Alexiadou and Tsoulas- Privative and Binary Features

17
Privative and Binary Features, Roots and Functional Vocabulary Artemis Alexiadou Universit¨atStuttgart artemis@ifla.uni-stuttgart.de George Tsoulas University of York [email protected] Features in Phonology, Morphology, Syntax and Semantics: What are they? University of Tromsø/CASTL 31 October and 1 November 2013 One chord is fine Two you’re pushing it Three you’re into Jazz. Lou Reed, (1942-2013) In memoriam 1 Introduction Features are fundamental building blocks of information that enter in the construction of elements such as lexical items (whatever the latter may be) in many theories of grammar. Their properties and inventory have been and are intensively discussed in the literature in various terms and with varying conclusions. Adger and Svenonius (2011) distinguish two basic ways in which features can be understood in the theory of grammar: 1 (1) A descriptive device (2) An object of the theory This distinction of course does not mean that there can be no theory of descriptive devices but while we remain at the theoretical (as opposed to the meta -theoretical) level our ap- proach is situated within the class of theories in (2), in other words we take features to represent substantive properties of linguistic elements, we do not feel it is necessary to a priori confine ourselves to a theory where features are necessarily only atomic elements (though in most cases they are). Another important aspect of the analysis of different pos- sible feature systems relates to the privative vs. valued nature of features 2 In this paper we are fundamentally concerned with the amount and type of semantic information that is made available, as features, to the syntactic derivation. We focus on nominals and more specifically, assuming a model whereby roots enter the derivation under- or un-specified and 1 There may be other ways to understand features but they are not relevant to us here. 2 We consider binary a subcase of valued feature systems. 1

description

Alexiadou and Tsoulas- Privative and Binary Features

Transcript of Alexiadou and Tsoulas- Privative and Binary Features

Page 1: Alexiadou and Tsoulas- Privative and Binary Features

Privative and Binary Features, Roots and Functional

Vocabulary

Artemis AlexiadouUniversitat Stuttgart

[email protected]

George TsoulasUniversity of York

[email protected]

Features in Phonology, Morphology, Syntax and Semantics: What are they?University of Tromsø/CASTL

31 October and 1 November 2013

One chord is fineTwo you’re pushing it

Three you’re into Jazz.Lou Reed, (1942-2013)

In memoriam

1 Introduction

Features are fundamental building blocks of information that enter in the construction ofelements such as lexical items (whatever the latter may be) in many theories of grammar.Their properties and inventory have been and are intensively discussed in the literature invarious terms and with varying conclusions. Adger and Svenonius (2011) distinguish twobasic ways in which features can be understood in the theory of grammar:1

(1) A descriptive device

(2) An object of the theory

This distinction of course does not mean that there can be no theory of descriptive devicesbut while we remain at the theoretical (as opposed to the meta-theoretical) level our ap-proach is situated within the class of theories in (2), in other words we take features torepresent substantive properties of linguistic elements, we do not feel it is necessary to apriori confine ourselves to a theory where features are necessarily only atomic elements(though in most cases they are). Another important aspect of the analysis of different pos-sible feature systems relates to the privative vs. valued nature of features2 In this paperwe are fundamentally concerned with the amount and type of semantic information that ismade available, as features, to the syntactic derivation. We focus on nominals and morespecifically, assuming a model whereby roots enter the derivation under- or un-specified and

1There may be other ways to understand features but they are not relevant to us here.2We consider binary a subcase of valued feature systems.

1

Page 2: Alexiadou and Tsoulas- Privative and Binary Features

much of the relevant information is provided exoskeletally (see, e.g. Acquaviva, 2012; Arad,2005; Borer, 2005a,b; Marantz, 1997; Embick, 2010) we consider the question of the exactsemantic feature(s) that roots carry and which is visible to the syntax and how the range ofsuch features is constrained. Within this class of theories, there are few, if any, overarchingconstraints on the nature and quantity of featural content a lexical root may have. We firstconsider one particularly restrictive recent proposal.

2 Kayne’s Conjecture

A recent proposal regarding the amount of semantic information that enters the syntax isdue to Kayne (2005, 2010) who proposes that every lexical item contains at most onesyntactically interpreted feature.3 The implication of this is that whatever meaning isassumed for any lexical item and is beyond the semantic contribution of its single syntac-tically interpreted feature must come from somewhere else. If this is on the right track, thequestion arises as to what the relevant feature will be for each lexical item and how doesone decide. We illustrate this approach with some of the items that Kayne analyses and amore detailed analysis of another item.

