Differential Object Marking and the properties of D in the dialects of the extreme south of Italy

This paper discusses two case studies of microvariation in accusative marking in the Italo-Romance varieties of the extreme south of Italy. In particular, the diatopic variation displayed by the dialects of southern Calabria gives rise to peculiar patterns of alternation between presence or absence of the marker a ‘to’ in flagging the accusative. The realisation of accusative case is partially governed by semantic and referential features, i.e. specificity and animacy. In addition, the nature of the realisation of the D head results in a degree of competition between zero marking and analytic accusative marking with a . Given the century-long co-existence of Latin/Romance and Greek in southern Calabria, the relevant morphosyntactic patterns in Case-marking will also be examined from a language contact perspective. We will highlight how the relevant outcomes do not simply involve borrowing mechanisms or template copying from the lending variety but, rather, produce hybrid structures no longer ascribable to a purely Romance or Greek grammar.


Differential Object Marking in Italo-Romance
Several languages of the world display differential marking whereby specific types of arguments are flagged by particular case and/or agreement strategies. One widely discussed example involves differential object marking (DOM; Bossong 1982: 580;, namely the morphosyntactic phenomenon for which, given two subsets of direct objects in a language, one is marked differently from the other (Bárány 2018: 3). According to traditional accounts, as well as recent typological approaches (cf. Aissen 2003;Carnie 2005), DOM is triggered by certain local features or individual factors of the relevant nominals. In particular, the use of DOM is generally explained as an epiphenomenon of the activation of at least three sets of semantico-syntactic properties. DOM may be a reflex of: (i) animacy, i.e. an intrinsic property of nominals (Silverstein 1976;Dixon 1979;Lazard 1984;de Swart & de Hoop 2007: 606); (ii) definiteness, i.e. the structural codification of features related to referentiality (Lyons 1999;Aissen 2003); and (iii) other properties which are subject to discourse-related requirements, such as specificity and topicality (de Swart & de Hoop 2007;Leonetti 2008; see García-García 2005 andIemmolo 2010 for the notion of topicality). It has been argued that DOM not only defines a subset of objects on the basis of specific semantico-syntactic nominal properties, but that it also co-occurs with a high degree of transitivity (Hopper & Thomson 1980;Torrego 1998;Cennamo 2003;Naess 2004;García-García 2005 a.o.). Given the wide range of triggering factors, languages which display DOM vary considerably depending on the nature of the single features involved and their interaction. Nevertheless, the relevant licensing factors for DOM do not define clear-cut subclasses of nominals, highlighting how the Glossa general linguistics a journal of Ledgeway, Adam, et al. 2019. Differential Object Marking and the properties of D in the dialects of the extreme south of Italy. Glossa: a journal of general linguistics 4(1): 51. 1-25. DOI: https://doi.org/10.5334/gjgl.569 distribution of a differential marking has to be assessed cross-linguistically and from a micro-variation perspective (cf. Ledgeway 2018).
In the Romance domain, the phenomenon of DOM is attested in Ibero-Romance (Campos 1999(Campos : 1529(Campos -1545Escandell-Vidal 2009: 837ff a.o.), Romanian (Dragomirescu & Nicolae 2016: 920-923), 1 several African varieties of French (Roberge 1990: 105-107, 120-121) and Sardinian (Jones 1993: 65ff;La Fauci 1997: 51-53). In these varieties DOM is generally realised through the marker a (<Lat. ad 'to'), with the exception of Romanian where it is marked by pe (<Lat. (su)per 'on'). 2 DOM is also widespread across the dialects of Italy. 3 The contrast in (1a-b) from Neapolitan is representative of DOM in Romance: the specific inanimate object in (1a) occurs in the unmarked form of the accusative, witness the doubling accusative clitic 'o on the verb, whereas the specific animate object in (1b) occurs in the marked accusative introduced by a and once again doubled by the accusative clitic 'o: (1) Neapolitan

Differential Object Marking in southern Calabrian varieties
