The semantics of prenominal possessives in Russian

Traditionally Russian prenominal possessives are classified as possessive adjectives (Isa č enko 1960; Townsend 1980). In more recent work by Babyonyshev (1997) it is claimed that prenominal possessives are functional elements that are hosted outside nominal phrase, which means that they cannot be adjectives. This claim is based on the ability of possessives to fill an agent argument of event nominals. In this paper I will provide evidence that syntactically prenominal possessives do behave like adjectives. I will propose a semantic analysis that will account for the peculiar properties that prenominal possessives have.


INTRODUCTION
This paper discusses the syntax and semantics of prenominal possessives in Russian.I will argue that syntactically expressions in bold in (1) are adjectives, despite claims to the contrary in Babyonyshev (1997), and I will offer a semantic analysis that will account for their grammatical behavior.
Like other adjectives in Russian, prenominal possessives precede the head noun and agree with it in number, gender and case.
Sosedk-in-a sobak-a zalajala.neighbor-poss-f.sg.nom dog-f.sg.nom barked 'The neighbor's dog barked.' As is well-known, possession is not the only type of relation that can hold between the individuals that are usually labelled as "the possessor" and "the possessum": (3) a. pet-in portret petya-poss.m.sg.nom portrait.m.sg.nom 'Petya's portrait' b.
pet-in-a mam-a petya-poss-f.sg.nom mother-f.sg.nom 'Petya's mother' c. petin nos petya-poss.m.sg.nom nose.m.sg.nom 'Petya's nose' In (3a) Petya can be the possessor of the portrait or the person portrayed, in (3b) petin is an argument of the 'mother ' relation, in (3c) the nose is a part of Petya's body -the possessive expresses a part-whole relation.Furthermore, in event nominals, possessive adjectives are able to fill an agent argument: (4) mam-in-o postojannoje vyraženije nedovol'tsva mom-poss-n.sg.nom constant.n.nom expression.n.nom displeasure.n.gen 'Mom's constant expression of displeasure' While petin 'Petya's' and mamin 'mom's' are classically analyzed as possessive adjectives, Babyonyshev (1997) presents an alternative explanation for why these possessives can fill an agent argument for event nominals.Babyonyshev claims that prenominal possessives are 1 The suffix -ov is no longer productive in Modern Russian (Townsend 1980; Koptjevskaya-Tamm and Shmelev 1994; Trugman 2007).Thus, in this paper I concentrate on the suffix -in.However, the analysis developed here extends to -ov as well.

PERMUTATIONS OF POSSESSIVES AND ADJECTIVES
Possessives can permute with other adjectives both when they occur attributively (6) and as predicates (7).( 6) a.
Novaja mam-in-a rabota svjazana s putešestvijami.new mom-poss-f.sgjob is tied with travelling 'Mom's new job involves travelling.' 2 In the rest of the paper I use, following Babyonyshev (1997), the term "prenominal possessives" as interchangeably with possessive adjective (mainly because this is accepted in the literature).The term comes from the distinction in English between the prenominal possessive -the Saxon genitive, and the postnominal genitive construction with the preposition of.
However, it is worth noting that this term does not seem to be precise enough, because on the one hand, a possessive relation is only one type of relation that is associated with this grammatical form, and on the other hand, these adjectives can occur in predicative position as well.

3
An anonymous reviewer brought to my attention the following examples: (i). so beautiful a book; (ii).how beautiful a book; (iii).that beautiful a book.
Bresnan (1973) argued that in these cases the AP is syntactically higher than the determiner.Cresswell (1976) claimed that due to compositionality requirements these examples should be assigned a more traditional structure with the determiner higher and the adjectival phrase originating within the nominal phrase.Gepner Glossa: a journal of general linguistics DOI: 10.5334/g jgl.1198 (7) a.
Vse gosti na etoj večerinke byli byvšimi pet-in-ymi odnoklassnikami.all guests on this party were former Petya-poss-pl.insclassmate.pl.ins 'The guests at the party were Petya's former classmates.' The examples in ( 6) and ( 7) are of special importance because they provide evidence that possessives in Russian and English are typologically different.English does not allow permutations, neither in attributive nor in predicate position.This kind of grammatical behavior is expected if we assume that possessives in English are part of DP, not NP (see Landman 2003 for the discussion of left-periphery effects in DPs in predicate position).In contrast, in Russian possessives can mingle with adjectives both when they occur attributively and as predicates.

POSSESSIVES AS ARGUMENTS OF QUANTIFIERS
Nouns combined with prenominal possessives can be arguments of quantifiers (každyj 'every').This pattern is typical of adjectives, not determiners.
(8) Bolnyje v palate lovili každyj mam-in žest.sick in room caught every.m.sg mom-poss.m.sg gesture 'The sick people in the room waited for every mom's gesture'.
The noun phrase in ( 8) is interpreted as 'every gesture of my mother', not 'the gesture of every mother'.

PERMUTATIONS WITH NUMERALS
Moreover, possessives can permute with numerals.
Pap-in-y dva velosipeda stojali na balkone.Dad-poss-pl.nomtwo bicycles stood on balcony 'Dad's two bicycles were on the balcony.' The data in (9) are problematic for the assumption that possessives are determiners.Numerals are generally considered to be hosted outside NP.If possessives were determiners, they should not be able to permute with numerals similarly to každyj 'every'.
(10) a. Každyje tri goda oni pokupajut novuju mašinu.every three years they buy new car 'Every three years they buy a new car.' b. #tri každyje goda three every year However, Landman (2003;2004) convincingly shows for English that numerals are better analyzed as adjectives that denote cardinal properties of plural individuals.Landman (2003) claims that they are adjectival predicates that originate within NP and raise to the DP area only in the absence of a lexical determiner.Khrizman (2016) shows that this analysis holds for Russian numerals as well.If both possessives and numerals are adjectives, then the data in (9) are explained: adjectives can permute.
Thus, with respect to their syntactic position, prenominal possessives pattern with adjectives.
If they are adjectives, then they originate inside the nominal phrase.Thus, the semantic type of the expression [possessive+noun] is naturally <e,t>.As a consequence, nouns modified by prenominal possessives can occur in predicative positions without undergoing any type-shifts.
In the next subsections I will provide evidence in support of this claim.

