Scholarly article on topic 'Reference to and via properties: the view from Dutch'

Reference to and via properties: the view from Dutch Academic research paper on "Languages and literature"

0
0
Share paper
Academic journal
Linguistics and Philosophy
OECD Field of science
Keywords
{""}

Academic research paper on topic "Reference to and via properties: the view from Dutch"

Linguist and Philos (2015) 38:315-362 DOI 10.1007/s10988-015-9173-0

RESEARCH ARTICLE

Reference to and via properties: the view from Dutch

Louise McNally1 • Henriette de Swart2

Published online: 15 August 2015

© The Author(s) 2015. This article is published with open access at Springerlink.com

Abstract Many languages offer a surprisingly complex range of options for referring to entities using expressions whose main descriptive content is contributed by an adjective, such as Dutch de blinde 'the blind (person),' het besprokene, 'the (matters) discussed,' or het ongewone van het niet roken 'the strange (thing) about not smoking.' In this paper, we present a case study of the syntax and compositional semantics of three such constructions in Dutch, one of which we argue has not previously been identified in the literature. The data and the analysis will shed light on our understanding of how reference using adjectives differs from that using nouns in languages that have the two categories, as well as on the differences between reference to entities via their properties vs. reference to properties themselves. Finally, we briefly discuss related work and indicate directions for future study of the typological variation found in this rich and highly understudied corner of natural language.

Keywords Adjective semantics • Noun semantics • Abstract objects • Nominalization • Inflected adjectives • Kinds • Properties • Tropes

& Henriette de Swart h.deswart@uu.nl

Louise McNally louise.mcnally@upf.edu

1 Department of Translation and Language Sciences,

Universitat Pompeu Fabra, c/Roc Boronat, 138, 08018 Barcelona, Spain

2 Utrecht Institute of Linguistics, Utrecht University, Trans 10, 3512 JK Utrecht, The Netherlands

1 Introduction

In formal semantic theories with roots in Montague Grammar, nouns and predicative adjectives are both generally assigned the same semantic type: they denote properties of individuals, typically modeled as sets of those individuals describable by the noun or adjective, or intensionally as functions from indices to such sets.1 Against such a theoretical background, we might expect there to be no substantive difference in the way the two types of expressions are used (in combination with determiners) for reference, and, indeed, it is somewhat surprising on such theories that there are distinct categories of adjectives and nouns at all. However, differences clearly do exist between the ways in which adjectives and nouns can be used for reference. This paper investigates three Dutch Determiner Phrase (DP) configurations with an inflected adjective (marked with a subscripted [+e] in the gloss) and without an overt nominal 'core.' To give just a few examples, een vreemde in (1a) can only be used for reference to humans (a stranger, i.e. someone you don't know), even though the adjective vreemd 'strange' can in principle apply to physical objects (e.g. een vreemd kunstwerk 'a strange piece of art'). Al het vreemde in (1b) can only have mass, and not count reference ('that which is strange'), even though there are uses of the definite article het that are compatible with count reference (as in het kind 'the child'). In contrast to (1b), het vreemde van het ongeluk in (1c) doesn't describe something strange in itself, but refers to the strange aspect of the accident:

(1) a. Hoe leer je een kind dat het niet met een vreemde mee how teach you a child that it not with a strange[?e] with mag gaan? may go

'How do you teach a child not to leave with a stranger?'

b. Ze moeten wennen aan al het nieuwe, al het vreemde they must get-used to all the new[+e] all the strange[+e] dat dit land hen biedt.

that this land them offers

'They must get used to everything new, everything strange that this land offers them.'

c. Het vreemde van de situatie vind ik dat politieke partijen the strange[?e] of the situation find I that political parties het maar laten gebeuren.

it just let happen

'The strange thing about the situation, I find, is that political parties just let it happen.'

1 On some analyses (e.g. von Stechow 1984) gradable adjectives denote relations between individuals and degrees or measure functions, and come to denote properties of individuals only in combination with a degree argument or degree morphology. The reader who prefers such analyses could substitute "adjectives" with "adjectives once combined with the relevant degree component'' in the text above; for simplicity, we set aside this complication.

The examples in (1) are extracted from Google searches, as are most of the examples discussed in this paper. The examples are sometimes shortened or adapted for reasons of space, but we prefer this methodology of naturally occurring examples over artificially constructed ones for the topic at hand. We find that the naturalness of the examples helps us to obtain good native speaker intuitions and capture the relevant generalizations.2

The premise of this paper is that we can gain insight into the syntactic and semantic differences between nouns and adjectives and the consequences of those differences for their referential uses by exploring in depth the constructions in (1) and the ways in which these constructions differ from DPs whose main descriptive content is provided by a noun (e.g. de hond 'the dog').3 We also hope to shed light on how reference to properties themselves is effected. To this end, we present an analysis of these three Dutch constructions, none of which, to our knowledge, has previously been given a comprehensive syntactic and semantic analysis (though see Kester 1996 and McNally and de Swart 2011 for partial descriptions).

The main claims of the paper can be summarized as follows. We argue that the construction in (1a) involves a null noun with human reference, and the adjective has its regular predicative interpretation. In contrast, we take the construction in (1b) to have the syntax of a small clause and the semantics of a free relative, with a semantics roughly paraphrased in English as 'all that is Adj(ective).' Finally, we assign the inflected adjective in (1c) a relational interpretation: it takes the DP complement of the PP as its inner argument and, in combination with the determiner, refers to a property, which we roughly paraphrase as 'the Adj aspect of/ thing about DP' or, in some cases, 'the Adj-ness of the DP.'4

The paper is organized as follows. We start with some background information on the patterns of adjectival inflection in Dutch (Sect. 2). Sections 3-5 are devoted, respectively, to the syntactic and semantic analysis of the three constructions illustrated in (1), including arguments that (1b) and (1c) do in fact constitute distinct constructions.

2 Of course, all examples have been checked against our respective native speaker intuitions for Dutch (de Swart) and English (McNally). The following abbreviations not mentioned in the text also appear in our glosses: Adj = adjective, Adv = adverb, dem = demonstrative, dim = diminutive, f(em) = feminine, inf = infinitive, m = masculine, part = particle, pl = plural, refl = reflexive.

3 For simplicity and where no confusion will arise, we will sometimes use the locution ''DPs headed by a [noun/adjective/...]'' for such cases, even though strictly speaking the head of the DP is the determiner on our analysis.

4 A note particularly for native speakers of English: This last construction lacks a single formal counterpart in English, which might make it intuitively difficult for native English speakers to grasp. We think the best way to deal with this potential difficulty is via the illustration of the specific properties of the construction as the paper develops. It is also therefore hard to find an exact English paraphrase for the inflected adjective in this construction.

2 Background on the uses of inflected adjectives in Dutch

We begin with some relevant background on Dutch. First, we review the basic patterns of adjectival inflection within nominals. We then distinguish the non-elliptical uses of inflected adjectives that will concern us from their elliptical uses, which will not further interest us here. This latter discussion will also provide a bit more introductory information about the three constructions to be discussed in Sects. 3-5.

2.1 Patterns of adjective inflection in prenominal position

Adjective inflection in Dutch depends in part on the gender of the noun that the adjective co-occurs with. Dutch nouns are classified as masculine, feminine or neuter. The distinction between masculine and feminine nouns is visible with natural gender, where the two lead to different anaphoric pick-ups: de man...hij 'the man...he' vs. de vrouw.zij 'the woman.she.' In the grammatical system of articles and adjective inflection, masculine and feminine nouns behave alike, in contrast to neuter nouns. Thus, Dutch grammars commonly establish a distinction between common gender and neuter gender. Dutch has two definite articles: het (glossed in 3 as 'the.SG.N'), which is singular and neuter; and de (glossed here as 'the'), which combines with all common gender (i.e. masculine and feminine) singular nouns, as well as all (common gender and neuter) plural nouns. There is one (singular) indefinite article een 'a,' which is used for all (common gender and neuter) singular nouns.

When a neuter noun appears with the indefinite article, any accompanying adjective appears in the basic uninflected form:

(2) Ik kocht een groen boek. [neuter noun, indefinite] I bought a green[ book

'I bought a green book.'

In all other cases, the adjective appears in its inflected form. That is, definite DPs containing neuter nouns (3 a), both definite and indefinite DPs containing common gender nouns (3b), and definite/bare plurals (3c), all use the long form of the adjective with the schwa -e at the end:

(3) a. Het groene boek was mooi. [neuter noun, definite]

the.SG.N green[+e] book was beautiful 'The green book was nice.'

b. Ik heb een/de groene tafel gekocht. [common gender noun, I have a/the green[?e] table bought indef/def] 'I bought a/the green table.'

c. Ik heb (de) groene boeken/tafels gekocht. [neuter/common gender I have (the) green[+e] books/tables bought noun, pl.] 'I bought (the) green books/tables.'

All adjectives show this alternation, except when they end in schwa (roze 'pink,' mauve 'mauve'), -a (lila, 'lilac'), or -en (gouden 'golden'); we therefore generally avoid such adjectives in the rest of the paper.

2.2 Elliptical vs. non-elliptical uses of inflected adjectives

Both inflected and uninflected adjectives have an elliptical use with the indefinite article een, the non-neuter definite article de, and the neuter definite article het, as illustrated in (4):

(4) a. Jan kocht een rode auto, en Karel een blauwe 0. Jan bought a red[+e] car and Karel a blue[+e] 0 'Jan bought a red car, and Karel a blue one.'

b. Jan kocht de rode auto, en Karel de blauwe 0. Jan bought the red[+e] car and Karel the blue[+e] 0 'Jan bought the red car, and Karel the blue one.'

c. Ik las het rode boek, en mijn buurvrouw het

I read thesg.n red[+e] book and my neighborfem thesg.n

groene 0. green[?e] 0

'I read the red book, and my neighbor the green one.'

There is no pronominal one here as in English, but elliptical use of inflected adjectives is subject to the same conditions on use as are DPs headed by one.5 Crucially, the DP must have an antecedent containing an overt nominal that identifies the elided descriptive content.

In the absence of an explicit or implicit antecedent in the preceding discourse, DPs headed by inflected adjectives have a use that is unrelated to ellipsis. In traditional grammars (e.g. Haeseryn et al. 1997; see also Broekhuis et al. 2013), it is common to distinguish between the so-called 'human' construction illustrated in (5a), and the so-called 'abstract object' construction in (5b). We will further argue

5 Sassoon (2013) discusses some facts involving English nationality expressions (e.g. Italian) that indicate that the use of one is somewhat more complex than it might seem at first glance, and that the conditions on its use are sensitive to the animacy of the DP referent. However, as our focus is on Dutch rather than English, and as we suspect that the facts are specific to nationality adjectives, we will not discuss these data here. More generally, we avoid using nationality expressions in this paper, as it seems possible to us that they have acquired a conventionalized nominal use alongside their adjectival one. Note also that there is a residual problem with indefinite neuter nouns in examples like (i):

(i) Ik kocht een groen schrift, en mijn broer een ?rood / ?rode 0. I bought a green notebook, and my brother a red / red[+ej 0 'I bought a green notebook, and my brother a red one.'

We would expect the uninflected form of rood here, as the antecedent is the neuter noun schrift. However, Dutch speakers are uncomfortable with the uninflected form in ellipsis, and some prefer the use of the inflected form in this environment (Kester 1996). Broekhuis et al. (2013) claim that ellipsis is disfavored without schwa, and therefore avoided with neuter nouns altogether. We will leave this problem aside, as it doesn't interact with the main issues addressed in this paper.

that examples like (5c), which Kester (1996) subsumes under the abstract object construction, are in fact syntactically and semantically quite different:

(5) a. De zieke werd thuis verzorgd en verpleegd. the sick[?e] was at-home cared-for and nursed 'The sick person was taken care of and nursed at home.'

b. Van het besprokene en de besluiten wordt een verslag of thesg.n discussed[?e] and the decisions is a report gemaakt.

'A report is made of what was discussed, and the decisions.'

c. Het mooie aan al deze activiteiten is dat ze voor en the beautiful[?e] at all these activities is that they for and door psychologiestudenten worden georganiseerd.

by psychology-students are organized

'The beautiful thing about all these activities is that they are organized

for and by psychology students.'

In the construction in (5a), an inflected human-describing adjective (e.g. rijk 'rich,' blind 'blind,' ziek 'sick') combines with the common gender article de (compare English the blind, the rich, etc.). In the abstract object construction in (5b), the inflected adjective generally takes the neuter article het (possibly preceded by the predeterminer al 'all'), though as we will see, in some cases demonstratives are also found. Note that we find the long form here even with adjectives that do not normally take the schwa, like besproken, which ends in -en (cf. above). English also allows reference to abstract objects by means of adjectives along the lines of (5b). Quirk et al. (1985, p. 424) list the mystical, the best, the sublime, the ridiculous and a few others, but observe that the construction is ''restricted chiefly to certain fixed expressions'' (though see Glass 2014 for data that indicate that the construction is more widespread than Quirk et al.'s characterization would suggest). In Dutch, there is a wide range of such inflected adjectives referring to what we can describe (pre-theoretically) as 'that which is Adj,' for (5b). The meaning of the abstract object construction in (5b) is similar to its English counterpart, but the process is fully productive in Dutch. Though, as we will see in some examples later on, the referent of this construction can sometimes be fairly concrete, we will follow tradition and continue to refer to it as the ''abstract object'' construction in order to distinguish it from the construction in (5c).

Het mooie aan al deze activiteiten in (5c) shares the use of the neuter definite article with the abstract object construction in (5b). However, as we will see, het is the only determiner that occurs in this construction. The intuition is that the construction doesn't refer to some thing or stuff that is attractive, but rather to an attractive feature or aspect of the activities. This interpretation points to a special relation between the inflected adjective and the PP, and indeed the latter is obligatory, while the PP in (5b) is not. The special role played by the obligatory complement motivates our decision to label this construction as involving a relational inflected adjective, as will be discussed in detail in Sect. 5.

We now turn to the syntactic and semantic differences between the three types of inflected adjectives illustrated in (5a-c), formulating a syntax-semantics interface for each of them. We start with the human construction, which is the most straightforward case.

