Why sentences are more complex than words Jeffrey Heinz 1 William - - PowerPoint PPT Presentation

▶
why sentences are more complex than words
SMART_READER_LITE
LIVE PREVIEW

Why sentences are more complex than words Jeffrey Heinz 1 William - - PowerPoint PPT Presentation

Sep 17, 2023 195 likes •689 views

Phonology Syntax Formal Learning Theories Conclusion Why sentences are more complex than words Jeffrey Heinz 1 William Idsardi 2 1 heinz@udel.edu University of Delaware 2 idsardi@umd.edu University of Maryland Parallel Domains University

slide-1
SLIDE 1

Phonology ∦ Syntax Formal Learning Theories Conclusion

Why sentences are more complex than words

Jeffrey Heinz1 William Idsardi2

1heinz@udel.edu

University of Delaware

2idsardi@umd.edu

University of Maryland

Parallel Domains University of Southern California May 6, 2011

1 / 29

slide-2
SLIDE 2

Phonology ∦ Syntax Formal Learning Theories Conclusion

Is phonology different from syntax?

Jean-Roger Vergnaud

No

2 / 29

slide-3
SLIDE 3

Phonology ∦ Syntax Formal Learning Theories Conclusion

Is phonology different from syntax?

Jean-Roger Vergnaud

No

Morris Halle

Yes (Bromberger and Halle 1989)

2 / 29

slide-4
SLIDE 4

Phonology ∦ Syntax Formal Learning Theories Conclusion

Is phonology different from syntax?

Jean-Roger Vergnaud

No

Morris Halle

Yes (Bromberger and Halle 1989)

2 / 29

slide-5
SLIDE 5

Phonology ∦ Syntax Formal Learning Theories Conclusion

Is phonology different from syntax?

Jean-Roger Vergnaud

No

Morris Halle

Yes (Bromberger and Halle 1989)

Elan Dresher, p.c.

If two things are different, make them similar. If they are similar make them the same.

2 / 29

slide-6
SLIDE 6

Phonology ∦ Syntax Formal Learning Theories Conclusion

This talk

There is an important computational difference between phonology and syntax that requires explanation.

Hypothesis

Humans make different kinds of generalizations over words than they do over sentences and this explains this difference.

Linguistics and Cognitive Science

We suggest this difference can play a key role in larger debates in cognitive science between domain-general and domain-specific learning.

3 / 29

slide-7
SLIDE 7

Phonology ∦ Syntax Formal Learning Theories Conclusion

Phonology ∦ Syntax Formal Learning Theories Conclusion

4 / 29

slide-8
SLIDE 8

Phonology ∦ Syntax Formal Learning Theories Conclusion

Strings

Strings are sequences of more basic units.

Sentences are sequences of morphemes.

John laugh ed while Mary talk ed.

Words are sequences of sounds.

b l i N

5 / 29

slide-9
SLIDE 9

Phonology ∦ Syntax Formal Learning Theories Conclusion

Language Patterns

Language patterns are sets of strings,

  • r relations among strings.

No coda: *Coda

  • {a, ka, ta, pi.koU, ba.du.pi} ⊂ *Coda
  • {bliN, mElp.ka, karp} ∩ *Coda = ∅

6 / 29

slide-10
SLIDE 10

Phonology ∦ Syntax Formal Learning Theories Conclusion

Language Patterns

Language patterns are sets of strings,

  • r relations among strings.

Word final obstruent devoicing: R=[-son]→[-voice]/ #

  • {pad→pat, pat→pat, pabaG→pabax} ⊂ R
  • {pad→pad, pad→dap} ∩ R = ∅

6 / 29

slide-11
SLIDE 11

Phonology ∦ Syntax Formal Learning Theories Conclusion

Language Patterns

Language patterns are sets of strings,

  • r relations among strings.

Conjunction: S → S and S

  • {John swam and Mary laughed, They talked and they

talked and they talked} ⊂ S

  • {John swam and Mary, They talked and they} ∩ S = ∅

6 / 29

slide-12
SLIDE 12

Phonology ∦ Syntax Formal Learning Theories Conclusion

Language Patterns

Language patterns are sets of strings,

  • r relations among strings.

Conjunction: S → S and S

  • {John swam and Mary laughed, They talked and they

talked and they talked} ⊂ S

  • {John swam and Mary, They talked and they} ∩ S = ∅

What kinds of sets and relations are natural language patterns?

