PSY 369: Psycholinguistics - the Department of Psychology at

Download Report

Transcript PSY 369: Psycholinguistics - the Department of Psychology at

PSY 369: Psycholinguistics
Language Acquisition
Announcements


Exam 2 moved to March 6th (the Thursday before
Spring Break).
Some other due dates have shifted:



http://www.routledge.com/cw/harley-9781848720893/s1/students/
Quiz 4, moved back a week (now Feb 28th)
Hmwk 3, moved to (Feb 27th)
Bonus videos (1 hr broken into 6 parts):






Why do we talk? The science of speech: part 1 (language acquisition
recording 3 years of language in a household)
Part 2 (chimps & language, language & brain)
Part 3 (language & brain cont., infant speech perception)
Part 4 (language and autism, language innateness [Chomsky], wild
children)
Part 5 (language in birds, language in our genes)
Part 6 (language and evolution, learning artificial languages)
Overview

General patterns and observations

Learning sounds



Learning words



Perception (e.g., categorical perception)
Production
Meaning
 Proposed Strategies
 Fast mapping
 Whole object
 Mutual exclusivity
Learning Syntax
Learning Morphology
Today’s focus
Language explosion continues

The language explosion is not just the result of simple
semantic development; the child is not just adding
more words to his/her vocabulary.

Child is mastering basic syntactic and morphological
processes.
Language explosion continues
Syntax

Mean length of utterance (MLU) in morphemes
6
5
4
MLU

3
2
1
0
0
20
40
age (months)
60
6
5
MLU
4
3
Language explosion continues
2
1
0
0
20
40
age (months)

Syntax

Mean length of utterance (MLU) in morphemes

Take 100 utterances and count the number of
morphemes per utterance
Daddy coming. Hi, car. Daddy car comed. Two car outside. It
getting dark. Allgone outside. Bye-bye outside.
# morphemes: 3, 2, 4, 3, 4, 2, 2
‘-ing’ and ‘-ed’ separate
morphemes
‘allgone’ treated as a single word
MLU = morphemes/utterances
= 20/7 = 2.86
60
Language explosion continues

Proto-syntax (??)

Holophrases (around 1-1.5 years)

Single-word utterances may be used to express more than the
meaning usually attributed to that single word by adults
“dog”
might refer to the dog is drinking water



Typically idiosyncratic, but some conventional/common (e.g.,
indicate the existence of an object, request recurrence of object
or event)
Often combined with intonation or gesture
Controversial claim: May reflect a developing sense of syntax,
but not yet knowing how to use it (e.g., see Bloom, 1973)
Language explosion continues

Syntax

Roger Brown (1973) proposed 5 stages

Stage 1: Telegraphic speech (MLU ~ 1.75; around 24 months)




Children begin to combine words into utterances
Limited to a small set of semantic relations (e.g., nomination,
recurrence, attribution, possession [see textbook table 4.5])
Debate: learning semantic relations or syntactic (position rules)
 “baby sleep” agent+action or Noun Verb
Children in telegraphic speech stage are said to leave out the ‘little
words’ and inflections:
 e.g. Mummy shoe NOT Mummy’s shoe
 Two cat NOT two cats
Language explosion continues

Syntax

Roger Brown (1973) proposed 5 stages

More than two words

Stages 2 through 5
 Stage 2 (MLU ~2.25)
 begin to modulate meaning using word order (syntax)
 Modulations for number, time, aspect
 Gradual acquisition of grammatical morphemes (“-ing”, “-s”
 Later stages reflect generally more complex use of syntax (e.g.,
questions, negatives)
Acquiring Syntax

Need to acquire (at least) two things:



The syntactic categories of the words (nouns, verbs, etc.)
How to order the words (or the syntactic categories)
General Theoretical Approaches

Innateness accounts


Universal grammar, Semantic bootstrapping
Learned accounts

Acquired from the linguistic input from the environment
Acquiring Syntax

Innateness account

Chomsky proposed that we have an innate facility for
language


Universal grammar: the rules that describe the fundamental properties
of all natural human languages
Chomsky’s arguments




The rules that underlie productive language use are highly abstract and
largely “unobservable”
We can understand and produce novel utterances (things that we’ve
never heard or said before)
Poverty of the stimulus: The gap between what language users know
and what they have access to during learning is too broad
Negative evidence: children rarely receive explicit feedback about
incorrect grammatical utterances
Acquiring Syntax

