from kinds to objects: prenominal and postnominal adjectives in polish

17
From Kinds to Objects Prenominal and Postnominal Adjectives in Polish Marcin Wągiel Palacký University, Olomouc, Czech Republic; Masaryk University, Olomouc, Czech Republic [email protected]; [email protected] Abstract: In this paper I discuss two types of adjectival modification in Polish, namely the semantics of prenominal and postnominal adjectives. I discuss different entailment patterns related to the placement of adjectival modifiers and the relation between adjectives in both positions and genericity. I postulate a unified intersective semantics for both prenominal and postnominal adjectives and argue that they are predicates denoting properties of objects and kinds respectively. I posit that the kind area is associated with the NP and in the process of semantic composition nouns first combine with postnominal modifiers and then with prenominal ones. The proposal is based on the syntactic analysis of Rutkowski and Progovac (2005) and the semantic framework of McNally and Boleda (2004). Keywords: prenominal adjectives; postnominal adjectives; modification; kinds 1. Introduction It is a well-known fact that Polish allows for both prenominal and postnominal placement of adjectives. 1 Though the syntax of Polish postnominal adjectives has been studied in detail, their semantics has not yet been accounted for from the formal perspective. In this paper I address the puzzle concerning different entailment patterns of adjectival modifiers with respect to their placement. I argue that both prenominal and postnominal adjectives are predicates denoting properties of individuals and have intersective semantics. The difference between them lies in that the prenominal adjectives are properties of objects, whereas postnominal adjectives are properties of kinds. The paper is outlined as follows. In Section 2 I review two standard semantic classifications of adjectives, namely a notionally-based typology developed in Bosque and Picallo (1996) and standard entailment-based typology of Parsons (1970), Kamp (1975) and others. In Section 3 I present novel data from Polish concerning entailment issues related to the prenominal and postnominal adjectival modification. I also examine the impact of the placement of adjectives on generic and existential interpretations of the whole NPs in which they appear. In Section 4 I present some theoretical background for the proposal, namely Rutkowski and Progovac (2005)’s syntactic analysis of classificatory adjectives in Polish and McNally and Boleda (2004)’s treatment of relational adjectives in Catalan as predicates denoting properties of kinds. In Section 5 I propose semantic interpretations for Polish NPs modified by prenominal and postnominal adjectives. Section 6 concludes. 1 I would like to sincerely thank Bożena Cetnarowska, two anonymous reviewers, and the audience at the Olinco 2014 conference for their helpful questions and comments.

Upload: muni

Post on 01-Feb-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

From Kinds to Objects Prenominal and Postnominal Adjectives in Polish

Marcin Wągiel

Palacký University, Olomouc, Czech Republic; Masaryk University,

Olomouc, Czech Republic [email protected]; [email protected]

Abstract: In this paper I discuss two types of adjectival modification in Polish, namely the semantics of prenominal and postnominal adjectives. I discuss different entailment patterns related to the placement of adjectival modifiers and the relation between adjectives in both positions and genericity. I postulate a unified intersective semantics for both prenominal and postnominal adjectives and argue that they are predicates denoting properties of objects and kinds respectively. I posit that the kind area is associated with the NP and in the process of semantic composition nouns first combine with postnominal modifiers and then with prenominal ones. The proposal is based on the syntactic analysis of Rutkowski and Progovac (2005) and the semantic framework of McNally and Boleda (2004). Keywords: prenominal adjectives; postnominal adjectives; modification; kinds

1. Introduction It is a well-known fact that Polish allows for both prenominal and postnominal placement of adjectives.1 Though the syntax of Polish postnominal adjectives has been studied in detail, their semantics has not yet been accounted for from the formal perspective. In this paper I address the puzzle concerning different entailment patterns of adjectival modifiers with respect to their placement. I argue that both prenominal and postnominal adjectives are predicates denoting properties of individuals and have intersective semantics. The difference between them lies in that the prenominal adjectives are properties of objects, whereas postnominal adjectives are properties of kinds.

The paper is outlined as follows. In Section 2 I review two standard semantic classifications of adjectives, namely a notionally-based typology developed in Bosque and Picallo (1996) and standard entailment-based typology of Parsons (1970), Kamp (1975) and others. In Section 3 I present novel data from Polish concerning entailment issues related to the prenominal and postnominal adjectival modification. I also examine the impact of the placement of adjectives on generic and existential interpretations of the whole NPs in which they appear. In Section 4 I present some theoretical background for the proposal, namely Rutkowski and Progovac (2005)’s syntactic analysis of classificatory adjectives in Polish and McNally and Boleda (2004)’s treatment of relational adjectives in Catalan as predicates denoting properties of kinds. In Section 5 I propose semantic interpretations for Polish NPs modified by prenominal and postnominal adjectives. Section 6 concludes. 1 I would like to sincerely thank Bożena Cetnarowska, two anonymous reviewers, and the audience at the Olinco 2014 conference for their helpful questions and comments.

2. Semantic Typologies of Adjectives McNally (forthcoming) distinguishes between three types of adjectival classifications, i.e. morpho-syntactic typologies, notionally-based typologies, and entailment-based typologies. Since this paper focuses on the syntax-semantics interface and does not deal with issues concerning morpho-syntax, for reasons of space I will omit the first classification and I will briefly introduce only the latter two.