2.1 Several, Few, and Many and NUMBER

The idea here us that these items are in fact adjectival modifiers of an unpronounced nounnumber or amount4 in the case of mass elements. So for items like several, few, many wewill have roughly (3):

(3) NP

AdjP Books

SeveralFewMany

NUMBER

Now the first question that arises is what is the semantic feature that these words have.For Kayne, the feature itself is directly grounded in the intuitive semantics of the items. Sofor few, the feature is [small], for many it is presumably [large] and something similarfor several. We return to the actual representation of these features.

2.2 TIME

Tsoulas (2013) analyses before-clauses as invariably involving a silent nominal time whichmay further be preceded by an empty demonstrative that.

Central in the analysis is the ungrammaticality of negation in Before-clauses:

3The hypothesis does have antecedents and can be thought of as the natural continuation of the traditionthat started, in a sense, with Pollock (1989). We will leave this aside for now.

4Elements appearing in this typeface are unpronounced.

2

Page 3: Alexiadou and Tsoulas- Privative and Binary Features

(4) * John left the store before Mary didn’t try the dress

The only interpretation that may be available for this sort of example, if accompanied byheavy focal stress on DIDN’T is a reading where Mary was somehow supposed to have triedthe dress on at a particular time and though the relevant event did not take place the timereference is to that time. Why do before-clauses disallow negation?

In the present framework there is an obvious statement of the analysis of before in thefollowing terms5

In somewhat informal terms the idea is that Before is always a preposition which selectsa DP, therefore there is only one Before. In the case of Before that appears alone the realstructure is:

(5) Before [THIS/THAT TIME]

In the case of Before followed by a clausal constituent the clause is a relative modifyingTIME.

(6) Before [THAT [TIME [CP . . . ]]]

The proposal in Tsoulas (2013) is that negation in a Before-clause leads to presuppositionfailure. If this idea is roughly on the right track, one thing that follows is:

(7) [THAT TIME] is a possible origin of the presupposition

What this means is that in the cases at hand the interpretation will halt as, clearly, thereare many times/intervals at which a given event did not take place, and therefore that timeis not uniquely identifiable as we would have expected.

The next question that then arises is:

(8) What is the single interpretable syntactic feature in Before given that it can’t beanything to do with TIME.

2.3 The feature of Before

(9) Q: What is the single syntactically interpretable feature of Before?

A: Comparison

2.3.1 Evidence

Italian Del Prete (2008) presents an analysis of the Italian equivalent of Before as a com-parative. Without going into the fine detail of his analysis, we can adopt similar stance andsuggest that roughly and informally, the single interpretable feature of Before is somethinglike Earlier.

5This is a particular implementation of an idea that goes back to Katz and Postal (1964). A similarapproach to before is also taken by Grønn and von Stechow (pear). For a recent criticism of this approachand by extension the approach taken here, see Sharvit (2013).

3

Page 4: Alexiadou and Tsoulas- Privative and Binary Features

Ancient Greek

(10) πρίν is originally a comparative adverb meaning before, i.e. sooner orformerly ; and seems to be connected with πρό, πρότερον before. Theadverbial force survives in Attic only after the article, as ἐν τοῖς πρίνλόγοις in the foregoing statements T. 2.62. The adverbial and originaluse appears also in Homer wherever πρίν occurs with the indicative, theanticipatory (futural) subjunctive, or the optative with κέ.Thus, τὴν δ΄ἐγὼ οὐ λύσω πρίν μιν καὶ γῆρας ἕπεισιν but her I will not release; soonershall old age come upon her A 29, οὐδέ μιν ἀνστήσεις: πρίν καὶ κακὸνἄλλο πάθῃσθα nor shalt thou recall him to life; sooner (before this) thouwilt suffer yet another affliction.

Smyth (1920, p.550, §2438)

English Modification by Many/Much:

(11) a. *More fasterb. Much faster

(12) a. *More before . . .b. Much before . . .

Modification by bare degree phrases:

(13) a. Three days/hours/minutes/seconds/years beforeb. Three days/hours/minutes/seconds/years earlier

(14) *Three days/hours/minutes/seconds/years earliest

Slovenian and Serbo-Croatian Moreno Mitrovic(p.c.): Slovenian has two Befores:

(15) prej = adverb derived from / based on a preposition pred- (locative ‘before’) –takes DP

(16) preden = relational / comparative adverb, cf. SC prije nego CP = ‘before thanCP’ (1) – takes CP

Serbo-Croatian (SC) has only one ‘before’, the prej typeWhen SC prije takes a CP, there’s comparative syntax involved:

(17) Prijebefore

negothan

stowhat

samaux

gahim.cl

videosee.part.sg

‘Before I saw him’

Also possible with DPs: comparative nego ‘than’ and the wh-item absent

(18) Dosaocame.part.m.sg

samaux

prijebefore

njega.him.acc

‘I came before him. (I was here before he was.)’

