External and internal possessors differ from each other in several properties. In contrast to internal possessors, external possessors do not form a constituent with the possessed noun and can participate in clause-level processes such as verb agreement and switch-reference. In this squib, we discuss “intermediate” possessors with both internal and external properties. In Tundra Nenets (Uralic), such possessors form a syntactic constituent with the possessed noun but show different types of clause-level behavior. They can bind and control out of their host DP and participate in an obviation system, a consequence of the possessor being adjoined to the host DP.
1 External, Internal, and Intermediate Possessors
Possessors that do not form a constituent with the noun they are in a possessive relation with (external possessors) are found in many languages and have been widely discussed in the literature (see, e.g., Lee-Schoenfeld 2016 on German, Landau 1999 on Hebrew, and Deal 2013 on Nez Perce). When external possessors have a different case from possessors that are internal to the noun phrase, this case is assigned by a head in the functional structure of the clause, which in some instances also assigns a semantic role such as BENEFACTIVE or MALEFACTIVE to the possessor, marking it as an affected participant. Not surprisingly, external possessors can participate in clause-level syntactic processes. For instance, they can control agreement with the predicate (e.g., Deal 2013) or participate in switch-reference (SR; see, e.g., Broadwell 2006, Munro 2016 on external possessors and SR in the Muskogean languages Choctaw and Chickasaw). Internal possessors, in contrast, do not generally show such “external” behavior. The distinct behavior of external and internal possessors follows from their distinct syntax.
. . . T [vP DPSBJ [v [Spec,FPPossessor [F [VP V DPOBJ ]]]]]
. . . v [VP V [DPOBJ D [PossPPossessor [Poss [NP N]]]]]
In (1a), the possessor is in the specifier of FP, the projection of a functional head F that assigns case to it. FP, which can represent different phrases across languages, is in the extended projection of the verb and its specifier position is thus accessible in the clause—for example, for agreement with v. This can give rise to object agreement with a possessor rather than the object DP (as, e.g., in Nez Perce). Similarly, in Choctaw and Chickasaw, possessors can be promoted to subject and thus participate in SR regularly, just like nonderived subjects.
In contrast, (1b) shows a DP-internal possessor, embedded below a D head. In this position, the possessor is not accessible for DP-external agreement or semantic role assignment and forms a constituent with the possessed noun. This is the standard analysis of internal possessors, applicable to many languages (see Szabolcsi 1994, Alexiadou, Haegeman, and Stavrou 2007:570–575).
In this squib, we discuss possessors that are in some sense “intermediate” with respect to the internal vs. external distinction: they form a syntactic constituent with the possessed noun but nevertheless show some clause-level behavior, as if they were in a DP-external position. Possessors that have both internal and external properties are discussed by (e.g.) Lødrup (2009) for Norwegian, but he does not relate this phenomenon to the structure of the noun phrase. In contrast, we argue, building on May 1985, Chomsky 1986, and Kayne 1994, that other languages with such “intermediate” possessors provide syntactic evidence for a representation like (2), in which the possessor is adjoined to the possessive noun phrase, rather than fully contained in it.
(2) [XPPossessor [XP Spec,XP [ . . . NP]]]
Although adjoined possessors are not widely accepted, similar structures involving adjunction have been proposed for Serbo-Croatian (Despić 2013), Hungarian (É. Kiss 2002,,Dékány 2015), and Turkish (Bošković and Şener 2014).
While we will briefly comment on these languages, our focus is on Tundra Nenets (TN; Samoyedic, Uralic; Nikolaeva 2014). The main contribution of the squib lies in providing a detailed and unified analysis of the DP structure and the behavior of adjoined possessors in TN in a Minimalist framework (in contrast to previous literature, in particular Nikolaeva and Bárány 2019). We will propose that XP in (2) is in fact a DP in TN, in contrast to NP in Serbo-Croatian, as argued by Despić (2013), and Turkish, as argued by Bošković and Şener (2014), but perhaps similarly to Hungarian. TN possessors in the adjoined position show hybrid behavior: on the one hand, they are not promoted to an argument function and not interpreted as affected or assigned DP-external case; on the other hand, they behave like arguments with respect to syntactic locality—for example, being able to bind into other DPs and participate in SR.
We discuss the position and DP-internal properties of possessors in TN in section 2, and their DP-external syntax in section 3. In section 4, we situate our analysis of TN in the existing literature on the structure of the DP and discuss how it addresses conceptual challenges to adjoined possessors that have been raised in the past.