3 The human construction

In English, the human construction is largely restricted to the sort of cases that would be expressed using definite generic plurals in Dutch (6a), but in Dutch we also find the human construction in indefinite generic and existential contexts, in the singular as well as the plural (6b, c, examples from Kester 1996, p. 231), and with a wide range of quantifiers (6c-e):

(6) a. De rijken worden alleen maar rijker. the rich[+e].PL become only just richer 'The rich only become richer.'

b. Een zieke heeft recht op een goede verzorging. a sick[+e] has right to a good care

'A sick person has a right to good care.'

c. Ik zag twee blinden de straat oversteken. I saw two blind[+e].PL the street cross

'I saw two blind people cross the street.'

d. Er komen weinig vreemden in dit dorp, dus there come few strange[+e].PL in this village so het is voor de lokale bevolking een hele belevenis it is for the local population a whole event om ons aan het werk te zien.

for us at the work to see

'Few strangers come to the village, so for the local population, it is quite an event to see us at work.'

e. De meeste armen leven in derde wereldlanden. the most poor[+e].PL live in third world countries 'Most of the poor people live in third world countries.'

In traditional grammars (like Haeseryn et al. 1997), it is generally assumed that the adjective in the human construction has been converted into a noun. One argument for this is that adjectives in prenominal position are not marked for plural (see 3c), but the -n appearing in (6a) and (6c) indicates plurality. That is, without the -n, the inflected adjective refers to a single person, as in (6b). The -n is an irregular plural, as most nouns ending in schwa in Dutch form the plural by means of the affix -s (dame-dames 'lady-ladies'), though -en is found as a plural nominal form with other classes of nouns (stoel-stoelen 'chair-chairs,' student-studenten 'student-students'). Note also that -n is used in special cases to indicate human reference. Certain indefinite plural quantifiers allow forms with and without -n, and in those cases, the form with -n is [?human], whereas the form without -n is [-human]. We see this

contrast in enkele/enkelen 'some,' sommige/sommigen 'some, certain,' vele/velen 'many.' The use of -n with rijke 'rich,' zieke 'sick,' blinde 'blind,' etc. may thus be explained by the human reference of these expressions.

A second argument for the conversion analysis of the adjective is that it also correctly predicts that adjectival modifiers can appear in the construction:

zielige

pitiful[+e]Adj

pitiful

verveelde

bored[?e]Adj

blinde

blind[+_e]

blind person' rijken

rich[+_e].PL

rich people'

Many adjectives double as adverbs, but the adverbial counterpart always occurs in the short form (without -e). Given that zielige 'pitiful' and verveelde 'bored' appear in the long form in (7), they must be adjectives. If blinde and rijken in (7) were adjectives, we might expect only adverbial modification, as adjectives normally do not modify other adjectives. The pattern in (7) therefore supports a nominal analysis of the inflected adjective in the human construction.

However, Kester (1996) challenges the claim that the adjective is converted into a noun, and maintains an adjectival analysis. She observes that inflected human adjectives as in (8c) do not allow for diminutive suffixes, unlike regular nouns (8a, b)6:

a. een vrouw - een vrouwtje

a woman - a woman.DiM

b. a dame - een dametje

a lady -a lady.DiM

c. een blinde - *een blindetje

a blind - *a blind.DiM

The contrast between (8b) and (8c) indicates that the problem with (8c) is not phonological: (8b) shows that regular nouns ending in schwa are compatible with the diminutive suffix. Kester takes the incompatibility with diminutive suffixes to be a major problem for the nominalization approach.

In support of an adjectival analysis Kester argues that the inflected form is compatible with degree modifiers and other adverbs as in (9). Furthermore, they can appear in comparative and superlative forms as in (10), and they can be preceded by adjuncts and complements as in (11) (examples from Kester 1996, p. 232):

6 This is not to say that no inflected adjectives are lexicalized as nouns. A likely example appears in (i). Note, in support of the lexicalization hypothesis, that there is a shift in meaning: the diminutive form has a somewhat pejorative interpretation. Unsurprisingly, as a noun, it admits the diminutive. The point is that such cases are the exception, not the rule.

(i) een blonde - een blondje a blond - a blond.dim

de the

veryAdv

'the very rich'

(10) a.

(11) a.

de the

onlangs recentlyAdv

rijken

rich[+e].PL

gestorvenen died[+e].PL

'the people who died recently'

de the

evenAdv

ziekeren

sicker[+e].PL

'the people that are even sicker'

de the

allerbesten all-best[+ej.PL

'the best of all'

de the

naar to

onze our

maatstaven standards

zeer veryAdv

vermögenden wealthy[+e].PL

'the people that are very wealthy by our standards'

de the

van of

uitkering levenden

social-security living[+e].PL

'the people that are living on social security'

de the

tot to

levenslang lifelong

veroordeelden sentenced[+e].P]

'the people serving a lifelong sentence'

On the basis of these facts, Kester concludes that the human construction in Dutch is best analyzed in terms of an adjective followed by the null noun pro (N-pro). The definite common gender article de identifies the semantic features of N-pro as [?human]. Recall that the Dutch human construction is much freer than its English counterpart. The English human construction is restricted to definite, generic plural uses like (6a). The Dutch human construction also appears in the singular (6b), and in non-generic uses (6c-e). Accordingly, Kester takes the N-pro in the English human construction to be specified as [?human, ?generic, ?plural]. In the Dutch human construction, N-pro only carries the feature [? human], and the special ending -n on the adjective is needed to license and identify it as [?plural] (Kester 1996, p. 238). Multiple adjectives can modify the N-pro, leading to configurations like (7), but the special plural ending -n appears only on the last one (7b). This adjective occurs in a specifier-head configuration with N-pro, and, as argued for other cases as well, specifier-head agreement in overt syntax is crucial for licensing (Kester 1996, p. 239).

We find Kester's (1996) arguments for a null noun in the human construction convincing, but she doesn't discuss the semantics beyond the features on the null noun. We will now be more explicit about the syntax-semantics interface so as to fine-tune the contrast with the abstract object construction and the relational inflected adjectives. Specifically, we propose that the null noun contributes a free variable and the sortal restriction that the value of this variable must be human. We will assume a variant of the ''layered'' approach to DP advocated in Zamparelli (1995), on which the descriptive core of DP (typically contributed by an NP)

denotes a property of kinds, while a functional projection Num(ber) adds number morphology and extensionalizes the NP denotation. On the basis of the syntactic structure in (12), the denotation of the null noun will have the logical representation in (12a): it denotes the property of human kinds. When an adjective modifies this noun, the resulting denotation will be effectively the same as that of the adjective (12b), only with the added condition that the individuals described by the adjective be human (12c). We take inflectional morphology (-e, -en) to sit in NumP; its job is to extensionalize the NP by means of Carlson's (1977) realization relation R (12d, e). In the case of the plural, we add Link's (1983) star operator (*) to the realization relation to indicate that the instances of the kind will be pluralities. In the absence of an overt noun, plural is marked on the adjective as -n, and fused with -e in NumP (12e, g).

(12) [NumP -e(n) [np AP [Npro]]]

a. H [n pro] ]]: kxk[human(xk)]

b. [[ [ap Adj] ]]: kPky[(Adj(P))(y)]

c. [[ [np AP [n pro]] ]]: kxk[AP[?human](xk)]

d. [[-e]]: kPkyAxk[R(y, xk) A P[+human](xk)]

e. [-en]]: kPkyAxk[R*(y, xk) A P[+human](xk)]

f. [[ [-e [np AP pro]] ]]: ky[R(y, xk) A AP[+human](xk)]

g. [ [-en [np AP pro]] ]]: ky[R*(y, xk) A AP[+human]fc)]

This semantics allows the inflected adjective to appear in the singular or the plural, with any kind of count article, determiner, or quantifier, as is in fact the case in Dutch. By way of example, we work out the semantics of rijke(n) in (13). (13a) is the output of (12f) applied to the adjective rijk. This corresponds to the set of rich (atomic) people. (13b) is the output of (12 g) applied to the adjective and denotes the set of pluralities of rich people.

(13) a. [[ [NumP -e [np AP pro]] ]]: ky[R(y, xk) A Rich[+toman](xk)] b. [[ [NumP -en [np AP pro]] ]]: ky[R*(y, xk) A Rich[+human]te)]

The semantics of -n is no different from that of the plural morphology on nouns. We assume Farkas and de Swart's (2010) semantics for the plural marker on nominal expressions, but note that the behavior of plural -n on inflected adjectives could also be compatible with other analyses of plural morphology.

With this basic analysis in hand, we can now dig a bit deeper into the semantics of the human construction with adjectives and adverbs preceding the inflected adjectives. (14) and (15) provide examples in a sentential context. Recall that the long form (with -e) zielige that appears in (14) indicates that it is an adjective. In contrast, ernstig 'serious' appears in the short form (without -e) in (15), indicating that it is an adverb, effectively equivalent to 'seriously':

(14) De vader van Luna heeft de twee zielige zieken opgehaald. the father of Luna has the two pitiful[+e]Adj sick[+e].PL up-picked 'Luna's father picked up the two pitiful sick persons.'

(15) Er zijn twee ernstig zieken overleden kort na de there are two seriousAdv sick[+e].PL died shortly after the inentingen.

vaccinations

'Two seriously sick people died shortly after the vaccinations.'

Preceding descriptive adjectives as in (7) and (14) always have an interpretation where they describe what might be called incidental properties of the referent: (14) conveys that two people who are identified as sick are picked up, and they are further described as pitiful. In contrast, preceding adverbs as in (9b) and (15) have an interpretation that serves to narrow down the kind of individuals picked out by the adjective. Thus, in (15) the two people who are claimed to have died shortly after the vaccinations are identified as of the seriously ill sort.

We take it as a fact that adjectives cannot modify other adjectives. Therefore, zielige cannot modify the adjective zieke directly, within the Adjective phrase that modifies the null pro head noun. In contrast, the adverb ernstig can modify zieke, resulting in ernstig zieke at the NP level, which denotes a property of kinds (of humans) that forms the core of the nominal description. The inflection -e on zieke apparently closes off the NP to any further internal modification, preventing zielige (or any other adjective) from appearing within that NP. Instead, the adjective in (14) must only be able to come in at a later point in the derivation, once Num(ber) has been added. The syntactic structures corresponding to (14) and (15) are thus as in (16):

(16) a. [dp twee [NumP [ap zielige] [№mP -en [np Ldj ziek] [n pro ]]]]] [=14] b. [dp twee [NumP -en [np [ap Ldv ernstig] [Adj ziek]] [n pro ]]]] [=15]

The adverb in (16b) is in a position to form a complex adjective phrase with ziek. Since the effect of adverbial modification occurs within NP (which, recall, denotes a property of kinds), we account for the subkind interpretation. The attachment of zielige above NumP in (16a) entails that the adjective applies after the kind description kx[sick[+human](x)] is closed off to further modification; the adjective thus can only modify the instances of the kind described by NP. Note that the adjective does not get a non-restrictive interpretation, that is, (16a) does not entail that there are only two sick individuals. On the contrary, we can use adjective modifiers of inflected zieke to contrast two sets of sick people, as in (17):

(17) Er liggen twee gemakkelijke, en twee moeilijke zieken there lie two easy[+e] and two difficulty] sick[+e].PL in kamer 10.

in room 10

'There are two easygoing patients in room 10, and two who are difficult to handle.'

In (17), the additional adjectives simply describe the realizations of the kind: in this example, there are four patients in total, two of them with one incidental property, and two with another.

The human construction also allows modification by a relative clause. As illustrated in (18) and (19), respectively, both restrictive and non-restrictive interpretations are available:

(18) De rijken die door hard werken hun rijkdom vergaard the rich[+e].PL who by hard work.INF their riches gathered hebben, bravo. De rijken die hun rijkdom vergaard hebben have bravo The rich[+e].PL who their riches gathered have op de rug van iemand anders...

on the back of somebody else

'The rich who acquired their riches by working hard, bravo. The rich who

acquired their riches by exploiting others...'

(19) De meerderheid van de bevolking kon zich ook geen vlees the majority of the population could REFL also no meat veroorloven. Dit was weggelegd voor de rijken, die vlees afford this was reserved for the rich[+e].PL who meat consumeerden in overvloed.

consumed in abundance

'The majority of the population could not afford meat. This was only possible for the rich, who consumed meat in abundance.'

The relative clause in (18) is used to contrast two groups of rich people who acquired their riches in different ways, so it restricts the denotation of the human construction. The relative clause in (19) has a non-restrictive interpretation, providing additional information about an independently identifiable group of rich people.

These two types of relativization are also straightforwardly expected on our account. The insight that restrictive relative clauses attach to (some projection of) the noun phrase rather than the DP goes back at least to Partee (1975). We take them to attach to NumP, like the adjective in (16a). We analyze non-restrictive relatives as adjoined to DP, following the treatment in Potts (2005) and earlier work by McCawley (1998) and Del Gobbo (2003). This leads to the structures in (20a) and (20b), respectively, for the restrictive and non-restrictive relative clauses in (18) and (19):

(20) a. [dp de [№mP [NumP -en [np [ap rijk] [n pro ]]] [cp die. (restrictive)]]]

b. [dp [dp de [NumP -en [np [ap rijk] [n pro ]]]][cp die. (non-restrictive)]]

In (20a), the relative clause serves to pick out a potentially proper subset of rich people, whereas in (20b) the relative clause serves as the predicate of an independent proposition, applied to the referent of de rijken, providing additional

information about the people identified as rich by the inflected adjective. See Potts (2005, Chap. 3) for the details of the semantic analysis of non-restrictive relatives.

In sum, the human construction involves a null noun, which denotes the property of human kinds. The inflectional marker -e sits in Num, and introduces the Carlsonian realization relation R that yields a set of individuals with the property Adj. The inflection marker -n pluralizes this set. As a result of the combination with the adjective, we obtain a set of people with the property Adj (singular) or Adj* (plural). This expression is of the right type to be taken as the argument of any determiner or article, which accounts for the wide range of determiners and quantifiers in (6). Both adjectival and adverbial modification is possible, depending on the attachment site: adverbials combine with the inflected adjective and build an adjective phrase below Num, whereas modifying adjectives combine at a later stage (after the inflected adjective has combined with the null noun). Both restrictive and non-restrictive relative clauses appear, with the same syntax-semantic interface as in any other DP.