2.1 Notionally-based Typology Notionally-based typologies classify adjectival modifiers with respect to what could be described as their descriptive content. A classification of adjectives postulated in Bosque and Picallo (1996) could serve as a typical example of such typology. This account distinguishes between two main classes of adjectives, i.e. qualifying adjectives and relational adjectives. Qualifying adjectives, e.g., black, female, name properties of entities denoted by the modified noun and express qualities of objects. On the other hand, relational adjectives constitute a class of denominal expressions, e.g., technical, molecular, which establish the relation between the modified noun and the entity denoted by the nominal root of an adjective.

Within relational adjectives the typology distinguishes between thematic adjectives and classificatory adjectives. Thematic adjectives saturate some thematic role licensed by the modified noun, e.g., the adjective Russian in (1a) saturates the Agent role and hence the phrase is equivalent to the NP invasion of Russia or invasion of Russians. On the contrary, classificatory adjectives do not saturate any thematic role, but rather classify objects in different domains, e.g., the adjective Russian in the example (1b) does not refer to the roulette played only by Russians, but rather to the particular type of a game of chance. (1) (a) Russian invasion (b) Russian roulette Since the same adjective can be interpreted either as thematic or classificatory, it is probably more adequate to speak about thematic and classificatory uses of relational adjectives.

2.2 Entailment-based Typology Entailment-based typologies are a different type of classification. Such typologies categorize modifiers with respect to the sorts of inferences they license. Since from the very beginning of formal semantics its essential concern was to account for inference in natural language, typologies of this kind date back to the early works of Parsons (1970) and Kamp (1975) and could be considered the most prevalent in the formal semantics tradition. The standard classification distinguishes between two major classes of adjectives, namely intersective and non-intersective adjectival modifiers. In addition, non-intersective modifiers divide further into subsective and intensional, i.e. privative and modal, modifiers.

The reason why adjectives such as black or lazy are called intersective is due to the fact that the denotation of the whole NP results from the intersection of sets denoted by the adjective and the modified noun. The general rule for intersective adjectives may be stated in terms of sets as in (2).

(2) 〚A N〛 = 〚A〛 ∩ 〚N〛

Intersective modifiers follow the entailment pattern in (3) and are traditionally treated as one-place predicates, i.e. expressions of type ⟨e, t⟩. Since the denotation of the NP lazy student is constituted by the intersection of the set denoted by the noun student and the set denoted by the adjective lazy, it contains entities that at the same time have the property of being a student and the property of being lazy. Therefore, the utterance of (3a) necessarily entails both (3b) and (3c). (3) (a) John is a lazy student. (b) ⊨ John is a student. (c) ⊨ John is lazy.

However, according to the standard entailment-based typology not all adjectival modifiers appear to be intersective. Adjectives such as molecular or theoretical are called subsective, because the denotation of the whole NP does not seem to be an intersection, but rather a subset of the denotation of the noun. (4) 〚A N〛 ⊆ 〚N〛 Subsective adjectives do not follow the entailment pattern given in (3). Though (5a) does entail (5b), it definitely does not entail (5c), which may seem somewhat confusing. (5) (a) John is a theoretical linguist. (b) ⊨ John is a linguist. (c) ⊭ #John is theoretical.

An early and probably most widely assumed solution to this puzzle (see Siegel 1976) posits that adjectives do not constitute a homogenous semantic class. Instead, intersective adjectives are treated as first-order predicates, i.e. expressions of type ⟨e, t⟩ denoting properties of individuals, whereas subsective adjectives are analyzed as predicate modifiers, i.e. expressions of type ⟨⟨e, t⟩, ⟨e, t⟩⟩ denoting properties of properties. Hence, although NPs modified by intersective adjectives, e.g., lazy student, get the standard semantic interpretation, see (6a), NPs modified by subsective adjectives, e.g., theoretical linguist, are translated as (6b). (6) (a) 〚lazy student〛 = λx[lazy(x) ∧ student(x)] (b) 〚theoretical linguist〛 = λx[(theoretical(linguist))(x)]

The undoubted advantage of the predicate modifier approach sketched out above is that it does account for different entailment patterns in (3) and (5). Nevertheless, as observed in Larson (1998), the problem with this analysis is that it postulates ambiguity for many adjectives, e.g., formal in (7a). According to the predicate modifier account the same adjective under the intersective reading, see (7b), is treated as a predicate and under the subsective interpretation, see (7c), it is interpreted as a predicate modifier. However, such kind of homophony could be difficult to justify since it is rather undesirable to posit that across languages there are many pairs of syntactically and

semantically separate items that happen to have identical phonological realizations and generally appear in the same positions. (7) (a) John is a formal linguist. (b) intersective reading: John is formal and John is a linguist. (c) subsective reading: John works within the scope of formal linguistics.

In addition, as pointed out in McNally and Boleda (2004), the predicate modifier analysis fails to explain why the presumably non-intersective reading is available in sentences such as (8)2, where the adjective seems to be predicated of an expression of type e. (8) Look at Olga dance – she’s beautiful! In the most natural reading of (8) beautiful does not refer to the property of the individual, but rather to the property of the event in which the individual is involved, i.e. it is Olga’s dancing that is beautiful and not necessarily Olga herself. This fact cannot be accounted for by postulating ambiguity between the predicate and predicate modifier interpretation of the adjective in question and seems to constitute a serious problem for the approach.

