Upload
oswald-phelps
View
237
Download
0
Tags:
Embed Size (px)
Citation preview
1
SYNTACTIC COMPLEXITY OF
DIFFERENT BASQUE WORD
ORDERS: EVIDENCE FROM
NEUROIMAGE (ERP)
Kepa Erdozia (University of the Basque Country)Itziar Laka (University of the Basque Country)Anna Mestres (University of Barcelona)Antoni Rodriguez-Fornells (ICREA and UB)
ESF/MCYT/EUROCORES: BFF2002-10379-E
2
BASQUE: Free Word Order
•Canonical Order: Subject-Object-Verb
(De Rijk, 1969, Ortiz de Urbina 1986, Elordieta 2001…)
•Derived Orders: Object-Subject-Verb
Subject-Verb-Object
Object-Verb-Subject…
•Phrases can be arranged in almost any order
3
vP
DP VP
DP V
gizon-ak emakume-a ikus-i du
man-the woman-the seen has Subject-Object-
Verb
SOV order
4
vP
DP VP
DP V
gizon-ak emakume-a ikus-i du
man-the woman-the seen has Object-Subject-
Verb
XP
emakume-a woman-the
OVS order
5
MORPHOLOGICAL AMBIGUITY
OBJECT PLURALPACIENT
SUBJECT SINGULAR
AGENT
EMAKUME-AK‘WOMAN-X’
Emakume-ak gizon-ak ikusi ditu woman-? man-? see has
‘The woman has seen the men’ or ‘The man has seen the women’
6
•SOV is processed faster•SOV is processed easier•Ambiguous Chains are processed like canonical word order (Kaan, 1997)
•LAN & P600: Syntactic Complexity in derived order (Matzke et al, 2001; Felser et al, 2003)•N400: Semantic desambiguation (Hagoort et al., 2004)
SOV and OSV ProcessingExperiments I & II:Self Paced Reading & comprehension task:
Experiment III: Event Related Potentials (ERP):
7
Experiment I: self-pace reading SOV vs OSV
Goal: to determine whether OSV sentences have a higher processing cost than SOV sentences:
(a) longer reading times
(b) comprehension problems
Participants: 23 native speakers
Materials: 2 conditions, 32 sentences per condition
2 lists :16 SOV sentences/16 OSV sentences
32 fillers (the same for two lists)
8
Experiment I: Materials
emakume-ak gizon-a ikus-i du
woman-the/Subj. man-the/Obj. seen has
‘the woman has seen the man’
gizon-a emakume-ak ikus-i du
man-the/Obj. woman-the/Subj. seen has
‘the woman has seen the man’
Subject
Object
Verb Aux
Subject
Object
Verb Aux
9
Mean Reading times of sentences: global score
•SOV order is processed faster than OSV order
Experiment I: Results
Reading time of two word orders
3500
3600
3700
3800
3900
4000
4100
4200
SOV OSV
Sentence Type
Tim
e (m
s)
p<0.005
10
•OSV order elicited more errors than SOV order
Errors in the comprehension task
Experiment I: Results
Comprehension Task: Errors
0%2%4%6%
8%10%12%14%16%
SOV OSV
Sentence Type
Per
cen
tag
e
p<0.001
11
Mean reading times Word by Word
•Unmarked form processed faster than marked
Experiment I: Results
•OSV requires a reanalysis of syntactic structure at subject position
Reading Times Word by Word
600
700
800
900
1000
1100
1200
1300
DP DP Vb Aux
Sentence Element
Tim
e (m
s)
sov
osv
12
Experiment I: SOV vs OSV
• Marked constituents are harder to processed
• Canonical SOV sentences require less processing time
• Derived OSV sentences require syntactic reanalysis
• Displaced constituents increase syntactic complexity
13
Experiment II: Processing of ambiguous chains
Goal: to determine whether there is any preference when processing ambiguous chains (SOV/OSV).
