LCMA 1.9rev

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 204

Course Outline & Introduction

Ishani Guha

HSL 747: Language Computations & Mental Architecture

July 25, 2023

HSL747/LCMA-1/IITDelhi I.Guha 1
Course Logistics

HSL747/LCMA-1/IITDelhi I.Guha 2
Course Logistics

▸ Instructors: Ishani Guha, Prof. Deepak Alok


▸ TA: Madhusmitha Venkatesan
▸ Time: Tu, Th, Fr - 11am-11:50am (note exceptions in Time Table)
▸ Room: LH-325
▸ TA sessions: to be decided
▸ Emails: (NamedotSurname@hss.iitd.ac.in)

HSL747/LCMA-1/IITDelhi I.Guha 2
Evaluation:

HSL747/LCMA-1/IITDelhi I.Guha 3
Evaluation:

▸ Minor: 30/100, Module 1 & 2; 15th September, Fr (tentative)

HSL747/LCMA-1/IITDelhi I.Guha 3
Evaluation:

▸ Minor: 30/100, Module 1 & 2; 15th September, Fr (tentative)


▸ Class Test: 10/100, Module 3; 27th October, Fr (tentative)

HSL747/LCMA-1/IITDelhi I.Guha 3
Evaluation:

▸ Minor: 30/100, Module 1 & 2; 15th September, Fr (tentative)


▸ Class Test: 10/100, Module 3; 27th October, Fr (tentative)
▸ Presentation: 5/100, Module 4, to be scheduled

HSL747/LCMA-1/IITDelhi I.Guha 3
Evaluation:

▸ Minor: 30/100, Module 1 & 2; 15th September, Fr (tentative)


▸ Class Test: 10/100, Module 3; 27th October, Fr (tentative)
▸ Presentation: 5/100, Module 4, to be scheduled
▸ Major: 50/100, Modules 1, 2 & 3, 18-25th November

HSL747/LCMA-1/IITDelhi I.Guha 3
Evaluation:

▸ Minor: 30/100, Module 1 & 2; 15th September, Fr (tentative)


▸ Class Test: 10/100, Module 3; 27th October, Fr (tentative)
▸ Presentation: 5/100, Module 4, to be scheduled
▸ Major: 50/100, Modules 1, 2 & 3, 18-25th November
▸ Attendance: 5/100

HSL747/LCMA-1/IITDelhi I.Guha 3
Grading matrices

Grading breakdown Attendance breakdown

a. More than 90: A+ Total no. of classes: 47


b. 70-89: A a. More than 35 classes: 5
c. 60-69: B+ b. 30-35 classes: 4
d. 55-59: B c. 24-29 classes: 3
e. 50-54: C+ d. 18-23 classes: 2
f. 45-49: C e. 14-17 classes: 1
g. 40-44: D f. Less than 14 classes: 0
h. 30-39: E
i. Less than 30: F

HSL747/LCMA-1/IITDelhi I.Guha 4
Course Content

HSL747/LCMA-1/IITDelhi I.Guha 5
Course Content

▸ Module 1: The Faculty of Language - Guha


▸ Theories of Language (Learning), Biolinguistic Program - Recur-
sion, Non-Human Communication Systems, Language Evolution,
Gesture creation by deaf children
▸ Module 2: The Structure of Words and Sentences (Tutorials re-
quired, HWs) - Deepak Alok
▸ Morphology & Phrase Structure Grammar/Syntax
▸ Module 3: The Meaning Component (Tutorials required) - Guha
▸ Truth-conditions, Entailment & Sets, Propositional Logic with some
proofs, Implicature, vs. Entailment, Primate Studies, Experimen-
tal Semantics/Pragmatics
▸ Module 4: Debates & Discussions - All of us
▸ Some debates on language, Student Presentations

HSL747/LCMA-1/IITDelhi I.Guha 5
Resources

HSL747/LCMA-1/IITDelhi I.Guha 6
Resources

▸ Google drive folder ‘HSL747-2023-Class’


▸ Excel sheet
Time Table, Important info.
▸ Required/Suggested readings:
Handouts, slides, book chapters, papers
▸ Bibliography & other materials

HSL747/LCMA-1/IITDelhi I.Guha 6
Language

HSL747/LCMA-1/IITDelhi I.Guha 7
Language

▸ Learning

HSL747/LCMA-1/IITDelhi I.Guha 7
Language

▸ Learning
▸ Acquisition - How does a child learn a language?

HSL747/LCMA-1/IITDelhi I.Guha 7
Language

▸ Learning
▸ Acquisition - How does a child learn a language?
▸ Uniqueness

HSL747/LCMA-1/IITDelhi I.Guha 7
Language

▸ Learning
▸ Acquisition - How does a child learn a language?
▸ Uniqueness
▸ Language vs. other animal communication systems?

HSL747/LCMA-1/IITDelhi I.Guha 7
Language acquisition

HSL747/LCMA-1/IITDelhi I.Guha 8
Language acquisition
How do children learn languages?

HSL747/LCMA-1/IITDelhi I.Guha 8
Language acquisition
How do children learn languages?
▸ Clearly not by memorization.

HSL747/LCMA-1/IITDelhi I.Guha 8
Language acquisition
How do children learn languages?
▸ Clearly not by memorization.

An average person knows about 50,000 words.

HSL747/LCMA-1/IITDelhi I.Guha 8
Language acquisition
How do children learn languages?
▸ Clearly not by memorization.

An average person knows about 50,000 words.


You can combine those in 50,000! ways. ; infinity

HSL747/LCMA-1/IITDelhi I.Guha 8
Language acquisition
How do children learn languages?
▸ Clearly not by memorization.

An average person knows about 50,000 words.


You can combine those in 50,000! ways. ; infinity
(100! is a number with 158 digits.)

HSL747/LCMA-1/IITDelhi I.Guha 8
Language acquisition
How do children learn languages?
▸ Clearly not by memorization.

An average person knows about 50,000 words.


You can combine those in 50,000! ways. ; infinity
(100! is a number with 158 digits.)
Even if one has to only remember strings with at most 50 words,
that is a very large number.

HSL747/LCMA-1/IITDelhi I.Guha 8
Language acquisition
How do children learn languages?
▸ Clearly not by memorization.

An average person knows about 50,000 words.


You can combine those in 50,000! ways. ; infinity
(100! is a number with 158 digits.)
Even if one has to only remember strings with at most 50 words,
that is a very large number.
▸ Most sentences are very rare.

HSL747/LCMA-1/IITDelhi I.Guha 8
Language acquisition
How do children learn languages?
▸ Clearly not by memorization.

An average person knows about 50,000 words.


You can combine those in 50,000! ways. ; infinity
(100! is a number with 158 digits.)
Even if one has to only remember strings with at most 50 words,
that is a very large number.
▸ Most sentences are very rare.

The most frequent words are very frequent, eg. the, am, is, are,
have, say, etc.

HSL747/LCMA-1/IITDelhi I.Guha 8
Language acquisition
How do children learn languages?
▸ Clearly not by memorization.

An average person knows about 50,000 words.


You can combine those in 50,000! ways. ; infinity
(100! is a number with 158 digits.)
Even if one has to only remember strings with at most 50 words,
that is a very large number.
▸ Most sentences are very rare.

The most frequent words are very frequent, eg. the, am, is, are,
have, say, etc.
But most words are very rare, they only occur once in a text/dis-
course.

HSL747/LCMA-1/IITDelhi I.Guha 8
Language acquisition
How do children learn languages?
▸ Clearly not by memorization.

An average person knows about 50,000 words.


You can combine those in 50,000! ways. ; infinity
(100! is a number with 158 digits.)
Even if one has to only remember strings with at most 50 words,
that is a very large number.
▸ Most sentences are very rare.

The most frequent words are very frequent, eg. the, am, is, are,
have, say, etc.
But most words are very rare, they only occur once in a text/dis-
course.
But the sentences change depending on the rare/ unique words.

HSL747/LCMA-1/IITDelhi I.Guha 8
Language acquisition
How do children learn languages?
▸ Clearly not by memorization.

An average person knows about 50,000 words.


You can combine those in 50,000! ways. ; infinity
(100! is a number with 158 digits.)
Even if one has to only remember strings with at most 50 words,
that is a very large number.
▸ Most sentences are very rare.

The most frequent words are very frequent, eg. the, am, is, are,
have, say, etc.
But most words are very rare, they only occur once in a text/dis-
course.
But the sentences change depending on the rare/ unique words.

see: “Productivity, and Zipf’s Law”, in Introduction to Linguistics by Ed


Stabler
HSL747/LCMA-1/IITDelhi I.Guha 8
Productivity

▸ Productivity

HSL747/LCMA-1/IITDelhi I.Guha 9
Productivity

▸ Productivity

Every human language has an unlimited number of sentences.

HSL747/LCMA-1/IITDelhi I.Guha 9
Productivity

▸ Productivity

Every human language has an unlimited number of sentences.


