Interactions between Differential Object Marking and Deﬁniteness in Standard and Heritage Romanian

: The observation that not all grammatical realizations in heritage languages can be attributed to transfer from a dominant language has been emphasized in several recent works. This paper provides further arguments in this direction from heritage Romanian. As opposed to standard Romanian, the heritage Romanian data examined here do not exhibit a restriction which blocks overt deﬁniteness on a differentially marked object (DOM), when the latter is unmodiﬁed but interpreted as deﬁnite. Moreover, in heritage Romanian there appear to be differences between the differential marker and (other) prepositions when it comes to interactions with overt deﬁniteness. It is shown that the preservation of overt deﬁniteness cannot be reduced to transfer; some of the dominant languages at stake, namely Serbian and Russian are determinerless, with nominals being used bare regardless of their syntactic function. The heritage data in turn give support to a theory under which the differential marker must be structurally set aside from (other) prepositions. If the latter spell out a P projection, the differential marker is the spell out of complex internal structure of certain classes of objects, which must project at least a DP. This structural complexity for DOM is transparent in other Romance languages, where deﬁniteness is equally obligatory on the surface, if a deﬁnite interpretation is intended. Thus, the DOM-overt deﬁniteness setting in the heritage data follows from predictable paths of language variation.

Despite the presence of a typical animacy based strategy, modern Romanian stands out in an important respect-an overt definite determiner is banned if the object under differential marking is unmodified.Romanian exhibits a definite morpheme which is a realized as a suffix (Cornilescu andNicolae 2011, 2012, a.o.).An example is in (2), which illustrates an unmarked animate object with the suffixed definite.
(2) Ai have.2SG privit watched fete-le.girl.F.PL-DEF.F.PL 'You have watched the girls.' modern standard Romanian Turning to sentences with DOM as in (1-a), we notice a definite interpretation despite the absence of overt definiteness morphology.In fact, if we try to add the suffixed definite, ungrammaticality ensues.As a result, an example similar to (1-a) but with the suffixed definite is strictly out in modern standard Romanian, as shown in (3-a).Note that, despite its showing up 'bare', the differentially marked object does not alternate between a definite and an indefinite interpretation.Instead only the definite interpretation is possible.An indefinite reading of the differentially marked object can be obtained only if overt indefiniteness morphology is present, as in (3-b); in this latter sentence two indefinite morphemes are illustrated, namely the plural indefinite nişte or an indefinite form constructed from a numeral/indefinite base to which definiteness morphology is added.
( The problem of the obligatory 'article drop' (Dobrovie-Sorin 2007) in the environment of DOM with unmodified nominals has been discussed under a variety of approaches for modern standard Romanian, as shown in Section 5.Here we will be providing further remarks about its nature, status in the grammar, and limits of variation by examining data from heritage Romanian, where overt definiteness is possible with an unmodified nominal which is differentially marked.The examples under analysis come from six (6) heritage Romanian speakers, aged 28-48, and whose primary languages include Serbian, Russian and English (additional details are provided in Section 3).For these speakers, examples such as (4) are the preferred option to construct a differentially marked nominal with a definite interpretation.In the heritage data, the interaction between DOM and overt definiteness shows another distinction as compared to standard Romanian.This difference points to a split between DOM and lexical prepositions, set differently from standard Romanian.Given that the surface deletion of the suffixal definite in standard Romanian does not affect the interpretation, a PF process is commonly assumed to be behind the DOM-definite cluster reduction (see Dobrovie-Sorin 2007 or Giurgea forthcoming, which will be discussed in more detail in Sections 5.1 and 5.1.1).In turn, this PF deletion process is a more general one in standard Romanian, as it affects most of the prepositions which select accusative case.In other words, standard Romanian has a rule which blocks the definite article on an unmodified nominal introduced by a preposition which selects the accusative.In (5) we see pe in its lexical locative use 'on' which blocks the overt definite morpheme. (5) Definiteness deletion on unmodified nominals with prepositions that select the accusative In the heritage Romanian data discussed here accusative inducing prepositions do not appear to have an uniform behaviour, when it comes to the interaction with the definite suffix.As opposed to DOM pe which shows up with with overt definiteness, other prepositions can be seen with bare nominals especially if the nominal is interpreted as generic, non indentified, non unique, as in ( 6).day.F.PL 'He was in the hospital for several days.'There are, however, instances in which lexical prepositions appear to be possible with the overt definite on unmodified nominals, especially if the latter refer to definite or specific entities.This distinction, which is obviated in standard Romanian (where examples such as (7) would be ungrammatical or at most substandard), is illustrated by a comparison of sentences such as (6) with those in (7).The data from heritage Romanian examined here thus show a slightly different picture from standard Romanian when it comes to the interaction between the differential marker, prepositions more generally, and definiteness.The questions we are interested in exploring is what lies at the core of these distinctions and what they tell us about the nature of DOM.
That the differential marker in (4) patterns with prepositions which introduce a nominal with a specific, individualized interpretation as in (7) could, prima facie, be attributed to the putative nature of differential marking as a specificity inducing mechanism (see especially López 2012 for recent discussion).Contrary to this assumption we will see data with differential marking on objects with non-specific interpretation in both standard and heritage Romanian.We will show that what triggers the obligatory spell out of the definite article on unmodified nominals under DOM in the heritage data is rather a matter of the syntactic complexity of differentially marked objects and not of semantic features such as 'specificity' per se.Differentially marked objects in heritage Romanian exhibit a structural pattern predicted by parametrization in this domain; they can be assumed to project KPs, with the K functional head hosting structural Case features above the DP (López 2012;Ormazabal and Romero 2013, a.o.for DOM as KPs in Spanish, where overt definiteness is obligatory under DOM, just like in heritage Romanian).As a result of this complex internal structure, the presence of overt definitenes is not surprising.We will also see that the presence of overt definiteness on unmodified nominals with DOM is not easy to attribute to transfer from the dominant languages; we have selected here speakers of Serbian and Russian as dominant languages which do not grammaticalize determiners.As overt definiteness morphology is absent, transfer would predict the non realization of the definite suffix, also marking the structures contingent to the target (standard Romanian).
The structure of the paper is as follows.To better ground the discussion, Section 2 provides more details on the interactions between the differential marker and the definite determiner, as well as on the similarities and differences between the differential marker and prepositions in standard Romanian.In Section 3, information about the heritage speakers included in this study is provided, alongside the data, emphasizing the differences from standard Romanian.As the dominant languages of the speakers examined here include Russian and Serbian, Section 4 presents some information on nominal structure and differential object marking in these languages.As already mentioned, given that overt morphology with definite interpretation is lacking in both Russian and Serbian, the extension of definite morphology with differential object marking cannot be attributed to transfer.Section 5 turns to three prominent accounts that have been proposed for the blocking of definiteness with DOM in standard Romanian, underlying their problems when extended to the heritage Romanian data.In Section 6 it is shown that a more plausible explanation for DOM-Def in heritage Romanian should take into account predictable patterns of variation in this domain; the complex structure of these types of nominals is relevant, as it contains functional projections above the DP, a realization otherwise seen in other Romance languages and even Old Romanian.Section 7 contains the conclusions.