6 / 29

slide-13
SLIDE 13

Phonology ∦ Syntax Formal Learning Theories Conclusion

The Chomsky Hierarchy

Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite

7 / 29

slide-14
SLIDE 14

Phonology ∦ Syntax Formal Learning Theories Conclusion

The Chomsky Hierarchy and natural language patterns

Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite Yoruba copying Kobele 2006 Swiss German Shieber 1985 English nested embedding Chomsky 1957 English consonant clusters Clements and Keyser 1983 Kwakiutl stress Bach 1975 Chumash sibilant harmony Applegate 1972

8 / 29

slide-15
SLIDE 15

Phonology ∦ Syntax Formal Learning Theories Conclusion

The Chomsky Hierarchy and natural language patterns

Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite Yoruba copying Kobele 2006 Swiss German Shieber 1985 English nested embedding Chomsky 1957 English consonant clusters Clements and Keyser 1983 Kwakiutl stress Bach 1975 Chumash sibilant harmony Applegate 1972

8 / 29

slide-16
SLIDE 16

Phonology ∦ Syntax Formal Learning Theories Conclusion

The Chomsky Hierarchy and natural language patterns

Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite Yoruba copying Kobele 2006 Swiss German Shieber 1985 English nested embedding Chomsky 1957 English consonant clusters Clements and Keyser 1983 Kwakiutl stress Bach 1975 Chumash sibilant harmony Applegate 1972

8 / 29

slide-17
SLIDE 17

Phonology ∦ Syntax Formal Learning Theories Conclusion

Phonology is regular (Kaplan and Kay 1994) F1 × F2 × . . . × Fn = P

  • 1. Optional, left-to-right, right-to-left, and simultaneous

application of rules A − → B / C D (where A,B,C,D are regular expressions) describe regular relations, provided the rule cannot reapply to the locus of its structural change.

  • 2. Rule ordering is functional composition (finite-state

transducer composition).

  • 3. Regular relations are closed under composition.
  • 4. SPE grammars (finitely many ordered rewrite rules of the

above type) can describe virtually all phonological patterns.

  • 5. Therefore, phonology is regular (both Fi and P).

9 / 29

slide-18
SLIDE 18

Phonology ∦ Syntax Formal Learning Theories Conclusion

Phonology is regular (Kaplan and Kay 1994) F1 × F2 × . . . × Fn = P

  • 1. Optional, left-to-right, right-to-left, and simultaneous

application of rules A − → B / C D (where A,B,C,D are regular expressions) describe regular relations, provided the rule cannot reapply to the locus of its structural change.

  • 2. Rule ordering is functional composition (finite-state

transducer composition).

  • 3. Regular relations are closed under composition.
  • 4. SPE grammars (finitely many ordered rewrite rules of the

above type) can describe virtually all phonological patterns.

  • 5. Therefore, phonology is regular (both Fi and P).

9 / 29

slide-19
SLIDE 19

Phonology ∦ Syntax Formal Learning Theories Conclusion

Phonology is regular (Kaplan and Kay 1994) F1 × F2 × . . . × Fn = P

  • 1. Optional, left-to-right, right-to-left, and simultaneous

application of rules A − → B / C D (where A,B,C,D are regular expressions) describe regular relations, provided the rule cannot reapply to the locus of its structural change.

  • 2. Rule ordering is functional composition (finite-state

transducer composition).

  • 3. Regular relations are closed under composition.
  • 4. SPE grammars (finitely many ordered rewrite rules of the

above type) can describe virtually all phonological patterns.

  • 5. Therefore, phonology is regular (both Fi and P).

9 / 29

slide-20
SLIDE 20

Phonology ∦ Syntax Formal Learning Theories Conclusion

Phonology is regular (Kaplan and Kay 1994) F1 × F2 × . . . × Fn = P

  • 1. Optional, left-to-right, right-to-left, and simultaneous

application of rules A − → B / C D (where A,B,C,D are regular expressions) describe regular relations, provided the rule cannot reapply to the locus of its structural change.

  • 2. Rule ordering is functional composition (finite-state

transducer composition).

  • 3. Regular relations are closed under composition.
  • 4. SPE grammars (finitely many ordered rewrite rules of the

above type) can describe virtually all phonological patterns.

  • 5. Therefore, phonology is regular (both Fi and P).

