Transcript Interaction between phonology and syntax in Icelandic

Interaction between phonology
and syntax in Icelandic
Arguments for a strongly parellel OT-analysis
A Phonological Workshop
University of Iceland
May 16, 2008
Anton Karl Ingason
University of Iceland
1
Introduction
• Generative grammar
– Generally serial and modular design (rules/transformations
within modules such as syntax and phonology)
– Optimality Theory (OT) (Prince and Smolensky. 1993/2002) is
more parellel, but usually within serially connected modules
(syntax still “happens before” phonology)
– In such a model syntax can affect phonology but phonology can
not affect syntax
– But what if there is evidence of the latter?
• Linguistic behaviour/performance
– Not what “real” linguistics should deal with?
– But what if variation has structure that can be predicted using an
already existing model?
2
Icelandic case variation
• Variation in case marking of subjects of impersonal
verbs, dative substitution (“dative sickness”)
– Native speakers are not consistent in judgements of the
grammaticality of subject case marking for certain verbs
– Effects of post-syntactic phenomena generally not considered
– Syntactic and semantic forces considered entirely responsible
– Which predicts that syntactically and semantically equivalent
subjects should behave identically with respect to case marking
• New hypothesis
– Speakers who show variation in their judgements will not have
syntactic preference for either accusative or dative case
– And will prefer the prosodically better case, ideally the least
marked prosodic word, a trochee (σσ)
3
Experiment
54 speakers judge the grammaticality of the following sentences:
Hvers vegna vantar Jón þessa nagla?
Why needs Jón-ACC-(σ) those nails?
Guðmund vantar nýjan jakka.
Guðmund-ACC-(σσ) needs a new jacket.
Það er ljóst að Jóni vantar betri hugmynd.
It is clear that Jóni-DAT-(σσ) needs a better idea.
Vantar ekki Guðmundi bara stærri jeppa?
Needs not Guðmundi-DAT-(σσσ) just a bigger jeep?
Not syntactic minimal pairs, but since dative sickness is a well known
phenomena it is necessary to disguise the experiment if possible
All comments from participants were about case marking
Current literature on syntax does not predict that the names Jón and
Guðmundur will behave differently with respect to case in those
examples
If prosody affects judgements and (σσ) is the least marked PrWd there
should be a tendency to accept Jóni-DAT-(σσ) rather than Jón-ACC-(σ)
and Guðmund-ACC-(σσ) rather than Guðmundi-DAT-(σσσ)
4
Positive judgements about
grammaticality of accusative subjects
Positive judgements about accusative subjects
number of speakers
35
30
25
20
15
10
5
0
Jón (þf)
Guðmund (þf)
subject
5
Positive judgements about
grammaticality of dative subjects
Positive judgements about dative subjects
number of speakers
45
40
35
30
25
20
15
10
5
0
Jóni (þgf)
Guðmundi (þgf)
subject
6
Modelling variation in OT
• Different dialects and social registers
– Constraint reranking works well and seems logical if
the grammar is indeed different
• Variation within a “fixed” system
– Variation as accessing non-optimal candidates
(Coetzee 2004; 2006)
– Extra information generated by OT used to make
predictions about variation
– Candidates are ordered by relative grammaticality
7
Coetzeean variation model
• Relative grammaticality
Const1
Const2
1 Cand 1
*
2 Cand 2
3 Cand 3
Const3
*
*
• Predictions:
– Better candidates are more frequent
– Possible variation patterns:
• (C1,*C2,*C3), (C1,C2,*C3), (C1,C2,C3)
– Impossible variation patterns:
• (*C1,*C2,C3), (*C1,C2,C3), (*C1,C2,*C3), (C1,*C2,C3)
8
Strongly parellel OT
• Using an entirely parellel grammar, with only one EVAL function syntax
and phonology are processed in the same step (Teeple 2007)
• Syntax can still outrank phonology (Golston 1995) but when two
syntactic options are equally grammatical – phonology can decide
which one is better
• Instead of serially connected modules we have layers of constraints
where any constraint can interact with any other constraint
SYNTAX
1 (σσ)
*/
2 (σσσ)
*/
PHONOLOGY
*
9
Trochee sickness?
•
•
•
•
29 out of 54 participants
were inconsistent in their
judgements about
grammaticality of dative
subjects
All of those had the same
judgement pattern
Current literature on syntax
does not predict one
pattern to be more frequent
than the other
Informal experiment using
syntactic minimal pairs
confirms the result
Explains some of the
results of Kristín Edda
Búadóttir (2007)
S
1
(σσ)
*/
2
(σσσ)
*/
P
Possible variation pattern:
(σσ) / *(σσσ)
*
Impossible variation pattern
*(σσ) / (σσσ)
Dative subjects with the verb vanta 'need'
35
Number of speakers
•
30
25
20
15
10
5
0
Jóni/*Guðmundi
*Jóni/Guðmundi
Judgement patterns
10
Conclusion
• Serial and modular grammar can not account for
phonology affecting syntax
– Strongly Parellel OT can
– Instead of serially connected modules we have layers of
constraints (syntax, phonlogy) where any constraint can interact
with any other constraint
– Syntax can still outrank phonology
• A Coetzeean model of variation can account for patterns
in linguistic performance while still predicting the optimal
“competence output” (1)
• Brings up questions
– Previous research on dative substitution asks questions like “how
common is dative sickness for the verb x?” (Given what subject?)
– Is it a good idea to design grammar as a one directional algorithm
if there is evidence of two way interaction between phonology and
syntax?
– Do similar patterns occur in other cases of variation?
11
References
• Búadóttir, Kristín Edda. 2007. Þágufallshneigð. [Dative substitution.]
Mímisþing. March 17, 2007.
• Coetzee, Andries W. 2004. What it Means to be a Loser. NonOptimal Candidates in Optimality Theory. University of
Massachusetts, Amherst. [ROA #687.]
• Coetzee, Andries W. 2006. Variation as Accessing “Non-Optimal”
Candidates – A Rank-Ordering Model of Eval. Phonology 23:337–
385. [Also ROA #863.]
• Golston, Chris. 1995. Syntax outranks phonology: evidence from
Ancient Greek. Phonology 12:343–368.
• Prince, Alan and Paul Smolensky. 1993/2002. Optimality theory:
Constraint Interaction in generative grammar. Manuscript. Rutgers
University and University of Colorado at Boulder. (Updated version
of the Technical Report from 1993) [ROA #537.]
• Teeple, David. 2007. Prosody Can Outrank Syntax. WCCFL 26,
Berkeley, April 27-29, 2007. University of California, Santa Cruz.
12