Differential Object Marking, Prepositions and the Definite in Standard Romanian
The puzzle of the obligatory 'disappearance' of the definite suffix in the context of the differential marker has given rise to a rich line of investigation in both descriptive functional studies (Panȃ Dindelegan 1997;Sala 1999;Mardale 2008;Nedelcu 2016, a.o.) and in more formal orientations (see especially Dobrovie-Sorin and Giurgea 2006;Dobrovie-Sorin 2007;Hill and Mardale 2021, or Giurgea forthcoming, a.o.).A further complication is given by the observation that the deletion process is seen only with objects which do not contain overt modification.Thus, the split between the examples in (8) and those in (9).The objects in (8) do not have overt modification, while those in (9) are modified either by an adjective (9-a), a prepositional phrase (9-b), or a relative clause (9-c).In turn, the presence of overt modification blocks a bare nominal under DOM; if a definite interpretation is intended on the differentially marked object which is modified, then the definite suffix is obligatory.These examples also show that the difference between modified and unmodified nominals holds both in the singular and in the plural.8), the bare objects must be interpreted as definite, an indefinite interpretation being excluded.An indefinite reading requires instead the obligatory presence of indefinite morphology, regardless of modification.To the example we have already seen in the introduction, namely (3-b), we can add the indefinite counterparts of the sentences in (8).(Hill and Mardale 2021, ex. 9a/b, p. 186) One apparent exception to the rule in (11) comes from some types of kinship and relational nouns, such as mother, father, cousin, boss, etc.These nominals can be used with overt definiteness under DOM, even if unmodified; but under this realization there is an implicit possessor ('my/your mother'), such as an understood speaker (more rarely the hearer or a third person entity to whom the relevant kinship relation applies), as in (14-a).On the other hand, when these types of nouns are used bare with DOM, they refer only to an entity which has the property of being a mother, father, etc., without having a relation to the speaker.Despite the absence of overt definiteness, in (14-b) the object must still be interpreted as definite.
( Although examples such as (14-a) might seem exceptional, the interpretation they receive indicates that the syntactic structure contains an implicit argument which acts as the possessor modifier.In this respect they can be syntactically unified with the nominals which contain overt modification, as in (9). 3

DOM and Definiteness Deletion as a PF Mechanisms
In a rich line of research (see Section 5), definiteness deletion under DOM is assumed to instantiate a phenomenon with a purely PF nature.Two properties are taken to support this conclusion.First, there is the interpretive side, which we have already pointed out: despite the absence of the definite suffix on the surface, the definite interpretation on the unmodified nominal must be obligatorily maintained.In the examples in (3), we have seen that the bare nominal only permits a definite interpretation, as in (3-a); an indefinite reading requires instead the obligatory presence of indefinite morphology, as in (3-b).
Secondly, a similar definiteness deletion process is seen with most of the prepositions in the language which select the accusative, as in (5).Another converging point between the accusative-introducing prepositions and DOM is that, in certain contexts, overt definiteness morphology becomes obligatory on modified nominals, if a definite interpretation is intended under the preposition.See the contrast between ( 5) and ( 15).The first two examples in (15) show the obligatory presence of the definite morphology, with a definite interpretation: in (15-a) the complement of the preposition is modified by a prepositional phrase, and in (15-b)  In the next section we will see some other syntactic differences between the differential marker and lexical prepositions.For now, it also needs to be mentioned that the process of definitenes deletion on unmodified nominals only affects those prepositions that select accusative case.In Romanian tonic pronouns for the first and second person singular have preserved three-way inflectional case morphology for the nominative, accusative and dative.The differential marker accepts only the accusative form 5 and in this respect is similar to other prepositions that take only the accusative, as seen in ( 17 A notable exception is the preposition cu 'with', which allows the definite on unmodified nominals under a definite interpretation, despite its selecting the accusative case.We leave aside, for the purposes of this paper, an explanation as to why cu shows a diverging behaviour in this respect.( 18 In turn, dative-introducing prepositions are not subject to the definite deletion rule.An example is below, with the prepositional-like element datoritȃ 'because of, thanks to'.As we see in ( 19), the absence of the definite marker on the nominal complement leads to ungrammaticality, as the dative desinence cannot attach directly to the base: reuşit succeeded datoritȃ thanks to mie/noroc-u-l-ui/*noroc-ui/*noroc.me.DAT/good luck-N.SG-DEF.N.SG-DAT.N.SG/good luck-DAT.N.SG/good luck 'They succeeded thanks to me/good luck.'