Weaknesses of the predicate modifier analysis indicated above lead some researchers, e.g., Larson (1998), McNally and Boleda (2004), Arsenijević, Boleda, Gehrke, and McNally (2014), to propose a unified intersective semantics for adjectives that previously seemed to be subsective, i.e. event-related, relational, and ethnic adjectives respectively.3 We will return to one of these accounts, namely the framework of McNally and Boleda (2004), later on to prepare some theoretical background for our proposal. Let us now, however, take a look at data concerning the placement of adjectives and Polish.

3. Polish Data As it has been mentioned before, Polish allows for both prenominal and postnominal placement of adjectives. In many cases qualitative adjectives tend to occur prenominally and by and large follow the entailment pattern associated with intersective adjectives. Unlike qualitative adjectives, relational adjectives predominantly appear postnominally and seem to behave subsectively. These facts could be summarized by the following examples: (9) (a) dobre oprogramowanie → prenominal placement good software qualifying adjective “good software” intersective semantics 2 The example is originally due to McNally and Boleda (2004); it is their (5). 3 Another attempt to unify the semantics of different classes of adjectival modifiers that should also be mentioned is Partee (2009, 2010)’s treatment of privative adjectives who proposes that such adjectives coerce the expansion of the denotation of the modified noun and are in fact subsective. I believe that Partee’s account could be reanalyzed in terms of intersectivity, nevertheless, for obvious reasons, I will not deal with this issue here and I leave it for further investigation.

(b) oprogramowanie komputerowe → postnominal placement software computer-ADJ classificatory adjective “computer software” subsective semantics

Examples like (9a) and (9b) led some researchers, e.g., Rutkowski and Progovac (2005), Rutkowski (2007, 2012), to generally treat Polish prenominal adjectives as qualifying adjectives and postnominal adjectives as classificatory ones. However, as we will see, the typological distinction between qualitative and classificatory adjectives (or more precisely classificatory uses of relational adjectives) should not be confused with the semantic effects resulting from the placement of Polish adjectival modifiers. In fact, it is common that even typical qualifying adjectives such as color adjectives can occur in both positions: (10) (a) czarny dzięcioł black woodpecker “black woodpecker” (b) dzięcioł czarny woodpecker black “black woodpecker” Though the lexical semantics of the noun dzięcioł (“woodpecker”) and the adjective czarny (“black”) in both examples does not differ, the meaning of (10a) and (10b) is not the same. In the predicative position (10a) would refer to the woodpecker whose color is black, whereas (10b) would refer to the representative of the species Dryocopus martius. The difference in the interpretation derives from the syntax, namely the word order, and similar doublets are numerous in Polish. For the rest of this section the phrases from (10a) and (10b) will serve as a litmus test for examining the relation between the placement of Polish adjectives and such issues as entailment patterns and genericity.

3.1 Entailments Interestingly, the placement of an adjective affects entailments of the sentence in which it occurs. As one can expect, the veracity of (11a) entails that both (11b) and (11c) are true. When the predicate in (11a) is applied to the entity denoted by the subject, it is predicated that the individual Kajtek belongs to the intersection of the set of all woodpeckers and the set of all black objects. (11) (a) Kajtek to czarny dzięcioł. Kajtek this black woodpecker “Kajtek is a black woodpecker (= a woodpecker whose color is black).” (b) ⊨ Kajtek to dzięcioł. Kajtek this woodpecker “Kajtek is a woodpecker.” (c) ⊨ Kajtek jest czarny. Kajtek is black “Kajtek is black.”

Surprisingly, when postposed, even typical intersective adjectives such as color adjectives seem to behave subsectively. The example (12a) does entail (12b), but it does not entail (12c). Thus, if (12a) is true, Kajtek is definitely a woodpecker, but he is not necessarily black. At first glance this fact is difficult to explain in terms of intersectivity and it may suggest that the adjective czarny (“black”) in (12a) is not a predicate. (12) (a) Kajtek to dzięcioł czarny. Kajtek this woodpecker black. “Kajtek is a black woodpecker

(= a representative of the species Dryocopus martius).” (b) ⊨ Kajtek to dzięcioł. Kajtek this woodpecker “Kajtek is a woodpecker.” (c) ⊭ Kajtek jest czarny. Kajtek is black “Kajtek is black.” Moreover, although the sentence in (13a) is inherently contradictory and thus extremely awkward, the sentence in (13b) is perfectly fine and it entails that Kajtek is white. This means that the phrase biały dzięcioł czarny (“white black woodpecker”) in (13b) should be understood as referring to the albino representative of the species Dryocopus martius. (13) (a) #Kajtek to biały czarny dzięcioł. Kajtek this white black woodpecker (b) Kajtek to biały dzięcioł czarny. Kajtek this white woodpecker black “Kajtek is a white black woodpecker.”

The data in (11), (12), and (13) may lead to the conclusion that unlike prenominal adjectives that are predicates and have intersective semantics, postnominal adjectives are predicate modifiers with subsective semantics. In this paper, however, I will argue that there is a more appropriate way to account for the puzzling facts described above. Nevertheless, before we outline the theoretical background for the proposal, there is another interesting piece of data that should be taken into consideration.