What is the longest sentence (in English)?
The Guinness Book of World Records once recognized the longest
English sentence (1300-words) in William Faulkner’s novel Absa-
lom, Absalom! , which begins like:
They both bore as though in deliberate flagellant exaltation…

HSL747/LCMA-1/IITDelhi I.Guha 9
Productivity

▸ Productivity

Every human language has an unlimited number of sentences.


What is the longest sentence (in English)?
The Guinness Book of World Records once recognized the longest
English sentence (1300-words) in William Faulkner’s novel Absa-
lom, Absalom! , which begins like:
They both bore as though in deliberate flagellant exaltation…
How can you make this sentence longer?

HSL747/LCMA-1/IITDelhi I.Guha 9
Productivity

▸ Productivity

Every human language has an unlimited number of sentences.


What is the longest sentence (in English)?
The Guinness Book of World Records once recognized the longest
English sentence (1300-words) in William Faulkner’s novel Absa-
lom, Absalom! , which begins like:
They both bore as though in deliberate flagellant exaltation…
How can you make this sentence longer?
Faulkner wrote “they both…”

HSL747/LCMA-1/IITDelhi I.Guha 9
Productivity

▸ Productivity

Every human language has an unlimited number of sentences.


What is the longest sentence (in English)?
The Guinness Book of World Records once recognized the longest
English sentence (1300-words) in William Faulkner’s novel Absa-
lom, Absalom! , which begins like:
They both bore as though in deliberate flagellant exaltation…
How can you make this sentence longer?
Faulkner wrote “they both…”
Can you make it even longer?

HSL747/LCMA-1/IITDelhi I.Guha 9
Productivity

▸ Productivity

Every human language has an unlimited number of sentences.


What is the longest sentence (in English)?
The Guinness Book of World Records once recognized the longest
English sentence (1300-words) in William Faulkner’s novel Absa-
lom, Absalom! , which begins like:
They both bore as though in deliberate flagellant exaltation…
How can you make this sentence longer?
Faulkner wrote “they both…”
Can you make it even longer?
I wonder if Faulkner wrote “they both…”/ Faulkner wrote “they
both…”emoji

HSL747/LCMA-1/IITDelhi I.Guha 9
Productivity

▸ Productivity

Every human language has an unlimited number of sentences.


What is the longest sentence (in English)?
The Guinness Book of World Records once recognized the longest
English sentence (1300-words) in William Faulkner’s novel Absa-
lom, Absalom! , which begins like:
They both bore as though in deliberate flagellant exaltation…
How can you make this sentence longer?
Faulkner wrote “they both…”
Can you make it even longer?
I wonder if Faulkner wrote “they both…”/ Faulkner wrote “they
both…”emoji
The Language Instinct by Steven Pinker

HSL747/LCMA-1/IITDelhi I.Guha 9
Language acquisition
27/07/2023

How do children get a grasp of this ‘unlimited’ knowledge?

HSL747/LCMA-1/IITDelhi I.Guha 10
Language acquisition
27/07/2023

How do children get a grasp of this ‘unlimited’ knowledge?


▸ By identifying discrete units in the continuous stream of speech
they are exposed to

HSL747/LCMA-1/IITDelhi I.Guha 10
Language acquisition
27/07/2023

How do children get a grasp of this ‘unlimited’ knowledge?


▸ By identifying discrete units in the continuous stream of speech
they are exposed to
▸ By learning to combine those units

HSL747/LCMA-1/IITDelhi I.Guha 10
Language acquisition
27/07/2023

How do children get a grasp of this ‘unlimited’ knowledge?


▸ By identifying discrete units in the continuous stream of speech
they are exposed to
▸ By learning to combine those units

Children must acquire the system of rules that can generate unlimited
number of sentences.
Puzzles of language acquisition:

HSL747/LCMA-1/IITDelhi I.Guha 10
Language acquisition
27/07/2023

How do children get a grasp of this ‘unlimited’ knowledge?


▸ By identifying discrete units in the continuous stream of speech
they are exposed to
▸ By learning to combine those units

Children must acquire the system of rules that can generate unlimited
number of sentences.
Puzzles of language acquisition:
▸ Children are exposed to particular sentences - but they learn the
underlying abstract rules of sentence formation

HSL747/LCMA-1/IITDelhi I.Guha 10
Language acquisition
27/07/2023

How do children get a grasp of this ‘unlimited’ knowledge?


▸ By identifying discrete units in the continuous stream of speech
they are exposed to
▸ By learning to combine those units

Children must acquire the system of rules that can generate unlimited
number of sentences.
Puzzles of language acquisition:
▸ Children are exposed to particular sentences - but they learn the
underlying abstract rules of sentence formation
▸ Every child is exposed to a different set of particular sentences of
a given language - yet they all learn the same rules…

HSL747/LCMA-1/IITDelhi I.Guha 10
Language acquisition
27/07/2023

How do children get a grasp of this ‘unlimited’ knowledge?


▸ By identifying discrete units in the continuous stream of speech
they are exposed to
▸ By learning to combine those units

Children must acquire the system of rules that can generate unlimited
number of sentences.
Puzzles of language acquisition:
▸ Children are exposed to particular sentences - but they learn the
underlying abstract rules of sentence formation
▸ Every child is exposed to a different set of particular sentences of
a given language - yet they all learn the same rules…
in roughly the same sequence.

HSL747/LCMA-1/IITDelhi I.Guha 10
Sequence of Language Acquisition

HSL747/LCMA-1/IITDelhi I.Guha 11
Sequence of Language Acquisition

The Newborns:

HSL747/LCMA-1/IITDelhi I.Guha 11
Sequence of Language Acquisition

The Newborns:
▸ Newborn infants can distinguish between speech sounds of their
mother tongue as opposed to the speech sounds of other languages.
(in fact babies can do that even before they are born!)

HSL747/LCMA-1/IITDelhi I.Guha 11
Sequence of Language Acquisition

The Newborns:
▸ Newborn infants can distinguish between speech sounds of their
mother tongue as opposed to the speech sounds of other languages.
(in fact babies can do that even before they are born!)
▸ They can distinguish between allophonic and phonemic contrasts
that are not present in their mother tongue. eg. babies of En-
glish speaking parents can distinguish between Hindi retroflex and
dental sounds, even though their parents cannot. Infants lose this
ability from around six months of age.

HSL747/LCMA-1/IITDelhi I.Guha 11
Sequence of Language Acquisition

The Newborns:
▸ Newborn infants can distinguish between speech sounds of their
mother tongue as opposed to the speech sounds of other languages.
(in fact babies can do that even before they are born!)
▸ They can distinguish between allophonic and phonemic contrasts
that are not present in their mother tongue. eg. babies of En-
glish speaking parents can distinguish between Hindi retroflex and
dental sounds, even though their parents cannot. Infants lose this
ability from around six months of age.
▸ Newborns do not distinguish between contrasts that are not phone-
mic in any human language, e.g. contrast in terms of loudness.

HSL747/LCMA-1/IITDelhi I.Guha 11
Sequence of Language Acquisition

HSL747/LCMA-1/IITDelhi I.Guha 12
Sequence of Language Acquisition

▸ Babbling ; 6-9/12 months


▸ First words/ Holophrastic phase ; after 1 year
▸ Telegraphic Stage ; 1 12 years onwards

HSL747/LCMA-1/IITDelhi I.Guha 12
Sequence of Language Acquisition

▸ Babbling ; 6-9/12 months


▸ First words/ Holophrastic phase ; after 1 year
▸ Telegraphic Stage ; 1 12 years onwards

How does the child segment the continuous flow of speech/ linguistic
input?

HSL747/LCMA-1/IITDelhi I.Guha 12
Sequence of Language Acquisition

▸ Babbling ; 6-9/12 months


▸ First words/ Holophrastic phase ; after 1 year
▸ Telegraphic Stage ; 1 12 years onwards

How does the child segment the continuous flow of speech/ linguistic
input? Is the input annotated?

HSL747/LCMA-1/IITDelhi I.Guha 12
Sequence of Language Acquisition

▸ Babbling ; 6-9/12 months


▸ First words/ Holophrastic phase ; after 1 year
▸ Telegraphic Stage ; 1 12 years onwards

How does the child segment the continuous flow of speech/ linguistic
input? Is the input annotated?
▸ Prosodic bootstrapping, Whole object principle - overextension
- one object one name, Syntactic bootstrapping, Semantic boot-
strapping, Word frames, etc.

HSL747/LCMA-1/IITDelhi I.Guha 12
Sequence of Language Acquisition

▸ Babbling ; 6-9/12 months


▸ First words/ Holophrastic phase ; after 1 year
▸ Telegraphic Stage ; 1 12 years onwards

How does the child segment the continuous flow of speech/ linguistic
input? Is the input annotated?
▸ Prosodic bootstrapping, Whole object principle - overextension
- one object one name, Syntactic bootstrapping, Semantic boot-
strapping, Word frames, etc.

