Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Language 2 – Language

Acquisition!

• Nature-nurture debate in language is related to a debate about


whether things like language rely on special-purpose processing
modules or are instead a result of more general mechanisms
that are used across many different domains of processing.!

• People on the Nature side of the nature-nurture debate


typically believe that the brain has special-purpose processing
modules that are used only for language.!

• People on the Nurture side of the debate typically believe the


much or all of language is the result of more general learning
mechanisms.!

Nancy Kanwisher! Isabel Gauthier! Noam Chomsky!


http://mcgovern.mit.edu/ http://www.vanderbilt.edu/
principal-investigators/nancy- psychological_sciences/bio/
kanwisher! isabel-gauthier!
• In language, the traditional leader of the Nature side of
the debate is Noam Chomsky.!

• We call people on this side of the debate “Nativists”,


because they believe that language is a native ability of
the human brain.!

Nativist Argument #1: Language is unique to humans! Nativist Argument #2: Lack of Negative Evidence!

• Negative evidence is feedback indicating that something a


child said is grammatically incorrect.!
• Humans acquire language very easily, whereas no other
species has a full-blown language.! • Lack of negative evidence refers to the fact that adults
don’t usually indicate that a child’s utterances are
• Even when gorillas and chimpanzees can be taught a grammatically incorrect.!
language, it requires a lot of intensive training, whereas
human children learn language almost effortlessly.!

© S. J. Luck

All rights reserved
1

Nativist Argument #2: Lack of Negative Evidence! Nativist Argument #2: Lack of Negative Evidence!

Child: “Look. Doggie running.”! Child: “Look. Doggie running.”!


Negative evidence from adult: “No. That’s not the right way Adult: “No. That’s a kitty. The kitty is running.” (This is a
to say it.” (This does not happen very often)! more typical adult response)!

Nativist Argument #2: Lack of Negative Evidence! Nativist Argument #2: Lack of Negative Evidence!

• Children don’t get much negative evidence when they’re learning


language.!

• If adults don’t correct children when they make grammatical errors,


how do the children learn to speak correctly?!

• According to nativists like Chomsky, normal learning mechanisms


would be insufficient given the lack of negative evidence.!
Child: “Can I be excused.”!
Adult: “That’s not right. You should say ‘May I be excused’.”! • Consequently, there must be a special, innate language system that
Child doesn’t actually change in response to this feedback! doesn’t need negative evidence.!

Nativist Argument #3: Poverty of the stimulus! Nativist Argument #3: Poverty of the stimulus!

Degenerate: “having fallen


below a normal or desirable
level… deteriorated or
degraded”(dictionary.com)!
Examples:!
False starts: where we start
saying something and then
change our minds and start over! • If you’re trying to learn the language just by listening to
Production errors: where you it, how can you figure out the underlying rules of grammar
mean to say one thing and from the degenerate sentences that you’re hearing?!
something else comes out!
Incomplete sentences:
These sentences are missing the • According to the nativists, learning a language from this
subject, so they’re not legal impoverished input would be impossible without a special,
English sentences! innate language system.!

© S. J. Luck

All rights reserved
2

Noam Chomsky!
“Language Acquisition
Device”!

Noam Chomsky!
“Universal Grammar”!
• To deal with these challenges in language acquisition, Chomsky
initially proposed that humans are born with what he called a • Chomsky later proposed a somewhat different explanation for
language acquisition device.! language learning, which he called a universal grammar.!

• A special brain circuit that allows rapid language learning.! • All languages share some core features, and languages differ in terms
of some of the details. According to Chomsky, we are born with
• Chomsky did not give any details about where in the brain this was knowledge of the set of possible language features, and children just
located, or how it would work.! need to figure out which features are present in their language.!

Nativist Argument #1: Language is unique to humans!


Empiricist Response to
Nativist Arguments!

Noam Chomsky! Jenny Saffran!

• Language in its full form does appear to be unique to


humans.!

• However, the fact that other apes can learn to use sign
languages quite well indicates that general-purpose
Katie Graf Estes! learning mechanisms can get you pretty far in language.!

Nativist Argument #1: Language is unique to humans! Nativist Argument #2: Lack of Negative Evidence!

• Humans are better than other apes at learning lots of things, so why
shouldn’t they also be better at learning language?!
• What about the fact that adults don’t provide children with
• A gorilla might be able to drive a car in a simplified manner but negative feedback when they speak ungrammatically?!
wouldn’t be able to navigate a complicated route in traffic.!
• Adults do provide negative feedback; they just do it
• This doesn’t mean that humans are born with an innate car driving
implicitly rather than explicitly.!
circuit.!