4

Page 5: Alexiadou and Tsoulas- Privative and Binary Features

CP-paraphrase of the latter:

(19) Dosaocame.part.m.sg

samaux

prijebefore

negothat

stowhat

jeaux

onhe.nom

(dosao).

‘I came before he did.’

In Slovenian, the two types of ‘before’ are in fact different only in presence/absence of thisadditional comparative+wh structure which is obligatory in SC. Slovenian preden is in factcomposed of pred+nego

(20) Prispelarrived.part.m.sg

semaux

predenbefore

/ *prej soaux

meme.acc

klicali.called.part.m.pl

‘I arrived before they called for me.’

DP/CP

‘before’ + DP

(21) Predbefore.prep

sestankom.meeting

‘before the meeting’ Slo

Slovenian uses a preposition here, not the basic adverb prej or the relational/comparativebefore preden

(22) Prijebefore

sastanka.meeting

‘before the meeting’ SC

‘before’ + CP

(23) Predenbefore.prep+com

jeis

JanesJohn

prispel.arrived

‘before John arrived’ Slo

(24) Prijebefor.textscprep

negothan.com

stois

jeJohn

Jovanarrived

dospio.

‘before John arrived’ SC

2.4 Syntactically Interpretable Features

Now, it is clear, if not completely formalized, that there is a relatively restricted domainof meanings including, as a first approximation number, amount, time, place as well as thedemonstrative elements this, that, here, there which can have the type of nominal exponentwe have just discussed. The analysis that Kayne proposes entails a rich decompositionwhen applied to items such as those above, whereby any extra meaning one would naturallyassociate with such elements must be expressed through unpronounced nominals such asNUMBER and AMOUNT. Beyond this somewhat limited domain, however, it is more

5

Page 6: Alexiadou and Tsoulas- Privative and Binary Features

difficult to understand what the relevant feature of nouns or verbs in general would be.It seems to us that this cannot be on a par with features such as [SMALL] or any otherfeature that relates directly to the conceptual semantics of the nominal. At the same timewe take Kayne’s analysis as suggestive evidence for the viability of the approach and thefact that at least one of the features visible to the syntactic derivation must be grounded inthe semantics. But how rich should this grounding be? Is there a single semantic propertythat the visible feature should be representing so to speak?

3 Nominal Ontology

The denotational space covered by nominals can be partitioned in three main subsets:Kinds, Singular Properties, and Number Neutral Properties (Including count plurals andMass terms). This is what Chierchia (2010) calls the semantic triad. There are definedmaps that link the three domains:

(25)

c the kind: type ⟨e⟩

∪ ∩

CAT plural/number neutral property

AT ∗

cat singular property

A natural proposal then is that nominal roots freely associate with one of these do-mains. Now, Building on Marantz (1997), Embick (2010) and Borer (2005a) we assumethat nominals are built in the syntax and involve combination of roots with elements fromthe functional vocabulary. We take the semantic information carried by the root to berelated to the denotational domain that the root is associated with (but does not reduceto its functional type). Now if this is true of nominals across the board, the question iswhether the non lexical (above the root) information is sufficient in order to identify andunambiguously map nominals onto these domains. We suggest that it is not. Note furtherthat if all information were available above the level of the root one would expect there tobe specific syntactically realised structure that identifies nominals as Kind denoting. Thisdoes not seem to be the case. Crucially, the syntax ”needs” to know that a nominal is kinddenoting (as it triggers other processes) but cannot determine it. It follows that the featurein question must be an inherent part of the root, i.e. preceding its introduction in thederivation. But this picture is unnatural in one respect, namely that it reduces effectivelythe singular/plural(number-neutral) distinction to a lexical root-level distinction. But theoperation that yields the plural or number neutral property does not seem to represent thesame type of feature as the one that codes for kind or property denotation. From one pointof view we can see the difference in terms of privativity. The root level features are privativewhereas other features must be somehow fixed in the syntax. The feature responsible for

6

Page 7: Alexiadou and Tsoulas- Privative and Binary Features

the plural property is such a feature, and it can take the values +/- plural (or whatever theway in which each language codes this distinction) . Where +plural is the count plural and-plural is the mass property, rather than the singular. Adopting the general stance outlinedabove turns out to have a number of consequences for nominal syntax and semantics as wellas for the crosslinguistic distribution of bare nominals, and the semantic status of ϕ-featuresets.