“Language Acquisition” in An Introduction to Language by Fromkin et al.

HSL747/LCMA-1/IITDelhi I.Guha 12
Sequence of Language Acquisition

HSL747/LCMA-1/IITDelhi I.Guha 13
Sequence of Language Acquisition

▸ Prosodic bootstrapping:

HSL747/LCMA-1/IITDelhi I.Guha 13
Sequence of Language Acquisition

▸ Prosodic bootstrapping:
Children learning English figure out onset of words by the stress
patterns in speech.
Trochaic: kítty, dóggy, mónster, rábbit, dóctor,…most English words
Iambic:delíght, amúse, guitár,…

HSL747/LCMA-1/IITDelhi I.Guha 13
Sequence of Language Acquisition

▸ Prosodic bootstrapping:
Children learning English figure out onset of words by the stress
patterns in speech.
Trochaic: kítty, dóggy, mónster, rábbit, dóctor,…most English words
Iambic:delíght, amúse, guitár,…
What about languages where stress is not a reliable cue?

HSL747/LCMA-1/IITDelhi I.Guha 13
Sequence of Language Acquisition

▸ Prosodic bootstrapping:
Children learning English figure out onset of words by the stress
patterns in speech.
Trochaic: kítty, dóggy, mónster, rábbit, dóctor,…most English words
Iambic:delíght, amúse, guitár,…
What about languages where stress is not a reliable cue?
▸ Strategies for learning words:

HSL747/LCMA-1/IITDelhi I.Guha 13
Sequence of Language Acquisition

▸ Prosodic bootstrapping:
Children learning English figure out onset of words by the stress
patterns in speech.
Trochaic: kítty, dóggy, mónster, rábbit, dóctor,…most English words
Iambic:delíght, amúse, guitár,…
What about languages where stress is not a reliable cue?
▸ Strategies for learning words:
▸ Whole object principle:
Children learn new words by assigning them to whole objects, not
to parts
▸ Overextension by shape/size:
Children use a word for a broader class of objects. dog ; sheep,
cow, any four legged animal
▸ One name one thing (mutual exclusivity bias):
Children assume one object has only one name. Helps in learning
names of parts.

HSL747/LCMA-1/IITDelhi I.Guha 13
Sequence of Language Acquisition
28/07/2023

HSL747/LCMA-1/IITDelhi I.Guha 14
Sequence of Language Acquisition
28/07/2023

▸ Semantic bootstrapping:

HSL747/LCMA-1/IITDelhi I.Guha 14
Sequence of Language Acquisition
28/07/2023

▸ Semantic bootstrapping:
Children use the meaning of a word to figure out its grammatical
category. Names of objects ; Nouns, Names of actions ; Verbs

HSL747/LCMA-1/IITDelhi I.Guha 14
Sequence of Language Acquisition
28/07/2023

▸ Semantic bootstrapping:
Children use the meaning of a word to figure out its grammatical
category. Names of objects ; Nouns, Names of actions ; Verbs
How do children learn the category of abstract words?

HSL747/LCMA-1/IITDelhi I.Guha 14
Sequence of Language Acquisition
28/07/2023

▸ Semantic bootstrapping:
Children use the meaning of a word to figure out its grammatical
category. Names of objects ; Nouns, Names of actions ; Verbs
How do children learn the category of abstract words?or even their
meaning?
▸ Syntactic bootstrapping:

HSL747/LCMA-1/IITDelhi I.Guha 14
Sequence of Language Acquisition
28/07/2023

▸ Semantic bootstrapping:
Children use the meaning of a word to figure out its grammatical
category. Names of objects ; Nouns, Names of actions ; Verbs
How do children learn the category of abstract words?or even their
meaning?
▸ Syntactic bootstrapping:
Children learn word classes from the syntactic environment the
words occur in.

HSL747/LCMA-1/IITDelhi I.Guha 14
Sequence of Language Acquisition
28/07/2023

▸ Semantic bootstrapping:
Children use the meaning of a word to figure out its grammatical
category. Names of objects ; Nouns, Names of actions ; Verbs
How do children learn the category of abstract words?or even their
meaning?
▸ Syntactic bootstrapping:
Children learn word classes from the syntactic environment the
words occur in.
▸ Word frames:
Children use frequently occurring frames in adult language to fig-
ure out the category of words. Verbs ; You _ it, Adjectives ;
the _ one

HSL747/LCMA-1/IITDelhi I.Guha 14
Sequence of Language Acquisition
28/07/2023

▸ Semantic bootstrapping:
Children use the meaning of a word to figure out its grammatical
category. Names of objects ; Nouns, Names of actions ; Verbs
How do children learn the category of abstract words?or even their
meaning?
▸ Syntactic bootstrapping:
Children learn word classes from the syntactic environment the
words occur in.
▸ Word frames:
Children use frequently occurring frames in adult language to fig-
ure out the category of words. Verbs ; You _ it, Adjectives ;
the _ one
Children are quite creative in the process of language learning.

HSL747/LCMA-1/IITDelhi I.Guha 14
Theories of Language Learning

HSL747/LCMA-1/IITDelhi I.Guha 15
Theories of Language Learning

▸ Behaviorist
▸ Connectionist

HSL747/LCMA-1/IITDelhi I.Guha 15
Theories of Language Learning

▸ Behaviorist ▸ Nativist
▸ Connectionist ▸ Social-Cognitive

HSL747/LCMA-1/IITDelhi I.Guha 15
Behaviorist Theories of Language Learning

HSL747/LCMA-1/IITDelhi I.Guha 16
Behaviorist Theories of Language Learning

▸ Before 1950s/60s

HSL747/LCMA-1/IITDelhi I.Guha 16
Behaviorist Theories of Language Learning

▸ Before 1950s/60s
▸ Behaviorism
▸ emphasizes behavior over mind or cognitive processes
▸ describes language as a behavior in terms of notions such as stim-
ulus, response, reinforcement and punishment

HSL747/LCMA-1/IITDelhi I.Guha 16
Behaviorist Theories of Language Learning

▸ Before 1950s/60s
▸ Behaviorism
▸ emphasizes behavior over mind or cognitive processes
▸ describes language as a behavior in terms of notions such as stim-
ulus, response, reinforcement and punishment
▸ Language learning is governed by general laws of learning - habit-
formation by imitation and reinforcement.

HSL747/LCMA-1/IITDelhi I.Guha 16
Behaviorist Theories of Language Learning

▸ Before 1950s/60s
▸ Behaviorism
▸ emphasizes behavior over mind or cognitive processes
▸ describes language as a behavior in terms of notions such as stim-
ulus, response, reinforcement and punishment
▸ Language learning is governed by general laws of learning - habit-
formation by imitation and reinforcement.
▸ B. F. Skinner (1957), Verbal Behavior
sought to explain the productivity and pragmatic appropriateness
of language

HSL747/LCMA-1/IITDelhi I.Guha 16
Behaviorist Theories of Language Learning

▸ Before 1950s/60s
▸ Behaviorism
▸ emphasizes behavior over mind or cognitive processes
▸ describes language as a behavior in terms of notions such as stim-
ulus, response, reinforcement and punishment
▸ Language learning is governed by general laws of learning - habit-
formation by imitation and reinforcement.
▸ B. F. Skinner (1957), Verbal Behavior
sought to explain the productivity and pragmatic appropriateness
of language
▸ Verbal behavior is shaped by the general learning mechanism of
Operant Conditioning - by differentially reinforcing or punishing
verbal output.

HSL747/LCMA-1/IITDelhi I.Guha 16
Behaviorist Theories of Language Learning

▸ Before 1950s/60s
▸ Behaviorism
▸ emphasizes behavior over mind or cognitive processes
▸ describes language as a behavior in terms of notions such as stim-
ulus, response, reinforcement and punishment
▸ Language learning is governed by general laws of learning - habit-
formation by imitation and reinforcement.
▸ B. F. Skinner (1957), Verbal Behavior
sought to explain the productivity and pragmatic appropriateness
of language
▸ Verbal behavior is shaped by the general learning mechanism of
Operant Conditioning - by differentially reinforcing or punishing
verbal output.

What is Operant Conditioning?

HSL747/LCMA-1/IITDelhi I.Guha 16
A Review of B. F. Skinner’s Verbal Behavior

Chomsky (1959)

HSL747/LCMA-1/IITDelhi I.Guha 17
A Review of B. F. Skinner’s Verbal Behavior

Chomsky (1959)
▸ Operant Conditioning cannot account for productivity

HSL747/LCMA-1/IITDelhi I.Guha 17
A Review of B. F. Skinner’s Verbal Behavior

Chomsky (1959)
▸ Operant Conditioning cannot account for productivity
▸ The linguistic environment of a child provides insufficient evidence
for the grammar of the language spoken around them.