Innateness account

Semantic bootstrapping – use meaning to help learn syntax

e.g., Pinker (1984, 1989)
Child has innate
knowledge of
syntactic categories
Child learnsChild
the constructs some
Child makes guesses
and linking rules:meaningssemantic
representations
of
about syntactic structure
• Nouns, and Verbs
of
simple
sentences
some content words
based on surface form
• Thematic roles
(Agents,
“Doggie
eats
the
ball.”
“Doggie” “eat” “ball”
and semantic meaning
Patients)
“Doggie” is N, subject
“eats” is V, predicate
“ball” is N, object
Acquiring Syntax

Innateness account

Semantic bootstrapping – use meaning to help learn syntax


e.g., Pinker (1984, 1989)
Some problems with this account (e.g., Bowerman, 1990; Gleitman,
1990)



Do kids really enough utterances with identifiable agents and
patients that can be mapped onto nouns and verbs (verbs are
especially hard)?
Many words appear as both nouns and verbs (e.g., “go for a
walk” & “walk out the door”).
Easier mappings aren’t always learned earliest (e.g., verbs like
“fall” and “chase” have themes that map onto the sentence
subject. This should be simpler than verbs like “have” and
“got” in which the them maps onto the object)
Acquiring Syntax

Learned account

“It is in the stimulus” accounts (e.g. Bates, 1979, Elman, 1993)
(Statistical learning, Usage-based approaches)


Speech to children is not impoverished (Snow, 1977)
In all languages there are multiple potential cues indicating
semantic/syntactic relations (e.g., word order, case marking)



Similar words occur in similar linguistic contexts
Acoustic information (e.g., prosody) may provide syntactic cues (phonological
bootstrapping)
Children do not need innate knowledge to learn grammar
(evidence largely from successful computer models)
Acquiring Morphology

Morphology


Typically things like inflections and prepositions start around
MLU of 2.5 (usually in 2 yr olds)
Remember the Wug experiment (Berko-Gleason, 1958)
Acquiring Morphology

Morphology
This person knows how to rick. She did the same thing yesterday.
Yesterday she ________.
Typically children say that she “ricked.”
Acquiring Morphology

Morphology: order of acquisition
Age (yrs)
Morpheme
Example(s)
2
Present progressive
I driving
2
Articles
A dog, the doctor
2
Plural
Balls
2
Uncontractible Copula
He is asleep, am, are
3
Third person singular
He wants an apple
3
Full progressive
Be + ing, I am singing
3
Regular past tense
She walked
Acquiring Morphology

Children sometimes make mistakes.
My teacher holded the baby rabbits.
Yes
She holded the baby rabbits.
No, she holded them loosely.
Did you say your teacher held the baby rabbit?
What did you say she did?
Did you say held them tightly?
Acquiring Morphology

Children sometimes make mistakes.
My teacher holded the baby rabbits.

This is ungrammatical in the adult language


Shows that children are not simply imitating
In this case, what they produce something that is not in their
input.
Acquiring Morphology

Children sometimes make mistakes.
My teacher holded the baby rabbits.

Why do they make errors like these?



In the case at hand, we have what is called overregularization
The verb hold has an irregular past tense form, held
Because this form is used, the regular past tense-- that with ed-- is not found (*hold-ed)
Acquiring Morphology

The case of verb past tense:


Regular verb forms require no stored knowledge of the
past tense form (wug test)
 Past tense is accomplished by applying a past tense
rule (e.g., add -ed) to the verb stem
With irregular verbs something must be memorized

Examples:



Horton heared a Who
I finded Renée
The alligator goed kerplunk
Acquiring Morphology

The case of verb past tense:

Stages in the acquisition of irregular inflections


With regular verbs, the default form -ed is used
With irregulars, lists associating the verb with a
particular form of the past tense have to be memorized:




Past tense is -t when attached to leave, keep, etc.
Is -> was
Dig -> dug
Has -> had
Acquiring Morphology

Stages in the acquisition of irregular inflections
time
Step
1
2
3
4
5

Examples
Description
Noun Verb
Adjective
No inflecti o n
Ma n G o
Bad
Adult form
Me n We n t
Wo r s e
Overregularization Ma n s G o e d Badder
Transition
Me n s Wented Worser
Adult form
Men
We n t
Wo r s e
On the face of it, learning these morphological quirks follows a
peculiar pattern:
 Early: correct irregular forms are used
 Middle: incorrect regular forms are used
 Late: correct forms are used again
Memory & Rules

Why do we find this type of pattern?
 Memory and rules

The use of overregularized forms starts at around the
same that that the child is beginning to apply the default
-ed rule successfully



Early: All forms-- whether regular or irregular-- are
memorized
Middle: The regular rule is learned, and in some cases
overapplied
Late: Irregulars are used based on memory, regulars use
the rule (the idea is that if the word can provide its own
past tense from memory, then the past tense rule is
blocked)
Memory & Rules

Why do we find this type of pattern?
 Memory and rules
 Other accounts

Maratsos (2000) – frequency explanation

It is possible to predict which verbs will be subject to
overregularization
 The more often an irregular form occurs in the input, the
less likely the child is to use it as an overregularization
 This is evidence that some part of overregularization
occurs because of memory failures
 Something about irregulars is unpredictable, hence
has to be memorized
What kind of “teaching” do kids get?

