Possessor raising and structural variations within the v P-domain

In this paper, we analyze external possessors in Brazilian Portuguese, showing that those realizing Nominative-Case (Nominative-possessors) are not syntactically uniform, varying derivationally. These variations are related to the v P internal structure. In agentive argument structures, Nominative-possessors are licensed in obligatory control configurations. Non-agentive argument structures give rise to raising configurations. In turn, raising configurations are not syntactically uniform either. Unaccusative v Ps, differently from inchoative ones, contain a dative position at the v P edge. This position is used as an intermediate landing site for the raised Nominative-possessor. Importantly, a copy of a possessor in this intermediate position alters information structure at LF, having three interpretative effects: (a) affectedness: the possessor is interpreted as being affected by the denoted event, (b) presupposition of existence: the possessor must be an existent entity, and (b) contextual confinement: the possessor-possessum is interpreted as an integral part-whole at the event time.


Introduction
External possessor constructions -where there is a mismatch between syntax and semantics as the possessor is spelled out in a position outside the possessive DP to which it is thematically associated -are quite productive in Brazilian Portuguese (BP). As shown below, different possessive configurations with an external possessor are licensed in the language. (1a) exemplifies Accusative-possessors, in which the possessor, the 2SG clitic, is marked with Accusative Case, surfacing within the vP shell, while the possessum is located within a locative prepositional phrase. (1b) is a case of Dative-possessors. The possessor, the 2SG clitic, is spelled out as Dative, with the possessum surfacing in object position. Dative-possessors are less available in the language; being mostly productive in southern dialects, particularly in the state of Rio Grande do Sul. 1 (1c), in contrast, is not subject to dialectal variation, being widely acceptable. These are Nominative-possessors; the preverbal DP is readily interpreted as the possessor of the inalienable possessum realized in object position. As far as we know, (1c) is unique to BP, being unavailable in other Romance languages.
(1) a. Ele te beijou nas bochechas. he 2sg.acc kissed.3sg in.the.pl cheeks.pl 'He kissed you on the cheeks.' b. Eu te massageei os pés. I 2sg.dat massaged.3sg the.pl feet 'I massaged your feet.' c. O relógio quebrou o ponteiro. the watch broke.3sg the hand 'The watch had its hand broken.' The present study investigates Nominative-possessors, aiming to show that these constructions do not have a uniform syntax. Focusing mainly on raising constructions such as (1c) above, we 1 A reviewer questioned this statement, reporting that speakers from non-southern dialects accept sentences as (i).
Our understanding is that Dative-possessors occur in non-southern dialects as well, but not in a productive way. For example, I am a native speaker of a dialect from Minas Gerais (southeast) and I accept and produce sentence (i).
Sentence (ii), however, which is part of my parents' production (dialect from Rio Grande do Sul), is not natural in my dialect. present evidence that the syntactic structure underlying these constructions varies as a function of the type of argument structure projected within the vP shell.
Raising Nominative-possessors occur with intransitive predicates expressing a change of state. In (1c), for example, the verb quebrar projects a mono-argumental structure, with the theta-marked internal argument (the possessive DP) undergoing a change of state as a result of the action denoted by the verb. Nevertheless, intransitive predicates of this sort do not form a homogeneous syntactic-semantic class. To begin, the predicates in (2) differ semantically from those in (3). In (2), the change of state that the internal argument undergoes has an external cause. In (3) the change of state is caused internally (Kemmer 1993;Levin 1993;Alexiadou et al. 2006;Cançado & Amaral 2010).
( Both unaccusative and inchoative verbs license Nominative-possessor raising, as shown in (10) and (11). Nevertheless, in what follows, we will show that (10) and (11)  The paper is organized as follows. We will first assess the general properties of Nominativepossessor structures, showing that there are two main types: raising and obligatory control (section 2). Focusing on raising, we will then investigate differences between unaccusative and inchoative predicates, observing that unaccusatives, as opposed to inchoatives, provide an intermediary landing site for the raised possessor (section 3). In section 4, we offer novel evidence suggesting that the intermediate position under consideration is a context-grounding position along the lines proposed by Uriagereka (2000). In section 5, we conclude the paper, summarizing our findings and their potential contributions to the study of possessive structures and information structure, while making note of an issue that requires further investigation.