HSL747/LCMA-1/IITDelhi I.Guha 17
A Review of B. F. Skinner’s Verbal Behavior

Chomsky (1959)
▸ Operant Conditioning cannot account for productivity
▸ The linguistic environment of a child provides insufficient evidence
for the grammar of the language spoken around them.
▸ insufficient evidence: unsystematic & weak

HSL747/LCMA-1/IITDelhi I.Guha 17
A Review of B. F. Skinner’s Verbal Behavior

Chomsky (1959)
▸ Operant Conditioning cannot account for productivity
▸ The linguistic environment of a child provides insufficient evidence
for the grammar of the language spoken around them.
▸ insufficient evidence: unsystematic & weak
▸ unsystematic
no negative evidence - no evidence that identifies ungrammatical
verbal input as ungrammatical

HSL747/LCMA-1/IITDelhi I.Guha 17
A Review of B. F. Skinner’s Verbal Behavior

Chomsky (1959)
▸ Operant Conditioning cannot account for productivity
▸ The linguistic environment of a child provides insufficient evidence
for the grammar of the language spoken around them.
▸ insufficient evidence: unsystematic & weak
▸ unsystematic
no negative evidence - no evidence that identifies ungrammatical
verbal input as ungrammatical
Can the absence of grammatical form be the evidence of ungram-
maticality?

HSL747/LCMA-1/IITDelhi I.Guha 17
A Review of B. F. Skinner’s Verbal Behavior

Chomsky (1959)
▸ Operant Conditioning cannot account for productivity
▸ The linguistic environment of a child provides insufficient evidence
for the grammar of the language spoken around them.
▸ insufficient evidence: unsystematic & weak
▸ unsystematic
no negative evidence - no evidence that identifies ungrammatical
verbal input as ungrammatical
Can the absence of grammatical form be the evidence of ungram-
maticality?
▸ weak
the positive evidence - evidence that identifies grammatical verbal
output as grammatical - is of low quality

HSL747/LCMA-1/IITDelhi I.Guha 17
A Review of B. F. Skinner’s Verbal Behavior

Chomsky (1959)
▸ Operant Conditioning cannot account for productivity
▸ The linguistic environment of a child provides insufficient evidence
for the grammar of the language spoken around them.
▸ insufficient evidence: unsystematic & weak
▸ unsystematic
no negative evidence - no evidence that identifies ungrammatical
verbal input as ungrammatical
Can the absence of grammatical form be the evidence of ungram-
maticality?
▸ weak
the positive evidence - evidence that identifies grammatical verbal
output as grammatical - is of low quality
▸ Despite insufficient stimulus they get from their verbal environ-
ment, children typically acquire their native language quite rapidly.

HSL747/LCMA-1/IITDelhi I.Guha 17
Nativist Theories of Language Learning
Principles and Parameters Theory 01/08/2023

HSL747/LCMA-1/IITDelhi I.Guha 18
Nativist Theories of Language Learning
Principles and Parameters Theory 01/08/2023

▸ Poverty of stimulus

HSL747/LCMA-1/IITDelhi I.Guha 18
Nativist Theories of Language Learning
Principles and Parameters Theory 01/08/2023

▸ Poverty of stimulus
▸ Innate syntactic knowledge and language specific learning proce-
dures.

HSL747/LCMA-1/IITDelhi I.Guha 18
Nativist Theories of Language Learning
Principles and Parameters Theory 01/08/2023

▸ Poverty of stimulus
▸ Innate syntactic knowledge and language specific learning proce-
dures.
▸ Universal Grammar: An innate genetic endowment for language,
consisting of Principles and Parameters

HSL747/LCMA-1/IITDelhi I.Guha 18
Nativist Theories of Language Learning
Principles and Parameters Theory 01/08/2023

▸ Poverty of stimulus
▸ Innate syntactic knowledge and language specific learning proce-
dures.
▸ Universal Grammar: An innate genetic endowment for language,
consisting of Principles and Parameters
▸ Core aspects of Universal Grammar and certain learning proce-
dures are unique to language and not shared with other cognitive
systems.

HSL747/LCMA-1/IITDelhi I.Guha 18
Principles and Parameters Theory
Chomsky 1981

HSL747/LCMA-1/IITDelhi I.Guha 19
Principles and Parameters Theory
Chomsky 1981

▸ Principles:
(i) common to all languages, (ii) do not depend on the specific
linguistic inputs received by the child
▸ Projection Principle, Structure Preservation Principle, Subjacency
Principle, Empty Category Principle, etc.

HSL747/LCMA-1/IITDelhi I.Guha 19
Principles and Parameters Theory
Chomsky 1981

▸ Principles:
(i) common to all languages, (ii) do not depend on the specific
linguistic inputs received by the child
▸ Projection Principle, Structure Preservation Principle, Subjacency
Principle, Empty Category Principle, etc.
▸ Parameters:
(i) specific linguistic features along which languages vary, (ii) these
vary according to the child’s linguistic input, (iii) setting a single
parameter can cause a cluster of apparently unrelated grammatical
features to appear in the language
▸ Head-directionality Parameter, Null-subject Parameter, Pro-drop
Parameter, etc.

HSL747/LCMA-1/IITDelhi I.Guha 19
Social-Cognitive Theory

▸ Core aspects of grammar are unique to language but learning


mechanisms are shared with other cognitive systems

HSL747/LCMA-1/IITDelhi I.Guha 20
Connectionist Models of Language Learning
Parallel Distributed Processing (PDP)

HSL747/LCMA-1/IITDelhi I.Guha 21
Connectionist Models of Language Learning
Parallel Distributed Processing (PDP)

▸ A model of the mind that closely resembles the brain.

HSL747/LCMA-1/IITDelhi I.Guha 21
Connectionist Models of Language Learning
Parallel Distributed Processing (PDP)

▸ A model of the mind that closely resembles the brain.

▸ Language, like all types of knowledge, is represented as patterns


of numerical activity across large sets of simple units (neurons).

HSL747/LCMA-1/IITDelhi I.Guha 21
Connectionist Models of Language Learning
Parallel Distributed Processing (PDP)

▸ A model of the mind that closely resembles the brain.

▸ Language, like all types of knowledge, is represented as patterns


of numerical activity across large sets of simple units (neurons).

▸ Learning occurs by adjusting the relative strengths of the connec-


tions in the network in response to external inputs.

HSL747/LCMA-1/IITDelhi I.Guha 21
Connectionist Models of Language Learning
Parallel Distributed Processing (PDP)

▸ A model of the mind that closely resembles the brain.

▸ Language, like all types of knowledge, is represented as patterns


of numerical activity across large sets of simple units (neurons).

▸ Learning occurs by adjusting the relative strengths of the connec-


tions in the network in response to external inputs.

PDP: McClelland & Rumelhart (1986), Joannise & McClelland (2015)

HSL747/LCMA-1/IITDelhi I.Guha 21
Two Types of Theories of Language Learning

▸ Behaviorist
▸ Connectionist

HSL747/LCMA-1/IITDelhi I.Guha 22
Two Types of Theories of Language Learning

▸ Behaviorist ▸ Nativist
▸ Connectionist ▸ Social-Cognitive

HSL747/LCMA-1/IITDelhi I.Guha 22
Two Types of Theories of Language Learning

▸ Domain-general
▸ Behaviorist ▸ Nativist
▸ Connectionist ▸ Social-Cognitive

HSL747/LCMA-1/IITDelhi I.Guha 22
Two Types of Theories of Language Learning

▸ Domain-general ▸ Domain-specific
▸ Behaviorist ▸ Nativist
▸ Connectionist ▸ Social-Cognitive

HSL747/LCMA-1/IITDelhi I.Guha 22
Two Types of Theories of Language Learning

▸ Domain-general ▸ Domain-specific
▸ Behaviorist ▸ Nativist
▸ Connectionist ▸ Social-Cognitive (mixed)

HSL747/LCMA-1/IITDelhi I.Guha 22
Two Types of Theories of Language Learning

▸ Domain-general ▸ Domain-specific
▸ Behaviorist ▸ Nativist
▸ Connectionist ▸ Social-Cognitive (mixed)
▸ No Rules

HSL747/LCMA-1/IITDelhi I.Guha 22
Two Types of Theories of Language Learning

▸ Domain-general ▸ Domain-specific
▸ Behaviorist ▸ Nativist
▸ Connectionist ▸ Social-Cognitive (mixed)
▸ No Rules ▸ Rules

HSL747/LCMA-1/IITDelhi I.Guha 22
Two Types of Theories of Language Learning

▸ Domain-general ▸ Domain-specific
▸ Behaviorist ▸ Nativist
▸ Connectionist ▸ Social-Cognitive (mixed)
▸ No Rules ▸ Rules

Reading: Domain-general and Domain-specific theories of language learning

HSL747/LCMA-1/IITDelhi I.Guha 22
English Past Tense

HSL747/LCMA-1/IITDelhi I.Guha 23
English Past Tense

(i) Regular past tense: Most (87%) English past tense verbs:
V + -ed

HSL747/LCMA-1/IITDelhi I.Guha 23
English Past Tense

(i) Regular past tense: Most (87%) English past tense verbs:
V + -ed
▸ walk - walked, blend - blended, chill - chilled, etc.

