10ho-raw [⌂ EvenYrbk]

10ho, PDF
Katalin Horváth

Possessive Interpretations

0 Introduction Possessors, being argument DPs in the possessive DP,1 need a thematic-role so that the exact relationship between them and the possessed can be determined. There are several approaches to the process of nominal theta-marking in the literature on possession. The present paper addresses the problem of theta-role assignment within the possessive DP and aims at arguing against the assumptions according to which the possessive interpretation originates in the semantics of the possessed noun alone. This essay intends to present an alternative approach to this problem by claiming that the possessor receives its theta-role from a theta-role assigning unit consisting of the possessed and the possessive morpheme.2 In other words, this paper approaches the issue of theta-marking from the perspective of Baker’s (1988) UTAH which will be adapted to the syntactic processes of the nominal domain. In order to be able to exploit the findings of Baker, section 1 introduces the UTAH in the verbal domain. Then, in the light of the CP-DP Parallelism,3 section 2 adapts this theory to the nominal domain. So, this part of the paper focuses on the argument structure of the possessed noun and the different kinds of possessive interpretations. Section 3 argues that the possessive interpretations presented in section 2 can be reduced in number and discusses the structure of the nP-shell and the role of the context in the disambiguation

1

2

3

The possessive DP is the DP hosting the possessed in its N head and the possessor in a specifier position. According to Szabolcsi (1994), the possessor is theta-marked by the complex head (N+I) hosting the possessed in N and the possessive agreement morpheme in I. Szabolcsi (1994) proposes that the structure of the CP and the DP is parallel. That is to say, both of them contain a similar set of functional and thematic projections. In other words, both of them are headed by a functional category (by C and D respectively) enabling them to function as arguments of a lexical category. In addition, both verbal and nominal agreement is made possible with the help of agreement projections responsible for case marking the subject/the possessor under spec-head agreement. Besides, Szabolcsi (1994) assumes that even the process of theta-role assignment is identical in the verbal and in the nominal domain. As a consequence of the CP-DP structural symmetry and the similarities between subject and possessor extraction, she concludes that the possessor can be conceived of as the ‘subject’ in the possessive DP.

Katalin Horváth

Possessive Interpretations 2

of the possessive relation. Finally, section 4 summarises the results and provides a conclusion.

1 The UTAH in the verbal domain According to Baker’s (1988) “Uniform Theta-role Assignment Hypothesis (UTAH), specific theta-roles are assigned to similar positions in all structures,” (Newson et al. 2006: 153). This means that a given theta-role is assigned to a given (specifier)4 position in all structures. Hence, there is a need for a multi-layered thematic projection so that all the arguments associated with a lexical category can be accommodated in separate specifier positions where they are theta-marked. Thus, the UTAH is based on the assumption that though the theta-grid of a lexical category contains all the thematic-roles associated with its argument structure, it is not able to theta-mark all of its arguments alone but only with the mediation of some heads having their own semantic contribution to make. For example, in the verbal domain there are vprojections on top of the thematic V’s maximal projection, as shown in (1). These vPs are the extended projections of V and they can be phrases headed by for instance an ‘agentive’/’causative’ or ‘experiencer’ v etc. mediating thematic-roles such as <agent>/<causer>, <experiencer> etc. For example, in (1) V directly assigns the theta-role <theme> to the argument hosted in the [Spec, VP] position, whereas the agent in [Spec, vP] gets its theta-role with the mediation of the agentive v. As stated above, the theta-role <agent> is idiosyncratically part of the theta-grid of the thematic V, but it can only be assigned to an argument if v equips V with an agentive interpretation. As a result, v’s task is to help mediate the theta-role to V’s argument in [Spec, vP]. (1)

4

The complement position is disregarded as it is irrelevant from the point of view of the present discussion.