Recent fieldwork investigations 4 confirm previous observations on the properties of DOM in southern Italian dialects, but also highlight some new patterns which crucially reveal significant correlations between the presence of the definite determiner and the accusative marker a. In the following section, we shall assess and discuss the patterns displayed by two subsets of dialects spoken in the extreme south of Calabria, in the province of Reggio Calabria. The system of accusative marking we describe below is attested in several southern Calabrian dialects spoken in an area roughly corresponding to the southern side of the Aspromonte, 5 e.g. Bagaladi, Bova, Cardeto, Embrisi, Melito di Porto Salvo, San Lorenzo, San Pantaleone, San Roberto, Scido. We shall henceforth refer to this subset of varieties as "Calabrese 1 ". Significantly the area encompassed by these localities hosts several Greekspeaking villages. 6 1 DOM is attested in standard Italian as well (Berruto 1985;Benincà 1988;Berretta 1989;Zamboni 1991;Lorenzetti 2002). 2 In the Gallo-Romance dialects of Sicily DOM is signaled by da < de + ab (Rohlfs 1969: 8, 15;Manzini & Savoia 2005, II: 502); in the central Italian dialects by ma/me < in medio (ad) (Rolhfs 1969: 15;Berizzi 2013). The markers a, da, ma/me also signal dative in these same varieties (Rohlfs 1971: 333-335). 3 See, among others, Rohlfs (1969: §632;1971); Jones (1995);Sornicola (1997a;b;2011);Ledgeway (2000: Chapter 2); Manzini & Savoia (2005, II: 515ff); Iemmolo (2010); Guardiano (2014); Andriani (2015). 4 The relevant data have been collected through interviews with native speakers of Italo-Romance varieties of the extreme south of Calabria from April to July 2016. The fieldwork investigations fall within the scope of the Research Project "Fading voices in southern Italy: investigating language contact in Magna Graecia", carried out at the University of Cambridge. Additional information can be found on the project's website: www.greekromanceproject.wordpress.com. 5 Thanks to the reference suggested by an anonymous reviewer, we also include the accusative marking system of Agnana Calabra as fully consistent with Calabrese 1 (Bentley et al. 2015: 141-142). Agnana Calabra is a village of the north-eastern side of the Aspromonte massif. 6 At the time of the investigations (April-July 2016), the Italo-Greek variety of Calabria was still spoken by a small number of predominantly elderly people in five villages: Bova (Marina), Chorío di Roghudi, Con-in the dialects of the extreme south of Italy Art. 51, page 3 of 25 In Calabrese 1 DOM surfaces on nominal expressions denoting humans. Typically, they correspond to highly referential nominals in definite and specific contexts (Abbott 2006), including DPs with overt definite determiners (2a), personal pronouns and proper names (2b), kinship terms modified by possessives (2c), DPs headed by universal and indefinite quantifiers (2d), bare demonstratives (2e) and bare quantifiers (2f) with human reference (cf. Ledgeway 2018 . The lengthening of /a/ is often not detectable, so that the quantitative difference between the bear fsg definite determiner and the articulated form is blurred. In order to graphically disambiguate the FSG definite determiner and the articulated form from the marker a, we represent the fsg definite determiner as 'a, the articulated form as â and the marker as a.
The marker a co-occurring with the non-aphaeresized forms of the definite determiners triggers the lengthening of the /l/: Vitti a-ll' atri ajeri. saw.1sg to the.mpl others yesterday 'I saw the others yesterday.' The evidence in (2a-f) shows that Calabrese 1 DOM occurs whenever Dᵒ is lexicalised by a determiner, a pronominal determiner or a raised N.
Although DOM is generally not licensed with nominal expressions prototypically conveying indefiniteness, it does occur when the direct object refers to a specific individual (Diesing 1992;Abbott 2006). Thus, the two sentences in (3) are not interchangeable. (3) San Pantaleone a. Petru mazzau du previti. Petru killed.3sg two priests 'Petru killed two priests (=two non-specific priests).' b. Petru mazzau a du previti chi canuscia jeu. Petru killed.3sg to two priests whom knew.1sg I 'Petru killed two priests (=two specific priests) whom I knew.' While du previti in (3a) refers to a set of two non-specified priests, a du previti in (3b) denotes two referents which are identified and particularised in the speaker's mind (Leonetti 2012: 296ff;Gianollo & Silvestri in press). Therefore, in Calabrese 1 DOM is a syntactic reflex of the specific interpretation of DPs and it occurs with highly referential DPs, including definite DPs and indefinite DPs when they refer to human, concrete, singular and count individuals (Table 1).