9 / 29

slide-21
SLIDE 21

Phonology ∦ Syntax Formal Learning Theories Conclusion

Phonology is regular (Kaplan and Kay 1994) F1 × F2 × . . . × Fn = P

  • 1. Optional, left-to-right, right-to-left, and simultaneous

application of rules A − → B / C D (where A,B,C,D are regular expressions) describe regular relations, provided the rule cannot reapply to the locus of its structural change.

  • 2. Rule ordering is functional composition (finite-state

transducer composition).

  • 3. Regular relations are closed under composition.
  • 4. SPE grammars (finitely many ordered rewrite rules of the

above type) can describe virtually all phonological patterns.

  • 5. Therefore, phonology is regular (both Fi and P).

9 / 29

slide-22
SLIDE 22

Phonology ∦ Syntax Formal Learning Theories Conclusion

Phonology is regular (Kaplan and Kay 1994) F1 × F2 × . . . × Fn = P

  • 1. Optional, left-to-right, right-to-left, and simultaneous

application of rules A − → B / C D (where A,B,C,D are regular expressions) describe regular relations, provided the rule cannot reapply to the locus of its structural change.

  • 2. Rule ordering is functional composition (finite-state

transducer composition).

  • 3. Regular relations are closed under composition.
  • 4. SPE grammars (finitely many ordered rewrite rules of the

above type) can describe virtually all phonological patterns.

  • 5. Therefore, phonology is regular (both Fi and P).

9 / 29

slide-23
SLIDE 23

Phonology ∦ Syntax Formal Learning Theories Conclusion

What about reduplication?

  • It’s morpho-syntax (Inkelas and Zoll 2000, Roark and

Sproat 2007).

10 / 29

slide-24
SLIDE 24

Phonology ∦ Syntax Formal Learning Theories Conclusion

Phonology is subregular

Regular Star-Free=NonCounting TSL LTT LT PT SL SP Proper inclusion relationships among subregular language classes (indicated from top to bottom). TSL Tier-based Strictly Local PT Piecewise Testable LTT Locally Threshold Testable SL Strictly Local LT Locally Testable SP Strictly Piecewise (McNaughton and Papert 1971, Simon 1975, Rogers and Pullum in press, Rogers et al. 2010, Heinz 2010, Heinz et al. 2011)

11 / 29

slide-25
SLIDE 25

Phonology ∦ Syntax Formal Learning Theories Conclusion

Phonology is subregular

Regular Star-Free=NonCounting TSL LTT LT PT SL SP Proper inclusion relationships among subregular language classes (indicated from top to bottom). TSL Tier-based Strictly Local PT Piecewise Testable LTT Locally Threshold Testable SL Strictly Local LT Locally Testable SP Strictly Piecewise (McNaughton and Papert 1971, Simon 1975, Rogers and Pullum in press, Rogers et al. 2010, Heinz 2010, Heinz et al. 2011)

11 / 29

slide-26
SLIDE 26

Phonology ∦ Syntax Formal Learning Theories Conclusion

The Chomsky Hierarchy and natural language patterns

Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite Yoruba copying Kobele 2006 Swiss German Shieber 1985 English nested embedding Chomsky 1957 English consonant clusters Clements and Keyser 1983 Kwakiutl stress Bach 1975 Chumash sibilant harmony Applegate 1972

So why the difference?

12 / 29

slide-27
SLIDE 27

Phonology ∦ Syntax Formal Learning Theories Conclusion

The problem of induction and generalization

Philosophy

(Plato, Aristotle, Hume, Mill, Russell, Carnap, Quine, Goodman, . . . )

Linguistics

(Chomsky 1957, 1965, Wexler and Cullicover 1980, Piattelli-Palmarini 1980, Berwick 1985, Morgan 1986, Yang 2000, Niyogi 2006, . . . )

Computer Science

(Gold 1967, Horning 1969, Angluin 1980, Valiant 1984, Osherson et al. 1984, Angluin 1988, Anthony and Biggs 1991, Kearns and Vazirani 1994, Vapnik 1994, 1998, Jain et al. 1999, Chater and Vitany´ ı 2007, de la Higuera 2010, Clark and Lappin 2011)

So how can language patterns be learned?

13 / 29

slide-28
SLIDE 28

Phonology ∦ Syntax Formal Learning Theories Conclusion

Define “Learning”

Learner Experience Languages

Figure: Learners are functions φ from experience to languages.