2 Lexical Possessors in Tundra Nenets
TN has two types of DP-internal lexical possessors. Both are in the genitive, but only one type triggers agreement on the possessum; see (3a–b), which have the same meaning. The distribution of demonstrative pronouns in the DP shows that the agreeing possessor is in a higher position, preceding the demonstrative.1
tʹuku˚ Wera-h ti / *te-da
this Wera-GEN reindeer reindeer-3SG.POSS
Wera-h tʹuku˚ te-da / *ti
Wera-GEN this reindeer-3SG.POSS reindeer
We call agreeing lexical possessors prominent internal possessors (PIPs). Here, we focus only on lexical possessors; while pronominal possessors in TN also control possessive agreement, they show distinct syntactic properties in other respects.
Evidence that both PIPs and regular possessors are internal to the DP comes from their identical behavior in standard constituency tests. This is demonstrated for PIPs in (4)–(6). Example (4) shows that the PIP cannot be separated from the possessed noun by sentence adverbs, even though their position is otherwise quite free. Other types of adverbs behave in the same way. In (5), a possessive phrase with a PIP is coordinated with another DP, using tədʹekəxət˚. This element can coordinate possessors inside a DP as well, and there is no evidence for verb ellipsis in (5), indicating that it involves real DP-coordination.
For coordinated subjects, as in (5), Nikolaeva (2014:416–417) shows that the verb agrees with the DP whose person is higher on the hierarchy 1>2>3 and in number with the coordinated phrase. Finally, (6) demonstrates that possessive phrases with PIPs can be contrastively focused as one constituent.2
(yetrʹi) Wera-h (*yetr i) te-x˚nəq-ta
always Wera-GEN always reindeer-LOC.PL-3SG.POSS
‘I (always) arrived on Wera’s reindeer.’
[Petʹa-hnʹa-da tədʹekəxət˚ pidər ] to˚-dʹih
Petya-GEN friend-3SG.POSS and 2SG come-2DU
‘Petya’s friend and you came (together).’
[Petʹa-h nʹa-m-ta ] yad˚btaə-d˚m, Maša-m
Petya-GEN friend-ACC-3SG.POSS meet-1SG Masha-ACC
‘I met [Petya’s friend]FOC, not Masha.’
These data indicate that PIPs form a constituent with the possessed noun. Since the Coordinate Structure Constraint holds in TN (Nikolaeva 2014:314), (5) also shows that PIPs are not extracted covertly, as in Deal’s (2013) analysis of covert possessor raising in Nez Perce, but are in fact DP-internal.
To the degree that external possession is possible at all in TN, it is only found when the host of the possessor is an (intransitive) subject and it involves topicalization of the possessor (Nikolaeva 2014:222). This is a different construction, not addressed here. In the data discussed in the rest of this squib, where a PIP is hosted by a subject, it is never a topicalized, external possessor; rather, it is an internal possessor, as confirmed by application of the tests cited above.
Possessors can be stacked, but stacking is not possible with more than one PIP, as in Petʹa-h tol˚-h ŋǣ-da (Petya-GEN table-GEN leg- 3SG.POSS) or Petʹa-h tolə-nta ŋǣ (*-da) (Petya-GEN table-GEN.3SG.POSS leg(*-3SG.POSS)) ‘the leg of Petya’s table’. This example also shows that PIPs can be nonhuman, although such constructions are not particularly frequent. While the examples below illustrate human PIPs, this is not a categorical condition, and there are generally no clear semantic restrictions on PIPs.
3 Evidence for PIPs C-Commanding out of DP
The syntax of PIPs differs from that of regular possessors. Their high DP-internal position allows them to show certain clause-level properties, and PIPs, in contrast to regular possessors, can c-command out of the DP that hosts them.
The first piece of evidence comes from binding. (7) shows that regular possessors of the subject cannot bind possessors of the object represented either by a possessive suffix (possibly with an associated pro possessor) only, (7a), or by a possessive suffix and an overt pronoun, (7b). Possessors referenced by a possessive suffix (and possibly pro) can be bound by the subject, (7a); overt pronominals must be free, (7b).3
[Wera-h nʹe˚ka ] [OBJ pro wenʹako-m-ta ]
Wera-GEN brother dog-ACC-3SG.POSS
‘Werai’s brotherj fed his/her*i/j/k dog.’
[Maša-h wǣsako ] [OBJpida xər˚-m-ta ]
Masha-GEN husband he knife-ACC-3SG.POSS
‘Mashai’s husbandj took his/her*i/*j/k knife.’