Nominative-possessors: control and raising
BP is a partial null subject language, with finite null subjects behaving as obligatorily controlled elements (Modesto 2000;Rodrigues 2004;Holmberg et al. 2009;Nunes 2019;Martins & Nunes 2021, among others). In tandem, null possessives in agentive structures like (12) are also obligatorily controlled (Rodrigues 2004;Floripi & Nunes 2009 Although differences between standard theories of obligatory control and Hornstein's (1999;) movement theory are irrelevant here, in more recent analyses, these null possessors are taken to be the residue of movement through a thematic position, as represented in (15). The possessor moves to Spec-TP to value its Case feature, passing through Spec-vP, where it collects the agent theta role (Rodrigues 2004;Floripi & Nunes 2009;Rodrigues & Dal Pozzo 2017 As mentioned in the introduction, and as (16) and (17)   In section 3 and 4, we will argue that (18) is the right structure for (17), but not for (16). For now, let us first focus on previous analyses, which do not endorse the scheme in (18) at all, arguing that in sentences like (16) and (17), the possessor is realized in a sentential topic position, above TP. The argumentation supporting this proposal is mistaken, however.
Brazilian researchers have analyzed sentences with possessor-raising as topic-subject constructions, with the possessor surfacing at the left periphery of the sentence, triggering verbal agreement morphology (Pontes 1987;Galves 1998;Lobato 2007;Kato & Duarte 2017;Munhoz & Naves 2012). 2 Assuming BP to be a topic-prominent language (Kiss 1995), and couching their analysis within the recent C-T feature inheritance framework proposed in  (16) and (17) as in (19), where the possessor is spelled out in Spec-αP, with α receiving φ-features and topic-focus feature from C. b. O assassino do Smith, se existe um, vai mofar a bunda na the assassin of.the Smith, if is one, will.3sg mold.inf the ass in.the cadeia. jail Idiomatically: 'Smith's assassin, if there is one, will be in jail for a long time.' Literally: 'Smith's assassin, if there is one, will get mold on his ass in jail.' c. Algum nenezinho está caindo o cabelo aqui no berçário. some baby.dim is.3sg falling the hair here in.the nursery 'Some little baby is losing his/her baby hair here in the nursery.' Nunes & Kato (2023) also observe that sentences like (16) and (17) are appropriate answers to out-of-the blue questions such as (21). This is not a well-suited property of topics. Sentential answers to out of-the-blue questions force a reading focused on the event itself, a statement about what happened, a thetic judgment in Kuroda's (1972)  For the sake of completeness, let us mention that the sentences in (20) can also serve as appropriate answers to out-of-the-blue questions, ruling out, thus, the possibility of a broad focus interpretation. 3 In view of the above, Kato & Ordóñez (2019), as well as Nunes & Kato (2023), assume that Nominative-possessors are spelled out in Spec-TP, although they do not uphold the structure in (18), as we will see shortly.
Important to our present discussion, Kato & Ordóñez observe that, even though BP had undergone diachronic changes in the C-T domain (loss of referential null subjects and impoverishment of verbal agreement morphology), the emergence of possessor raising in the language is unrelated to these changes. This observation results from comparisons between BP and Dominican Spanish. Dominican Spanish is comparable to BP in that it lost null subjects too A este reloj se le rompió la aguja. to this clock refl 3sg.dat broke.3sg the need 'The clock's hand broke.' (23) *El reloj rompió la aguja. the clock broke.3sg the needle ('The clock's hand broke.') An obvious difference between BP and Dominican Spanish is the clitic system. While Dominican Spanish kept 3Person clitics, BP lost them. To exemplify this loss, consider the Dativepossessor constructions in (1b) again. In addition to being productive only in some dialects, these constructions are restricted to 1Person and 2Person clitics because the 3Person clitic is not available anymore. As a matter of fact, BP does not license 3Person Dative-possessors at all, independently of them being clitics or not. As Torres Morais & Lima-Salles (2016) observe, 3Person Dative-possessors, clitics or full DPs, are readily licensed in other Romance Languages, but not in BP. The sentences in (24), for example, are acceptable in European Portuguese, but not in BP. Thus, by and large, BP has lost Dative-possessors. We will not discuss it, but the availability of dative-possessors with 1Person and 2Person clitics might be just a diachronic residue.  (16) and (17) (2020), following Landau (1999), point out that the movement of the possessor through the edge of the vP (Spec-DatP in (25)) has, as a consequence, the semantics of affectedness: the possessor is interpreted as being affected by the event denoted by the verb.