Differential Marking, Prepositions and the Definite in Heritage Romanian
After having introduced the background on the interaction between DOM, prepositions and definiteness, we can now turn to the data from heritage speakers.The consultants examined for this project meet the classification as heritage speakers in that Romanian is not their dominant language, nor the dominant language of the community they are part of, nor at societal level, being used only in informal contexts (Rothman 2009).The speakers have command of Romanian, but their competence diverges from that of native speakers as a result of language contact.For all the speakers in this study, the acquisition of Romanian is a result of exposure at home, and generally outside of institutional use.
The six speakers included in this survey have as dominant languages Serbian (3 speakers), Russian (2 speakers) and English (1 speaker).Turning to the Serbian speakers, their ages are 28, 40, 46, they were all born in Serbia from Romanian parents and are female.They have received instruction in Serbian, with limited institutional exposure to Romanian (1 or at most two hours of Romanian classes a week up to high school), but with extensive interaction in Romanian at home.The two speakers with Russian as a dominant language were born in Moldova (one, male, in the capital Chişinȃu and the other one, female, in Bȃlţi).They both have received institutional instruction in Russian at school.The male consultant, now aged 30, subsequently moved to Toronto at the age of 11 and continued their instruction in Russian and English; while he was still in Chişinȃu, he also had instruction in Romanian for at most two hours a week, but used Romanian exclusively at home, with his grandparents.The female consultant, now aged 45, followed institutional instruction in Russian at school, with limited exposure to Romanian in institutionalized settings, but used Romanian exclusively at home.At the age of 30, she moved to Toronto, continued using Russian as a dominant language, and then also English.The English-dominant speaker, a male, aged 48, was born in Bucharest, but moved to Toronto with his family at the age of 10; while still in Romania, he had institutional instruction only in standard Romanian.
The data reported here come from both freely occurring speech and a questionnaire addressing various aspects related to differential object marking in Romanian.The questionnaire asked the speakers to judge the grammaticality of various examples with DOM on a scale from 1 (ungrammatical) to 5 (grammatical).Data have been collected over a period ranging from the summer of 2020 to the summer of 2022.A unifying point of the examples presented here is the presence of the definite suffix on unmodified nominals in the context of DOM.In fact, the speakers mention explicitly that, for them, if a definite interpretation of the differentially marked nominal is intended, the overt definite is either the only option or a much better option than leaving the nominal bare.Various examples are presented below, first with singular objects and then with plurals (indicating in parentheses the dominant language and the age of the speaker).The examples are immediately followed by their standard Romanian counterparts.Note that the heritage speakers from Serbia normally omit clitic doubling on differentially marked nominals.

DOM and Prepositions in Heritage Romanian
In the heritage Romanian sentences above we can immediately notice that the differential marker has a different status from the prepositions when it comes to the presence of definite morphology on unmodified objects.In (22-c) the locative preposition la 'at' has a bare nominal complement.In general, the speakers explicitly mention that definiteness morphology is not necessary if the nominal embedded in the prepositional phrase is interpreted as non-specific, generic, non identifiable in the context.We repeat here the example in (6) presented in the introduction, to which we add other similar realizations.
( We thus notice an important distinction between standard Romanian and heritage Romanian; in heritage Romanian, the absence of overt definiteness is correlated with indefiniteness and non-specificity in the case of lexical prepositions.DOM normally requires the overt definite on unmodified nominals in heritage Romanian, as opposed to standard Romanian. As DOM has been claimed to give rise to readings related to specificity (López 2012; Rodríguez-Mondoñedo 2007, or Hill and Mardale 2021 for discussion), the preservation of definite morphology might not be surprising.Definite markers more easily ensure specificity and the impossibility of existential, narrow scope readings.However, more recent work (see especially López 2012 for extensive details) has conclusively shown that, contrary to traditional assumptions, DOM is not a specificity inducing mechanism.Heritage Romanian confirms this observation.First, similarly to standard Romanian, the differential marker is seen on categories which cannot be analyzed in terms of specificity, such as the animate negative quantifier (nimeni 'nobody' in standard Romanian, nimenea for some of the heritage Romanian speakers) in (25-a) or a with nominal which contains overt material signalling non-specificity.In (25-b) we include an example from Irimia (2022) 20) and ( 22) appears to function as a true definiteness marker.It imposes restrictions that are characteristic to definiteness and not to specificity.As such the entities (X) under DOM and definiteness entail that: (i) the speaker has at least 'mental contact' with X; (ii) X is unique and maximal in the current discourse setting; (iii) X is also known to the addressee.All these features are characteristic to definiteness (Lyons 1999;Chesterman 1991, a.o.).By contrast, specificity does not presuppose that an entity be known to both the speaker and the addressee, as we will also see in Section 4, in examples such as ( 40) or (41).
To sum up, heritage Romanian DOM has a distinct behaviour from lexical prepositions when it comes to interactions with overt definiteness, even if it is homophonous with a preposition on the surface.Moreover, it also preserves other important syntactic differences from lexical prepositions; in this regard it is similar to standard Romanian.We will illustrate a few important syntactic distinctions in the next subsection.

DOM and Prepositions: Syntactic Distinctions
We can start with clitic doubling.It is true that not all the heritage speakers obligatorily use accusative clitic doubling with differentially marked objects.In standard Romanian clitic doubling of DOM (leaving aside quantifiers, etc.) seems to be the preferred option.However, both heritage and standard Romanian speakers judge ungrammatical the presence of clitic doubling on lexical prepositional phrases.In (27) we adapt an example from Irimia (2021) where the pe preposition does not function as a differential marker but fulfills a lexical requirement of the predicate conta (pe) 'count (on)', being thus obligatory on inanimates too in both standard and heritage Romanian.Clitic doubling is ungrammatical in this context.Another important syntactic difference between DOM and lexical prepositions refers to passivization; while differentially marked objects are subject to passivization under the be periphrastic form, lexical prepositional phrases cannot undergo this process.The proper name object in (28-a) needs obligatory differential marking in active constructions; in the periphrastic passive the nominal is realized as a subject and the differential marker is lost, as seen in (28-b).The use of pe as required by the selectional properties of certain predicates (in this case, the predicate conta pe 'count on' we already saw above) does not permit periphrastic passivization, no matter whether the preposition is maintained or not.Thus, the contrast between (28-b) and (28-d).
( Interactions with the pronominal marker SE, in its medio-passive realization (SE MP ) confirms the same results.In Romanian, differential object marking is not possible in this context (Dobrovie-Sorin 1998; Cornilescu 1998 or Giurgea 2019, a.o.contain comprehensive exemplification and discussion related to this restriction, which is known as the SE-PERSON restriction).As a result, the sentence in (29-a) is ungrammatical in both standard and the heritage Romanian data illustrated here.On the contrary, sentence (29-b) is grammatical, as here pe is not used as a differential marker but as a lexical preposition which is not affected by SE MP .Hill and Mardale (2021, Sections 3.1.3, 5.4, 7.2.1, 7.2.2) summarize other differences between the differential marker and prepositions, which we cannot include here for lack of space.
( To summarize, this section has shown that the differential marker, despite its prepositional appearance on the surface, shows two important differences with respect to 'true' prepositions in heritage Romanian.First, it eludes a rule that imposes definiteness deletion on unmarked nominals under prepositions, under a definite interpretation.To refresh our memory, this distinction has been illustrated by the contrast between examples such as ( 20) or ( 22) vs. ( 24).Secondly, there are also syntactic differences; these hold in standard Romanian too: differentially marked objects, as opposed to prepositional objects, are subject to periphrastic passivization, but are not possible under SE MP .These latter two differences result from the differential marker acting as a structural accusative syntactically (see also Section 6), which is affected by the periphrastic passive, as the be auxiliary cannot license a structural accusative (under a sub-type of Burzio's Generalization).Prepositional phrases, on the other hand, do not realize structural accusatives, but inherent/lexical accusatives.
The former difference, however, is more puzzling.Definiteness deletion is not active in the heritage Romanian varieties discussed here, but applies in standard Romanian.We have also seen that DOM-definite sequences cannot be simply attributed to a putative nature of differential marking as a strategy for introducing specificity or individuation on the embedded object.Given this, another explanation to explore is whether the presence of overt definiteness on unmarked nominals under DOM in heritage Romanian could be due to transfer from the dominant language.In the next section we will see that the answer to this question is negative.

