyka Pontic, Guardiano et al. 2016, pp. 134-35). Detailed investigation of their diachronic distribution would probably shed light on their structure and nature, but we have to leave this to future work.
Possessives display high internal variability across the Romance dialects of Southern Italy. One finds tonic possessives that agree with the head noun in (gender and) number and have the same distribution as adjectives 43 . As remarked above, in the dialects of Southern Italy, adjectives are mostly postnominal: thus, expectedly, the unmarked position for such "adjectival" possessives is the postnominal one 44 .
Finally, it is worthwhile observing that there seems to be no trace, in Italiot Greek, of "adjectival" possessives, which by contrast are available in Ancient Greek. 42 These items agree with the head noun in gender, number and case and have the same distribution as adjectives. They have not been preserved in standard Greek either. In contrast, possessive items with analogous properties are found in Asia Minor Greek (e.g., in Romeyka Pontic, Guardiano et al. 2016, pp. 134-35). Detailed investigation of their diachronic distribution would probably shed light on their structure and nature, but we have to leave this to future work.
Possessives display high internal variability across the Romance dialects of Southern Italy. One finds tonic possessives that agree with the head noun in (gender and) number and have the same distribution as adjectives 43 . As remarked above, in the dialects of Southern Italy, adjectives are mostly postnominal: thus, expectedly, the unmarked position for such "adjectival" possessives is the postnominal one 44 . Besides adjectival possessives, Salentino features clitic possessives not agreeing with the head noun, albeit they have a very limited distribution and are subject to the following constraints: the noun must be a kinship/relational noun 45 and must be in the singular, the possessive can only cliticize on a noun (not on an adjective or any other modifier) and never co-occurs with an article. Besides adjectival possessives, Salentino features clitic possessives not agreeing with the head noun, albeit they have a very limited distribution and are subject to the following constraints: the noun must be a kinship/relational noun 45 and must be in the singular, the possessive can only cliticize on a noun (not on an adjective or any other modifier) and never co-occurs with an article. 42 From Guardiano (2003;cf. also Guardiano and Stavrou 2019b): Languages 2021, 6, x FOR PEER REVIEW 13 of 23 Finally, it is worthwhile observing that there seems to be no trace, in Italiot Greek, of "adjectival" possessives, which by contrast are available in Ancient Greek. 42 These items agree with the head noun in gender, number and case and have the same distribution as adjectives. They have not been preserved in standard Greek either. In contrast, possessive items with analogous properties are found in Asia Minor Greek (e.g., in Romeyka Pontic, Guardiano et al. 2016, pp. 134-35). Detailed investigation of their diachronic distribution would probably shed light on their structure and nature, but we have to leave this to future work.
Possessives display high internal variability across the Romance dialects of Southern Italy. One finds tonic possessives that agree with the head noun in (gender and) number and have the same distribution as adjectives 43 . As remarked above, in the dialects of Southern Italy, adjectives are mostly postnominal: thus, expectedly, the unmarked position for such "adjectival" possessives is the postnominal one 44 . Besides adjectival possessives, Salentino features clitic possessives not agreeing with the head noun, albeit they have a very limited distribution and are subject to the following constraints: the noun must be a kinship/relational noun 45 and must be in the singular, the possessive can only cliticize on a noun (not on an adjective or any other modifier) and never co-occurs with an article. 42 From Guardiano (2003;cf. also Guardiano and Stavrou 2019b In Salentino, possessives are invariable for gender but non for number: mia = my (masc., fem. sg.); toa = your (masc., fem. sg.); soa = his, her, its; mei = my (pl.); toi = your (pl.); soi = his, her, its (pl.).  * lu meššu-ma * lu cumpari-ma ii. * sire mia * meššu mia * cumpari mia c. i. lu sire mia fessa lu meššu mia fessa lu cumpari mia fessa ii. * sir-ma fessa * meššu-ma fessa * cumpari-ma fessa iii. * cumpari fessa-ma * meššu fessa-ma * cumpari fessa-ma In Southern Calabria (and in Sicily), a further type is found, with the following properties: it occurs prenominally, it is uninflected, and it systematically attaches to articles; because of their distribution, these possessives have been dubbed "Wackernagel" in Guardiano et al. (2018, pp. 118-24 Table 6 sums up the distribution of possessives in Italiot Greek, Ancient Greek, standard Greek and Romance of Southern Italy: Italiot Greek does not have any of the possessives found in the neighboring Romance dialects.