UNDER THE SCOPE OF A NEGATIVE OPERATOR
Prenominal possessives can occur in genitive case under the scope of negation.It is well known that in Russian verbs under negation can take arguments in accusative or genitive case (Timberlake 1975;Babby 1980;Neidle 1982).Genitive case is usually associated with a decrease in referentiality.Genitive NPs get non-specific/indefinite interpretation, while accusative NPs tend to be interpreted as specific/definite.Partee and Borschev (2004) and Kagan (2005;2007;2013) account for the semantic contrast by arguing that NPs in genitive case are predicative expressions at type <e,t>, while accusative NPs are arguments at type e.
The natural interpretation for (11a) is 'there is a specific ring that he does not wear'.This contrasts with (11b) that implies that he is not married.
On ne nosit obručal'noje kol'c-o. he not wear wedding.n.sg.acc ring-n.sg.acc 'He does not wear the/his wedding ring.' b.
On ne nosit obručal'nogo kol'c-a. he not wear wedding.n.sg.gen ring-n.sg.gen 'He does not wear a wedding ring.' Nouns modified by prenominal possessives are felicitous in this position.
Ja ne slušala mam-in-yx sovet-ov.I not listen mom-poss-pl.genadvice-pl.gen'I did not listen to my mother's advice.'In (12a) maminy sovety is in the accusative and gets a specific interpretation at the argument type.It means 'the pieces of advice that my mother gave me'.The Genitive NP in (12b) gets a non-specific interpretation, the sentence roughly means 'I did not listen to any pieces of advice that my mother gave me', as predicted by Partee (2008) and others.This strongly suggests that maminy sovety is an expression of type <e,t>, since, as Partee shows, the non-specific interpretation follows from the fact that the genitive is a predicative NP.
If [possessive+noun] were a DP that underwent a type-shift from <<e,t>,t> to <e,t>, then the prediction would be that any quantificational DP can undergo the same kind of shift and occur in genitive case under the scope of a negative operator. 4  However, this is not the case in Russian.Každyj 'every', shown to be a quantifier in Gepner (to appear), is infelicitous in this position.
(13) #ja ne slušala každ-ogo sovet-a I not listen every-m.sg.gen advice-m.sg.genIntended: 'I did not listen to every piece of advice' The data discussed in this subsection provide evidence that nouns modified by prenominal possessives can occur in genitive of negation because the expression [possessive+noun] is born as an expression of type <e,t>.The ability to occur in this position does not result from a semantic shift from argument type <<e,t>,t>.
At this stage the possibility of the shift from e to <e,t> for [possessive+noun] has not yet been excluded.In the next subsection I will show that this is not a plausible analysis either.2.5 UNDER THE SCOPE OF MEASURE PREFIXES Filip (2005) claims that only expressions of type <e,t> can occur under the scope of measure prefixes.In this subsection I will show that possessives can modify nouns that are complements of measure operators.I will argue that the expression [prenominal possessive+noun] is born at type <e,t> and, thus can occur in this position.
In principle there could be a different explanation of why nouns modified by possessives are grammatical in this predicative position.Following Babyonyshev (1997) one could assume that prenominal possessives are determiners, thus [possessive+noun] is an expression of type e.It is well-known that expressions of type e can shift to <e,t> interpretation.Therefore, [possessive+noun] could shift to a predicative interpretation and occur under the scope of measure prefixes.However, I will show that this is not a plausible analysis.It is not the case that possessives head DPs at type e and can occur under the scope of measure prefixes because they shift to <e,t> interpretation.Filip (2005) analyzes the prefixes na-and po-in Russian as measure phrases and claims that their nominal arguments are predicative nominal phrases with non-specific indefinite interpretation.na-and po-first combine with a property-denoting nominal argument (of type <e,t>) and only after this grammatical operation the expression is able to combine with a verbal root.
(14) Filip (2005) On kak-to varenya na-varil iz čerešni žut' kak he somehow jam.sg.gen na-cook.pst.3.sg from cherry.sg.gen horror how mnogo: desjat' veder.much: ten bucket.pl.gen 'He made/cooked up a (relatively) large quantity of jam -from cherries -boy, did he make a lot of it: ten buckets!' Jam is marked genitive in ( 14) and has an indefinite non-specific interpretation.na-incorporates a measure function that contributes to an implication that the quantity of jam that was cooked was large.
In (15) po-first combines with the predicate brandy identifying quantities of brandy that are more than null, but small, and only then it is attached to a verbal root pit' 'drink'.
In (16a) natašinyx pirogov is a NP at type <e,t> which combines with a measure prefix na-to form a measure predicate that will then combine with a verb base.The same holds for (16b)maminyx is part of the nominal phrase headed by the predicate kotlet 'chops'.
On s udovolstvijem po-jel mam-in-yx kotlet.he with pleasure po-eat mom-poss-pl.genchops.pl.gen 'He ate some of mom's chops with pleasure.'Now the crucial observation is that proper names cannot be arguments of measure prefixes that require their arguments to be of type <e,t>.