DOM-Def as Transfer from Dominant Languages?
The presence of the overt definite under differential marking with unmodified objects has been reported in other works on heritage Romanian.For example, Montrul and Bateman (2020) mention several such cases produced by Romanian heritage speakers in the United States, having English as the dominant language.An example is in (30-a), followed by its standard Romanian counterpart in (30-b), where the presence of the overt definite is not grammatical.
( In these instances, the presence of the definite with unmodified nominals under DOM can easily be attributed to transfer from the dominant language.In both English and French, prepositions are not grammatical with bare nominals in the singular (leaving aside some very limited exceptions, such as the English 'to school', etc.).As we see in the examples below, the definite must be present, even if the interpretation of the nominal is not definite nor specific.The same explanation cannot, however, be extended to all the dominant languages examined here.Crucially, Serbian and Russian are determinerless languages, which lack overt morphology for the definite.In the two sections below we provide some brief details on the structure of nominal phrases in these two languages.

Some Notes on Nominal Structure in Serbian
Serbian, just like closely related Russian, lacks a determiner system of the type seen across Romance (e.g., Romanian and French illustrated in this paper) or Germanic.Nominals contain inflectional affixes encoding number, gender and case but they show up bare in argumental positions (subjects, objects, indirect objects), irrespective of whether they are singular or plural (Trenkic 2003;Bošković 2005Bošković , 2009;;Progovac 1998Progovac , 2005, among many others).When present, adjectives agree in number, gender and case with the noun.Additionally, adjectives present a morphological alternation between a short (SF) and a long form (LF), which has traditionally been related to definiteness and specificity.The examples in (34) illustrate bare subjects, a bare direct object and a bare indirect object.As a determiner-less language, Serbian is famous for exhibiting the process known as left branch extraction (LBE); this explains why the adjective can be dislocated from the nominal phrase in (34-c), without producing ungrammaticality.
( In Serbian, the absence of articles entails that nominals introduced by prepositions are used bare too, regardless of their interpretation.This is indeed borne out, as we see in the example below, where the nominals carry only gender, number and case inflectional marking.This is different from Romanian, where as we have seen, there are complex interactions between prepositions and (in)definiteness marking.
( Despite the absence of grammaticalized (in)definiteness in Serbian, it should not be concluded that these categories cannot be constructed at the semantic-pragmatic level.For example, it has been claimed that the definiteness status of the nominal can be encoded via word order (Zlatić 1997;Willim 2000): generally a subject in the theme or leftmost position tends be interpreted as definite, while the rheme is restricted to indefiniteness (Raskin 1980, a.o.).Trenkic (2000Trenkic ( , 2003) ) notices this generalization is not without exception, as there are examples, such as (37-a) below, where the subject in theme position can be interpreted as an indefinite, while the rheme in (37-b) accepts a definite reading.A safer conclusion sees the (in)definite interpretation of bare nouns in Serbian as dependent on a variety of factors, with word order being just one variable.suit.'The gentleman in the blue suit won the first prize.'(Trenkic 2003, ex. 2 & 3)