• The ability of humans to learn language so readily may just be a • Parents frequently repeat back what their children say, but
result of better general-purpose learning mechanisms.! with corrected grammar.!

© S. J. Luck

All rights reserved
3

Nativist Argument #3: Poverty of the stimulus! Nativist Argument #3: Poverty of the stimulus!

Degenerate: “having fallen


below a normal or desirable • Infant-directed speech
level… deteriorated or (“motherese”)!
degraded”(dictionary.com)! – Simple, highly
grammatical sentences!
Examples:!
– Simplified vocabulary!
False starts!
– Slow rate / lots of pauses!
Production errors!
– Exaggerated prosody
Incomplete sentences! (variations in pitch and
loudness)!
• What about the idea that young children are hearing an – Lots of repetition!
impoverished, degenerate version of the language?! – Focused on concrete
aspects of here and now!

• How can they learn the actual grammar of the language if


Adult: “Look at the bunny! See the bunny! It’s eating! What is it
they’re constantly hearing incorrect grammar?!
eating? That’s right! It’s eating a carrot! The bunny is eating a
carrot! Bunnies like carrots!”!

Nativist Argument #3: Poverty of the stimulus!

• Recent research has shown that this


kind of infant-directed speech actually
makes a difference in language
learning.!

• Infants who heard a lot of infant-


directed speech when they were 1 had
substantially larger vocabularies when
they were 2.!

• This is just a correlation, but it’s one of


Patricia Kuhl!
http://ilabs.uw.edu/institute-faculty/ many findings suggesting that infant-
Father’s also engage in infant-directed speech.! bio/i-labs-patricia-k-kuhl-phd! directed speech helps babies learn
language.!

• Imagine that you’re an infant, and you’re trying to learn


Saffran Experiments! words.!

• How do you know where the word boundaries are when


you hear this sentence?!

This is an! example! of me speaking!

Jenny Saffran! No sound gap before or after “example”!

© S. J. Luck

All rights reserved
4

Transitional Probabilities! Using transitional probabilities to
learn word boundaries!
• Hypothesis: Infants can use transitional
probabilities to make good guesses about word/
morpheme boundaries!
• Transitional probability: Probability of a
transition between two phonemes or syllables!
– Some sequences of phonemes or syllables occur within
many words, but rarely occur between words!
• Examples! schwa!
– “tǝ” occurs within “tub” “tomato” “fetal”!
– Occurs between words only when first word ends in “t”
and next word begins with “ǝ” (“rust under”)!
– When you hear “_ _ _ t ǝ _ _ _”, it is likely that the
“tǝ” lies within a word rather than at a word boundary!
– Syllables: In pretty#baby, pre-tty more common than
tty-ba!

Using Transitional Probabilities to Implicit vs. Explicit Memory!


Learn Word Boundaries! Statistical information such as transitional
probabilities may be learned implicitly!
• If infants can learn the transitional probabilities of various pairs
of syllables, this could help them figure out where the word
boundaries are in continuous speech.!

In this learning game you are the weather


• But can infants actually do this? ! forecaster. You will learn how to predict
rain or shine using a deck of four cards:

• For a long time, people thought that infants had terrible


memory.!
!
• If infants have poor memory, how could they learn transitional
probabilities?!

(75% Sun) (57% Sun) (43% Sun) (25% Sun)

(Knowlton, Mangels, & Squire, 1996)!

Using transitional probabilities to


learn word boundaries!

Jenny Saffran! Richard Aslin! Elissa Newport!

• Needed to use an artificial language so they could make


sure the infants weren’t using prior knowledge.!
• Needed to use computer-generated speech to be sure
• Can infants actually learn these transitional that there weren’t any other cues, like pauses or changes
probabilities, just from hearing language in the in pitch, that the infants could use.!
background while they’re doing other things?!
Saffran, J.R., Aslin, R., & Newport, E.L. (1996). Statistical learning by 8-month-old infants. Science, 274, 1926-1928.!

© S. J. Luck

All rights reserved
5

Words!
bidaku!
padoti!
golabu!
bidaku#padoti#golabu#padoti#bidaku#bidaku#golabu…!