In Calabrese 1 only specific animate direct objects are marked through DOM.  Direct objects introduced by a definite determiner do not bear a differential accusative marker even when they denote referents which are explicitly presupposed in the speech participants' knowledge: San Luca Vitti i cotrari fimmini i Riggiu chi canusci tu. saw.1sg the.pl kids.pl female of Reggio that know.2sg you 'I saw the girls from Reggio whom you know.' (ii) San Luca U previti, a mastra e i cotrari mi salutaru. the.msg priest the.fsg seamstress and the kids me.obj= wave.3sg.prt 'The priest, the seamstress and the kids have waved at me.' Non-aphaeresized forms of the definite articles occur when the following word begins with a vowel: Mi cicasti l' occhiu. me= blinded.2sg.prt the.msg eye 'You have blinded me (lit.= you have blinded my eye).'

(iv) San Luca
Vippi l' acqua fridda. drank.1sg.prt the.fsg water cold.fsg 'I have drunk some cold water.' (v) San Luca Vitti l' Armandu / la Ida / l' atri. saw.1sg.prt the.msg Armando / the.fsg Ida / the.mpl others 'I saw Armando/Ida/the others.' Also, non-aphaeresized forms of the definite article surface when the preposition a (<ad), in its locative function, precedes the definite DPs. In these cases, a triggers the lengthening of /l/ in Gioiosa Ionica (vi) whereas in San Luca /l/ undergoes also palatalisation (vii).

(vi) Gioiosa Ionica
Vaju a-lla casa i tata / a-llu mari. go.1sg.prs to-the.fsg house of dad / to-the.msg sea 'I am going to my dad's/to the seaside.'

(vii) San Luca
Staiu jendu a-glia casa d'a Maria / a-gliu mari. stand.1sg.prs do.ger to-the.fsg house of.the.fsg Maria / to-the.msg sea 'I am going to Maria's/to the seaside.' Crucially, DOM in Calabrese 2 is found with some of the configurations that license DOM in Calabrese 1 (2 a-f), i.e. whenever the head Dᵒ is lexicalised by a pronoun or a raised N and whenever Dᵒ is empty: 9 Among Calabrese 2 dialects, the conservative variety of San Luca exhibits the requirement that proper names be preceded by an expletive determiner (Longobardi et al. 2013). Given such requirement, DPs with proper names fail to license DOM (cf. also De Angelis in press), witness the example (7a). On the contrary, in the Calabrese 2 dialect of Gioiosa Ionica, which does not license expletive articles, proper names occur with the marker a (7b): 10 To sum up, the two subsets of southern Calabrian varieties exhibit two different types of accusative marking: in Calabrese 1 DOM marks DPs denoting human individuals, which are highly referential (i.e. denoted by pronouns, proper names or definite DPs) or, if denoted by indefinite DPs, encode specific individuals; in Calabrese 2 DOM is licensed with the same set of DPs if and only if no definite determiner occurs. From a pragmatic 9 The evidence brought up by Calabrese 2 finds a striking parallel in patterns of DOM displayed in Sardinian (Jones 1993: 65-76; and Corsican (Neuburger & Stark 2014). 10 The phonetic changes triggered by /a/ (<ad), as described in fn.8, help prove that in San Luca an underlying ad as a mark of the direct object is ruled out: (i) San Luca Vitti *a-glia Maria / *a-gliu Petru. saw.1sg.prt to-the.fsg Maria to-the.msg Petru 'I saw Maria/Pietro.' point of view, the DPs denoting highly referential and specific entities express a presupposition, i.e. a piece of information that the speaker implicitly assumes to be already known by all discourse participants (cf. Stalnaker 1974;Schwarz 1977: 247;Enç 1991;Diesing 1992; see also Jäger 1995;Raposo & Uriagereka 1995;Leonetti 2008;Gianollo & Silvestri in press). Specifically, we adopt a pragmatic notion of presupposition (Dryer 1996: 487;Schwenter 2005; see also Larrivée 2014: 116) whereby presupposition corresponds to activated and discourse-old information, i.e. information available to both the speaker and the hearer at a given point in discourse. We therefore assume that in the varieties of southern Calabria DOM is a reflex of the semantico-pragmatic entailment of presuppositionality.
The patterns of accusative realisation of Calabrese 2 exhibit peculiar properties which need to be accounted for. First of all, the direct object is never differentially marked if the definite determiner occurs (7a). Contrastingly, if the definite determiner is absent, the accusative is marked by a when conveying a presuppositional reading (8a vs 8b): In what follows a structural interpretation of the accusative configurations attested in both sub-sets of Calabrese varieties is provided.