PIPs, in contrast, can bind both null and overt possessive pronouns, as shown in (8a–b).
[Wera-h nʹe˚ka-da ] [OBJ pro
wenʹako-m-ta ] ŋəwla˚
‘Werai’s brotherj fed hisi/j/k dog.’
[Wera-h nʹabako-da ] [OBLpida
Wera-GEN sister-3SG.POSS 3SG
mʹa-k˚nanata ] yilʹe˚
‘Werai’s sisterj lives in his/heri/*j/k tent.’
Additional evidence comes from SR. The same-subject converb in -(sʹ/cʹ)˚ requires its null subject to corefer with the subject of the superordinate clause. This null subject is arguably PRO: it can be bound by the c-commanding superordinate subject but no other super-ordinate argument, and it cannot have free reference either, as shown in (9). These properties are characteristic of obligatory control (OC) (Landau 2013:29).
[tol˚-h tʹax˚na ŋamtʹo-˚ ] Wera Petʹa-m məneqŋa
table-GEN at sit-SS.CVB Wera Petya-ACC see.3SG
‘Werai saw Petyaj while
Structure (2) for DP-internal possessors in TN predicts that PIPs should also be able to control the subjects of converbial clauses, while lower possessors should not. This is true. In (10), where the subject hosts a low lexical possessor, the possessed subject itself, but neither its possessor nor the object, can corefer with embedded subject. The order of the converbial clause and the matrix clause does not affect the possibility of control, as Nikolaeva (2014:383–384) also shows for another same-subject converb.
[tol˚-h tʹax˚na ŋamtʹo-˚] Wera-h nʹīsʹa
table-GEN at sit-SS.CVB Wera-GEN father
‘Werai’s fatherj saw Petyak while
Wera-h nʹīsʹa Petʹa-m məneqŋa [tol˚-h
Wera-GEN father Petya-ACC see.3SG table-GEN
tʹax˚na ŋamtʹo-˚ ]
‘Werai’s fatherj saw Petyak while
(Second author’s field notes)
However, PIPs can control the subject of a converbial clause. In (11), the agreeing possessor precedes the demonstrative, showing that it must be a PIP, and the null subject of the converbial clause can co-refer with the main clause subject’s possessor.
Petʹa-h tʹuku˚ŋǣ-da / *ŋǣ yesʹ˚mʹa
Petya-GEN this leg-3SG.POSS leg start.hurting.3SG
[tol˚-h tʹax˚na ŋamtʹo-˚ ]
table-GEN at sit-SS.CVB
‘This leg of Petyai’s started hurting when hei was sitting at the table.’
An anonymous reviewer notes that (11) might involve nonobligatory control (NOC), suggesting that ŋǣ-da ‘leg-3SG.POSS’ cannot be a controller due to the odd semantics resulting if it were. But the control relation between the possessor and the embedded subject in (11) is only possible because the possessor is not a regular possessor, but a PIP: replacing ŋǣ-da with ŋǣ does not allow the same interpretation.
(12) provides further evidence that PRO in the converbial clause is OC PRO. Landau (2013:chap. 7) argues that PRO in NOC must have a [+human] controller. If, therefore, a [−human] referent can control PRO, the relation cannot be NOC. In (12a), with a low lexical possessor, the embedded subject is controlled by the possessed noun kniga ‘book’, giving rise to an odd interpretation rather than an interpretation with the possessor as the controller. The same logic holds for (11): in the absence of a PIP, SR constructions are not rescued if the controller yields an unusual interpretation. This establishes that the relation is not NOC. In (12b), with a PIP, the possessor of the main clause subject controls the embedded subject. This contrast indicates that the relation involved here is not logophoric control (Williams 1992, Sichel 2010), since the inanimate controller kniga ‘book’ cannot be a logophor. Rather, the contrast involves a categorial morphosyntactic difference.
#[tol˚-h tʹax˚na ŋamtʹo-˚ ] ŋəcʹeki˚-h kniga
table-GEN at sit-SS.CVB child-GEN book
‘Sitting at the table, the child’s book fell.’
[tol˚-h tʹax˚na ŋamtʹo-˚ ] ŋəcʹeki˚-h
table-GEN at sit-SS.CVB child-GEN
‘When iti was sitting at the table, the childi’s book fell.’
