language acquisition and cross-language variation colin phillips cognitive neuroscience of language...

Language Acquisition and Cross-Language Variation

Colin PhillipsCognitive Neuroscience of Language Laboratory

Department of LinguisticsUniversity of Maryland

Overview of Talks

1. The Unification Problem

2. Building Syntactic Relations

3. Abstraction: Sounds to Symbols

4. Linguistics and Learning

In-situ

1 2 3 4 5 6 7 8

Region

Reading Time

どの生徒に…

Outline

• Background

• Constraints on pronoun interpretation

• Argument structure

• Aspectual interpretation

• Verbal morphosyntax

• Conclusions

The Big Idea (not mine)

• Many properties of language seem hard-to-observe, hence hard-to-learn

• Typology may inform learning…i. Universals need not be learnedii. Parametric clusters: hard-to-observe properties can be linked to easy-to-observe properties

• Goal is to use typology to drive a deductive learning theory which requires simple choices

• “…we no longer consider UG as providing a format for rule systems and an evaluation metric. Rather, UG consists of various subsystems of principles […] Many of these principles are associated with parameters that must be fixed by experience. The parameters must have the property that they can be fixed by quite simple evidence, because this is what is available to the child.”

(Chomsky, 1986: 146)

• This is a new twist on Jakobson’s proposal in Kindersprache, Aphasie, und allgemeine Lautgesetze (1941) that language development should track cross-linguistic patterns of markedness

(Tomasello, 2000, Cognition)

Predictions

• Universal constraints respected early• Minimally different non-universal constraints

appear at a later age• Areas of parametric consistency also early (if

linked to an easy-to-learn property)• No violation of universals• Language-specific knowledge (relatively) delayed

Notice that…

• Relation between theories of adult language and

– Development - widely accepted

– Real-time Computation - widely rejected

Outline

• Background

• Conclusions

Children Distinguish Universal and Language Particular Constraints on Coreference

Nina KazaninaColin Phillips

A Constraint on Interpretation

• When can a pronoun and a name refer to the same person?

i.e. when can they corefer

a. While John was reading the book, he ate an apple

b. While he was reading the book, John ate an apple

c. John ate an apple while he was reading the book

d. *He ate an apple while John was reading the book

Japanese

[Pooh-ga hon-o yonde-iru aida] (kare-wa) ringo-o tabeta.

[(Kare-ga) hon-o yonde-iru aida] Pooh-wa ringo-o tabeta.

Pooh-wa [(kare-ga) hon-o yonde-iru aida] ringo-o tabeta.

*Kare-wa [Pooh-ga hon-o yonde-iru aida] ringo-o tabeta

ate the apple

whileS

was reading the book

While John was reading the book, he ate the apple

ate the apple

whileS

was reading the book

While he was reading the book, John ate the apple

A Constraint on InterpretationS

ate the apple

S’VP

whileS

was reading the bookJohn ate the apple while he was reading the book

A Constraint on InterpretationS

ate the apple

S’VP

whileS

was reading the bookHe ate the apple while John was reading the book

Principle C (informal)

• A name cannot be c-commanded by a pronoun that co-refers with it

Principle C in Other Languages

a. While he was reading the book, John ate an appleb. *He ate an apple while John was reading the book

• French, Italian, German, Greek, Amharic, Hindi, Hebrew, Spanish, etc.

• Mohawk

MohawkNative American language, Quebec & upstate New York

• Free Word OrderSak ra-núhwe’-s ako-[a]tyá’tawiSak MsS-like-hab FsP-dress‘Sak likes her dress.’

• Ra-núhwe’-s Sak ako-[a]tyá’tawi• Sak ako-[a]tyá’tawi ra-núhwe’-s• Ra-núhwe’-s ako-[a]tyá’tawi Sak • Ako-[a]tyá’tawi ra-núhwe’-s Sak• Ako-[a]tyá’tawi Sak ra-núhwe’-s

• Omission of arguments

Ra-núhwe’-sMsS-like-hab‘He likes it.’

• Discontinuous constituents

Ne kíke wa-hi-yéna-‘ ne kwéskwesne this fact-1sS/MsO-catch-punc ne pig‘I caught this pig.’

Condition C Effects

• Wa-ho-nakuni-‘ tsi Sak wa-hi-hrewaht-e’fact-NsS/MsO-anger-punc that Sak fact-1sS/MsO-punish-punc‘That I punished Saki made himi mad.’ (coreference possible)

• Wa-shako-hrori-‘ tsi Sak wa-hi-hrewaht-e’fact-MsS/FsO-tell-punc that Sak fact-1sS/MsO-punish-punc‘Hei told her that I punished Saki.’ (coreference impossible)

Language Acquisition

• How could a child ever learn that Principle C applies?

• Particularly in a language like Mohawk, where its effects are rather obscure

• Why does Principle C apply in every language?