DE RE AND DE DICTO AMBIGUITY IN INTENSIONAL CONTEXT
It is known that indefinites in English can have narrow or wide scope interpretation in intensional context.
(18) Peter wants to find a French woman.
When 'a French woman' in ( 18) has a de re interpretation, it is naturally followed by (19a) -there is a specific woman he wants to find.When 'a French woman' has a de dicto interpretation, (19b) is an acceptable continuation.
He met her at a party.b.
He has not met any French woman yet, anyone who is both French and female will suit him The same distinction holds for definite DPs in English: (20) Peter wants to find the department chair.
(20) also allows for a de dicto and a de re interpretation.
The literature following Longobardi (1994) links de re interpretations to referentiality and referentiality to DPs.Thus, it is well known that proper names only allow de re interpretations, as in ( 21): (21) Peter wants to find June.
Longobardi proposed that proper names undergo N to D raising, and end up in the D position, where they are interpreted as referential (de re).Following this strategy, the lack of a de dicto reading for examples like ( 21) can be attributed to the lack of analysis where the proper name stays an NP.
Following this, one can assume that the de dicto-de re ambiguity for the definites is a question of whether the definite is interpreted as an NP at the predicate type (<e,t>), or whether the definite article also raises to D, and the definite gets a referential interpretation like the proper name.
This idea forms the background for the analysis of Babyonyshev (1997).Babyonyshev observes that in intensional context prenominal possessives are always associated with a specific possessor (Babyonyshev 1997).
(22) Petr xočet najti sosedk-in-u podrugu.Peter wants to find neighbor-poss-f.sg.acc friend.f.sg.acc 'Peter wants to find the neighbor's friend.'Gepner Glossa: a journal of general linguistics DOI: 10.5334/g jgl.1198 In ( 22) there is a specific neighbor that the speaker refers to.Babyonyshev (1997) uses this fact as an argument in support of the claim that prenominal possessives are determiners -the possessor is specific, consequently, sosedkin 'neighbor's' is part of a DP, not NP.The idea, then, is that prenominal possessives only have de re interpretations because the possessive is in D position.
However, looking at the Russian facts in more detail shows that Babyonyshev's identification of de re readings with referentiality is problematic; in Russian the expression [possessive+noun] shows de dicto-de re ambiguities despite the fact that the prenominal possessive can only have a de re interpretation.
Let us imagine a situation: Peter is a young man who likes dreaming and hates working.Peter often thinks about how different his life would be if he married a rich woman.One day Peter finds out that a very rich single woman bought a house in the neighborhood.Peter was told that this woman only meets new people if they are introduced to her by someone she already knows.Peter does not know yet whether the woman has any friends in the neighborhood.However, we can felicitously say the sentence in ( 22): he wants to find someone who has the property of being the neighbor's friend so that she could introduce him to his new neighbor.
Later on, it turns out that the neighbor has three friends in the neighborhood and Peter goes to the bakery with one of them.In the updated context ( 22) is still a felicitous sentence, however, the interpretation is different -Peter has a specific friend in mind, but we do not know which one.
Thus, ( 22) is ambiguous (analogously to indefinite nominal phrases in English): either there is a specific individual who Peter wants to find (transparent reading) or Peter does not care as long as the person fits the description 'the neighbor's friend' if there is one (opaque interpretation).
We observe that, like indefinites and definites in English, Russian prenominal possessives show de dicto-de re ambiguities, and this is independent from the referentiality that Babyonyshev observes (and that the analysis will have to deal with).This means that there isn't really an argument for linking de re to the N to D analysis, in particular, since it would have to involve a lowering type shift for indefinites (from type <e,t> to type e), which one would want to avoid if one can do without.
However, if prenominal possessives are assumed to be adjectives, then the lifting in argument position for [possessive+noun] is a standard operation as predicted by the Upward Partee Triangle (Landman 2004): an expression of type <e,t> shifts to <<e,t>,t>.The details of the derivation will be provided in Section 5.
In the next section I will discuss literature on possessives and claim that English-like analyses of possessives are not appropriate for Russian and a new approach is needed.

BACKGROUND ON POSSESSIVES
Many linguists have worked on the semantics and syntax of possessives (e.g.Hellan 1980;Partee 1983Partee /1997;;Jensen and Vikner 1994;2004;Partee and Borschev 1998;1999;Peters and Westerståhl 2013a;b).Most research has been done on English.Let us first look at the data and then discuss the questions that the data raise and what approaches there exist to analyzing possessives in English.
There are three main components that play an important role in constructing and interpreting a possessive: two individuals -the possessor and the possessum -and a relation that holds between the two (Barker 2011).
In John's brother John is the possessor, the possessum is some male individual, and the relation that holds between the two is a brother of relation.In this case the relation is part of the lexical semantics of the head noun brother.However, in John's book the relation that holds between John and the book largely depends on the context and is not encoded in the semantics of the head noun.John can be the author of the book, its possessor, his research project can be about this book, and there are many other possibilities.Thus, the type of relation in a possessive construction depends on the properties of the head noun, to be more precise, the type of relation depends on whether the head noun is sortal like book or relational like brother.If there are two types of head nouns, does this mean that there should be two different possessives?If we propose a unified analysis for the possessive, how are we going to deal with the fact that the same possessive construction should combine with two different types of nouns?Where does the relation come from?These are the questions that the researchers try to answer.
There are different strategies discussed in the literature aiming at providing an analysis of possessives.Partee and Borschev (2003) argue against a unified analysis of genitives and related constructions claiming that many genitives have mixed properties both of arguments and modifiers.Jensen and Vikner (1994;2004) and Hellan (1980) propose a uniform analysis for possessives.
Jensen and Vikner argue that possessives are uniformly arguments, while Hellan suggests that they are uniformly modifiers.These analyses differ in positing different shifting rules -whether sortal nouns shift to a relational interpretation or whether, vice versa, relational nouns get detransitivized.
If the possessive can combine only with relations, then sortal nouns have to undergo a semantic shift to get a relational interpretation.In this case the possessive (Mary's) is uniformly an argument of the relation denoted by the head noun.The second possibility is to assume that the possessive is a modifier of the head noun applying only to predicates.In that case, relational nouns must be detransitivized.
In this section I will discuss these two possibilities in detail and explain why neither of them seems to be adequate for prenominal possessives in Russian.