To recapitulate, BP has two well-defined types of Nominative-possessors: obligatory control and raising. There is compelling evidence that in both configurations, the possessor surfaces in sentential subject position (Spec-TP). However, assuming (25), in raising, but not in obligatory control, the possessor passes through Spec-DatP. In what follows, we will show that this is partially true. It is arguably true that Spec-DatP does not involve obligatory control, but it seems firmly false that all raising configurations are like (25)

Affectedness and movement through Spec-DatP
Affectedness is understood as a change of state, location or existence obtained and maintained by the internal argument due to the event denoted by the verb (Fillmore 1970;Anderson 1979;Tenny 1987;Dowty 1991;Beavers 2006;2011;Arsenijević et al. 2020). It is implicated in many syntactic-semantic processes, including those required for nominal passive and middle formation (Anderson 1979;Fagan 1988;Condoravdi 1989;Doron & Rappaport 1991). The contrast in grammaticality between (26) and (27)  (28) a. The enemies will not sink Black Pearl easily. b. Black Pearl does not sink easily.
Affectedness has been linked to the aspectual properties of the predicate. Verbs denoting delimited events select for direct objects that are event delimiters, and the degree by which the object is affected measures out the duration of the event (Tenny 1987;1994). More recent investigations have proposed that affectedness is a relational notion involving the event, the direct object and an aspectual scale projected by the verb, defining the progress of the change impinged on the object (Kennedy & Levin 2008;Beavers 2011;Arsenijević et al. 2020). Beavers (2011) proposes that the aspectual scale is an argument of the verb. In an event-decomposition analysis, this can be syntactically modeled by positing that the argument structure of affecting predicates contains an extra projection in which the direct object and the aspectual scale interact structurally in a Spec-Head relation.
Affectedness imposes a presupposition of existence on the direct object (Anderson 1979;Beavers 2011). In (26) and (28) Tim hat der Nachbarin das auto gewaschen. Tim had.sg the neighbor.dat.fem the car washed 'Tim washed the neighbor's car.' As considered above, it has been suggested that affectedness is a thematic relation, and the literature on Dative-possessors offers analyses along the same lines. Within Government & Binding Theory and the Minimalism, it had been proposed that Dative-possessors are obligatory control configurations, where the possessor is base generated, externally merged, in its surface position, where it receives an 'affected' theta-role (Guéron 1985;Borer & Grodzinsky 1986;Tellier 1991;Authier 1992;Hole 2004, among others). Subsuming Dative-possessors as obligatory control configurations is a possibility, although we concur with Landau's (1999) observation that affectedness is a diffuse notion that does not fit well the linguistic characterization of theta roles.
We will revisit this issue in section 4, but, for the time being, we will neutralize it, concentrating on the observation that Dative-possessors are subject to the affectedness constraint, which gives rise to the affected-possessor entailment in (30) and to the presupposition of existence in (31).
We will use the affected-possessor entailment and the presupposition of existence as tests to detect affectedness.
Unsurprisingly, residues of Dative-possessors in BP, illustrated in (1b) -repeated below as (32) -behave as described above with respect to the affectedness constrain. The possessor is affected by the denoted event, holding an affected-possessor entailment. (32) entails I massaged you. As for the presupposition of existence, it cannot be easily tested in BP since in this language Dative-possessors are restricted to 1Person and 2Person clitics.
Eu te massageei os pés. I 2sg-dat massaged.3sg the.pl feet 'I massaged your feet.' In Nominative-possessors, the affectedness constraint applies to the possessor in raising configurations whenever the predicate is headed by an unaccusative verb. (33), for example, forcedly triggers a reading in which the possessor is affected by the burning event, resulting in an affected-possessor entailment. It also responds well to the presupposition of existence test.
(34) is infelicitous if used to describe a situation in which the pig is dead. Notice the contrast in acceptability between (34) and (35) Eu vou crescer o seu cabelo um pouquinho para você ficar mais I will.1sg grow.inf the your hair a little.bit for you stay.inf more bonita. beautiful 'I will make your hair grow a little bit so that you can look prettier.' In addition to these observations, Nunes & Kato (2023)   (46) As mulheres 1 vão todas 1 dançar com os maridos na festa. the.pl women will.3pl all.pl dance.inf with the.pl husbands in.the party 'All the women will dance with their husbands at the party.' In raising configurations with unaccusative verbs, a similar linear sequence is obtained, with the quantifier floating between the future modal and the main the verb, as (47) exemplifies. Since these are not agentive structures, the quantifier is in a higher position within the vP, but it is not Spec-vP. We take it to be Spec-DatP.