That PIPs are able to c-command out of the possessive DP arguably explains restrictions on their distribution. PIPs cannot cooccur with certain other third person DPs in the same minimal clause (Nikolaeva and Bárány 2019). First, a PIP cannot appear if the subject of the clause is third person, unless the subject itself hosts the PIP. PIPs are compatible with first and second person subjects, however. This is shown in (13).
Maša [Wera-h ti-m /
Masha Wera-GEN reindeer-ACC
*te-m-ta ] ladə˚
‘Masha hit Wera’s reindeer.’
mənʹ˚ [Wera-h ti-m /
I Wera-GEN reindeer-ACC
te-m-ta ] ladə˚-d˚m
‘I hit Wera’s reindeer.’
Second, a PIP cannot occur when the third person object controls agreement on the verb, unless the object hosts the PIP.
Third, a PIP cannot occur if there is a free-standing third person pronoun, nʹa˚nta in (15a), in any of the core grammatical cases NOM, ACC, or DAT in the same clause.
[Petʹa-h nʹe˚ka-m / *nʹe˚ka-m-ta ]
Petya-GEN brother-ACC brother-ACC-3SG.POSS
‘I sent Petyai’s brotherj to him/herk.’
[Petʹa-h nʹe˚ka-m / nʹe˚ka-m-ta ]
Petya-GEN brother-ACC brother-ACC-3SG.POSS
‘I sent Petya’s brother (somewhere).’
Regular low possessors are not subject to these distributional restrictions.
Since coreference between the PIP and the other third person DP blocking the PIP’s presence is not necessary for these sentences to be ungrammatical, an account of these data in terms of binding theory seems unlikely at first glance. Rather, the fact that the restrictions all involve third person DPs suggests an analysis in terms of obviation (Dahlstrom 1986a,b, Jeanne and Hale 1987, Aissen 1997, Brittain 2001, Bruening 2001, Lochbihler 2012). Obviation governs the cooccurrence of third person DPs in a given syntactic domain, usually a finite clause, and crucially, only a single referent in the relevant domain can be proximate (Proximate Uniqueness).
Following Bruening (2001), we will represent proximate status as a formal feature in syntax, which we call [uProx]. Bruening suggests for Passamaquoddy that the proximate feature of third person DPs needs to be checked by a functional head H. We interpret this checking as the valuation of [uProx] by H via Reverse Agree (Zeijlstra 2012, Wurmbrand 2014).
In Nikolaeva and Bárány 2019, we argue that TN shows the effects of a syntactic obviation system (Aissen 1997). This means that not all DPs are overtly marked for their obviation status, but obviation nevertheless has syntactic consequences. To capture the distribution of PIPs with respect to other third person DPs in the clause, we suggest that there are two ways of receiving a proximate feature [uProx] in TN.
First, argument DPs can, but need not, be proximate. They are never overtly encoded as proximate or obviative, but one of the arguments of the verb is assigned unvalued [uProx] in the verbal extended projection (in the sense of Grimshaw 2000). Which argument receives [uProx] from v is determined by pronominal status and syntactic position. If there are several third person arguments, the structurally highest one is [uProx]; for lexical DPs, either a subject or an agreeing (ACC) object is assigned [uProx], in that order. Third person pronouns in TN always have animate referents, and so animacy is a grammaticalized feature for such pronouns. If present, a third person pronoun is assigned [uProx], on the basis of its inherent animacy, even if there are higher third person lexical DPs; we take third person pronouns to be lexically specified as [uProx]. These assumptions model that third person pronouns are highly ranked discourse participants and that this ranking is represented in syntax via syntactic features, as Bruening (2001: 119) argues. This implementation also matches observations by Dryer (1992) and Aissen (1997) that animacy and grammatical function can determine proximate status.
Second, we propose that PIPs are always proximate and that they are merged already carrying a [uProx] feature. That PIPs are inherently [uProx] arguably reflects their prominent role in discourse: as we demonstrate in Nikolaeva and Bárány 2019, PIPs show properties of discourse topics; that is, they behave like aboutness topics over longer stretches in discourse. By treating [uProx] as inherent on PIPs, we model that TN has grammaticalized the proximate vs. obviative distinction on possessors (PIPs are proximate while regular lexical possessors are obviative). In this respect, TN differs from Passamaquoddy (Bruening 2001:128–130) and Plains Cree (Dahlstrom 1986a:116, Aissen 1997:711–714), for example, where all possessors are (inherently) proximate and possessed nouns are always obviative. In TN, when a PIP is present, the possessed noun must be obviative, as in other obviation systems. In contrast, since regular possessors are obviative, they do not impose restrictions on the obviation status of their possessed noun: it can be either proximate or obviative, so there are situations in which the possessed noun outranks the possessor in obviation status.