Discussion
The data presented in Section 3 support the hypothesis that the current structure of Italiot Greek DPs has been shaped by different concurrent factors, crucially including both social factors (as for instance the pressure of Romance as the dominant language and the prolonged coexistence of the two groups) and structural similarity with Romance. The differences in terms of sociolinguistic settings between Salento and Calabria (see Section 2) are probably responsible for the different pace at which the two communities have stabilized their innovations. This is in line with Thomason and Kaufman's (1988) predictions about the impact of social factors on structural horizontal change. However, there are differences in the impact that the pressure of Romance had on specific structural 46 Some speakers accept this sequence with a strong focus intonation on bella.
In Southern Calabria (and in Sicily), a further type is found, with the following properties: it occurs prenominally, it is uninflected, and it systematically attaches to articles; because of their distribution, these possessives have been dubbed "Wackernagel" in Guardiano et al. (2018, pp. 118-24 * lu meššu-ma * lu cumpari-ma ii. * sire mia * meššu mia * cumpari mia c. i. lu sire mia fessa lu meššu mia fessa lu cumpari mia fessa ii. * sir-ma fessa * meššu-ma fessa * cumpari-ma fessa iii. * cumpari fessa-ma * meššu fessa-ma * cumpari fessa-ma In Southern Calabria (and in Sicily), a further type is found, with the following properties: it occurs prenominally, it is uninflected, and it systematically attaches to articles; because of their distribution, these possessives have been dubbed "Wackernagel" in Guardiano et al. (2018, pp. 118-24 Table 6 sums up the distribution of possessives in Italiot Greek, Ancient Greek, standard Greek and Romance of Southern Italy: Italiot Greek does not have any of the possessives found in the neighboring Romance dialects.

Discussion
The data presented in Section 3 support the hypothesis that the current structure of Italiot Greek DPs has been shaped by different concurrent factors, crucially including both social factors (as for instance the pressure of Romance as the dominant language and the prolonged coexistence of the two groups) and structural similarity with Romance. The differences in terms of sociolinguistic settings between Salento and Calabria (see Section 2) are probably responsible for the different pace at which the two communities have stabilized their innovations. This is in line with Thomason and Kaufman's (1988) predictions about the impact of social factors on structural horizontal change. However, there are differences in the impact that the pressure of Romance had on specific structural 46 Some speakers accept this sequence with a strong focus intonation on bella. Table 6 sums up the distribution of possessives in Italiot Greek, Ancient Greek, standard Greek and Romance of Southern Italy: Italiot Greek does not have any of the possessives found in the neighboring Romance dialects.

Discussion
The data presented in Section 3 support the hypothesis that the current structure of Italiot Greek DPs has been shaped by different concurrent factors, crucially including both social factors (as for instance the pressure of Romance as the dominant language and the prolonged coexistence of the two groups) and structural similarity with Romance. The differences in terms of sociolinguistic settings between Salento and Calabria (see Section 2) are probably responsible for the different pace at which the two communities have stabilized their innovations. This is in line with Thomason and Kaufman (1988) predictions about the impact of social factors on structural horizontal change. However, there are differences in the impact that the pressure of Romance had on specific structural domains: our data suggest that some domains have almost completely embraced the Romance patterns, while others have remained more immune to massive change/full substitution. These domain-specific differences cannot be explained with appeal to the 46 Some speakers accept this sequence with a strong focus intonation on bella.
intervention of sociolinguistic forces (which are assumed to have acted uniformly on all domains): therefore, we make the hypothesis that they depend on the internal structure of each domain.