MARY'S AS AN ARGUMENT
This approach is advocated in Jensen and Vikner (1994;2004).They assume that in genitive constructions all the head nouns denote a relation and the noun in the genitive saturates an argument of this relation.Sortal nouns are coerced into a relational interpretation.The semantic shift from a sortal to relational interpretation is licensed by the lexical semantics of the noun.
Jensen and Vikner ( 2004) follow Pustejovsky's (1995) theory of lexical structure.According to this theory there exist four levels of lexical representation: argument structure, event structure, qualia structure and lexical inheritance structure.What we are interested in is the qualia structure because it underlies the lexical mechanism that makes it possible for sortal nouns to shift to a relational interpretation.The qualia structure represents the typical characteristics of a real-world object denoted by a lexical item.These characteristics are defined by the way human beings interact with the world around them and originate from Aristotle's "modes of explanation".People use things, interact with things, produce things and, in general, do things (Jensen and Vikner 2004).This mode of conceptualization of the world is reflected in the four qualia roles that describe different aspects of lexical meaning of words.
Qualia roles: Formal -being an instance of, distinguishing an object within a larger domain; Constitutive -being a part of, the relation between the object and its parts; Telic -being used in/for, the function of the object; Agentive -being the result of, how the objects come into being.
From qualia roles, Jensen and Vikner (2004) derive four possibilities for the lexical interpretation of a genitive relation: inherent (e.g. a mother), part-whole (e.g. a nose as part of somebody's body), agentive (e.g. a poem that was written by someone) and control (e.g. a house that is built/owned/designed by somebody).
There is one more possible type of interpretation -pragmatic.In this case the relation is provided by the context (this is analogous to "free R/weak possession relation" discussed in Partee 1983Partee /1997)).Partee and Borschev (1999) extend Jensen and Vikner's (1994) analysis to postnominal genitive construction in Russian (e.g.mama Peti 'mother of Petya', kniga mamy 'my mother's book').
Genitives are arguments of the relation denoted by the head noun.Any sortal noun has to undergo a meaning shift to a relational interpretation.The meaning shift is allowed by the lexical semantics of the noun.
Partee and Borschev (1998) distinguish between three "sorts" of nouns that can take postnominal genitives: 1. inherently relational (e.g.brother); 2. relativized -they do not express relations directly, but "it is typical for these nouns to stand in some relation to other objects/ individuals" (e.g.portrait); 3. non-relational -only strong contextual support can make the genitive construction grammatical (ex.nebo Andreja Bolkonskogo 'the sky of Andrey Bolkonsky').
Jensen and Vikner's analysis cannot be straightforwardly extended to prenominal possessives in Russian.Despite the fact that they seem to be interacting with the argument structure of the head noun (this will be discussed in detail in the next section), syntactically they are adjectives, i.e. they originate within NP.Thus combining a prenominal possessive with a nominal results in a predicate at type <e,t>.

MARY'S AS A MODIFIER
The second possibility discussed in the literature is that a noun in genitive is a modifier that applies to predicates.The possessive modifier naturally combines with sortal nouns.The relation is provided by the context ("pragmatically controlled", see Barker 2011).
(25) Mary's car: To be modified by a possessive, relational nouns must undergo a semantic transformation into a sortal noun.Barker (2011) discusses the detransitivization type-shifter Ex = λRλx.∃y[R(x,y)]that binds one of the arguments and allows for non-relational usages of relational nouns.
The availability of this type-shifting operation allows for a uniform analysis of the possessive as a modifier that combines only with sortal nominals, the relation being pragmatically supplied.
Brother shifts to brother of someone via the type-shifter Ex.Then it can combine with the possessive Mary's.However, if there is a contextually supplied relation, e.g.writing an article about the Kennedy brothers, then the detransitivized brother undergoes an additional meaning shift to a relational interpretation with the relation pragmatically supplied.In this case it is interpreted as 'the brother of somebody who stands in R relation to Mary' and R='is writing an article about'.Partee and Borschev (2003) claim that prenominal possessives in Russian should be analyzed as modifiers, while postnominal genitive NPs are arguments of a relation denoted by the head noun.
However, it seems to be the case that detransitivized usages of relational nouns and relational nouns occurring with arguments (i.e. in the relational interpretation) are both common in the possessive context in Russian.
The interpretation of prenominal possessives depends on the argument structure of the noun it combines with.When prenominal possessives combine with relational nouns (i.e. a relation is provided by the lexical semantics of a noun), the pure possessive interpretation (ownership) is not available.If a prenominal possessive applies to a sortal noun, the most natural interpretation is possessive.Thus, prenominal possessives in Russian can apply both to non-detransitivized relational nouns and sortal nouns.Consequently, the English-like analysis in which possessives are modifiers of <e,t> type nominal expressions is not applicable in Russian.
Let us look at the noun portret 'portrait' that is ambiguous between three different interpretations: sortal (27a), relational with two arguments (27b) and relational with three arguments (27c).
( Prenominal possessives can combine with portret 'portrait' in all the three interpretations.The meaning of the prenominal possessive will depend on how many arguments there are and how many of them are explicitly expressed by the postnominal genitive.
Petin portret 'Petya's portrait' is ambiguous between Petya being the theme of the portrait and the possessor.
Mne ne nravitsja etot pet-in portret.Petya na sebja me not like this Petya-poss-m.sgportrait.m.sg Petya on himself sovsem ne poxož.absolutely not similar 'I do not like this portrait of Petya.He does not look like himself in it.'b.
Pet-in-y portrety mne nravjatsja bolše drugix jego kartin.Petya-poss-pl.nom portrait.pl.nom me like more others his pictures 'I like Petya's portraits more than his other pictures.' In (28a) Petya is the person in the picture, an argument of the relation between the picture and its theme.In (28b) the portrait is not associated with the objects in the picture, it simply names Gepner Glossa: a journal of general linguistics DOI: 10.5334/g jgl.1198 the kind of the picture, not landscapes or still life.Petya in (30b) can be interpreted as the owner of the pictures or as the author, i.e. he is in some way in control of the portrait.5 Petin portret Petrova 'Petya's portrait of Petrov' is ambiguous between Petya being the author or the possessor of the portrait (assuming that Petrov denotes the person in the picture).
It is worth noting here that being the author and being the possessor are different semantic relations.This is reflected in the fact that a prenominal possessive has to be interpreted differently depending on the meaning of the postnominal genitive, which in its turn is affected by the argument structure of the head noun.
If we force the author interpretation on the postnominal genitive, (e.g.we replace Petrov by a famous painter Vasnetsov) then the prenominal possessive must be interpreted as the person in the picture -as in ( 29).
(29) Aljonušk-in portret Vasnecova prozvodit silnoje vpečatlenije.Alyonushka-poss.m.sg portrait.m.sg Vasnetsov.genproduces strong impression 'Alyonushka's portrait by Vasnetsov is very impressive.'However, if we create a context (as in 30) in which Alyonushka is the possessor of the picture, e.g.she owns a huge collection of portraits from different ages, then Vasnetsov in (29) must be the theme of the portrait, and the author interpretation is unavailable.
(30) U mojej sestry Aljonuški ogromnaja kollekcija portretov.at my sister Alyonushka huge collection portraits 'My sister Alyonushka possesses a huge collection of portraits.'When the head noun is followed by two postnominal genitives, one denoting the theme and the other -the author, the only possible interpretation for the prenominal possessive is being the possessor of the picture.This holds even if we use famous portrait and artist's names.Vasnetsov's portrait of Alyonushka is well known -Alyonushka is a girl in the picture.Despite this fact, in (31) Alyonushka is understood as the possessor, not as the theme.
(31) Aljonuškin portret Ivanuški xudožnika Vasnecova Alyonushka-poss.m.sg portrait.m.sg Ivanushka.genpainter.genVasnecov.gen'Alyonushka's portrait of Ivanushka by the painter Vasnetsov' It seems to be the case that the interpretation of prenominal possessives in Russian depends on the argument structure of the noun it combines with.Consequently, we cannot assume that possessives are modifiers that apply only to sortal nouns and that as a result, relational nouns have to undergo detransitivzation.
We conclude that applying the analyses that have been given for the English Saxon genitive (either as an argument or as a modifier) do not seem to be on the right track for Russian prenominal possessives.A proper analysis should incorporate two important features of prenominal possessives: morphologically they are adjectives, and semantically they are able to interact with the argument structure of the modified noun.We now provide such an analysis.