• Universal Principles may not need to be learned - they may be part of the child’s innate knowledge of language

• This would explain why the principle is universal

• It would also set aside the language acquisition problem

• Predicts that young children should know constraints like Principle C

a. While he was in the box, the smurf ate a hamburger

d. *He ate a hamburger while the smurf was in the box

(Crain & McKee, 1985)

English

(1) While Poohi was reading a book, hei ate the apple.

(2) Poohi ate the apple while hei was reading a book.

(3) * Hei ate the apple while Poohi was reading the book. Pr. C sentence

(4) While hei was reading a book, Poohi ate the apple. while-

sentence

English

(1) While Poohi was reading a book, hei ate the apple.

(2) Poohi ate the apple while hei was reading a book.

(3) * Hei ate the apple while Poohi was reading the book. Pr. C sentence

(4) While hei was reading a book, Poohi ate the apple. while-sentence

Russian

(1R) Poka Poohi chital knigu, oni s'el yabloko. while Pooh was reading the book he ate the apple

(2R) Poohi s'el yabloko, poka oni chital knigu. Pooh ate the apple while he was reading the book

(3R) * Oni s'el yabloko, poka Poohi chital knigu. Pr. C sentence he ate the apple while Pooh was reading the book

(4R) * Poka oni chital knigu, Poohi s'el yabloko. poka-sentence

while he was reading the book Pooh ate the apple

Russian

• Backwards Anaphora is allowed in adult Russian: (5) Rasskaz, kotoryy onai prochitala, rasstroil devochkui.

The story which shei read upset the girli.

• No c-command between the pronoun and r-expression in poka-sentences

The poka-constraint is independent of Principle C

• The poka-constraint is a discourse-level constraint; applies to sequences of (agent-) subjects in Russian

• Language-specific, but minimally different from Principle C

English Russian

While Poohi was reading a book, hei ate the apple. Poohi ate the apple while hei was reading a book.

Hei ate the apple while Poohi was reading the book. * *Pr. C sent Pr. C

While hei was reading a book, Poohi ate the apple. * while-sent poka-

– English-speaking children know Principle C at 3;0

– English-speaking children allow Backwards Anaphora in while-sentences

(Crain & McKee 1985)

Methods and Design

• 50 Russian speaking monolingual children aged 2;8 - 4;11

• Truth Value Judgment Task

• 2x2 between-subject design

• 4 experimental stories per child

• Filler story after each tested story

• Child hears stories in which the coreference interpretation is TRUE

• Child then judges the truth of a sentence which is TRUE under the coreference interpretation

• If the child says that the sentence is FALSE, the relevant interpretation must be blocked by a linguistic constraint.

Truth Value Judgment Task

“I know what happened in this story…”

“Hello, Eeyore! I see that you’re reading a book.”

“What a fine-looking apple.”

“No, Pooh. You can’t eat the apple - that’s my apple.”

“Ok, I’ll have to eat a banana instead.”

“Ok, Pooh. I’ve finished reading. Now you can read the book.”

“Great. Now that Pooh is reading the book, I can eat this delicious apple.”

“I shouldn’t be such a greedy donkey - I should let Pooh eat the apple.”

“I suppose I have to eat a banana instead.”

“Here you are, Pooh. You can have the apple.”

“Oh, I’m such a lucky bear! I can read the book, and I can eat the apple, at the same time.”

Apple is eaten up.

OK, that was a story about Eeyore and Winnie-the-Pooh. First Eeyore was reading the book and then Winnie-the-Pooh was reading the book. I know one thing that happened...

While Pooh was reading the book, he ate the apple.

While he was reading the book, Pooh ate the apple.

Pooh ate the apple while he was reading the book.

He ate the apple while Pooh was reading the book.

3-4 Year Old English Speakers

a. While Pooh was reading the book, he ate an apple

b. While he was reading the book, Pooh ate an apple

c. Pooh ate an apple while he was reading the book

d. *He ate an apple while Pooh was reading the book

How the Task Works

• Child is not being judged

• Identical story for all test sentences

• Avoids child’s ‘yes’ bias

• Story favors the ungrammatical meaning

• Plausible denial

Plausible Denial

TRUE - but ungrammatical

Grammatical - but FALSE

clearly FALSE, since it almost happened, but then didn’t

Eeyore

5-6 Year Old Russian Speakers

3 Year Old Russian Speakers

0102030405060708090

% Rejection

Pr. C poka Control1

Control2

Russian, n=50

Overall Results

Russian Judgments: Breakdown by Age

< 3 years 3-5 years > 5 years Adults

Principle C

Poka/whilesentence

3-Year Olds

English: Crain &McKee, ave. age 3;1Russian: this study, ave. age 3;2

Rejection in Russian vs. English kids

Rus: N=39, mean age = 4;2 Eng: N=62, mean age = 4;2 (Crain&McKee 1985)

Principle C Poka/while sentCondition

Russian children

English children

Interim Conclusion

• 3-year old Russian children clearly distinguish two constraints on backwards anaphora which have very similar surface properties

• At age 3, English-speaking and Russian-speaking children show almost identical judgments - they respect the universal constraint

• Important further questions:(i) Why do Russian and English differ?(ii) How do Russian children become adult-like?

