Transcript Event orientated adnominals and compositionality

Event Orientated Adnominals
and Compositionality
Yoad Winter
Joost Zwarts
Workshop Syntax and Semantics of the Nominal Domain
February 4-5, 2016
Frankfurt University
Types of Arguments
Semantic Arguments
Referential
Syntactic Arguments
Thematic
Expletive
John surprised us
dog: <e,t>
It rained
Hypothesis:
no other
types of arguments.
ate (it)
Fido
is a dog John
the dog
Orientation of Intersective Modifiers
Generalization: intersective modifiers are
orientated to the referential argument.
[[blue door]] =
λx.x is a door & x is blue
[[short giraffe]] = λx.x is a giraffe & x is short
(relative to giraffes)
[[sang loudly]] = λe.e is a singing event & e is loud
(relative to singing)
Witness:
This is a blue door  This door is blue
This is a short giraffe  This giraffe is short
John sang loudly  John's singing was loud
The Puzzle of -er Nominals
Bolinger 1967, Larson 1998:
John is a heavy smoker  John is heavy
 John smokes heavily
Hebrew: ha-potrim
nexona
the-solver-PL correctly-ADV
hem potrim nexona
they solve correctly
Hebrew’s so-called Beynoni (“participle”) form for
verbs is ambiguous between a verb and a noun.
Our approach: Hebrew and English are more similar
than what overt morpho-syntax reveals, due to the
special morpho-syntactic properties of -er nominals.
Note on Coercion
Bolinger 1967, Larson 1998:
John is a heavy smoker  John is heavy
 John smokes heavily
Pustejovsky 1995:
This Ferrari is a fast car
(i.e. it moves fast)
 This Ferrari is fast
This waltz is a fast dance  This waltz is fast
(i.e. it has a fast tempo)
Coercion is a lexical phenomenon, independent
of the -er puzzle.
Modifier Orientation: Preliminaries (1)
Why do adjuncts modify the RA, and
not any thematic argument?
John surprised us enormously
The verb surprise has 2 thematic arguments and a
referential argument (RA) – the event argument.
What orientates the modifier to the RA?
x.y.e.surprise(e,y,x) + z.enormous(z)
= x.y.e.surprise(e,y,x)enormous(e)
≠ x.y.e.surprise(e,y,x)enormous(x)
≠ x.y.e.surprise(e,y,x)enormous(y)
Modifier Orientation: Preliminaries (2)
John’s polite friend
The relational noun friend has one thematic
argument (the possessor) and one RA (the friend).
What orientates the modifier to the RA?
x.y.friend(x,y) + z.polite(z)
= x.y. friend(x,y)  polite(x)
≠ x.y. friend(x,y)  polite(y)
Modifier Orientation: Preliminaries (3)
Why do adjuncts modify the RA, and
not any thematic argument?
Standard answer: Syntax does.
Somehow – the syntax singles out the event
argument for adverbial modifiers, the possesed
argument for adnominal modifiers in posessives, etc.
How precisely?
W&Z 2011: The new formal theory of Abstract
Categorial Grammar (De Groote 2001,
Muskens 2003).
Event Orientated Adnominals
Adjectives: (Bolinger 1967, Larson 1998, Coppock 2009, Alexeyenko 2011)
beautiful dancer
heavy smoker
hard worker
violent campaigner
smooth operator
clean fighter
shallow breather
PPs: (possibly new)
destroyer of the city in 1735
killer of JFK in 1963
founder of Apple in a Silicon Valley garage
defeater of the Armada in 1588
Restrictions: (Levin/Rappaport 1988,1992, McIntyre 2010)
inducer of protein growth (*with a new technique)
Hebrew Participles (Beynoni)
(ha)potrim nexona
(the)solve-PL correctly
maxrivey ha-ir bishnat 1735
(the) destroy-PL (of the) city in-year 1735
rokdim yafe
dance-PL beautifully
rocxo shel kenedi be-1963
kill-he of Kenndy in-1963
me'aSnim bixvedut
Smoke-PL heavily
meyasda shel epel be-musax
found-she of Apple in-garage
ovdim kashe
work-PL hard
mevisey ha-armada be-1588
defeat-PL the-Armada in-1588
mesaxkim bealimut
play-PL violently
nilxamim behaginut
Fight-PL decently
mesoxaxim biydidutiyut
speak-PL amicably
Larson’s Proposal
Nominals, including -er nominals, may
have a Davidsonian event argument.
[[dancer]] = λx.λe.x is the agent of dancing event e
[[beautiful dancer]]
λx.λe.x is the beautiful agent of dancing event e
λx.λe.x is the agent of beautiful dancing event e
Questions for Larson´s proposal
1. What specifies the agent as the RA in beautiful
dancer?
Thus, why can´t it mean ``a beautiful dancing event´´?
2. What blocks ambiguity in dance beautifully?
Thus, why can’t it mean “person who dances
beautifully”?
These possible problems would appear if nonreferential arguments (e.g. events with dancer)
can be freely modified as in Larson’s proposal.
Revising Larson’s Account
Maintaining the tripartite typology of arguments
Semantic Arguments
Referential
Syntactic Arguments
Thematic
Expletive
-ER NOMINALS (cf. Williams 2003, Egg 2004):
1- have verbal projection within NP,
2- which is low within the NP,
3- and whose referential arg. is the event
1- Verbal Projection
beautifully
dance
1- Verbal Projection
nexona
potrim
2- Low within NP (Hebrew)
ha
ha-nexonim
nexona
potrim
2- Low within NP (English)
the
beautiful
beautiful
dance er
3- RA is the event (Hebrew)
ha
ha-nexonim
nexona
potrim
3- RA is the event (English)
the
beautiful
beautiful
dance er
Improvements over Larson’s proposal (1)
Questions for Larson:
1. What specifies the agent as the RA?
2. What blocks ambiguity in dance beautifully?
Answers:
1. Only one covert argument at each level – the RA.
2. Syntax does, as usual with modification and
thematic arguments.
Improvements over Larson’s proposal (2)
No event modification with non-er
nominals:
beautiful dancer
?beautiful ballerina
hard worker
?hard employee
light traveler
?light passenger
just king, stray bullet, fast horse, daily newspaper –
probably coercion
or more complex modification than predication over events
Ordering of modifiers
Larson and Takahashi 2007
Olga is a blonde beautiful dancer.
 Olga is blonde and beautiful
 Olga is blonde and she dances beautifully
Olga is a beautiful blonde dancer.
 Olga is blonde and beautiful
blonde beautiful [NP [NPer [ dancer ]]]
blonde [NP beautiful [NPer [ dancer ]]]
beautiful blonde [NP [NPer [ dancer ]]]
?beautiful [NP blonde [NPer [ dancer ]]]
Relational Nouns
champion in 1981
new president
mayor till 2014
Speculation: Relational nouns always
come with an eventuality.
Summary
 -ER nominals have a verbal layer.
 Within this layer, event orientated adnominals
act as adverbial modifiers, (only) sometimes
disguised under adjectival morphology.
 Multiple verbal + nominal layers allow us to
maintain the RA hypothesis, without
complicating compositional processes.
 Event-orientated modification of relational,
non-deverbal, nominals, requires further study.