Essas crianças 1 vão todas 1 queimar o pé na fogueira. these children will.3pl all.pl burn.inf the foot in.the bonfire 'All these children will burn their feet in the bonfire.' Unsurprisingly, in raising configurations with inchoatives predicates, a floating quantifier between the future model and the main verb is not fully acceptable. 4 (48) ?*Essas crianças 1 vão todas 1 cariar os dentes nesse verão. the.pl children will.3pl all.pl decay.inf the teeth in.this summer ('All the children will get cavities this summer.') Lunguinho (2006) and Cançado & Negrão (2010) propose that a raised possessor escapes the possessive DP passing through Spec-DP, where a quantifier may get stranded. Taking this to be a possibility, let us add, however, that this analysis does not account for the contrast above, which indicates a structural distinction between unnacusative and inchoative predicates.
To close this section, let us consider distributive effects. Vergnaud & Zubizarreta (1992) pointed out that in Dative-possessors where the possessor is plural (49), a possessum denoting a body-part is interpreted as a plurality even though it is morphologically singular. Although this effect is quite strong in Dative-possessors, it seems unrelated to the Dative position itself. As shown 4 One of the reviewers made two observation about (47) and (48). We will go through them separately. Firstly, it was observed that some speakers of BP pointed out that cariar is not a frequent verb. This might be a dialectal difference. In my dialect, cariar is a regular verb. (i), for example, is a perfect sentence.
(i) Cara, eu estou com alguma coisa. Todos os meus dentes cariaram.  (48), the floating quantifier might be in a position higher than vP, given that quantifiers can also float in sentences like (ii) and (iii). This is potentially interesting and indicates that careful research on floating quantifiers in different dialects of BP is in order. Neither (ii) nor (iii) is grammatical in my dialect. Thus, it might be that we reject (48) because we cannot float a quantifier in a position outside vP.
for sure the.pl women will.3pl all.pl have.inf danced with the husbands.
'For sure, all the women will have danced with their husbands.' (ii) Com certeza as mulheres vão todas estar dançando com os maridos.
for sure the.pl women will.3pl all.pl be.inf dancing with the husbands.
'For sure, all the women will be dancing with their husbands.' in (50)-(52), in BP the same distributive effect is obtained for Nominative-possessors in general: control (50), raising with unaccusative predicate (51) and raising with inchoative structure (52).
In all these sentences, the possessum is forcedly singular. The presence of a plural possessum leads to a somewhat awkward interpretation. Thus, it seems to us that in constructions with external possessors, the distributive reading is a reflex of the possessor having scope over the possessum. not have a full understanding of the syntax and semantics of affectedness yet, in this paper we remain neutral with respect to the thematic status of Spec-DatP, but we revisit the issue in the following section, providing some insights on its nature.
Next section, we focus on the pragmatic import of Spec-DatP, showing that movement through it alters information structure. Uriagereka (2000) observes that Spanish possessive constructions like (53) have two layers of semantic information: a conceptual one related to the expression of an inalienable possession (i.e., the relation between she/ella and the umbilical cord) and an intentional one, relating the seeing event and the umbilical cord. (54), with a dative-clitic possessive, adds another type of intentional information: this proposition is not about any umbilical cord, but about an umbilical cord that was attached to its owner (ella) when the sight event took place. That is, double dativepossessors force an integral possessum-possessor (part-whole) reading that canonical possessives do not. Although (53) forces an inalienable possession reading, it is felicitous even in a situation in which the sight event took place when the umbilical cord was not attached to its owner.  Higginbotham (1988), assumes that predicates, including nominals, introduce contextual variables, which are set pragmatically by the speaker, and these variables can ground each other in a process of contextual confinement. The intuitive integral reading of (54) is a consequence of this. The speaker confines the context of the possessum (cord) to that of the possessor (her) at the time of the sight event. Assuming a neo-davidsonian semantics, (55a) (translated as (55b)) is, thus, assigned to (54), where X, Y and Z are free contextual, second order variables predicated of e, e′, and x respectively. b. There is a sub-event e of EXPERIENCING, a sub-event e′ of SIGHT, and a relation x such that I is the subject at e and e′ is the object at e, and x is the subject at e′, and cord is a part at x for her a whole at x, where the speaker confines the range of the quantification over e′ to contexts Y defined over her. (Uriagereka 2000: 410) Clearly, the pragmatic process at issue is not an extra-linguistic phenomenon, as it depends on the syntactic structure projected within the vP shell, and, most important to our present discussion, it involves a double-possessor structures with a dative clitic. (54), but not (53), triggers contextual confinement. That is to say that the internal structure of a possessive DP translates into conceptual information relative to possession, while the internal structure a vP containing a filled Dative translates into an intentional information relative to contextual grounding.