Participants: 23 native speakers
Materials: 3 conditions, 48 sentences per condition
3 lists: 16 SOV sentences
16 OSV sentences
16 AMBIGUOUS
48 fillers
14
MORPHOLOGICAL AMBIGUITY
OBJECT PLURALPACIENT
SUBJECT SINGULAR
AGENT
EMAKUME-AK‘WOMAN-X’
Emakume-ak gizon-ak ikusi ditu woman-? man-? see has
‘The woman has seen the men’ o ‘The man has seen the women’
15
Experiment II: Materials
Emakume-ek gizon-ak ikus-i dituzte
women the Subj. men the Obj. seen have
‘the women have seen the men’
Emakume-ak gizon-ak ikus-i ditu woman-? man-? seen has‘the woman has seen the men’ or ‘the man has seen the women’
Gizon-ak emakume-ek ikus-i dituzte
man-? women-the Subj. seen have
‘the women have seen the men’
Subject
Object
Verb+aux
Subject
Object
Verb+aux
Ambiguous Chain
16
Results: Mean Reading time of the sentences
•Ambiguous chain is processed as SOV sentence
40004200
440046004800
50005200
54005600
Tim
e (
ms
)
SOV AMB OSV
Sentece Type
Reading time of sentences
n.s.p<0.001
17
Results: SOV vs OSV, word by word
•Comparing SOV and OSV orders: replication of the first experiment, syntactic reanalysis
SOV vs OSV. Reading time word by word
700
900
1100
1300
1500
1700
1900
DP1 DP2 Verb Aux
Constituent of the sentence
Tim
e (m
s)
SOV
OSV
18
Results: SOV vs Ambiguous Chain, Word by Word
SOV vs Ambiguous. Reading time, word by word
700
900
1100
1300
1500
1700
1900
DP1 DP2 Verb Aux
Sentence constituent
Tim
e (m
s)
SOV
AMB
•No evidences of syntactic reanalysis•Ambiguous chains are processed as SOV order sentences•SOV is the simplest processing solution.
19
EXPERIMENT III: Electrophysiological evidences using event-related brain potentials (ERPs)
This experiment suggests that:
Participants: 24 right-handed native speakers
a) SOV is the canonical, underived word order in Basque, and it constitutes the simplest choice for sentence parsing
Materials:
4 conditions, 240 sentences per condition
2 conditions, unambiguous
2 conditions, temporally ambiguous
b) OSV word order is syntactically derived and therefore more complex to parse
20
Experiment III: Materials
Subject
Verb+aux
Object
Subject Verb+aux
Object
‘the wolf has eaten the sheep’
otso-ak
wolf-the/Subj
Ardi-a
Sheep-the/Obj
jan
eaten
du
has
‘the wolfs have eaten the sheep(pl)’
Otso-ek
Wolfs/the/Subj
ardi-ak
sheep-the/Obj
jan
eaten
dituzte
have
21
Experiment III: Materials
SOV temporally ambiguous
OSV temporally ambiguous
‘the wolf has eaten the sheep(pl)’
Otso-ak
wolf ?
ardi-ak
sheep ?
‘the wolf has eaten the sheep(pl)’
otso-ak
wolf ?
Ardi-ak
sheep ?
jan
eaten
ditu
has
jan
eaten
ditu
has
22
Experiment III: ERP Results
Constituents of unambiguous SOV/OSV sentences:
Left Anterior Negativity was obtained for displaced subjects and objects
Syntactically displaced constituents increase the processing costs. (Kluender y Kutas, 1997)
23
ERP RESULTS: SOV vs OSV
SOV OSV Left Anterior Negativity
LAN
24
SOV OSV
ERP RESULTS: SOV vs OSV
Left Anterior Negativity
LAN
25
Experiment III: ERP Results
Verb of unambiguous SOV/OSV sentences:
Syntactic integration of displaced constituent at verb position increases processing cost (Felser et al, 2003)
At verb position of OSV non-canonical sentences we obtained a syntax related P600 component.
26
SOV OSV
ERP RESULTS: SOV vs OSV
P600/SPS
27
Experiment III: ERP Results
Unambiguous SOV/OSV sentences:These components, LAN and P600 showed that non-canonical OSV sentences are syntactically more complex than canonical SOV sentences in the grammar of Basque
28
Experiment III: ERP Results
Temporally ambiguous SOV/OSV:
The semantic disambiguation at the verb of OSV elicited a N400 component, related to semantic analysis and complexity
We didn´t find any difference between the two first constituents.
29
SOV OSV
ERP RESULTS: DISAMBIGUATION
N400
30
Basque’s underived word order is SOV
Conclusions
Displaced elements increase syntactic complexity
Non-canonical OSV order is syntactically more complex
31
SummaryWe have provided processing and ERP evidence that gives support to the claim that Basque grammar is built upon a basic, head final sentence structure (SOV word order), despite the fact that this language allows almost all constituent combinations in a sentence. Building upon results from other languages, where it has been shown that underived word orders yield shorter reading times and less comprehension difficulties, we have shown that these results also replicate in Basque, although the specific word order signaled as less complex and easier to understand is systematically SOV.
ERP results also showed that OSV generated LAN and P600, both signaling syntactic complexity