Katalin Horváth

Possessive Interpretations 3

So, Baker (1988) proposes that verbs can be classified based on the system of their vP-projections (i.e. based on their morphological structure) because they reflect the event structure with which the verb is associated. In other words, a predicate’s event structure correlates with its valency.5 The examples in (2)(5) illustrate these assumptions (Newson et al. 2006: 166). (2) He rolls the ball. e6 = ei ej : ei = ‘he does something’ ej = ‘the ball rolls’ (3)7

(4) (Ő) gur-ít-ja a labdá-t. He-NOM roll-CAUS-3.SG the ball-ACC ‘He rolls the ball.’ e = ei ej : ei = ‘he does something’ ej = ‘the ball rolls’

5

6

7

In English the thematic v heads are manifested by null-morphemes. Thus, the valency of the English verbs is reflected in the system of their thematic null-morphemes, i.e. in their morphological structure. Predications can be decomposed into subevents represented by light verbs (vs) in the structure: e=event structure, ex=subevent. Case and agreement relations are not represented in the trees because they are not in the focus of the present discussion.

Katalin Horváth

Possessive Interpretations 4

(5)

The complex event structure in (2) and (4) suggests a complex vP-shell in the structures presented in (3) and (5). It can be supposed that the agentive/causative element in v (which is phonologically non-empty in Hungarian, see (5)) equips the V with an agentive/causative interpretation which helps V mediate an <agent>/<causer> theta-role to the subject. In other words, gur- without the bound causative morpheme cannot assign the thetarole <agent> present in its theta-grid. Consequently, the theta-marking of the subject in (5) is only possible if the agentive/causative suffix helps the verb assign the theta-role <agent>/<causer> to the [Spec, vP] position. Hence, V and v can only assign theta-roles together to the arguments accommodated in the vP-shell. As a consequence, the verb must undergo V-to-v movement to pick up the affix base generated in v. A similar process is assumed to take place in English, although there are no overt morphemes associated with v.

2 Theta-role assignment in the nominal domain In the light of the CP-DP Parallelism, theta-role assignment in the nominal domain could be conceived of as a process similar to theta-role assignment in the verbal domain. Szabolcsi’s (1994) and Olsen’s (1989) analysis of thetarole assignment within the possessive DP are reminiscent of the UTAH because they propose that the possessor’s theta-role originates partly in the semantics of the possessive morpheme which makes the possessed N capable of assigning a theta-role to the possessor. In other words, as opposed to verbs, N is able to theta-mark the possessor only together with the possessive affix. (Hence, the [Spec, NP] position is not a theta-position.) Under the guidance of the UTAH Roehrs (2005) proposes that possessive markers can be conceived of as light nouns (ns) similar to light verbs (vs) in the clausal domain, being able to assign a theta-role together with the possessed (N) to the possessor

Katalin Horváth

Possessive Interpretations 5

base generated in a [Spec, nP] of the nP-shell.8 Thus, it can be supposed that the theta-role of the possessor does not originate in the semantics of the possessed noun alone, but the possessive marker (n) has its own semantic contribution to make, too. Hence, N and n together can be assumed to be the unit responsible for theta-marking the possessor. According to Szabolcsi (1994) and Olsen (1989), derived and non-derived nouns need separate treatment when their argument structure is examined. Derived Ns inherit the argument structure of the verbs (or adjectives in some languages (e.g. in Hungarian)) they are derived from, whereas referential nouns9 do not have an argument structure, but the possessive affix in n has its own semantic contribution to make, so the non-derived noun together with n can assign theta-roles.10 2.1 The argument structure of the possessed noun (N) As mentioned above, referential nouns do not have an argument structure. Consequently, like derived nominals, they can never assign theta-roles without possessive morphemes in n. That is why Szabolcsi (1994) and Olsen (1989) assume that the theta-role of the possessor does not originate in the semantics of the possessed noun alone but the semantics of the possessive morpheme attached to N is also indispensable for the possessive interpretation. According to Roehrs (2005), these thematic affixes are similar to the verbal thematic affixes in v (e.g. agentive suffix in (4) and (5)) helping V assign theta-roles to its arguments. According to Szabolcsi (1994), the structures of DPs and CPs are identical. In addition, she also argues that possessors can be analysed as subjects inside possessive DPs. Thus, the nP projections presented in Roehrs (2005) are the mirrors of the vPs in (3) and (5). In other words, the possessor gets its theta-role from N and n in [Spec, nP] (see (6)). 8

