Linguistics final

Syntax

Focuses on the structure between words

Word order

Subjects typically precede verbs, then objects follow

Ex: Sally ate an apple

*Ate Sally an apple

*Sally an apple ate

*Ate an apple Sally

Word order

English word order pattern is SVO (subject word object)

40% of world’s languages have this word order pattern

SOV is more common

English is not a SVO language across the board

Is Sally a student? VSO in yes/no questions

OSV requires special context but it is a possible sentence of English

English adjective order

Determiner: (a, an, the, my, your, numbers)

Opinion: (beautiful, ugly)

Size: (big, small, tall, tiny)

Age: (old, new)

Shape: (round, square)

Color: (red)

Origin: (French,

Material (wooden, silk)

Purpose or Qualifier: (what it’s being used for: sports car, walking shoes)

Noun: thing being described

Drunk Old Sailors Always Sing Country on Muddy Piers, Naked

Syntax is structure, not just order

Language specificity- Englisyh is SVO, Japanese is SOV

Ambiguity: I saw the man with the telescope (does he have the telescope, or were you using the telescope)

Syntactic knowledge is highly productive, never heard some sentences but instantly know they are grammatical

Properties of syntactic rules

Recursive: ability to inbed structures within itself to create infinite possibilities

Based on syntactic categories: syntactic knowledge relates to syntactic categories(parts of speech)
“colorless green ideas sleep furiously” syntactic well-formedness is independent of semantic plausibility

Hierarchical: collections of words behave as a single functional unit

Lexical vs structural ambiguity

Lexical: comes from a single word having more than one meaning

Structural: comes from the way words are grouped or attached in the syntax- multiple possible structures for the same string of words

Phrase structure grammars PSGs

Hierarchical grammars that utilize syntactic categories

X→YZ

-Element on the left arrow is composed of the elements on the right

Proper noun: Ann, Tufts University

pronoun: he, she, they

Mass noun:

-words like water, sand, furniture

-can be preceded by ‘some’ but not ‘a’ or number words

-do not take plural suffix

-semantically: (usually) things that can’t easily be counted, often substances

Nouns (N) are things that must be preceded by a determiner

Count noun:

-words like ‘water bottle’, ‘chair’

-can be preceded by ‘a’ and number words

-take the plural suffix when plural

-semantically (usually) things that can be counted

VP= verb phase

Adverb: quickly, gently

Intransitive verb (ITV): express action but do not require object: sleep, dream, look, followed by adverbs or pp

Transitive verb (TV): see, kick, want, transfer action to direct object

Ditransitive verg (DTV): requires thing being acted upon and recipient, direct object and indirect, ask, wish, perscribe

Sentential verb (SV): requires whole clause as object: believe, think doubt

Prepositional phrase PP

Construct S to terminal node

New approach is go backwards and merge:

X bar theory

Every phrase in every sentence in every language is organized the same way

Pronouns in sign language

Have fixed positions, simply individuals themselves change in location

Using space to denote subject/object

Spatial agreement: the use of abstract space to indicate semantic roles

A directional verb’s starting and ending location agrees with the location of the agent and patient

Spatial verb agreement is avbailable in a subset of signs “directional verbs”

1: Establish local referents

2: Use location in signs to refer to referents

Referents can be assigned locations in signing space by making the sign and indicating a location

Semantic relations

Synonyms- same meaning

Antonyms- opposite meaning

Hypernym: a word whose meaning includes the meaning of a subordinate concept- furniture is a hypernym of chair

Hyponym: a word whose meaning is included or is a subset of the meaning of another concept- chair is hyponym of furniture

Meronym: a word that is part of another word: ‘mast’ is a meronym of ‘sailboat’

holonym: a word that contains another word: ‘sailboat’ is a holonym of ‘mast’

Types of antonyms

Complementary: represent two opposing and mutually exclusive categories, where the presence of one implies the absence of the other: married/ single, on/off

Gradable: represents point on a continuum (something may be neither one nor the other): warm/cold small/big

Reverses: One ‘undoes’ the other: expand/contract, ascend/descend, are like complementary but relate specifically to verbs

Converses: For one to occur, the other must occur as well: lend/borrow, employer/employee

Words with multiple meanings

Polysemy: a word that has several related meanings

-paper=article/material for writing

-chicken=a domesticated bird/meat/coward

Homophony: word that has several unrelated meanings

-light=not heavy/form of illumination

-bank=financial institution/side of a river

-club=social organization/weapon

Reference

One of the things that words do is refer to things in the world

Reference/denotation are the things in the world it refers to

Referent is a specific thing that a word refers to

Reference

Proper nouns have a single referent

-Sunil Kumar, etc.