3.2 Genericity Similarly to bare common nouns Polish NPs modified by postnominal adjectives can shift between existential and generic interpretations freely. The NP in (14a) is ambiguous and can occur in both existential and generic contexts, see (14b) and (14c) respectively. (14) (a) dzięcioł czarny → existential reading woodpecker black → generic reading “black woodpecker”

(b) Dzięcioł czarny siedział na moim parapecie. woodpecker black sat-IMPERF on my-LOC parapet-LOC “A black woodpecker was sitting on my parapet.” (c) Dzięcioł czarny wyginął w XXI wieku. woodpecker black died-out-PERF in 21st century-LOC “The black woodpecker became extinct in the 21st century.”

Interestingly, NPs modified by prenominal modifiers do not show any ambiguity with respect to generic and existential readings and seem to lack generic interpretations, see (15a). Prenominal adjectives in existential contexts such as (15b) are perfectly fine, but sentences like (15c), where the NP with prenominal adjective appears as the argument of the generic predicate are clearly anomalous. (15) (a) czarny dzięcioł → existential reading black woodpecker → *generic reading “black woodpecker” (b) Czarny dzięcioł siedział na moim parapecie. black woodpecker sat-IMPERF on my-LOC parapet-LOC “A black woodpecker was sitting on my parapet.” (c) #Czarny dzięcioł wyginął w XXI wieku. black woodpecker died-out-PERF in 21st century-LOC

Besides the constraints on the distribution of NPs modified by prenominal adjectives presented above there is one more interesting fact worth consideration. Postnominal adjectives cannot combine with proper names (unless the postnominal adjective is the part of a proper name, i.e. the last name) as can be witnessed by the anomalous phrase in (16b).4 (16) (a) czarny Kajtek black Kajtek “black Kajtek” (b) #Kajtek czarny Kajtek black For reasons of space I will completely omit all the problematic issues concerning modification of proper names which would probably require some sort of type-shifting to be accounted for. Nevertheless, the contrast between (16a) and (16b) may further suggest that there is in fact some crucial difference between the prenominal and postnominal placement of adjectives with respect to genericity since only adjectives in the prenominal position, where they force the unambiguously existential interpretation of the whole NP, can modify expressions denoting object-level individuals.

4 The # sign in (16b) indicates that the reading on which the adjective is not the part of the proper name is unavailable. In writing the difference between these two interpretations is emphasized orthographically by the use of capital letters, cf. #Kajtek czarny and Kajtek Czarny.

The data presented in this section is intriguing, because it shows a clear connection between the placement of adjectives in Polish and two semantic phenomena, namely entailments and genericity. Although the classificatory flavor of postnominal adjectives has been already studied, their puzzling behavior described above has not been observed. In the following section I will develop some theoretical tools that will allow us to account for the semantic properties of Polish prenominal and postnominal adjectives in terms of intersectivity.

4. Theoretical Background To begin with, there are several preliminary assumptions to be made. First of all, I presume that the intersective semantics for all classes of adjectival modifiers is both possible and desirable. In this paper, however, I will not deal with intensional, i.e. privative and modal, adjectives. Second, I assume that bare common nouns enter the composition ready to denote kinds and that it is legitimate to model kinds as individuals (Carlson 1977). This assumption is mainly motivated by the intuition that kinds seem to be ontologically prior to specimens and by the cross-linguistic fact that in languages allowing for bare NPs they are basically used as generic terms (Krifka 1995). By all means, common nouns can denote sets of object-level entities, technically via the realization relation R (Carlson 1977). Third, I assume that the order of the semantic composition of an expression corresponds to its syntactic structure.

4.1 Syntax of Polish Adjectives In this paper I adopt the so-called ClassP model (henceforth CPM)5 of the syntax of Polish postnominal adjectival modifiers developed in Rutkowski and Progovac (2005) and Rutkowski (2007, 2009, and 2012). The approach in question contributes to the discussion of the universal DP hypothesis by arguing that there are sound reasons to posit a DP structure even in determiner-less languages, e.g., Polish, and to postulate a universal functional head on top of the NP projection.

In the CPM analysis prenominal adjectives are generally treated as qualifying adjectives, while postnominal adjectives are regarded as classificatory adjectives. It should be noted that the distinction is not so clear-cut, as pointed out in Cetnarowska, Pysz, and Trugman (2011), and that there do exist some NPs such as (17)6, in which the adjective definitely has a classificatory flavor and at the same time it has to surface in the prenominal position. Nevertheless, for sake of clarity in this paper I will ignore all the intricacies and assume a sharp semantic distinction between prenominal and postnominal adjectives in Polish. (17) (a) boża krówka god-ADJ cow-DIM “ladybird” (b) *krówka boża cow-DIM god-ADJ

5 The term was coined in Cetnarowska, Pysz, and Trugman (2011) and used later on in Linde-Usiekniewicz (2013). 6 The example is due to Cetnarowska, Pysz, and Trugman (2011); it is their (14).

According to the CPM approach the syntactic difference between qualifying and classificatory adjectival modifiers lies in that qualifying APs are merged in functional projections (αPs) between the DP and NP, whereas classificatory APs are base generated in the SpecNP and are associated with the strong feature [+class] that has to be checked by the noun. As a result the noun moves to the head of the special projection being an immediate functional extension of the NP, namely to Classo (Rutkowski and Progovac 2005) or no (Rutkowski 2007, 2012)7 and ends up necessarily preceding the classificatory adjective in the surface syntax as in (18).