The distributional, syntactic and semantic properties of the human construction are quite straightforward, and although the syntax-semantics interface we presented is new, it will not strike anyone as highly unusual. In contrast, the abstract object and relational constructions are more challenging. In general, these two constructions are not distinguished in the literature, presumably because they are both introduced by the neuter definite article het, and—to our knowledge—no one has delved more deeply into their syntactic and semantic properties. It will be easier to distinguish the two constructions after we have built the syntax-semantics interface for the abstract object construction, to which we now turn.

4 The abstract object construction

Kester (1996) proposes an analysis of the abstract object construction in (5b) in terms of an adjective modifying a null noun pro with the features [?mass, -animate]. In support of the adjectival analysis of the abstract object construction, we find that het + inflected adjective can be preceded by degree modifiers (21a) and other adverbs (21b), can appear in comparative and superlative forms (21c), and can be preceded by adjuncts and complements (21d), all of which we have already seen for the human construction (examples from Kester 1996, p. 244):

(21) a. het zeer onverwachte

the veryAdv unexpected[?e]

'that which is/was very unexpected'

b. het onlangs gebeurde the recentlyAdv happened[?e] 'that which recently happened'

c. het allerduurste

the all-expensive-superlative[?e] 'that which is super-expensive'

d. het voor mij onacceptabele the for me unacceptable^] 'that which is unacceptable to me'

The inflected adjective in the abstract object construction is also not compatible with diminutives, though this could also be due to the fact that the diminutive suffix is restricted to count nouns, and Kester qualifies the null noun in these constructions as mass. On this analysis, then, the different articles in each construction identify different semantic features of pro: the common gender article de identifies animate or human count pro, whereas neuter het identifies inanimate or non-human mass pro.

Besides the definite article het, we find the neuter form of the proximal and distal demonstrative articles dit (22a) and dat (22b) in the abstract object construction. The predeterminer al 'all' combines either with the definite (22c) or with the demonstrative (22d). It indicates maximality, as opposed to uniqueness, in the interpretation of the determiner.

(22) a. Maar toch merk je dat dit onzekere in zijn karakter zit.

but yet notice you that this uncertain^ in his character sits 'But yet you notice that this uncertainty is in his character.'

b. Ze haatte dat onverschillige in zijn houding. she hated that indifferent[+ej in his attitude 'She hated that indifference in his attitude.'

c. We willen een gemeenschap zijn die met beide benen in de we want a community be that with both legs in the samenleving staat, met al het goede en mooie verder society stands with all the good[+e] and beautiful[+e] on gaat en werkt aan al het moeilijke dat er is.

goes and works on all the difficult^] that there is 'We want to be a community that stands with both feet in society, continues all that is good and beautiful, and works on everything that is difficult.'

d. Als hij er over zou beginnen zou jij misschien if he there about would start would you maybe herinnerd worden aan al dat nare van gister. reminded become of all that nasty[+e] from yesterday 'If he would start talking about it, it might remind you of all that nasty stuff from yesterday.'

However, we do not think these parallels with the human construction are sufficient to motivate the same null noun analysis. An important difference with the human construction is that the abstract object construction can only be headed by the definite article or a demonstrative (23a, b), and not by the indefinite article, a numeral or a quantifier of any sort:

(23) *een onzekere / *twee moeilijke(n) / *ieder nare /

an uncertain[?e] / two difficult[?e](.PL) / every nasty[?e] /

*een beetje onverschillige / *veel moeilijke / *weinig goede /

a bit indifferent[?e] / much difficult[?e] / little good[?e] /

*het meeste mooie

the most beautiful[?e]

If abstract object constructions involved a null mass noun, as proposed by Kester (1996), this would explain the incompatibility with count indefinites, numerals, plurality and distributive ieder 'every.' However, it would not explain why the adjective does not combine with partitive expressions such as een beetje 'a bit' that take mass nouns (een beetje zout 'a bit of salt'), or indefinites like veel 'many/much' and weinig 'few/little' that have both count and mass uses (veel boeken 'many books,' veel olie 'much oil'; weinig kinderen 'few children,' weinig peper 'little pepper'). Although we find al het 'all the' with a maximality interpretation in (22c), we do not find het meeste 'most,' even though meeste is otherwise compatible with both mass and count nouns (het meeste water 'most of the water,' de meeste studenten 'most of the students'). The null noun analysis of the human construction provided an attractive account of the wide range of possible determiners and quantifiers in examples like (6). The lack of this freedom with the abstract object construction therefore provides a strong argument against a null noun analysis.

Instead, we propose that the null pro that Kester saw evidence for in this case is a pronominal DP, not a null noun, and contributes a free variable that is sortally restricted to inanimate mass reference. The inflected adjective predicates over this free variable, creating an open proposition of the form Adj(x). This difference is small but crucial. Syntactically, the proposal implies that the inflected adjective builds a small clause in combination with pro, rather than building an NP. The small clause is embedded under het in a DP structure.7 This reflects our intuition that this structure bears a strong semantic similarity to a free relative: 'that which is Adj.' The use of the definite article fits with the observation that free relatives are definite descriptions (Dayal 1995; Jacobson 1995; and others). The PP in de mens 'in the human being' functions as a locative modifier of the small clause (SC). The result is shown in (24):

7 We should point out that adjectives with predicative uses normally appear in the short form, rather than the long form:

(i) De tafel is rood /*rode. the table is red / red[?e] 'The table is red.'

(ii) Ik vind dit mooi /*mooie.

I find this nice / nice[?e] 'I find this nice.'

What unites the examples in (i) and (ii) in contrast to the constructions we discuss in the text is that the adjective in them is selected by a higher predicate (here is, vind). Though we must leave this issue for future research, we suggest that the inflection marker -e in the abstract object construction may be the counterpart of the higher predicate in (i) and (ii).

(24) [dp het [sc [Ldj goede] [dp pro]][pp in [dp de mens]]]]

The definite article het introduces closure over the open proposition. For its semantics, we adapt Hinterwimmer's (2013) implementation of the iota operator in his analysis of free relatives. Hinterwimmer provides both extensional and intensional variants, which we use for het in (25c) and (26c), respectively. Here s is a situation or world variable (we use an intensional version of the adjectives here). The extensional definition of het is used for non-generically interpreted DPs, as in (25), whereas the intensional definition is used when the DP has a generic interpretation, as would be possible with (26). The difference is simply that the extensional definition introduces a specific situation or world variable, indicated as s1, which saturates the lambda-bound situation variable and gets valued by the assignment function g with respect to which the sentence is evaluated (25c, d). Technically, the iota operator forces lambda abstraction over the variable x[-anim, -count] of the open proposition in the (a) examples to have taken place prior to application (25d, 26b). As in (13) above, P* (e.g., here, Discussed*) is the closure of P under sum formation to account for mass nouns (Link 1983). The PP is added as an intersective modifier in the (b) examples.

(25) het besprokene in de vergadering the discussed[+e] in the meeting 'what was discussed in the meeting'

a. [[besprokene pro]]g,s: ks[Discussed*(x[-anim, _count])(s)]

b. [besprokene pro in de vergadering[\g:

ks[Discussed*(x[_anim, -count])(s) A In(x[_anim, -count], the meeting)(s)]

c. [het! si]]g: kP(e,<s,i»lx[P(x[-anim, -count])(g(s1))]

d. [het1 si besprokene pro in de vergadering[\g:

kP<e,<s,i»ix[P(x[_anim -count])(g(s1))](kxks[Discussed*(x[-anim, -count])(s) A In(x[_anim, -count], the meeting)(s)])

= tx[Discussed*(x[_anim, -counqXg^O) A In(x[ -anim, -count],

the meeting)(g(s1))]

(26) het goede in de mens

the good[+e] in the human-being 'what is good in man'

a. [[goede pro]]g: ks[Good*(x[_anim, -count])(s)]

b. [[goede pro in de mens]\g: ks[Good*(x[-anim, _count])(s)

A In(x[_anim, -count], man)(s)]

c. [[het2]]g = kP (e,(s,t» ksix[P(x[-anim, -count])(s)]

d. Hhet2 goede pro in de mens[\g: kstx[Good*(x[-anim, _count])(s)

A In(x[_anim, -count], man)(s)]

The DP in (26d) has the semantic type of an individual concept and corresponds to the denotation that Chierchia (1998) provides for kind-denoting bare plurals. The intensional definition of het in (26c) returns a function that for each possible world or situation returns the maximal sum individual satisfying the predicate in that world/situation. In combination with the semantics of the PP as a conjunctive

modifier, introducing a general locative relation between the variable x and the denotation of de mens, we obtain the kind-denoting interpretation paraphrasable as 'everything that is good in humankind.' In contrast, the extensional definition in (25) returns the maximal sum individual that satisfies the predicate with respect to a contextually given situation or the world of evaluation. With the modifier added in, this yields a non-generic interpretation paraphrasable as 'what was discussed in the meeting.'

Under the analysis of the abstract object construction as similar to a free relative, the restriction to the definite or demonstrative article follows from the maximal sum interpretation. The predeterminer al can be added because it further emphasizes maximality. The abstract object construction does not contain a NumP, and thus does not involve reference to a property that just any determiner or quantifier could take as an argument. This accounts for the infelicity of other quantifiers in this configuration (cf. 23), in contrast to the much wider range of determiners we find in the human construction (cf. 6).

We might expect the abstract object never to accept adjectival modifiers and relative clauses, unlike the human construction, since the Adj + pro is a small clause rather than an NP or NumP and thus has a propositional, rather than property-type, semantics that should be incompatible with adjectival modification. In fact, Kester (1996) claims that the abstract object construction does not accept adjectival modifiers. It turns out that restrictively interpreted descriptive adjectives are, indeed, excluded: only the short (adverbial) forms are possible in (27). However, a small subset of adjectives is attested: the inflection on enige, andere, eerste, vermeende, mooie, vele, and weinige in (28) shows that these modifiers are not adverbs8:

(27) a. het recent / *recente besprokene

the moralAdv / moral[+e]Adj good in the other 'the moral good in the other'

(28) a. Het enige bijzondere in Pisa is de toren en the only special[?e] in Pisa is the tower and het is daar enorm toeristisch, dus wellicht dat je it is there very touristy so maybe that you liever toch een andere stad bezoekt?

rather yet another city visit

'The only special thing in Pisa is the tower, and it is very touristy there, so maybe you prefer to visit another town after all?'

8 These are examples from the internet, tested against native speaker intuitions. Incidentally, Ellen-Petra Kester (p.c.) accepts them as well.

the recentAdv / recent[? 'the matters recently b. het moreel / *morele

the recentAdv / recent[

[+e]Adj

discussed discussed'

goede in de ander

b. Een blog over eten, drinken, genieten, muziek, films, voetbal a blog about eat.iNF drink.iNF enjoy.iNF music movies soccer en al het andere leuke in het leven!

and all the other[+e] nice[+e] in the life

'A blog about eating, drinking, enjoying, music, movies, soccer and all

the other nice things in life!'

c. Met Spike was het ook anders, die kwam uit zo'n ellende with Spike was it also different DEM came out such misery en ik was het eerste leuke in zijn leven.

and I was the first[+e] fun[+e] in his life

'With Spike it was different, he came out of such misery, and I was the

first fun thing in his life.'

d. In die zin vind ik het vreemd, en ook heel jammer, dat in that sense find I it strange and also very pitiful that een discussie die ik startte over het vermeende

a discussion that I started about the alleged[+e] antisemitische van Luther op slot is gezet. anti-semitic[+e] of Luther on lock is put 'In that sense, I find it strange, and also a pity, that a discussion that I started on the alleged anti-Semitism of Luther has been put on hold.'

e. En ook een paar echte Woerdenaren die het and also a couple real Woerden-inhabitants who the mooie oude in Woerden zelf ter harte nemen. beautiful[+e] old[+e] in Woerden self to heart take kaaspakhuisje hier, boerderij daar. cheese-warehouse.DIM here farm there

'And also a couple of real inhabitants of Woerden, who take to heart the beautiful old things in Woerden. Here a cheese warehouse, there a farm.'

f. Waarna we weer vrolijk de bus ingingen en de tweede whereafter we again happily the bus in went and the second helft van de reis voldaan van het vele goede van de half of the trip satisfied of the many[+e] good[+e] of the hele dag met een dutje afsloten.

whole day with a nap off-closed

'After which we happily got back into the bus, and completed the second half of the trip satisfied by the many good things of the day.'

g. een gedeelte van het weinige nieuwe onder de zon a part of the little[+e] new[+e] under the sun 'a part of the few new things on Earth'

Crucially, though, the categories of adjectives that we find are limited to DP-internal only (28a), other (28b), ordinals like first/last (28c), intensional adjectives (28d), non-restrictively used evaluative adjectives (28e) and adjectival many/much and few/little (28f, g). The evaluative adjectives can be analyzed extending Potts' (2005) analysis of epithets (e.g. damned in the damned dog). Like epithets, non-restrictive adjectives do not form part of what is at issue in utterances containing them. Potts

accounts for this by composing the predication of the adjective with the entity denoted by the DP on the whole as an independent, conventionally implicated proposition in parallel to the composition of the rest of the descriptive content of the DP. Thus, the modified small clause in (28e), setting aside the PP for simplicity, would be analyzed as in (29)9:

(29) a. [[oude pro]]: ks[Old*(x)(s)]

b. [[mooie oude pro]]: At issue content: ks[Old*(x)(s)]

Conventional implicature: ks[Beautiful*(x)(s)]

If, as suggested in footnote 7, the inflection serves to license predication in the absence of an overt licensing predicate such as the copula, we can make sense out of the fact that it appears on the adjective in this case.

Adjectival vele and weinige are necessarily interpreted as cardinal or amount oriented (de Hoop 1992). Just like the evaluative adjectives, their use in the abstract object construction is non-restrictive in nature, so we take them to be composed in parallel to the descriptive adjective, rather than interacting with it, as in (29). We do not work out the formal details here, but the approach allows adjectival vele and weinige in (28f, g) to describe the maximal sum of everything that is Adj as a high amount.