Common nouns refer to many things

mug, cat, table

Things that do not exist have no referent

unicorn, 4-sided triangles, Disney character

Different words/linguistic expressions can refer to the same thing

Sense

Sense/connotation-mental representation of a linguistic expression’s meaning

-Definition/defining characteristics

-Associations

-Experiences

ex:

book: reference- set of all objects containing bound pages

sense: object collected in libraries, education, knowledge, entertainment

Sense+Reference: An expression can have sense without any referent, the same referent can be associated with different sense

Talmy

Primitive conceptual categories

Manner- describes the specific motion encoded by a verb

Path- describes a physical or abstract trajectory, trajectory or route of a motion event, focusing on where something starts, goes, or ends

Verbs encode manner or path information

Manner- jump, hop, skip, prance

Path- enter, exit, ascend

Typology

Languages tend to lexicalize either path or manner in verbs

Manner of Motion- english, greek

Path- Turkish, etc.

When a language lexicalizes one kind of info, that info is carried by the verb, the other info is expressed by other means, using via PPs

Talmy, Jackendoff

FROM(start)

TO(end)

VIA(path passes near place)

EVENT

GO

FROM (THING

TO (THING:

Principle of compositionality

The meaning of a linguistic expression is a systematic function of the meaning of its parts and their syntactic organization

-an expression’s meaning is built from the meaning of its parts

Non compositionality

The meaning of some expressions is not compositional, meaning that the overall meaning is not a systematic meaning of its parts (It’s raining cats and dogs)

Speakers have to memorize holistic meaning

NP VP

Ordinary declarative sentences assert a proposition

Propositions have:

truth values- is the fact that is being asserted true?

truth conditions- the conditions that must hold in the world for the proposition to be true

Truth values

Propositions have truth values, meaning they can be true or false

Ordinary declarative sentences assert a proposition, assert fact about world

ADJ NP

Adjs specify some subset of the entities referred to by the NP

Some do this by intersecting two sets

ADJ NP=the set of entities that are in both the ADJ set and the NP set

Non-intersection: refers to individuals who may or may not be in the set referred to by the NP

ex: possible, alleged, likely

Anti-intersection: refers to individuals not described by NP

ex: fake, pretend, artificial

A fake gun is not in the set of things referred to by “gun”

ADJ picks out something that is not NP but similar to NP

Thematic roles

agent: individual which performs action

patient: something which is acted upon

theme: something which moves as a part of the action

source: location/individual from which the movement occurs

goal: location/individual to which movement occurs

location: location at which something happens

instrument: something the agent uses to make something happen

cause: something that causes something to happen

stimulus: something that causes an experience

Pragmatics

study of how we use language in context, and how context contributes to meaning

Key concepts:

-Meaning is context-dependent

-Conversations are structured and follow rules

-We make inferences from language

-People use language to do things

Context dependence- deictic terms

What a sentence means and does depends on the context

D terms: linguistic expressions that by themselves don’t mean anything specific. Their meaning is entirely dependent on the context in which they appear.

-Pronouns

-Temporal terms

-Directional terms

Ex: She is there now

could mean a friend of yours is in class, could mean that Ruth Bader Ginsburg is in the Supreme Court in 1995

Different types of context

Linguistic- what has been said earlier in the conversation

Extra-linguistic: context outside of language

-Situational-local environment of the utterance

-Social- the relationships and norms of the social situation

Given vs. new information

Given- information that is assumed by the speaker to be known or inferable by the addresse at the time of the speaker’s utterance because it is

-common knowledge

-part of the extralinguistic context

-previously established in the discourse

Tends to be referred to with single words and deictic terms

New- information that is assumed by the speaker to be unknown or not likely to be inferred by the addressee

New info tends to be expressed with full phrases

Context-dependence summary

Linguistic expressions vary in the amount of their meaning that is determined by context

Deeictic terms have almost all meaning determined by context

Context can be discourse, environment, etc

How we formulate an utterance in conversation depends on whether the information is given or new

Grice’s cooperative principle

Grice: mental framework exists that allows listeners to make inference

Cooperative: People in a conversation assume that what people say is intended to contribute to the purposes of the conversation

Gricean maxims

Assumptions that hold when we engage in convos

Maxims of quality: expectations of honesty, don’t say what you believe to be false, don’t say that for which you lack evidence