Note that in English the possessive light nouns are manifested by null-morphemes, whereas they are phonologically realised in Hungarian, see (i). (i)

9 10

a te kalap-ja -i -d the you-GEN hat -POSS-PL-2.SG ‘your hats’

Referential nouns are taken to be proper names and common nouns. According to Newson et al (2006), the thematic elements which are not case assigning categories cannot theta-mark their arguments alone. To put it differently, [Spec, NP] and [Spec, AP] are not theta-positions. Hence, there is always a need for n in possessive DPs no matter whether the N head (the possessum) is derived or not. Otherwise, theta-role assignment cannot take place.

Katalin Horváth

Possessive Interpretations 6

(6)11

In sum, Szabolcsi (1994) claims that deverbal nouns inherit the argument structure of the verb they are derived from, so the arguments of the noun get traditional theta-roles (<agent>, <theme>, <experiencer> etc.) from the possessed and the possessive morpheme. On the other hand, in the case of referential nouns the possessive n together with N is responsible for the thetamarking of the possessor with an arbitrary theta-role <poss>. Some might doubt that affixes (as functional categories) are able to assign theta-roles. However, it can also be said that v and n are not functional categories. That is to say, they bear the characteristics of functional and thematic categories at the same time. Moreover, these thematic affixes assign theta-roles together with the lexical category they are attached to. Consequently, N and n together are responsible for theta-marking in the possessive DP. In other words, none of them is able to theta-mark the possessor alone. 2.2 Possessive interpretations Greenbaum and Quirk (1990) distinguish seven types of genitive12 meaning, i.e. according to them, ‘genitive’ in English has seven kinds of interpretation: 11

12

The present paper deals exclusively with the process of theta-role assignment in the nominal domain. Hence, the functional layers of the possessive DPs are disregarded in the trees. That is why case marking and agreement relations are not represented graphically throughout this essay. Greenbaum and Quirk’s (1990) term ‘genitive interpretation’ is confusing because it implies that the English possessive of-construction contains the case genitive. As in the present paper this kind of construction is not regarded as a manifestation of genitive case, the term ‘possessive interpretation’ is used throughout the argumentation when possible. (The of-construction is assumed to be an instance of P-insertion. In other words, a semantically empty preposition is inserted into the structure to assign accusative to the postnominal possessor.)

Katalin Horváth

Possessive Interpretations 7

(i) ‘subjective genitive’, (ii) ‘objective genitive’, (iii) ‘attributive genitive’, (iv) ‘possessive genitive’, (v) ‘partitive genitive’, (vi) ‘descriptive genitive’ and (vii) ‘genitive of origin’. Quirk et al (1989) has an additional category: (viii) ‘appositive genitive’.13 In the next subsections of the paper these eight genitive interpretations will be illustrated and discussed in the light of the distinction between derived and referential nouns presented above. Subsection 2.2.1 deals with Greenbaum and Quirk’s (1990) ‘genitive interpretations’ (i)(iii), whereas subsection 2.2.2 discusses the interpretations (iv)-(viii). 2.2.1 Derived nominals Previously it has been argued that in possessive DPs the n and N heads are responsible for theta-marking the possessors, so Greenbaum and Quirk’s (1990) ‘genitive interpretations’ are associated with different kinds of ns in the possessive DP. It has also been mentioned that deverbal nouns inherit the argument structure of the verb they are derived from. Thus, they and their ‘light nouns’ (ns) together are able to assign traditional theta-roles to their arguments, see (7)-(8). (7) a.