THE BASIC MECHANISM
-in/-ov is an operator that applies to an individual and results in a function that, in its turn, applies to a relation and gives a set of individuals.In other words, possessive adjectives modify relations via saturating one of the arguments of this relation.
(32) shows the derivation of petin 'of Petya'.The possessive morpheme -in attaches to the proper name Petya.
-in/-ov can combine with predicates like neighbor, mother or actress: sosedkin 'of the neighbor', mamin 'of the mother', aktrisin 'of the actress'.
However, quantificational noun phrases cannot participate in the formation of a possessive adjective.In English "every candidate's dissertation" is perfectly felicitous.The Russian translation of this expression is impossible with a possessive adjective, -in/-ov can attach only to grammatical expressions that denote individuals, a plurality of individuals (even on a collective, i.e. singular interpretation) cannot be the possessor as is shown in ( 33).For cases when the possessive suffix combines with a predicate (like in sosedkin 'of the neighbor', aktrisina 'of the actress') we will assume, following Winter (1997), that a choice function picks out a specific individual from a set -f(ACTRESS a ).The possessor is always a specific individual salient in the discourse (Koptjevskaya-Tamm and Shmelev 1994;Babyonyshev 1997), thus the possessor is always anchored in the real world. 6  I assume that w is a variable of type s, and in ACTRESS w , relative to assignment g, w stands for the index of evaluation.Out of blue the evaluation assignment function g maps variable w onto the real world w 0 .
Thus, relative to this assignment, ACTRESS w will denote the set of actresses in w 0 .
I assume that a (for actual) is a name of type s, a is the name of the real world w 0 . ( Thus, ACTRESS a will denote the set of actresses in w 0 independently of assignment functions.Since w is a variable, it can be abstracted over. Note the difference: Even if g(w) = w 0 , λw.ACTRESS w denotes the function that maps every world v onto the set of actresses in that world v. λw.ACTRESS a , on the other hand, denotes the function that maps every world v onto the set of actresses in the real world w 0 .
(37) a. petina mama 'the mother of Petya': λRλx.R(x,PETYA) (λyλx.MOTHER w (x,y)) λx.MOTHER w (x,PETYA) b. mamina podruga 'a friend of my mother': 7 λRλx.R(x,f(λx.MOTHER a (x,me))) (λyλx.FRIEND w (x,y) λx.FRIEND w (x,f(λx.MOTHER a (x,me))) 6 I use a choice function to derive an individual from a predicate as this operation does not seem to have impact on the theory of definiteness in Russian (as little as it is known about how definiteness works in Russian).Our analysis can be compatible with the existence of a null definite determiner.