Maxim of relevance: be relevant

Quantity: provide the right amount of information, make your contribution as informative as required, don’t make more informative than required

Manner: make contribution in straightforward, common way, avoid obscurity of expression, be brief, avoid ambiguity

Flouting maxim

When a speaker says something that in its most literal form apppears to violate a maxim, but the listener is expected to understand the intended communication

Play a large role on humor

Entailment

Facts that must also be true if a given proposition is true

Only statements that are logically necessary

Monique has a red car

Proposition: Monique owns a red car

Entailment: Monique owns a car

Monique owns things

BUT: Monique owns a sports car is NOT entailed since it is not logically necessary

Maria stole the diamond

Entailment: maria is a theif

Maria allegedly stole the diamond

Does not entail: Maria is a theif

Conversational implicature

Inferences made based on language, information inferred from an utterance that is not the literal proposition or an entailment of that proposition

The non-literal meaning of an utterance

Implying information through language

Using maxims to make inference

1: Speaker conforms to the maxims

Listener makes inferences from maxims

2: Speaker flouts the maxims

Listener assumes they are being cooperative

Listener reinterprets utterance

Direct speech act

Locution: the actual utterance itself

I apologize for being late

Illocution: communicative intention

Apology- the act of apology is performed simply by saying it

Perlocution: effect on the listener

the listener may feel appeased, forgive the speaker

Indirect speech act

Someone is blocking a doorway

Locution: could you open the door?

Illocution:

Request- you are asking them to open the door, not questioning their ability

Perlocution-the listener may open the door

Performative speech acts

Utilize specialized verbs to accomplish the speech act denoted by the verb

I order you to buy that house

Language progression

Language development proceeds according to a set of milestones

Proceeds at a regular pace in a particular sequence no matter which language the child is learning

Milestones

0-1 month Cooing

6 months babbling, infants produce consonant vowel sequences with segments from both ambient and non ambient languages

8-11 months, first signs of comprehension of words

11-13 months first word production

12-18 months one word stage- single word utterances used to convey a variety of message types, phonological simplifications: potato [tedo] juice [dus]

18-22 months two word speech

2-4 years telegraphic stage: generally correct syntactic order, primarily for content words, few function words

5-6 years: most basic syntactic structures, about 12K-14K words, fluent production

Development

1. Comprehension precedes production

2. All linguistic levels develop simultaneously

3. Competence and performance develop simultaneously

   1. Immature language can be due to immature competence, perofrmance,both

How do children learn

Recievbe positive evidence (sentences consistent with the grammar to be learned)

To not recieve negative evidence (that’s not correct) or instruction (this is how you form a question)

Produce productive patterns (grammars)

Over -regularization

Errors reflect over-application of rules

Application of grammatical rules to an irregular items

-holded

-runned

-When she be’s in the kindergarden

Critical period

Period of time in development where input from environment is necessary for typical functioning to development

Genie

Isolated since 20mo, rarely if ever spoiken to, Found at age 13 yr 9 months

Language performance following exposure- vocab grew, syntax never developed typically

Chelsea

Born deaf

Never diagnosed as deaf, as a result, never exposed to language

First fitted with hearing aids and exposed to speech at 32

Language performance following exposure:

After 12 years vocab 9/10th grade, syntactic ability never approached that of a typical adult

Poverty of the Stimulus

In cases where linguistic input is impoverished in some way, children learn linguistic grammars that are more sophisticated, more complex, more language-like than what they are exposed to

Nicaraguan sign language

First cohort- children 10 or older

Second cohort: children who joiuned later in time

First cohort developed a signing system that was like a pidgin, second cohort had more sophisticated constructions- signing increased in fluidity, more compact, more stylized

Youngest more advanced

Young children who recieve impoverished input regularize and improve on input

Synchronic

Diachronic

Pertaining to language at a particular point in time

Diachronic- pertaining to language across time

Word level change

Adding words

Semantic change

Word choice

Word specific phonological change

Non-morphological ways of adding words

Acronym: combining initial letters or sounds of a phrase into a pronounceable word

Blend: combining the phonemes of two or more words

Clipping: shorteneing a word: app

Coining: creating a brand new word by combining phonemes

Borowing: from other languages

Semantic change

Words constantly changing their meaning

Semantic extensions/reductions

Pertains to changes in the number of things a word refers to and how specific/general it is

Broadening/extension: the meaning of a word broadens, the number of things it refers to increases, becomes less specific