13

The enemy destroyed the city. <agent> <theme>

There are several ways of grouping possessive interpretations. Barker (2008: 9) classifies the possessive relations from a semantic point of view, as shown in (i). His ‘inherent’ and ‘part-whole’ relations constitute lexically predetermined possessive interpretations with Ns such as mother or arm. In other words, he proposes that these kinds of possessive relation are lexically encoded in the semantics of the possessed noun. The ‘agentive’ category in (i) can be conceived of as the subcategory of Greenbaum and Quirk’s (1990) ‘subjective genitive’. ‘Control’ means legal or physical control equalling ‘possessive’, whereas ‘others’ cover some additional interpretational categories of possessive constructions. With his umbrella term ‘pragmatic’, Barker (2008) regards the latter three ‘possession relations’ to be unspecified. In other words, Barker claims that the exact meaning expressed by these categories can only be determined with the help of some contextual information which makes the semantic disambiguation of the possessive structure possible. (i) lexical pragmatic

Possession Relation inherent part-whole agentive control others

Katalin Horváth

Possessive Interpretations 8

b. 14

c.

the city’s destruction <theme>

d.15

e.16

the enemy’s destruction <agent>

f.

14

15

16

As noted, the present paper does not deal with agreement and case relations. Thus, the trees do not contain information about these phenomena. As proposed, as opposed to case assigning verbs, nouns in themselves cannot assign thetaroles to their arguments even if they are derived from verbs. As a consequence, there is always a need for ns (for all arguments a separate one) in the possessive DP. That is why the theme the city cannot be in [Spec, NP]. In this example the enemy can also be conceived of as a <theme> argument in the DP. The issue of ambiguity is examined later.

Katalin Horváth

Possessive Interpretations 9

(8) a. The boy runs. <agent> b.

c. the boy’s running <agent> d.

In the examples in (7) a transitive verb was nominalised, so the possessor in the derived DP is either the agent (cf. (7f)) or the theme (cf. (7d)). In (8) the intransitive verb run was nominalised. Hence, the possessor in (8c) is the agent. In sum, (7) and (8) illustrate that in the case of deverbal Ns the noun inherits the theta-grid of the verb it is derived from. (7c) is an instance of the objective interpretation, whereas (7e) and (8c) are examples for structures with a subjective n (Greenbaum and Quirk 1990). These two kinds of n are labelled on the basis of the function the possessor originally has in the clause. To put it differently, ‘subjective’ and ‘objective’ are defined functionally, i.e. reflecting the syntactic functions the possessors have in the original active clause. Greenbaum and Quirk’s (1990) ‘attributive genitive’ occurs in constructions synonymous with clauses containing a verb+adjective predication illustrated in (9a). In this case the noun inherits the argument structure of the adjective, consider (9). The Even Yearbook 9 (2010), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061-490X, http://seas3.elte.hu/delg/publications/even, © 2010, Katalin Horváth

Katalin Horváth

Possessive Interpretations 10

(9) a.17 The victim was very courageous. <experiencer> b.18 the victim’s courage <experiencer> In sum, there are three types of possessive interpretations occurring within DPs of derived nouns: (i) subjective, (ii) objective and (iii) attributive. The Ns in these structures inherit the theta-grid of verbs or adjectives, so they are able to assign traditional theta-roles (<agent>, <experiencer>, <theme> etc.) together with their light nouns. To put it differently, the inherited theta-grid gets reflected in the nP-shell in the course of the nominalisation. 2.2.2 Referential nouns As opposed to Greenbaum and Quirk’s (1990) and Barker’s (2008) account, referential nouns such as table do not inherit or do not idiosyncratically have an argument structure. As mentioned in the introductory part of section 2, the possessive morpheme and the possessed noun together assign theta-roles to the possessor. In other words, the theta-role of the possessor comes from the unit containing the possessive morphology in n and the possessed noun, as shown in (6). Szabolcsi (1994) argues that these theta-roles are arbitrary, and according to Barker (2008), they are specified in context. To put it differently, the exact relationship between the possessor and the possessed can only be deduced from the context. However, Greenbaum and Quirk (1990) classify these context dependent interpretations of possession (or using a traditional term ‘genitive constructions’) into four groups: (i) ‘possessive genitive’ expressing 17

18

Copulas are zerovalent verbs. Thus, in this sentence the adjective has an argument structure. Hungarian demonstrates the way a noun can inherit the argument structure of an adjective more straightforwardly because this language derives nouns from adjectives with the help of suffixes (-ság and –ség), consider (i). (i)

a.

b.

