Lexical semantics and syntactic structure

LEXICAL SEMANTICS AND
SYNTACTIC STRUCTURE
BETH LEVIN & MALKA RAPPAPORT HOVOV
1
OUTLINE
 0. Introduction
 1. A Brief Introduction to the Unaccusative Hypothesis
 2.Verbs of Sound: an Introduction
 3. The Variable Behavior of Verbs of Sound
 4. Further Confirmation for the Semantic Class Shift
 5. The Scope of the Semantic Class Shift
 6. Implication s for the Theory of Linking
 7. The Source of the Multiple Meanings
 8. Conclusion
2
INTRODUCTION
 The syntax of a sentence is determined by the meaning of the
predicator in that sentence
 “linking regularities”  “linking rule”
 Associate arguments bearing certain semantic roles with certain
syntactic expression
 In English, arguments bearing agent are commonly expressed as the
syntactic subject
3
 Three Qs
1. The extent to which the syntactic expression of arguments is
predictable
 Fully predictable or idiosyncratic?
2. The nature of the lexical semantic representation
 “role-centered” vs. “predicate-centered”
3. Cross-linguistic variation
4
Q1: To what extent the syntactic expression is predictable
 Chomsky: subcategorization frames (i.e. syntactic
expression of arguments) are fully predictable.
 Jackendoff and Rosen: more skeptical
 Author’s assumption: the mapping between lexical
semantic representation and syntactic expression is fully
predictable
 Proof: the case study in the paper
5
Q2 The nature of the lexical semantic representation
 Role-centered:
 representations were formulated in terms of the name of the semantic
roles, such as agent, patient, theme.
 Fillmore’s case grammar (1968) and Gruber’s thematic relations (1965)
 Predicate-centered:
 Focus on aspects of the meaning of a predicator that is relevant to the
syntax.
 Jackendoff (1983, 1990), Carter (1997), and Pinker (1989), Levin and
Rappaport (1988)
6
 Lexicalization Patterns: generalizations concerning the types of
meaning that can be associated with the verbs of a language.
Verbs of the same class often share a common core but differ in
other meaning components.
 “It appears that the syntactically relevant components of
meaning can be better expressed in predicate-centered
approaches to lexical semantic representation.”
7
Q3: Cross-linguistic Variation
 A certain degree of variation exists between languages
 The variation can be attributed to differences in lexicalization patterns, rather than to the
set of meaning components which are relevant to the linking rules.
 Translation equivalents in two languages may differ in the syntactic expression of arguments due to the
fact that the elements of meaning lexicalized along with the core meaning may differ
 Languages may not give the same weight to each component so that a particular argument may not
receive the same expression in different languages.
 E.g., Could you suggest some books to me?  mentioned a few books
可否建議我一些書?  告訴我一件事
8
WHAT’S IN THIS PAPER
 What are the syntactically relevant semantic components?
 A case study: verbs of sound and verbs of manner of motion
 Consider the behavior of these verbs in the context known as
the Unaccusative Hypothesis
9
A BRIEF INTRODUCTION TO
THE UNACCUSATIVE HYPOTHESIS (UH)
 Intransitive verbs fall into two sub-classes: “unaccusative” vs.
“unergative”
(1) a. Unaccusative Verb: __ [VP V NP]  telic, derived subject
(1) It melts/drops/falls  __ melts/drops/falls it
b. Unergative Verb:
NP [VP V]
 agentive, subject at all levels
(2) He eats/drinks
 There are semantic generalizations regarding class membership
10
VERBS OF VARIABLE BEHAVIOR
 There are verbs which sometimes
display unaccusative behavior and
sometimes unergative:
 Still semantically predictable or not?
 A close look at such verbs reveals that
they are always associated with more
than one meaning:
Variable
behavior
Unaccusative
Unergative
Unacc.
member
Unerg.
member
 When displaying unaccusative behavior, it
is independently correlated with
membership in the unaccusative class,
and vise versa.
11
VERBS OF SOUND: AN INTRODUCTION
 A semantic class: describe the emission or production of a sound
 Differentiated by the physical properties of the sound and by its
manner of production
 Members: beep, buzz, creak, gurgle. Jingle, ring, roar, rumble, screech…
Cars honked and hummed in the road
I hissed and snarled at them.