Interestingly, the remaining classes of adjectives have in common that they can all be analyzed as propositional, rather than predicate, modifiers. Consider intensional adjectives first. Larson (2002) argues for a general theory of intensionality as what he calls a 'sentential' phenomenon. The intuition, as we understand it, is that intensional predicates always express relations to thoughts or proposition-like objects. On such a view, which we adopt here, intensional adjectives such as vermeende 'alleged' express a relation to a proposition about the applicability of a(n adjectival) property to some individual. Though Larson does not provide a fully explicit semantics for intensional adjectives, he says the following:

Consider the case of fully intensional adjectives. We suggested a sententialist analysis of their corresponding adverbs in which the latter combined with a clauselike VP: one that contained both subject and predicate. To duplicate this idea, we would evidently need to view the nominal in cases like alleged dancer as containing a subject, which, for concreteness, we might construe as a silent pronoun (pro):

Olga interviewed an alleged [№ pro dancer].

'Olga interviewed an x such that it is alleged that x is a dancer'

(Larson 2002, pp. 253-254)

Though Larson acknowledges that in the ordinary case it is not trivial to justify the sort of syntax he proposes for an alleged dancer, for the Dutch cases that interest us, this is exactly what the small clause analysis provides us. The composition of the DP in (28d) (leaving out the PP for the sake of simplicity) can thus be sketched as in (30):

9 For readability, hereafter, unless necessary, we suppress the assignment function and situation indices on the denotations as well as the sortal restriction subscripts on the variable contributed by pro.

(30) a. [vermeende^Dg: kphs,t)ks[Alleged(p)(s)]

b. [[antisemitische pro]]g: ks[Anti-Semitic*(x)(s)]

c. [vermeende antisemitische pro]]g: ks[Alleged(Anti-Semitic*(x)(s))]

d. [het sj vermeende antisemitische pro]]g: tx[Alleged(Anti-Semitic*(x)(g(s1))]

The semantics of vermeende in (30a) allows it to combine directly with the small clause in the abstract object construction as an open proposition in (30b), as shown in (30c). (30d) gives us the maximal sum of what is alleged to be anti-Semitic about Luther in the world under consideration. With respect to the presence of the inflection, we suggest that, as with the evaluative adjectives, it is present to license predication, in this case between the adjective and its propositional argument. We extend the same explanation to account for the presence of inflection on enige 'only' and ordinals, to which we now turn.

Consider first enige. The classic analysis of English only in Rooth (1985) treats it as a focus-sensitive propositional operator that quantifies over the set of alternative propositions that constitutes the focus semantic value of its argument. For example, the set of alternatives for John introduced [Bill]F to Sue is the set of all propositions obtained by substituting all of the alternatives x to the focused constituent Bill (31a) into the argument position occupied by Bill (31b). Only quantifies over this set of alternatives, and asserts that the sole value of x that makes the proposition true is Bill (31c):

(31) John only introduced [Bill]F to Sue

a. Alt(Bill): {a, b, c, d, ...}

b. Alt(John introduced [Bill]F to Sue):

{p : Ax e Alt(Bill)[p = Introduce(s)(x)(j)]}

c. Introduce(j,b,s) A Vp[p e Alt(Introduce(s)(b)(j)) A True(p) ? p = Introduce(s)(b)(j)]

We can give exactly the same sort of semantics to enige in the abstract object construction. The only difference is that the alternatives over which enige quantifies are determined by the different possible values of the free variable x in the open proposition corresponding to the small clause; we also intensionalize the semantics of enige (vs. what is sketched for only in 31c) to fit with the rest of our analysis of the DP. By way of example, we work out the analysis of the het DP from (28a) in (32):

(32) Het enige bijzondere in Pisa is de toren. the only special[+e] in Pisa is the tower

a. [enige]]g: kp^ks^s) A Vq[q e ks'[Alt(p)(s')] A q(s) ? q = p]]

b. [bijzondere pro]]g: ks [Special*(x)(s)]

c. Alt(bijzondere pro): {q: Ay e Alt(x)[q = ks' [Special*(y)(s')]]}

d. [enige bijzondere pro]]g:

ks[Special*(x)(s) A Vq[q e Alt(ks' [Special*(x)(s')]) A q(s) ? q = ks'[Special*(x)(s')]]

e. Hhetj sj enige bijzondere pro]]g: tx[Special*(x)(g(s1)) A

Vq[q e Alt(ks'[Special*(x)(s')]) A q(g(s1)) ? q = ks'[Special*(x)(s')]]

The analysis of andere in (28b) is similar, except that it involves existential, rather than universal quantification over alternatives.

For the analysis of (28c), we need a semantics for ordinals. To start, slightly adapting a proposal in Yee (2010), we rely on the notion of well-ordering in (33):

(33) A set X is well-ordered by < iff

a. X linearly ordered by <

b. VY.Y ( X [Y = 0 ? Ay e Y.Vz e Y.y < z]

In prose: every non-empty subset of a linearly ordered set has a first/least element. The intuition we have about ordinals is that they combine well-ordering with the exclusion of other values, a property they share with only. The difference is that alternatives have the structure of a well-ordered set, as indicated by the subscripted Altwo in (34):

(34) a. [eerste]]: kp<s,t)ks[p(s) A Vq[q e ks'[Altwo(p)(s')] A q(s)

? p(s) < q(s)]] b. [laatste]]: kphs,t»ks[p(s) A Vq[q e ks'[Altwo(p)(s')] A q(s) ? q(s) < p(s)]]

The denotation of (28c), repeated in (35), can now be spelled out as in (35a-d), again using the extensional definition of the iota operator given the episodic context:

(35) Het eerste leuke (in zijn leven) the first[+e] nice[+e] (in his life)

a. [leuke pro]]g: ks[Nice*(x)(s)]

b. Altwo(leuke pro): {q: Ay e Alt(x)[q = ks'[Nice*(y)(s')]]}, a well-ordered set.

c. Heerste leuke pro]]g: ks[Nice*(x)(s) A Vq[q e Altwo(ks'[Nice*(x)(s')]) A q(s) ? Nice*(x)(s) < q(s)]]

d. [het1 s1 eerste leuke pro]]g: tx[Nice*(x)(g(s1)) A Vq[q e Altwo(ks'[Nice*(x)(s')] A q(g(s0) ? Niee*(x)(g(s{)) < q(gis{))]]

The ordering of the set of alternative values for p, the open proposition denoted by the abstract object construction, is strongly context dependent. In the case of (35b), we can assume that the order is identified with the linear structure assumed for the time axis. The first nice thing in Spike's life is then the nice thing that happened earlier in time than any other nice thing that happened in his life. However, we can also imagine more abstract orderings, e.g. in terms of importance. Given that the exact interpretation depends on the context, we do not think it needs to be fixed as part of the semantics of the ordinal.

In conclusion, the small clause analysis we propose correctly predicts that restrictively interpreted descriptive adjectives are not acceptable in the abstract object construction (cf. 27). The empirical data confirm that adjectival modification

is restricted to evaluative adjectives, adjectival much/little, intensional adjectives, only, other, and ordinal first/last (cf. 28). These adjectives all have either a non-restrictive semantics or a propositional semantics that combines easily with the free relative interpretation of the abstract object construction. We take this to provide further support for the small clause analysis, and against the null noun analysis used for the human construction.

Kester (1996) does not discuss the possibility of modification by relative clauses for the abstract object construction. Recall that both restrictive and non-restrictive relative clauses are found in the human construction (cf. 18, 19). We find both in the abstract object construction as well:

(36) a. Al het vervelende dat hoort bij zijn ziekte kon hij die

all the annoying[+e] that comes with his illness could he that dag even vergeten, en dat is denk ik precies de day shortly forget and that is think I exactly the bedoeling voor zo'n jochie. intention for such-a kid

'All the annoying things that come with his illness, he could forget for a moment that day, and that is exactly what such a kid needs, I think.' b. Japanners die de galerie bezochten, verbaasden zich over Japanese who the gallery visited surprised refl about het gewone dat van der Kleij fotografeerde. the ordinary[+e] that van der Kleij photographed 'Japanese visitors of the gallery were surprised about the ordinary things that van der Kleij took pictures of.'

(37) a. Want het goede dat ik wil, doe ik niet, maar het kwade,

for the good[+e] that I want do I not but the evil[+e] dat ik niet wil, dat doe ik. that I not want that do I

'For good, which I want, I do not do, but evil, which I do not want, that I do.'

b. Het is wel goed dat we weer eens even niet enkel het it is though good that we again once just not only the verdrietige of sombere naar boven halen maar ook het leuke sad[+e] or gloomy[+e] to up hold but also the nice[+e] waarvan we genoten hebben. whereof we enjoyed have

'It is good though that for once we do not just remember the sad and gloomy things, but also the nice things, which we enjoyed.'

The relative clauses in (36) restrict the denotation of the abstract object to annoying things that come with the illness in question (rather than annoying things in general) and ordinary things that van der Kleij took pictures of (rather than ordinary things in general), respectively. In contrast, the non-restrictive relative clauses in (37) provide elaborations on the abstract objects denoted by

het goede, het kwade, het leuke. Note that the relative clause in (37b) is introduced by a so-called R-pronoun, a wh-expression which shows up as a relative pronoun linked to the complement of a preposition (van Riemsdijk 1985). We can analyze the non-restrictive relative clauses as attached at the DP level, just as we did for the human construction. For the restrictive relative clauses modifying inflected adjectives in the human construction, we posited that they attach at a lower level. However, given the small clause analysis we have adopted for the abstract object construction, we cannot use exactly the same semantics as we would for a property-type modifier of NumP. Instead, we hypothesize that what differentiates relative clauses from adjectives is the fact that they, like the small clause formed in the abstract object construction, semantically correspond to an open proposition, as in (38b). This allows for a conjunctive analysis of their semantic composition, as in (38) (we gloss over the other details of the internal semantics of the relative clause, as they are orthogonal to the main argumentation):

(38) het gewone dat van der Kleij fotografeerde the ordinary[+e] that van der Kleij photographed 'the ordinary things that van der Kleij took pictures of'

a. [[gewone pro]]: ks[Ordinary*(x)(s)]

b. [dat van der Kleij fotografeerde[\: ks[Photographed*(x)(vdK)(s)]

c. [gewone pro dat van der Kleij fotografeerde[\: ks[Ordinary*(x)(s) A Photographed*(x)(vdK)(s)]

We conclude that Kester (1996) was essentially correct in observing that ordinary descriptive adjectival modification is unavailable in the abstract object construction, and that both this restriction and the exceptions to it are naturally accounted for under the small clause syntax, with its corresponding semantics in terms of open propositions.

5 The relational inflected adjective construction

We now turn to the last of the three constructions, which we have dubbed the relational inflected adjective construction. At first sight, there are many similarities between the abstract object construction and the relational inflected adjective, to the point that Kester (1996) and Broekhuis et al. (2013) do not distinguish the two. However, there are also crucial differences in their distribution and semantic properties that lead us to reject for the relational adjective construction the free relative-like semantics that we posited for the abstract object construction. In Sect. 5.1, we focus on distributional similarities and differences between the abstract object construction and the relational inflected adjective. Section 5.2 develops the syntax and semantics of relational inflected adjectives.

5.1 The relational inflected adjective as a separate construction

We begin by reviewing the similarities between the two constructions. First, as in the abstract object construction, the inflected form in the relational construction is really an adjective. For example, it takes adverbial rather than adjectival modification (39a-c). In (39a) we find the adverbial degree modifier al te 'all too'; intens 'intense' in (39b) qualifies as an adverb, as shown by the fact that it cannot bear inflection in a context where an adjective would have to bear it (cf. 2 above). The adjective can also appear in synthetic as well as analytical comparative/superlative forms (39c, d):

(39) a. Het al te zure van citrusvruchten is in deze confiture

the all too sour[+e] of citrus-fruits is in this jam

verdwenen.'

disappeared

'The all too sour (aspect) of citrus fruits is lacking in this jam.'

b. Zonder licht, om het intens / *intense rode van de without light in-order the intenseAdv / intenseAdj red[+e] of the ondergaande zon niet te verstoren.

setting sun not to disturb

'Without light, so as not to disturb the intense redness of the sunset.'

c. De acteerprestaties zijn eigenlijk het minst interessante aan the acting-performances are in fact the least interesting[+e] at deze film.

this movie

'The acting is in fact the least interesting aspect of this movie.'

d. Hoe hij gestalte geeft aan deze koning vind ik nog het how he portrait gives of this king find I yet the sterkste aan de hele film.

strongest[+e] at the whole film

'The way he portrays this king, I find to be the strongest feature of the entire movie.'

Second, just like the abstract object construction, the relational inflected adjective resists indefinite articles, quantifiers and plurality (40):

(40) *een / *ieder vreemde / *de vreemdes / *de vreemde(n) van

a / every strange[+e] / the strange[+e].PL / the strange[+e](.PL) of deze situatie (. ) this situation (. )

Third, the relational inflected adjective shares the restrictions on adjectival modification with the abstract object construction. Recall that we only find short (adverbial) modifiers in (27), not long (adjectival ones), and only a few subclasses of adjectival modifiers appear in (28). We find the same restrictions with the relational inflected adjective construction: enige (41a), andere (41b), ordinals (41c),

intensional adjectives (41d), and non-restrictively used evaluative adjectives (41e, f), though we note that the latter two sorts of adjectives have turned up extremely rarely in corpus searches.

(41) a. Het enige leuke aan niet naar school kunnen gaan is niet the only nice[+e] at not to school can.INF go.INF is not naar school gaan. Voor de rest is het vooral 'n hoop to school go.INF for the rest is it especially a heap administratie naast ander werk. administration next other work

'The only nice thing about not being able to go to school is not going to school. For the rest, it is mostly a lot of administration and other work.'

b. Het andere leuke aan schilderen is dat ik zelf altijd the other[+e] nice[+e] of paint.INF is that I self always het meest onder de verf zit.

the most under the paint sit

'The other nice thing about painting is that most of the paint ends up on me.'

c. Het eerste vervelende van de maatregel is dat het voor the first annoying[+e] of the measure is that it for de scholen lastig is om daar rekening mee te houden. the schools tricky is to there account with to take 'The first annoying thing about the measure is that it is tricky for the schools to take that into account.'

d. ...bijvoorbeeld over het vermeende goede van hele hoge for-example about the alleged[+e] good[+e] of very high eiwitpercentages

protein-percentages

'. for example about the alleged goodness of very high protein percentages'

e. Dit is een jonge African Spoonbill waaraan goed te zien is this is a young African Spoonbill which-of good to see is dat hij nog niet op kleur is. Hij is nog wat flets, en that he yet not on color is he is still somewhat pale and heeft niet het mooie rode van een volwassen exemplaar. has not the beautiful[+e] red[+e] of an adult exemplar 'This is a young African Spoonbill, which is not yet showing his full colors. It is still somewhat pale and doesn't have the beautiful red color of an adult exemplar.'

f. Dit is een heerlijke couscous met het kruidige vlees en het this is a delicious couscous with the spicy meat and the lekkere zoete van de pruimen en abrikozen. delicious[+e] sweet[+e] of the plums and apricots.