Az áldozat bátor volt. <experiencer> the victim-GEN courageous was ‘The victim was courageous.’ az áldozat bátorsága <experiencer> the victim-GEN courage-POSS-3.SG ‘the victim’s courage’

Katalin Horváth

Possessive Interpretations 11

ownership, see (10), (ii) ‘partitive genitive’ expressing inalienability, consider (11), (iii) ‘descriptive genitive’ (cf. (12)) and (iv) ‘genitive of origin’ (cf. (13)). Quirk et al (1989) add a fifth group called ‘appositive genitive’, in which the possessor is the possessed at the same time, see (14). (10) Peter’s house (11) a. Peter’s arm b. the ship’s funnel (12) Children’s shoes (13) a. the mother’s letter (the letter is from the mother) b. England’s cheeses (14) the city of Gödöllő So, Greenbaum and Quirk (1990) and Quirk et al (1989) list five types of possessive meanings associated with referential nouns. However, it is important to note that the possessive relation is unspecified (it is determined in context). Consequently, even referential nouns can be associated with the interpretational categories listed in the precious subsection. That is to say, for example, the context may specify the possessive relation in the example in (10) as subjective meaning that ‘Peter built the house’ in which case the possessor gets the theta-role <agent> from N and n. Further interpretational alternatives of the examples above are presented in the next section.

3 The thematic layers of the possessive DP 3.1 Reclassification of the possessive interpretations – n-types At this point there are 8 types of possessive interpretations: (i) subjective, (ii) objective, (iii) attributive, (iv) possessive, (v) partitive, (vi) descriptive, (vii) ‘genitive of origin’ and (viii) appositive. Thus, there should be 8 types of ns responsible for the derivation of DPs representing them. But some of the groups introduced in the previous two subsections can be collapsed into one. First of all, the ‘attributive genitive’ presented in (9) can be conceived of as a subgroup of the ‘subjective genitive’ because it is the subject of (9a) that bears the theta-role <experiencer> and it is inherited by the possessor in the derived DP in (9b). Analogously, constructions like (15b) can be conceived of as further representatives of the subjective interpretation. In sum, in the light The Even Yearbook 9 (2010), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061-490X, http://seas3.elte.hu/delg/publications/even, © 2010, Katalin Horváth

Katalin Horváth

Possessive Interpretations 12

of the examples in (7e), (8c) and (9b) the possessor can be interpreted as an agent or as an experiencer if n is subjective. (15) a. The girl is beautiful. b. the girl’s beauty (13a) provides an additional potential reading of subjective constructions if the interpretative restrictions provided in brackets are disregarded. In (13a) the letter is from the mother, i.e. she wrote it. Consequently, mother is an agent. Thus, (13a) could be conceived of as subjective. In addition, (13a) can also have several interpretations (not only ‘genitive of origin’): the mother <beneficiary> got the letter, she <experiencer> dreamt about it. All of these are subjective readings specified by the context. The problem of this kind of ambiguity is discussed later. (13b), the other example representing a ‘genitive of origin’, gives a description about the cheeses at issue, i.e. that they are English.19 In other words, the examples like (13b) can be reclassified as instances of the descriptive possessive relation or, using Sinclair’s (1993) term, as a ‘classifying’ interpretation. So there is no need for a separate group termed ‘genitive of origin’ because its members can be distributed among other ntypes. At this point it is important to note that there are cases when a ‘descriptive genitive’ does not have a possessive reading at all, see (16). In this paper structures like the one in (13b) (with a clear possessive semantic component) are termed ‘classifying’ (adapting Sinclair’s (1993) term), whereas constructions like the ones in (12) and (16) below are labelled ‘descriptive’. As the latter group and the appositive expression in (14) are assumed not to have anything to do with possession, they are not represented as nPs and, therefore, are not discussed in this essay further. There are cross-linguistic arguments for not assigning a possessive reading to ‘descriptive genitives’: for instance in Hungarian and in German this kind of meaning is not expressed with a “possessive-like” construction. Instead of a ‘descriptive genitive’ measure NPs (17), PPs (18) or compound nouns (19) are used. (16) a. a house of stone b. a cup of tea

19

(13b) can also be regarded as an instance of possessive genitive: England has famous cheeses.