 Intransitive verbs: taking the emitting argument as subject
 Take a range of SUBJ: animate and inanimate concrete nouns
12
VERBS OF SOUND AND UH
 Two semantic determinants of unaccusativity:
telicity and agentivity
 Unaccusative: telic
Unergative: agentive
 Sound verbs: do not fall clearly into either group
 As atelic verbs  unergative verbs
He is whistling.
 If take non-agent SUBJ  unaccusative verbs
The door bell rings (>I rang the door bell.)
13
 Verbs of sound are basically taken to be unergative:
 Do not select the Aux BE: cf. It is broken
 May take an object and assign accusative case
 Unaccusative V: cannot assign case  cannot take any
direct object  *The vase broke the ??
 Unergative V: can assign case  The bell jangled its first
summons (external argument)
 May occur with non-subcategorized object X’s way:
 he whistled his way out.
14
THE VARIABLE BEHAVIOR OF VERBS OF SOUND
 How verbs of sound show variable behavior:
 The resultative construction:
 Two patterns
1. The bottle broke open.  unaccusative verbs, with no object
2. …the other officers laugh themselves helpless.  unergative, with an OBJ
 Vs of sound are found in both patterns:
 Unassuative: The curtains creak open.
 Unergative: We yelled ourselves hoarse.
15
 Corpus data:
(12) a. We searched the wood and cliffs, yelled ourselves hoarse and imagined you
drowned…  unergative
(13) a. …the curtains creak open and radiant evening light steams into the cluttered
room  unaccusative
 Problem: Given the initial assumption that verbs of sound are unergative, the use of
verbs of sound in the unaccusative resulative pattern demands an explanation.
 Proposal: Verbs of sound have two different but related meanings, each correlated
with a different classification.
16
BECOMING VERBS OF DIRECTED MOTION
 Verbs of sound have the option of becoming verbs of directed motion:
 Verbs of directed motion are known to be unaccusative  The bus arrived/moved.
 Require a directional phrase as a complement
(14) a. …the elevator wheezed upward
b. …a flatbed truck…rumbled through the gate
 Describes the motion of an entity, characterized by the concomitant emission of the
sound  The elevator moved upward while wheezing.
 When verbs of sound are used as directed motion Vs  show unaccusative behavior
17
 Resulative state as resulative position:The door banged shut.
FURTHER CONFIRMATION FOR
THE SEMANTIC CLASS SHIFT
 Agentive verbs of sound cannot in general become verbs of
directed motion
 The sound is emitted via vocal tract, not emitted as a consequence of
motion  not qualify for a directed motion sense
(19) a. *He yelled/shouted down the street
b. *The frogs croaked to the pond
18
 But occasionally
(20) a. …Sedgwich often clanked into town…
b. She rustled out of the room…
 These sounds, while have an agentive subject, are never emitted by the vocal tract,
but by articles of clothing and accessories
 Verbs appear in the unaccusative resultative pattern only when the
sound is not emitted by the vocal tract:
 No unaccusative resultative with direction or position, if the sound is emitted
via the vocal tract (internal sound):
(22) a. *He yelled down to the falling rock.  They yelled themselves coarse.
b. *The frogs croaked apart.
(23) a. We splashed clear of the oncoming boat.
19
THE SCOPE OF THE SEMANTIC CLASS SHIFT
 There is a more widespread phenomenon in English: “Verbs
from several semantic classes may become verbs of directed
motion”
 Verbs of sound
 Verbs of manner of motion (e.g. run, fly, shuffle)
(24) a. The children ran into the room
 Verbs of exerting force (e.g. push, pull)
(25) a. Kim pushed the stroller into the store
20
VERBS OF MANNER OF MOTION
 Verbs of manner of motion are unergative in the basic sense (i.e.
nondirectional motion), as expected from their agentive nature
 But they display unaccusative behavior in their directed motion sense:
 Resultative construction:
(28) a. Don’t expect to swim yourself sober!  unergative, contain result state
(29) a. She danced/swan/sprinted free of her captors  unaccusative, result position
 The verbs appear in the unaccusative resultative pattern only on the their
directed motion sense
21
IMPLICATION FOR THE THEORY OF LINKING
 Traditionally
 Agentivity determines unergative classification
 Telicity determines unaccusative classification
 How about verbs which are neither telic nor agentive and verbs
which are both telic and agentive?