3.1 Greek Kinds: Alexopoulou et al. (2013)

Two main but related proposals about kinds can be found in the literature. First, Carlson’soriginal proposal treats kind denoting expressions as names of kinds. On the other hand,Chierchia’s neo-Carlsonian approach treats them as (intensional) maximal collections ofindividuals. The extent to which these approaches are mutually exclusive is unclear andit is conceivable that both can be realised. In other words, it is possible that naturallanguages provide both names for kinds but also, in other circumstances, functions thatyield denotations equivalent to that of a kind (qua maximal collection of instances of thekind).

In Greek the situation is as follows: A number neutral root can be used as the com-plement to a Num head which is specified say as [+singular] which triggers an atomicityimplicature or [-singular] which triggers a multiplicity implicature. Let us now assume, fol-lowing Krifka (2010) that implicatures may sometimes be reified, or folded into the meaningof certain elements. As implicatures, they may be cancelled. Let us assume that cancella-tion does not only amount to overtly contradicting them but also to a choice of the speakersimply not to reify the relevant implicatures. This would be one case of neutralising theimplicature. We thus have four possibilities:

(26) a. Number Neutral Root - [+ singular] - [+atomicity]b. Number Neutral Root - [+ singular] - No implicaturec. Number Neutral Root - [- singular] - [+multiplicity]d. Number Neutral Root - [- singular] - No implicature

What would each of these cases give us? Restricting attention to bare nouns, (26-a) willyield a bare singular which is able to be used argumentally since Num is the argumentizinghead but cannot denote a kind as it is a predicate of individual level entities. It can,nonetheless be the basis of a kind expression through the application of the and in thatcase:

(27) the(N[+singular,+atomicity]) = The name of a kind

The same is true of (26-c). The bare plural in this case will be possible in argument positionsbut not as a kind denoting element. In this case

(28) the(N[−singular,+multiplicity]) = A kind,qua maximal collection of individuals.

The two remaining cases are those where, theoretically, the bare noun should be ableto refer to kinds. As we saw in the previous sections such cases do exist, but are limited.However, such readings must be instances of (26-b) and (26-d) above, that is, cases where

7

Page 8: Alexiadou and Tsoulas- Privative and Binary Features

the implicatures are neutralised (or cancelled). Such examples then should be possibleonly when contextually relevant factors force either the neutralisation of the implicature or,equivalently, allow the speaker to choose not to reify the relevant implicatures.

The case of coordinated nouns is a case in point:

(29) a. Dinosavridinosaurs-nom

keand

tiranosavrityrannosaurs-nom

ehounhave-3pl

eksafanistidisappeared

edohere

kaiand

ekatomiriamillions

xroniayears

Dinosaurs and Tyrannosaurs have disappeared millions of years ago.b. (Karharies

(sharks-nomkai)and)

faleneswhales-nom

spanizounare-rare

stoin-the

SaronikoSaronic

Sharks and whales are rare in the Saronic gulf.

The picture emerging from Alexopoulou et al. (2013)is the following:

Noting the crosslinguistic connection between kind reference and nominal bare-ness we have reached the following conclusions for Greek: first, determinerlessnumber phrases may be type shifted but carry multiplicity/atomicity implica-tures that must be neutralised for the bare NumP to denote a kind. Whenthese implicatures are neutralised bare nouns can refer to kinds as the exampleslike (29) show. When this is not the case, the definite article with its maxi-mal/universal quantificational force must be used. Interestingly, in the case ofa definite singular, we are led to the conclusion that it names a kind, whereasin the case of a definite plural it denotes a maximal collection of individuals,which is another way to get to the kind.

The upshot of the discussion on the expression of Kinds in Greek is that there is adifference between number neutrality/numberlessness and the plural denotation. As aresult one might wish to reduce the semantic triad to a pair a la Chierchia (1998)

4 Nominalizations

Grimshaw (1990) argues in detail that de-verbal nouns do not form a homogeneous class.6

They are argued to be ambiguous between a complex event reading that supports argu-ment structure (AS), and a result/referential (R)-reading that does not. (30-a) instantiatesthe AS-interpretation of the nominal, while (30-b) instantiates the R one.

(30) a. the examination of the patients took a long time AS

6Note that Grimshaw actually distinguishes between three classes of nominals:

(i) Complex event nominals that license AS

(ii) Event nominals that do not license AS but still have an eventive interpretation and

(iii) Result nominals that do not license AS and lack an eventive interpretation.