If language is learned (and
not innate), how do kids do
it?


What kind of feedback do
they get?
Claim: Positive evidence is
not sufficient for learning a
language.
What kind of “teaching” do kids get?

Are the kids even aware of mistakes?

The children are apparently aware of the fact that their
forms are strange:




Parent: Where’s Mommy?
Child: Mommy goed to the store
Parent: Mommy goed to the store?
Child: NO! Daddy, I say it that way, not you
Positive and negative evidence

What kind of feedback is available for learning?


Positive evidence: Kids hear grammatical
sentences
Negative evidence: information that a given
sentence is ungrammatical

Kids are not told which sentences are ungrammatical
(no negative evidence)

Let’s consider no negative evidence further…
What kind of “teaching” do kids get?

How much Positive Evidence is there?





Estimated 5000 – 7000 utterances a day
Between ¼ and 1/3 are questions
Over 20% are not “full” adult sentences (typically Noun
or prepositional phrases)
Only about 15% have typical English SVO form
Roughly 45% of all maternal utterances began with one
of 17 words (e.g., “what”, “that”, “it”, “you”)
Cameron-Faulkner, et al (2003)
•
So what kids do hear may be somewhat limited.
Negative evidence

Negative evidence could come in various
conceivable forms.


“The sentence Bill a cookie ate is not a sentence in
English, Timmy. No sentence with SOV word order
is.”
Upon hearing Bill a cookie ate, an adult might
 Not understand
 Look pained
 Rephrase the ungrammatical sentence
grammatically
Kids resist instruction…
McNeill (1966)



Child: Nobody don’t like me.
Adult: No, say ‘nobody likes me.’
Child: Nobody don’t like me.
[repeats eight times]


Adult: No, now listen carefully; say ‘nobody likes me.’
Child: Oh! Nobody don’t likes me.
Kids resist instruction…
Cazden (1972) (observation attributed to Jean Berko Gleason)








Child: My teacher holded the baby rabbits and we patted them.
Adult: Did you say your teacher held the baby rabbits?
Child: Yes.
Adult: What did you say she did?
Child: She holded the baby rabbits and we patted them.
Adult: Did you say she held them tightly?
Child: No, she holded them loosely.
So there doesn’t seem to be a lot of explicit negative evidence, and
what there is the kids often resist
Negative evidence via feedback?

Do kids get “implicit” negative evidence?


Do adults understand grammatical sentences and not understand
ungrammatical ones?
Do adults respond positively to grammatical sentences and
negatively to ungrammatical ones?
Negative evidence via feedback?
Brown & Hanlon (1970):
Case study of “Adam” - looked at things that were said to
him by adults, and what he said to them




Adults understood 42% of the grammatical sentences.
Adults understood 47% of the ungrammatical ones.
Adults expressed approval after 45% of the
grammatical sentences.
Adults expressed approval after 45% of the ungrammatical
sentences.
Suggests that there isn’t a lot of good negative evidence.
In a way, it’s moot anyway…

One of the striking things about child language is how few
errors they actually make.



For negative feedback to work, the kids have to make the errors
(so that it can get the negative response).
But they don’t make enough relevant kinds of errors to
determine the complex grammar.
Pinker, Marcus and others, conclude that much of this stuff
must be innate.

But this isn’t the only view. There is an ongoing debate about
whether there are rules, or whether these patterns of behavior
can be learned based on the language evidence that is available
to the kids
Critical (sensitive) periods
Critical (sensitive) periods

Certain behavior is developed more quickly
within a critical period than outside of it. This
period is biologically determined.

Examples:

Imprinting in ducks (Lorenz, ; Hess, 1973)



Ducklings will follow the first moving thing they see
Only happens if they see something moving within the first
few hours (after 32 hours it won’t happen) of hatching
Binocular cells in humans


Cells in visual system that respond only to input from both
eyes.
If these cells don’t get input from both eyes within first year
of life, they don’t develop
Critical (sensitive) periods

Certain behavior is developed more quickly
within a critical period than outside of it. This
period is biologically determined.

Some environmental input is necessary for normal
development, but biology determines when the
organism is responsive to that input.