Interpreting Calabrese 1
In Calabrese 1 DOM represents the analytic realisation of structural accusative Case (Vergnaud 2008(Vergnaud [1977; Chomsky 1981). Accusative is licensed when v enters into an Agree relation with an accessible DP able to value its unvalued features (Chomsky 2000;2001), giving rise to both long-distance feature-checking operations of Case valuation and agreement. By way of illustration, consider (9a), where the v head functions as a Probe and exhibits unvalued φ-features which need to be valued in the course of the derivation. DPs that display an unvalued Case feature (uCase) are accessible Goals for the Agree operation, which makes them active Goals. In (9a), v and the direct object fulfill these requirements and establish an Agree relation.
The feature valuation is bidirectional. The Probe's unvalued φ-features are valued by the Goal and the Goal's unvalued Case feature is valued by the Probe (9b). Agree results in the valuation of accusative on the direct object DP and valuation of person and number on v. This operation is essential in order for the derivation not to crash. In particular, the DP must enter into the Agree relation with the Case ACC -valuing Probe, otherwise the Case Filter is violated (Vergnaud 2008(Vergnaud [1977; Chomsky 1981).
We observed that in Calabrese 1 the marker a obligatorily occurs whenever a pronominal D or a raised N occupy Dᵒ. 11 In southern Italian dialects personal pronouns are the syntactic expression of Person (Longobardi et al. 2013 (Harley & Ritter 2002). Both first and second persons are also intrinsically human. The accusative of the pronouns is realised both through the marker a and the selection of the morphological non-nominative form (11a). It follows that there is a close connection between the semantic feature of first/second person, i.e. humanness, the pragmatic entailment of presuppositionality (2.1.2) and the marked accusative. 12 In Calabrese 1 demonstratives functioning as third person object pronouns behave like the first/second personal pronouns if denoting a human referent 13 (11a), whereas they do not exhibit DOM if referring to [-human]  Based on this evidence, two semantic and syntactic types of third person pronouns can be identified. Type A (Table 2) corresponds to human referents and, by implication, presuppositional third person pronouns (11a). We assume that Type A is endowed with a Person feature, which in our account includes humanness by default. 14 This assumption departs from Harley & Ritter's (2002) account whereby the third person pronoun is not a true personal form. Type B third person pronouns refer to non-human entities (11b), therefore failing to convey the default presuppositionality entailments shown by 1 st and 2 nd pronouns and lacking Person (see also Anagnostopoulou 2003: 267-272;Richards 2008). 11 The same generalisation cannot be put forward for Calabrese 2 as a whole, as no N-to-D movement is attested in the dialects of San Luca and Gioiosa Ionica. 12 Indexicals are prototypically presuppositional in that they determine a referent only in conjunction with elements of the utterance context, i.e. the speech act presupposes a speaker and an addressee (Abbott 2010: 180). For implicational relations between Person and animacy see Adger & Harbour (2007); for Person and definiteness/specificity see Richards (2004;. 13 Some Sardinian varieties avail themselves of parallel DOM patterns (Jones 1995: 48-61). In (26) we record two different patterns of Sardinian. 14 However, in Ariellese (a variety of Abruzzese) the accusative of the animate third person pronouns is not marked with a (D'Alessandro 2012; see also §4.1).
Proper names and singular kinship terms equally denote [+human/+definite] referents and are presupposed by the speech act participants. Therefore, they assimilate to third person DPs of the type A and are endowed with [+Person].
When specified [+human], in Calabrese 1 other third person nominals (e.g. definite DPs, indefinite pronouns, bare universal quantifiers and demonstratives) display the same behaviour as personal pronouns. In such cases, they semantically and syntactically cluster with the first/second personal pronouns, i.e. type 3A in Table 2, and display DOM. By contrast, third person nominals of type B are indefinite, non-human or inanimate and nonpresuppositional on the basis of their inherent features, such that DOM consistently fails to obtain with them. Given these assumptions, a strong correlation holds for Calabrese 1 Theoretically, we therefore claim that, in order for Agree to occur, v is endowed with a strong unvalued D-feature, i.e. Person, semantically endowed with humanness. Direct object DPs with third person morphology can display two types of values, i.e. [± Person]. In order to derive (13a), the feature [+Person] needs to be probed by v in the course of the accusative valuation.