8

Page 9: Alexiadou and Tsoulas- Privative and Binary Features

b. the examination was on the table R

The table in (31) summarizes the criteria Grimshaw introduced to distinguish betweenthe two types of nominals in English (Alexiadou, 2009; Borer, 2001, 2013):

(31)

R Nominals AS Nominals

1. Non-θ -assigner, No obligatory arguments θ-assigners, Obligatory arguments

2. No event reading Event reading.

3. No agent-oriented modifiers Agent-oriented modifiers

4. Subjects are possessives subjects are arguments

5. by phrases are non-arguments by phrases are arguments

6. No implicit argument control Implicit argument control

7. No aspectual modifiers Aspectual modifiers.

8. Modifiers like frequent, constant only withplural

Modifiers like frequent, constant appear withsingular

9. May be indefinite Must be definite

10. May be plural Must be singular

Let us concentrate on properties (9) and (10): according to Grimshaw, AS nominals canonly be definite, while result nouns may be modified by the indefinite determiners a, one.

(32) a. *An examination of the cat was interrupted by the fireworks.b. One exam was rejected because it was written in red ink.

Moreover, the claim is made that AS nouns behave like mass nouns, they cannot plu-ralize. R- nouns are count nouns, and they may pluralize. When AS nominals pluralize,they are no longer AS nominals, rather they carry the R-interpretation (33-c):

(33) a. *{the, some, a lot of} examinations of the catb. one exam, two examsc. one examination, two examinations

However, this claim has been refuted in the more recent literature. Across languages,AS nominals can pluralize:

(34) French (Roodenburg 2006, also on Italian):

a. lesthe

destructionsdestructions

frequentesfrequent

deof

quartiersquarters

populairespopular

pourfor

eleverraise

des tourstowers

staliniennesstalinistThe frequent destruction of popular quarters to build stalinist towers

b. *the destructions of the city by the enemies

Romanian (Iordachioaia and Soare, 2007): Romanian has two types of event nominal-izations: infinitive and supine. As (36) shows, the infinitival one can pluralize, while thesupine one cannot:

9

Page 10: Alexiadou and Tsoulas- Privative and Binary Features

(35) demolarea/demolish-Inf-the/

demolatuldemolish-Sup-the

cartierelorquarters-Gen

vechiold

deby

catrecommunists

comunisti

(36) demolarile/demolish-Inf-Pl/

*demolaturiledemolish-Sup-Pl

frecventefrequent-Pl

alequarters-Gen

cartierelorold

vechiby

decommunists

catre comunisti

Crucially, in Romanian, the infinitival nominalization can pluralize both under an ASand a non-AS supporting reading.

4.1 English

Mourelatos (1978), Borer (2005b, 78f), Alexiadou et al. (2010): Telic event nominals canpluralize and can appear with indefinite determiners. This concerns the –ing of nominals(nominal gerunds) and -ion nominals, but not verbal gerunds. Verbal gerunds cannotpluralize (37-d):

(37) a. There were three late arrivals of a trainb. There was (*a) pushing of the cart by Johnc. There was one pushing of the cart to New York by Johnd. John’s pushing(*s) the cart to New York

The event entailed by the gerund is imperfective (Pustejovsky, 1995), irrespectively ofthe Aktionsart involved. Verbal gerunds cannot be interpreted as R-nominals and they arenecessarily definite.

4.2 German

Pluralization only with -ung nominals, infinitival nominalizations do not pluralize: -U ngnominals project either the theme or the agent of an activity verb (see (), so they are notnecessarily telic. In this case, a plurar under the R-reading is possibible (b). However, with“-ung” nominals derived from telic verbs which project the theme obligatorily, plural ASNscan occur just like in English:

(38) a. diethe

Beobachtungobserve-UNG

desthe-Gen

Verdachtigen/suspect/

derthe-Gen

Polizeipolice

b. diethe

Beobachtungenobserve-Ung-Pl

derthe-Gen

Polizeipolice

(Ag) R-plural

(39) a. diethe

Totungkill-UNG

desthe-Gen

Feindes/enemy-Gen/

desthe-Gen

Verbrecherscriminal-Gen

(Th/*Ag)

b. diethe

gezieltentargeted

Totungenkill-UNG-Pl

derthe-Gen

politischenpolitical

Fuhrerleaders

durchvia

diethe

Armeearmy

ASN plural

(40) a. diethe

jahrlicheannual

berprufungencontrolling

desof-the

Betriebsfirm

10

Page 11: Alexiadou and Tsoulas- Privative and Binary Features

b. ReinholdsReinhold’s

Besteigungenclimbings

desof-the

K3K3

c. diethe

Entsorgungdisposals

derof-the

Atomfirmanuclear-firm

Ehrich (2002): while the singulars denote single events, the corresponding plurals denotesequences of iterated events. ’Plurals are thus homogeneous objects comparable to thedenotations of mass nouns’.