That “when” is the critical period
Critical period for language

Lenneberg (1967) proposed that there is a critical
period for human language

It assumes that language acquisition must occur
before the end of the critical period

Estimates range from 5 years up to onset of puberty
Evidence for critical period for language

Feral Children



Children raised in the wild or with reduced exposure to
human language
What is the effect of this lack of exposure on language
acquisition?
Two classic cases


Victor, the Wild Boy of Aveyron
Genie
Victor, The Wild Boy of Aveyron

Found in 1800 near the outskirts of Aveyron, France





Estimated to be about 7-years-old
Considered by some to be the first documented case of autism
Neither spoke or responded to speech
Taken to and studied by Dr. Jean-Marc-Gaspard Itard, and
educator of deaf-mute and retarded children
Never learned to speak and his receptive language ability was
limited to a few simple commands.

Described by Itard as “an almost normal boy who could not speak”
Genie

Found in Arcadia, California in 1970, was not
exposed to human language until age 13.5.

Raised in isolation a situation of extreme abuse

Genie could barely walk and could not talk when
found

Dr. Susan Curtiss made great efforts to teach her
language, and she did learn how to talk, but her
grammar never fully developed.

Only capable of producing telegraphic utterances
(e.g. Mike paint or Applesauce buy store)

Used few closed-class morphemes and function
words

Speech sounded like that of a 2-year-old
Genie

By age of 17 (after 4 years of extensive training)

Vocabulary of a 5 year old

Poor syntax (telegraphic speech mostly)

Examples

Mama wash hair in sink

At school scratch face

I want Curtiss play piano

Like go ride yellow school bus

Father take piece wood. Hit. Cry.
What Do These Cases Tell Us?

Suggestive of the position that there is a critical
period for first language learning (in particular for
syntax and phonological development)


If child is not exposed to language during early childhood
(prior to the age of 6 or 7), then the ability to learn syntax will
be impaired while other abilities are less strongly affected
Not uncontroversial: Victor and Genie and children like them
were deprived in many ways other than not being exposed to
language

Genie stopped talking after age 30 and was institutionalized
shortly afterward (Rymer, 1993)
What Do These Cases Tell Us?




Suggestive of the position that there is a critical
period for first language learning (in particular for
syntax and phonological development)
Why?
Nativist explanation (see pg 79 of text)
Maturational explanation: “less is more”
Second language learning

Learning a new language



What if we already know one language, but want to learn
another?
Adults learning another language typically have a persistent
foreign accent – perhaps a critical period for phonology
(Flege & Hillenbrand, 1984)
Adults typically do better initially at learning a new language
compared to kids, but kids typically do better over the long
term (Krashen, Long, & Scarcella, 1982)
Second language learning
Johnson and Newport (1989)

Native Chinese/Korean speakers moving to US
Task: Listen to sentences and judge whether
grammatically correct
Age and Second-language acquisition
mean score on
English grmmar test
R = -.87
Test score

280
270
260
250
240
230
220
210
200
2
Age of arrival
17
R = -.16
native
3 to 7
8 to 10
age of arrival
11 to 16
17 to 39
Test score

17
Age of arrival
40
Second language learning

Johnson and Newport (1989)





Native Chinese/Korean speakers moving to US
Task: Listen to sentences and judge whether
grammatically correct
Concluded that around the age of 16 something
happens
Different factors operate on language acquisition before
and after the age of 16
Birdsong and Molis (2001)

Replicated the Johnson and Newport study in
Spanish/English speakers.

Did not find a discontinuity around the age of 16
Effects of the Critical Period

Learning a language:




Under 7 years: perfect command of the language possible
Ages 8- c.15: Perfect command less possible progressively
Age 15-: Imperfect command possible
But these claims are far from universally accepted
Bilinguals & Polyglots

Many people speak more than one language


Tucker (1999) - multilinguals outnumber monolinguals
What is the impact of knowing/using more than one
language?



Factors affecting second language acquisition?
What does the lexicon look like?
Interesting effects in bilinguals
 Interference
 Code switching
 Cognitive advantages
Second language acquisition

Contexts of childhood bilingualism

Simultaneous

Both languages are acquired at the same time





Vocabulary growth of bilinguals is similar to that of monolinguals
Some aspects of acquisition may be slowed, but by age of 4
typically caught up
Doesn’t seem to matter whether languages are “related” or not
(e.g., English - French versus English Japanese)
Can achieve “fluency” in both languages
Sequential acquisition

The second language is learned after a first language

When the second language (L2) is acquired is important
 Early versus late learning (e.g., see the Johnson and
Newport study)
Second language acquisition

Frequency of usage of both languages



Mode of acquisition




How often and in what contexts do you use the two languages
“Use it or lose it” - language attrition
Native bilingualism - growing up in a two language environment
Immersion - schooling provided in a non-native language
Submersion - one learner surrounded by non-native speakers
Language dominance effects

Relative fluency of L1 and L2 may impact processing
Bilingual Representations

How do we represent linguistic information in a
bilingual lexicon?