Long and Short Forms of Adjectives
In turn, of particular interest in Serbian is a category traditionally referred to as the 'aspect' of adjectives.Given that this marking has been related to 'definiteness', we will say a few words about it here.A split that is seen with various adjectives (although not all of them, see especially Trenkic 2003Trenkic , p. 1405 for examples and exceptions) is that between the 'indefinite aspect' and the 'definite aspect' (Hlebec 1986;Progovac 1998;Leko 1999, a.o.).As the discussion goes, adjectives which show up in their short form, as in (38), tend to more easily be interpreted as indefinite.The long form of the adjective, which is obtained by the addition of the -i suffix on masculine singular nouns in the nominative or by tonal alternations in the plural and in other genders (Trenkic 2003, a.o.) yields an interpretation which is commonly equated with 'definiteness'.Two examples are in (39).See also the following two statements regarding the interpretation of the long form: 'in this sense, the only SC (Serbo-Croatian, our note) form for which correspondence with the could be claimed, seems to be the long 'definite' adjectival form or rather the inflections which make it longer... ' (Hlebec 1986, p. 33), or 'in SC (Serbo-Croatioan, our note), the so-called 'definite aspect' is marked on adjectives, and often corresponds to the use of definite articles in English.' (Progovac 1998, p. 174 (Progovac 1998;Trenkic 2003, adapted) However, this can hardly be a source of transfer, when it comes to presence of the overt definite under DOM in heritage Romanian.First, the distinction is only seen on adjectives in Serbian, while differential marking is not possible on adjectives in standard Romanian nor in heritage Romanian.Secondly, as Trenkic (2003) correctly points out, the difference between the short and the long form of the adjective should not be understood in terms of definiteness, but rather in terms of specificity.As such, the possible interpretations of an example such as (39-b) cover cases in which reference is made to a specific wise man, while the short form in (38-b) can refer to any wise man.As Trenkic (2003Trenkic ( , p. 1406) ) demonstrates, in order to correctly grasp the interpretive differences between the two adjectival realizations, one needs to take into account more refined distinctions when it comes to the notion of identifiability.Two subtypes are relevant: (i) contexts in which a category is identifiable to both the speaker and the hearer, and which corresponds to the classical domain of definiteness (as we have also seen in the previous section); (ii) contexts in which a category is identifiable only to the speaker, and which fall under the domain of specificity.
Additionally, Trenkic (2003) provides salient examples where the long form of the adjective is used, even if a given entity is identifiable only to the speaker, but not to the hearer, and thus cannot be linked to definiteness.Let us look at the context below proposed by Trenkic (2003); this context corresponds to a scenario in which a person is required to describe remotely what they see as they enter a room neither themselves nor their interlocutor have seen before. (40) As you enter, there is a big round table in the middle of the room.To the left, is an old piano.Next to the piano, there is a black desk, and there is a blue globe on the desk.(Trenkic 2003(Trenkic , ex. 8, p. 1406) ) The entities made salient in this discourse are clearly not known to the hearer.However, as Trenkic (2003) notices and as several native speakers of Serbian confirm, the short form of the indefinite would be at least odd here, if not even ungrammatical.The long form must be used instead so that the speaker can signal a (specific) entity they have in mind.
(  Trenkic 2003Trenkic , ex. 9, p. 1406) ) On the contrary, the overt definite with differentially marked objects in heritage Romanian examples such as ( 20) is only grammatical if the entity is identifiable to both the speaker and the hearer, thus corresponding to definiteness, and not specificity.If an entity is intended that is known only to the speaker, but not to the hearer, then the indefinite must be used under differential marking.Two examples are below, but it should also be noted that not all the heritage speakers might uniformly accept the differential marker with indefinites (thus, they restrict the special marking only to definites, and use only unmarked indefinites as direct objects).( 42 Russian is similar to Serbian as an article-less Slavic language, which does not grammaticalize definiteness (Comrie 1989;Franks 1995, a.o.). 7Russian nominals show up bare regardless of their syntactic function, for example the direct object in (45-a), the complement to the preposition in (43-b), or the subject in (44-c) and (44-d).These examples also show that the interpretation obtained by the bare nominal can be either indefinite or definite.Of course, even if not grammaticalized, definiteness can be constructed as a semantic and pragmatic category, just like we have seen for Serbian.In the absence of overt determiners, a definite interpretation can be obtained by the manipulation of various grammatical means (Franks 1995, a.o.), such as case alternations, word order, information structure (topic, focus) marking (King 1995).For example, the sentences in ( 44) indicate that a definite interpretation is more easily available on the objects that carry accusative case, as opposed to those marked with genitive-the latter tend to be non-specific or interpreted as weak non-specific partitives, as also seen in the contrast between (44-c) and (44-d).However, rules of this type are not exception-less, mirroring similar complications as seen for Serbian.
( In any case, this is a completely different system from Romanian, where case alternations are not a means to signal splits between definiteness and indefiniteness. Another aspect that has to be mentioned about Russian is the presence of a limited type of differential object marking, which affects masculine direct objects, if they are animate.The morphological means recruited for DOM is genitive morphology, with unmarked nominals using the expected accusative (see especially Mel'čuk 1980;Brecht and Levine 1986;Bossong 1991Bossong , 1998;;Halle and Marantz 1993;Rappaport 2003;Wiese 2004;Glushan 2010;Franco and Manzini 2017, a.o.).An illustration is below-the masculine animate direct object must use the genitive in (45-b), while the inanimate in (45-a) shows up with accusative morphology (which is homophonous with the nominative).However, even if both Romanian and Russian exhibit the so-called 'oblique DOM strategy' (the use of oblique morphology to signal objects that need to be differentially marked, Bossong 1991;Manzini and Franco 2016), there are crucial differences between the two languages: Romanian DOM is not gender-dependent, is a much more robust grammatical mechanism which otherwise overrides the expected animacy restrictions in various contexts, and is dependent on multidimensionality (animacy by itself is not enough).Taking all this into account, it is clear that transfer from Russian cannot be the source of the overt definite with DOM and unmodified nominals in heritage Romanian.mal'čika .boy-GEN=DOM 'He sees a/the boy.' (Bossong 1998, adapted)

Interactions between DOM and Definiteness. Possible Analyses
Two main lines of accounts have been proposed for definite article drop in standard Romanian, namely PF accounts and syntactic accounts.We will briefly examine each in this section and then turn to the heritage data in Section 6.

Definite Article Drop: A PF Process (Dobrovie-Sorin 2007)
As maximality and uniqueness entailments are preserved in standard Romanian examples such as (3-a), despite the non-pronounciability of the definite article, a robust line of research attributes the absence of the definite determiner to a process applying at Phonological Form (PF), and which does not affect the semantic side.Dobrovie-Sorin (2007) assumes a morphophonological deletion rule applying to elements which belong to the same Complex Head (a constituent obtained from two or more X 0 heads).
However, what types of complex heads, more precisely?Various rules and processes that have been formulated to account for the creation of complex heads (head raising or lowering, etc.) will fall short when applied to standard Romanian as they cannot explain why modified nominals are not subject to article drop.Dobrovie-Sorin (2007) proposes instead a process which creates an Extended Head, that is 'a type of Complex Head which is not formed by displacement', as defined in (46).Prepositions and D form an Extended Head, given that the process does not make reference to linear order, but to structural conditions.One stipulation is that at least one of the heads involved in this process must be functional; thus, article drop with nominals which are modified is blocked, given that modifiers (adjectives, relative clauses, etc.) are not functional categories. (46) Extended head where F 0 is a functional head, L 0 is a lexical head and F 0 /L 0 is an extended head (Dobrovie-Sorin 2007, ex.9) Extended Head formation applies twice, as in (47).First comes the creation of a D and N complex head, and secondly the formation of a complex head involving P and the result of the previous step.The rule in (48) subsequently applies that deletes the article, as it is found in the same Extended Head as the Preposition.As this rule applies to accusativetaking prepositions, it will also cover DOM; the latter is homophonous with a preposition and can assign only accusative case (remember the pronominal examples in ( 17)).Unification between DOM and prepositions that govern lexical cases is maintained in another PF analysis, as proposed by Giurgea (forthcoming).The author notices that the rule in (48) might run into problems in contexts involving coordination.Let us look at specific examples, such as those in (49).The problem are examples such as (49-b): here the first conjunct is a nominal which contains modification and thus the definite suffix must be preserved, while the second conjunct is not modified, and thus the definite article must be dropped.Giurgea (forthcoming) shows that one easy solution that comes to mind, namely postulating that the second conjunct is also introduced by a preposition which is not pronounced, cannot be extended across-the-board.One case which appears to indicate that the second conjunct is not individually introduced by a preposition is provided by the preposition între ('between').As Giurgea (forthcoming) notices, the overt repetition of this preposition before the second conjunct would lead to ungrammaticality.As article drop applies across members of a coordination (that can also have phrasal status), the conclusion is that P-Def adjacency is not required.Therefore, Giurgea (forthcoming) rephrases the conditions of article drop as in (51).The only two requirements are that an accusative-taking preposition has as its complement a maximal projection of N and that the maximal projection of N bundles D and N. (51) Article drop conditions (Giurgea forthcoming, ex.24 adapted) (i) The maximal projection of N occurs in the complement position of accusativetaking prepositions (with limited exceptions) (ii) The maximal projection of N consists of [D +D + N] Subsequently, Giurgea (forthcoming) assumes that definiteness deletion is triggered by the absence of a Case feature in D. Thus, the conditions in ( 52): (52) The definiteness morpheme has a null realization iff (i) it has no Case feature (ii) it occurs in a complex head [D 0 D 0+Def N 0 ] (Giurgea forthcoming, ex.43 adapted) Given that the differential marker looks like a preposition (on the surface), Giurgea (forthcoming) assumes that it will be subject to both ( 51) and ( 52); it is, thus, subject to a PF rule which deletes the definite in a complex head.Now the question is-what do these types of PF analyses tell us about heritage Romanian where the differential marker has a distinct status from prepositions, as it does not impose definiteness deletion?In the next subsection we turn to an analysis for article deletion under which the differential object marker is separated from prepositions, at least at the syntactic level.