7
We assume, of course, that MOTHER w is a relation which is a function.Possessives can modify sortal nouns like in petin telefon 'the/a telephone of Petya' or aktrisino platje 'a/the dress of the actress'.With Jensen and Vikner (1994;2004), Partee and Borschev (1998;1999) I assume that sortal nouns can undergo a semantic shift to a relational interpretation.Whenever a sortal noun is modified by a prenominal possessive, it shifts to a relational interpretation.The shifting rule is λPλxλy.P(x) ∧ of w (x,y).After applying this rule to dress, we get the relational interpretation, i.e. 'dress of', the content of of-relation being contextually provided. 8 (38) λPλxλy.P(x) ∧ of w (x,y) (λx.DRESS w (x)) λxλy.DRESS w (x) ∧ of w (x,y) After the shift a relational noun can combine with the possessive aktrisino 'of the actress'.
(39) aktrisino platje 'a dress of the actress': λRλx.R(x,f(ACTRESS a )) (λxλy.DRESS w (x) ∧ of w (x,y)) λx.DRESS w (x) ∧ of w (x,f(ACTRESS a )) We will assume one more semantic shift that will turn an adjective that modifies sortal nouns into an adjective that can modify relational nouns.This shift is used in the analysis to account for the cases when possessives permute with property adjectives.
Most adjectives are predicates of type <e,t>: they denote a property as in (40a).In attributive position they standardly shift to a modificational type <<e,t>,<e,t>> as in (40b).
This car is black.Black: λx.BLACK w (x) b.This is a black car.Black: λPλx.P(x) ∧ BLACK w (x) There are adjectives that are inherently relational -they always result in a relational noun phrase.Partee and Borschev (1999) claim that favorite "obligatorily produces a relational output".This adjective can modify both sortal and relational nouns, thus, they postulate two types of favorite -favorite 1 combines with sortal nouns (41a), and favorite 2 -with relational nouns (41b).Favorite 2 is derived from favorite 1 .
Let LB w be the relation that holds in w between x and y and P if x is in P and y likes x best of all the elements in P.
(41) a. favorite 1 λPλyλx.P(x) ∧ LB w (y,x,P) favorite 1 book λyλx.BOOK w (x) ∧ LB w (y,x,BOOK w ) b. favorite 2 λRλyλx.R(x,y) ∧ favorite 1 (x,y,R(y)) favorite 2 sister λyλx.SISTER w (x,y) ∧ favorite 1 (x,y,SISTER w (y)) λyλx.SISTER w (x,y) ∧ LB w (y,x,SISTER w (y)) 8 It seems to be the case that possessive adjectives with -in/-ov suffixes are not the only adjectives in Russian that interact with the argument structure of the modified noun.Adjectives in -ovskiy (Vendlerovskiy 'Vendlerian', Oruellovskoje 'Orwell's', Rassellovskoye 'Russell's' etc.) are classified as possessive adjectives together with prenominal possessives by Paducheva (2000).Indeed, both Petin 'Petya's' and Vendlerovskiy 'Vendlerian' are associated with individual possession in contrast to relational-possessive adjectives like materinskiy 'of the kind that mothers experience/have' that express a property as being typical of a class of individuals.Both adjectives naturally modify event nominals saturating an agent argument.Russell-ovsk-f.sginterpretation descriptions 'Russell's interpretation of descriptions' However, relational nouns like mother or friend cannot be modified by -ovskiy adjectives.
(ii).#Vedler-ovsk-aja mama Vendler-ovsk-f.sgmother Intended: 'Vendler's mother' A more detailed study of the data is needed.We will leave these issues for further research.Gepner Glossa: a journal of general linguistics DOI: 10.5334/g jgl.1198 For cases when a property adjective combines with a relational noun (e.g.young sister) we will assume the shifting rule given in (42) that will change the type of the adjective from <<e,t>,<e,t>> to <<e,<e,t>>,<e,<e,t>>>.This rule shifts the interpretation of an adjective like young.
(42) Relational shift operation: λZλRλy.Z(λz.R(z,y)) λZλRλy.Z(λz.R(z,y)) (λPλx.P(x) ∧ YOUNG w (x)) = λRλy.(λPλx.P(x) This shift is available in the grammar independently of our analysis.We are going to use it for cases when a relational noun is first modified by a property adjective and then by a prenominal possessive, as in: natašin glupyj muž 'Natasha's silly husband'.
According to compositionality principles, muž first combines with glupyj, and then glupyj muž is combined with the possessive.Thus, glupyj 'silly' undergoes the same semantic shift from a property adjective to a relational adjective, the result of which is shown in ( 43).
(43) λZλRλy.Z(λz.R(z,y)) (λPλx.P(x) After the shift glupyj 'silly' can modify a relational noun muž 'husband ' -(44).natašin 'of Natasha' applies to glupyj muž 'silly husband' to give a set of individuals who have the property of being Natasha's silly husband -( 45).In principle there could be one more possibility to solve the type mismatch in examples like natašin glupyj muž 'Natasha's silly husband' that would not require any shifts in the adjectival interpretation.The relational noun husband could be detransitivized via existentially binding one of its arguments, λx.∃y [HUSBAND(x,y)].This operation would give a sortal noun of type <e,t> that can be modified by the <<e,t>,<e,t>> adjective silly.
However, here silly husband of type <e,t> would have to undergo another shift back to a relational interpretation to be able to combine with the prenominal possessive Natasha's.Thus, this approach is more costly -it requires more shifts.Landman (2003) convincingly argues that semantic shifts 'down' from type <<e,t>,t> into <e,t> or e, and from <e,t> into e are not natural.
(46) ?A sister is walking in the street.
Without a strong contextual support ( 46) is infelicitous.The reason is that sister is relational, and the grammar does not allow to shift it to a sortal interpretation.However, the mismatch can be resolved contextually, by finding an appropriate value for the argument that is still open in the discourse: (47) I know that John's brothers are all inside, but I see a sister walking down the street.
In sum: on the analysis we assume, prenominal possessives are adjectives that modify relations via saturating one of the arguments of this relation.Relational nouns are the right input for prenominal possessives to apply to.Sortal nouns must undergo a semantic shift and become relational.Property denoting adjectives can shift to a relational interpretation and modify relational nouns.Within the framework described in this section, the data discussed earlier in the paper can be naturally accounted for.In the following subsections I will show how the grammatical and semantic behavior of prenominal possessives follows from the semantics that has been assigned to them.