(14) As the direct object DP u previte 'the priest' is [+Person], it functions as a non-defective Goal for the head v to probe the Person feature. In order for the uCase feature to be valued through the D head on the [+Person] DP, the marker a must be inserted. Conversely, the Person specification is redundant for inanimate, generic, non-presuppositional DPs ( Richards 2008: 14ff). We assume that they exhibit [-Person], i.e. the absence of the feature. Therefore, the DP acts as a defective Goal (cf. Roberts 2018: 118) and Person is left unvalued in the derivation. As a result, the absence of Person is spelt out as third person morphology, i.e. the default morphological exponence (cf. Benveniste 1966: 256), no marker is inserted and accusative is left unmarked. Nevertheless, if the DP lacks Person, it still behaves as a non-defective Goal for v to value Case. As a result, the DP's Case is valued. In sum, in Calabrese 1 the accusative of [+Person] DPs is always licensed through DOM. 15 In particular, the accusative is valued through Agree: v values its D-feature on the animate DPs which are endowed with [+Person] and the resulting structure, which is spelled out with the accusative marker, is a scattered configuration due to the projection of KP (Giorgi & Pianesi 1997; see also Kremers 2009). Person is left unvalued as the nonhuman and generic DPs do not display a Person feature. This leads to default third person morphological exponence and no DOM.

Interpreting Calabrese 2
The patterns of accusative realisation of Calabrese 2 exhibit peculiar properties, as the direct object is never differentially marked if the definite determiner occurs (16a). Conversely, if the definite determiner is absent, the accusative is marked by a when conveying a presuppositional reading (16b vs 16c):  (Jaeggli 1982;Torrego 1998;Leonetti 2008) and in Romanian (Dobrovie-Sorin 1990), the specific interpretation of internal arguments often requires clitic doubling. This correlation, which is valid for the set of data discussed in this paper as well, proves stronger when the object is a tonic pronoun (e.g. Neapolitan; Sornicola 1997a; b; Ledgeway 2009: 831ff).
In Calabrese 2 , as in Calabrese 1 , generic and inanimate DPs do not require DOM, whereas DPs conveying a presuppositional entailment require a on condition that the D head is not lexicalised by a definite determiner. DOM is also obligatory with indefinite DPs if bearing a presuppositional interpretation. The point of variation between the two systems is represented by the cases in which definite DPs are headed by the definite determiner and interpreted as presuppositional: this configuration leads to the realisation of the marker a in Calabrese 1 (18a) and to the sole presence of the definite determiner in Calabrese 2 (18b): In Calabrese 2 the opposition between [±presuppositional] is morphologically blurred when the definite determiner is lexicalised. The accusative in Calabrese 2 is expressed through three different configurations of the direct object DP, which is either marked with a (<ad), is headed by the definite determiner (det), or has no mark (Ø), as summarised in Table 3.
The accusative valuation on the DPs endowed with [+Person] can only result in two types of configurations, either one showing the definite determiner or the other requiring a. Therefore, the definite determiner and the marker a are in complementary distribution. In order to explain this micro-variation in accusative Case marking, we propose that accusative DPs in Calabrese 2 undergo the same licensing mechanism as in Calabrese 1 , but with the crucial difference that, when Agree takes place in Calabrese 2 , the lexicalisation of the D head through the definite determiner is the necessary and sufficient condition for v to value the uCase feature on the [±Person] DP. Arguably, the Agree operation between v and the DP in which the D-position is filled with the definite determiner is the default mechanism of accusative valuation. The Person feature is valued covertly in the case of [+Person] DPs and no differential marker is required (16a, 18b). The DP is able to mark the accusative with the sole presence of the definite determiner and the KP layer is not projected. Therefore, the definite determiner spells out the syncretic head of a syntactically definite accusative DP, as shown in (19): The accusative is realised through a scattered DP structure where the marker a is merged in the head of KP (21). Finally, when the DP exhibits no Person, no marker a can be merged in the structure, resulting in the default third person morphology.

(21)
In Calabrese 2 the lexicalisation of the D head through the definite article is sufficient for Agree to check a bundle of features, including [+Person] and accusative Case. This property is related to the definite article's ability to head a DP that acts as a non-defective Goal for Case valuation. At the same time, the definite article is also the marker of [+Person]. Therefore, its presence makes the insertion of a redundant. This pattern of accusative licensing is available to other Romance varieties, i.e. Sardinian, Corsican and Catalan. However, it is significant that such a mechanism of accusative valuation is found in southern Calabria, an area of pervasive and long contact between Greek and Latin/Romance. In the same varieties, the lexicalisation of Dᵒ also affects the realisation of dative. In order to mark dative, Calabrese 2 displays the Greek-style genitive marker di only when D is filled with the definite article. 16 For a detailed description and a structural account of the complementary distribution of the dative Case realisation in southern Calabria we refer to Ledgeway et al. (2016; forthcoming; see also Chilà 2017). In the next section we discuss the possibility that language contact determined the complementary distribution between the definite article and DOM.