Reduction: the meaning of a word narrows

The number of thing is refers to decreases, becomes more specific

Semantic elevations and degradations

Pertains to changes in the sense/connotation of a word

Elevation: word becomes more positive

Degradation: word becomes more negative

Word choice

What word we use to express a concept changes over time

Vocabulary convergence influences word choice- natural tendency to express a concept with a single word

-Social cachet

-Cross-generational opacity

-Formal considerations: brevity, phonotactic frequency, lexical support

Word specific vs language wide change

Changes to specific words involve changing specific lexical entries

Changes that are language wide involve changes to grammars

A paradigm is a set of inflected forms of a word

Paradigm levelling

Instance of morphological regularization: pressure/ for uniformity causes idiosyncrasies to become erased in favor of dominant pattern

climb, past: clomb—> climbed

Scalar implicature

Scalar implicatures are inferences that arise when a weak expression is used instead of a stronger alternative. For example, when a speaker says, “Some of the children are in the classroom,” she often implies that not all of them are.

Algospeak

Example of how social pressures can accelerate lexical innnovation and semantic shifts

Speech to text (perception) Bottom-up info

Goal: turn audio signals into words

Key steps-

Acoustic analysis- break audio into small frames and extract features

Phoneme recognition: identify basic units of sound (phonemes)

Word recognition: map sequences of phonemes to words using language models

Ex Siri, google ai

Speech to text (perception): Top-down info

Pronunciation modeling guides- phoneme identification, what phonemes are likely in this context

Lexical information

Phonological information, allophonic and phonotactic

Language modeling guides which words are likely given:

Frequency of use

Other words in sequence

Syntactic structure

Text to speech (production)

Goal: convert written text into natural-soundng sound

Text analysis

Prosody modeling- decode how to stress words, pause and intonate

Waveform synthesis: generate actual audio waveform using algorithms or neural networks

Ex: Alexa

Distributional semantics

Distributional hypothesis: words that appear in similar contexts tend to have similar meanings

By looking at these patterns across huge text corpora, computers can build vector representations (embeddings) that capture meaning mathematically

Vector representations

1. Count how many times each target word occurs

2. Build vectors out of (a function out of) these context occurence counts

3. Similar words have similar vectors

Input: Word collocation counts

Output: Vector for each word (length=# words in language)

Each word becomes a vector where dimensions represent co-occurence with other words

Embeddings

a function that takes words as inputs and produces vectors as outputs

System performs task of predicting next word or character based on current word or character

How likely is x to appear near y

Static embeddings: Lears which vector dimensions are useful for task

System performs task of predicting next word based on current word

Problem: Vector collapses multiple meanings into one representation, do not encode context

LLMS

Meaning is contextual- when LLM processes text, each word becomes vector that encodes what word means, how it relates to surrounding words

meaning = context dependent patterns in vector space

N-grams

Sequence of n words that can be used to approximate which word should go next in the sequence

N-gram models choose the most likely word to go next in a sequence given data on which words appear next to each other

Require a lot of data to identify word relationships, diverse datasets to generate novel text, specific datasets relefant to use case of model

Transformers

Neural network architecture proposed by google in 2017, LLMs like ChatGPT only possible because of transformers

have attention mechanisms that make them good at generating human language

Encode context sensitivity- subject verb agreement and homonyms

Social implications

Ignoring linguistic diversity

Embeddings reproduce implicit biases

Major research areas: psycholinguistics

Speech perception: How we decode acoustic signals into meaningful sounds

Lexical access: How we retrieve word meanings from mental storage

Sentence processing: How we parse grammatical structure and build meaning

Language production: How we plan and articulate our thoughts into speech

Speech perception

We can identify speech sounds even amidst noise- this is a difficult task because we don’t produce speech sounds exactly the same every time: lack of invariance problem

Speaker normalization- the modification of our expectations or judgements about linguistic input to account for what we know about the speaker

McGurk effect

the brain combines conflicting visual and auditory speech information, creating a third, merged sound perception

Other factors involved in speech perception

Phonotactic constraints- slips of the ear errors- mistake in percieving an acoustic signal but it still yields a possible word

Lexeme knowledge- words that our in our mental lexicon can help us identify individual sounds

Linguistic context- word heel is most likely to appear in context of shoes, whereas word peel is more likely to appear in context of oranges

Phoneme restoration effect

when a phoneme is replaced by noise, listeners hear it anyway, brain fills in missing information based on context

Syntactic parsing

As we hear a sentence unfold, we assign expressions to syntactic categories and build syntactic structure that’s updated as a new word comes in- brain doesn’t wait until end of sentence to start understanding, parse as you go