4.3 Greek

Derived nominals in Greek can have two forms. One of the instantiations involves a certainspecial affix that attaches to a verbal stem and creates a deverbal noun. The most commonaffixes are: -m-, -sim-, -s-. A second formon involves forms that basically attach theclass/number marking affixes to the verbal stem/root, which might undergo vowel gradation(41-c):

(41) a. kathariz-oclean-verb

→ kathariz-m-acleaning

b. tragudosing-verb

→ tragudis-t-issinger

c. katastrefodestroy-verb

→ katastrof-idestruction

Kolliakou (1995): there are certain restrictions on -m- affixation. Specifically, she notesthat prototypical state and accomplishment predicates do not give grammatical nominal-izations when they combine with the affix -m-.

(42) a. *agapima (loving) *skepsimo (thinking)b. *dolofonima (assassinating) *katastrema (destroying) Koliakou (1995, 211f.)

She notes that a sub-set of -m- nouns denote activities, as in (43), while others denoteconcrete nouns, as in (44):

(43) Activities

a. perpataowalk

→ tothe

perpati-m-awalk

b. sprohnopush

enaa

karotsicart

→ tothe

sprok-sim-opushing

tuof-the

karotsiucart

(44) Concrete nouns

a. paraskevazoproduce

→ paraskevasmaproduct/concoction

However, a closer look at the possible and impossible formations reveals finer details(Alexiadou, 2011). Certain accomplishment predicates can build -m- nouns, as illustratedbelow. On the other hand, achievement predicates cannot built -m- nouns at all:

11

Page 12: Alexiadou and Tsoulas- Privative and Binary Features

(45) Accomplishments

a. htizobuild

enaa

spitihouse

→→

tothe

htisimobuilding

enosof-a

spitiuhouse

b. zografizodraw

enaa

kiklocircle

→→

tothe

zografismadrawing

enosof-a

kiklucircle

(46) Achievements

a. anagnorizorecognize

→→

ithe

anagnorisi/*anagnorismarecognition

tuof-the

kleftithief

b. ftanoarrive

→→

ithe

afiksiarrival

//*tothe

afigmaarriving

c. krignioexplode

→→

ithe

ekriksi/*toexplosion

ekrigma

Achievement nominalizations with -m- do not receive a different interpretation. Theyare just out. This suggests that -m- affixation does not introduce aspectual shift, it ratherintroduces non-culmination, extension of activity. Hence it gives grammatical results onlywith those accomplishments which can receive an extended interpretation. On the samevein, destruction as well as assassination are out with -m-, as they cannot be interpreted asactivities. The state of affairs presented is reminiscent similar to what Borer (2005b) notesfor the English nominalizer -ing. As already noted by Kolliakou, plural marking is illicit on-m- nouns. Only telic nouns can surface with plural marking:

(47) a. *tathe

plisimatawashings

tonthe

piatondishes-gen

b. *enaa

plisimowashing

tonthe

piatondishes-gen

c. ithe

dolofoniesmurderings

nearonyoung

koritsiongirls-gen

4.4 General Conclusions:

• Across languages, telic/bounded nominals can pluralize.

• In agreement with Borer (2005b): Grimshaw was wrong about telic AS nominals, butshe was right about atelic AS nominals.

• Atelic AS nominals are mass nouns and thus cannot pluralize unless they are inter-preted as R-nominals.

• This predicts that if an R-interpretation is independently unavailable, the nominal-izations will not have plurals.

4.4.1 On the internal structure of ASNs

Formation of ASNs is syntactic (see Marantz 1997; Alexiadou 2001; Borer 1993, 2013).

(48) What does the internal structure of ASNs that pluralize look like?

12

Page 13: Alexiadou and Tsoulas- Privative and Binary Features

There seems to be a correlation between a certain kind of nominal internal structure, theability to pluralize and the availability of R-readings.

(49) What does it mean to have a verbal internal structure one as opposed to a nominalor a mixed one (verbal + nominal)?

All nominals contain the DP layer.