HSL747/LCMA-1/IITDelhi I.Guha 23
English Past Tense

(i) Regular past tense: Most (87%) English past tense verbs:
V + -ed
▸ walk - walked, blend - blended, chill - chilled, etc.
▸ neologisms: google - googled, LOL - LOL’d

HSL747/LCMA-1/IITDelhi I.Guha 23
English Past Tense

(i) Regular past tense: Most (87%) English past tense verbs:
V + -ed
▸ walk - walked, blend - blended, chill - chilled, etc.
▸ neologisms: google - googled, LOL - LOL’d

(ii) Irregular past tense: A number of English verbs: *V-ed

HSL747/LCMA-1/IITDelhi I.Guha 23
English Past Tense

(i) Regular past tense: Most (87%) English past tense verbs:
V + -ed
▸ walk - walked, blend - blended, chill - chilled, etc.
▸ neologisms: google - googled, LOL - LOL’d

(ii) Irregular past tense: A number of English verbs: *V-ed


▸ go - *goed - went
sleep - *sleeped -slept
take - *taked - took
bring - *bringed - brought

HSL747/LCMA-1/IITDelhi I.Guha 23
English Past Tense

(i) Regular past tense: Most (87%) English past tense verbs:
V + -ed
▸ walk - walked, blend - blended, chill - chilled, etc.
▸ neologisms: google - googled, LOL - LOL’d

(ii) Irregular past tense: A number of English verbs: *V-ed


▸ go - *goed - went
sleep - *sleeped -slept
take - *taked - took
bring - *bringed - brought
▸ Universal Grammar: The learning of the regular and the irregular
past tense forms involve distinct mechanisms.

HSL747/LCMA-1/IITDelhi I.Guha 23
English Past Tense

(i) Regular past tense: Most (87%) English past tense verbs:
V + -ed
▸ walk - walked, blend - blended, chill - chilled, etc.
▸ neologisms: google - googled, LOL - LOL’d

(ii) Irregular past tense: A number of English verbs: *V-ed


▸ go - *goed - went
sleep - *sleeped -slept
take - *taked - took
bring - *bringed - brought
▸ Universal Grammar: The learning of the regular and the irregular
past tense forms involve distinct mechanisms.
▸ Connectionist Models: The same mechanism can account for the
learning of both regular and irregular past tense forms.

HSL747/LCMA-1/IITDelhi I.Guha 23
Learning English Past Tense

Universal Grammar Connectionist Models

(i)Regular ; application of Rules


Both Regular and Irregular
(ii)Irregular ; memorization
forms involve learning by
Reading: Box5.4, pp.77-79,
analogy or pattern association
Language in Mind

HSL747/LCMA-1/IITDelhi I.Guha 24
Learning English Past Tense

Universal Grammar Connectionist Models

(i)Regular ; application of Rules


Both Regular and Irregular
(ii)Irregular ; memorization
forms involve learning by
Reading: Box5.4, pp.77-79,
analogy or pattern association
Language in Mind
▸ children:

HSL747/LCMA-1/IITDelhi I.Guha 24
Learning English Past Tense

Universal Grammar Connectionist Models

(i)Regular ; application of Rules


Both Regular and Irregular
(ii)Irregular ; memorization
forms involve learning by
Reading: Box5.4, pp.77-79,
analogy or pattern association
Language in Mind
▸ children:
▸ overgeneralization of regular: goed, bringed, holded, weared, etc. -
rules

HSL747/LCMA-1/IITDelhi I.Guha 24
Learning English Past Tense

Universal Grammar Connectionist Models

(i)Regular ; application of Rules


Both Regular and Irregular
(ii)Irregular ; memorization
forms involve learning by
Reading: Box5.4, pp.77-79,
analogy or pattern association
Language in Mind
▸ children:
▸ overgeneralization of regular: goed, bringed, holded, weared, etc. -
rules
▸ overextension of irregular: brang (bring), bat (bit) - analogy

HSL747/LCMA-1/IITDelhi I.Guha 24
Learning English Past Tense

Universal Grammar Connectionist Models

(i)Regular ; application of Rules


Both Regular and Irregular
(ii)Irregular ; memorization
forms involve learning by
Reading: Box5.4, pp.77-79,
analogy or pattern association
Language in Mind
▸ children:
▸ overgeneralization of regular: goed, bringed, holded, weared, etc. -
rules
▸ overextension of irregular: brang (bring), bat (bit) - analogy
▸ regularity in irregular forms:
lie-lay, give-gave, forgive-forgave, bid-bade;
fling-flung, sling-slung, sting-stung, hang-hung;
blow-blew, grow-grew, know-knew, throw-threw;
find-found, bind-bound, grind-ground, wind-wound
- analogy/pattern association

HSL747/LCMA-1/IITDelhi I.Guha 24
English Past Tense

HSL747/LCMA-1/IITDelhi I.Guha 25
English Past Tense
▸ Rule:
(i) Vstem + -ed = VPast
e.g.: slip + -ed = slipped
Needs decomposition of words into word-parts

HSL747/LCMA-1/IITDelhi I.Guha 25
English Past Tense
▸ Rule:
(i) Vstem + -ed = VPast
e.g.: slip + -ed = slipped
Needs decomposition of words into word-parts
▸ Analogy:
(i) Notice the difference in meaning between slip and slipped
(ii) Notice the difference in sounds between slip and slipped
(iii) Apply the sound difference to flip to yield the similar difference
in meaning.
Does not need decomposition of words into word-parts

HSL747/LCMA-1/IITDelhi I.Guha 25
English Past Tense
▸ Rule:
(i) Vstem + -ed = VPast
e.g.: slip + -ed = slipped
Needs decomposition of words into word-parts
▸ Analogy:
(i) Notice the difference in meaning between slip and slipped
(ii) Notice the difference in sounds between slip and slipped
(iii) Apply the sound difference to flip to yield the similar difference
in meaning.
Does not need decomposition of words into word-parts
▸ Pattern Association:
(i) Learn the sound patterns in pairs like slip-slipped or bind-bound.
(ii) Make predictions for new inputs like flip or grind based on their
sequence of sounds and assuming they follow regular statistical
patterns.
Does not need decomposition of words into word-parts

HSL747/LCMA-1/IITDelhi I.Guha 25
Wug test
Jean Berko Gleason (1958)

HSL747/LCMA-1/IITDelhi I.Guha 26
Wug test
Jean Berko Gleason (1958)

HSL747/LCMA-1/IITDelhi I.Guha 26
Wug test
Jean Berko Gleason (1958)

HSL747/LCMA-1/IITDelhi I.Guha 26
Wug Test
Jean Berko Gleason (1958)

HSL747/LCMA-1/IITDelhi I.Guha 27
Wug Test
Jean Berko Gleason (1958)

▸ What is this Wug Test testing?

HSL747/LCMA-1/IITDelhi I.Guha 27
Wug Test
Jean Berko Gleason (1958)

▸ What is this Wug Test testing?


It is testing pluralization in English.

HSL747/LCMA-1/IITDelhi I.Guha 27
Wug Test
Jean Berko Gleason (1958)

▸ What is this Wug Test testing?


It is testing pluralization in English.
In particular: It was designed to test whether the children have
memorized particular plural forms or whether they know the plu-
ralization rule.

HSL747/LCMA-1/IITDelhi I.Guha 27
Wug Test
Jean Berko Gleason (1958)

▸ What is this Wug Test testing?


It is testing pluralization in English.
In particular: It was designed to test whether the children have
memorized particular plural forms or whether they know the plu-
ralization rule.
▸ Based on what you have seen for the English Past tense, identify
the different kinds of data we need to consider for Pluralization.

HSL747/LCMA-1/IITDelhi I.Guha 27
Wug Test
Jean Berko Gleason (1958)

▸ What is this Wug Test testing?


It is testing pluralization in English.
In particular: It was designed to test whether the children have
memorized particular plural forms or whether they know the plu-
ralization rule.
▸ Based on what you have seen for the English Past tense, identify
the different kinds of data we need to consider for Pluralization.
▸ How would the different kinds of theories of learning account for
pluralization?

HSL747/LCMA-1/IITDelhi I.Guha 27
Wug Test
Jean Berko Gleason (1958)

▸ What is this Wug Test testing?


It is testing pluralization in English.
In particular: It was designed to test whether the children have
memorized particular plural forms or whether they know the plu-
ralization rule.
▸ Based on what you have seen for the English Past tense, identify
the different kinds of data we need to consider for Pluralization.
▸ How would the different kinds of theories of learning account for
pluralization?
▸ Can you think any other kinds of datasets (English or any lan-
guage) where the same debate can arise?