The distributional restrictions on PIPs illustrated above follow from their being proximate: a sentence with a PIP cannot contain another noncoreferential proximate DP; therefore, PIPs cannot cooccur with third person subjects, objects controlling agreement, or third person personal pronouns. Thus, only PIPs are able to compete with argument DPs for proximate status; regular possessors cannot do so.
The three phenomena presented in this section have illustrated the clause-level behavior of PIPs in TN. Given the evidence from constituency, an analysis of PIPs in terms of external possession is unlikely. Instead, the distinct behavior of PIPs and low possessors with respect to binding and control indicates that PIPs are in a peripheral position in the DP from which they can c-command out of it.
4 Proximate Possessors as Adjoined Possessors
C-command out of DP is possible if the possessor is not properly contained in but adjoined to it (May 1985, Chomsky 1986, Kayne 1994). Despić (2013) and Bošković and Şener (2014) argue for Serbo-Croatian and Turkish, respectively, that adjoined possessors of the subject induce violations of Conditions B and C if the object is bound by the possessor and that possessors of subjects can bind possessive pronouns in objects, indicating that they c-command out of the possessive phrase. While Despić (2013) and Bošković and Şener (2014) analyze these data in light of the universality of the DP hypothesis and argue that possessors are adjoined to NP rather than DP, here we focus only on the possibility of adjunction.
We distinguish specifier positions from adjoined positions (like May 1985, but in contrast to Kayne 1994), so that it is only PIPs that can c-command out of DP. This leaves open the possibility that pronominal possessors are in Spec,DP. (16) shows the structure we propose for PIPs in TN (adapted from Alexiadou, Haegeman, and Stavrou 2007:575). In TN, there are two DP-internal surface positions for lexical possessors. The PIP is adjoined to DP, meaning that not all of the DP’s segments (boldfaced) dominate it; the PIP is thus not fully contained in it and can c-command out (see Despić 2013:244, Bošković and Şener 2014:111).
Position of possessors in the Tundra Nenets DP
[DP PIP-GEN [DP D [AgrP
PIP[Agrʹ Agr [PossP Low possessor-GEN [nP n NP]]]]]]
LaTerza (2016:749–751) considers, but rejects, an analysis for possessor adjunction to DP for Bulgarian and Macedonian in part because possessors follow determiners in these languages. The reverse order of possessors and determiners in TN makes LaTerza’s (2016) argument irrelevant for this language and arguably supports an adjunction analysis. Further evidence for adjunction comes from Bošković and Şener’s (2014) discussion of Turkish demonstratives and possessors. Bošković and Şener argue that both possessors and demonstratives are adjoined to the NP in Turkish, since possessors can c-command out of the NP in either order. In TN, in contrast, PIPs must precede demonstratives, suggesting that they are adjoined but demonstratives are not.
Alexiadou, Haegeman, and Stavrou (2007:551–560, 572–575) suggest that adjoined positions for possessors make it difficult to derive similarities between subjects and possessors, in part because they are not case positions. However, low lexical possessors in TN are also case-marked, showing that the high adjoined position is not a case position, but a lower position is. In (16), this position is Spec,PossP, which we assume to be the final position of low possessors and an intermediate position of PIPs, as both bear genitive case. This position precedes adjectives, quantifiers, and other nominal modifiers (Nikolaeva 2014:171–173). From Spec,PossP, PIPs move to Spec,AgrP to agree with the possessed noun after being probed by Agr. Finally, PIPs are adjoined to the DP to be able to satisfy their [uProx] feature. We assume that this feature attracts PIPs to Agr in the first place. The PIP’s ϕ-features value Agr, but its [uProx] feature is valued later (see below). A probe on Agr that is sensitive to discourse features (as in Miyagawa 2010, 2017) can also explain why overt pronominal possessors are higher than regular lexical possessors (in Spec,DP or Spec,AgrP) and why pronominal possessors usually control possessive agreement. It is also natural to assume that pronominal possessors have some discourse feature, as overt pronominal possessors are generally interpreted as contrastive, while null possessors generally have topical referents. Regular lexical possessors remain low, as they are not seen by the probe on Agr. Functionally, this would reflect that proximate DPs are generally assumed to be more prominent and/or topical than obviative DPs and thus more likely to control agreement.