Outline

• Background

• Conclusions

Cross-language Variation in Syntax-Semantics Linking Rules

Meesook KimColin PhillipsBeth Rabbin

Barbara Landau

Learning Verb Syntax

• “Locative Verbs”

• Verbs which refer to an action in which a substance moves to a particular location

• pour, spill, stuff, pile, fill, load, cover, decorate, spray, bandage, soak, sprinkle, spread, etc.

• Similar verbs and similar constructions found in very many languages

Learning Verb Syntax

• “Locative Verbs”

• Sally poured the water into the glass.Sally poured the glass with water.

• Sally filled the water into the glass.Sally filled the glass with water.

• Sally loaded the boxes into the truck.Sally loaded the truck with boxes.

Japanese

John-ga gurasu-ni mizu-o sosoi-da figure-frame

John-ga mizu-de gurasu-o sosoi-da ground-frame

John-ga ki-ni raito-o kazatta figure frame

John-ga raito-de ki-o kazatta ground frame

John-ga kabe-ni penki-o nutta figure-frame

John-ga penki-de kabe-o nutta. ground Frame

‘Overgeneralization’

• Well-known errors with locative verbs (Bowerman 1982)

I didn't fill water up to drink it; I filled it up for the flowers to drink it.Can I fill some salt in the bear? [= a bear-shaped salt shaker]I'm going to cover a screen over me.

(see also experimental data in Gropen et al. 1991a, b)

• Why do children make these errors?

Terminology

Sally poured the water into the glass

Terminology

moving objectFIGURE

Terminology

moving objectFIGURE

locationGROUND

Terminology

Sally filled the glass with the water

moving objectFIGURE

locationGROUND

Terminology

moving objectFIGURE

locationGROUND

moving objectFIGURE

Terminology

moving objectFIGURE

locationGROUND

moving objectFIGURE

Figure Frame

Ground Frame

Classes of Verbs

• Verbs with syntax like pour– dribble, drip, spill, shake, spin, spew, slop, etc.

• Verbs with syntax like fill– cover, decorate, bandage, blanket, soak,

drench, adorn, etc.

• Verbs with syntax like load– stuff, cram, jam, spray, sow, heap, spread, rub,

dab, plaster, etc.

Classes of Verbs

drench, adorn, etc.

dab, plaster, etc.

manner-of-motion

Classes of Verbs

drench, adorn, etc.

dab, plaster, etc.

manner-of-motion

change-of-state

Classes of Verbs

drench, adorn, etc.

dab, plaster, etc.

manner-of-motion

change-of-state

manner-of-motion & change-of-state

Learning Syntax from Semantics

Manner-of-motion

V NP PPfigure ground

V NP PPfigureground

Change-of-state

FigureFrame

GroundFrame

Linking RulesSEMANTICS SYNTAX

Learning

• Linking Rules can be used to ‘bootstrap’ verb syntax or verb meanings, provided that Syntax-Semantics Linking Rules are– consistent across languages (i.e. verbs with

same meaning should have same syntax across all languages)

– innate (i.e. children know the connections from the outset)

• Assumption: linking generalizations are universal

• Shared by opposing accounts of learning verb syntax & semantics

But Languages Vary

• English*John decorated the flowers in the room. John decorated the room with flowers.

But Languages Vary

Change-of-state--> Ground Frame

But Languages Vary

• KoreanYumi-ka ccoch-ul pang-ey cangsikha-yess-ta Nom flowers-Acc room-Loc decorate-Past-Dec‘John decorated the flowers in the room.’Yumi-ka pang-ul ccoch-ulo cangsikha-yess-ta Nom room-Acc flowers-with decorate-Past-Dec‘John decorated the room with flowers.’

But Languages Vary

Korean is more liberal than English

• English John piled the books on the shelf. John piled the shelf with books.

But Languages Vary

• Korean Yumi-ka chaek-lul chaeksang-ey ssa-ass-ta. Nom book-Acc table-Loc pile-Past-Dec ‘Yumi piled books on the table.’

But Languages Vary

• Korean Yumi-ka chaek-lul chaeksang-ey ssa-ass-ta. Nom book-Acc table-Loc pile-Past-Dec ‘Yumi piled books on the table.’*Yumi-ka chaeksang-lul chaek-elo ssa-ass-ta.

Nom table-Acc books-with pile-Past-Dec ‘Yumi piled the table with books.’

But Languages Vary

• Korean Yumi-ka chaek-lul chaeksang-ey ssa-ass-ta. Nom book-Acc table-Loc pile-Past-Dec ‘Yumi piled books on the table.’*Yumi-ka chaeksang-lul chaek-elo ssa-ass-ta.