Case-marking and the Greek-Romance contact in southern Calabria
Some Romance dialects of southern Calabria display indisputable structural influences from the Italo-Greek varieties spoken in the same area. 17 In the domain of morphosyntax, a property exhibited by Italo-Greek which has been transferred into some surrounding Romance varieties concerns the requirement that Dᵒ be lexicalised with an expletive article in conjunction with proper names. Variation in the use of the article with proper names is the reflex of a deeper abstract property of nominals, namely Strong Person in Longobardi & Guardiano's (2009) terms. This property requires D to be visible in order to license a referential interpretation of its associated noun (Longobardi 1994;. Southern Italian dialects display this property as proper names overtly raise to D. Yet, such movement is blocked in the variety of San Luca (22a), which belongs to the Calabrese 2 sub-set in our discussion, in the Romance dialects of Salento (22b), as well as in all Greek varieties of southern Italy (23), where the expletive definite article is used instead: Such an evident contact feature between Greek and Romance has arguably played a crucial role in the patterns of case marking observed in Calabrese 2 . Italo-Greek determiners display cumulative overt exponence of features, as not only are they able to express gender and number, but they also convey Case as part of a portmanteau realisation (Ledgeway et al. in prep.). We can suppose that the tripartite inflectional Case system of Calabrian Greek ("Case System I" in Table 4) was transferred into Romance. Within the syntax of the Italo-Romance dialects, two possible scenarios in the Case marking arose, "Case System II" and "Case System III" in Table 4.
"Case System II" occurs if nouns cannot move to the D area, as Dᵒ is filled with the definite determiner. The definite article lexicalises j-features and Case at the same time, exactly as in Calabrian Greek. In this configuration, no further marker is needed to mark accusative. In "Case System III", nominals lexicalise Dᵒ in order to satisfy referential requirements. In this configuration, no definite determiner can be lexicalised. Hence, accusative is marked by DOM.
Given this account of the case systems of Greek and Romance varieties of southern Calabria, the point of interference from contact becomes evident: in both Case System I and Case System II the mechanism of structural accusative valuation requires the merger of a DP whose head syncretically realises jand Case features. Given that Case System II is found in dialects spoken in the past by bilingual Romance/Greek speakers, we can suppose that it emerged as an innovative configuration, representing a hybrid structure in which the realisation of accusative and dative cases results from the combinations of (morpho)syntactic properties of Calabrian Greek, the competing language. The emergence of such forms is extremely revealing for the debate on the notion of transfer in the context of language contact and change. According to Aboh (2015), interfaces play the relevant role in the selection and recombination of linguistic features. In particular, the syntax-discourse/semantics interface within the noun phrase represents a vulnerable area of the grammar for transfer to happen (Aboh 2015: 171ff). In the competing grammars of southern Calabria, features such as Definiteness, Person and Specificity/Presuppositionality, visible at the syntax-discourse/semantics interface, have determined two different mechanisms, which can be defined as pattern transmission 18 (24) and feature transmission (25) (Aboh 2015: 172). In the varieties of southern Calabria pattern transmission is witnessed in Case System II. Such hybrid Case configurations are the result of selecting and retaining Person feature (F x ) as well as Definiteness features (F x ) from Greek (L x ), the historically competing language, and adopting their semantics and licensing properties: Bilingual speakers in the villages of southern Calabria were exposed to at least two grammars and can therefore be identified as the locus of linguistic change 19 (cf. Roeper 1999;Lightfoot 2006 18 Cf. the mechanism of PAT(tern) transmission in Matras & Sakel (2004;. 19 In the Aspromonte area (Bovesìa), Greek was still spoken extensively alongside the Romance varieties at the end of the 18 th century (Martino 1980: 311;Piromalli 1996: 419-420; see also Rohlfs 1972;Fanciullo 2005Fanciullo -2006. The same chronology cannot be maintained for the area where Calabrese 2 is spoken (Locride area). From the scarce sources available, one can speculate that there Greek ceased to be used by Greek-Romance bilingual speakers by the end of the 17 th century (Alessandro De Angelis p.c.).