We start from adjectives. In this domain, two major changes have taken place in Italiot Greek: the introduction of overt movement of the noun across prenominally merged adjectives (which are consequently linearized postnominally) and the loss of polydefiniteness. Guardiano and Stavrou (2019a) made the hypothesis that these two phenomena are connected to one another, because they are triggered by the same interrelated factors, as we discuss immediately below. First of all, postnominal articleless adjectives are not ungrammatical in Greek: they are actually the rule in indefinite DPs. Hence, both Greek and Romance produce [N A] strings that are linearly identical but structurally different. In Greek, adjectives linearized postnominally originate in a position that is different from that where prenominal ones are merged. In contrast, in Romance, adjectives linearized postnominally are generated from two different sources: the prenominal one (that appears postnominally as a consequence of noun movement) and a predicative-like position similar to that where postnominal adjectives are merged in Greek. According to Guardiano and Stavrou (2019a), a potential trigger for the increase in the amount of postnominal adjectives in Italiot Greek and the disappearance of prenominal ones is that, in the Romance dialects of Southern Italy, prenominal adjectives are very rare: hence, adjectives borrowed from Romance are systematically realized in postnominal position (as for instance in antrepo/athropo gioveno in 9a: the adjective gioveno occurs in prenominal position in Italian but not in the Romance dialects of Southern Italy; in Italiot Greek it is systematically realized postnominally) 47 . A further trigger for the realization of postnominal adjectives as articleless in Italiot Greek is the weakening of the conditions that generate polydefinite DPs, namely the weakening of case exponence on nouns and adjectives. According to Stavrou (2012Stavrou ( , 2013Stavrou ( , 2019, the article that appears before postnominal adjectives in polydefinite DPs in Greek is the spell-out of a functional head (Pred) that realizes case agreement between the noun and the adjectives originally merged postnominally. According to Guardiano and Stavrou (2019a), the weakening of case morphology on nouns and adjectives makes case agreement dispensable; as a consequence, the overt realization of Pred is no longer required, and polydefiniteness disappears. In turn, the effect of the loss of the structural configuration that generates polydefinite DPs induced structural reanalysis of the linear strings originally emerged from it.  suggest that the loss of polydefiniteness is also responsible for the changes that took place in the domain of demonstratives. In what follows, we sum up their arguments. In Greek, demonstratives are generated in the same structure as postnominal adjectives; unlike adjectives, they are able to spell out Pred: thus, no additional copy of the definite article is required. Hence, [Art N Dem] sequences correspond to polydefinite DPs like [Art N Art A]. In Greek, the [Art A] cluster can be fronted DP-initially (usually generating informationally marked DPs), thus giving rise to [Art A Art N] sequences; similarly, Dem can be fronted from the postnominal position, generating [Dem Art N] sequences, which are often associated to deictic interpretation (Manolessou and Panagiotidis 1999). Additionally, in some non-standard Greek dialects, it is often the case that, under certain phonological conditions (Guardiano and Michelioudakis 2019), the (fronted) demonstrative and the definite article are fused into a single item. In Italiot Greek, this process was generalized, to the point that the demonstrative and the article, when fused, were no more perceived as two separate elements, and [Dem-art N] sequences were reanalysed as [Dem N]. Thus, as in the case of postnominal adjectives, [Dem N] strings became available in both Greek and Romance (although originally stemming from two different sources). In Italiot Greek, when the original source of demonstratives disappeared as a consequence of the loss of polydefiniteness, the reanalysis of DP-initial demonstratives as demonstratives of the Romance type (see  for a detailed analysis) started precisely from these strings.
As a consequence of these processes, the distribution of adjectives and demonstratives in Italiot Greek ended up identical to that of the neighboring Romance dialects (where adjectives are mostly postnominal and articleless, and demonstratives are DP-initial and never co-occur with definite articles).
Diachronically, the loss of polydefiniteness in Italiot Greek seems to have taken place after the establishment of the postnominal position for adjectives and of the DP-initial position for demonstratives: as shown in Guardiano and Stavrou (2019a), in the written sources of Calabria Greek, adjectives are overwhelmingly postnominal and are systematically articulated in definite DPs; similarly, demonstratives are mostly DP-initial and co-occur with definite articles. The persistence of the polydefinite pattern in Calabria Greek, and more generally the higher degree of conservativism of this variety until recent times, can be explained in terms of the geographic and social isolation of many Greek speaking communities of Calabria.
