Transcript Summary and Discussion:

Summary and
Discussion:
Intertheoretical Approaches to
Complex Verb Constructions
C. Bowern, 18th March 2006.
Abbreviations
 SVC
= Serial Verb Construction
 LVC = Light Verb Construction
 (Won’t be talking about the other types)
Issues (= Recurring Themes)
 Finiteness
 Lexicality
 Sources
of SVCs and LVCs, and what
they turn into
 Continua
Back to the definition

Complex Predicates have the properties of a
single predicational head spread over several
items in the predicate.



overlapping arg structure
single event described
monoclausal (monopredicational? ie a single complex
predicate, or a set of predicates in a single clause?)
Challenge to the unity of the ‘word’ as an atom
of syntax (cf. Woodbury 2003).
 However, not all definitions invoked this – some
relied on monoclausality.

Aside: Does Bardi have
Serialization under this definition?
 LVCs
(head extensively)
 Australian Serialisation (not common)



Pilbara (“Crow is sitting chasing them” – Peter
Austin)
Arnhem Land – e.g. Yan-nhangu binmunu ‘do
thus’, garama ‘go’, gabatthun ‘run’ and a few
others.
Bardi (??)
Bardi serialization?
 Multiple
verbs under a single intonation
contour
 No conjunctions or other markers of
subordination (although these also exist)
 Argument sharing.
 (But not conventionalised? Only narrative
use)
 (Similar
difficulty in identifying LVCs in
some languages) – e.g. just because
something is preverbal, doesn’t mean it’s a
preverb
 Any solution apart from more native
speaker linguists? (these are problems
that linguists with nonnative fluency
probably can’t solve)
Finiteness
=
Inflected for person
 = Inflected for tense/aspect
 (gradient feature)
 Therefore,
SVCs are mostly symmetric,
LVCs never (?) are – but cf Thepkanjana’s
data – several Thai SVCs appear to have
similar functions to light verbs.
Finiteness, cont.
 Ambient
serialisation
 Still serialisation, but verb not necessarily
marked for tense/aspect.
 In process of grammaticalisation of SVCs
into (e.g.) PPs, must lose those features at
some point.
Loniu and Titan (Admiralty Islands)
 deverbal
preps agree for tense
Ala Yap ala-me sos, ala-sos i-ti Nauna
3pl Yapese 3pl.nfut=come meet,
3pl.nfut=meet nfut-’on’ Nauna.
‘The people from Yap met together on
Nauna.’
Lexicality

more or less lexical? based on citeria of



arg structure,
‘full semantics’ (ie comparable semantics to main verb
use)
lexicality in the sense of the phrase undergoing
lexical rules (e.g. nominalization - Bodomo)



Thai
Bardi
Farsi
 ‘Lexical’ verb
implies ‘listed in the mental
lexicon with its own entry’
 lexical = heavy verb (vs light verb)
 but some LVCs ‘lexical’ in that they’re
formed in the lexicon vs in the syntax
 If formed in syntax can’t be formed from
full verb w. argument structure changes
(implies non-monotonic theory)
 Therefore,


need two verbs
sayfull
saylight
 How
different is that from common or
garden polysemy?
 How to tell polysemy from
underspecification? [non-rhetorical
question]
Diachrony (1): ‘Results’
 The


dead-end idea.
Classifiers (Gooniyandi – McGregor 1990)
Classifiers (Udi – Harris)
 Not
great even for Indic (Pardeshi and
Hook)
 Non-uniform behaviour synchronically
(Shibatani for Atayal and Japanese; cf
Bardi)
Diachrony (2)




Sources
Other complex preds (cf Butt and Lahiri)
Parataxis
Generalisation of a rather specific construction;
e.g. English go X





go eat
come go eat
went ate
‘go suck on a neutron star’/’go bang some rocks
together four-eyes’ (Zaphod Beeblebrox)
go vituperate
Continua, Clines and the like
 What
does it mean to say that a particular
property is gradient?
 What should we do about it?
Identifying a category as gradient isn’t the end of
the story
 A variable can be gradient for many different
reasons.




Different speakers with different internal grammars
(multiple different discrete grammars)
Different but related phenomena which are
themselves discrete
Other factors which interact and give the appearance
of scalarity
Solutions

Give up
 Recognize the gradience but recognize too that
many aspects of the world are gradient but can
still be modelled discretely.



phonemes
taxa
Develop scalar models
 Be more specific about the scalarity – what
causes it (cf Shibatani on motion serialisation)
 Recognise that it’s a fundamental difference in
approach to research and the nature of
explanation and agree to differ
Areality and borrowability

Tend to cluster areally
 Therefore,
easily borrowed? (e.g.
Schultze-Berndt)
 BUT, what gets borrowed and how does it
arise?




embedding
coverb borrowing
coverb + light verb borrowing?
calquing? code-switching?
 i.e.
is this lexical borrowing, or syntactic
borrowing, or neither?
E.g. North Australia

Rather different underlying organisation, e.g. in
Northern Australia
 (Most of the differences in N. Aus. fall out from
how ‘eventive’ the coverb is – that is, how well
it’s able to act as a full predicate on its own and
what licenses the coverb as a predicate)
 Bardi coverbs can’t appear without a light verb –
not fully lexical themselves? Can’t license the
appearance of all their arguments without
another part of the predicate?
Similarities
 Semantics
 function
of marking event perspectives
Differences b/w LVCs and SVCs
 Sources

Givón: LVCs (e.g. Wagiman) typically arise
through embedding (especially where there is
asymmetric finiteness)
 Argument

sharing properties
(not discussed, but different)
 Narrative
light verbs? (NO?)
 LVCs aren’t usually (ever?) event chains
Event chains
 Wagiman:
ngi-ya-nggi woerrkge-ma maman // garatjjin
dorroh-dorroh // denh-na wirin.
1pl.go-past work well // grass pull // cut tree.
We worked well, pulling out grass, cutting trees.
 Bardi:
cf ‘serialisation’ above – NEVER coverb
chaining.
Some things we didn’t talk about (much)



Headedness
Argument sharing
Valency and valency determinates




Slave, Koyukon, Warlpiri – Valency is straightforward
Bardi – valency is messy
Semantic roles can arise through the construction
Relationship between grammaticalization and
frequency – and between stability and change
 Butt and Lahiri – LVCs are stable once they
arise. BUT LVCs aren’t uniform in many
languages.
Where to?
 Have


to look at
cognitive underpinnings – cognitive versus
grammatical definitions
processing issues
 Social/Interactive
issue – information
gain?
 the universals versus the language/culture
specific items
 Concetration
on variation makes
identification of universals difficult
 OT/LFG
 The nature of explanation