22
WHAT DETERMINES THE CLASSIFICATION?
 The verbs of sound with inanimate emitters exemplify verbs
which are both atelic and nonagentive, therefore:
 The notion of “agent” does not figure directly in the linking rule
which maps certain arguments into d-structure SUBJ
23
 Telicity may not be relevant to the classification here,
either:
 Verbs of sound with inanimate emitters
 Supposed to be unergative  atelic and nonagentive  may telicity
determines the classification?
 If so, then unergative Vs are atelic? No, because there are atelic
unaccusative verbs, such as rock and roll which participate in the
causative alternation.
 Telicity does not necessarily determine an unergative classification
24
“INTERNALLY CAUSED” VS. “EXTERNALLY CAUSED” EVENTUALITY
 Then, what is responsible for inanimate emitters being
expressed as the d-structure subject so that verbs of sound
would be classified as unergative?
 The “causer” argument is more important here!
25
 Verbs, including intransitive verbs, should be subdivided into
those which denote “internally caused eventuality” and those
which denote “externally caused eventuality”
 Intransitive verbs that denote internally caused eventuality
 Causer of the sound, be it animate or inanimate, is taken to be the
subject.
 Causer Linking Rule: associating internal causer with the d-structure
SUBJ
26
THE NOTION OF “CAUSER”
 Agent is one type of causer argument
 Causer is not equivalent to agent
 Encompass the emitter argument of verbs of sound
 The causer may be the argument of an intransitive verb, and the verb
may even be stative
 The behavior of verbs of sound supports that the notion of causer,
instead of agentivity, is a determinant of which intransitive verbs are
unergative.
27
EXTERNALLY CAUSED EVENTUALITY
 These eventuality typically involve two sub-events:
 The causing event, which includes the causer as an argument
 The caused event
 Causer Linking Rule
 Transitive verbs (e.g. I break the vase)
 Associating the causer, be it an agent, an instrument, or a natural force, with the
d-structure SUBJ
 Intransitive verbs (e.g. The vase break)
 Describe externally caused eventuality; they arise from a process which allows
their causer argument not to be expressed syntactically  Causer Linking Rule
cannot be applied
 Unaccusative verbs
28
WHAT IS INVOLVED?
If telic, then unaccusative
Atelic V?
If atelic, then unergative in general, but…
Causer ?
If Internally caused (yell, hum, grumble),
then Internal causer is the underlying
subject 
I yelled myself coarse.
If externally caused (honk, roar, crack), then
external causer is the underlying subject 
I honked the car down the road.
The door cracked open. 
behaves like Directed Motion Vs
SUMMARY
 A telic single-argument verb is unaccusative, independent of whether it
denotes an internally or an externally caused eventuality
 For atelic single-argument verbs:
 We can identify unergative verbs with internally caused intransitive verbs, and
unaccusative verbs with the intransitive variant of externally caused verbs
 The syntactic behavior of verbs belongs to particular semantic classes
30
THE SOURCE OF THE MULTIPLE MEANINGS
 Meaning shifts are regular
 Verbs of sound,Verbs of manner of motion  manner of directed motion (“regular
polysemy”)
 Meaning shifts are productive across semantically coherent classes of
verbs
 Not all intransitive activity verbs can become verbs of directed motion in English
(36) a. *Kelly laughed/sang/swore/cried out of the room
b. *The boy laughed/sang/swore/cried clear of oncoming traffic
 Meaning shifts are rule-governed processes: they are regular and
productive
31
CROSS-LINGUISTIC EVIDENCE
 Verbs of manner of motion  Verbs of directed motion
 In English, German, Hebrew
 But not in languages, as French
 Sound verbs  Verbs of directed motion
 In English, German, Hebrew
 But not in languages, as French
 If a language allows one class of verbs to shift, it will allow the other class
to shift as well
32
CONCLUSION
 Many facet of the syntax of a sentence are determined by the
meaning of its predicator
 The variable behavior of sound verbs and verbs of manner of
motion makes no problem  it is semantically determined
 The meaning of a verb determines its syntactic expression
33