14 / 29

slide-29
SLIDE 29

Phonology ∦ Syntax Formal Learning Theories Conclusion

Results: Do feasible learners exist?

  • 1. Identification in the limit from positive data (Gold 1967)
  • 2. Identification in the limit from positive and negative data (Gold 1967)
  • 3. Identification in the limit from positive data from r.e. texts (Gold

1967)

  • 4. Learning context-free and r.e. distributions (Horning 1969, Angluin

1988, Chater and Vitany´ ı 2007)

  • 5. Probably Approximately Correct learning (Valiant 1984, Anthony and

Biggs 1991, Kearns and Vazirani 1994)

Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite 15 / 29

slide-30
SLIDE 30

Phonology ∦ Syntax Formal Learning Theories Conclusion

Results: Do feasible learners exist?

  • 1. Identification in the limit from positive data (Gold 1967)
  • 2. Identification in the limit from positive and negative data (Gold 1967)
  • 3. Identification in the limit from positive data from r.e. texts (Gold

1967)

  • 4. Learning context-free and r.e. distributions (Horning 1969, Angluin

1988, Chater and Vitany´ ı 2007)

  • 5. Probably Approximately Correct learning (Valiant 1984, Anthony and

Biggs 1991, Kearns and Vazirani 1994)

Recursively Enumerable

Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite

15 / 29

slide-31
SLIDE 31

Phonology ∦ Syntax Formal Learning Theories Conclusion

Results: Do feasible learners exist?

  • 1. Identification in the limit from positive data (Gold 1967)
  • 2. Identification in the limit from positive and negative data (Gold 1967)
  • 3. Identification in the limit from positive data from r.e. texts (Gold

1967)

  • 4. Learning context-free and r.e. distributions (Horning 1969, Angluin

1988, Chater and Vitany´ ı 2007)

  • 5. Probably Approximately Correct learning (Valiant 1984, Anthony and

Biggs 1991, Kearns and Vazirani 1994)

Recursively Enumerable

Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite

15 / 29

slide-32
SLIDE 32

Phonology ∦ Syntax Formal Learning Theories Conclusion

Results: Do feasible learners exist?

  • 1. Identification in the limit from positive data (Gold 1967)
  • 2. Identification in the limit from positive and negative data (Gold 1967)
  • 3. Identification in the limit from positive data from r.e. texts (Gold

1967)

  • 4. Learning context-free and r.e. distributions (Horning 1969, Angluin

1988, Chater and Vitany´ ı 2007)

  • 5. Probably Approximately Correct learning (Valiant 1984, Anthony and

Biggs 1991, Kearns and Vazirani 1994)

Recursively Enumerable

Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite

15 / 29

slide-33
SLIDE 33

Phonology ∦ Syntax Formal Learning Theories Conclusion

Results: Do feasible learners exist?

  • 1. Identification in the limit from positive data (Gold 1967)
  • 2. Identification in the limit from positive and negative data (Gold 1967)
  • 3. Identification in the limit from positive data from r.e. texts (Gold

1967)

  • 4. Learning context-free and r.e. distributions (Horning 1969, Angluin

1988, Chater and Vitany´ ı 2007)

  • 5. Probably Approximately Correct learning (Valiant 1984, Anthony and

Biggs 1991, Kearns and Vazirani 1994)

Recursively Enumerable

Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite

15 / 29

slide-34
SLIDE 34

Phonology ∦ Syntax Formal Learning Theories Conclusion

Positive Results

Many classes which cross-cut the Chomsky hierarchy and exclude some finite languages are feasibly learnable in the senses discussed (and others).

Recursively Enumerable

Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite

(Angluin 1980, 1982, Garcia et al. 1990, Muggleton 1990, Denis et al. 2002, Fernau 2003, Yokomori 2003, Clark and Thollard 2004, Oates et al. 2006, Niyogi 2006, Clark and Eryaud 2007, Heinz 2008, to appear, Yoshinaka 2008, Case et al. 2009, de la Higuera 2010) 16 / 29

slide-35
SLIDE 35

Phonology ∦ Syntax Formal Learning Theories Conclusion

Lessons from formal learning theories

Learning requires a structured hypothesis space, which excludes at least some finite-list hypotheses. Gleitman 1990, p. 12: ‘The trouble is that an observer who notices everything can learn nothing for there is no end of categories known and constructable to describe a situation [emphasis in original].’

