ISSN 2385-4138 (digital) Isogloss 2022, 8(4)/9 https://doi.org/10.5565/rev/isogloss.174 1-30 DOM co-occurrence restrictions and their repair strategies: evidence from Romanian and Galician Brian Gravely Emory University brian.gravely@icloud.com Monica Alexandrina Irimia Università di Modena e Reggio Emilia irimiamo@unimore.it Received: 30-07-2021 Accepted: 12-10-2022 Published: 15-10-2022 How to cite: Gravely, Brian & Monica Alexandrina Irimia. 2022. DOM co-occurrence restrictions and their repair strategies: evidence from Romanian and Galician. RLLT18, eds. Jonathan MacDonald, Zsuzsanna Fagyal, Ander Beristain & Robin Turner. Special Issue of Isogloss. Open Journal of Romance Linguistics 8(4)/9, 1-30. DOI: https://doi.org/10.5565/rev/isogloss.174 Abstract Since the pioneering discussion in Ormazabal & Romero (2007), work on PCC-effects that extend beyond clitic clusters has been the focus of much investigative research (Cornilescu 2020, Sheehan 2020, Irimia in press, a.o.). Many of these effects have been shown to arise with animate nominals, in Romance particularly those bearing differential object marking (DOM). More work, however, is needed on documenting how these types of co-occurrence restrictions can be repaired. In this article, we discuss repair strategies in contexts containing DOM from Romanian and Galician as compared to Spanish; we show that an understanding of these facts revolves around the typology of available [PERSON] licensing positions along the clausal spine. We show that Romanian and Galician exhibit multiple licensing positions within various licensing domains, both within and beyond the vP. We also discuss data from both Peninsular and 2 Isogloss 2022, 8(4)/9 Brian Gravely & Monica Alexandrina Irimia Argentine Spanish that differ in this regard and demonstrate that they have fewer licensing positions than Romanian and Galician. Keywords: Person-Case Constraint, differential object marking, PCC, DOM, animacy, licensing, repair strategies. 1. Introduction Many languages present splits in the morphosyntactic behavior of direct objects as instantiations of DOM (Comrie 1989, Bossong 1991, 1998, Aissen 2003, De Swart 2007, a.o.). Generally, specifications related to humanness, animacy, specificity, definiteness, etc., trigger the obligatory presence of dedicated marking. A typical example comes from Spanish where the animate definite in (1a) requires a marker which is homophonous with the dative preposition; the inanimate in (1b), on the other hand, does not permit the same marker (see especially Torrego 1998, López 2012, Fábregas 2013, Ormazabal and Romero 2013, a.o.). (1) Spanish direct objects (López 2012:12) 1 a. Juan vio *(a) la chica Juan see.PST.3 SG DOM the girl ‘Juan saw the girl.’ b. Juan vio (*a) la casa Juan see.PST.3 SG DOM the house ‘Juan saw the house.’ Given its robustness cross-linguistically, DOM has been a topic of inquiry for linguists from various orientations, both descriptive-typological and formal generative. In a pioneering contribution, Ormazabal & Romero (2007, 2013) (henceforth O&R) showed that important insights into the nature of DOM come from the types of co-occurrence restrictions it gives rise to. As observed by O&R (2013), in many varieties of Spanish, DOM is not possible in a configuration that also contains a dative which is clitic doubled (2).2 (2) Spanish: DOM ungrammatical with clitic doubled dative (O&R 2013:224) Le=enviaron (*a) todos los enfermeros CL. DAT.3SG=send. PST.3PL DOM all the sick a la doctora Aranzabal DAT the doctor Aranzabal ‘They sent all the patients to the doctor Aranzabal.’ 1 Abbreviations: ACC = accusative, CL = clitic, DAT = dative, DOM = differential object marking, EA = external argument, F = feminine, GEN = genitive, LOC = locative, M = masculine, NEG = negative, PL = plural, PRES = present, PST= past, REFL = reflexive, SG = singular. 2 It should be noted that not all speakers of Peninsular Spanish accept the lack of DOM in (2) as grammatical. For reasons of space, we do not address this distinction or how it possibly ties in with the notion of scales typically used in determining when DOM surfaces (e.g. Aissen 2003). DOM co-occurrence restrictions Isogloss 2022, 8(4)/9 3 O&R (2013) also observed that these co-occurrence restrictions can be repaired in some specific ways. This paper is interested in precisely these problems, going beyond Spanish into the larger Romance picture: when does DOM create co- occurrence restrictions and what are the repair strategies? We compare Spanish with Romanian and Galician, emphasizing that the three languages show similar co- occurrence restrictions triggered by DOM. 3 However, the latter two languages appear to have more extensive repair strategies than Spanish. Specifically, they boast the ability to (accusative) clitic double the direct object in order to avoid co-occurrence restrictions with clitic doubled datives. These types of differences have not been investigated formally and require more detailed analysis. This paper aims at making a step in that direction. The structure of this paper is as follows. In §2, we briefly review the literature on DOM with respect to the approaches that support individual licensing positions along the clausal spine (López 2012, Irimia 2020, a.o.). Additionally, we outline how the availability of these licensing positions affects the repair strategies that we find with certain DOM+cliticization patterns. In §3, we present the co- occurrence restrictions in Romanian and argue for multiple [ PERSON] licensing positions within the vP and beyond. In §4, we discuss the data from Galician. As Galician has very restricted differential marking patterns and no clitic doubling with full DPs, we resort to the data of clitic doubling with personal pronouns. In §5, we address the case of Argentine Spanish and Peninsular Spanish, both of which show the lack of repair strategies observed in Romanian and Galician. We discuss specific patterns that arise in these Spanish dialects and how they fit into the theoretical contributions we make in this paper. In §6, we make concluding arguments regarding the licensing positions shown in the aforementioned sections. 2. DOM licensing and co-occurrence restrictions Ormazabal & Romero’s seminal work has brought to the fore a topic that had previously received less attention, namely co-occurrence restrictions involving DOM. The theoretical proposal made by these authors is predicated on the lack of multiple Case licensing positions within the vP; this makes the licensing of multiple arguments specified with a Case feature impossible within vP. As we have seen in (2), ungrammaticality arises when differential marking remains on the direct object in the presence of a clitic-doubled indirect object. O&R show that such examples produce PCC-like effects related to animacy, or what we consider here to be the licensing of [PERSON], with a direct link to the overt exponent that is DOM. As these authors note, without clitic doubling of the indirect object, regular animate DPs may not avoid DOM and grammaticality ensues, as demonstrated by the contrast between (3a) and (3b). 3 Throughout §2-4, we make reference to data from Peninsular Spanish which we refer to simply as ‘Spanish’. We discuss data from Argentine Spanish in §5. 4 Isogloss 2022, 8(4)/9 Brian Gravely & Monica Alexandrina Irimia (3) DOM and co-occurrence restrictions in Spanish a. Enviaron *(a) Mateo a los doctores send.PST.3PL DOM Mateo DAT the doctors b. *Les=enviaron a Mateo a los doctores CL.DAT.3PL=send.PST.3PL DOM Mateo DAT the doctors Intended: ‘They sent Mateo to the doctors’ Much like the data in (2), both Galician and Romanian rule out DOM in the presence of a clitic-doubled dative in certain contexts. (4) Galician a. *Mandáron=llei a el ói médico send.PST.3PL=CL.DAT.3SG DOM he DAT.the doctor Intended: ‘They sent him to the doctor.’ Romanian (Cornilescu 2020:119) b. *Comisia lei=a repartizat pe mai mulţi board.the CL.DAT.3PL=have. PRS.3SG assigned DOM more many medici rezidenți unori foști profesori de-ai lor medical residents some.DAT.3PL former professors of-GEN theirs Intended: ‘The committee assigned several medical residents to some former professors of theirs.’ Our investigation focuses on the fact that both Romanian and Galician have a repair strategy that involves the doubling of the direct object. In contrast to the examples in (4), the differentially marked objects in (5a) and (5b) are clitic doubled by a clitic; as such, they are grammatical with a clitic-doubled dative.4 (5) Galician a. Mandáron=ll=oi,k a elk send.PST.3PL=CL.DAT.3SG= CL.ACC.M.3SG DOM he ói médicoi DAT.the doctor ‘They sent him to the doctor.’ Romanian (Cornilescu 2020:119, adapted) b. Comisia ii lk=a repartizat board.the CL.DAT.3SG CL.ACC.M.3SG=have.PRS.3SG assigned pe fiecarek rezident uneii foste profesoare DOM each resident some.DAT.F.SG former professor a lui GEN his ‘The committee assigned every medical resident to some former professor of his.’ 4 It should be noted that although i in (5b) is a dative clitic, it is tonic in nature and, thus, does not exhibit the same type of dependency on a host as commonly found in other clitics across Romance. DOM co-occurrence restrictions Isogloss 2022, 8(4)/9 5 Although there are individual differences between the two languages, we claim the differences in grammaticality that arise from (4a) vs (5a) and (4b) vs (5b) may be accounted for by positing the need to resort to further licensing heads outside of the vP domain. As shown in Saab (2022), this repair strategy is absent in Spanish, even in the dialects in which clitic doubling of full accusative DPs is possible (e.g. Argentine Spanish). (6) Argentine Spanish DOM co-occurrence restrictions (Saab 2022:34) *Ayer sei=lak=presenté yesterday CL.DAT.3SG=CL.ACC.F.3SG=introduce.PST.1SG (a) Maríak a Juani DOM María DAT Juan Intended: ‘Yesterday I introduced María to Juan.’ Thus, there is something to be said regarding why Romanian and Galician are able to resort to this repair strategy while Spanish is not. We leave the discussion of the specifics for Spanish to §5. Before moving on to the data, we wish to make several theoretical assumptions in our approach. First, we take DOM to involve the licensing of a [PERSON]5 feature (Cornilescu 2000, Rodríguez-Mondoñedo 2007, Richards 2008, a.o.) beyond u[Case]. From this perspective, differential marking, thus, is the Spell- Out of a licensing operation between a probe on a functional head and a goal DP with a complex structure, containing not only u[Case] but also [PERSON]. Furthermore, we take clitic doubling to entail the same operation, as some types of clitics similarly entail an additional [PERSON] feature beyond the u[Case] of their double. As [PERSON] is one of the interpretable features that requires valuation within the narrow syntax (Preminger 2019, Coon & Keine 2021, Deal 2021, a.o.), all DPs marked with a [PERSON ] feature must have this feature licensed by a probe within a local domain. Second, following the arguments made in Irimia (2020, in press), we consider the sensitivity to licensing domains key to accounting for our data here. Specifically, a certain domain (e.g. the vP domain) may only bear one supplemental licenser in addition to the standard licenser of that domain, which normally licenses u[Case]. We show that the need for an element to raise into a higher licensing domain is crucial when considering repair strategies and the necessity for additional heads able to license [PERSON ] features. Finally, although we discuss instances of clitic doubling, our focus will not be on how doubling is realized within the syntax. For purposes of simplicity, we assume 5 One reviewer asks whether this should be encoded as [PERSON] or as some other feature. In our view, what is relevant for DOM in the languages discussed here is that it should be linked to a specification beyond uninterpretable Case per se (as otherwise the co-occurrence restrictions we address cannot be derived straightforwardly, see fn. 13). We have followed the ‘DOM as [PERSON]’ line of investigation mentioned in the text given that it is not problematic for the data and is independently motivated. As the reviewer themselves note, in the end, this is a mere notational issue. Whatever specification beyond u[Case] we would like to link DOM to, the facts would still be correctly explained. 6 Isogloss 2022, 8(4)/9 Brian Gravely & Monica Alexandrina Irimia a ‘Big DP’ analysis (Torrego 1992, Uriagereka 1995, a.o.).6 It should be noted, however, that approaches that deal with doubling in other ways (e.g. Preminger 2019) should not hinder the derivational claims made throughout this paper. Building off of the work in López (2012), we preliminarily adopt a modified vP shell as in (7).7 (7) vP shell (modified from López 2012:45) XP 3 P3 vP 3 EA v’ 3 vº αP 3 P2 α’ 3 α ApplP 3 IO Appl’ 3 Applº VP CLIO 3 Vº P1 As described by López, the distinct P-positions represent the licensing sites of different objects at the syntax-pragmatics interface. The base-generated position of theme arguments, P1, is where bare nouns are interpreted, leaving them in a position 6 A reviewer correctly notes that our adoption of the ‘Big DP’ hypothesis only extends to the accusative argument, whereas the dative argument follows the original layout by Cuervo (2003) in which the lexical DP is base generated in the specifier of the dative clitic found on the head of ApplP (cf. (7)). Recall that one of the original hypotheses upon which the ‘Big DP’ hypothesis was posited supported accusative Case assignment to both the clitic and the lexical DP. Following Cuervo’s original arguments, however, dative Case is assigned by Applº itself with no need for both syntactic elements to be base generated as sisters and subsumed under a lone functional projection that receives their respective Case (see also fn. 23). However, were we to assume a ‘Big DP’ hypothesis for dative arguments, nothing critical about our approach would be forced to change. 7 A reviewer questions our implementation of ApplP in the spirit of Cuervo (2003), Pylkkänen (2008), i.a., based on the observations in, e.g. Pineda (2020). Taking Pineda (2020) as a reference of argument, this author proposes a VP-internal (or √P-internal in her words) structure in which an applicative phrase (LowApplP) contains the lexical dative in its specifier, the corresponding dative clitic as its head, and the accusative argument as its complement. Aside from the position in which ApplP base generates, many of Pineda’s assumptions align with our own. For example, Pineda claims that the accusative argument must move to an intermediate position above ApplP (for us [Spec, α], for her [Spec, v]) rather than base generate above the applicative argument, a view that we support and follow in our subsequent argumentation. Much of her data deals with the identification of non- doubled lexical IOs as true dative arguments rather than prepositional phrases as they have previously been identified in the literature (e.g. Demonte 1995). However, as we neither question the status of non-clitic-doubled datives as true datives nor do we focus on the same data set as Pineda, we see no true theoretical divide that discourages us from following the approach we have laid out below. DOM co-occurrence restrictions Isogloss 2022, 8(4)/9 7 from which they might never bear DOM.8 The P2 position is where objects are said to receive differential marking, although we will show some counterevidence for this in §4 regarding Galician. Finally, the P3 position relates direct objects to their connection to information structure (e.g. their discourse anaphor behavior). As we shall show, this position is strongly linked to [PERSON ] licensing, as well. 3. Romanian DOM and its repair strategies Just like Spanish and many other Romance languages, Romanian presents a split in the morphosyntactic marking of direct objects, generally regulated by animacy and specificity (Dobrovie-Sorin 1994, Cornilescu 2000, Tigău 2011, Mardale 2015, Irimia 2020, Hill and Mardale 2021, a.o.). The animate definite in (8a) can be introduced by a preposition which is homophonous with the locative. 9,10 The inanimate in (8b), on the other hand, is not grammatical with the same preposition and must remain unmarked. (8) Romanian direct objects a. (Le)=am văzut (pe) fetele frumoase CL.ACC.F.3PL=have.PRS.1SG seen DOM girls.the beautiful.PL ‘I saw the beautiful girls.’ b. (*Le)=am văzut (*pe) casele frumoase CL.ACC.F.3PL=have. PST.1SG seen DOM houses.the beautiful.PL ‘I saw the beautiful houses.’ The sentences in (8) also show that nominals which are differentially marked can be clitic doubled using the accusative form of the clitic. For many speakers, in fact, clitic doubling is the preferred option with lexical DPs. For all speakers, clitic doubling is required with full pronouns (as well as some other categories, such as demonstratives, etc.), as illustrated in (9).11 (9) Romanian DOMed pronouns and obligatory clitic doubling *(L)=am văzut *(pe) el CL.ACC.M.3SG=have.PRS1SG see.PRT DOM he ‘I saw him.’ An important observation which will be of interest to us is that there are syntactic differences between marked objects with and without clitic doubling. As 8 In fact, whether direct objects that are left in-situ can show up with differential marking or not is a debated issue (see Irimia 2020 for details). For reasons of space, we do not address this problem here. See also Gravely (2021c) for evidence regarding bare nouns that raise and take DOM in Galician. 9 The differential marker is obligatory with various types of DPs such as proper names, animate quantifiers, etc. Moreover, in some instances it does also extend to certain types of inanimates. 10 For reasons of space, we gloss pe simply as ‘DOM’ despite its existence as a locative preposition, as well. 11 There are also certain types of DOM (e.g. that which is found on quantifiers), where accusative clitic doubling is not grammatical, as shown with the animate negative quantifier in (37e). 8 Isogloss 2022, 8(4)/9 Brian Gravely & Monica Alexandrina Irimia highlighted by Cornilescu (2020), a.o., DOMed objects which are clitic doubled allow binding into the external argument (EA). Marked objects without clitic doubling, however, do not permit this reading, as seen in the examples in (10). The interpretation of (10b), according to which the subject muzica (‘music’) co-varies with the direct object, is not allowed in (10a). This contrast indicates that at least some component in the makeup of differentially marked objects, if not the DOMed nominal itself, is found in a higher position above vP in the case of clitic-doubled direct objects. (10) Romanian DOM and accusative clitic doubling (Cornilescu 2020, 24-25) a. Muzica lor*i plictisește pe mulții music.the their annoy.PRS.3SG DOM many ‘Their music annoys many people.’ b. Muzica lori îi=plictisește pe mulții music.the their CL.ACC.M.3PL=annoy.PRS.3SG DOM many ‘Their own music annoys many people.’ Given that non-clitic doubled DOM cannot take scope over the EA, it must be the case that it is licensed in a position below [Spec, v] where the EA is introduced. We follow López (2012) and assume that DOM in Romanian is licensed by αº below vº. Following the observations in Irimia (2020), as well as Cornilescu (2000), we also assume that differentially marked objects contain a discourse-linking [PERSON] feature, beyond u[Case]. The main licenser in the verbal domain, vº, licenses the uninterpretable Case on the DO. As objects with special features involve an additional [PERSON] specification, an additional functional projection, in this case αº, is also relevant. As seen below in (11), αº licenses [PERSON] in differentially marked objects. The result is that DOM cannot raise above the vP and thus cannot take scope over the EA. As we show below, clitic doubling involves a yet distinct licensing operation on DOM: the licensing operation spelled out by the clitic double which allows raising of differential objects above vP. (11) DOM co-occurrence restrictions Isogloss 2022, 8(4)/9 9 3.1. Romanian DOM and co-occurrence effects A less discussed fact about Romanian DOM is that it gives rise to various types of co-occurrence restrictions. As Cornilescu (2020) has observed, Romanian DOM leads to ungrammaticality in contexts that contain a clitic-doubled indirect object when binding from the direct object is attempted into the clitic doubled dative, as in (12). (12) Romanian DOM and clitic doubled dative (Cornilescu 2020:119) *Comisia le=a repartizat pe mai mulţi board.the CL.DAT.3PL=have.PRS.3 SG assigned DOM more many medici rezidenţi unor foşti profesori de-ai lor medical residents some.DAT former professors of-GEN theirs Intended: ‘The board assigned several medical residents to some former professors of theirs.’ One way to lift this co-occurrence restriction is by clitic doubling the accusative argument, as shown in the example repeated in (13). It is worth noting that an explanation according to which the ungrammaticality of (12) is due to the fact that marked nominals always need clitic doubling cannot be on the right track. The sentence in (12) is equally ungrammatical for speakers who accept DOM without clitic doubling. Thus, this begs the question of how the contrast between (12) and (13) may be explained. Also note that in Romanian, similarly to what we have seen for Spanish, DOM is grammatical with an indirect object which is not clitic doubled, as seen in (14) (adapted from Cornilescu’s original examples). This indicates that the problem is the co-occurrence of DOM with a clitic-doubled dative, a picture which is similar to what O&R (2007, 2013) observed for Spanish (3). (13) Accusative clitic doubling in Romanian (Cornilescu 2020:119) Comisia i l=a repartizat pe board.the CL.DAT.3SG CL.ACC.M.3SG=have.PRS.3SG assigned DOM fiecare resident unei foste profesoare a luii each resident some.DAT former professor GEN his ‘The board assigned each medical resident to a former professor of his.’ (14) Comisia a repartizat pe fiecare rezident board.the have.PRS.3SG assigned DOM each resident unei foste profesoare a luii some.DAT former professor GEN his ‘The board assigned each medical resident to a former professor of his.’ In Romanian, indirect objects show inflectional dative morphology. 12 For the types of datives that are relevant in the configurations analyzed in this paper, the most straightforward hypothesis is that dative Case is licensed by a low Applº head. Following the remarks in Irimia (2020, in press) and in accordance with the observations Cornilescu (2020), we take clitic-doubled datives to involve the 12 There are also prepositional datives, which we leave aside here, due to the fact that in some contexts they are a hallmark of colloquial registers. 10 Isogloss 2022, 8(4)/9 Brian Gravely & Monica Alexandrina Irimia presence of an additional feature beyond Case, which we consider to be [PERSON] and which requires a separate licensing operation. Based on the observations above, we take αº to be involved in this licensing operation.13 Going beyond López (2012) and adapting recent observations by Pancheva & Zubizarreta (2018), we claim that αº is the locus of perspectivization and sentience within the vP domain. [PERSON] can only be licensed by αº, unless other features in the composition of the DP force it to raise and this feature is able to find an adequate licenser above vP. With this in mind, we can now turn to the co- occurrence restrictions caused by DOM (12). Under the current analysis, the problem with such examples is that both the DOMed accusative argument and the clitic- doubled dative contain a [PERSON ] feature which requires licensing beyond u[Case]. Given that there is only one [ PERSON] licenser within the vP domain, only one [PERSON] feature may be licensed: either that on the clitic-doubled dative or that of the DOMed DO but not both (15).14 (15) Why is clitic doubling of a DOMed argument a repair strategy? We have seen in (10) that there are crucial syntactic differences between clitic-doubled accusatives with and without clitic doubling related to the possibility of binding into the external argument. These facts appear to support the observation that in Romanian the clitic double is linked to another [PERSON] feature that we specify as [PERSON]Cl for convenience. As its licensing requires it to raise to a domain above vP, the nominal 13 O&R (2013) assume that Spanish clitic-doubled datives are licensed by αº in terms of a Case feature, while datives without clitic doubling are PPs and do not require licensing. The same hypothesis cannot be extended to Romanian, as inflected datives do not behave like unlicensed PPs structurally. See also Pineda (2020) for arguments against equating Spanish datives which are not clitic doubled with adjunct PPs that do not need licensing. We will see a similar observation in §5 with respect to non-clitic-doubled datives in Argentine Spanish as argued for by Saab (2022). An analysis under which clitic doubling on datives indicates licensing of an additonal [ PERSON] feature yields better results. 14 We leave aside u[Case] licensing on DOMed objects in order to simplify the representation. DOM co-occurrence restrictions Isogloss 2022, 8(4)/9 11 containing it will have to raise above the external argument. 