HSL747/LCMA-1/IITDelhi I.Guha 27
Structure Preservation
A Principle & Its Testing

▸ Formation of Yes/No Questions from Declaratives

Declaratives Yes/No Questions

The girl is tall. Is the girl tall?


There was a book on the table. Was there a book on the table?
She can go. Can she go?
Rumi has got pets. Has Rumi got pets?
Rumi will be working then. Will Rumi be working then?
It was raining. Was it raining?
▸ Hypothesis I:
(i) Identify the first verb starting from the left of the sentence.
(ii)Put this verb in the left-most position of the sentence.

HSL747/LCMA-1/IITDelhi I.Guha 28
Structure Preservation

▸ Hypothesis I will fail if we try to apply it on a sentence with a


Subject Relative Clause.
Declaratives Yes/No Questions

The girl [who is tall] is standing Is the girl [who is tall] standing
at the end of the line. at end of the line?
▸ Hypothesis II:
(i) Identify the auxiliary verb of the main clause of a sentence.
(ii)Swap the position between the subject of the main clause and
this auxiliary verb.

HSL747/LCMA-1/IITDelhi I.Guha 29
Structure Preservation Principle

▸ Formation of Yes/No Questions requires analyzing the sentence


into different levels of structure: main clause, subordinate
clause, etc.
▸ Transformations can singularly affect one part of the structure
(main clause) while bypassing over other parts (relative clause).
Thus syntactic transformations are structure preserving.

HSL747/LCMA-1/IITDelhi I.Guha 30
Affix Hopping

▸ What about sentences without auxiliary verbs?


Declaratives Yes/No Questions

Ravi liked cupcakes. Did Ravi like cupcakes?


Mili works in the lab. Does Mili work in the lab?
Do Mira and Vivek attend
Mira and Vivek attend the talks.
the talks?
▸ Notice how the Tense+Number marking travels!
▸ Notice sentences with the subject relative:

(1) a. The guy who joined the PhD program works in the lab.
b. Does the guy who joined the PhD program work in the lab?
(2) a. The guys who joined the PhD program work in the lab.
b. Do the guys who joined the PhD program work in the lab?

HSL747/LCMA-1/IITDelhi I.Guha 31
Crain & Nakayama (1987)

The objective was to verify if children form a structure-independent


Hypothesis I, while learning English Yes/No questions.
▸ Participants: Group I: mean age 4;3 (4 years 3 months), Group II:
mean age 5;3.
▸ Game-like Experimental setup: two experimenters, one child (par-
ticipant), one doll (named Jabba), a set of pictures. One experi-
menter manipulated the doll Jabba. The other experimenter asked
the child embedded questions related to the pictures.
▸ Ask Jabba if the boy who is watching Mickey Mouse is happy.
▸ The child had to ask Jabba a Yes/No question, which Jabba an-
swered with ‘Yes’ or ‘No’. If Jabba answered correctly the child
gave Jabba a treat.

HSL747/LCMA-1/IITDelhi I.Guha 32
Crain & Nakayama (1987)
Materials: Exp 1

HSL747/LCMA-1/IITDelhi I.Guha 33
Crain & Nakayama (1987)
Materials: Exp 1

(3) Practice
a. The girl is tall.
b. The man is tired.
c. The pig next to the tree is tall.

HSL747/LCMA-1/IITDelhi I.Guha 33
Crain & Nakayama (1987)
Materials: Exp 1

(3) Practice
a. The girl is tall.
b. The man is tired.
c. The pig next to the tree is tall.
(4) Test
a. The dog that is sleeping is on the blue bench.
b. The ball that the girl is sitting on is big.
c. The boy who is watching Mickey Mouse is happy.
d. The boy who is unhappy is watching Mickey Mouse.
e. The boy who is being kissed by his mother is happy.
f. The boy who was holding the plate is crying.

HSL747/LCMA-1/IITDelhi I.Guha 33
Crain & Nakayama (1987)
Materials: Exp 1

(3) Practice
a. The girl is tall.
b. The man is tired.
c. The pig next to the tree is tall.
(4) Test
a. The dog that is sleeping is on the blue bench.
b. The ball that the girl is sitting on is big.
c. The boy who is watching Mickey Mouse is happy.
d. The boy who is unhappy is watching Mickey Mouse.
e. The boy who is being kissed by his mother is happy.
f. The boy who was holding the plate is crying.
(5) Prospective Responses
a. Is the dog that sleeping is on the blue bench? (S-I)
b. Is the dog that is sleeping on the blue bench? (S-D)

HSL747/LCMA-1/IITDelhi I.Guha 33
Crain & Nakayama (1987)
Results: Exp 1

▸ Errors : 40%, Group I: 62%, Group II: 20%

▸ Error Types:
▸ Type I ‘Prefix’ errors: Is the dog that is sleeping is on the blue
bench? (58%)
▸ Type II ‘restarting’: Is the boy that is watching Mickey Mouse, is
he happy? (22%)
▸ Type III: S-I errors (3a). (None)

HSL747/LCMA-1/IITDelhi I.Guha 34
Crain & Nakayama (1987)
Analyzing Type I Errors

(6) Is the boy who is watching Mickey Mouse is happy?

▸ structure independent
▸ not structure independent
Move = copy + delete
(7) a. Did you came home?
b. What did you got?
c. Could you get me an apple for me?
d. Whose is that is?
In (6) the children have copied is but they haven’t deleted it.
▸ Is the copying in (6) S-D or S-I?

HSL747/LCMA-1/IITDelhi I.Guha 35
Crain & Nakayama (1987)
Exp 2

HSL747/LCMA-1/IITDelhi I.Guha 36
Crain & Nakayama (1987)
Exp 2

▸ It was not clear whether is in (6) copied from the main clause or
the relative clause.

HSL747/LCMA-1/IITDelhi I.Guha 36
Crain & Nakayama (1987)
Exp 2

▸ It was not clear whether is in (6) copied from the main clause or
the relative clause.
▸ How can you get rid off this confusion?

HSL747/LCMA-1/IITDelhi I.Guha 36
Crain & Nakayama (1987)
Exp 2

▸ It was not clear whether is in (6) copied from the main clause or
the relative clause.
▸ How can you get rid off this confusion? By using different auxiliary
verbs in the two clauses.

HSL747/LCMA-1/IITDelhi I.Guha 36
Crain & Nakayama (1987)
Exp 2

▸ It was not clear whether is in (6) copied from the main clause or
the relative clause.
▸ How can you get rid off this confusion? By using different auxiliary
verbs in the two clauses.
Exp 2 Materials:
(8) a. The boy who is happy can see Mickey Mouse.
b. The boy who can see Mickey Mouse is happy.

HSL747/LCMA-1/IITDelhi I.Guha 36
Crain & Nakayama (1987)
Exp 2

▸ It was not clear whether is in (6) copied from the main clause or
the relative clause.
▸ How can you get rid off this confusion? By using different auxiliary
verbs in the two clauses.
Exp 2 Materials:
(8) a. The boy who is happy can see Mickey Mouse.
b. The boy who can see Mickey Mouse is happy.
Prospective errors:
(9) a. Is the boy who is happy can see Mickey Mouse? (Is-Can)

HSL747/LCMA-1/IITDelhi I.Guha 36
Crain & Nakayama (1987)
Exp 2

▸ It was not clear whether is in (6) copied from the main clause or
the relative clause.
▸ How can you get rid off this confusion? By using different auxiliary
verbs in the two clauses.
Exp 2 Materials:
(8) a. The boy who is happy can see Mickey Mouse.
b. The boy who can see Mickey Mouse is happy.
Prospective errors:
(9) a. Is the boy who is happy can see Mickey Mouse? (Is-Can) ;
S-I (copy of is) or S-D (insertion of is as Question-particle)

HSL747/LCMA-1/IITDelhi I.Guha 36
Crain & Nakayama (1987)
Exp 2

▸ It was not clear whether is in (6) copied from the main clause or
the relative clause.
▸ How can you get rid off this confusion? By using different auxiliary
verbs in the two clauses.
Exp 2 Materials:
(8) a. The boy who is happy can see Mickey Mouse.
b. The boy who can see Mickey Mouse is happy.
Prospective errors:
(9) a. Is the boy who is happy can see Mickey Mouse? (Is-Can) ;
S-I (copy of is) or S-D (insertion of is as Question-particle)
b. Can the boy who see Mickey Mouse is happy? (Can-Is)

HSL747/LCMA-1/IITDelhi I.Guha 36
Crain & Nakayama (1987)
Exp 2

▸ It was not clear whether is in (6) copied from the main clause or
the relative clause.
▸ How can you get rid off this confusion? By using different auxiliary
verbs in the two clauses.
Exp 2 Materials:
(8) a. The boy who is happy can see Mickey Mouse.
b. The boy who can see Mickey Mouse is happy.
Prospective errors:
(9) a. Is the boy who is happy can see Mickey Mouse? (Is-Can) ;
S-I (copy of is) or S-D (insertion of is as Question-particle)
b. Can the boy who see Mickey Mouse is happy? (Can-Is) (None)

▸ By negative evidence - absence of Can-Is errors - they concluded


structure dependency.
HSL747/LCMA-1/IITDelhi I.Guha 36
Some follow up studies

Ambridge et al 2010 argued against Crain & Nakayama’s conclusion


based on the results of the experiments they conducted.