As far as adnominal genitives are concerned, in Italiot Greek, their syntax, as shown above, has not been affected by major changes and remains very similar to that of Greek. In this domain, in spite of the massive exposure to Romance, where prepositional Free genitives are the productive type, Italiot Greek has kept the Greek inflected postnominal genitive (GenO), thus differring from Romance. Interestingly, as noted in Section 3, Italiot Greek has developed a prepositional structure that is sometimes used to realize nominal possessors. This structure has not evolved into a prepositional Free genitive of the Romance type, in spite of some similarities/overlapping strings. 48 Actually, there are structural differences between the two systems, which might have blocked the reanalysis of prepositional phrases headed by atse as prepositional Free genitives and may have acted as barriers against change, thus favoring the persistence of the GenO type. The first is that, in Greek, genitive DPs are inflected: overt genitive morphology is visible, if not always on the noun, at least on definite articles. This feature has remained unchanged since ancient times. In contrast, there is no trace of genitive morphology in Romance. We make the hypothesis that it is precisely the preservation of genitive morphology that acted as a barrier against the spread of prepositional (uninflected) Free genitives, which is also in agreement with those proposals that emphasize the role of morphology in shaping structural contact (cf. for instance the hypothesis recently made by Poletto and Tomaselli 2020 about "resilient" morphosyntax). A further difference between Greek and Romance is that, while more than one genitive modifying the same noun (in one and the same DP) is possible (although quite rare) in Romance, such "multiple" genitives are ungrammatical in Greek. Yet, since multiple genitives are not expected to be frequently found in the E-languages accessible to the speakers, it is unlikely that they are used as triggers (or barriers) for structural reanalysis. Another difference between GenO and prepositional genitives is positional freedom: prepositional genitives can be realized in various different positions with respect, e.g., to (postnominally generated) adjectives, relative clauses and other prepositional modifiers of the noun; by contrast, GenO is linked to a fixed position (after structured adjectives). This aspect is actually less transparent in Italiot Greek, because of noun raising: as already mentioned, in Italiot Greek, the noun raises over GenO and over most prenominal adjectives; as a consequence, adjectives are realized both before (if they are generated prenominally and crossed over by the noun) and after (if they are generated postnominally) GenO.
The realization of possessives as inflected clitics is well-attested in Greek, and remains unchanged in the history of the language. This strategy has been preserved in Italiot Greek as well: Italiot Greek seems not to have been affected, in this domain, by heavy interference effects due to the pressure from Romance, in spite of potential similarities. Greek and Romance possessives are superficially similar in at least two respects. First of all, in both Greek and Romance, there are [N Poss] sequences. Second, clitic possessives not agreeing with the head noun are available both in Italiot Greek and in some Romance dialects of Southern Italy. Yet, concerning the latter, the two groups exhibit important differences. First, in Greek, enclitic possessives have genitive morphology, while in Romance they do not. Second, in the Romance dialects where clitic possessives are available, they are incompatible with articles (that is probably the consequence of raising to D of the sequence [kinship noun+enclitic possessive], Giorgi and Longobardi 1991): thus, [Art N-poss] sequences are ungrammatical. By contrast, in Italiot Greek, the article is required. Third, in the Romance dialects where clitic possessives are available, these are accepted only in very limited structural configurations, while in Italiot Greek they can appear in all types of DPs. Finally, "weak" possessives of the Wackernagel type, in Romance, systematically attach to D; by contrast, enclitic possessives in Greek only cliticize on nouns or adjectives. These patterns are summarized in Table 7 49 .

Conclusions
Two major types of structural factors must be considered when analyzing potential instances of syntactic contact. One is the availability, in the empirical evidence accessible to the speakers of the target language, of sequences/items that are superficially identical in the source and in the target language ("overlapping sequences"). This aspect was explored in Guardiano et al. (2016. The other factor is the internal processes affecting the structure of specific domains in the target language. These processes can either combine with overlapping sequences, thus acting as triggers for structural reanalysis, or block the effect of overlapping sequences, thus acting as barriers against structural change 50 . The scenarios emerged from our data with respect to the interaction of these two types of factors are summarized in Table 8. If these conjectures are on the right track, further investigation is required in order to measure the amount of overlapping strings necessary for reanalysis, to define the nature of the internal processes interacting with them, and to formalize the dynamics of their interaction. A further important question that should be addressed in future work is whether our findings have broader implications for other already known (or unknown) contact situations. 49 In this domain, one interesting aspect that is worth of deeper investigation is the syntactic nature of dikommu, which in Italiot Greek appears less transparent than in standard Greek and whose distribution is partially similar to that of postnominal adjectival (and pronominal) possessives in Romance. In the absence of more detailed data, we leave this issue for future investigation. 50 Interestingly, a very similar conclusion, in the same "language-contact situation involving South Italian Greek as recipient and Italo-Romance as donor", but on a different domain (i.e. borrowing and integrating of nouns), has been reached by Manolessou and Ralli (2020, pp. 274-75): "the accommodation of loans in a language is not only the product of extra-linguistic factors (e.g., among others, degree of bilingualism [ . . . ]), but follows specific language-internal constraints which are at work throughout the process."