17 / 29

slide-36
SLIDE 36

Phonology ∦ Syntax Formal Learning Theories Conclusion

Lessons from formal learning theories

Learning requires a structured hypothesis space, which excludes at least some finite-list hypotheses. Gleitman 1990, p. 12: ‘The trouble is that an observer who notices everything can learn nothing for there is no end of categories known and constructable to describe a situation [emphasis in original].’

17 / 29

slide-37
SLIDE 37

Phonology ∦ Syntax Formal Learning Theories Conclusion

Hypothesis spaces for language learning

Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite Yoruba copying Kobele 2006 Swiss German Shieber 1985 English nested embedding Chomsky 1957 English consonant clusters Clements and Keyser 1983 Kwakiutl stress Bach 1975 Chumash sibilant harmony Applegate 1972

18 / 29

slide-38
SLIDE 38

Phonology ∦ Syntax Formal Learning Theories Conclusion

Strategy #1: learn everything (e.g. Chater and Vitany´

ı 2007)

Recursively Enumerable

Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite Yoruba copying Kobele 2006 Swiss German Shieber 1985 English nested embedding Chomsky 1957 English consonant clusters Clements and Keyser 1983 Kwakiutl stress Bach 1975 Chumash sibilant harmony Applegate 1972

Problems

  • 1. Possible in principle, not feasible in practice
  • 2. Predicts any pattern is possible with sufficient data

19 / 29

slide-39
SLIDE 39

Phonology ∦ Syntax Formal Learning Theories Conclusion

Strategy #2: Single hypothesis space for language

(e.g. Clark 2010)

Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite Yoruba copying Kobele 2006 Swiss German Shieber 1985 English nested embedding Chomsky 1957 English consonant clusters Clements and Keyser 1983 Kwakiutl stress Bach 1975 Chumash sibilant harmony Applegate 1972

Problems

  • 1. Predicts syntactic patterns ought to be found within words.

20 / 29

slide-40
SLIDE 40

Phonology ∦ Syntax Formal Learning Theories Conclusion

Strategy #3: Distinct hypothesis spaces for phonology and syntax

Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite Yoruba copying Kobele 2006 Swiss German Shieber 1985 English nested embedding Chomsky 1957 English consonant clusters Clements and Keyser 1983 Kwakiutl stress Bach 1975 Chumash sibilant harmony Applegate 1972

Syntax learner Phonology learner

  • 1. The complexity differential between phonology and syntax can be

explained if language-learning itself is modular.

  • 2. People make different kinds of generalizations over words than they do
  • ver sentences.

21 / 29

slide-41
SLIDE 41

Phonology ∦ Syntax Formal Learning Theories Conclusion

Strategy #3 accords with recent research within linguistics

  • Recent computational models for learning phonology are

successful in part because the generalization strategies employed do not consider every finite pattern nor do they extend beyond the regular boundary (Hayes and Wilson 2008, Albright 2009, Heinz 2010, Goldsmith and Riggle to

  • appear. . . ).
  • Likewise, the learners for syntax are successful in part

because the learners’ generalizations are constrained to the right, non-superfinite classes of nonregular patterns (Yang 2000, et seq., Clark and Eryaud 2007, Yoshinaka and Clark 2010, Becerra-Bonache et al. 2010, . . . )

22 / 29

slide-42
SLIDE 42

Phonology ∦ Syntax Formal Learning Theories Conclusion

Advocates of general purpose learners

Recursively Enumerable

Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite Yoruba copying Kobele 2006 Swiss German Shieber 1985 English nested embedding Chomsky 1957 English consonant clusters Clements and Keyser 1983 Kwakiutl stress Bach 1975 Chumash sibilant harmony Applegate 1972 Context- Sensitive Mildly Context- Sensitive Context-Free Regular Finite Yoruba copying Kobele 2006 Swiss German Shieber 1985 English nested embedding Chomsky 1957 English consonant clusters Clements and Keyser 1983 Kwakiutl stress Bach 1975 Chumash sibilant harmony Applegate 1972

Challenges

  • 1. They must present a single learner capable of learning

phonological and syntactic patterns from reasonably-sized sets of words and sentences, respectively (to our knowledge no such demonstration exists).

  • 2. They must also either offer an explanation for the

complexity differential or deny it.

23 / 29

slide-43
SLIDE 43

Phonology ∦ Syntax Formal Learning Theories Conclusion

One possibility: articulatory/perceptual grounding

Hypothesis

Sound sequences within words are constrained by psychophysical properties of the human nervous, motor, and auditory systems in ways that word sequences within sentences are not.