15 The other [PERSON] feature, the one linked to animacy, will get licensed, together with the [ PERSON]Cl feature in domain above vP. 16 In our view, co-occurrence restrictions with DOM, which are less studied, provide further support for the existence of this [ PERSON]- licensing position above vP (see also Cornilescu 2020). Given that in accusative clitic doubling contexts the [PERSON] feature on the DOMed direct object can be licensed above vP, this leaves º available for licensing the [PERSON] feature of clitic-doubled datives. As a result, we find that grammaticality arises in contexts such as (13), as represented in (16).17 (16) XP 3 Clº [PERSON ]Cl X’ [PERSON] =DOM 3 Xº vP 3 vº αP 3 α ApplP 3 IO Appl’ [PERSON] 3 u[C:_]→[DAT] Applº VP [DAT] 3 Vº DP 3 Clº DP [PERSON]Cl u[C:_] [PERSON] The theoretical stances we have taken with respect to the Romanian data in this section open the door for further examination of licensing positions both within and beyond the vP domain. 18 In the subsequent section, we elaborate on the latter and 15 A question is whether the lexical DP raises, or whether it can remain in-situ and only the relevant category containing [PERSON]Cl needs to raise. In order to keep the Romanian data systematic with the facts from Galician and Spanish, we will assume, for convenience, that the lexical DP raises. However, nothing hinges on this. A thorough examination of the configurational possibilities and how to discriminate among them is orthogonal to the point we are interested in making here. See also the remarks at the end of §5. 16 Extensive motivation for the [PERSON]-related field above vP can be found in Săvescu Ciucivara (2009), Belletti (2005), a.o. 17 While there have been works that assume the order of clitics is derived based on which argument agrees first (cf. Laenzlinger 1993), it is most commonly assumed in present-day syntactic theory that clitic order is templatic (cf. Deal 2021), a view we adopt here without argument. 18 A reviewer asks what we define as a ‘domain’. Thus far, we have seen that there are clear cross-linguistic limitations for [PERSON]-licensing below vº with microparametric variation amongst Romance varieties regarding whether an argument or its clitic counterpart may raise and be licensed above this functional head. We are hesitant to call these phase-related effects in the sense of Chomsky 12 Isogloss 2022, 8(4)/9 Brian Gravely & Monica Alexandrina Irimia show that Galician has an even more extensive licensing strategy above the vP domain due to its position-based restriction on DOM assignment. 4. Galician DOM and its repair strategies Differential marking in Galician surfaces in a much more limited way than that which we have seen in Romanian, in addition to the discrepancy between the features that license differentially marked objects. In Gravely (2021a), it was shown that Galician objects do not typically undergo DOM, regardless of word order. (17) Galician direct objects a. Xan viu Uxía (SVO) Xan see.PST.3 SG Uxía ‘Xan saw Uxía.’ b. Podou o pexegueiro Manolo (VOS) trim.PST.3SG the peach-tree Manolo ‘Manolo trimmed the peach tree.’ c. Colleron os rapaces moitas troitas (VSO) catch.PST.3PL the boys many trouts ‘The boys caught a lot of trout.’ d. Meus paisi, recollemo=losi mañá (CLLD) my parents collect.PRS.1PL=CL. ACC.M.3PL tomorrow ‘My parents, we pick them up tomorrow.’ However, when it does surface, DOM is limited to i) a specific word order, and ii) the featural makeup of both the object and subject. Specifically, DOM only arises in VOS strings in which the object nominal has the same or higher animacy specification as the subject, as seen in (18). (18) Galician DOM Saudaron ós nenos as nenas greet.PST.3PL DOM.the boys the girls ‘The girls greeted the boys.’ The relevance of these data to our proposal lies in fact that the object must raise above the subject in order to receive DOM, unlike what we find in other Romance varieties such as Romanian and Spanish. As we show below, this has direct implications for our theoretical proposal for Galician. As in Spanish, DOM in Galician is morphologically parasitic on the dative marker and the adposition a (‘to’ or ‘at’). When the object bears a definite determiner, there is obligatory incorporation of the determiner into the a-preposition, as in (18). However, when the DP is indefinite (19a), headed by a quantifier (19b), or a personal pronoun (19c), it surfaces independently of the DP. (2000, 2008, i.a.). In particular, the Galician data and theory presented in §4 would entail not one but two domains above vP based on the work in Fernández-Rubiera (2013), Gravely (2020), a.o., which would take us too far afield. Thus, we limit our reference to ‘domains’ as either below or above vº. DOM co-occurrence restrictions Isogloss 2022, 8(4)/9 13 (19) Galician a. Coñecen a unhas persoas pícaras os teus amigos know.PRS.3PL DOM some people hateful the your friends ‘Your friends know some hateful people.’ b. Despediron a toda=las enfermeiras os médicos release.PST.3 PL DOM all=the nurses the doctors ‘The doctors sent home all of the nurses.’ c. Viches a ela cando pasou? see.PST.2 SG DOM she when pass.PST.3SG ‘Did you see her when she came by?’ The position of the DOMed DP is telling both for full DPs, as well as for pronouns. Unlike what has been claimed for Spanish (López 2012) or Portuguese (Costa 2000), Gravely (2021c) shows that this is an A-position in Galician from which the marked object may bind into the EA.19 (20) Galician: DOM binding into the EA Axudou a cadai neno seui pai help.PST.3SG DOM each kid his father ‘His parent helped each child.’ Following Gravely (2021a), we assume that because the object only receives differential marking above the subject, there is no vP-internal licensing position, contrary to López’s (2012) proposal and differently than what we have seen in Romanian (§3). The position that we refer to as P3, thus, will be crucial for our accounting for additional [PERSON] licensing positions further up the clausal spine. 4.1. Galician DOM and co-occurrence effects Unlike what has been shown for Romanian, Galician does not have clitic doubling with full DPs.20 Therefore, in order to discuss similar repair strategies observed in §3 19 In the contrast in (10), we saw that Romanian DOM without clitic doubling cannot bind into the EA. López (2012) presents extensive evidence supporting the same conclusion regarding DOM in Spanish, although see arguments in Gallego (2013) against these views. As illustrated more extensively in §5, Spanish does not generally allow clitic doubling of full DPs bearing DOM, thus binding with clitic doubled DOM cannot be tested. 20 A reviewer asks what differentiates languages that only double pronouns from those that double lexical DPs, as well. We are not aware of a consensus in the literature on the structural differences, although it seems that this would largely depend on the theoretical assumptions made regarding doubling more generally. An approach such as that put forth by Preminger (2019), which considers clitic doubling the overt Spell-Out of the same φ-set in two different places along the clausal spine, defends that doubling is not optional but depends on whether an argument has raised to a position from which it may be probed and, subsequently, doubled. There are several problematic aspects with this type of approach in Romance, particularly when the consensus in the literature favors consistent object movement in Romance even for languages that lack clitic doubling of full DPs. Instead, it may be a microparametric difference amongst languages of the same typology, as initial hypotheses regarding these patterns seem to vary. Whereas it is likely the case that for Galician doubling may occur with any Xº-element (i.e., pronouns, lone demonstratives and quantifiers, etc.; Gravely 2021a,c), this seems unlikely to account for the vast variation found in Romanian. We leave this question for further research. 14 Isogloss 2022, 8(4)/9 Brian Gravely & Monica Alexandrina Irimia for Romanian, the construction we focus on here is one with pronominal direct objects, which must be differentially marked (21a) and may undergo clitic doubling (21b).21 (21) Galician object pronouns: DOM and clitic doubling a. Mandamos *(a) el ó médico send.PRS.1PL DOM he DAT.the doctor b. Mandámo=loi a eli ó médico send.PST.1PL=CL.ACC.M.3SG DOM he DAT.the doctor ‘We sent him to the doctor.’ Much like the observations in O&R (2013) for Spanish, as well as those for Romanian in §3, there are DOM+clitic doubling restrictions. First, although clitic doubling of the direct object pronoun is licit in (21b), doubling of the indirect object renders the sentence ungrammatical when DOM is present and is not clitic doubled itself, as in (22). This is precisely what we see in Spanish (3) and Romanian (12). (22) Galician: DOM and clitic-doubled datives *Mandámos=lle a el ó medico send.PST.1PL=CL. DAT.3SG DOM he DAT.the doctor Intended: ‘We sent him to the doctor.’ Much like in the case of Romanian, we claim this is due to the lack of licensers for the two vP-internal arguments. Importantly, we must recall that the P2 position is not an A-position in Galician, unlike what we showed for Romanian. Therefore, the model adopted from (7) for Galician is as in (23).22 21 The obligatory marking of personal pronouns does not dissuade us from claiming that P2 is not a position that licenses differential marking in Galician. An approach that distinguished the licensing position of full DPs from that of personal pronouns would be forced to explain why the syntax obligatorily marks these pronominal objects in a distinct position from that of their DP counterparts, and we are currently aware of no such reasoning. 22 A reviewer questions the need for DOs in Galician to pass through the P2 position and also questions whether we only claim this out of uniformity with Romanian. Object shift, albeit it in various flavors with respect to the landing site, is widely assumed throughout all Romance varieties that license VSO order (cf. Gallego 2013). Our proposal of a vP-internal object shift captures aspects of agreement that Gallego (2013) cannot. We follow observations in Walkow (2012, 2013) and claim that there is a ‘DO preference’ in the ordering of agreement operations of the DO and IO. This inherently requires the accusative argument to be structurally higher than, and thus closer to the probe borne by its respective functional head, the dative one. For additional agreement effects that this ordering of a derivation’s internal arguments may have, see the observations in Deal (2021). DOM co-occurrence restrictions Isogloss 2022, 8(4)/9 15 (23) [PERSON] licensing positions in Galician (version 1) XP 3 P3u[PERSON] vP 3 EA v’ 3 vº αP 3 P2 α’ 3 α ApplP 3 IO[PERSON] Appl’ 3 Applº VP CLDAT 3 Vº DO[PERSON] Although the direct object does not receive differential marking in P2, we claim that it must move to P2 for reasons related to the order in which the direct object and indirect object are probed (see Pineda 2020 and Cornilescu 2020 for evidence in favor of this in Romance, as well as Deal 2021 for a view on this cross- linguistically). As we assumed for Romanian, the [PERSON] feature of the dative clitic is licensed in its base-generated position by Applº.23 When the direct object moves to P2, it is structurally higher than the indirect object in the specifier of ApplP. When the probe in the P3 position searches for a goal bearing [PERSON], it will find and bring the direct object above the subject where it receives differential marking. This leaves the indirect object without a probe to license its [PERSON] feature within the local vP domain. 4.2. Licensing beyond the vP shell In order to account for the data in (5), repeated below in (24a) for convenience, we propose that Galician has additional [ PERSON ] licensing positions, although they differ slightly from what we showed for Romanian in §3. We claim that the YP position in (24b) hosts a probe that may check this feature. 23 A reviewer suggests that our claim regarding the dative clitic being licensed in-situ is difficult to substantiate. We reference Cuervo’s (2003:34) original proposal regarding (low) applicative datives in Spanish and argue the same for Romanian and Galician: “[…] the head assigns inherent dative case to the argument it licenses, and the dative clitic is the Spell-Out of the applicative head” (emphasis ours). This position is in line with much of the work done on applicatives more generally (especially in Romance), thus we follow these claims without argument. 16 Isogloss 2022, 8(4)/9 Brian Gravely & Monica Alexandrina Irimia (24) a. Mandáron=ll=oi,k a elk send.PST.3PL=CL.DAT.3SG= CL.ACC.M.3SG DOM he ói médicoi DAT.the doctor ‘They sent him to the doctor.’ b. [PERSON] licensing positions in Galician (version 2) YP 3 P4u[PERSON] XP 3 P3u[PERSON] vP 3 EA v’ 3 vº αP 3 P2 ApplP 3 IO Appl’ 3 Applº VP 3 Vº P1 In Gravely (2021a,c), it was claimed that this position is where the subject is probed in VSO constructions, strengthening our reasoning for a [ PERSON] licensing head. This site, which we consider to be in a domain above that in which the direct object and indirect object DPs are licensed, is able to license the IO. Similarly to our claim for Romanian, albeit in a higher point along the clausal spine, the doubled accusative clitic also requires licensing. Due to the position of the direct object (e.g. above the external argument), we claim that from this position the clitic may be probed directly by the [PERSON ] feature on f where clitics in Galician cluster (Raposo & Uriagereka 2005, Gupton 2014, Gravely 2021b). DOM co-occurrence restrictions Isogloss 2022, 8(4)/9 17 (25) FP 3 f u[PERSON] TP CL[PERSON] 3 Tº YP 3 IOk[PERSON] Y’ 3 Yºu[PERSON] XP wo DPi X’ 5 3 DO[PERSON] Xºu[PERSON] vP 3 EA v’ 3 vº αP 5 ti t k In (25), we show that the clitic leaves the DP in P2 in order to be probed by f. However, it may also be that it departs from the DP that hosts it once the latter has undergone further movement to a position above Tº. Let us consider the position of the direct object in both a doubled scenario as well as without clitic doubling. (26) Galician a. Mandaron xa a el ó médico send.PST.3PL already DOM he DAT.the doctor b. Mandáron=ll=o a el send.PST.3PL=CL. DAT.3SG=CL.ACC. M.3SG DOM he xa ó médico already DAT.the doctor c. ???Mandáron=ll=o xa a el send.PST.3PL=CL. DAT.3SG=CL.ACC. M.3SG already DOM he ó médico DAT.the doctor Intended: ‘They send him to the doctor already.’ In (26a), the DOMed pronoun is structurally lower than the Tº-oriented adverb xa (‘already’). When both DOM and the IO are doubled, however, raising of the direct object greatly ameliorates the grammaticality of this construction, as in (26b). Based on the syntactic limitations of the clitic doubling of direct objects in Galician, it is difficult to derive accurate testing that would make a definite prediction regarding from which position (above or below Tº) the clitic is probed from f. However, as the accusative clitic is the last [PERSON]-bearing argument to be 18 Isogloss 2022, 8(4)/9 Brian Gravely & Monica Alexandrina Irimia licensed, all that matters is that it is not the same position as that in which clitic doubling is licensed. 24 5. Spanish and its dialectal variation As we discussed briefly in §2, the Peninsular Spanish data in O&R (2013) shows some differences from what we have seen for Romanian and Galician. Based on the theory put forth in §3-4, a quick conclusion is that Peninsular Spanish lacks additional [PERSON] licensing positions outside of the vP, which explains the ungrammaticality of DOM with dative clitic doubling as in (3) and its impossibility to be repaired by accusative clitic doubling. We provide another example here. (27) Spanish: DOM and clitic-doubled datives a. Enviaron *(a) Mateo a los doctores send.PST.3PL DOM Mateo to.DAT the doctors b. *Les=enviaron a Mateo a los doctores CL.DAT.3PL=send.PST.3PL DOM Mateo to.DAT the doctors Intended: ‘They sent Mateo to the doctors.’ Although Peninsular Spanish does not have clitic doubling of full DPs, it does boast doubling of DOMed pronominal arguments as in Galician (21). (28) Peninsular Spanish: accusative clitic doubling of DOMed pronouns Lo=envié a él al médico CL.ACC.M.3SG=send.PST.1SG DOM he DAT.the doctor ‘I sent him to the doctor.’ However, as opposed to Romanian and Galician, pronominal DOM arguments which are clitic doubled still give rise to ungrammaticality when a clitic doubled dative is added. This is seen in (29). (29) Peninsular Spanish: co-occurrence restrictions with DOM a. *Le=envié a él al médico CL.DAT.3SG=send. PST.1SG DOM he DAT.the doctor b. *Se=lo=envié a él CL.DAT.3SG=CL.ACC. M.3SG=send.PST.1 PL DOM he al médico DAT.the doctor Intended: ‘I sent him to the doctor.’ 