Nom table-Acc books-with pile-Past-Dec ‘Yumi piled the table with books.’

But Languages Vary

Korean is more restrictive than English

A Problem for Learners?

• If syntax-semantics Linking Rules are not uniform across languages, then how can they help learners?

• If each language had different Linking Rules, would this be any use to a child?

Cross-Language Survey

• Survey I

English TurkishKorean LugandaFrench HindiJapanese HebrewChinese Malay Thai Arabic

• Survey II

ItalianYorubaPolishEweJapanese RussianFrench EnglishBrazilian Portuguese Spanish (Argentinian) Spanish (Castilian)

• Survey I

• Survey II

ItalianYorubaPolishEweJapanese RussianFrench EnglishBrazilian Portuguese Spanish (Argentinian) Spanish (Castilian)Less detailed

classification used(~15 verbs)

• Survey I

• Survey II

ItalianYorubaPolishEweJapanese RussianFrench EnglishBrazilian Portuguese Spanish (Argentinian) Spanish (Castilian)More detailed

classification used(~30 verbs)

Consistent Properties...

A Universal

• EnglishJohn poured the water into the glass.*John poured the glass with water.

• SpanishJuan vertí agua en el vaso.John poured water into the glass*Juan vertí el vaso con agua.John poured the glass with water

A Universal

• HebrewDanny shafax mayim letox ha-kos.John poured water into the glass‘John poured water into the glass.’*Danny shafax et ha-kos be-mayin.John poured Acc the glass with water‘*John poured the glass with water.’

A Universal

• JapaneseTaro-ga mizu-o baketu-ni sosoi-da. Nom water-Acc bucket-Loc pour-Past‘Taro poured water into a bucket.’*Taro-ga baketu-o mizu-de sosoi-da. Nom bucket-Acc water-with pour-Past‘*Taro poured a bucket with water.’

A Universal

Manner-of-motion

FigureFrame

SEMANTICS SYNTAX

Consistent Variation...

A Two-way Split

English

*He decorated lights on the tree

He decorated the tree with lights

French

Spanish

Arabic

Hebrew

Korean

He decorated lights on the tree

Chinese

Japanese

Turkish

Luganda

‘Serial Verbs’ (Verb Compounds)

• JapaneseJohn-ga Bill-o osi-taosi-ta. Nom Acc push-topple-Past‘John pushed Bill down.’

• Igbo (W. Africa)Adha si-ri anu ri-eAda cook-asp meat eat-asp‘Ada cooked the meat and ate it.’ (Igbo)

Easy to observe!

A Parameter

English

French

Spanish

Arabic

Hebrew

Korean

Chinese

Japanese

Turkish

Luganda

A Parameter

English

French

Spanish

Arabic

Hebrew

Korean

Chinese

Japanese

Turkish

Luganda Allow Serial Verbs

A Parameter

English

French

Spanish

Arabic

Hebrew

Korean

Chinese

Japanese

Turkish

Luganda Allow Serial Verbs

Don’t Allow Serial Verbs

A ParameterVP

V NP PPfigureground

Change-of-state

FigureFrame

GroundFrame

SEMANTICS SYNTAX

A ParameterVP

V NP PPfigureground

Change-of-state

FigureFrame

GroundFrame

SEMANTICS SYNTAX

SerialVerbs?

Quantitatively...

• Sample of ~2000 judgments in 20 languages• A small number of principles & parameters allows

us to predict ~90% of judgments• In some classes accuracy is much higher: basic

Figure class, Ground class, etc.• In some classes accuracy is somewhat lower at

present: Ground Alternator, ‘Pure’ Alternator

‘Overgeneralization’

• Well-known errors with locative verbs (Bowerman 1982)

I didn't fill water up to drink it; I filled it up for the flowers to drink it.Can I fill some salt in the bear? [= a bear-shaped salt shaker]I'm going to cover a screen over me.

(see also experimental data in Gropen et al. 1991a, b)

• Why do children make these errors?

Elicited Production Study

• Adult and child (age 3-4) speakers of English & Korean (10 in each group)

• Describing 30 videotaped scenes - 14 verbs(video clips preceded by a contrasting scene, to encourage production of full V NP PP structure)

• Are same errors found as in spontaneous speech?• If so, do we find evidence for mis-set parameter?

Elicited Production Study

• 300 locative structures elicited from each group• Adult speech is fully grammatical

Restricted Errors

• Errors with fill were extremely common; few otherwise

0102030405060708090

Fill (bypouring)

Fill (byloading)

Cover Decorate

% figure frames

Ground Verbs (children)

0102030405060708090

% Figure Frames

English Korean

• On other change-of-state verbs, English & Korean children showed very different production

• Both groups know native language syntax for these verbs

Successes & Failures

• No evidence that errors due to mis-set parameter• Korean 2-year olds use serial verb constructions in

spontaneous speech (Kim & Phillips, 1998); could support early knowledge of change-of-state verbs

• Prevalence of Fill errors remains puzzling

Interim Conclusions• Knowing the meaning of a verb does not predict

the verb’s syntax, BUT…

• Knowing the meaning of a verb, together with further syntactic knowledge about the language, does predict the verb’s syntax rather well

• Typological research contribute to explanation of (i) how linking rules are available in principle, and (ii) how children succeed in practice

• Many questions remain unanswered...