'This is a delicious couscous with the spicy meat and the delicious sweet taste of the plums and apricots.'

Despite these similarities, the relational inflected adjective construction differs in a few important respects from the abstract object construction, which leads us to develop different analyses of the two constructions.

First, we observe that the relational inflected adjective is strictly restricted to the definite article het, even to the exclusion of demonstratives like dit and dat and the predeterminer al (cp. 22). We see this in the impossibility of creating variants of examples like (1c) or (5c) with demonstratives and al in (42a). Moreover, we cannot use adjectival vele or weinige in this configuration (42b, c). Note that this does not reflect a general incompatibility with a degree interpretation of the relational adjective, as the felicity of the adverbial use of weinig in (42c) illustrates.10

(42) a. Het/ *dit/ *al het vreemde van de situatie vind ik dat

the/ this/ al the strange[+e] of the situation find I that

politieke partijen het maar laten gebeuren.

political parties it just let happen

'The / *this / *all the strange thing about the situation, I find, is that

political parties just let it happen.'

b. *het vele goede van zwemmen in de winter

the much[+e] good[+e] of swimming in the winter

c. Het weinig / *weinige vreemde van de situatie the little / little[+e] strange[+e] of the situation 'the not very strange aspect of the situation'

Intuitively, al is blocked in (42a) because the DPs do not refer to potentially maximal parts of something, which also suggests that we are not dealing with a free relative semantics here. Vele and weinige in (42b, c) are blocked because they can only be ascribed to measurable entities, but relational inflected adjectives are not perceived as mass-like. The difference between the abstract object construction and the relational inflected adjective is particularly clear when we consider the interpretation of examples involving inflected adjectives that describe concrete properties, such as the examples in (43) with color and taste adjectives:

(43) a. Het rode van de aardbeien, het witte van de

The red[+e] of the strawberries the white[+e] of the mascarpone en de slagroom en de groene mintblaadjes mascarpone and the cream and the green mintleaves.DIM.PL kleurden prachtig bij elkaar. colored beautifully with each-other 'The red (aspect) of the strawberries, the white (aspect) of the mascarpone and the cream and the green mint leaves are a wonderful color combination.'

10 Dutch weinig has a determiner use (both mass and count), an adjectival use, and an adverbial use. Dutch veel shares with weinig the determiner and adjectival use, but doesn't have an adverbial use. See Doetjes (1997) for discussion of quantifiers and categories.

b. Het bittere van het bier is een mooi contrast met het the bitter[+e] of the beer is a nice contrast with the zoete van de mout.

sweet[+e] of the malt

'The bitter (aspect) of the beer is a nice contrast with the sweet (aspect) of the malt.'

The het DPs in (43a) do not refer to colored parts of the strawberries or mascarpone, or to colored stuff in/on them, as would be expected if we extended the mass-like semantics of the abstract object construction to relational inflected adjectives. Nor (even more clearly) do those in (43b) refer to the parts of the beer or malt with a particular taste. Rather, they describe properties of the object contributed by the PP, understood as a whole.11 Adding al to these DPs, or using a demonstrative instead, if felicitous at all, would imply, for instance, that the strawberries and mascarpone could have had parts that were not red and white, respectively, or that the beer and the malt could have parts that were not bitter and sweet, respectively. In other words, the reference in the relational inflected adjective construction is truly unique, as opposed to the maximal reference of the abstract object construction.

Second, and relatedly, we observe that restrictive relative clauses do not appear with relational inflected adjectives, although they are fine with the abstract object construction. (44a) shows that we cannot replace the PP complement of a relational inflected adjective with a restrictive relative clause, and (44b) that we cannot add one to a relational inflected adjective with the PP.

(44) a. Het zoete van de aardbeien / *dat de aardbeien toevoegen the sweet[+e] of the strawberries / that the strawberries add wordt getemperd door het frisse van de wilde becomes moderated by the fresh[+e] of the wild bessen / *dat de wilde bessen hebben. berries / that the wild berries have 'The sweetness of the strawberries/*that the strawberries add is moderated by the freshness of the wild berries/*that the wild berries add.' b. *het zwarte van de nacht dat al het licht absorbeert the black[+e] of the night that all the light absorbs

In contrast, non-restrictive relative clauses do seem to be possible in principle, though there are few attested examples. They typically do not appear with the relative pronoun dat 'that,' but rather with the so-called R-pronoun waar (cf. also 37b).12

11 See McNally and de Swart (2011) for some discussion of how the DPs in (43) differ both from uninflected color/taste nouns and deadjectival derived nominals like roodheid 'redness.'

12 It may be the case that relative clauses with dat 'that' are not felicitous because its use creates agreement problems between het and dat/die that do not show up with the R-pronoun waar. We leave this open for further research.

(45) a. Het mooie van Twilight waar ik WEL naar uitkijk

the beautiful[+e] of Twilight where I part to out.look zijn altijd de reacties op recensies en persberichten are always the reactions to reviews and press releases hier op Filmtotaal. here at Filmtotal

'The beautiful thing about Twilight, which I AM looking forward to, is always the reactions to reviews and press releases here on Filmtotal.' b. Nu komt het vervelende aan mediteren waar elke now comes the annoying[+e] at meditate.INF where every beoefenaar mee te maken krijgt. practitioner with to make gets

'Now comes the annoying thing about meditation, which everyone practicing it has to deal with.'

The relative pronoun waar 'where' in (45a) does not refer to Twilight but rather picks up on the beautiful thing about Twilight as what I am looking forward to. Similarly, what everyone has to deal with in (45b) is the annoying aspect of meditation, not meditation in general. Thus, unlike restrictive relative clauses, non-restrictive relative clauses are not excluded, although they are rarely attested. The fact that restrictive relatives are impossible indicates that the adjective extension is subject to a uniqueness condition in the relational construction and cannot be interpreted as a mass with parts that could be distinguished via added information from a restrictive relative, as we have proposed for the abstract object construction.

A third indication of the special status of relational inflected adjective construction resides in the observation that the PP in the construction has a different status from that in the abstract object construction. In the abstract object construction, the PP is strictly optional. Examples without PPs appear in (5b), (22c), (36) and (37). When a PP does appear in the abstract object construction, it can be introduced by any number of locative prepositions, such as in ('in,' 22a, b), van ('of,' 22d, 28d, 43), bij 'with,' op 'on,' rond 'around,' or aan 'at,' as we see in the examples in (46):

(46) a. De kunst is om zo te leven dat we dit goede bij

the art is to so to live that we this good[+e] with elkaar tevoorschijn roepen. each-other forward call

'The art is to live in such a way that we bring out that which is good in each other.'

b. Achteraf gezien, had ik beter nog een zomerlang genoten back looked, had I better still a summer-long enjoyed van al het leuke met mijn man, vele fiets-of all the nice[+e] with my husband, many bike en leuke wandeltochtjes... and nice hiking-tours

'Looking back, I would have been better off spending another summer enjoying all the nice things with my husband, making bike and fun hiking tours.'

c. Het zijn mild geformuleerde stellingen, die ons tot tevredenheid it are mildly formulated theses that us to contentment manen en oproepen te genieten van al het goede

urge and up-call to enjoy of all the good[+e] rond ons heen. around us part

'They are mildly formulated theses, which urge us to be content and call upon us to enjoy all the good things around us.'

d. Voor al het mooie op deze aarde hoef je niet ver for all the beautiful[+e] on this earth need you not far te reizen.

to travel

'For all the beautiful things on this earth, one does not need to travel far.'

e. Hier vindt je al het leuke aan een vakantie: goed here find you all the nice[+e] at a vacation good onthaal, lekker ontbijt, leuke kamers.

reception delicious breakfast nice rooms

'Here you find everything that is nice about a vacation: good reception, delicious breakfast and nice rooms.'

The PPs in these examples contribute the location (e.g. each other in 46a or the earth in 46d) where we can find the abstract object (e.g. that which is good or beautiful). Thus, we posited in Sect. 4 that the PP is a locative adjunct and gets a conjunctive analysis as a regular modifier (cf. 25 and 26).

With the relational inflected adjective construction, the PP complement is most often introduced by van 'of' (47a, e), aan 'at' (47b, d), and more by rarely bij 'with' (47c) or met 'with' (47f). More importantly, the status of the PP in the relational inflected adjective construction is, we maintain, that of an argument rather than an adjunct, as it is needed to identify the referent of the DP. This is perhaps most clearly illustrated by examples in which the complement of the preposition is a nominalized infinitive (47a-c, f), but we also see it with nouns denoting events or properties (47d, e) in the PP complement of the relational inflected adjective:

(47) a. Het goede van zwemmen in de winter is dat je in

the good[+e] of swim.INF in the winter is that you in conditie blijft terwijl je de fietsbeweging even los laat.

shape stay while you the bike-motion for-a-moment go let 'The good thing about swimming in winter is that you stay in shape while you let go of the biking motion.' b. Het vervelende aan pesten is dat het vaak en met the annoying[+e] at bully.INF is that it often and with opzet gebeurt.

intention happens

'The annoying thing about bullying is that is happens frequently and intentionally.'

c. Het gevaarlijke bij rechts inhalen is dat je bij

the dangerous[+e] with right overtake.INF is that you with

het inhalen even de andere baan gebruikt en dan

the overtake.INF for-a-moment the other lane use and then

weer op de oorspronkelijke baan komt.

again on the original lane come

'The dangerous thing with passing on the right is that while overtaking

you use the other lane for a moment, and then return to the original lane.'

d. Het goede aan deze campagne is dat de ondernemer zelf the good[+e] at this campaign is that the entrepreneur self aan de slag kan.

at the go can

'The good thing about this campaign is that the entrepreneur can get started himself.'

e. Het vervelende van talent is dat je het vaak zelf niet ziet. the annoying[+e] of talent is that you it often self not see 'The annoying thing about talent is that you often do not see it yourself.'

f. Het leuke met het meerijden in een vrachtwagen is dat the fun[+e] with the drive-along.INF in a truck is that je overal in Nederland terecht komt.

you everywhere in Netherlands right come.

'The fun thing about driving along in a truck is that you get everywhere

in the Netherlands.'

Though goede is an inflected adjective that can appear in abstract object constructions like (26) above, the abstract object analysis is not available for (47a) or (47d), as evidenced by the fact that neither al/vele nor a demonstrative determiner could be used in these sentences, as exemplified in (48):

(48) a. *Dit vervelende van pesten is dat het vaak en met this annoying[+e] of bully.INF is that it often and with opzet gebeurt. intention happens. b. *Al het leuke / *Het vele leuke met het meerijden

all the nice[+e] / the much[+e] nice with the drive-along.INF in een vrachtwagen is dat je overal in Nederland

in a truck is that you everywhere in Netherlands

terecht komt. right come

The source of the problem appears to be that an eventuality cannot function as a location for the abstract object—in other words, it cannot provide the ground in a figure-ground relation. Note, in support of this, that the prepositions that are

excluded from the relational inflected adjective construction are locative prepositions (e.g. in 'in,' op 'on,' rond 'around'), while those that are accepted are often used to express more abstract relations (van 'of,' aan 'at').13

At this point we do not have a deep explanation for what blocks eventuality-denoting expressions from the abstract object construction, but one possibility is that eventualities fail to have an existence independent of their participants and spatiotemporal location. It might be the case that for purposes of locating something, those participants or locations would be more salient options than the transitory relations between them or the actions that they perform. In contrast, something like humankind, though also in a sense abstract, can provide a ground against which things can be located, perhaps because what really serves as the location is the instances of humankind. The instance-kind relation is different from the participant-eventuality relation insofar as properties attributed to kinds are routinely (if not always) shared by their instances, whereas that is arguably not so for participants and the eventualities they participate in.

Whatever the deeper explanation, the crucial fact is that the impossibility of an eventuality like swimming to serve as a location does not prevent it from being the target of property ascription: zwemmen in (47a) is described as an activity that is good to do in order to stay in shape, and the DP refers to that property of swimming. Similarly, deze campagne 'this campaign' in (47d) is what is described as having a good aspect, and the DP refers to that aspect. We can make similar remarks about het vervelende in (47b) in comparison to its use in the abstract object construction illustrated in (36a) above. In (36a), al het vervelende dat hoort bij zijn ziekte refers to 'all the annoying things that come along with his disease.' Het vervelende aan pesten in (47b) does not have such a maximal sum interpretation, but rather refers to a property of bullying ('bullying is annoying'). If a figure-ground relation is ruled out, we conclude that the PP in the relational construction cannot be analyzed as a conjunctive locative modifier. We will return shortly to the question of exactly what DPs like those in (47) denote and how their semantics is computed.

Additional support in favor of the special status of the PP in the relational inflected adjective construction comes from the fact that its complement can include an overt negation:

13 What is less clear to us is why the choice of prepositions is not more radically limited in the case of the relational construction, as might be expected if the PP is semantically selected by the adjective. The second author observes that there is a certain amount of speaker variation in the choice of preposition, noting, for example, a preference for van or aan in (47f), rather than met. We also note that something similar can be seen in the English translations of the construction: For the first author, there is not a big difference between using about or with in (47a, c). We must leave further exploration of the conditions governing the choice of preposition in the relational construction for future research.