DE DICTO AND DE RE READINGS
It has been shown in section 3 that prenominal possessives are referentially linked, but at the same time the expression [possessive+noun] is ambiguous between a de re and a de dicto interpretation in intensional contexts.Within the framework of our analysis this behavior is naturally explained.The referentiality stems from the fact that the possessive operator -in/-ov can apply only to individuals, giving a possessor who is specific in the context and interpreted only relative to the actual world.
We have analyzed this via the choice function mechanism: -in with interpretation λyλRλx.R(x,y) combines with sosedka/neighbor with interpretation f(NEIGHBOR a ), yielding sosedkin/neighborposs with interpretation λRλx.R(x,f(NEIGHBOR a )).Here f(NEIGHBOR a ) is a rigid, de re expression, in every world it denotes the neighbor in w 0 chosen by f.
However, combining a prenominal possessive with a noun results in a predicate -a property that relates to the possessor in the actual world.While the possessor itself is always de re, in intensional contexts this property can be interpreted either de dicto or de re.
I will now show how the two interpretations are derived for ( 22) from section 3, repeated here as ( 48): (48) Petr xočet najti sosedk-in-u podrug-u.Peter wants to find neighbor-poss-f.sg.acc friend-f.sg.acc 'Peter wants to find the neighbor's friend.' The de dicto interpretation can be derived in the standard way analogously to the interpretation of try to find in Montague (1973), but adapted to the peculiarities of Russian: sosedkina podruga 'the neighbor's friend' is of type <e,t>, but occurs in the argument position of najti 'find', and has to shift to the argument type <<e,t,>,t> according to the Partee Triangle (Partee 1987, but in the version of Landman 2003), and najti 'find' shifts in the standard way to combine with <<e,t>,t> arguments.
(50) najti sosedkinu podrugu 'find a friend of the neighbor' λx.∃y[FRIEND w (y,f(NEIGHBOR a )) ∧ FIND w (x,y)] In the intensional context, this interpretation is lifted to type <s<e,t>>: (51) najti sosedkinu podrugu 'find a friend of the neighbor' λvλx.∃y[FRIENDv (y,f(NEIGHBOR a )) ∧ FIND v (x,y)] want is of type <<s<e,t>>t> and it applies to to find a friend of the neighbor: (52) xočet najti sosedkinu podrugu 'want to find a friend of the neighbor' λx.WANT w (x, λvλx.∃y[FRIENDv (y,f(NEIGHBOR a )) ∧ FIND v (x,y)]) WANT relates in w 0 the subject to worlds where the subject finds what is in that world a friend of what is in w 0 the neighbor.This is a de dicto reading of a friend of the neighbor, even though the neighbor itself is rigid, and hence de re.For de re interpretations we will assume a mechanism for deriving wide scope readings.We will not make a choice here on the nature of this mechanism (you can pick your own favorite, like LF movement, storage and retrieval, type shifting, etc.).
The wide scope mechanism does the following: instead of combining najti 'find' with the <e,t> interpretation of sosedkinu podrugu 'a friend of the neighbor', you combine najti 'find' with a free variable x n of type e.The interpretation of najti 'find' can apply to this without lifting, and the derivation builds up: (53) xočet najti x n 'want to find x n ' λx.WANT w (x, λvλx FIND v (x,x n )) The wide scope mechanism abstracts over this variable x n at the relevant level of scope taking, creating what is sometimes called a derived predicate, and combines this with the given interpretation of sosedkinu podrugu/a friend of the neighbor.The derived predicate operates on the wide scope NP as if it were in argument position.
Thus, in our case, the interpretation of the wide scope NP lifts to <<e,t>,t> by (49a), and the relation at type <e,<e,t>> lifts to <<<e,t>,t>, <e,t>> by (49b), and hence we derive: (54) xočet najti sosedkinu podrugu 'want to find a friend of the neighbor' λx.∃y[FRIEND w (y,f(NEIGHBOR a )) ∧ WANT w (x, λvλx.FIND v (x,y))] This interpretation relates the subject in w 0 to worlds in which the subject finds what is in w 0 a friend of the neighbor in w 0 , a de re reading.
Despite the fact that the possessor is always interpreted relative to the actual world, the property of being related to this possessor can have either a wide or narrow scope with respect to the intensional operator.

EVENT NOMINALS
Event nominals are nouns morphologically derived from verbs.They denote sets of events (or states) and inherit from the verb its thematic relations with its arguments (Grimshaw 1990).
One of Babyonyshev's (1997) arguments in support of the determiner analysis of prenominal possessives in Russian is the fact that these expressions can fill an agent argument for event nominals.In this subsection I will take the original sentence from Babyonyshev (1997), (slightly modifying it in order to make the analysis more transparent) and show that the adjectival analysis can naturally deal with these kinds of examples.
Event nominals encode a relation in their lexical semantics, therefore they are the right type of input for the possessive adjectives and can be modified by them.Prenominal possessives modify relations saturating an argument of those relations.Grimshaw (1990) claims that event nominals do not occur without the theme.Thus, a theme argument is saturated first, and the only relation for a possessive to apply to is the relation between an event argument and the agent.
In (55) there are two event nominals -expression and displeasure.Displeasure is one of the arguments of expression. (