(18) stary brzydki biały dzięcioł czarny old ugly white woodpecker black “old ugly white black woodpecker”

The syntactic derivation of (18) may be illustrated by the phrase marker diagram in (19), which is just an adjustment of the figure (5) from Rutkowski (2012). (19)

7 In the subsequent part of the paper I will follow the labelling from Rutkowski (2007, 2012).

Though iterative projections of qualifying APs are possible and hence there is no

limitation to the number of prenominal adjectives, the immediate extension of the NP is unique, i.e. non-recurrent, which results in that Polish allows for only one postnominal, i.e. classificatory, adjective inside the DP.8 However, what is crucial for purposes of this paper is the fact that in the structure in (19) it is the postnominal adjective that is closer to the noun than any of the prenominal adjectives. “Being closer” should be understood in the way that the AP projection containing the postnominal adjective is the first branching node that c-commands No.

4.2 Composition One of the main goals of this paper is to argue that despite of apparent differences in entailment patterns both prenominal and postnominal adjectives in Polish are in fact predicates. To justify this claim we need to account for the acceptability of phrases such as (18), or more precisely, its crucial part repeated here as (20) which seems to be difficult to explain on the intersective account. (20) biały dzięcioł czarny white woodpecker black “white black woodpecker”

For purposes of the analysis I adopt Heim and Kratzer (1998)’s rule of Predicate Modification generalized to events in Morzycki (forthcoming): (21) PREDICATE MODIFICATION (GENERALIZED TO EVENTS)

If a branching node α has as its daughters β and γ, and 〚β〛 and 〚γ〛 are either both of type ⟨e, t⟩ or both of type ⟨v, t⟩, then 〚α〛 = λX.〚β〛(X) ∧ 〚γ〛(X), where X is an individual or an event (whichever would be defined).

Of course Predicate Modifiacation defined above is intersective modification and in terms of sets it is equivalent to (2).

Having in mind the assumption that the order of the semantic composition corresponds to the syntactic derivation, the structure in (19) posits the following sequence of the composition of the phrase in (20): 8 Rutkowski and Progovac (2005)’s claim that Polish postnominal adjectives are obligatorily non-iterative is somewhat controversial since there are a relatively small number of examples such as msza święta żałobna ‘lit. mess holy memorial; memorial mess’ (Cetnarowska, Pysz, and Trugman 2011) which seem to be potentially problematic for the adopted analysis. In effect, Cetnarowska, Pysz, and Trugman (2011) and Cetnarowska (2013) propose a different account for the placement of classificatory adjectives in Polish. Following the representational model of Bouchard (2002) they posit the Linearization Parameter determining whether the functor, i.e. the head, precedes or follows its dependent. Since this paper comes with different solution, I will not comment further on this issue here. For the detailed discussion see Linde-Usiekniewicz (2013).

(22) The noun first combines with the postnominal adjective, i.e. a predicate of type ⟨e, t⟩, by Predicate Modification defined in (21), so the resulting expression is also of type ⟨e, t⟩, and not until then the composed NP combines with the prenominal adjective. The result of the composition illustrated in the diagram in (22) is the conjunction-based semantics: (23) 〚biały dzięcioł czarny〛 = λx[woodpecker(x) ∧ black(x) ∧ white(x)]

The formula in (23) could be generalized to all similar structures as in (24), where N, A1, and A2 are predicates corresponding to the meaning of a noun, prenominal adjective, and postnominal adjective respectively. (24) 〚AP1 N AP2〛 = λx[N(x) ∧ A2(x) ∧ A1(x)]

Given that conjunction is commutative and associative the order of conjuncts in (23) and (24) itself of course does not yet explain why the whole NP is a perfectly normal Polish phrase since the intersection of the sets denoted by the expressions dzięcioł (“woodpecker”), czarny (“black”), and biały (“white”) is still necessarily the empty set. As a result the whole phrase should be inherently contradictory and thus semantically anomalous that is obviously not the case. For the purpose of the proper explanation of this challenging data we need though some more ingredients.

4.3 Properties of Kinds The semantic framework adopted in this paper is McNally and Boleda (2004)’s analysis of relational adjectives in Romance. This approach refuses the standard analysis of subsectively interpreted adjectives as predicate modifiers denoting properties of properties (Siegel 1976) and argues for their intersective semantics. McNally and Boleda build on standard theories of genericity (e.g., Carlson 1977; Krifka et al. 1995) and Larsonian intersective semantics for event-related adjectives (Larson 1998) to provide a semantic interpretation of Catalan relational adjectives as denoting properties of kinds.

Inspired by Larson (1998)’s analysis McNally and Boleda (2004) postulate that all common nouns have an implicit kind-level argument that gets saturated by a contextually-determined kind and is associated with the object-level argument via the standard Carlsonian realization relation R (Carlson 1977). In this analysis all common nouns translate as follows:

(25) 〚N〛 = λxkλyo[R(yo, xk) ∧ N(xk)] In the formula in (25) N is a predicate associated with the noun, whereas xk and yo represent individual variables. The superscript k indicates a kind-level entity while the superscript o marks an object-level entity. In other words, the formula in (25) states that objects realize the kind of entities denoted by the common noun.