HSL747/LCMA-1/IITDelhi I.Guha 37
Some follow up studies

Ambridge et al 2010 argued against Crain & Nakayama’s conclusion


based on the results of the experiments they conducted.

▸ Make presentations on Ambridge et al.’s experiments or on Lewis


& Elman 2001, Reali & Christiansen 2005

HSL747/LCMA-1/IITDelhi I.Guha 37
Some follow up studies

Ambridge et al 2010 argued against Crain & Nakayama’s conclusion


based on the results of the experiments they conducted.

▸ Make presentations on Ambridge et al.’s experiments or on Lewis


& Elman 2001, Reali & Christiansen 2005
▸ Chomsky himself has co-authored papers arguing against Lewis
& Elman or Reali & Christiansen. (Berwick, Pietrosky, Yankama
& Chomsky 2011) It seems we have not yet seen the end of this
debate.

HSL747/LCMA-1/IITDelhi I.Guha 37
Principles of Universal Grammar

‘Universal grammar may be thought of as some system of principles,


common to the species and available to each individual prior to expe-
rience.’ ∼ Chomsky (1981)

Universal grammar (UG) is the basis for acquiring language. It un-


derlies all human languages. All and only human beings are equipped
with UG and they are all able to learn languages. Other systems…(non-
human beings or artificially built systems) …are not equipped with UG
and therefore will not be able to learn human languages. The linguistic
endowment characterized as UG is species-specific. (Haegeman 1994,
Introduction to GB Theory, p.13)

HSL747/LCMA-1/IITDelhi I.Guha 38
Parameters

▸ Can we speak any language whenever we want?

HSL747/LCMA-1/IITDelhi I.Guha 39
Parameters

▸ Can we speak any language whenever we want? Evidently, no.

HSL747/LCMA-1/IITDelhi I.Guha 39
Parameters

▸ Can we speak any language whenever we want? Evidently, no.


Why?
▸ Languages differ in important respects: Word Order, Null Sub-
jects, Expletive subjects, Overt Wh-movement etc.
(10) Word Order / Head-directionality / VO - OV
a. [Ravi]S [saw]V [Venu]O . English
b. [Ravi-ne]S [Venu-ko]O [dekha]V Hindi
Ravi-erg Venu-dat saw.Msg
‘Ravi saw Venu.’

HSL747/LCMA-1/IITDelhi I.Guha 39
Parameters

Tensed sentences in English must have overt subjects. But in many


languages of the world, e.g. in all Indian languages, in Hebrew, Italian,
Japanese, Korean, Mandarin, Persian, Spanish, Turkish, etc., there is
no such requirement.
(11) Null Subjects
a. What happened?
b. *(She) doesn’t want to pay the taxes. English
c. (o) tax dite caichhe na Bangla
(Pron.3sg) tax give.inf want.prog.3 neg
‘She is not willing to pay the taxes.’

*( x ) = x cannot be dropped / it is obligatory in this position


( x ) = x can be dropped / it is optional in this position

HSL747/LCMA-1/IITDelhi I.Guha 40
Expletive Subjects

Languages that do not allow null-subjects (English, French, Germanic


languages like Dutch, Danish, German, Icelandic, Norwegian, Swiddish,
Yiddish) place a ‘dummy’ pronoun in the subject position of tensed
clauses with certain verbs like exit/be, rain, seem, be likely, etc. that do
not have thematic/ referential subjects. These placeholder or pleonastic
elements are called ‘expletive’ pronouns.
English Expletive pronouns: there, it (these are different from referen-
tial deictic pronouns there or it.)
(12) a. There is a white board right in front of me. [exist/be]
b. It is raining. [rain]
c. It seems to me that they are very aggressive. [seem]

HSL747/LCMA-1/IITDelhi I.Guha 41
Parameters

In some languages (English, German, Bulgarian, Dutch, etc.) to form


wh-questions, the wh-word has to be moved to front of the sentence.
However, in many languages of the world (Indian languages) the wh-
word occurs in the same position as the ordinary, non-question words.
(13) English
a. Ravi saw Venu.
b. Who did Ravi see _?
(14) Hindi
a. Ravi-ne Venu-ko dekha
b. Ravi-ne kis-ko dekha ?

HSL747/LCMA-1/IITDelhi I.Guha 42
Parameters

HSL747/LCMA-1/IITDelhi I.Guha 43
Parameters

▸ Parameters are language-specific properties which are not fully de-


termined by Universal Grammar, and these vary cross-linguistically.

HSL747/LCMA-1/IITDelhi I.Guha 43
Parameters

▸ Parameters are language-specific properties which are not fully de-


termined by Universal Grammar, and these vary cross-linguistically.
▸ Parameters are like switches - properties with (usually) a binary
choice.

HSL747/LCMA-1/IITDelhi I.Guha 43
Parameters

▸ Parameters are language-specific properties which are not fully de-


termined by Universal Grammar, and these vary cross-linguistically.
▸ Parameters are like switches - properties with (usually) a binary
choice.
▸ Children are born with the parametric properties of languages set
at a default or unmarked value.

HSL747/LCMA-1/IITDelhi I.Guha 43
Parameters

▸ Parameters are language-specific properties which are not fully de-


termined by Universal Grammar, and these vary cross-linguistically.
▸ Parameters are like switches - properties with (usually) a binary
choice.
▸ Children are born with the parametric properties of languages set
at a default or unmarked value.
▸ Exposure to actual input from language X sets the value of a pa-
rameter according to the value in X, which helps acquire the gram-
mar of X.

HSL747/LCMA-1/IITDelhi I.Guha 43
Simultaneous setting of multiple parameters

Hyams (1986) proposed that setting a single parameter can cause a


cluster of apparently unrelated grammatical features to appear in the
language. Thus, turning the Null Subject parameter ‘off’ can lead to
the availability of Expletive Subjects, Subject-Aux Inversion, and un-
availability of Pro-drop (having null subjects and objects in a language).
Note: Crain & Nakayama concludes against with this position based
on their Exp 3.

HSL747/LCMA-1/IITDelhi I.Guha 44
Principles and Parameters Theory
Chomsky 1981

Each individual is born with a genetic endowment for language, termed


Universal Grammar, which is an innate system of computational knowl-
edge.

HSL747/LCMA-1/IITDelhi I.Guha 45
The Biolinguistic Program

HSL747/LCMA-1/IITDelhi I.Guha 46
The Biolinguistic Program

▸ seeks to unravel what exactly the genetic endowment for language


(aka. Universal Grammar) is…

HSL747/LCMA-1/IITDelhi I.Guha 46
The Biolinguistic Program

▸ seeks to unravel what exactly the genetic endowment for language


(aka. Universal Grammar) is…
…from the perspectives of Linguistics, Biology and Mathematics

HSL747/LCMA-1/IITDelhi I.Guha 46
The Biolinguistic Program

▸ seeks to unravel what exactly the genetic endowment for language


(aka. Universal Grammar) is…
…from the perspectives of Linguistics, Biology and Mathematics
▸ It seeks to answer questions like:
(i) What are the components of the faculty of language (FL)?
(ii) Which components of FL are uniquely human?
(iii) Which components of FL are uniquely adapted for language?

HSL747/LCMA-1/IITDelhi I.Guha 46
Recursion

The idea since P&P theory ;

HSL747/LCMA-1/IITDelhi I.Guha 47
Recursion

The idea since P&P theory ;


▸ What is the core computational property that underlies the lin-
guistic capacity of humans (or FL)?

HSL747/LCMA-1/IITDelhi I.Guha 47
Recursion

The idea since P&P theory ;


▸ What is the core computational property that underlies the lin-
guistic capacity of humans (or FL)?
▸ Recursion
A recursive system can take as input the output it generates. This
leads to the capacity of discrete infinity (what we have known as
productivity in this course).

HSL747/LCMA-1/IITDelhi I.Guha 47
Recursion

The idea since P&P theory ;


▸ What is the core computational property that underlies the lin-
guistic capacity of humans (or FL)?
▸ Recursion
A recursive system can take as input the output it generates. This
leads to the capacity of discrete infinity (what we have known as
productivity in this course).
▸ One can add linguistic units indefinitely to form sentences of ar-
bitrary length, recall the fallacy of ‘the longest sentence’ discussed
at the beginning of the course.