Katalin Horváth

Possessive Interpretations 13

(17) egy csésze tea a cup tea ‘a cup of tea’ (18) ein Haus aus Stein a house from stone ‘a house of stone’ (19) a. egy kő- ház a stone house ‘a stone house’ b. gyerek- cipő child- shoe ‘children’s shoe’ The rest of the possessive interpretations, the possessive (from this point on referred to as ‘control’ in order to avoid confusion with the general umbrella term ‘possessive interpretation’)20 (10) and the partitive (11), cannot be united. The reason for this is that partitive constructions cannot be paraphrased as have-sentences, but the constructions containing a control n can, see (20)-(21). The other possessive constructions lose their original meaning in havesentences similarly to partitives, consider (22). That seems to support the assumption that these two possessive meanings (control and partitive) are semantically too different to be collapsed into one category. It is important to mention that the partitive interpretation may also comprise more types than it does in Greenbaum and Quirk (1990), but then it would have to be renamed. As mentioned above, the category of partitive expresses inalienability, see (11a). Besides, relational nouns such as daughter idiosyncratically imply some kind of inalienable relationship between two people, as every woman must be a daughter of someone.21 That is why it is better to label the category ‘partitive’ as ‘inalienable’. (20) a. Peter’s arm b. Peter has an (it is his distinguishing feature or he has arm. somebody else’s arm) (21) a. Peter’s house b. Peter has a house. 20 21

The term ‘control’ is taken from Barker (2008). Some other relational nouns include e.g. friend, boss, sister, teacher, student etc.

Katalin Horváth

Possessive Interpretations 14

(22) a. Children’s shoes (shoes designed for children) b. The children have shoes. (The children possess shoes.) Given this type of reasoning, there are 5 types of n in possessive DPs: subjective, objective, control, classifying and inalienable. All of these categories are homomorphous in English. In other words, all of them are realised as possessive null-morphemes in n attached to the possessed N, cf. the tree in (23c) below. Consequently, (10) may have several readings, too, because of the homomorphy of the affixes in n. In other words, the listener cannot decide which type of n is projected in the structure. So, apart from ‘Peter owns the house’ it can also mean that ‘Peter designed it’, ‘built it’, ‘lives in it’, ‘wants to buy it’, ‘bought it’ or ‘exploded it’. But all of these imply “subjectivity” on behalf of Peter (just like (13a) on behalf of mother). In such cases and in the case of descriptive constructions the specific meaning can only be deduced based on the context. To put it differently, n is responsible for the type of possessive meaning and the context specifies the theta-role to be assigned to the possessor by N and n. As Storto (2005) claims, “the semantics of the possessive relation is unspecified, leaving it to the context to determine directly the relation holding between possessor and possessum,” (Storto 2005: 83). In other words, he insists that possessive DPs have a semantic core (the nP-shell) “playing a role in licensing the availability of certain contextually determined interpretations,” (Storto 2005: 83). All in all, based on the argumentation above it can be concluded that there is no need for eight types of ns representing Greenbaum and Quirk’s (1990) and Quirk et al’s (1989) eight types of possessive interpretations in the possessive DP because it can be shown that they can be reclassified. First, the subjective and the attributive interpretation can be collapsed into one category labelled ‘subjective’. Second, the ‘partitive genitive’ can cover more meanings than in Greenbaum and Quirk (1990), so the new category proposed is named ‘inalienable’. Third, it was claimed that the ‘appositive genitive’ and some instances of the group ‘descriptive genitive’ have no possessive reading at all, and as such they do not involve a light noun projection in the DP. Finally, the ‘genitive of origin’ and the representatives of the ‘descriptive genitive’ with a possessive reading can be collapsed into one category: a ‘classifying interpretation’. As a consequence, there are five types of light nouns that can