A Syntactic Account for Article Drop: Hill and Mardale (2021)
A different take on 'article drop' is proposed by Hill and Mardale (2021).The two authors support the conclusion that pe as a differential marker should be set aside from regular, lexical prepositions, despite similarity on the surface.Pe as a differential marker does not project a prepositional phrase (PP), but is instead a D-related functional element that merges inside the DP.
The general structure for nominals assumed in this work is a counterpart of the CP.Just like the CP, the extended projection of nominals contains three important fields: argument structure (nP), the inflectional field (AgrP) and the structural domain in which discourse pragmatic effects interact with the syntactic structure (Top/Foc, DP 1 ).Following Giusti (2005), when TopP and FocP are absent, the two Ds collapse into a larger D head.
(53) DP: D 1 P > TopP > FocP > D 2 P > AgrP > nP (Hill and Mardale 2021, ex. 1, p. 182;based on Roca 2009) Given that for Hill and Mardale (2021) DOM-pe is a D-related functional element, differential object marking does not entail the merge of P in syntax.In the case of DOM, instead, the higher discourse pragmatic layer in the extended projection of nominals contains two crucial types of features: (i) [F mark ], related to notions such as speaker's affectedness, speaker anchoring, secondary topic (Dalrymple and Nikolaeva 2011), etc.;(ii) [particularize], ensuring stable reference and uniqueness interpretations.This is schematically represented in (54).In turn, these two types of features that are responsible for DOM may or not be merged separately from the D head; importantly, DOM does interfere with the D layer, as it triggers an obligatory specific reading (under Hill and Mardale 2021 assumption).Given these observations, 'definite article drop' reduces to competition between the Drelated features that characterize differential marking and the D head itself.As both these types of features are related to referentiality, definiteness and specificity, they can be assumed to collapse into one single head such that on the surface only the differential marker will be pronounced (as pronouncing just the definite would not be sufficient for overtly signaling the differently marked objects, where grammaticalized animacy is also relevant, at least in some instances).In the case of differentially marked nominals which contain modification, the D layer must be preserved such that an adequate merging site is made available for the latter.Therefore, in the context of overt modifiers both pe DOM and the definite will be pronounced, as they are both present in the structure.
In sum, the take home lesson from this analysis is that pe as a differential marker does not imply the presence of the P functional head, and thus there is no PP in this configuration.Differential marking on objects implies a D-related functional element that is merged inside the DP and which competes with the D head given that they carry similar features.Lexical prepositions, on the other hand, do contain a P head.As we have seen, in the case of prepositions a bare nominal can be interpreted either as definite or indefinite (generic, non-specific).Thus, it must be the case that the [Def] feature is either absent or deficient.For Hill and Mardale (2021) this indicates that the D head is not projected in these structures; definiteness and specificity being interpreted contextually.As a result, what looks like 'article drop' with lexical prepositions is, in fact, the absence of definiteness altogether.Instead, in these configurations, there is a P head which merges with the argumental domain of the nominal, namely nP.The N head is restructured with P and as a result P can check Case on N. Due to restructuring, there cannot be intervening material between P and N; a telling example Hill and Mardale (2021) use is the impossibility of parenthethicals breaking the P and N sequence, as in ( 55).

Doamne)
God.VOC parc.park Intended: 'S/he walks through, apparently, park.' (the park, or any park) (Hill and Mardale 2021, ex. 6b, p. 185) In turn, the elimination of the inflectional load introduced by the D layer does not interfere with the presence of modifiers; if these are not dependent on definiteness or specificity, they are possible even if the noun does not contain the definite article nor the DP layer.Such modifiers can be merged in the argumental domain, that is inside the nP.Some examples with determinerless nominals with adjectival modification under lexical prepositions have been provided in (16); another example with PP modification is below, from Hill and Mardale (2021) (Hill and Mardale 2021, ex. 7, p. 186) Note that if the nominal contains modification, the parenthetical becomes possible, indicating that there is no P-N restructuring; the nP phrasal layer must be preserved such that the modifiers can be merged.See the contrast between ( 55) and ( 57).wondeful.N.SG Intended: 'S/he walks through, apparently, the wonderful park.' (Hill and Mardale 2021, ex. 6a, p. 184) In summary, under Hill and Mardale's (2021) account, despite uniformity when it comes to definite article drop on unmodified nominals, lexical prepositions and the DOM marker have diverging syntactic natures.This leaves open the path to the possibility of distinct behavior at PF too, for example the situation we see in the heritage data, where DOM must use the definite on unmodified nominals, while lexical prepositions might be able to drop it.The more precise question is how to derive the presence of the definite with DOM as a structural realization in the heritage varieties.The next subsection contains some remarks about this point.