Spec-DatP and contextual confinement
In Uriagereka's syntactic analysis, (54) involves movement of the possessor to Spec-vP.
Thus, although (54) is an instance of clitic doubling, Uriagereka's analysis, mutatis mutandis, is similar to ours in (44). Both analyses posit a copy of the possessor at the edge of the vP domain.
Therefore, contextual confinement is expected to occur in BP possessor raising.
Contextual confinement is indeed observed in external possessive constructions in BP, but not in all of them. It applies to Dative-possessors. (56a), for example, is appropriate only if the hair forms an integral part-whole with its owner, the 2SG dative-clitic. This is not observed in (56b), for which a non-integral reading is equally possible. The utterance in (56b)  Accusative-possessors also display obligatory contextual confinement. (57a), as opposed to (57b), cannot be used in a situational context in which I have a mechanical leg that is not attached to me when the shot hit it. This suggests, given the analysis above, that what we are calling Accusative-possessors in this paper, similar to Dative-possessors, is spelled out in a high position within the vP. 5 As for Nominative-possessors, the integral reading depends on the structure. In obligatory control configurations, it is not forced. If, the verb pintar is used transitively, as in (58), with the DP 5 A reviewer pointed out that the reading contrast between (57a) and (57b) does not emerge in Spanish. According to them, (57a) can refer to a mechanical leg. We acknowledge this observation, but emphasize that contextual confinement is a tricky matter in that it requires making reading judgments under a particular, well-set situational context, as we will exemplify shortly. Our point is that (57a), in comparison to (57b), is infelicitous only if the mechanical leg under consideration is NOT attached to its owner's body (I) at the time it was hit by the bullet. Consider, for instance, the oddness of (i), uttered by a driver, sitting in the driver's seat when the implied shooting event took place. But expectedly contextual confinement distinguishes raising in unaccusative structures from raising in inchoatives. To see this clearly, let us set an appropriate scenario. Suppose we work in a laboratory that is conducting research on new species of exoskeleton animals, including snails, which are able to leave behind their shells when they go out looking for food and mates. They also abandon old shells. Our work in the lab involves writing short sentences that accurately report everything that happens to the animals under observation. The sentences in (59) are both grammatical, but (59a) is an accurate report only if the shells were on the snails when they broke. That is, (59a), as opposed to (59b), forces a possessor-possessum integral reading. (59b), in contrast, does not have this reading restriction; its truth conditions do not depend on the shell being on the snail when it actually broke. There is an interaction between the contextual confinement under discussion and the presupposition of existence presented in section 3. To see this clearly, add the following information to the laboratory-scenario set above: to avoid spreading diseases, we incinerate dead animals every single day. In this context, (63a) is a very odd report, while (63b) is fine. The possessor raised in (63a) forces an integral reading that is impossible, given that the DP os caramujos (the snails) denotes a set of no longer existent snails. This is not an unexpected observation.   If the line of reasoning followed above is on the right track, the presupposition of existence under discussion is a consequence of assigning an affected reading to the possessor's copy at the edge of the vP, with contextual confinement being another layer of information projected by this syntax. As already said, our current theoretical understanding of the affectedness constraint is rather nebulous. But, while acknowledging that we do not have a firm basis to discuss this issue in depth here, we should ask whether the affected-possessor interpretation at issue is somehow associated to the contextual confinement process. Is it a corollary of contextual confinement?
We think the answer is yes. Here is a piece of evidence. As the ungrammatically of (65), in comparison to the grammatically of (66), indicates, external possessors seem to be available only in structures in which the possessive DP is itself interpreted as being affected by the verbal action. In contrast to all the examples of external possessor presented above, the examples in (65) are formed by verbs that do not trigger object affectedness. Therefore, we have the following: by raising a possessor to or through of the edge of vP, speakers ground the context variable value assigned to the possessum to that assigned to the possessor, and, as a result, they conceptualize the possessum-posssessor as a whole, undivided entity at the moment the (sub)event affecting the possessum occurs, as represented in (55). Consequently, under contextual confinement, the affectedness of the possessum is carried over to the possessor.