The positions in (16) are attested crosslinguistically (see Alexiadou, Haegeman, and Stavrou 2007:575 for examples), and the derivation proposed here resembles that of internal possessors in Hungarian passing through specifiers in the DP in some respects, but differs in others (Szabolcsi 1994, É . Kiss 2002:161–169, Georgi 2014:198–204, Dékány 2015). One difference between TN (and Serbo-Croatian/Turkish), on the one hand, and Hungarian, on the other hand, lies in how adjunction is motivated. In contrast to lexical possessors in TN, Hungarian internal possessors can be dative or nominative, and both types control agreement on the possessed noun. Nominative possessors are assumed to be in a position corresponding to Spec,AgrP in (16), while dative possessors are higher, either adjoined or in Spec,DP, depending on the analysis. Szabolcsi (1994) suggests that Spec,DP serves as an “escape hatch” for the dative possessor, a position from where it can be extracted. É. Kiss (2002:169) argues that dative possessors project a KP and cannot be hosted in Spec,DP but must be adjoined for licensing reasons. She cites the possible cooccurrence of dative, but not nominative, possessors with demonstratives as evidence in favor of this analysis. As all lexical possessors in TN are in the genitive and can cooccur with demonstratives, this reasoning does not extend to TN.
Moreover, the adjunction analysis in (16) provides a better account of the TN data because, unlike in Hungarian, high possessors in TN show DP-external behavior. Both nominative and dative possessors in Hungarian are ungrammatical when they corefer with an argument in the clause; É. Kiss (2002:33–40, 2008) explains this in terms of VP structure, not in terms of c-command out of the DP. Independently of the Hungarian facts, the adjoined position of PIPs in (16) accounts for their binding properties (see (8)) as well as for the fact that PIPs but not low possessors can control OC PRO (see (11), (12)).
The distributional restrictions of PIPs can be derived from the interaction of the feature [uProx] with a functional head H. This head, located below CP, carries an [iProx] feature. DPs with unvalued [uProx] enter a Reverse Agree relation with H in order to value [uProx], resulting in [uProx: Prox]. Reverse Agree naturally derives multiple valuation by a single head, as in (17).
Examples (13)–(15) showed, however, that there cannot be two noncoreferential DPs with valued [uProx] features in a single clause (Proximate Uniqueness). We propose that the adjoined position of PIPs in the possessive DP derives this restriction as well. Aissen’s (1997:710) principle of “co-linking” states that two DPs in an obviation domain can both be proximate only if they corefer. Thus, if co-linking holds and more than one DP with [uProx] is valued, these two DPs must corefer. This can feed violations of binding theory if the syntactic relation between the two DPs is local enough. As we argued above, due to their position PIPs are local enough to enter binding relationships with other DPs in the clause. However, since PIPs are lexical DPs, any such binding relationships will lead to ungrammaticality: either the PIP will bind another lexical or pronominal DP that is construed to be coreferential, giving rise to a violation of Condition B or C, respectively (as in (14a) and (15a)), or the PIP is bound by another DP, giving rise to a violation of Condition C (as in (13a) and examples involving stacking of PIPs).
To conclude, we have shown that lexical possessors that trigger agreement on the possessed noun in Tundra Nenets are internal to the possessive DP in terms of surface constituency, yet their syntactic properties show aspects of external behavior with respect to binding, control, and obviation. We argued that this hybrid profile provides evidence that such possessors are adjoined to the possessive DP rather than being external or fully contained in it, and we developed an analysis that derives all aspects of the special syntactic behavior of these possessors from their position at the edge of the DP, using independently motivated mechanisms like Agree and binding. Our analysis supports the view that possessors can be adjoined, as has been proposed for Serbo-Croatian, Turkish, and Hungarian.
1 Abbreviations: 1 = first person, 2 = second person, 3 = third person, ACC = accusative, CVB = converb, DAT = dative, DU = dual, FOC = focus, GEN = genitive, LOC = locative, NEG = negative, NOM = nominative, OBJ = object, OBL = oblique, PL = plural, POSS = possessive, PST = past, REFL = reflexive, SBJ = subject, SG = singular, SS = same subject.
We want to thank Éva Dékány, two anonymous reviewers, and the Squibs and Discussion editors for valuable comments. The research reported here was partly carried out as a portion of the Arts and Humanities Research Council (AHRC)–funded project “Prominent Possessors” (AH/M010708/1). The AHRC’s support is gratefully acknowledged.