Such younger speakers act as the agents of the linguistic change in the Case system, as they received the inputs necessary to complete the pattern transmission (25) in their emerging grammar. Even though Greek later ceased to be spoken, the language transmission was successful and resulted in the observed configuration of accusative and dative.

A parameter hierarchy of accusative Case checking
The Case marking systems of southern Calabrian varieties display fine-grained morphosyntactic variation. To understand it, we adopt a theory that departs from the notion of Universal Grammar as a defined set of parametrised options (Borer 1984), but, rather, takes parameters to be mutually dependent: the ones which are highly local and related to the surface-oriented variation (microparameters) cluster together to form parameters of greater import (macroparameters). More specifically, macroparameters are composed of aggregates of microparameters acting in unison. 20 The hierarchy in (27) is conceived as a typological classification based on the valuation of a Case feature (acc) and [+Person] ([+per]). If the DP is not able to have [+per] valued through the D head, the marker a must be inserted in Kᵒ. The relevant varieties of the hierarchy show different patterns of accusative marking depending on whether Dᵒ is lexicalised by a nominal element and what the nature of such an element is. The Person-licensing property of the DP is given even if Dᵒ is empty. 21 Given the evidence discussed above, when v cannot value Person as the DP is [-per] (i.e. generic or inanimate), the spelt-out structure expresses a default third person morphology. The classification in (27) 21 We assume that in the relevant varieties of the hierarchy, with the exception of standard Modern Greek, Dᵒ is empty also in DPs introduced by demonstratives (see 21) which occupy the SpecDP position (Brugè 1996;Giusti 1997;Guardiano 2009 a.o.). The empirical evidence we discuss here proves that, for the same given variety, DPs introduced by demonstratives show the same patterns of accusative realisation as the DPs introduced by definite articles: One can argue that the Spec-Head relation between the demonstrative and Dᵒ allows the demonstrative to inherit the Person-licensing property of Dᵒ. so that Dᵒ is also lexicalised by proper names. Lexicalisation of Dᵒ through the definite article is the sufficient condition for Calabrese 2 (30a), Sardinian 2 (30b), Corsican (30c) and Catalan (30d) to value [+per] on the accusative DP. Some varieties of Sardinian (Sardinian 1 (31)) insert the marker a in Kᵒ if Dᵒ is occupied by a definite article (31a), whereas if Dᵒ is empty (31b) [+per] is valued without projecting the KP layer. In the Abruzzese dialect of Arielli ((32); see fn.14), the lexicalisation of the D head is able to value [+per] only when it hosts third person nominals, i.e. proper names and pronouns. 22

Conclusions
In the Romance dialects of southern Calabria, DOM is required to convey a presuppositional interpretation on the direct object which, in order to be marked as accusative, triggers the appearance of the marker a. This generalisation is not valid for a subgroup of varieties (Calabrese 2 ) which exhibit a complementary distribution between the marker a and the definite determiner. The lexicalisation of the latter rules out the realisation of DOM, irrespective of the presuppositional entailments. We have shown that this can be accounted for as a reflex of the valuation process of Case and Person features. More specifically, D bears a [+Person] feature to be valued by v, which in turn values an accusative Case feature on the DP. Once this feature bundle is checked through Agree, the accusative is licensed without the insertion of a. Independent evidence from Sardinian, Corsican and Catalan clearly suggests that this configuration is a parametric option available across Romance. As for the trigger of such a configuration in southern Calabria, the long-standing Romance-Greek contact has played a highly relevant role. The structural parallel with the properties of D exhibited by the surrounding Greek varieties is striking. The definite determiner in Calabrian Greek is the morphosyntactic outcome of a bundle of features, including Case, such that no accusative marker surfaces. The Romance dialects of southern Calabria have integrated this property of Italo-Greek syntax, filtering the grammar of accusative valuation through the Romance case licensing strategies. The result is therefore an accommodation of a Greek syntactic pattern within Romance syntax, ultimately leading to the emergence of hybrid patterns. In this respect, accusative marking follows a trend that the present-day Romance varieties of the extreme south of Italy exhibit through several morphosyntactic aspects: accusative is realised through differential marking in accordance with specific factors, among which we singled out the checking of a [+Person] feature. Finally, the variation in patterns of case marking in such a long-standing contact area contributes to the debate on the emergence and structural configurations of hybrid systems.