DOM and the Definite on Unmodified Nominals in Heritage Romanian
Although numerous aspects related to the true nature of differential object marking in Romanian (and Romance more generally) are still in need of explanation, recent research agrees on two important formal characteristics of this mechanism.First, similarly to Hill and Mardale's (2021) observations above, it is widely assumed that differentially marked objects are types of nominals with a complex structure, that need to project at least the DP layer (see, among others, Dobrovie-Sorin 1994;Torrego 1998;Cornilescu 2000;Rodríguez-Mondoñedo 2007;Tigȃu 2011;López 2012;Ormazabal and Romero 2013).Secondly, this structural complexity has a correlate on the semantic contribution of the differential marker too.Various proposals (Brugè and Brugger 1996;de Hoop 1996;Cornilescu 2000;Bleam 2005;López 2012, a.o.) have connected the dedicated morphology seen with special objects to the presence of a semantic mechanism which turns nominals into true arguments (of type e); the result is that categories of type <e,t> which act as predicates and are interpreted as part of complex predicate with V (at least at the semantic level) are blocked under differential marking.
As it is clear by now, postulating just the DP layer does not seem to be sufficient for isolating differentially marked nominals.As we have already seen, in Romanian (just like in other Romance languages), objects can show up with overt definiteness morphology but without differential marking.This holds for both standard Romanian and the heritage speakers examined here.In (58-a) we present another example of a definite animate object which is not differentially marked.The same object, given its animacy that can be grammaticalized, can be used with differential marking.If unmodified, it must show up bare (that is, without the definite suffix) in standard Romanian (58-b), but it exhibits both definiteness and DOM in the heritage data (58-c) under analysis here.
( However, the rich theoretical literature on DOM has pointed out to another parametric option in the make-up and spell-out of differentially marked categories.According to an alternative structural realization, differentially marked objects involve additional structure above and beyond the DP layer.The relevant domain is a KP layer where structural Case features are hosted and licensed.This hypothesis is prominent in accounts for differential object marking in Spanish (Rodríguez-Mondoñedo 2007;López 2012;Ormazabal and Romero 2007 for extensive discussion and motivation, a.o.).The KP hypothesis correctly derives the empirical observation that (standard) Spanish DOM is not possible in the absence of overt definiteness, if the nominal is to be interpreted as a definite.In fact, various other Romance languages with robust DOM systems pattern with Spanish in requiring the overt definite.More generally, Spanish nominals cannot show up bare under differential marking; they need to contain overt definite or indefinite determiners, as seen in the examples below.As the KP domain cannot be projected in the absence of the DP domain, the KP hypothesis makes the correct prediction: differential marking should not be possible with nominals that lack overt (in)definite markers, as those nominals lack the projection of the DP layer.In summary, a full exploration of the parametric options made available by the grammar of differential object marking can immediately derive the heritage Romanian pattern under scrutiny here.Under one of their structural realizations, differentially marked objects are categories with a complex structure, that need to project a KP layer and thus a DP layer too.As a result, in these configurations there is no competition between KP and D. If definiteness is intended, it must be spelled-out.The definite interpretation requires the presence of the D head even under differential marking; the latter is not a marker of definiteness, but it indicates the presence of a structural Case feature on the nominal that needs adequate licensing.The Case feature is independent of specificity or definiteness, and as we have seen, DOM can show up with non-specific categories.This implies that under this parametric option, if a definite interpretation in intended on the marked nominal, definiteness morphology must be overt.In standard Romanian, on the other hand, there is competition between the KP and the D head (as Hill and Mardale 2021 also show), as the Case feature might not be independent of specificity/definiteness, and this leads to the non-pronunceability of the overt definite under DOM. 9

Heritage Grammars as Autonomous Systems
Before concluding, let us say a few words about an important question: can the obligatory presence of the definite on unmarked nominals with DOM be explained in other terms, going beyond the options made available by the syntactic parametrization of DOM?For example, a possibility that comes to mind is to say that it is not just parametric syntax per se.There might be additional parsing and syntax-pragmatics interface constraints that come into play and which might force the presence of the definite on marked unmodified nominals.In a review of the acquisition of DOM, Parodi and Avram (2018) have shown that discourse-related features are the most difficult to constrain by both L1 and L2 speakers.Moreover, in this domain, interface features are particularly vulnerable.The interactions between overt definiteness and the differential marker is clearly an interface issue; in standard Romanian we see a mismatch between what is pronounced, namely the bare nominal, and the interpretation, namely the definite reading which normally involves overt definiteness morphology.In heritage Romanian, on the other hand, this mismatch is avoided.
One hypothesis would go along these lines: given that DOM projects a KP which is dependent on the DP, dropping the definite article when a definite interpretation is intended under differential marking will lead to a mismatch between syntactic structure and PF.On the one hand, the syntactic structure, and thus the semantic interpretation, require the presence of the D head with definiteness features; on the other hand, not pronouncing the definite article at PF will render the nominal similar to bare unmarked objects as in (63) on the surface, which cannot be interpreted as definite.Preserving the definite article overtly would thus amount to a simplification strategy when compared to article drop in standard Romanian: the syntactic and the phonological representation are kept uniform in the heritage data.In other words, what you pronounce is what you have in the syntactic structure, without applying processes that are characteristic just to PF (such as deletion), and not to core syntax.This also avoids referential ambiguity, as categories that are interpreted as definite are not rendered opaque by unification with bare nouns, which, in turn, cannot be interpreted as definite.Strategies of this type have been shown to backtrack outputs in heritage grammars that diverge from the target (Sorace 2011;Montrul 2008, a.o., for discussion).Adding to this the observation that the syntactic mechanisms and the PF processes required to explain obligatory article drop in Romanian (such as formation of extended heads in Dobrovie-Sorin 2007, etc.) might have a marked status in the grammar, preserving the definite article in heritage Romanian might be yet another step towards a simplification strategy.In other words, the overt definite avoids the possibility of referential ambiguity.
A potential counterargument to the simplification hypothesis comes from heritage data itself: if DOM-Def involved a 'simpler' option at the syntax-PF-pragmatics interface, less costly in terms of processing, one would expect to see it across heritage grammars.That is definitely not the case.As various studies on DOM in Romanian heritage speakers have shown, the more common pattern is the standard Romanian one, which involves definite article drop on unmodified objects (Montrul et al. 2015;Montrul and Bateman 2020, a.o.).Additionally, article drop itself could be seen as 'simpler', in the sense that it matches both the target and the dominant language input, where definiteness is not pronounced at PF, and it also avoids the morphological complexity introduced by realizing the definite suffix overtly.
Instead, rendering the overt definite is an option permitted by the parametrization of DOM as a syntactic mechanism and the expectation is that it should be actualized, as Old Romanian indeed shows.This is to be expected under theories that see heritage grammars as autonomous grammatical systems (Frason 2022), internally consistent (Polinksy 2018), and which can follow autonomous paths of development and thus diverge from other varieties of the same language (Kupisch and Rothman 2018).