(50) a. [DP [............]]b. [DP [NominalFP....]] nominal internal structurec. [DP [VerbalFP...]] verbal internal structured. [DP [NominalFP... [VerbalFP...]]] mixed internal structure

Non-derived nominals contain only nominal functional layers internally (50-a). A nom-inal internal structure is characterized by the presence of nominal functional projectionsbelow D ((50-a)/(50-c)). A verbal internal structure is characterized by the presence ofverbal layers below D. A mixed internal structure contains both nominal and verbal layersand nominals with that structure show a mixed behavior .

• ASNs that pluralize contain whaterver layer of structure plural morphology is intro-duced in: NumberP, ClassP.

• Most ASNs that do not pluralize lack this layer, e.g. verbal gerunds:

(51) Correlation to be explainedAS/R-nominals and the mass vs. count noun distinction

4.5 Aspect and Pluralization

As known, the mass/count distinction in the nominal domain has often been compared toaspectual distinctions in the domain of verb phrases see e.g. (Mourelatos, 1978; Bach, 1986;Krifka, 1992; Borer, 2005b) to mention just a few:

Certain amount of consensus:

(52) count nouns are similar to bounded events

(53) mass nouns are similar to (Un)bounded events

As known, activities can become bounded, when endpoints are introduced. The same appliesin the domain of NPs:

(54) a. Sue is running. Sue has runb. Sue is running a mile / Sue has run a milec. There was one pushing of the cart to New York by Johnd. There was (*one) pushing of the cart by John

Let us assume that the problem lies in ’matching the information’ of the lower verbalstructure and the upper nominal structure, for those nominalizations that have an internalnominal structure.

Introducing (un)boundedness (see also Engelhardt 2000:

13

Page 14: Alexiadou and Tsoulas- Privative and Binary Features

(55) a. Pluralization (inflectional plural, nominal structure)b. Aktionsartc. Morphological Aspect

If the function of pluralization, is to introduce unboundedness, pluralization of structuresthat are already unbounded via Morphological Aspect, realised in AspP, and/or Aktionsart,realised within the Voice/vP would be non-sensical. This is the case with the verbal gerundsin English (and supine in Romanian).

(56) What happens to mass nouns and atelic ASNs?

Pluralization of atelic ASNs is out for the same reason pluralization of a mass noun isanomalous. It is only allowed if the noun is able to be construed as picking out distinct units.In the case of derived nominals pluralization is possible only under an R-interpretation, i.e.when no AS is licensed.

Why?

As known, there is a difference concerning boundednes in the verbal domain as opposedto boundedness in the nominal domain: the former makes reference to VPs and not to

’lexical’ items. Since atelic derived nominals are VPs, which are already unbounded, furtherpluralization is out. However, the ‘nominal’ part can pluralize, in the absence of AS, sinceit is not unbounded.7

5 General Conclusions

Regarding the notion of features we are led to the following conclusions. Given

(57) The framework that we have assumed

(58) The idea that there is a true difference between the notions of privative and valuedfeatures (i.e. they are not notational variants)

(59) That there is a distinction between content and specification

Then:

(60) “Content” features are privative and limited to the denotational sort of nominals.

(61) The denotational sorts may be simply properties and kinds (i.e. the number neutralproperty is not basic)

(62) Specificational features are valued

Finally turning to the question of innateness. Our current thinking leads us to a positionwhere would be able to live with the suggestion that content (privative) features are innatewhereas specificational ones are not as such. Rather they represent a way to interact withthe data and they are in some sense emergent.

7Remark: Greek -m- functions as a kind of lexical plural and thus blocks pluralization.

14

Page 15: Alexiadou and Tsoulas- Privative and Binary Features

References

Acquaviva, P. (2012). The roots of nominality, the nominality of roots. Ms UniversityCollege Dublin.

Adger, D. and P. Svenonius (2011). Features in minimalist syntax. In C. Boeckx (Ed.),The Oxford Handbook of Linguistic Minimalism, pp. 27 – 51. Oxford: Oxford UniversityPress.

Alexiadou, A. (2001). Functional Structure in Nominals. Nominalization and Ergativity.Amsterdam: John Benjamins.

Alexiadou, A. (2009). On the role of syntactic locality in morphological processes: the caseof (greek) derived nominals. In A. Giannakidou and M. Rathert (Eds.), Quantification,Definiteness and Nominalization, pp. 253–280. Oxford: Oxford University Press.

Alexiadou, A. (2011). The aspectual properties of nominalization structures. In A. Galani,G. Hicks, and G. Tsoulas (Eds.), Morphology and its Interfaces, pp. 195–220. John Ben-jamins.

Alexiadou, A., G. Iordachioaia, and E. Soare. (2010). Number/aspect interactions in thesyntax of nominalizations: a distributed morphology approach. Journal of linguistics 46,537–574.