(49) a. Het mooie van het niet hebben van zo'n tool is dat

the beautiful[+e] of the not have.INF of such tool is that

je gewoon het zekere voor het onzekere neemt en

you just the certain for the uncertain take and

alles vervangt door goede materialen.

everything replace by good materials

'The nice thing about not having such a tool is that you just err on the

side of caution, and replace everything by good materials.'

b. Geef toe dat uw irritatie komt door het ongewone acknowledge that your irritation comes through the unusual[+e] van het niet roken.

of the not smoke.INF

'Acknowledge that your irritation is caused by the unusual experience of not smoking.'

c. Dat is natuurlijk ook het mooie van niet geloven: that is naturally also the beautiful[+e] of not believe.INF je hoeft nergens van gered te worden en er gebeurt you need nowhere of saved to become and there happens niks naars met je na je dood. nothing bad with you after your death

'That is of course the nice thing about not believing: you don't need to be saved from anything, and nothing bad happens to you after you die.'

That these are not abstract object constructions can be seen in the fact that, as with (47), it is impossible to modify any of the examples in (49) with the predeterminer al, or to replace the definite article with a demonstrative article:

(50) *Geef toe dat uw irritatie komt door al het / dit

acknowledge that your irritation comes through all the / this ongewone van het niet roken. unusual[+e] of the not smoke.INF

It is hard to imagine what sort of location the denotation of the negated infinitive could provide for locating the property. We therefore take the presence of negation to indicate that the complement to the PP is not providing the ground in a figure-ground relation.

A final piece of evidence that the PP is an indispensable ingredient of the semantics of the relational inflected adjective construction is the fact that it is obligatory, unless its content can be unambiguously retrieved from the immediate context, as in (51), where the adjective rare 'funny' describes (the landscape of) the town mentioned immediately before:

(51) En dan komen we aan op de top, en wat doe je dan. and then arrive we PART at the top and what do you then je geniet van het uitzicht. Hier vandaan kijken we op de you enjoy of the view Here from look we at the stad Wernigerode, en het rare is dat het daar weer city Wernigerode and the funny[+e] is that it there again helemaal plat als een dubbeltje is.

entirely flat as a dime is

'And then we arrive at the top, and then what do you do.. you enjoy the view. From here, we look down upon the city of Wernigerode, and the funny thing is that there, it is flat as a dime.'

Intuitively, the information contributed by the PP is essential because without it we wouldn't know what object the adjective is describing a feature of. We conclude that the PP with the relational inflected adjective is an obligatory argument, and that the preposition does not denote a general locative figure-ground relation, hence the greater limit on the variety of prepositions attested in the construction. In contrast, in the abstract object construction the PP is an optional locative adjunct, in which we can use any preposition that denotes a general locative figure-ground relation.

Summarizing, there is clear evidence that the relational construction constitutes a third non-elliptical use of the inflected adjective alongside the human and the abstract object constructions. The fact that it occurs exclusively with the definite neuter article het, rejecting demonstratives as well as the predeterminer al, and that it rejects restrictive relative clauses casts doubt on the possibility of positing a free relative semantics for the construction. We therefore develop a different syntax and semantics for this construction in the following section.

5.2 The syntax and semantics of the relational inflected adjective construction

The fact that the PP is obligatory and that the DP intuitively refers to the property itself and not to an individual or stuff that manifests the property leads us to adopt a rather different syntactic analysis for this construction than for the abstract object construction. Most importantly, we see no place for a null nominal or pronominal pro, as the target of the adjective's predication is provided by the PP complement. The structure we propose is thus quite simple: the inflected adjective directly complements the head of a DP, taking the PP as its own complement, as in (52b) for (52a):

(52) a. het vreemde van de situatie

the strange[+e] of the situation 'the strange thing about the situation' b. [DP het [AP [A vreemde] [PP van de situatie]]]

As with the abstract object construction, the fact that we do not posit conversion of the adjective into a noun immediately explains the tendency to use adverbial, rather

than adjectival, modifiers. Initially, it might seem surprising that we assign the adjective a PP complement in this structure when it does not take one in general. In fact, the construction is deeply similar to infinitival constructions in Dutch, such as (53).14

(53) het / *de zingen van Jan the.SG.N / the sing.INF of John 'John's singing'

What, in the sentential use of the predicate, would be the external argument of the predication appears as a van-marked PP complement in the nominalized infinitive in (53). If the PP complement can indicate argumenthood in (53), it can do the same in (52). In addition, the infinitival construction in (53) refers necessarily to a specific eventuality of singing by John, in an exactly parallel fashion as the relational inflected adjective construction in (52) refers to a specific instance of strangeness manifest in the relevant situation.

We have seen that the semantics of the relational inflected adjective construction is crucially different from that of the abstract object construction. Most importantly, we need to make explicit a variable in the semantic representation that corresponds to the property denoted by the entire DP (e.g. the strangeness manifest in the situation). The simplest way to do this is to posit that the adjective has an additional argument analogous to the event argument associated with verbs, represented in (54a) as x.15 This amounts to the claim that the adjective constitutes a description of a property, much in the same way that a verb denotes a description of an eventuality. We take the inflection -e, which we suggested in footnote 7 may mediate in predication DP-internally, to increase the valence of the adjective by one and to introduce a relation (Pasp) indicating 'the P aspect' of an argument (54b). The adjective provides a value for the P in Pasp, as in (54c), which can be roughly paraphrased as ''y's strange aspect'' or ''y's strangeness'':

14 However, in contrast to what we find with the relational inflected adjective construction, we can use the predeterminer al with infinitives, and the demonstrative articles are fine as well:

(i) dit / dat / al dat zingen van Jan this / that / all that sing.inf van Jan

Thus, the comparison only goes so far.

15 One anonymous reviewer expressed concern that this semantics for -e might create problems for the analysis of adjectival inflection in other sorts of nominals. We see no obstacle to including the extra argument introduced in (54b) in the representations for inflected adjectives more generally (thus allowing for a fully uniform treatment of -e and the inflected adjective), including in the human and abstract object constructions, though we will not revise the analyses of the latter here to conform with (54b). If we were to do this, in other uses of the inflected adjective, including the human and abstract object constructions, we would need to existentially close the additional variable corresponding to the property itself (x in 54b) so as to prevent it from interacting with the rest of the semantics of these constructions and, in the case of the human and abstract object constructions, allow the y variable in (54b) to directly or indirectly identify the referent of the constructions as a whole. A similar strategy is adopted for English in Glass (2014), though we make different use of it here than Glass does. See Sect. 6 for further discussion of Glass's proposal.

(54) a. [vreemd]]: kzks[Strange(z)(s)]

b. [[-e]]: kPhS:(e,t»kykxks[Pasp(y)(x)(s)]

c. [vreemd+e]]: kykxks[Strangeasp(y)(x)(s)]

The resulting relation takes an argument syntactically introduced by a preposition that we consider semantically vacuous, similar to the van in (53) or the of in relational noun configurations like mother of John. (55) illustrates the composition. We represent the denotation of the complement to the preposition as in (55a). In the absence of a PP, the argument is saturated contextually, as in (51). The result in (55b) is the property of being the Adj aspect of the denotation of the DP complement to the P. If we apply het as defined in (25b) to the AP (assuming the nominal is extensionally interpreted here), we get (55c).

(55) het vreemde van de situatie the strange[+e] of the situation [DP het [ap [a vreemde] [PP van de situatie]]]

a. [de situatieJig: ty[Situation(y)]

b. [[ [AP vreemde+e van de situatie] ]]g: kxks[Strangeasp

(iy[Situation(y)])(x)(s)]

c. [[ [DP[Dhet s1 [AP vreemde+e van de situatie]] ]]g: kPhe,hs,t»

tx[P(x)(g(s1))] (kxks[Strangeasp(iy[Situation(y)])(x)(s)) = ix[Strangeasp(ty[Situation(y)])(x)(g(s1))]

(55c) gives us the intuitive meaning of (55), in which the adjective indicates a particular aspect of the situation.

This analysis directly accounts for various properties of this construction. First, the obligatoriness/necessary recoverability of PP is due to the relational semantics of the adjective. This obligatoriness might eventually be given a deeper account based on the idea that we access or ground specific properties in their bearers. Indeed, it is an old philosophical idea that particular instances of properties (sometimes referred to as 'tropes') depend for their existence on the individuals that manifest or bear them (see e.g. Moltmann 2004 for discussion and references to earlier literature).16

Second, it is now clear why infinitival complements, including negated ones, are acceptable in the PP. Though their denotations might not be able to serve as locations for the stuff denoted by the abstract object construction, there is ample linguistic evidence that they serve as the arguments to property predications (see e.g. Chierchia 1984; Zucchi 1993).

16 Note, interestingly, that certain cases of reference to eventualities also seem to be mediated through their participants in a similar fashion, further underscoring the parallel between infinitival expressions like (53) and the relational inflected adjective construction. Grimm and McNally (2013) present corpus evidence that the expression of participant-denoting PPs with deverbal nouns is obligatory precisely when the event referred to by the noun is introduced for the first time in discourse and the relevant participant(s) cannot be recovered from the context. This generalization, in turn, has antecedents in Rappaport Hovav and Levin's (2001) Argument-per-Subevent Condition, which was proposed to account for a broad range of argument realization facts based on the event structures associated with the verbs in question.

Third, as was the case with the abstract object construction, the analysis accounts for the fact that, as a rule, adverbial rather than adjectival modification is found (cf. 39), since the inflected adjective is, indeed, an adjective and not converted to a noun. Since the syntax of the relational construction is slightly different than that of the abstract object construction, we illustrate the semantic composition of adverbial modification.

Consider (39a), repeated here in (56). The adverb and adjective together form a complex description that is ascribed to the referent of the PP complement to the adjective; in other words, the adverb composes semantically with the adjective before the latter combines with the PP. This is unsurprising given that te 'too' is a degree adverb, and, on an analysis of adjectives as measure functions (e.g. Kennedy 1997), degree morphology is essential to licensing the adjective as a predicate of or relation between individuals (56a, b). Following e.g. Pinon (2005) and Kennedy and Levin (2008), we assume that degree morphemes can apply both to measure functions and (as is needed here) measure relations, in order to convert them into predicates of the corresponding valence. In other words, the expression in (56b), denotes a function not to propositions but rather to degrees, while that expression combined with te as in (56d) does denote a function to propositions. Adapting a proposal in Meier (2003), we take te to introduce a standard whose value exceeds the maximum degree compatible with the argument of the adjective being able to manifesting some specific property (56c). The specific property is contextually identified as Pi in (56c), but may in other cases be identified by an infinitival clause. The rest of the derivation in (56) is like that in (55).

(56) het al te zure van citrusvruchten is in deze confiture the all too sour[+e] of citrus-fruits is in this jam verdwenen. disappeared

'the all too sour (aspect) of citrus fruits is lacking in this jam.'

a. [DP het [DegP [D te][A zure]] [PP van citrusvruchten]]

b. Hzurehe,{e,{s,d)»]]g: kykxks[Sourasp(y)(x)(s)]

c. [[te]]: kG(e,{e,{s,4»kykxks[G(y)(x)(s) > maxd[G(y)(x)(s)

= d ? e Pi(x)]]

d. [te zure]]g: kykxks[Sourasp(y)(x)(s) > maxd[Sourasp(y)(x)(s)

= d ? OPi(x)]]

e. \[citrusvruchten]\g: c

f. [te zure van citrusvruchten]]g:

kxks[Sourasp(c)(x)(s) > maxd[Sourasp(y)(x)(s) = d ? ePi(x)]]

g. [het sj te zure van citrusvruchtenl\g:

tx[Sourasp(c)(x)(g(s:)) > maxd[Sourasp(y)(x)(s) = d ? OPi(x)]]

The final representation in (56g) states that the sour aspect of the citrus fruits has a value that exceeds the maximum degree of sourness acceptable in the context. Thus,

this sort of adverbial modification can be handled straightforwardly.17 For the remainder of the paper, however, we will abstract away from the complexities of degree semantics introduced here and will maintain the simpler analysis of adjectives that we have used up to this point.

Let us now turn to the remaining modification facts. Recall that the relational inflected adjective construction resembles the abstract object construction in licensing enige 'only' (41a), the ordinals (41b), intensional adjectives such as vermeende 'alleged' (41c) and non-restrictively used evaluative adjectives (41d, e) and relative clauses (45). It differs in not licensing restrictive relative clauses. The analysis of the non-restrictive modification (whether via adjectives or relative clauses) works exactly as with the abstract object construction: these modifiers intervene on the conventional implicature level and ascribe properties directly to the referent of the full DP (e.g. the strange aspect of the situation). We will therefore not illustrate again the derivation of such examples. The remaining cases are worth discussing because the denotation we assign to the adjective phrase in the relational construction is of a different semantic type than the one we assigned to the small clause in the abstract object construction: it denotes a property, rather than a proposition. We might therefore have expected the modification facts to be different, and that restrictive adjectival or relative clause modification would be more systematically available when, in fact, the options are even more limited than in the case of the abstract object construction.

The key to understanding this combination of similarities and differences, as noted in Sect. 5.1 above, lies in the unique extension attributed to the Adj ? PP in the relational construction, evidence for which can be found in the inability of the construction to take any determiner other than het, including demonstratives, the maximizing predeterminer al and adjectival vele, weinig ('many/much, few/little'). This uniqueness blocks restrictive modification but will not block those adjectives with other semantic contributions, as is the case precisely with enige, the ordinals, and intensional adjectives. All of these are independently attested as properties of property-type expressions in Dutch (i.e. common nouns), so a property-type alternant for them must be independently assumed, even though their semantics is fundamentally propositional in nature, as argued e.g. by Larson (2002). Therefore, their composition with the Adj ? PP in the relational inflected adjective construction should work exactly as it does with those, modulo the difference in semantic type. For the sake of comparison with the abstract object construction, we now briefly go through sample derivations with these adjectives in the relational construction.

Consider the use of vermeende in (41c), repeated below in (57). In (35a) we assigned vermeende the denotation in (57a), but here we use the alternative, equivalent denotation in (57b). This will combine with the Adj ? PP (57c, where p, for 'protein percentages', is the translation of the PP) as in (57d).

17 The adverbial modification we have found in this construction appears to be limited to degree modification, in which category we include adverbs such as intens 'intensely' in (39b) and weinig 'little' in (42c).