24 / 29

slide-44
SLIDE 44

Phonology ∦ Syntax Formal Learning Theories Conclusion

Long-distance patterns in phonology

Long-distance agreement (Ringen 1988, van der Hulst 1994, Hansson

2001, Rose and Walker 2004)

Samala Chumash (Applegate 1972) StoyonowonowaS ‘3s stood upright’ *stoyonowonowaS *Stoyonowonowas

Long distance disagreement (Suzuki 1998)

Grassman’s Law thr´ ık-s ‘hair’ tr´ ıkh-es ‘hairs’ *thr´ ıkh-es Latin Liquid dissimilation (Jensen 1974, Odden 1994) nav-alis ‘naval’ lun-aris ‘lunar’ flor-alis ‘floral’ *flor-aris

25 / 29

slide-45
SLIDE 45

Phonology ∦ Syntax Formal Learning Theories Conclusion

Is “long-distance” the right generalization?

Perhaps all long distance cases can be reduced to chained instances of strictly local generalizations.

  • 1. Research exists which examines to what extent

intermediary sounds in long-distance assimilation patterns are truly transparent and finds in many instances that the posture of the relevant articulator is maintained throughout pronunciation (Gafos 1996, N´ ı Chios´ ain & Padgett 1997, Gordon 1999, Gafos and Benus 2003, Walker et al. 2009)

  • 2. On the other hand, in Guaran´

ı nasal harmony, research also exists which confirms the oral obstruent realization for voiceless stops that act transparent (Walker 1998).

  • 3. What about the dissimilation cases?

26 / 29

slide-46
SLIDE 46

Phonology ∦ Syntax Formal Learning Theories Conclusion

If they deny the complexity differential. . .

We expect to find synactic patterns in phonology.

  • 1. Nested embedding patterns in phonological words

C V C . C V C C V . C V V . C V C

  • 2. Multiple crossing dependencies in phonological words

C V . C V C V . C V . C V . C V

27 / 29

slide-47
SLIDE 47

Phonology ∦ Syntax Formal Learning Theories Conclusion

Testable Hypothesis

Artificial Language Learning Experiments

Cleeremans & McClelland 1991, Stadler & Frensch 1998, Dell et al. 2000, Gom´ ez 2002, Onishi, Chambers, & Fisher 2002, Chambers, Onishi, & Fisher 2003, Pycha, Nowak, Shin & Shosted 2003, Wilson 2003, 2006, Fitch and Hauser 2004, Goldrick 2004, Newport & Aslin 2004, Petersson et al. 2004, Onnis, Monaghan, Richmond, & Chater 2005, Perruchet & Rey 2005, Bahlmann & Friederici 2006, De Vries et al. 2006 Peperkamp, Skoruppa & Dupoux 2006, Friederici, Bahlmann, Heim, Schubotz, & Anwander 2006, Finley 2008, submitted, in revision, Finley & Badecker 2008, 2009, Folia et

  • al. 2008, Forkstam, Elwer, Ingvar, & Petersson 2008, Moreton 2008, Seidl,

Cristi` a, Bernard, & Onishi 2009, Udd´ en, Araujo, Forkstam, Ingvar, Hagoort, & Petersson 2009, Koo & Callahan submitted, Moreton and Pater, MS, . . .

28 / 29

slide-48
SLIDE 48

Phonology ∦ Syntax Formal Learning Theories Conclusion

Conclusion

There are substantial similarities between phonology and syntax.

  • 1. Both are generative.
  • 2. Both are richly structured domains which subsequently

limit the cross-linguistic variation.

But there is a significant difference.

  • 1. Phonological patterns can be described with regular

grammars, but syntactic patterns cannot.

  • 2. The hypothesis that language-learning itself is modularized

currently offers the best explanation for this fact.

29 / 29

slide-49
SLIDE 49

Phonology ∦ Syntax Formal Learning Theories Conclusion

Conclusion

There are substantial similarities between phonology and syntax.

  • 1. Both are generative.
  • 2. Both are richly structured domains which subsequently

limit the cross-linguistic variation.

But there is a significant difference.

  • 1. Phonological patterns can be described with regular

grammars, but syntactic patterns cannot.

  • 2. The hypothesis that language-learning itself is modularized

currently offers the best explanation for this fact.

Thank You.

29 / 29