Katalin Horváth

Possessive Interpretations 15

occur in possessive DPs: (i) subjective, (ii) objective, (iii) control, (iv) inalienable and (v) classifying.22 3.2 The nP-shell In the previous section it was concluded that there are 5 types of ns in the nPshell. The question is how many of them can occur in a DP at a time. As a first step, derived Ns will be examined. As it was already mentioned, derived Ns inherit the theta-grid of the verb they are derived from or the argument structure of the adjective occurring in clauses like (9a) or (15a). This suggests that all the arguments present in the vP-shell are present in the nP-shell of the noun derived from the bivalent verb, too. Taking (23) as an example it becomes clear that the two arguments of destroy can occur in the structure of one nP-shell at the same time if they are case marked by different case assigners, i.e. the prenominal argument receives genitive (from an agreement head), the postnominal one receives accusative from the inserted preposition of. Consequently, all kinds of nominalised bivalent verbs stay bivalent in the course of the derivation. In the case of non-derived nouns it is impossible to combine two kinds of ns within the same nP-shell, see (24). The reason for the incompatibility of ‘genitive types’ in constructions with a non-derived N is the fact that referential nouns do not inherit and do not idiosyncratically have an argument structure. In other words, they cannot be bivalent because they do not refer to an action expressed with the help of more than one argument. Hence, only one argument slot can be opened with the help of the possessive suffix hosted in n. All in all, control, inalienable and classifying ns cannot be combined with any kinds of n. Nevertheless, they can be combined in separate DPs, see (25). In this case multirelational possession is expressed. In other words, in (25) there are two possessed entities (house and mother) and two possessors (Peter and Peter’s mother). Thus, one of the possessor DPs is a possessive DP itself. (23) a. 22

[DP the enemy’s destruction of the city]

In the light of the argumentation above, Barker’s (2008) alternative way of arranging possessive meanings (presented in footnote 13) seems to have severe disadvantages. Barker’s (2008) categories of possessive relations are exclusively semantic in nature, as seen in (i) (in footnote 13) above. This results in abstract categories and the recognition of the role of context in determining the exact relationship between the possessor and the possessed. However, Barker (2008) does not address the distinction between subjective and objective interpretations of possession. In addition, Barker’s (2008) ‘agentive’ category is too specific to cover all the interpretations with which the subjective n is associated in this paper.

Katalin Horváth

Possessive Interpretations 16

b.*23

c.

(24) a. [nP Peter [NP house]] b. * [nP Peter [nP the enemy [NP house]]] c. * [nP Peter [nP England [NP house]]] (25) a. [DP[DP Peter’s mother]’s house] b. [DP das Haus [PP von Peters Mutter]] the-3.SG-NEUT house-SG of Peter-GEN Mutter-DAT ‘Peter’s mother’s house’

23

As proposed, as opposed to verbs, nouns alone (without ns) cannot assign theta-roles to their arguments even if they are derived from verbs. As a consequence, there is always a need for ns (for all arguments a separate one) in the possessive DP. That is why the theme the city cannot be in [Spec, NP]. ([Spec, NP] is assumed to be occupied by postdeterminers in Newson et al (2006).)

Katalin Horváth

Possessive Interpretations 17

c. [DP[DP Péter anyjának ] a háza] 24 Peter-GEN mother-POSS-3.SG-LD the house-POSS-3.SG ‘Peter’s mother’s house’ As for the co-occurrence restrictions on n types, it can be concluded that they can be freely combined and there is no limit on their number in the structure as long as they are accommodated in the same way as described in the previous paragraphs,25 see (26), where an inalienable (in the nation’s enemy), a control (in Caesar’s city), a subjective (in the enemy’s destruction) and an objective n (in destruction of the city) are combined. Limitations are set only by processing capacity. (26) [DP[DP the nation’s enemy]’s [DP destruction [PP of Caesar’s city]]]