24 Another potential avenue that may be yield positive results is the fact that OVS order in Galician is exceptionally productive compared to, e.g. Spanish, even with [PERSON]-bearing DPs. In non-clitic-doubling scenarios, it has been shown that the verb remains on Tº (Gupton 2014). Therefore, we may assume that the accusative argument is in a position higher than Tº in both OVS constructions as well as in (26b). For reasons of space, we do not explore this possibility further but wish to note that there may be a further connection regarding non-subject-DP licensing positions above Tº. DOM co-occurrence restrictions Isogloss 2022, 8(4)/9 19 We claim that these data may be explained under the same licensing terms we have argued for in the case of Romanian and Galician. We will do so by also comparing the data with doubling patterns in Argentine Spanish, which differs from those of Peninsular Spanish. 5.1 The case of Argentine Spanish: only one doubled argument Argentine Spanish, much like Romanian, permits clitic doubling of full DP direct objects that bear an [ANIMATE] feature, as seen in (30). Saab (2022), referencing the observations made by O&R (2007, 2013), claims that the dative clitic doubling ruled out in the presence of a DOMed object in Peninsular Spanish is indeed grammatical in Argentine Spanish. (30) Argentine Spanish: DOM and clitic doubling Lai=vimos a la niñai CL.ACC.F.3SG=see.PST.1 PL DOM the girl ‘We saw the girl.’ (31) Argentine Spanish: DOM and clitic doubled IO (Saab 2022:34) Ayer le=presenté (a) María a Juan yesterday CL.DAT.3SG=introduce.PST.1SG DOM María DAT Juan ‘Yesterday I introduced María to Juan.’ Saab explains that Argentine Spanish has undergone a process of grammaticalization, resulting in obligatory clitic doubling of indirect objects. Following the observations in Saab (2022), which we adapt to our system, dative clitic doubling equates with an additional licensing position in P3. What distinguishes Argentine Spanish from Romanian and Galician is its ability to double the dative argument but not both the accusative and the dative. (32) Argentine Spanish (Saab 2022:34) *Ayer se=la=presenté a María yesterday CL.DAT.3SG=CL.ACC.F.3SG=introduce.PST.1SG DOM María a Juan DAT Juan Intended: ‘Yesterday I introduced María to Juan.’ Saab claims that the ban on simultaneous clitic doubling of both the direct object and indirect object is predicated on two factors. 25 First, doubled objects, be those direct or indirect, entail A-movement above vP. Specifically, he claims that the clitic is a probe for its doubled DP to undergo A-movement itself. Second, he references an ‘only one clitic probe per sentence’ restraint, highlighting that A- movement of the direct or indirect object bleeds clitic doubling to a low CliticP projection. However, the doubling mechanism may only apply to one of the arguments. For Saab, this clitic position immediately c-commands VoiceP, which we 25 Saab’s work is built on doubling and its interaction at the syntax-semantics interface. For reasons of space, we cannot review the full account here. However, the syntactic similarities and differences relevant to tying in the data from Argentine Spanish will be addressed. 20 Isogloss 2022, 8(4)/9 Brian Gravely & Monica Alexandrina Irimia interpret as our proposed XP that c-commands the external argument. Our adoption of Saab’s model (33a) is shown in (33b). (33) a. CliticP 3 Cliticº VoiceP CL 3 a-DP Voice’ 3 EA Voice’ 3 Voiceº vP 5 b. YP 3 Yº XP CL 3 a-DP X’ 3 Xº vP 3 EA v’ 3 vº αP 5 Finally, Saab claims that because A-movement of the direct object is in complementary distribution with that of the indirect object, only one of them may raise via this mechanism.26 Due to the fact that the indirect object asymmetrically c- commands the direct object, his prediction is that the indirect object will always be doubled in ditransitive constructions. 26 A slight difference in Saab’s proposal that we do not adopt here is movement of the raised DP to [Spec,Clitic], the position where this author argues Case is checked. Movement to P3 (i.e., XP) is what he refers to as ‘blind movement’, following Rodríguez-Mondoñedo (2007). As we have shown ample argumentation for the P2 and P3 position being an A-position in Romanian and Galician respectively, we maintain our stance and leave aside this theoretical divergence. DOM co-occurrence restrictions Isogloss 2022, 8(4)/9 21 5.2 What Argentine Spanish tell us about Peninsular Spanish Building off of these observations, we may now make several conclusions that fall in line with what we have argued for throughout this paper. First, there seems to be competition for the [PERSON] licensing positions in both dialects of Spanish, just as we have shown for direct objects and doubled indirect objects in Romanian and Galician. The initial assumption that Peninsular Spanish has fewer [PERSON] licensing positions than Romanian and Galician seems correct. Upon comparing this variety to that of Argentine Spanish, the obvious conclusion is that Peninsular Spanish boasts only one vP-internal [PERSON ] licensing position, that which is found in P2. For instances of clitic doubling of personal pronouns as shown in (28), we claim that the direct object may Ā-move to P3 where its clitic double may be licensed. 27 (34) XP 3 CL+a-DPi vP 3 EA v’ 3 vº αP 3 ti … What prevents the indirect object clitic from reaching the same site and be licensing by the probe in P3? Building off of the observations in Saab, we claim that the indirect object does not undergo A-movement in Peninsular Spanish, contrary to what has been shown for Argentine Spanish. 28 Due to the obligatory nature of DOM with the type of direct objects we have addressed, as well as the ability of personal pronouns to undergo doubling while being differentially marked, Peninsular Spanish differs from Argentine Spanish in that the former does not have an obligatory licenser in the P3 position whereas the latter does. 29 27 Because clitic doubling of either argument is not obligatory in Peninsular Spanish as it is in Argentine Spanish, there is no reason to tie in A-movement with clitic doubling. 28 Recall that, for Saab, only DP objects that have undergone A-movement may be clitic doubled. Therefore, due to the obligatory nature of indirect object clitic doubling, the DP that A- moves must be that of the indirect object. 29 P3 is left in parenthesis in (35a) as a means of showing optionality of additional licensing, reserved solely for the pronominal clitic doubling of direct objects. 22 Isogloss 2022, 8(4)/9 Brian Gravely & Monica Alexandrina Irimia (35) a. Peninsular Spanish vP licensing positions XP 3 (P3)u[PERSON] vP 3 EA v’ 3 vº αP 3 P2u[PERSON] α’ 3 α ApplP 5 b. Argentine Spanish vP licensing positions XP 3 P3u[PERSON] vP 3 EA v’ 3 vº αP 3 P2u[PERSON] α’ 3 α ApplP 5 What is proposed for Peninsular Spanish in (35a) is that a clitic-doubled dative and a differentially-marked object do indeed compete for the same licenser in the P2 position and that there is no further position available beyond vP for full DP objects. In contrast, based on the theoretical claims for Argentine Spanish in Saab (2022), the object that raises to P2 and has its [ PERSON] feature licensed (be this the direct or indirect object) will proceed to P3 where its clitic double is licensed (35b). The licensing in P2, thus, bleeds the licensing in P3 by the same argument. What should be noted, however, is that although clitic doubling of a full direct object DP is grammatical in Argentine Spanish (36a), doubling of the indirect object takes precedence over doubling of the direct object (36c). (36) Argentine Spanish a. Lai=vimos a la niñai CL.ACC.F.3SG=see.PST.1 PL DOM the girl ‘We saw the girl.’ DOM co-occurrence restrictions Isogloss 2022, 8(4)/9 23 b. *Lai=mandamos a la niñai a Juan CL.ACC.F.3SG=send.PST.1PL DOM the girl DAT Juan c. Lei=mandamos la niña a Juani CL.ACC.F.3SG=send.PST.1PL the girl DAT Juan Intended: ‘We sent the girl to Juan.’ Thus, as predicted from the model in (35), both the direct object and its clitic are able to license their [PERSON] features. However, an additional argument bearing [PERSON] will leave one of the three items without licensing, as in (36b). Therefore, when the indirect object is doubled, the lack of available licensers forces the direct object to appear without differential marking in (36c). 