NOUNS WITH OPTIONAL ARGUMENTS
Our analysis correctly predicts the patterns of behavior of nouns with optional arguments, e.g.'picture' nouns.A prenominal possessive will aim at saturating an argument of a relation.What kind of relation is available for the prenominal possessive to apply to depends on the number of arguments that the noun has and which of these arguments are explicitly expressed by other grammatical forms (e.g. a postnominal genitive construction).We are now in a position to explain the observations that were made in section 2 of this paper.
I claimed in this paper that a prenominal possessive petin 'of Petya' expresses a function λRλx.R(x,PETYA) that modifies relations and saturates one of the arguments of this relation.The relation is either denoted by the head noun like mother in (37) or expression in (55) or derived from a predicate via the OF-shift as it was shown in (38) -a sortal noun gets reinterpreted as a relation.
Using the predictions that can be made based on the analysis given and the ambiguity of portrait between sortal and relational (with two or three arguments) interpretations, we are going to generate an unrestricted system of possible interpretations for portrait in combination with a prenominal possessive and postnominal genitives.We will then compare the interpretations that the unrestricted analysis predicts with the data in Russian to check whether the predictions are correct and reject some of them.Then we will formulate a more restricted analysis.
The most unrestricted assumption about the lexical meaning of portrait would be that in Russian the lexicon assigns to portrait the following 5 interpretations: There is another argument -an argument of the of w -relation, the possessor.This relation is derived via a semantic shift that sortal nouns undergo to be modified by a prenominal possessive, and this relation is supplied pragmatically.An argument of this relation is always saturated last, as this relation is derived from a predicate nominal (after all arguments of a relation are filled in).
As we will see, I am going to argue that the Russian lexicon does not assign all of these 5 interpretations to portrait.I will claim that the interpretations in (64c) and (64d 2 ) do not exist in Russian.Before arguing that, we are going to check what readings would be derived if portrait were 5 ways ambiguous in this way.
Portrait can be preceded by a prenominal possessive, take one postnominal genitive or two postnominal genitives.
Thus, if portrait is 5 way ambiguous, the following 9 readings are derived.In the first place, there is no existing reading based on putative lexical meaning (64d 2 ): Petin portret Petrova xudožnika Ivanova has only reading (8), but not (9).Moreover, the interpretation of Petin portret Petrova in which Petya is the painter and Petrov is a theme exists, but it cannot be represented as in reading (7), i.e. it does not follow from (64d 2 ).I will claim that it is derived via a different mechanism.Thus, this inverse lexical reading (64d 2 ) can safely be dismissed.
Secondly, there is also good reason to assume that lexical reading (64c) does not exist either.This follows from the fact that reading (5) does not exist.
If portret denoted a relation between the picture and the individual who painted it (λyλx.PORTRAIT w (x) ∧ by(x)=y), then we would expect the following to be a felicitous interpretation: Petin [OF] portret Petrova [by] 'Petya's portrait by Petrov'.Even with a strong contextual support this interpretation is unavailable in Russian.
The context: my grandmother has an impressive collection of portraits that were painted by different artists.The walls of her house are covered with these pictures.In this situation it should be natural to interpret the postnominal genitive Van Goga 'of Van Gogh' in (65) as denoting the painter.However, this does not happen.apply to being provided contextually.The default contextually provided relation is a possessive relation.However, with proper contextual support other relations can become available.
Furthermore, our analysis predicts correctly that if one of the arguments gets a possessive interpretation, then this argument is always saturated last.There are no nouns that incorporate a possessive relation in their lexical semantics.Therefore, the possessive relation is contextually provided via the OF-shift of a sortal noun of type <e,t>.
roditel-in-y ključi parents-poss-pl.nomkeys.pl.nomIntended: 'the parents' keys' b. #semj-in dom family-poss.m.sg.nom house.m.sg.nomIntended: 'the family's house' Thus, the possessive operator can apply only to individuals.Proper names inherently denote individuals and are the most natural input for the possessive morpheme.
Arguably, the interpretations of proper names are generated at type e.If shifting from e to <e,t> is freely available, proper names should be felicitous in the scope of these prefixes.So we GepnerIn the next section I will show that the expression[possessive+noun], being born at type <e,t>, does not always stay at type <e,t> and can shift to <<e,t>,t> in argument position, as predicted by Partee Triangle.
However, the options of contextually available relations are numerous: the car that Mary has dreamed of, the car that Mary has stolen, the car that Mary saw last week, etc.
If we assume that Mary's is an argument, then the head noun must denote a relation.Some nouns are inherently relational like mother, teacher, birthday.Mary's will naturally combine with brother saturating an argument of the brother of relation denoted by the noun.w(x,y)(MARY)λx.BROTHER w (x,MARY)If the head noun is sortal, it would have to undergo a meaning shift into a relational noun.The shift can be driven by different mechanisms: either the lexical semantics of the noun or pragmatics/context.To derive the meaning of Mary's car in (24) there is a shifting rule (OF-shift) for car to get a relational interpretation 'car of'.Then Mary will fill an argument of the relation that is provided by the context, i.e. the most salient relation.The possessive relation is usually the most natural interpretation in a neutral context.
∧ R(x,MARY) (λx.∃y[BROTHER w (x,y)]) λx.∃y[BROTHER w (x,y)] ∧ R(x,MARY)If there is no strong contextual relation between Mary and 'the brother of someone', then the only salient relation is the relation denoted by the head noun.Mary's brother is interpreted as 'the individual who stands in brother relation to someone and Mary is this someone'.
We will start by showing how the meaning of displeasure with Ann is derived.Displeasure is derived from a verbal relation 'x is displeased with y' and denotes a set of states.The noun inherits the thematic relation of the event (state) with its experiencer (Exp) and its theme (Th): w (s) ∧ Exp(s)=x ∧ Th(s)=y (57) shows the derivation for displeasure with Ann.Ann saturates a theme argument: Gepner Start with (64b) λyλx.PORTRAIT w (x) ∧ Th(x)=y Apply to Petrov λx.PORTRAIT w (x) ∧ Th(x)=PETROV Shift with of, apply to Petya λx.PORTRAIT w (x) ∧ Th(x)=PETROV ∧ of w (x,PETYA) w (x) ∧ of w (x,y) Apply to Petya λx.PORTRAIT w (x) ∧ of w (x,PETYA) Reading 2: Start with (64b): λyλx.PORTRAIT w (x) ∧ Th(x)=y Apply to Petya: λyλx.PORTRAIT w (x) ∧ Th(x)=PETYA Gepner