(57) het vermeende goede van hele hoge eiwitpercentages the alleged[+e] good of very high protein-percentages 'the alleged goodness of very high protein percentages'

a. [vermeende]]: kphs,t»ks[Alleged(p)(s)]

b. [[vermeende]]: kP^,^»»kxks[Alleged(P(x))(s)]

c. [goede van hele hoge eiwitpercentages]\: kxks[Goodasp(p)(x)(s)]

d. [vermeende goede van hele hoge eiwitpercentages]\:

kP he, (s,t») kxks [ Alleged(P(x))(s)](kxks [Goodasp(p)(x)(s)]) = kxks[Alleged(Goodasp(p)(x))(s)]

The examples with enige and ordinals are more interesting. In our corpus searches, we have found these adjectives exclusively in DPs that serve as the fronted predicate in specificational copular sentences. We repeat two examples from above in (58):

(58) a. Het enige leuke aan niet naar school kunnen gaan

the only fun[+e] at not to school can.INF go.INF is niet naar school gaan. is not to school go.INF

'The only fun thing about not being able to go to school is not going to school.'

b. Het eerste vervelende van de maatregel is dat het

the first annoying[+e] of the measure is that it voor de scholen lastig is...

for the schools tricky is

'The first annoying thing about the measure is that it is tricky for the schools...'

We pointed out in Sect. 4 that the function of both enige and the ordinals is to quantify over alternatives. In (58), these alternatives are intuitively the potential bearers of the property of being fun or annoying. There is, thus, a sortal mismatch between the preverbal, predicate DP in the specificational copular sentence (the relational inflected adjective construction) and its postverbal subject argument. To take (58a) as an example, on our analysis the preverbal DP (ignoring enige) should denote a property, namely the fun (or funness) in not going to school; however, its argument DP—with which it is specificationally identified—identifies that aspect or part of not going to school that actually manifests that property or, to put it in other terms, supports the ascription of the property. Though this might look like a reason to think that perhaps the examples in (58) involve the abstract object construction, rather than the relational one, we note that the Adj ? PP forms in (58) without enige or the ordinal cannot be preceded by al het or dit, which would be clear diagnostics for the abstract object construction:

(59) *Al het/ dit leuke aan niet naar school kunnen gaan

all the/ this fun[+e] at not to school can.INF go.INF

We therefore conclude that analyzing the preverbal, predicate DPs in (58) as involving the abstract object construction is not correct. Rather, it seems that the predicative context coerces the Adj ? PP from denoting a property of a particular object to denoting the set of parts or aspects of that object that manifest the property. For example, in (58a) the coercion would be from denoting the fun in not going to school to the set of those parts or aspects of not being able to go to school that manifest this fun. The unacceptability of (59) suggests that these manifestations are not conceived of as a maximal sum of a mass, as is the case for the referent of the abstract object construction. It also suggests that the referent of the relational inflected adjective construction in (58a) is not, strictly speaking, conceived of as a part of the referent of the complement to the preposition in the construction. This may seem initially surprising, especially since one can find in English similar examples with the overt noun part, as in (60):

(60) a. About the only good part of not making it to Philadelphia was that I got to

go back to sleep for a bit.

b. So, the only bad part of not taking the test is that she may not be able to apply for some college scholarships.

c. The only negative part of not having a kid with me was that I had no one to blame for all the little color samples I kept taking.

However, observe that it is difficult to imagine how the referent of the expression on the right-hand side of the copula in these examples could be considered a part of the referent of the complement to of on the corresponding left-hand side, except in the most abstract of senses.18 Note that even if we construe the negated VPs as denoting propositions or facts—that is, as pieces of information, whose entailments could be considered parts—it is not clear that we can take the propositions or facts on the right-hand side to constitute informational parts, as in no case are they strictly entailed by the negated VP. Thus, while it is beyond the scope of this paper to explore more deeply the sort of predication relation that holds in (58) and (60), we consider examples such as (60) to confirm rather than call into question the need for an analysis that is distinct from the one we used for enige and ordinals in the abstract object construction.

Once we accept that Adj ? PP in the relational inflected adjective construction can be coerced from describing a property to describing something that manifests the property, the example in (58a) can be composed straightforwardly as in (61). We follow the semantic analysis of specificational sentences in Geist (2007) (who draws

18 We also observe that such examples are significantly rarer than their counterparts of the form the only good/bad/negative thing about.... To give an idea of the difference in frequency, a Google search (26/5/ 2014) for ''the only good part of not'' turned up 40 results before Google warned that the rest were highly similar; in contrast a search for ''the only good thing about not'' turned up 268 results before that warning emerged. Good was by far the adjective that yielded the largest number of examples in this construction with part; examples with other adjectives are anecdotal.

on earlier work by Williams 1983 and Partee 1986), on which the verb be has the option of combining first with an entity-denoting expression, and then a property-denoting one (see 61f). We assign enige a version of the semantics of DP-internal only from McNally 2008 (see 61d).19 We take the coercion step mentioned above to apply to the Adj ? PP as in (61c), indicated by a coercion operator Man (for 'manifestations') and the corresponding subscript on x in the output. The relational inflected adjective construction is converted into the specificational predicate via the ident type shift (Partee 1987, 61i).

(61) Het enige leuke aan niet naar school kunnen gaan is

the only fun[+e] at not to school can.INF go.INF is niet naar school gaan. not to school go.INF

'The only fun thing about not being able to go to school is not going to school.'

a. [niet naar school kunnen gaan]]g: nscg ('not school can go')

b. [leuke aan niet naar school kunnen gaan]]g: kxks[Niceasp(nscg)(x)(s)]

c. Man([7euke aan niet naar school kunnen gaan]]g): kxManks [Niceasp

(nscg)(xMan)(s)]

d. [[enige]]g: kPkxks[P(x)(s) A Vy[P(y)(s) ? y C x]]

e. Hhetj sj enige leuke aan niet naar school kunnen gaan]]g:

ixMan[Niceasp(nscg)(xMan)(g(s:)) A

Vy[Niceasp(nscg)(y)(g(s:)) ? y C x]]

f. [[is]]g: kvkP[P](v)

g. [niet naar school gaan]]g: nsg ('not school go')

h. [is niet naar school gaan]]g: kvkP[P](v)(nsg) = kP[P](nsg)

i. ident([[61e]]g): kz[ixMan[Niceasp(nscg)(xMan)(g(s1))

AVy[Niceasp(nscg)(y)(g(s:)) ? y C xMan]] = z]

j. kP.P(nsg)([[61i]]g): txMan[Niceasp(nscg)(xMan)(g(s:)) = nsg

AVy[Niceasp(nscg)(y)(g(s:)) ? y C xMan]]

In prose: (61e) provides us with the unique manifestation of what is nice about not being able to go to school. After application of Partee's ident operator in (61i), we obtain the singleton set corresponding with this manifestation. Application of the predicate is niet naar school gaan from (61h) to this singleton set results in the identification of the unique manifestation of what is nice about not being able to go to school with not going to school in (61j). A fully analogous derivation can be given for (58b), assuming we assign the ordinal a property- (rather than proposition-) selecting variant of the interpretation in (34).

This concludes our description and analysis of the Dutch facts. We now turn to a very brief discussion of how these facts and their analysis differ from what has been

19 See Coppock and Beaver (2012) for a very similar analysis of DP-internal only, which differs from McNally's analysis mainly in putting the first conjunct in the presuppositional part of the interpretation. Our claims concerning the contribution of DP-internal only are independent of the choice between the two proposals.

observed and proposed in the literature for similar constructions in other languages, and of what we see as the larger implications of our study.

6 The bigger picture

We have argued that the abstract object construction involves reference to a mass of stuff via a property that this stuff manifests. It does not refer to the property itself. The same, mutatis mutandis, is (more obviously) true of the human construction. In contrast, the relational inflected adjective construction does apparently refer to a property. However, even though they are crystal clear in certain contexts (cf. 42-47), the differences between the abstract object and relational constructions can be difficult to detect in other contexts (such as the instances of coercion in 61), and indeed prior to now the difference between the two has gone unnoticed in the literature.

Although our discussion has focused exclusively on Dutch, adjective-based constructions that refer specifically to humans, to maximal sums of stuff bearing the adjectival properties, and/or to the properties themselves exist in a number of other languages. Without going far afield we find such constructions in French (Lauwers 2008; Beauseroy 2009; Borer and Roy 2010), Spanish (Villalba 2009; Villalba and Bartra-Kaufmann 2010), Greek (Giannakidou and Stavrou 1999; Alexiadou 2011), Serbo-Croatian (Arsenijevic 2011), Slovak (Richtarcikova 2014), German, Italian (Marzo and Umbreit 2013), and Romanian (Alexiadou and Iordachioaia 2013), in addition to English, as mentioned in Sect. 1. As there is considerable variation across languages in the syntax-semantics interface of these constructions and their use, it would be impossible to discuss them all here. We therefore limit ourselves to briefly comparing our proposal to the two semantic analyses that we know of, namely that for Spanish in Villalba (2009, 2013) and for English in Glass (2014), and we will focus mainly on the constructions that do not have human reference, as they seem to present more cross-linguistic variation and greater complexity. Since the constructions in these two languages and the analyses proposed differ in interesting ways from what we have argued for Dutch, this comparison will serve as a useful illustration of the cross-linguistic terrain, though it does not by any means exhaust the range of variation.

We begin with Spanish. Examples of human reference, reference to what is arguably a mass of (abstract) stuff, and to a property are illustrated in (62a-c), respectively.

(62) a. dos ciegos, los ciegos, la pija

two blind.PL.M, the.PL.M blind.PL.M the.SG.F posh.SG.F 'two blind people, the blind people, the posh woman' b. ...todo lo aprendido o estudiado, ...iba teniendo gravedad... .all lo learned or studied went having seriousness

'. everything that was learned or studied,. gradually took on seriousness.'

c. „.el cielo despejado, limpio, .the sky clear clean

hiciera perder la hermosura made.suBj lose the beauty bóveda celeste. heavens

'.the sky clear, clean.and without a single cloud that could mar the beauty of the heavenly blue.'

The human construction allows for generic or specific count reference, as in Dutch, and invariably involves the presence of a determiner (in some cases inflected for number and gender) of the sort that typically appears with nouns. Everything indicates that this construction could be analyzed as we have proposed for Dutch, and we will have nothing more to say about it here.

Non-human reference is more interesting. It invariably involves neuter lo instead of a typical determiner.20 Villalba (2009) argues that examples like (63) denote tropes, i.e. particular instances of a property (citing Moltmann 2004); in a later work (Villalba 2013) he adopts the analysis of het-constructions in McNally and de Swart (2011), which, setting aside the treatment of the article, is identical to the one we have proposed here for the relational abstract object construction.

(63) lo honesto de Juan lo honest of John 'John's honesty'

Curiously, however, though examples of reference to properties involving lo-constructions, like (62c), are attested in Spanish, and some further examples like

(63) are provided in Bosque and Moreno (1990), the Peninsular Spanish speakers we have consulted generally do not find (63) itself or the examples in Bosque and Moreno acceptable, preferring instead a deadjectival nominalization such as (64).

(64) la honestidad de Juan the.SG.F honesty of John 'John's honesty'

Informal Google searches for color and taste adjectives with clear property reference yield very few results. Note that (62c) involves the adjective azul 'blue,' which, unlike other common color terms in Spanish lacks a nominalized form (cp. e.g. rojo/rojez 'red/redness,' blanco/blancura 'white/whiteness,' negro/negrura 'black/blackness'). Interestingly, corpus examples with lo rojo or lo blanco are very hard to find with a relational interpretation, suggesting that their use is dispreferred. Furthermore, we observe that lo heads constructions that are the equivalent of free

20 The status of lo is controversial: It has been claimed to be a determiner, a pronoun, and it has even been argued that perhaps one cannot tell (see Bosque and Moreno 1990; Leonetti 1999 for an overview of the literature). We set aside this question here, as a fruitful comparison between these lo-constructions and the constructions in Dutch can be made without resolving the status of lo, and gloss lo as lo.

y sin ninguna nube que

and without no cloud that

de lo azul de la

of lo blue of the

relatives in English (65 a) and related degree relative clauses (65b), as well as other degree-referring expressions (65c, from Bosque and Moreno 1990, p. 34):

(65) a. lo que dije ayer

LO that I-said yesterday 'What I said yesterday'

b. Lo bueno que era! LO good that was 'How good it was!'

c. Juan come lo necesario... no mas cantidad de la Juan eats lo necessary no more quantity of the aconsejable.

recommendable

'Juan eats the necessary amount. not more than is recommendable.'

Two of the nine attested sentences that Villalba (2009) provides containing lo-constructions could clearly be translated using degree expressions as in (66c) or a non-exclamative version of (66b), and three others could clearly be translated as 'that which is Adj,' like (65a).

The discussion of examples (62) to (65) suggests that the semantics we proposed for the abstract object construction in Dutch would be the obvious choice for the lo ? Adj constructions (62b), and actually its preferred interpretation in Peninsular Spanish more generally. We hypothesize that the relational interpretation of lo ? Adj, although available in principle, is only recruited for property reference in actual language use when a deadjectival noun is not available (as in 62c).

Note that the Dutch het constructions do not have the free relative and degree relative uses illustrated in (65). In contrast, they are productively used for reference to individuals (through the human construction), stuff (through the abstract object construction) and properties (through the relational construction). Whereas derived nominals in Spanish are commonly used to refer to properties in general (64), derived nominals in Dutch are specialized for expressing a degree or quantity of a property (66), as observed by McNally and de Swart (2011).

(66) hooguit een beetje roodheid op mijn wangen at-most a bit redness on my cheeks 'at most a bit of redness on my cheeks'

No such restriction to degree/quantity interpretation is observed for deadjectival nouns like (64) in Peninsular Spanish. It thus appears that, while the two languages share a similar basic set of forms and express a similar set of meanings, the distribution of the meanings over the different forms is distinct in the two languages.

