Novel phonotactic learning: Syllable-level and co-occurrence representations?, Amélie Bernard

Novel phonotactic learning: Syllable-level and co-occurrence

representations?

Amélie Bernard

Phonotactics

dahdap

2

✗

✓

in English...

Phonotactic learning

•  starts early

•  remains flexible

•  affects speech processing

e.g., PLok vs PTak in label learning (e.g., MacKenzie et al., 2012)

3

How are phonotactics represented?

dap vs dah

syllable-level

P often ends syllables H does not

A & P often co-occur A & H do not

co-occurrences

4

nAPAPpleadAPtcAPableimPAct

naPcuPsleePnaPkinrecePtion

How are phonotactics represented?

dap vs dah

syllable-level

P often ends syllables H does not

syllable

coda a pd

co-occurrences

a pd

5

A & P often co-occur A & H do not

How are phonotactics represented?

✓ ouTPut ✓

(✗) PermiT ✓ (✓) sePTic ✗

c v C.C v c

baT.Pev

6

syllable

onset P e v

syllable

coda b a T .

co-occurrences

syllable-level

cluster

P e vb a T

7

Support for syllable-level representation from artificial languages

onset coda

syllable syllable

(Bernard & Onishi, 2013)

b a F . P e v

t i D . Z e k

t u F . Z e b

t u Z . F e bvs.

train:

test:

coda coda

syllable syllable

(e.g., Dell et al., 2000)

some consonants are better syllable codas than others

from natural language

f a N G f a H

k e N G

k e Hvs.

8

Support for co-occurence representation

(e.g., Richtsmeier et al., 2009)

within words, some consonant co-occurrences are more

frequent than others

t u F P e b

t u F Z e bvs.

test:

(Bernard & Onishi, 2013)

from artificial languages from natural language

b o P K e mvs.

co-occur

b a F P e v

t i D Z e k

train:

__KT__ __PK__

b o K T e m

syllable-level co-occurrences syllable

onset

syllable

coda b a F . P e v

9

P is an onset

✗

✓ P _ _ _ _ P

✓

✗ _ _ FP _ _ _ _ FD_ _

cluster

b a F P e v

(e.g., Bernard, 2015)

F and P co-occur

syllable-level co-occurrences

10

adults ✓ adults ✓

syllable

onset

syllable

coda b a F . P e v

cluster

b a F P e v

EXPERIMENT 1

Can 5-year-old children track co-occurrence regularities?

11

12

Key Manipulation

•  Train with medial consonants: •  restricted syllable positions

onsets codas unrestricted

P , Z F , D b , k , t , v

CVCsyllable

onset (start)

coda (end)

13

•  Train with medial consonants: •  restricted syllable positions

•  restricted cluster pairings

Key Manipulation

clusters

FP , DZ

onsets codas unrestricted

P , Z F , D b , k , t , v

14

Key Manipulation

•  Train with medial consonants: •  restricted syllable positions

•  restricted cluster pairings

baF.Pev tiD.Zek

15

Key Manipulation

•  Test in medial positions comparing: •  same clusters

•  different clusters

•  Train with medial consonants: •  restricted syllable positions

•  restricted cluster pairings

baF.Pev tiD.Zek

vuF.Pet teD.Zev

vuF.Zet teD.Pev

baF.Pevvib.ket

kut.veb tiD.ZekbaF.Pev

…vib.kettiD.Zek

TRAINING (presented 3x)

48 training items 24 filler items

TEST repeating:

8 training items 8 filler items novel: 8 same items 8 different items

16

experimenter

child

YES , we heard it when playing

baF.Pevtrial 1

YES , we heard it when playing

vuF.Pet

trial 2

YES , we heard it when playing

vuF.Zet

trial 3

YES , we heard it when playing

vib.ket

trial 4

FALSE

recognition

0.00

0.25

0.50

0.75

1.00

Prop

ortio

n Ye

s Res

pons

es

.

correct

recognition FALSE

recognition correct

recognition FALSE

recognition

n = 21

*

*

n = 24

5-year-olds: all items

17

Experiment 1 Ex 2.2- manipulating syllable-position

(new word position)

repeating novel repeating novel

(preliminary data)

0.00

0.25

0.50

0.75

1.00

Prop

ortio

n Ye

s Res

pons

es

.

n.s.

different cluster

vuF.Zet

same cluster

vuF.Petn = 21

5-year-olds: novel items

*

n = 24

different Fet.vuZ

same

Zet.vuF

18

Ex 2.2- manipulating syllable-position

(new word position)

baF.Pev tiD.Zek

Experiment 1

(preliminary data)

syllable-level syllable

onset

syllable

coda b a F . P e v

co-occurrences

cluster

b a F P e v

19

adults ✓ adults ✓ 5yos (✗) Ex 1

EXPERIMENT 2

Can 5-year-old children track syllable-position regularities?