Furthermore, McNally and Boleda (2004) posit that Catalan relational adjectives are in fact generic predicates, i.e. they denote properties of kinds similarly to English adjectives such as extinct or widespread. Thus, in their analysis APs containing relational adjectives receive the translation as in (26), where AR is a one-place predicate related to the relational adjective. (26) 〚APR〛 = λxk[AR(xk)]

Such semantics allows relational adjectives to modify any kind since they can be truthfully applied to whatever kind-level entity introduced by the modified noun via the composition rule defined in (27): (27) If noun N translates as λxkλyo[R(yo, xk) ∧ N(xk)] and adjectival phrase AP

translates as λxk[A(xk)], then [N AP] translates as λxkλyo[R(yo, xk) ∧ N(xk) ∧ A(xk)]

As a result we obtain means to preserve the intersective semantics of relational

adjectives and at the same time to explain their apparent subsective behavior. The analysis is furthermore justified by the fact that contrary to the predictions formulated by the predicate modifier account, and deceptively corroborated by the awkwardness of examples such as (5c), relational adjectives can be used predicatively provided that the argument of an adjective denotes something related to a kind. This holds not only for Catalan but also for English and Polish as can be witnessed by the acceptability of (28a)–(28c), where the triplet of equivalent sentences in the languages in question is given.9 (28) (a) This conference is international. (b) Aquest congrés és internacional. (c) Ta konferencja jest międzynarodowa.

With these tools in place, we can now proceed to propose a unified semantics for both prenominal and postnominal adjectives in Polish.

5. Proposal Put informally, the essential idea of this paper is that the closer (in terms of hierarchical structure) to the modified noun the adjective is the more generic it is. Since Polish postnominal adjectives sit inside the NP, namely in the SpecNP, see (19), they combine with the noun earlier and denote properties kinds. On the contrary, prenominal adjectives 9 The examples in (28a) and (28b) are due to McNally and Boleda (2004); they are their (10b). The equivalent Polish sentence in (28c) is mine.

are merged in the region between the DP and NP and hence they enter the semantic composition later and denote properties of object-level individuals. It should be noted, however, that it is not a mere proximity of an adjective to the noun that results in the kind or object interpretation, but rather the fact whether it is located inside or outside the NP. In other words, the boundary between the kind and object area is situated in the nP. Everything inside the NP is generic while the region of αPs belongs to the area of objects. Due to the Carlsonian realization relation R both common nouns and NPs modified by postnominal adjectives can shift to the object denotation any time, however there is no means for NPs modified by prenominal adjectives to shift their denotation to the sort of kinds.

Following Krifka et al. (1995) and McNally and Boleda (2004) I assume that common nouns enter composition ready to denote kinds, i.e. nominal predicates contain an implicit kind argument, see (25) repeated here as (29a). At the same time, I posit that all Polish adjectives but intensional adjectives are first order predicates, see (29b). The sort of the argument an adjective requires is not defined by its semantics, but rather it results from its placement. APs that are base generated inside the NP and enter the composition just after nouns, require kind-level individuals as their arguments, see (29c). In the surface syntax such APs occur postnominally. On the other hand, APs which are merged higher, i.e. in the region between the DP and NP, enter the composition later and select object-level arguments, see (29d). This is the case of Polish prenominal adjectives. Finally, in the semantic interpretation of phrases with both prenominal and postnominal adjectives such as (20) the order of conjuncts is always as in (29e), i.e. the predicate associated with the postnominal AP (A2) directly follows the noun and precedes the predicate related to the prenominal AP (A1). (29) (a) 〚N〛 = λxkλyo[R(yo, xk) ∧ N(xk)] (b) 〚AP〛 = λx[A(x)] (c) 〚N AP〛 = λxkλyo[R(yo, xk) ∧ N(xk) ∧ A(xk)] (d) 〚AP N〛 = λxkλyo[R(yo, xk) ∧ N(xk) ∧ A(yo)] (e) 〚AP1 N AP2〛 = λxkλyo[R(yo, xk) ∧ N(xk) ∧ A2(xk) ∧ A1(yo)]

As one can see in (29b), I assume a unified semantics for Polish adjectives (with the probable exception of intensional adjectives) and posit that they always denote properties, but it is the syntax that specifies the sort of properties, i.e. properties of kinds or properties of objects. This accounts for the empirical facts discussed in Section 3.2, see examples in (14) and (15), showing that unlike Polish NPs modified by postnominal adjectives that are ambiguous between generic and existential readings, NPs modified by prenominal adjectives have existential interpretations only. Though the realization relation R can shift the first from the domain of kinds to the domain of objects, the latter cannot go the opposite way, i.e. get the kind denotation.