Outline

• Background

• Conclusions

Developing Understanding of Events and Aspect

Nina KazaninaColin Phillips

Completion Entailments

Simple Past

John-ga ie-o tate-ta

Past Progressive

John-ga ie-o tate-tei-ta

Frame of Reference + past/progressive

Mary-ga NY-ni ryokoo-si-teiru aida, John-ga ie-o tateteita / tateta

Imperfective Paradox

(1) a. Mary built a house.b. Mary was building a house.

How come we can say (1b) when no house gets built? Is build a house about building walls?

(2) a. Mary drove from DC to Boston.b. Mary was driving from DC to Boston.

Is drive to Boston about getting to NYC?

What does the IMP/PROG denote?

a. Mary was drawing an arc.

a. Mary was drawing an arc. b. Mary was drawing a circle.

c. Mary was drawing a face. d. Mary was drawing a bike.

• “Present activities are the whole story”• Allows both complete & incomplete events in the

denotation of the verb: "a verb such as 'cross' is true of all crossings independently of whether they culminate."

• An eventuality may – culminate

Cul(e,t) - e is an event that culminates at time t

– hold for a whileHold(e,t) - e is an event which is in progress (in its developmental portion) at t

Parsons (1989)

• An incomplete event in the actual world w is related to a

complete version of the same event in a certain possible world

w’ (Dowty 1979, Landman 1991, Portner 1998 among others)

Dowty (1979): [PROG ] is true at I and w iff there is an interval I’ such

that I’ I [and I is not a final subinterval of I’] and there is an inertia world

w’ for which is true at I’ and w’, and w is exactly w’ at all times preceding

and [including] I

BostonDC

IMP/PROG

Actual world w Possible world w’

Dowty-Landman Approach

Research Question:

• Do children know how to deal with the

IMP Paradox?

Russian: Perfective vs. Imperfective

• Rus Perfective Eng simple past– refers to completed events

postroil dom ‘built a house’ sobral kartinku ‘do a puzzle’

• Rus Imperfective Eng past progressive– can refer to completed or incomplete events– used to describe ongoing events (past, present or

future)

stroil dom ‘was building a house’ sobiral kartinku ‘was doing a puzzle’

Previous Research

Previous findings suggest early mastery of aspect

• Spontaneous Speech:

Children produce both aspectual forms from a very

young age (< 2 years) (Brun et al., 1999; Gvozdev,

1961; Bar-Shalom&Snyder 2000)

Previous Research• Picture-matching task (Vinnitskaya&Wexler, 2001)

Mal’chik chitalI knigu. Mal’chik prochitalP knigu.The boy was reading the book. The boy read the book.

3-4 year olds appear to use IMP vs. PERF to correctly distinguish ongoing from completed events

Our Experiments

• Do Russian children appropriately make use of aspectual morphology to distinguish completed from incomplete events?

• Tested verbs were Creation verbs(enable a clear-cut difference between complete and incomplete events)

• 4 stories per child, 44 trials total

• Within-subject design

• 11 Russian monolingual children, aged 3-5, tested in Moscow preschools

Creation Experiment

Creation Expt: Design

• In each story, an event occurs at 3 landmarks:a flower-bed, a castle and a tree

• In each story, an event occurs

(i) completely(ii) incompletely randomized order(iii) not at all

• Children were asked where an event happened, using PERF and IMP verbs; encouraged to give more than one location as answer

• Monkey assemble a smurf obez’yanka sobrala/sobirala gnomika

• Lion build a house l’venok postroil/stroil domik

• Tiger make a puzzle tigrenok sostavil/sostavlyal kartinku

• Puppy mould a bear sh’enok vylepil/lepil medvedya

Creation Expt: Scenarios

• Monkey assemble a smurf obez’yanka sobrala/sobirala gnomika

• Lion build a house l’venok postroil/stroil domik

• Tiger make a puzzle tigrenok sostavil/sostavlyal kartinku

• Puppy mould a bear sh’enok vylepil/lepil medvedya

Creation Expt: Scenarios

A road with 3 landmarks: a flower-bed, a castle and a tree. There are parts of a smurf at each location.

A monkey starts her journey down the road.

The monkey arrives at the flower-bed.These are nice flowers. Oh, look there are the pieces of a smurf down here. Let me try to revive this guy.

OK, the body goes on top of the legs, what’s next...

A bug bites the monkey. Ouch, that hurts!!! I don’t want to stay here any longer. I’m going to leave all of it like this and continue down the road.

The monkey reaches the castle.Oh, look, what a beautiful castle! And there are pieces of a smurf next to it. Let me try this one too!