• Infants heard a continuous stream of these words, in random order,


• Pparents brought their 8-month-old infants into the lab.! for 2 minutes.!

• No gaps between the words -- a continuous stream of syllables.!


• The infants just played quietly.!
• The pound signs indicate word boundaries, but there were no silent
• While they were playing, the artificial language was played in gaps between the words.!
the background.!
• Transitional probabilities between syllables provided the only
• The infants heard just 2 minutes of the artificial language, and
information that could be used to divide this continuous stream into
then the researchers tested whether they had learned the words.!
transitional probabilities.!

• If the infants learned the transitional probabilities between the


bidaku#padoti#golabu#padoti#bidaku#bidaku#golabu…! various syllable pairs, they would know that the probability from bi-da
and da-ku was high, whereas the transition from ku to pa was low!

• Whenever the syllable “bi” was presented, it had • And they could use this information to learn that bidaku was a
a 100% probability of being followed by “da”, separate word from padoti, etc.!
because they’re part of the same word, bidaku!
• But how can we tell if these nonverbal 8-month-old infants were able
to learn the words based on these transitional probabilities?!
• However, “ku” could be followed by either “pa”
or “go” or “bi”, depending on whether the next • To answer questions like this, studies of infant cognition take
word was padoti, golabu, or bidaku.! advantage of the fact that infants get bored if they see or hear the
same thing over an over again.!

• So the ku-pa transition had a probability of only • They prefer things that are novel.!
33%!

• The ti-go and bu-pa transitions also had 33%


probabilities!

Words! Nonwords!
bidaku! kupado! Test trials start with all lights off!
padoti! tigola! Center light blinks!
golabu! bubida! bidaku! Brings infant’s attention to center!
bidaku …!
One of the side-lights blinks!
bidaku#padoti#golabu#padoti#bidaku#bidaku#golabu…!
When infant looks at side-light, !
words or nonwords are presented!
as long as infant keeps looking!
• If the infants learned the words from the artificial
language, you might expect that they would be bored Word or non-word stops when
the infant looks away!
with them because they kept hearing them over and over.!
Infant will look longer if stimulus
is novel than if it is familiar!
• They wouldn’t be as bored with nonwords created by
taking sequences that crossed the word boundaries, like Finding: Infants looked away faster for words (e.g., bidaku) than for
how kupado is the last syllable of bidaku followed by the nonwords (e.g., kupado). This means that they could differentiate
first two syllables of padoti.! between words and nonwords on the basis of transitional probabilities.!
Courtesy of Bob McMurray!

© S. J. Luck

All rights reserved
6

Using transitional probabilities to
learn word boundaries!

• Amazing result, because the infants heard the artificial


language for only two minutes.!

• They did this while the artificial language was presented in • This finding makes it quite plausible that infants use
the background, without being directed to listen to it.! general-purpose learning mechanisms when learning many
aspects of language during infancy.!
• This was a very simple language, with only three words,
but this study shows that infants can learn transitional • Of course, nativists would say that domain-general
probabilities quite readily.! learning mechanisms might be useful for simple aspects
of language learning, but not for more complex aspects
such as syntax.!

• However, more recent research has indicated that


transitional problems can also play a role in syntax
learning (Thompson & Newport, 2007).!

Noam Chomsky!
“Language Acquisition Device”!
“Universal Grammar”!

• Unlikely that we have a language acquisition


device or universal grammar system that is • On the other hand, the ability of young children to learn
specified entirely by our DNA.! language effortlessly is amazing, especially given how
hard it is for them to learn other things that seem a lot
• DNA is just a code for making proteins, and it
simpler.!
would be difficult to specify a detailed set of
grammatical rules by means of proteins.!

Immanuel Kant (1724-1804)! • Language researchers are trying to


understand exactly what is built in,
and exactly what is acquired from
Pure concepts of the
experience.!
understanding!
• There is a lot of research going on to
answer this question, and a lot of
vigorous debate. !

• Some researchers think that a great


deal of language-specific information
is built in, whereas other researchers
• This brings us back to Immanuel Kant, who proposed that think that general-purpose learning
we have innate “pure concepts of the understanding” that mechanisms do most of the work.!
we use to organize our empirical observations of the
• But either way, it’s clear that both
world.!
nature and nurture play important
roles in the ability of young children
• We need some built-in cognitive abilities to learn things, to learn a language simply by being
including language.! immersed in it.!

© S. J. Luck

All rights reserved
7

You might also like