Let us now discuss the denotation of the Polish noun dzięcioł “woodpecker” and the phrases czarny dzięcioł, dzięcioł czarny, and biały dzięcioł czarny, as presented in (10a), (10b), and (20) respectively: (30) (a) 〚dzięcioł〛 = λxkλyo[R(yo, xk) ∧ woodpecker(xk)](kj) = = λyo[R(yo, kj) ∧ woodpecker(kj)]

(b) 〚czarny dzięcioł〛 = λxkλyo[R(yo, xk) ∧ woodpecker(xk) ∧ black(yo)](kj) = = λyo[R(yo, kj) ∧ woodpecker(kj) ∧ black(yo)] (c) 〚dzięcioł czarny〛 = λxkλyo[R(yo, xk) ∧ woodpecker(xk) ∧ black(xk)](kj) = = λyo[R(yo, kj) ∧ woodpecker(kj) ∧ black(kj)] (d) 〚biały dzięcioł czarny〛 = λxkλyo[R(yo, xk) ∧ woodpecker(xk) ∧ black(xk)

∧ white(yo)](kj) = = λyo[R(yo, kj) ∧ woodpecker(kj) ∧ black(kj) ∧ white(yo)] Following McNally and Boleda (2004) I assume that the noun’s implicit kind argument gets saturated by the contextually-determined kind represented in the formulae by kj. McNally and Boleda argue that the kind is uniquely identifiable in the particular context and hence this step is justified. As one can see in (30a), the noun dzięcioł denotes a function from object-level entities realizing the kind “woodpecker”, i.e. Dryocopus, to truth values, or in other words, the set of objects realizing this kind. The prenominal adjective czarny modifies the denotation of (30a) by introducing an additional truth condition concerning object-level individuals. Hence, (30b) represents the set of all objects that are black and that realize the kind Dryocopus. This crucially differs from (30c), where the set denoted by the postnominal adjective intersects with the set denoted by the kind, i.e. it is the kind “woodpecker” that has been assigned the property of being black and not necessarily the objects that realize it. As a result, a subkind of the woodpecker is established, namely “black woodpecker”, i.e. Dryocopus martius. Therefore, the NP in (30c) denotes the set of all objects realizing the kind Dryocopus martius. This accounts for the classificatory flavor of postnominal adjectives that has not been captured formally until now. Finally, in (30d) it is the postnominal adjective that first combines with the noun to assign the property to a kind-level entity and thus establish the subkind “black woodpecker” and not until then the resulting expression is modified by the prenominal adjective biały denoting the property of objects realizing the kind. Hence, the formula in (30d) represents the set of all objects that are white and realize the kind “black woodpecker”. Such semantics predicts that it should be possible to paraphrase (13b), repeated below as (31a), as (31b) and in fact the sentence in (31b) expresses the exact meaning of (13b) stating that Kajtek is an albino representative of the species Dryocopus martius. (31) (a) Kajtek to biały dzięcioł czarny. Kajtek this white woodpecker black “Kajtek is a white black woodpecker.” (b) Kajtek to biały okaz dzięcioła czarnego. Kajtek this white representative woodpecker-GEN black-GEN “Kajtek is a white representative of the black woodpecker.”

The proposed semantics also explains why the sentence (13a) from Section 3.1 is anomalous. The reason is that since in the NP #biały czarny dzięcioł both adjectives

occur prenominally, they select object-level entities as their arguments and the intersection of the sets denoted by the expressions biały, czarny, and dzięcioł is the empty set which results in necessarily tautological truth-conditions of any sentence in which the NP would occur.

An additional argument in favor of the proposed analysis comes from the fact that properties can be assigned to kinds not only by generic predicates such as widespread or extinct, but also by other predicates, e.g., qualitative adjectives. As can be witnessed in (14), it is possible to use color adjectives predicatively not only with respect to object-level individuals denoted by proper names, see (14a), but also to kind-level entities denoted by generic terms, see (14b). (14) (a) Kajtek jest czarny, a Krzyś zielony. Kajtek is black and Krzyś green. “Kajtek is black and Krzyś is green.” (b) Jeden gatunek dzięcioła jest czarny, a drugi zielony. one species woodpecker-GEN is black and second green. “One species of the woodpecker is black and one is green.” Hence, it seems that there is nothing in the lexical semantics of adjectives typically characterized as not being generic that would prevent them from combining with kinds. The only constraint is imposed by syntax, i.e. in order to do so they need to occur postnominally or in post-copular phrases.

According to the standard rule for intersective modifiers given in (2) it appears that Polish adjectives in both positions in question are in fact intersective. Different entailment patterns in (11) and (12) result from the fact that they select arguments of different sorts. Since postnominal adjectives denote properties of kinds, (12a) cannot entail (12c) simply because the adjective czarny (“black”) in the postnominal position says nothing about the property of the particular object-level individual realizing the kind denoted by the noun.

6. Conclusion In this paper I have dealt with the problem concerning different entailment patterns of prenominal and postnominal adjectives in Polish. I have presented novel data showing that while NPs modified by postnominal adjectives are systematically ambiguous between generic and existential readings, NPs modified by prenominal adjectives get existential interpretations only. I have argued that all Polish adjectives except intensional ones are predicates and as such denote properties of individuals and have intersective semantics. Nevertheless, their placement has an impact on whether they take object-level or kind-level arguments. According to my proposal, since postnominal adjectives sit inside the NP, they combine with the noun first and denote properties of kinds, whereas prenominal adjectives, merged in the area above the NP, enter the semantic composition later and denote properties of objects. This explains supposed subsective behavior of postnominal adjectives and acceptability of phrases such as biały dzięcioł czarny (“white black woodpecker”) that seemed to be challenging for the intersective account.

The analysis presented in this article gives an explanation for the Polish data, however, an obvious question arises: how to account for the cross-linguistic variation in the placement and interpretation of adjectives within the proposed framework?