OK, the body goes on top of the legs, what’s next...

A bug bites the monkey. Oh no, a bug bit me again! Why am I so unlucky today?No, this time, I’m going to finish this thing anyway!

The monkey assembles the smurf completely and continues along the road.

The monkey reaches the tree.What a great tree, it’s so nice to sit here. And there are some smurf pieces here again. But I guess I have to go home now.

The scene at the end of the story.

INCOMPLETE

COMPLETE

INCOMPLETE

COMPLETE

Gde obez’yanka sobrala gnomika?assemble-PERF

Where did the monkey assemble the smurf?

ALL CHILDREN

Gde obez’yanka sobrala gnomika?assemble-PERF

Where did the monkey assemble the smurf?

Gde obez’yanka sobirala gnomika? assemble-IMPWhere was the monkey assembling the smurf?

ADULTLIKE children vs.

Gde obez’yanka sobirala gnomika? assemble-IMPWhere was the monkey assembling the smurf?

ADULTLIKE group vs. NON-ADULTLIKE group

Adultlike group Non-adultlike group(N=5) (N=6)

Creation Expt: Results from Children

20/20 acceptances

22/24 rejections

20/20 acceptances

Interruptions• Interruptions occur twice per story; allows

independent test of ability to give 2 locations as answer:

Gde obez’yanku ukusil zhuk?Where was the monkey stung by a bug?

All children answered with 2 locations

build a smurf

Past incomplete

sobiralaI gnomikawas building a smurf

Where we are...

build a smurf

Present ongoing Past incomplete

build a smurf

sobiraetI gnomikais building a smurf

Where we are...

build a smurf

Present ongoing Past incomplete

build a smurf

Completion Hypothesis: children require possibility of completion in the actual world

sobiraetI gnomikais building a smurf

Where we are...

What is the cause of children’s error on IMP?

(A) Presence of the Object is needed?

• Creation verbs raise a separate problem:no object in the scene unless the event is completed

(e.g., Parsons notion of Incomplete Objects)

(A) Children require the presence of the object in the scene

• Change-of-state verbs (e.g. color in a flower) do not have this problem - the object (a flower) is present throughout the event independent of its completion

• Will the error from the Creation experiment persist with Change-of-state verbs? If children again reject IMP with incomplete events, then the problem is not (solely) due to the absence of the object in the scene

• Same task as in the Creation expt

• 34 children age 2;7 - 6;0• 4 stories per child

• Run in Moscow & Moscow region in Jan’02 & Aug’02

Change-of-state Expt: Design

• Tigrenok perevorachivalI/perevernulP kartinku Tiger turn over a picture

• Zaychik napolnyalI/napolnilP stakanchik Rabbit fill a glass

• Sh’enok razvorachivalI/razvernulP podarok Puppy unwrap a gift

• Kotenok zakrashivalI/zakrasilP cvetokKitty color in a flower

Change-of-state Expt: Scenarios

INCOMPLETECOMPLETE

Change-of-state Expt: Results from Children

Adultlike group

(N=13, mean age = 5;2)

88% acceptance (38/43 trials)

Change-of-state Expt: Results from Children

Adultlike group Non-adultlike group (N=13, mean age = 5;2) (N=16, mean age = 4;2)

86% rejection(49/57)

88% acceptance

(38/43 trials)

• PERF - 92% correct

• IMP - 52% correct

Mean age

Adultlike group N=13 5;2(accept IMP for INC in 88% trials)

Non-adultlike group N=16 4;2(accept IMP for INC in 14% trials)

(remaining 5 children - hard to classify due to inconsistent responses)

Change-of-state Expt: Summary of Results

(error due to Creation verbs)

NO: same error on IMP in the Change-of-state as in Creation Expt

color in a flower

Where we are...

Past incomplete / conative

zakrashivalaI cvetokwas coloring in a flower

color in a flower

Present ongoing

color in a flower

zakrashivaetI cvetokis coloring in a flower

Where we are...

Non-counterfactual Counterfactual

color in a flower

Present ongoing

color in a flower

Where we are...

The event is counterfactual iff - not completed by now or - cannot be completed in the future

color in a flower

Present ongoing

color in a flower

Complete Event Hypothesis: children fail to license IMP with counterfactual events

Where we are...