In other words, external possessor configurations are similar to middles and nominal passives in that they are formed from argument structures in which the internal argument is affected by the verbal action. Affectedness of a possessor in a dative position is not an inherent semantic feature of these configurations, but a by-product of contextual confinement.
Let us turn to another remaining issue in our discussion. There is not yet a principled explanation for why a specific syntactic configuration (presence of a copy of a possessor at the edge of vP) triggers contextual confinement. Raposo & Uriagereka (1995), working on individual level and stage level readings within small clauses, put forward an analysis according to which contextual confinement piggybacks on scope interactions between associated operator variables, as stated in (67), where X and Y are contextual variables. According to (67), in individual level interpretations, the predicate context variable is confined to that of the subject because the subject has scope over the predicate. In stage level interpretations, it is the other way around: the predicate has scope over the subject. Thus, the subject's context variable is confined to that of the predicate. (67) Given a structure … [ … Xx … Yy…] …., the value of Y is set relatively to the value of X only if the operator Ox takes scope over the operator Oy. (Raposo & Uriagereka 1995: 194) However, as Uriagereka pointed out to me (p.c.), trying to reduce contextual confinement to scope interaction must address some problems relative to the nature of the variables at issue.
While scope relations involve binding of first order variable values, contextual confinement is a relation among second order variable values, and it is an open empirical question whether or not this type of relation is subject to the same restrictions as scope relations (e.g. c-command).
Furthermore, the BP data discussed above provide evidence that contextual confinement is not carried out by scope interactions. If it were, all cases of Nominative-possessors, including control and raising in inchoative structures, should trigger contextual-confinement because in all of them the possessor is spelled out in Spec-TP, having, thus, scope over the material inside vP.

Conclusion
Focusing primarily on Nominative-possessor in raising configurations in BP, we have shown that these configurations are syntactically heterogeneous. Their derivational history varies as a function of the argument structure built within the vP. In contrast to inchoatives, unaccusative structures provide an intermediate position for raised possessors at the edge of the vP (Spec-DatP).
Although there is strong evidence that in BP, due to diachronic changes, DatP lost its ability to license 3Person dative possessors, our observations indicate that, in unaccusative structures with Nominative-possessors, Spec-DatP is filled by a non-pronounced copy of a raised possessor.
This syntax has three interpretative effects. Besides conceptual information on the nature of the possessive relation, the possessor is interpreted as being affected by the action denoted by the unnacusative verb, being, thus, subject to a presupposition existence. Also contextual confinement is forced, with the possessum-possessor being conceptualized as a whole, integral entity at the time the event takes place. On more speculative grounds, we have considered that affectedness and the resultant presupposition of existence on the possessor might be a corollary of contextual confinement.
Our conclusions contribute to research on BP grammar by reinforcing and extending the hypothesis that the so-called BP topic-subjects involve structural changes not at the left periphery of the sentence, but at the vP level. Also, although we do not yet know why movement to or through the edge of vP triggers contextual confinement (a phenomenon first observed by Uriagereka (2000) with respect to double dative-clitic possessives in Spanish), our findings contribute to the study of information structure at the vP level. We have shown that the syntax of possessor raising, particularly in structures involving Spec-DatP, alters pragmatics, adding intentional information. Hence, contextual confinement is a broad pragmatic effect, maybe universal, of A-movement through the edge of vP.
To finalize, let us bring back Nunes & Kato's (2023) remark that resumption of a Nominativepossessor arises only in unnacusative structures, not in inchoative structures. For convenience, the relevant data are repeated in (68). The acceptability of (68a) puts unnacusative Nominativepossessors on a par with control Nominative-possessors, which support pronominal resumption as well, as shown in (69) This observation resurrects the hypothesis that affectedness is a theta-role. But, putting on hold the temptation to treat unnacusative Nominative-possessors as control, let us add a final note. Resumption on raised Nominative-Possessors changes the interpretation. (68a) does not force contextual confinement. It is felicitous even in a scenario in which the foot at issue was disconnected from its owner's body (Maria) when it got burned. Under this non-integral reading, the possessor is not affected. Therefore, an analysis of (68a) as control, with the possessor DP receiving an affected theta-role, is not straightforward.