Answering such a question would certainly lead far beyond the scope of this paper and hence the extension of the idea developed here has to be left for the future investigation. Works Cited Arsenijević, Boban, Gemma Boleda, Berit Gehrke, and Louise McNally. 2014. “Ethnic

Adjectives are Proper Adjectives”. In CLS 46-I: The Main Session. Proceedings of the 46th Annual Meeting of the Chicago Linguistic Society, edited by Rebekah Baglini, Timothy Grinsell, Jonathan Keane, Adam R. Singerman, and Julia Thomas, 17-30. Chicago: The Chicago Linguistic Society.

Bosque, Ignacio, and Carme Picallo. 1996. “Postnominal Adjectives in Spanish DPs”. Journal of Linguistics 32: 57–78.

Bouchard, Denis. 2002. Adjectives, Number and Interfaces: Why Languages Vary. Oxford: Elsevier Science.

Carlson, Gregory. 1977a. “A Unified Analysis of the English Bare Plural”. Linguistics and Philosophy 1: 413–456.

Carlson, Gregory. 1977b. Reference to Kinds in English. Ph.D. dissertation. Amherst: University of Massachusetts.

Cetnarowska, Bożena. 2013. “The Representational Approach to Adjective Placement in Polish”. Linguistica Silesiana 34: 7–22.

Cetnarowska, Bożena, Agnieszka Pysz, and Helen Trugman. 2011. “Accounting for Some Flexibility in a Rigid Construction: On the Position of Classificatory Adjectives in Polish”. In Generative Investigations: Syntax, Morphology, and Phonology, edited by Piotr Bański, Beata Łukaszewicz, Monika Opalińska, and Joanna Zalewska, 24–47. Newcastle upon Tyne: Cambridge Scholars Publishers.

Chierchia, Gennaro. 1998. “Reference to kinds across languages”. Natural language semantics 6: 339–405.

Heim, Irene, and Angelika Kratzer. 1998. Semantics in Generative Grammar. Oxford: Blackwell Publishing

Kamp, Hans. 1975. “Two Theories about Adjectives”. In Formal Semantics of Natural Language, edited by Edward L. Keenan, 123–155. Cambridge: Cambridge University Press.

Krifka, Manfred. 1995. “Common Nouns: A Contrastive Analysis of Chinese and English”. In The Generic Book, edited by Gregory N. Carlson and Francis J. Pelletier, 398–411. Chicago: The University of Chicago Press.

Krifka, Manfred, Francis Pelletier, Gregory Carlson, Alice ter Meulen, Gennaro Chierchia, and Godehard Link. 1995. “Genericity: An Introduction”. In The Generic Book, edited by Gregory N. Carlson and Francis J. Pelletier, 1–124. Chicago: The University of Chicago Press.

Larson, Richard. 1998. “Events and Modification in Nominals”. In Proceedings from Semantics and Linguistic Theory (SALT) VIII, edited by Devon Strolovitch and Aaron Lawson, 145–168. Ithaca, NY: CLC Publications.

Linde-Usiekniewicz, Jadwiga. 2013. “A Position on Classificatory Adjectives in Polish”. Studies in Polish Linguistics 8/3: 103–126.

McNally, Louise. Forthcoming. “Modification”. In The Cambridge Handbook of Semantics, edited by Maria Aloni and Paul Dekker. Cambridge: Cambridge University Press.

McNally, Louise and Gemma Boleda. 2004. “Relational Adjectives as Properties of Kinds”. In Empirical Issues in Formal Syntax and Semantics, edited by Olivier Bonami and Patricia Cabredo Hofherr, 179–196. http://ww.cssp.cnrs.fr/eiss5.

Morzycki, Marcin. Forthcoming. Modification. Cambridge: Cambridge University Press. Parsons, Terence. 1970. “Some Problems Concerning the Logic of Grammatical

Modifiers”. Synthese 21: 320–324. Partee, Barbara. 2009. “Formal Semantics, Lexical Semantics, and Compositionality:

The Puzzle of Privative Adjectives”. Philologia 7: 11–23. Partee, Barbara. 2010. “Privative Adjectives: Subsective plus Coercion”. In

Presuppositions and Discourse: Essays Offered to Hans Kamp, edited by Rainer Bäuerle, Uwe Reyle, and Thomas E. Zimmermann, 273–285. Bingley: Emerald Group Publishing Limited.

Rutkowski, Paweł. 2007. “The Syntactic Properties and Diachronic Development of Postnominal Adjectives in Polish”. In Formal Approaches to Slavic Linguistics: The Toronto Meeting 2006, edited by Richard Compton, Magdalena Goledzinowska, and Ulyana Savchenko, 326–345. Ann Arbor, MI: Michigan Slavic Publications.

Rutkowski, Paweł. 2009. Fraza przedimkowa w polszczyźnie. Warsaw: Wydział Polonistyki Uniwersytetu Warszawskiego.

Rutkowski, Paweł. 2012. “Is nP Part of Universal Grammar”. Journal of Universal Language 13-2: 119–144.

Rutkowski, Paweł, and Ljiljana Progovac. 2005. “Classification Projection in Polish and Serbian: The Position and Shape of Classifying Adjectives”. In Formal Approaches to Slavic Linguistics: The South Carolina Meeting 2004, edited by Steven Franks, Frank Y. Gladney, and Mila Tasseva-Kurktchieva, 289–299. Ann Arbor, MI: Michigan Slavic Publications.

Siegel, Muffy. 1976. Capturing the Adjective. PhD diss. Amherst: University of Massachusetts.