Conclusions
This paper has addressed aspects of Romanian grammar which involve the interaction between differential object marking and the definite suffix.In standard Romanian, overt definiteness marking leads to ungrammaticality if an unmodified object is differentially marked, and despite the preservation of the definiteness interpretation.The heritage Romanian varieties examined here permit the co-occurrence of the overt definite suffix under DOM, raising the question of how this structural realization is to be explained.
We have shown that the presence of overt definiteness on DOM in heritage Romanian is not a result of transfer: the data involve dominant languages such as Serbian or Russian, which are determinerless.Against this background, evidence from both standard Romanian and determinerless dominant languages would predict article drop in heritage Romanian too as the default, simpler option.The preservation of the definite article in the heritage varieties follows instead from parametric options made available by the grammar of differential marking, and instantiated in other Romance languages such as Spanish and even older stages of Romanian.This provides further support to the hypothesis that heritage grammars are grammatical systems which follow systematic principles of organization and predictable but autonomous diachronic paths.
The data addressed here also strengthen the need to structurally disambiguate the differential marker from lexical prepositions, despite their homomorphism on the surface and despite their both being subject to the article drop process in standard Romanian.
Funding: The work reported here benefited from a research grant from the University of Modena and Reggio Emilia.
Institutional Review Board Statement: Not Applicable.
Informed Consent Statement: Not applicable.
privit CL.3F.PL.ACC-have.2SGwatched pe LOC=DOM fete.girl.F.PL 'You have watched the girls.' b. (*Le-)ai CL.3N.PL.ACC-have.2SGprivit watched (*pe) LOC=DOM filme.movie.N.PL 'You have watched movies.' modern standard Romanian Intended: 'You have watched some girls.' birou-l.desk/desk-DEF.N.SG 'You left the keys on the desk.' .F.SG Lit.'I have been to girl.' (Intended-'I have been to my girl's place/I have been to a specific girl's place.')b.I-a CL.3SG.DAT-have.hehas given a gift to the girl.' one) cannot count on these people.' (standard and heritage Romanian) Ion has always been counted on.' (standard and heritage Romanian) and definiteness-singular a. Johnny goes to *(the) market every week.(could be a different market) b.Cars cannot run on *(the) sidewalk.(could be different sidewaks) c.Residential areas are normally far from *(the) train station.(could be different train stations) d.The teachers need to give the book to *(the) student.(could be different students) paints in the garden.' (could be a different garden) to marry a (specific) blonde/the blonde.' D and N form a complex head: [P [Det [N ] ] ] → [P [ D 0 /N 0 Det ⊕ N ] ] (ii) P forms a complex head with D and N: [P [ D 0 /N 0 Det ⊕ N ] ] → [ P 0 /D 0 /N 0 P ⊕Det ⊕ N ] ] (adapted from Dobrovie-Sorin 2007, ex.10) (48) The definite article is deleted whenever it is governed by a preposition that belongs to the same extended head.(Dobrovie-Sorin 2007, ex.11) 5.1.1.Giurgea (forthcoming): An Updated PF Rule .M.SG 'the view towards the vegetable gardens and the mountain.' between my house and the park.' DOM D DOM:[Fmark], [particularize] [ D 1 P D 1 [ TopP/FocP TOP/FOC [ D 2 P ...]]]](Hill and  Mardale 2021, ex.47) The volunteers have helped the old lady who was having health problems.'Despite the obligatory absence of definiteness morphology in examples such as ( by an adjectival phrase. The examples in (22) contain differentially marked objects with overt definiteness in the plural: LOC=DOMhoţ/*hoţul. thief/thief.DEF.M.SG 'He will not be able to catch the thief.'LOC=DOM copii/*copiii.child.M.PL/child.M.PL.DEF.M.PL 'He scolded the children.' which contains the anti-specificity quantifier oarecare 'no matter who'.The heritage speakers confirm that DOM is not ungrammatical in this context.In turn, the overt definite in examples such as ( Some examples that have been kindly provided to us are below (we indicate in parentheses the child's age, date of utterance, and dominant language).
Similarly, in a project examining heritage Romanian children in Paris, Alexandru Mardale (p.c.) has noticed that even if the overt definite is not robust with differentially marked unmodified nominals, such occurrences do occur.
By contrast, in Romanian subjects cannot be bare, especially if found in a preverbal position.When it comes to direct objects, an obligatory (in)definite determiner is needed with count nouns in the singular (seeDobrovie-Sorin et al. 2006for exceptional, but limited count noun singulars which might be used bare) and indirect objects marked with inflectional dative case must equally use an overt determiner, as already seen in Section 2, in (19).The Romanian grammatical correspondents to the examples in (34) are below: /child-DAT.M.SG Intended 'My friend sends a book to a/the child.' ).
. As we can see, in these examples the nominal can be determinerless or can contain the suffixed definite: The data from heritage Romanian under examination here are similar to Spanish, in that overt definiteness morphology is necessary.A structural representation presupposing the KP domain immediately explains the facts.Thus, heritage Romanian reduces to the selection of a parametric representation for DOM, which is different from standard Romanian, but which is predicted by patterns of variation in this domain and which was in fact instantiated in Old Romanian too (as shown by examples such as (13)).Additionally, assuming the presence of the KP can also derive another property of DOM, which appears mysterious at first sight.We have seen that differential marking can extend even to categories that are interpreted as non-specific.The K head hosts a structural (accusative) Case feature and signals a complex internal architecture of the nominal.Features such as specificity are not relevant by themselves; what counts is the projection of the KP layer, even if specificity is not present.For example, animate negative quantifiers as in (60) cannot be interpreted in terms of specificity.However, they do contain a quantificational domain to which the KP layer is merged.Thus, they have a complex structure and need to be licensed in terms of accusative Case.Taking this into account, the structure of differentially marked objects can be represented as in (61).Unmarked nominals, on the other hand, do not contain a KP layer; they can be just NPs, for example bare indefinites as in (63) with the corresponding structure in (62); or they can be DPs, as the unmarked definites in (58-a) with their representation in (64).