Glass (2014) carried out a corpus study of English the ? Adj expressions, and argues that they have two sorts of denotations: a plural count denotation referring to (mainly but not necessarily always animate) individuals bearing the property, and a mass denotation referring to the property itself. Representative examples of each type from her paper appear in (67) and (68), respectively:

a. New Swarm Theory: The Weak Can Lead the Strong [topic: insects]

b. In Tacloban, the dead are being taken to a mass grave in a public cemetery

c. How about because few rich are philanthropic?

d. The stylish young are reclaiming the necktie as their own.

a. The cute is perhaps the dominant aesthetic category of our late-capitalist times.

b. ''progress'' always seems to go in one direction-toward the dead and the dull.

c. A little pretty is just what the doctor ordered!

d. You're on the upside of the healing and most of the mean nasty is behind you

Glass formalizes the notion of 'property' as a neo-Davidsonian state and proposes that adjectives have the representation in (69a). She then proposes that a nominalizing type-shifter, nom-indiv (69b), can convert the adjective (or adjective phrase) into an NP that denotes a set of entities of which the state holds, with existential closure over the state variable (69c). Alternatively a related type shifter, nom-mass (69d), can convert it into a mass of the property, with existential closure over the entity variable (69e; irrelevant details of the formalization are changed here):

(69) a. [happy]]: kxks[s < happiness A holder(x, s)]

b. [nom-indiv]]: kP^,^kxAs[P(x)(s)]

c. [nom-indiv(happy)]]: kxAs[s < happiness A holder(x, s)]

d. [nom-mass]]: kP^,^»ksAx[P(x)(s)]

e. [nom-mass(happy)]]: ksAx[s < happiness A holder(x, s)]

The resulting nominals can combine with any sort of determiner that is compatible in count/mass features.

The denotation that Glass provides for the individual interpretation is, except for the human restriction, effectively identical to that we provide for the human construction in Dutch. She substitutes type-shifting for our use of a null noun. However, her semantics for the mass construction is different from our analysis both of the abstract object construction and of the relational inflected adjective construction, and is in fact much closer to Kester's (1996) proposal. This is appropriate to the extent that the determiner facts in English are different from those in Dutch, as we see in the comparison of (68c) with (23). It may be the case that the specifically neuter form of the definite article, combined with adjectival inflection, sets the Dutch construction apart from ordinary nominals in other languages.21 Clearly further comparison of the English and Dutch data is called for.

While we cannot do justice to Villalba's and Glass's interesting discussions here, this brief comparison with Spanish and English leads to two main observations, which we hope will inspire future lines of research. First, while very similar

21 A similar compatibility with mass quantifiers for reference to 'stuff' via properties is found in French (Lauwers 2008), German (Berit Gehrke, p.c.), and Italian (Denis Delfitto, p.c.).

morphosyntactic resources are used cross-linguistically for reference to and via properties, languages vary in the degrees to which these are productive and in the ways in which the individual constructions become specialized. For reference to properties, Spanish relies heavily on derived nominals, but for reference to stuff via their properties and to degree, lo-constructions appear to be the productive option. In English, -ness affixation is tremendously productive for reference to properties themselves and to degree, while for reference to stuff via their properties there is no single preferred option, with a combination of free relatives, relatively empty nouns such as part, and occasionally the ? Adj being used. Finally, in Dutch het constructions are highly productive for reference to properties and stuff, while derived nominals are used to express degree. This cross-linguistic variation opens up interesting possibilities for exploring how different options for reference are established, interact as a system, and evolve over time.

The second observation resulting from our brief comparison of Dutch, Spanish and English brings us back to one of the main questions we raised in the introduction: that of the division of labor between nouns and adjectives as the descriptive core of referring expressions. Our analysis of Dutch assigns multiple syntactic and semantic roles to adjectives in the constructions where they are the primary source of descriptive content for reference. However, on our account they never directly replace common nouns: adjectives serve as modifiers of a null noun in the human construction, as predicates of individuals in the free relative-like small clause structure associated with the abstract object construction, and as predicates over aspects of the referent identified by the PP complement in the relational inflected adjective construction. The diversity is perhaps surprising, but it strongly indicates that the grammar of Dutch exploits the specific features of adjectives to enrich the expression of reference, rather than obliterating the distinction between adjectives and common nouns. We thus hope that our analysis contributes to a more fine-grained semantic analysis of adjectives that goes beyond the <e,t>-type assignment they share with common nouns, and helps to articulate the relation between the abstract notion of a property and the less abstract ways in which property particulars are actually manifested.

7 Conclusion

In this paper we have argued that Dutch has three distinct DP constructions whose main visible descriptive element is an adjective: alongside the well-known human construction, there is not only the abstract object construction previously discussed in the literature, but also what we call the relational inflected adjective construction. We have provided the first formal semantic analysis that we know of that accounts for the distinct properties of these three constructions. We have also pointed to some of the larger implications of this study for our understanding of both typological and within-language strategies for reference to and via properties. Though we have had to leave various deeper questions unanswered along the way, we hope that this study serves to inspire further work on this rich and underexplored area of natural language semantics.

Acknowledgments We are grateful to Anna Szabolcsi, Ellen Petra Kester, Xavier Villalba, and audiences at the 2011 Amsterdam Colloquium, Universitat Pompeu Fabra, Universitat Stuttgart, Universite Paris 8, Utrecht University and ZAS Berlin for comments and suggestions. We also acknowledge the support of Spanish MINECO Grants FFI2010-15066 and FFI2013-41301, AGAUR Grant 2009SGR00763 and an ICREA Academia award to the first author, and the support of the Dutch NWO Grant 360-70-340 to the second author.

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

References

Alexiadou, A. (2011, March). Adjectival nominalizations: Qualities and properties. Paper presented at the Workshop on Nounhood and Adjectivehood. Barcelona.

Alexiadou, A., & Iordachioaia, G. (2013, February). Two syntactic strategies to derive (abstract) deadjectival nominalizations. Paper presented at the workshop on adjectives and their nominalizations, Stuttgart.

Arsenijevic, B. (2011). The semantic ontology of deadjectival nominalizations in Serbo-Croatian. Recherches linguistiques de Vincennes, 40, 53-72.

Beauseroy, D. (2009). Syntaxe et sémantique des noms abstraits statifs: Des propriétés verbales ou adjectivales aux propriétés nominales. Dissertation, Universite Nancy.

Borer, H., & Roy, I. (2010). The name of the adjective. In P. Cabredo Hofherr & O. Matushansky (Eds.), Adjectives: Formal analyses in syntax and semantics (pp. 85-114). Amsterdam: John Benjamins.

Bosque, I., & Moreno, J. C. (1990). Las construcciones con lo y la denotación del neutro. Lingüística, 2, 5-50.

Broekhuis, H., Keizer, M., & den Dikken, M. (2013). The syntax of Dutch. Amsterdam: Amsterdam University Press.

Carlson, G. (1977). Reference to kinds in English. Dissertation, University of Massachusetts, Amherst. (Published 1980 by Garland Press, New York).

Chierchia, G. (1984). Topics in the syntax and semantics of infinitives and gerunds. Dissertation, University of Massachusetts, Amherst. (Published 1989 by Garland Press, New York).

Chierchia, G. (1998). Reference to kinds across languages. Natural Language Semantics, 6, 339-405.

Coppock, E., & Beaver, D. (2012). Exclusivity, uniqueness, and definiteness. In C. Piñon (Ed.), Empirical issues in syntax and semantics 9 (pp. 59-76). Paris: CNRS Publications.

Dayal, V. (1995). Quantification in correlatives. In E. Bach, E. Jelinek, A. Kratzer, & B. Partee (Eds.), Quantification in natural language (pp. 179-206). Dordrecht: Kluwer.

de Hoop, H. (1992). Case configuration and noun phrase interpretation. Dissertation, University of Groningen.

Del Gobbo, F. (2003). Appositives at the interface. Dissertation, University of California, Irvine.

Doetjes, J. (1997). Quantifiers and selection: On the distribution of quantifying expressions in French, Dutch and English. Dissertation, Leiden University.

Farkas, D., & de Swart, H. (2010). The semantics and pragmatics of plurals. Semantics and Pragmatics, 3, 1-54. Retrieved January 15, 2015, from http://semprag.org/article/view/sp.3.6.

Geist, L. (2007). Predication and equation in copular sentences: Russian vs. English. In I. Comarovski & K. von Heusinger (Eds.), Existence: Semantics and syntax (pp. 79-105). Cambridge: Cambridge University Press.

Giannakidou, A., & Stavrou, M. (1999). Nominalization and ellipsis in the Greek DP. The Linguistic Review, 16, 295-332.

Glass, L. (2014). Deriving the two readings of English determiner + adjective. In U. Etxeberria, A. Falau§, A. Irurtzun, & B. Leferman (Eds.), Proceedings of Sinn und Bedeutung 18 (pp. 164-181). Retrieved January 15, 2015, from http://semanticsarchive.net/sub2013/.

Grimm, S., & McNally, L. (2013). No ordered arguments needed for nouns. In M. Aloni, M. Franke, & F. Roelofsen (Eds.), Proceedings of the 19th Amsterdam Colloquium (pp. 123-130). Amsterdam: ILLC Publications.

Haeseryn, W., Romijn, K., Geerts, G., de Rooij, J., & van den Toorn, M. C. (1997). Algemene Nederlandse spraakkunst. Second revised version. Groningen/Deurne: Martinus Nijhoff uitgevers/ Wolters Plantyn.

Hinterwimmer, S. (2013). Free relatives as kind-denoting terms. In A. Mari, C. Beyssade, & F. Del Prete (Eds.), Genericity (pp. 140-156). Oxford: Oxford University Press.

Jacobson, P. (1995). On the quantificational force of English free relatives. In E. Bach, E. Jelinek, A. Kratzer, & B. Partee (Eds.), Quantification in natural language (pp. 451-586). Dordrecht: Kluwer.

Kennedy, C. (1997). Projecting the adjective: The syntax and semantics of gradability and comparison. Dissertation, University of California at Santa Cruz. (Published 1999 by Garland Press, New York).

Kennedy, C., & Levin, B. (2008). Measure of change: The adjectival core of degree achievements. In L. McNally & C. Kennedy (Eds.), Adjectives and adverbs: Syntax, semantics, and discourse (pp. 156-182). Oxford: Oxford University Press.

Kester, E.-P. (1996). The nature of adjectival inflection. Dissertation, Utrecht University.

Larson, R. (2002). The grammar of intensionality. In G. Preyer & G. Peter (Eds.), Logical form and language (pp. 228-262). Oxford: Oxford University Press.

Lauwers, P. (2008). The nominalization of adjectives in French: From morphological conversion to categorial mismatch. Folia Linguistica, 42, 135-176.

Leonetti, M. (1999). El artículo. In I. Bosque & V. Demonte (Eds.), Gramática descriptiva de la lengua española (Vol. 1, pp. 787-890). Madrid: Espasa Calpe.

Link, G. (1983). The logical analysis of plurals and mass terms: A lattice-theoretic approach. In R. Bauerle, C. Schwarze, & A. von Stechow (Eds.), Meaning, use and interpretation of language (pp. 302-323). Berlin: de Gruyter.

Marzo, D., & Umbreit, B. (2013, February). Deadjectival nominalizations in Romance languages from a lexicalist perspective. Talk presented at the Workshop on Adjectives and their Nominalizations, Stuttgart University.

McCawley, J. (1998). The syntactic phenomena of English (Vol. 2). Chicago: University of Chicago Press.

McNally, L. (2008). DP-internal only, amount relatives, and relatives out of existentials. Linguistic Inquiry, 39, 161-169.

McNally, L., & de Swart, H. (2011). Inflection and derivation: How adjectives and nouns refer to abstract objects. Pre-proceedings of the 18th Amsterdam Colloquium (pp. 425-434).

Meier, C. (2003). The meaning of 'too', 'enough', and 'so.that'. Natural Language Semantics, 10, 69-107.

Moltmann, F. (2004). Properties and kinds of tropes: New linguistic facts and old philosophical insights. Mind, 113, 1-43.

Partee, B. (1975). Montague grammar and transformational grammar. Linguistic Inquiry, 6, 203-300.

Partee, B. (1986). Ambiguous pseudoclefts with unambiguous be. Proceedings of NELS, 16, 354-366.

Partee, B. H. (1987). Noun phrase interpretation and type-shifting principles. In J. Groenendijk, D. de Jong, & M. Stokhof (Eds.), Studies in Discourse Representation Theory and the theory of generalized quantifiers (pp. 115-143). Dordrecht: Foris.

Pinon, C. (2005). Adverbs of completion in an event semantics. In H. J. Verkuyl, H. de Swart, & A. van Hout (Eds.), Perspectives on aspect (pp. 149-166). Dordrecht: Springer.

Potts, C. (2005). The logic of conventional implicatures. Oxford: Oxford University Press.

Quirk, R., Greenbaum, S., Leech, G., & Svartvik, J. (1985). A comprehensive grammar of the English language. London: Longman.

Rappaport Hovav, M., & Levin, B. (2001). An event structure account of English resultatives. Language, 77, 766-797.

Richtarcikova, V. (2014, September). DPs with adjectival core in Slovak: A corpus based study. Paper presented at the 9th annual meeting of the Slavic Linguistic Society, Seattle.

Rooth, M. (1985). Association with focus. Dissertation, University of Massachusetts at Amherst.

Sassoon, G. (2013). Vagueness, gradability and typicality: The interpretation of adjectives and nouns. Leiden: Brill.

van Riemsdijk, H. A. (1985). A case study in syntactic markedness: The binding nature of prepositional phrases. Dissertation, University of Amsterdam.

von Stechow, A. (1984). Comparing semantic theories of comparison. Journal of Semantics, 3, 1-77.

Villalba, X. (2009). Definite adjective nominalizations in Spanish. In M. T. Espinal, M. Leonetti, & L. McNally (Eds.), Proceedings of the IV Nereus international workshop definiteness and DP structure in Romance languages. Arbeitspapier 124 (pp. 139-153). Konstanz: Fachbereich Sprachwissenschaft, Universitat Konstanz.

Villalba, X. (2013, June). Eventualities under deadjectival nominalizations. Paper presented at JeNom 5, Barcelona.

Villalba, X., & Bartra-Kaufmann, A. (2010). Predicate focus fronting in the Spanish determiner phrase. Lingua, 120, 819-849.

Williams, E. (1983). Semantic vs. syntactic categories. Linguistics and Philosophy, 6, 423-446.

Yee, C. (2010). Building DRT lexical entries for superlatives and ordinal numbers. Ms., Stuttgart University. Retrieved January 2015, from http://semanticsarchive.net/Archive/jg0YzQ4M/.

Zamparelli, R. (1995). Layers in the determiner phrase. Dissertation, University of Rochester (Published 2000 by Garland Press, New York).

Zucchi, A. (1993). The language of propositions and events. Dordrecht: Kluwer.