Alexopoulou, T., R. Folli, and G. Tsoulas (2013). Bare number. In R. Folli, C. Sevdali,and R. Trusswell (Eds.), Syntax and its limits. Oxford: Oxford University Press.

Arad, M. (2005). Roots and Patterns. Berlin: Springer.

Bach, E. (1986). The algebra of events. Linguistics and Philosophy 9 (1), 5–16.

Borer, H. (1993). Parallel morphology. Ms. University of Massachussetts, Amherst.

Borer, H. (2001). Morphology and syntax. In S. Andrew and Z. Arnold (Eds.), The handbookof morphology, pp. 151–190. Oxford: Blackwell Publishers.

Borer, H. (2005a). Structuring Sense: In Name Only, Volume 1. Oxford: Oxford UniversityPress.

Borer, H. (2005b). Structuring Sense: The normal course of events, Volume 2. Oxford:Oxford University Press.

Borer, H. (2013). Structuring Sense: Taking Form, Volume 3. Oxford: Oxford UniversityPress.

Chierchia, G. (1998). Reference to kinds across languages. Natural Language Seman-tics 6 (4), 339–405.

Chierchia, G. (2010). Mass nouns, vagueness and semantic variation. Synthese 174 (1).

15

Page 16: Alexiadou and Tsoulas- Privative and Binary Features

Del Prete, F. (2008). A non-uniform semantic analysis of the italian temporal connectivesprima and dopo. Natural Language Semantics 16 (2), 157–203.

Ehrich, V. (2002). The thematic interpretation of plural nominalizations. In ZASPIL 27.

Embick, D. (2010). Localism versus Globalism in Morphology and Phonology. Number 60in Linguistic Inquiry Monographs. Cambridge, Mass: MIT Press.

Engelhardt, M. (2000). The projection of argument-taking nominals. Natural Languageand Linguistic Theory 18 (1), 41–88.

Grimshaw, J. (1990). Argument Structure. Cambridge, MA: MIT Press.

Grønn, A. and A. von Stechow (To Appear). Tense in adjuncts. Language and LinguisticsCompass.

Iordachioaia, G. and E. Soare (2007). Two kinds of event plurals: evidence from romaniannominalizations. Paper presented at the Syntax and Semantics 7 Conference in Paris,October 2007.

Katz, J. J. and P. M. Postal (1964). An integrated theory of linguistic descriptions. Cam-bridge, Mass.: MIT Press.

Kayne, R. S. (2005). Movement and Silence. Oxford and New York: Oxford UniversityPress.

Kayne, R. S. (2010). Comparisons and Contrasts. Oxford and New York: Oxford UniversityPress.

Koliakou, D. (1995). Definites and Possessives in Modern Greek: an HPSG Syntax forNoun Phrases. Ph. D. thesis, University of Edinburgh.

Krifka, M. (1992). Thematic relations as links between nominal reference and temporalconstitution. In I. Sag and A. Szabolsci (Eds.), Lexical Matters, pp. 30–52. University ofChicago Press.

Krifka, M. (2010). How to interpret ”expletive” negation under bevor in german. In T. H.. G. Fanselow (Ed.), Language and logos. Studies in theoretical and computational lin-guistics, pp. 214–236. Berlin: Academie Verlag.

Marantz, A. (1997). No escape from syntax: Don’t try morphological analysis in the privacyof your own lexicon. In A. Dimitriadis, L. Siegel, C. Surek-Clark, and A. Williams (Eds.),Proceedings of the 21st Annual Penn Linguistics Colloquium, Volume 4.2 of University ofPennsylvania Working Papers in Linguistics, Philadelphia, pp. 201–225.

Mourelatos, A. (1978). Events, processes and states. Linguistics and Philosophy (2), 415–434.

Pollock, J.-Y. (1989). Verb movement, universal grammar and the structure of IP. LinguisticInquiry 20, 365–424.

16

Page 17: Alexiadou and Tsoulas- Privative and Binary Features

Pustejovsky, J. (1995). The Generative Lexicon. Cambridge, Massachuesetts: The MITPress.

Roodenburg, J. (2006). On the existence of plural event nominalizations: a view fromromance. MS Universitat Stuttgart.

Sharvit, Y. (2013). On the universal principles of tense embedding: The lesson from before.Journal of Semantics.

Smyth, H. W. (1920). Greek Grammar for Colleges. American Book Company.

Tsoulas, G. (2013). The grammar of silent time. Ms. University of York.

17