4 Conclusion The aim of this paper is to argue that nouns in themselves are not theta-role assigning elements even though derived nominals inherit the argument structure of the verb (or adjective) they are derived from. Therefore, in possessive DPs the source of the possessive interpretation is the possessive morpheme and the possessed noun together. In the light of the CP-DP Parallelism and under the guidance of the UTAH it has been proposed that the possessive morpheme heads a thematic nP projection in the DP which is the extended projection of N. There are five types of n (subjective, objective, control, inalienable and classifying), which, together with N, assign theta roles to the possessors. The n heads are manifested by null-morphemes in English, so the context has an important role in the determination of the possessive relation. In other words, the system of the nP projections constitutes the semantic core of the unspecified possessive interpretation disambiguated in context. It has also been shown that the structure of the nP-shell is semantically and syntactically restricted. The semantic restriction lies in the difference between derived and non-derived Ns regarding their argument structure. In other words, the semantic restrictions determine which ns are compatible with each other. On the other hand, the availability of only two structural cases in 24 25

LD stands for “Left-Dislocation” marked by the –nak suffix in Hungarian possessive DPs. In the case of deverbal bivalent nouns at most two possessors (a prenominal and a postnominal one) can be accommodated in the same nP-shell, whereas in the case of derived nominals multirelational possession is expressed with the help of separate possessive DPs hosted in the specifier positions of each other (i.e. in these structures the possessor can be a possessive DP itself too).

Katalin Horváth

Possessive Interpretations 18

the English possessive DP (prenominal genitive from the agreement head and postnominal accusative from the inserted preposition of) determines the number of the available structural positions for the possessors within the possessive DP. To put it differently, in the case of deverbal (bivalent) Ns at most two possessors (a prenominal and a postnominal one with different theta roles) can be accommodated in the same nP-shell. On the contrary, in DPs with non-derived possessed nouns only one possessor can be accommodated in one nP-shell. Thus, in this case the only way to combine n heads is to host them in separate nP-shells in separate DPs exhibiting multirelational possessive constructions. References Baker, Mark. 1988. Incorporation. Chicago: Chicago University Press. Barker, Chris. 2008. Possessives and relational nouns. In: Claudia Maienborn, Klaus von Heusinnger and Paul Portner (eds.). Semantics: An International Handbook of Natural Language Meaning. Berlin. de Gruyter. Retrieved from http://semanticsarchive.net/Archive/WYxOTc5M/barker-possessives.pdf (last accessed: October 22, 2009). Greenbaum, Sidney and Randolph Quirk. 1990. A Student’s Grammar of the English Language. Harlow: Longman. Newson, Mark et al. 2006. Basic English Syntax with Exercises. Budapest: Bölcsész Konzorcium. Olsen, Susan. 1989. Das Possessivum: Pronomen, Determinans oder Adjektiv? Linguistische Berichte 120: 133-153. Quirk, Randolph et al. 1989. A Grammar of Contemporary English (18th impression). Harlow: Longman Group Ltd. Roehrs, Dorian. 2005. Possessives Consist of Heads and Complements: Some Notes. Indiana University Working Papers in Linguistics 5. Retrieved from http://ling.auf.net/lingBuzz/000302 (last accessed: October 22, 2009). Sinclair, John. 1993. English Usage. London: Collins Cobuild. Storto, Gianluca. 2005. Possessives in Context. In: Ji-yung Kim, Yury A. Lander and Barbara H. Partee (eds.). Possessives and Beyond: Semantics and Syntax. Amherst: GLSA, University of Massachusetts. 59-86. Retrieved from http://semanticsarchive.net/Archive/mYOMTMwN/storto_UMass02.pdf (last accessed: October 22, 2009). Szabolcsi, Anna. 1994. The Noun Phrase. In: Ferenc Kiefer and Katalin É. Kiss (eds.). Syntax and Semantics 27. The Syntactic Structure of Hungarian. San Diego: Academic Press. 179-231.

Katalin Horváth Department of English Linguistics, Eötvös Loránd University, Budapest horvath.katalin1986@indamail.hu