Probably depends on many of the factors just discussed
Let’s look at some models and research focusing on the
situation where L1 is dominant relative to L2
Models of the bilingual lexicons
Potter et al (1984): Separate Stores Models – separate
lexicons for each language
Word Association Model
Concept Mediation Model
L1=First Language
L2=Second Language
CONCEPTS
L1
L2
CONCEPTS
L1
L2
Models of the bilingual lexicons
Paivio, Clark, & Lambert (1988): Common Stores
Models – words from both languages in same store
L1=First Language
L2=Second Language
CONCEPTS
L1 & L2
Revised Hierarchical Model

The results are mixed,
supporting more complex
models

May be different in different bilinguals
depending on things like age of acquisition,
relative proficiency, etc.

Kroll & Stewart (1994)

Proposed that the fluency of
L2 needs to be considered in
the processing model
concepts
conceptual
links
conceptual
links
lexical
links
L1
L2
Interesting effects in bilinguals



Interference
Code switching
Cognitive advantages
Interesting effects in bilinguals

Interference


Does knowing two languages lead to interference?
When found, interference is at multiple levels


Phonological - least amount of interference
Lexical - mixing words from different languages



Initially, appear to use a one word per thing strategy
But as they realize there that they’re speaking two language,
then they’ll use words from both languages simultaneously
Syntactic



Until year two, may use only one syntactic system which is
common to both languages
Then a brief period with two sets of lexical items, but still a
common syntax
Finally, two lexicons and two sets of syntax
Interesting effects in bilinguals

Determine who or what is the one performing the action.





The waitress pushes the cowboys.
The telephones pushes the cowboys.
Kisses the table the apple.
The baskets the teacher kicks.
As a native speaker of English we can use many
cues:




Word order
Animacy
Verb agreement
Not all languages use the same cues to the same
extent

e.g., German doesn’t rely as much on word order, but
relies more on agreement processes
Interesting effects in bilinguals

Determine who or what is the one performing the action.





The waitress pushes the cowboys.
The telephones pushes the cowboys.
Kisses the table the apple.
The baskets the teacher kicks.
Kilborn (1989, 1994)


Found that bilinguals (English as second language)
typically carry over the dominant processing strategies
from their native languages.
This interacts with their level of fluency in the second
language
Code switching

When bilinguals substitute a word or phrase from one
language with a phrase or word from another language
“I want a motorcycle VERDE”

Switching is systematic, not random
Code switching

When bilinguals substitute a word or phrase from one
language with a phrase or word from another language
“I want a motorcycle VERDE”

The Spanish adjective “verde” follows a grammatical rule that is
observed by most bilingual speakers that code-switch
“I want a VERDE motorcycle”

Would be incorrect

because language switching can occur only if the adjective is placed
according to the rules of the language of the adjective

In this case, the adjective is in Spanish; therefore, the adjective must
follow the Spanish grammatical rule that states that the noun must
precede the adjective
Code switching

When bilinguals substitute a word or phrase from one
language with a phrase or word from another language
“I want a motorcycle VERDE”

Generally, bilinguals take longer to read and comprehend
sentences containing code-switched words

May be due to a “mental switch mechanism” that determines which
of the bilingual’s two mental dictionaries are “on” or “off” during
language comprehension.

This mental switch is responsible for selecting the appropriate mental
dictionary to be employed during the comprehension of a sentence.
 E.g., if reading an English, a Spanish code-switched word is
encountered, the mental switch must disable the English linguistic
system, and enable the Spanish linguistic system.
Code switching

When bilinguals substitute a word or phrase from one
language with a phrase or word from another language
“I want a motorcycle VERDE”

Generally, bilinguals take longer to read and comprehend
sentences containing code-switched words

This time difference depends on similarity of the languages


Chinese-English bilinguals take longer to recognize English codeswitched words in Chinese sentences only if the English words contain
initial consonant-consonant (e.g., flight) clusters, simply because the
Chinese language lacks this phonotactic structure.
Another current view suggests that language dominance (i.e.,
which language is used more frequently) plays an important role in
code-switching
Cognitive advantages

Some evidence suggest that being bilingual can have
an impact on cognition outside of language

Bialystok and colleagues


Bilinguals are very proficient at switching between languages
Bilinguals also have to be good at suppressing the contextually
inappropriate language