Temporary ambiguity

Present up until some point during the processing of a sentence, but that is resolved by the end of the sentence (because only one of the original passes is consistent with the entire sequence of words) The rock band *banned played all night

Garden path effect

phenomenon by which people are fooled into thinking a sentence has a different structure than it actually does because of a temporary ambiguity

Global ambiguity

ambiguity is not resolved by the end of the sentence and require context to determine intended structure and meaning

Late closure: incoming material should be incorporated into the phrase currently being processed

Constraint based models: models of sentence parsing in which context, frequency, and specific lexical info can influence decisions about ambiguities

Prosody

Intonation and pausing help resolve ambiguity

(commas)S

Stages of speech production

Conceptualization: deciding what to say (the message)

Formulation: selecting the sounds, words and grammatical structure

Articulation: motor execution (moving the articulators)

Production errors

Anticipations: a later unit is substituted for/added to an earlier unit

Perserations: an earlier unit is substituted for/added to a later unit

Metathesis: switching up two units each taking the place of the other

Spoonerisms: Metathesis that involves the first sounds of two separate words

Shifts: unit is moved from one location to another

Blends: Two words fuse into a single item

Substitutions: One unit is replaced with another

Reveal that speech is planned in advance, there are distinct levels of planning, phonotactic constraints are for the most part obeyed

Slips of the hand

Signed production errors are systematic

Types of non-literal language

Metaphor: understanding one thing in terms of another

Idiom: Fixed expressionn with non-compositional meaning

Sarcasm: saying opposite of what you mean

Sequential processing theory

Brain computes literal meaning, detects literal meaning does not fit context, searches for alternative meaning, figurative language takes longer to process

Direct access theory

brain uses context from the start, accesses figurative meaning directly, familiar metaphors/idioms are processed just as fast as literal language

Rhoticity

Whether a variety contains an /r/ like consonant in syllable coda position

NYC rhoticity- prior to WWII, non-rhoticity was prestigious

After WWII, rhoticity became prestige norm

Rhoticity and Labov

Study asked to produce words with /r/ in coda position given a variety of tasks, rate of /r/ production correlated with socioeconomic status

Age and gender

Women and young adults are often linguistic innovators

Young adults- young people’s speech is still plastic and identity formation drives differentiation from older generations

Women- may use language as social capital, being more attuned to subtle presitge markers. Typically maintain larger, more diverse social networks, increasing exposure to new variants

Gay speech

Beliefs/stereotypes- there is a gay speech style, gay men lisp, gay men have effeminate voices

Research conclusions: there does not appear to be a gay speech style, listeners are capable of discerning these cues, different for lesbians/bis and gay men, does not involve mimicry of other sex

Style switching

A stylistic switch in language use to align with one’s context or with one’s perceived identity in a given situation

Register

a collection of linguistic behaviors and features associated with a particular social situation

What about the situation matters?

Topic, purpose, modality, genre, social relations

Theories of variation

Variation as grammar: Variation is systematic, so it makes sense to think of it as grammatical rules- speakers of different varieties learn different grammatical rules. Speakers learn different rules based on their social communities/affiliations

Theories of style switching:

Attention to speech- style varies according to the amount of attention the speaker pays to their speech

Audience design: speakers adjust their speech according to the needs of and their relationship with interlocutor

Convergence- speaker will attempt to match speech of interlocutor

Divergence: speaker will shift speech away from interlocutor

Chain shift

Many vowels in a variety change their pronunciation

Change in one vowel leads to change in the next:

Ex: Northern cities vowel shift

Joos (1961) classification of registers

Frozen: unchanging language- fixed language: Pledge of Allegance

Formal: One way participation among strangers ex: presentations

Consultative: two-way participation (teacher/student)

Casual: conversation among in group friends

Intimate: family members, close friends

Discrimination where language is a conduit for identity

Purnell housing discrimination study: landlords/sellers screen out unwanted applicants on the basis of their accents callbacks varied on the dialect the user used- anecdotally black and hispanic renters got fewer call-backs about housing opportunities

Linguistic prejudice

People infer identity from speech and may discriminate based on it: attitudes towards southern varieties, Black English

Vocal fry

Vocal fry is the lowest vocal register, a low, creaky, rattling sound made when vocal cords are loose and air bubbles through

Key

Linguistic constructions have no inherent value

A judgement of a linguistic feature is a hidden judgement of a group of people- it is a reflection of our social biases