The event is counterfactual iff - not completed by now or - cannot be completed in the future

Parsons (1989)

(3) a. Mary was coloring in the flower. b. (e)[Coloring(e) & Subject(e,Mary) &

Object(e, the flower) & (t)[ t<now & Hold(e,t)]]

(4) a. Mary is coloring in the flower. b. (e)[Coloring(e) & Subject(e,Mary) &

Object(e, the flower) & (t)[ t=now & Hold(e,t)]]

Children accept the imperfective with present ongoing events,but reject it with past ongoing events =>

unexpected for Parsons’ theory

Parsons (1989)

(3) a. Mary was coloring in the flower. b. (e)[Coloring(e) & Subject(e,Mary) &

Object(e, the flower) & (t)[ t<now & Hold(e,t)]]

(4) a. Mary is coloring in the flower. b. (e)[Coloring(e) & Subject(e,Mary) &

Object(e, the flower) & (t)[ t=now & Hold(e,t)]]

Children accept the imperfective with present ongoing events,but reject it with past ongoing events =>

unexpected for Parsons’ theory

(B) Children mistakenly equate the semantics of Imperfective to that of Perfective (IMP = PERF)

• Children fail to make distinction between the semantics of IMP

and PERF in one particular situation - with conative (past

permanently incomplete) events (as in the Creation & Change-of-

state experiments)

• Maybe they will distinguish semantics of IMP from that of PERF

in some other situation

Children: semantics IMP = semantics PERF ?

bikewater the flowers

clean the table

Ongoing-success Experiment

(i) Poka mal’chik polival cvety, devochka vytiralaI stol. While the boy was watering flowers, the girl was cleaning the table.

clean the table

Adult Response

(ii) Poka mal’chik polival cvety, devochka vyterlaP stol. While the boy was watering flowers, the girl cleaned the table.

clean the table

Adult Response

clean the table

evaluation of Matrix event

Adult Response

If children behave like adults =>

they know some semantic difference between IMP & PERF

clean the table

evaluation of Matrix event

Adult Response

• Truth Value Judgment Task (Crain&Thornton 1998)

• 12 children age 3 - 5;10; 4 stories each

• Each story was such that

IMP sentence is correct

PERF sentence is wrong

• 39 trials total: 19 trials – IMP20 trials - PERF

Ongoing-success Expt: Design

Ongoing Experiment Results

39 trials total: 20 trials - PERF, 19 trials - IMP

Ongoing Experiment Results

PERF IMP

Children

Adults

NO: distinguish IMP from PERF in the ongoing-success situation

Landman (1992): IMP(A) – the imperfective form of the predicate A with a denotation - is true in a given event e iff

(i) E, such that e E, E (E – complete event/event type)

(ii) E CON (e, w)

CON (e, w) = w

The actual world is enough to find E

CON (e, w) ≠ w

Need to appeal to possible worlds to find E

e - non-counterfactual e - counterfactual

CON (e, w) = w

CON (e, w) ≠ w

Complete Event Hypothesis: children fail to license IMP with counterfactual events...

(ii) E CON (e, w)

CON (e, w) = w

CON (e, w) ≠ w

(ii) E CON (e, w)

Complete Event Hypothesis: children fail to license IMP with counterfactual events...

because they fail to properly deal with non-actual worlds

• Children incorrectly reject IMP in conative situations (Creation & Change-of-state expts)Okolo dereva Obez’yanka perevorachivalaI kartinku.

At the tree Monkey was turning over a picture.

Complete Event Hypothesis

turn over the picture

Fail if counterfactual e(no E can be found in the actual world)

• Children correctly accept IMP in the ongoing-success situation

Poka mal’chik polival cvety, devochka vytiralaI stol

While the boy was watering flowers, the girl was cleaning the table.

Complete Event Hypothesis

BOYGIRL

bikingwatering the flowers

cleaning the table

Succeed if non-counterfactual e

(E can be found in the actual world)

BOYGIRL

cleaning the table

BOYGIRL

cleaning the table

Follow-up:

Ongoing-failure

Ongoing-success

BOYGIRL

cleaning the table

BOYGIRL

cleaning the table

Follow-up:

Ongoing-failure

Ongoing-success

BOYGIRL

cleaning the table

BOYGIRL

cleaning the table

Follow-up:

Ongoing-failure

Ongoing-success

BOYGIRL

cleaning the table

BOYGIRL

cleaning the table

NON-counterfactualChildren accepted IMP

Counterfactual accept/reject IMP???

Follow-up:

Ongoing-failure

Ongoing-success

do the puzzle

clean the table

water flowers now

water flowers

clean the table

Ongoing-failure

Ongoing-success

Conative

Russian children

do the puzzle

clean the table

water flowers now

water flowers

clean the table

Ongoing-failure

Ongoing-success

Conative

Russian children

Complete EventHypothesis

do the puzzle

clean the table

water flowers now

water flowers

clean the table

Ongoing-failure

Ongoing-success

Conative

Russian children

• conative reading is NOT available

# Het meisje maakte een puzzel (bij het huis). The girl was doing a puzzle (at the tree).

Adult Dutch

do the puzzle

clean the table

water flowers now

water flowers

clean the table

• ongoing reading IS available

Terwijl Hans de bloemen aan het water geven was, maakte Maria de tafel schoon.