5.3 More on clitic doubling as a repair strategy We have seen that Saab (2022) has proposed a ‘one clitic per sentence’ constraint to explain the fact that clitic doubling of both the direct and indirect object still results in ungrammaticality in Argentinian Spanish. Romanian and Galician, on the other hand, are able to avoid this constraint and, by avoiding this constraint, co-occurrence restrictions imposed by DOM are lifted in the two languages. We have shown that this is a result of the fact that both Galician and Romanian have more than one [PERSON]-licensing domain active. However, additional data from Romanian co-occurrence restrictions involving DOM indicate that there are also contexts in which clitic doubling of DOMed accusative arguments is not a repair strategy, while also not being a result of the ‘one clitic per sentence’ constraint. Let us examine the examples in (37). (37) Romanian DOM and the dative clitic ‘își’ a. *Şi/*mi=l=ajută pe prieten DAT.REFL.3SG/CL. DAT.1SG=CL.ACC. M.3SG=helps DOM friend Lit. ‘She helps the friend to herself/to me.’ Intended: ‘She helps her own/my friend.’ b. *Ȋși/*îmi ajută pe prieten DAT.REFL.3SG/CL.DAT.1SG helps DOM friend Lit. ‘S/he helps the friend to himself/to me.’ Intended: ‘She helps her own/my friend.’ c. Ȋși/îmi ajută prietenul DAT.REFL.3SG/CL.DAT.1SG help.PRS.3SG friend.the Lit. ‘She helps the friend to herself/to me.’ ‘She helps her own/my friend.’ d. Ȋl ajută pe prietenul meu/său CL.ACC.M.3SG helps DOM friend.the my.M.SG/his. M.SG ‘She helps her (own)/my friend.’ e. Nu își/îmi trimite *(pe) nimeni în ajutor NEG DAT.REFL.3SG/CL.DAT.1SG sends DOM nobody in help 1. ‘She doesn’t send anybody to herself/to me as an aid.’ 2. ‘She doesn’t send anybody as an aid, and she is negatively affected by this/I am negatively affected by this.’ 24 Isogloss 2022, 8(4)/9 Brian Gravely & Monica Alexandrina Irimia In (37), we see a dative clitic interpreted as a possessor. This type of dative possessor clitic gives rise to ungrammaticality if a differentially marked argument is present in the sentence, even if the latter is clitic doubled.30 To ensure grammaticality, either the differential marker or the dative clitic interpreted as a possessor must be removed, as shown in (37c) and (37d). In (37d), the possession interpretation is obtained by using the adjectival realization of the possessive’s full pronominal form. In (37b), we see that this dative possessor triggers ungrammaticality with DOM which is not clitic doubled, as expected. 31 However, if the dative clitic is not interpreted as a possessor, the co-occurrence restriction with DOM does not arise (37c). The animate negative quantifier nimeni (‘nobody’), which requires DOM, does not allow a possessor interpretation, at least not one which is realized with the dative clitic. 32 The interpretation of the dative clitic in (37e) can be that of a goal (i.e., not sending anybody to himself/to me) or that of an affected applicative (i.e., not sending anybody as an aid and the intended goal being negatively affected). As expected, these latter two interpretations do not give rise to a co-occurence effect with DOM. These data indicate that: i) accusative clitic doubling is not a repair strategy across-the-board, as (37a) is ungrammatical despite the clitic doubling of the DOMed argument; ii) the problem is not the ‘one clitic per sentence’ constraint either, as (37b) is still ungrammatical even if accusative clitic doubling is missing. Then, what would the best explanation be for these facts? We believe that there are two possibilities. On one hand, it could be that the dative possessor clitic and the DOMed direct object compete for the same position within the extended projection of the DP, as depicted in (38). (38) DOM and dative possessor inside KP a. KP = [PERSON]P 3 PERSON DP ↓ DOM b. KP = [PERSON]P 3 PERSON DP ↓ Possessor On the other hand, it could be the case that both the feature relevant for DOM and the feature relevant for the Possessor are merged in the KP, but they cannot both 30 Co-occurrence restrictions involving DOM and the dative possessor clitic have not been extensively investigated, especially in the formal literature. For some recent discussion see Onea and Hole (2017), Onea (2018), or Irimia (in press). 31 Alternations in the realization of this dative clitic (interpreted as possessor) are purely phonetic. The shorter form in (37a) is due to the presence of the accusative clitic. 32 The negative quantifier cannot be clitic doubled using an accusative form of the clitic. DOM co-occurrence restrictions Isogloss 2022, 8(4)/9 25 be licensed. If the KP does not raise above vP, the two [PERSON] features are both found by a single licenser, namely αº, which can license only one of them (39). (39) As one [PERSON ] feature remains unlicensed, ungrammaticality ensues. The interesting fact is that the same problem stays if the KP hosting the two [PERSON] features raises above vP. As opposed to what we have seen with accusative clitic doubling in (13) and (16), the [PERSON] feature in the possessor clitic and the [PERSON] feature responsible for DOM cannot ‘coalesce’ and be licensed together, as shown in (40). (40) XP wi K=[PERSON]P X’ 3 3 [PERSONDOM] DP Xºφ:(PERSON) vP [PERSONPOSS] 3 vº αP 3 α VP 3 V KP = [PERSON]P 3 [PERSONDOM] DP [PERSONPOSS] We hypothesize that this may be due to the fact that the possessor clitic also needs to have its u[Case] checked. If the nominal stays within the vP, the dative will be licensed by Applº above VP. If the nominal raises, the dative Case will probably be checked by an applicative head above vP. Given its clitic status, the dative possessor cannot have its u[Case] licensed below vP and its [PERSON] feature above vP. At the same time, the [ PERSON]Poss feature cannot be separated from the other [PERSON] feature, the one responsible for DOM.33 Although interactions between 33 Onea and Hole (2017) provide an alternative analysis for the ungrammaticality between DOM and the possessor dative, building on the need of both these categories to be licensed above VP. We have not followed this account here as it cannot derive the observation that accusative clitic 26 Isogloss 2022, 8(4)/9 Brian Gravely & Monica Alexandrina Irimia DOM and the dative possessor clitic require more attention, they preliminarily support the hypothesis that two [PERSON ] features cannot be licensed when they are too local within the same domain that contains only one relevant licenser. 6. Conclusion The analysis presented in this paper centers around the idea that differential marking patterns and categories containing [ PERSON] are bound by the presence and availability of a set of licensers within a given domain. We have shown that both Romanian and Galician permit a repair strategy in which clitic doubling of both internal arguments lifts ungrammaticality of one clitic-doubled argument and a lexical DP. Although these patterns are similar on the surface, the limitations on DOM in Galician with full DPs force us to adopt distinct licensing positions in each respective language, where Romanian is able to license a vP-internal [PERSON]- bearing argument but Galician is not. Moreover, in a comparative light, we have shown that multiple varieties of Spanish, even Argentine Spanish which permits clitic doubling of full DPs as in Romanian, are unable to repair ungrammatical strings by adding structure/additional licensing requirements. Based on the observations in Saab (2022) and Ormazabal & Romero (2013), we have shown that both Argentine Spanish and Peninsular Spanish have licensing limitations both within the vP and in its periphery. In Peninsular Spanish, this amounts to the inability to license DOM in the presence of a clitic-doubled dative, and for Argentine Spanish it entails the restrictions on which argument may be clitic doubled (i.e., either the accusative or dative, but not both). Throughout this paper, we have shown that an approach along these lines is able to account for the lack of multiple licensing domains as in Peninsular Spanish while explaining how languages like Galician and Romanian are able to resort to repair strategies that entail the employment of additional licensers in a domain further up in the phrase marker. Finally, we have extended our investigation to a phenomenon in Romanian that concerns a possessor dative clitic that systematically outputs ungrammaticality in a configuration that contains DOM. We have shown that the absence of expected repair strategies is a result of the nature of this clitic itself, in that its [ PERSON ] feature cannot be licensed separately from other person features within the same KP. We believe that this variation is instructive of the types of DOM, clitic doubling, and [PERSON]-licensing strategies available across Romance. Our work is just the beginning of a comparative exploration which hopefully will receive further attention. Acknowledgments We are grateful for the valuable feedback provided by three anonymous reviewers, as well as by the series editor Jonathan MacDonald, whose observations have improved this paper. We would also like to thank the audience at LSRL 51 for extensive discussion on some of the material presented here. doubling is not a suitable repair strategy in this context. Also, Onea and Hole (2017) do not address examples such as (37e) where there are no co-occurrence restrictions. DOM co-occurrence restrictions Isogloss 2022, 8(4)/9 27 References Aissen, Judith. 