HSL747/LCMA-1/IITDelhi I.Guha 47
Recursion

The idea since P&P theory ;


▸ What is the core computational property that underlies the lin-
guistic capacity of humans (or FL)?
▸ Recursion
A recursive system can take as input the output it generates. This
leads to the capacity of discrete infinity (what we have known as
productivity in this course).
▸ One can add linguistic units indefinitely to form sentences of ar-
bitrary length, recall the fallacy of ‘the longest sentence’ discussed
at the beginning of the course.
▸ Take a sentence S, we can add a Subject and a Verb on top of it
and make it a complex sentence S’. These steps can be infinitely
iterated.
Subj Verb + S = S’
Subj Verb + S’ = S”... (think of the actual examples discussed in
class)

HSL747/LCMA-1/IITDelhi I.Guha 47
Externalization & Interpretation

A crucial property of the human linguistic capacity is that it interfaces


with an externalization system as well as an interpretation system
(a) linguistic expressions are externalized by a system of sounds or
signs, and
(b) the meaning of the linguistic expressions are interpreted by a des-
ignated component of the mind

▸ The system for externalization is called the Sensory-Motor system


(SM)
▸ The interpretation component is called the Conceptual-Intentional
system (CI)

HSL747/LCMA-1/IITDelhi I.Guha 48
Components of FL

The Faculty of Language (FL) has three key components:


(i) The computational system
(ii) The Sensory Motor system
(iii) The Conceptual Intentional system

Hauser, Chomsky & Fitch (2002) put forward the hypothesis that:
▸ The computational system is unique to humans and unique to
language.
▸ The sensory motor and the conceptual systems are unique to hu-
mans but not unique to language.
▸ All three components together form the Faculty of language in the
Broad sense (FLB), while the computational system is the core
and unique property of the linguistic system - it is the Faculty of
language in the Narrow sense (FLN)

HSL747/LCMA-1/IITDelhi I.Guha 49
Some Questions

HSL747/LCMA-1/IITDelhi I.Guha 50
Some Questions

▸ What is in the computational component or FLN?

HSL747/LCMA-1/IITDelhi I.Guha 50
Some Questions

▸ What is in the computational component or FLN?


▸ Recursion

HSL747/LCMA-1/IITDelhi I.Guha 50
Some Questions

▸ What is in the computational component or FLN?


▸ Recursion
…in particular the operation of putting two syntactic units together
to form a set. The resulting set can be combined with another
syntactic unit to form a new set and this can go on infinitely (this
operation is called ‘Merge’, which is the basic structure building
operation)

HSL747/LCMA-1/IITDelhi I.Guha 50
Some Questions

▸ What is in the computational component or FLN?


▸ Recursion
…in particular the operation of putting two syntactic units together
to form a set. The resulting set can be combined with another
syntactic unit to form a new set and this can go on infinitely (this
operation is called ‘Merge’, which is the basic structure building
operation)
▸ The operation is trying to describe how linguistic expressions or
‘phrases’ are formed.

HSL747/LCMA-1/IITDelhi I.Guha 50
Some Questions

▸ What is in the computational component or FLN?


▸ Recursion
…in particular the operation of putting two syntactic units together
to form a set. The resulting set can be combined with another
syntactic unit to form a new set and this can go on infinitely (this
operation is called ‘Merge’, which is the basic structure building
operation)
▸ The operation is trying to describe how linguistic expressions or
‘phrases’ are formed.
▸ How does it relate to the idea of language acquisition according to
Principles & Parameters theory?

HSL747/LCMA-1/IITDelhi I.Guha 50
Some Questions

▸ What is in the computational component or FLN?


▸ Recursion
…in particular the operation of putting two syntactic units together
to form a set. The resulting set can be combined with another
syntactic unit to form a new set and this can go on infinitely (this
operation is called ‘Merge’, which is the basic structure building
operation)
▸ The operation is trying to describe how linguistic expressions or
‘phrases’ are formed.
▸ How does it relate to the idea of language acquisition according to
Principles & Parameters theory?
▸ The children are born with the knowledge of the basic structure
building operation

HSL747/LCMA-1/IITDelhi I.Guha 50
Some Questions

▸ What is in the computational component or FLN?


▸ Recursion
…in particular the operation of putting two syntactic units together
to form a set. The resulting set can be combined with another
syntactic unit to form a new set and this can go on infinitely (this
operation is called ‘Merge’, which is the basic structure building
operation)
▸ The operation is trying to describe how linguistic expressions or
‘phrases’ are formed.
▸ How does it relate to the idea of language acquisition according to
Principles & Parameters theory?
▸ The children are born with the knowledge of the basic structure
building operation
▸ How does FLN interact with SM or CI systems?

HSL747/LCMA-1/IITDelhi I.Guha 50
Some Questions

▸ What is in the computational component or FLN?


▸ Recursion
…in particular the operation of putting two syntactic units together
to form a set. The resulting set can be combined with another
syntactic unit to form a new set and this can go on infinitely (this
operation is called ‘Merge’, which is the basic structure building
operation)
▸ The operation is trying to describe how linguistic expressions or
‘phrases’ are formed.
▸ How does it relate to the idea of language acquisition according to
Principles & Parameters theory?
▸ The children are born with the knowledge of the basic structure
building operation
▸ How does FLN interact with SM or CI systems?
▸ Via the interfaces PF and LF respectively

HSL747/LCMA-1/IITDelhi I.Guha 50
Unique to humans

HSL747/LCMA-1/IITDelhi I.Guha 51
Unique to humans

Does any other species possess a communication system like the hu-
mans?

HSL747/LCMA-1/IITDelhi I.Guha 51
Unique to humans

Does any other species possess a communication system like the hu-
mans?

There are different ways to answer this question. One way is to define
what the definitional properties of language are and then compare with
non-human communication systems to see how many of those properties
these systems possess.

▸ Design Features of Language:


There are 9 key features that define human language (according
to Charles Hockett)

HSL747/LCMA-1/IITDelhi I.Guha 51
Design Features of Language
1. Mode of communication (transmitted via speech and sign)
2. Semanticity (signals mean or refer to objects, actions, situations etc.)
3. Pragmatic Function (have useful purpose)
4. Interchangeability (everybody can both send and receive messages)
5. Cultural Transmission (learn only by interacting with other humans)
6. Arbitrariness (The relationship between a signal and its meaning is
arbitrary. The word ‘big’ has more syllables than ‘teeny-tiny’ and that
does not tell us anything about their respective meanings.)
7. Discreteness (Linguistic signals are made up of discrete units, changing
which can change the meaning. Consider ‘phonemes’ like /t/, /n/, /i/,
/e/. The word /tin/ contains three of the phonemes mentioned.
Replacing the vowel phoneme /i/ with another vowel /e/, yields /ten/,
which has a completely different meaning from /tin/.)
8. Displacement (we can communicate about things, ideas not present in
time or space)
9. Productivity

HSL747/LCMA-1/IITDelhi I.Guha 52
Language vs. Animal communication systems
▸ All non-human communication systems have the first three fea-
tures, i.e., mode-of-communication, semanticity and pragmatic func-
tion
▸ But no other communication system has all 9 features.

HSL747/LCMA-1/IITDelhi I.Guha 53
Language vs. Animal communication systems
▸ All non-human communication systems have the first three fea-
tures, i.e., mode-of-communication, semanticity and pragmatic func-
tion
▸ But no other communication system has all 9 features.
▸ Productivity/The computational capacity is to some extent exhib-
ited by song birds, gibbons, whales. Their songs (calls for gibbons)
have rule based structure.
▸ Bird songs, which are different from calls, are typically longer and
they are usually seasonal. Bird songs can be structurally very com-
plex. For example, the songs of branded wren have four component
phrases - A, B, C, D, which have been found to occur in various
combinations like the following: (find out the rule)
ABABABABCCCCD
ABABABABABABABCCD
ABCCCD
But the meaning of the song does not change with the complexity
of the song structure.

HSL747/LCMA-1/IITDelhi I.Guha 53
Language vs. Animal communication systems

HSL747/LCMA-1/IITDelhi I.Guha 54
Language vs. Animal communication systems

▸ Vervet monkeys use a variety of alarm calls to signal the presence


of different predators: loud-bark ; leopard, two-part cough like
sound ; eagle, soft whirring sound ; snake. Other than these
they have specific set of calls for territorial dispute, mating rituals,
social hierarchy arguments, etc. They do have a rigid vocabulary,
but their calls have been found to have no computational complex-
ity.

HSL747/LCMA-1/IITDelhi I.Guha 54
Language vs. Animal communication systems

▸ Vervet monkeys use a variety of alarm calls to signal the presence


of different predators: loud-bark ; leopard, two-part cough like
sound ; eagle, soft whirring sound ; snake. Other than these
they have specific set of calls for territorial dispute, mating rituals,
social hierarchy arguments, etc. They do have a rigid vocabulary,
but their calls have been found to have no computational complex-
ity.
▸ Thus no other species seems to have integrated the computational
system with meaning!

listen to this podcast: https://www.bbc.co.uk/programmes/b05tz9jr

HSL747/LCMA-1/IITDelhi I.Guha 54

You might also like