While the boy was watering the flowers, the girl did/was doing the puzzle.

do the puzzle

clean the table

water flowers now

water flowers

clean the table

Ongoing-failure

Ongoing-success

Conative

Rus adultsDutch

simple past =Rus children

• What may have looked at first like a typological anomaly, may turn out to reflect a form of ‘parametric learning’, moving from the ‘Dutch’ state to the adult Russian state

Outline

• Background

• Conclusions

Verbal Morphosyntax

Cross-Language Contrasts at Age 2

• ‘Root Infinitives’ in spontaneous speech of 2-year olds in many languages; alternate with finite forms

doggie wants snack doggie want snackHans ißt Brot Hans Brot essen

• Striking regularities in distribution of Root Infinitives across languages (update of Phillips, 1995)

• Root Infinitives absent from children’s wh-questions and topicalizations in German, Dutch, Swedish, etc.

• These are languages where the adult language disallows embedded infinitival wh-clauses

• Root Infinitives very rare in sentences with overt subjects in some languages

• These are languages where the adult language disallows ECM (e.g., I want John to leave); e.g. Dutch, German, Russian

• In languages which allow ECM, children produce overt subjects with RI’s; e.g. English, Danish, Icelandic

• Although the cause of RI’s remains unclear, the distribution of RI’s across languages closely tracks the language-specific syntax of infinitival clauses

• Why does language-specific knowledge appear so rapidly in this case?

• Surface syntax is easy-to-observe

Outline

• Background

• Conclusions

Conclusions

• Cross-language typology can help to predict developmental trajectories … imperfectly

• Must be combined with an independent understanding of what a child is equipped to easily observe in language input

• ‘Deep typology’ is not a replacement for ‘observational learning’; it enhances observational learning by making observations more powerful

• In this light, developmental trajectories could be projected rather more accurately

Unification Problem

Overview of Talks

1. The Unification Problem

2. Building Syntactic Relations

3. Abstraction: Sounds to Symbols

4. Linguistics and Learning

In-situ

1 2 3 4 5 6 7 8

Region

Reading Time

どの生徒に…

Prospects

• The Unification Problem is becoming a problem, not a mystery

• We can generate detailed hypotheses about real-time linguistic computation … and test them

• We can probe different levels of representation of the same external events

• We can draw close connections between theories of the adult state and theories of development

with help from ...University of Maryland

Shani AbadaSachiko Aoshima

Daniel Garcia-PedrosaAna Gouvea

Nina KazaninaMoti LiebermanLeticia PablosDavid PoeppelBeth RabbinSilke Urban

Carol WhitneyMasaya Yoshida

University of Delaware

Evniki EdgarBowen HuiBaris KabakTom Pellathy

Dave SchneiderKaia Wong

Alec Marantz, MITElron Yellin, MIT

National Science FoundationJames S. McDonnell Foundation

Human Frontiers Science ProgramJapan Science & Technology Program

Kanazawa Institute of Technology

http://www.ling.umd.edu/colin

colin@umd.edu

Experimental Results: Breakdown by Condition

Yes/No

Principle CN=12

Poka-sentN=12

FA 1N=10

FA 2N=10

0/4 7 1 0 0

1/3 2 3 0 0

2/2 3 1 1 2

3/1 0 5 2 3

4/0 0 2 7 5

FA1: While Poohi was reading a book, hei ate the apple.FA2: Poohi ate the apple while hei was reading a book.Pr_C: * Hei ate the apple while Poohi was reading the book. Poka-sent: *While hei was reading a book, Poohi ate the apple.

How the Task Works

• Child is not being judged

• Identical story for all test sentences

• Avoids child’s ‘yes’ bias

• Story favors the ungrammatical meaning

• Plausible denial

Plausible Denial

Eeyore

Plausible Denial

Grammatical - but FALSEEeyore

Plausible Denial

Grammatical - but FALSE

clearly FALSE, since it almost happened, but then didn’t

Eeyore

“Great. Now that Pooh is reading the book, I can eat this delicious apple.”

“I shouldn’t be such a greedy donkey - I should let Pooh eat the apple.”

“I suppose I have to eat a banana instead.”

language acquisition and cross-language variation colin phillips cognitive neuroscience of language...

Documents

a cognitive neuroscience view of schizophrenic … cognitive...

language the great equalizer education neuroscience language...

the neuroscience of inspirational leadership the...

cognitive neuroscience of language: 18: memory and...

linking language and cognition to neuroscience via ... ·...

language and...

neuroeconomics: how neuroscience can inform …...journal of...

escience meeting, edinburgh, november 2006. slide 1 carmen...

(cambridge companions to philosophy) pulvermueller f.-the...

take-home message for today from grammatical …...6/17/15!...

cognitive neuroscience in second language acquistion

social neuroscience...neuroscience of animal behaviors...

human language: chap31 - a cognitive neuroscience

the psycholinguistics of ellipsis colin phillips & dan...

a review of the neuroscience of second language acquisition

cognitive neuroscience of language

linguistic neuroscience: extending perceptual neuroscience...

the neuroscience of language. on brain circuits of words and...

mind and language: ontology and neuroscience

gray matters: how neuroscience can inform...