2003. Differential object marking: Iconicity vs. economy. Natural Language & Linguistic Theory 21(3): 435-483. https://doi.org/10.1023/A:1024109008573 Belletti, Adriana. 2005. Extended doubling and the VP periphery. Probus 17(1): 1- 35. https://doi.org/10.1515/prbs.2005.17.1.1 Bossong, Georg. 1991. Differential object marking in Romance and beyond. In Douglas A Kibbee & Dieter Wanner (eds.), New analyses in Romance linguistics, 143-170. Amsterdam: John Benjamins. https://doi.org/10.1075/cilt.69.14bos Bossong, Georg. 1998. Le marquage différentiel dans les langues de l’Europe. In Jacques Feuillet (ed.), Actance et valence dans les langues d’Europe, 193–259. Berlin/New York: Mouton de Gruyter. Comrie, Bernard. 1989. Language universals and linguistic typology. Chicago, IL: University of Chicago Press. Coon, Jessica & Stefan Keine. 2021. Feature gluttony. Linguistic Inquiry 52(4) 655- 710. https://doi.org/10.1162/ling_a_00386 Cornilescu, Alexandra. 2000. On the interpretation of the prepositional accusative in Romanian. Bucharest Working Papers in Linguistics 2:91-106. Cornilescu, Alexandra. 2020. Ditransitive constructions with differentially marked direct objects in Romanian. In A. Pineda & J. Mateu (eds.), Dative constructions in Romance and beyond, 117-142. Berlin: Language Science Press. Costa, João. 2000. Word order and discourse-configurationality in European Portuguese. In João Costa (ed.), Portuguese syntax: New comparative studies, 94- 115. Oxford (New York): Oxford University Press. Cuervo, María Cristina. 2003. Datives at large. PhD thesis, Massachusetts Institute of Technology. De Swart, Petrus Jacobus Franciscus. 2007. Cross-linguistic variation in object marking. Ph.D. thesis, LOT. Deal, Amy Rose. 2021. Interaction, Satisfaction, and the PCC. Linguistic Inquiry (Early Access Corrected Proof), 1-56. https://doi.org/10.1162/ling_a_00455 Demonte, Violeta. 1995. Dative alternation in Spanish. Probus 7, 5-30. https://doi.org/10.1515/prbs.1995.7.1.5 28 Isogloss 2022, 8(4)/9 Brian Gravely & Monica Alexandrina Irimia Dobrovie-Sorin, Carmen. 1994. The syntax of Romanian. Berlin: De Gruyter. Fábregas, Antonio. 2013. Differential object marking in Spanish: State of the art. Borealis: An International Journal of Hispanic Linguistics (2)2: 1-80. http://dx.doi.org/10.7557/1.2.2.2603 Gallego, Ángel. 2013. Object shift in Romance. Natural Language and Linguistic Theory 31: 409-451. https://doi.org/10.1007/s11049-013-9188-6 Gravely, Brian. 2020. Only syntactic movement will do: Morphosyntactic variation in Galician and Icelandic. Talk presented at Cambridge Comparative Syntax 9 (CamCos9). University of Newcastle (Newcastle, UK). Gravely, Brian. 2021a. DOM and non-canonical word order in Romance: The case of Galician. Bucharest Working Papers in Linguistics 22(2): 5-31. http://doi.org/10.31178/BWPL.22.2.1 Gravely, Brian. 2021b. Language acquisition and endogenous grammar change: The rise of Galician complementizer agreement. Ph.D. thesis, University of Georgia. Gravely, Brian. 2021c. ‘X’ marks the spot but not why you think: a reply to Gallego (2020). Ms., Emory University. Gupton, Timothy. 2014. The syntax-information structure interface: Subjects and clausal word order in Galician. Berlin/Boston: de Gruyter/Mouton. https://doi.org/10.1515/9781614512059 Hill, Virginia & Alexandru Mardale 2021. The Diachrony of Differential Object Marking in Romanian. Oxford: Oxford University Press. https://doi.org/10.1093/oso/9780192898791.001.0001 Irimia, Monica Alexandrina. 2020. Types of structural objects: Some remarks on differential object marking in Romanian. In A. Bárány & L. Kalin (eds.), Case, Agreement, and Their Interactions: New Perspectives on Differential Argument Marking, 77-126. Boston/Berlin: de Gruyter. https://doi.org/10.1515/9783110666137-003 Irimia, Monica Alexandrina. in press. DOM and the PCC: How many types? DOM and co-occurrence restrictions. How many types? In C. Cinzia, C. E. Bullock & A. J. Toribio (eds.), A half century of Romance Linguistics. Selected Proceedings of the 50th Linguistic Symposium on Romance Languages, 157-182. Berlin: Language Science Press. Laenzlinger, Christopher. 1993. A syntactic view of Romance pronominal sequences. Probus 5, 241-270. https://doi.org/10.1515/prbs.1993.5.3.241 DOM co-occurrence restrictions Isogloss 2022, 8(4)/9 29 López, Luis. 2012. Indefinite objects: Scrambling, choice functions, and differential marking. Cambridge, MA: MIT Press. https://doi.org/10.7551/mitpress/9165.001.0001 Mardale, Alexandru. 2015. Differential object marking in the first original Romanian texts. In V. Hill (ed.), Formal approaches to DPs in Old Romanian, 200-245. Leiden: Brill. https://doi.org/10.1163/9789004292550_007 Onea, Edgar. 2018. DOM and dative arguments in Romanian. Talk presented at the Datives in Discourse Workshop, University of Cologne (Cologne, Germany). Onea, Edgar & Daniel Hole. 2017. Differential object marking of human definite direct objects in Romanian. Revue Roumaine de Linguistique 62(4): 359-376. http://doi.org/10.13140/RG.2.2.34224.25608 Ormazabal, Javier & Juan Romero. 2007. The Object Agreement Constraint. Natural Language & Linguistic Theory 25(2): 315-347. https://doi.org/10.1007/s11049-006- 9010-9 Ormazabal, Javier & Juan Romero. 2013. Differential object marking, case and agreement. Borealis: An International Journal of Hispanic Linguistics 2(2), 221-239. https://doi.org/10.7557/1.2.2.2808. Pancheva, Roumyana & María Luisa Zubizarreta. 2018. The Person Case Constraint. The syntactic encoding of perspective. Natural Language and Linguistic Theory 36(4). 1291-1337. https://doi.org/10.1007/s11049-017-9395-7 Pineda, Anna. 2020. Double-Object Constructions in Romance: The Common Denominator. Syntax 23(3): 203-240. http://doi.org/10.1111/synt.12193 Preminger, Omer. 2019. What the PCC tells us about “abstract” agreement, head movement, and locality. Glossa: a journal of general linguistics 4(1), 13. http://doi.org/10.5334/gJgl.315 Pylkkänen, Liina. 2008. Introducting arguments. Cambridge, M.A.: MIT Press. https://doi.org/10.7551/mitpress/9780262162548.001.0001 Raposo, Eduardo & Juan Uriagereka. 2005. Clitic placement in Western Iberian: a Minimalist view. In G. Cinque & R. Kayne (eds.), The Oxford handbook of comparative syntax, 639-697. Oxford: Oxford University Press. Richards, Marc. 2008. Defective agree, case alternations, and the prominence of person. In M. Richards & A.L. Malchukov (eds.), Scales (Linguistische Arbelts Berichte 86), 137-161. Leipzig: Universität Leipzig. Rodríguez-Modoñedo, Miguel. 2007. The syntax of objects: Agree and differential object marking. Ph.D. thesis, University of Connecticut. 30 Isogloss 2022, 8(4)/9 Brian Gravely & Monica Alexandrina Irimia Saab, Andrés. 2022. From pronouns to probes. A theory for a subset of Spanish clitics. Ms., CONICET-UBA. Săvescu Ciucivara, Oana. 2009. Disentangling Romance object clitics. The view from Romanian: Ph.D. thesis, New York University. Sheehan, Michelle. 2020. The Romance Person Case Constraint is not about clitic clusters. A. Pineda & J. Mateu (eds.), Dative Structures in Romance and Beyond, 143-171. Berlin: Language Science Press. http://doi.org/10.5281/zenodo.3776543 Tigău, Alina M. 2011. Syntax and interpretation of the direct object in Romance and Germanic languages. Bucureşti: Editura Universităţii din Bucureşti. Torrego, Esther. 1992. Case and agreement structure. Ms., University of Massachusetts, Boston. Torrego, Esther. 1998. The dependencies of objects. Cambridge, MA: MIT Press. Uriagereka, Juan. 1995. Aspects of the syntax clitic placement in Western Romance. Linguistic Inquiry 26(1): 79-123. Walkow, Martin. 2012. Goals, big and small. Doctoral dissertation, University of Massachusetts Amherst. Walkow, Martin. 2013. Locating variation in person restrictions. In K. Shwayder (ed.), Proceedings of the 36th Annual Penn Linguistics Colloquium, 247-256. University of Pennsylvania Working Papers in Linguistics 19(1). https://repository.upenn.edu/pwpl/vol19/iss1