20

21

Key Manipulation

•  Test in EDGE positions comparing: •  same syllable positions

•  different syllable positions

•  Train with medial consonants: •  restricted syllable positions

•  restricted cluster pairings

baF.Pev tiD.Zek

Zet.vuF Pev.teDFet.vuZ Dev.teP

0.00

0.25

0.50

0.75

1.00

Prop

ortio

n Ye

s Res

pons

es

.

correct

recognition FALSE

recognition correct

recognition FALSE

recognition

n = 21

*

*

n = 24

5-year-olds: all items

22

Experiment 2

repeating novel repeating novel

Experiment 1

(preliminary data)

0.00

0.25

0.50

0.75

1.00

Prop

ortio

n Ye

s Res

pons

es

.

n.s.

n = 21

5-year-olds: novel items baF.Pev tiD.Zek

*

n = 24

23

Experiment 1 Experiment 2

different cluster vuF.Zet

same cluster

vuF.Pet

different syllable-position Fet.vuZ

same syllable- position Zet.vuF

(preliminary data)

syllable-level syllable

onset

syllable

coda b a F . P e v

co-occurrences

cluster

b a F P e v

24

task sensitivity? some orthography?

adults ✓ 5yos ✓ Ex 2

adults ✓ 5yos (✗) Ex 1

EXPERIMENT 3

Can 11-month-old infants track co-occurrence regularities?

25

baF.Pevvib.ket

kut.veb tiD.ZekbaF.Pev

…vib.kettiD.Zek

TRAINING (presented 3x)

48 training items 24 filler items

26

kiF.PebvaD.Zek…

parent

baby in saucer

parent

baby

(Head Turn Preference Paradigm; e.g., Kemler Nelson et al., 1995)

TEST (novel)

3 same trials 3 different trials

27

Key Manipulation

•  Test in medial positions comparing: •  same clusters

•  different clusters

•  Train with medial consonants: •  restricted syllable positions

•  restricted cluster pairings

baF.Pev tiD.Zek

vuF.Pet teD.Zev

vuF.Zet teD.Pev

0.0

2.0

4.0

6.0

8.0

10.0

Orie

ntat

ion

time

(sec

)

.

n.s.

n =23

*

n =24

different Fet.vuZ

same Zet.vuF

11-month-olds: novel items baF.Pev tiD.Zek

28

Ex 2.4- manipulating syllable-position

(new word position)

Experiment 3

different cluster vuF.Zet

same cluster

vuF.Pet

(preliminary data)

EXPERIMENT 3

Can 11-month-old infants track co-occurrence regularities?

29

30

Key Manipulation Experiment 4

•  Test in EDGE positions comparing: •  same syllable positions

•  different syllable positions

•  Train with medial consonants: •  restricted syllable positions

•  restricted cluster pairings

baF.Pev tiD.Zek

Zet.vuF Pev.teDFet.vuZ Dev.teP

0.0

2.0

4.0

6.0

8.0

10.0

Orie

ntat

ion

time

(sec

)

.

n.s.

n =23

*

n =24

11-month-olds: novel items

31

baF.Pev tiD.Zek

Experiment 3 Experiment 4

different cluster vuF.Zet

same cluster

vuF.Pet

different syllable-position Fet.vuZ

same syllable- position Zet.vuF

(preliminary data)

32

co-occurrences

cluster

b a F P e v

syllable-level syllable

onset

syllable

coda b a F . P e v

adults ✓ 5yos ✓ Ex 2

11mos ✓ Ex 4

adults ✓ 5yos (✗) Ex 1

11mos (✗) Ex 3

Summary Syllable-level regularities •  learned with limited, or no orthography •  learned from variable context (33% filler items)

•  spontaneously transferred to new word positions

Co-occurrence regularities •  no evidence for tracking by children/infants

33

Word-recognition task •  can be used with 5-year-olds •  results converge with headturn preference results •  permits multiple comparisons

•  repeating vs novel •  same vs different

Summary

34

How are phonotactics represented?

dap vs dah

P often ends syllables H does not syllable

coda c u P

syllable

coda

syllable

r e c e P t i o n

syllable

n a P k i n

syllable

coda

syllable

syllable

coda s l e e P

35

Thanks to

Research supported by

Participating adults, parents, and children

Kris Onishi

Ramsey Cardwell Julie Chobert

Melissa Ferland Cindy Fisher

AJ Orena Esther Schott

to AB, to MF, and to KO to AB to ES

Richard H. Tomlinson Fellowship

to KO