Psycholinguisctics

Budapest Semester in
Cognitive Science
Cognitive Psychology Day 2.
Today’s themes

1. What is language?(Thursday) How do

we recognize its parts and types?
2. What levels of linguistic studies are
there? – some psycholinguistic
experiments
3. What is the relationship between
language and thought?
 Linguists and psychologists talk about different things…
Grammarians are more interested in what could be said than
in what people actually say, which irritates psychologists, and
psychologists insist on supplementing intuition with objective
evidence, which irritates linguists.
(Miller, 1990)

Chomsky’s Competence vs. perfomance

What is a language?

 Which one of these are the same

laguages?
 English – German
 American English – British English
 Black English – American English
 Jamaican Creole – Jamaican English

Mutual intelligibility
 Dialect continuum
 Standard language – written language
 Chinese – Japanese?

Dialect 1

Dialect 2

Dialect 3
 A disoputed case - German

•Isophone Isoglosses:
•Isolex
Ik - ich
•boot - trunk
•Isoseme Maken -
• dinner machen
•Isomorph
•Dived - dove
Language – same and
different

 Typology
 Based on morphological constructions
 Based on default word order
 Universality
Language typology

 Configurational – nonconfigurational
 Analytic-synthetic
 Agglutinating - inflecting
 Typology
 Configurational and non-
configurational languages

 Arwen
 Nazgǔl
 chase
 Typology
The Nazgǔl are chasing Arwen.

A Nazgullok kergetik Arwent.

Arwent kergetik a Nazgullok.

 Arwen
 Nazgǔl
 chase
 In the box  A dobozban
 At the table  Az asztalnál

 You could have  (ti) megcsináltathattátok

got it done (volna)

•Free word order

•Null anaphora
•Syntactically discontinuous
expressions
Language typology
 Word Order
 S – subject
 V – verb
 O – Object

 What is the default word order in English?

 The cat
 Mouse
 Little
 Chase
 Linguistic universals
 Joseph Greenberg – 30 languages
 Absolute – substantive
 Lexicon and grammar
 Nouns, verbs, pronouns (deictics – time, space,
number)
 First person
 Vowel, consonant
 Rules of intonation
 No language without /a/
 Antonymy – categorial thinking?
 Roman Jakobson – Linda Waugh: i sound
 What do you call a small cat? (Mackó – maci)
 Cecil H Brown: body part namings
 Body, head, eyes, arms, nose, mouth
 IF foot > hand
 IF individual toes > individual fingers
 Some languages lack the term for
‘body’
 Implicational – statistical
 Trial grammatical number > dual
grammatical number
 VSO languages > adjectives come after
nouns
 SOV languages > postpositions
A war that never ends
 Descriptive and prescriptive linguistics

 third-person singular /s/: "she goes," - "she go."

 no double negatives: "he didn't see anybody," - "he didn't
see nobody." "who/whom did you see"

 "Winston tastes good like/as a cigarette should"

 "the data is/are unreliable"
 "I disapprove of him/his doing it"
 "get it done as quick/quickly as possible"
What is a word?

 Meaningful units (It is the light I switched on)

 potential pause – or is there?
 Undivisibility (absobloominglately)
 Phonetical boundaries (vowel harmony,
stress) – statistical learning!
 Minimal free morphemes (the, of)

 How do children ever learn to distinguish

words? (then the gavagai problem)
 Statistical learning might be one answer!
Statistical learning and Implicit
learning

 Initially very different

 Acquisition of syntax – remember artificial
grammars
 Acquisition of vocabulary

 Later converged – now their

interpretations are different, but
reconciliable
Saffran, Aslin & Newport 1996

 Can 8-month old infants extract information

about word boundaries solely on the basis
of the sequential statistics of concatenated
speech?
 Familiarisation-preference procedure
(Jusczyk & Aslin 1995)

Infants exposed to auditory material that

serves as potential learning experience.
Saffran, Aslin & Newport 1996
 Test stimuli
Items contained within the familiarisation
material
Items highly similar but weren't within the
familiarisation material.

 Hypothesis
If they have extracted the crucial info from
the data, there will be a differential fixation
time.
Saffran, Aslin & Newport 1996

 Speech stream
Length 2 mins
Speed 270 syllables/min
Content 4 trisyllabic nonsense words
(repeated in random order)
TP 1 within words; 0.33 across
words.
No effect of co-articulation, stress...
Saffran, Aslin & Newport 1996

Sample
bidakupadotigolabubidaku...
Saffran, Aslin & Newport 1996
 What is a Transitional Probability (TP)? P= x/xy

X A TP(XA) = 1.0

A TP(XA) = 1/3 = 0.33

X B TP(XB) = 0.33
C TP(XC) = 0.33
Saffran, Aslin & Newport 1996

 Which one has a higher TP?

Pre.tty Ba.by

1. TP (Pre, tty) > TP (tty, Ba)

2. TP (tty, Ba) > TP (Pre, tty)

Classical English example: „tp”

say a word that contains this sequence
Saffran, Aslin & Newport 1996
 Test items

4 items in total
- 2 of the 'words' from familiarisation.
- 2 with the same syllables from
familiarisation but not same order.

Infants can distinguish between novel/familiar

orderings; So, they can extract serial-
ordering info. (after 2 mins!!)
Word classes

 How do you know what a noun is?

 Semantic definition – what designates
something
 What about happiness or love
 Grammatical definitions – comes after
„the”
What part of speech is it? /
Guess

1. Can occur after „to be”

2. They can occur after articles (the, an)
and before nouns
3. They can occur after very
4. Can haver superlative forms (er, est)
5. Can form an adverb with -ly
Or are things really as clear-
cut?

 Try these
 Happy
 Old
 Top
 Two
 Asleep
 Want
Why is psycholinguistics
interesting?

 Speech is natural

 All humans learn a language – no culture

without speech (numbers - Piraha! Colours -
dani)

 Any human baby can learn any human

language – what about deaf children?
What’s the big deal?

 It is actually sg of a miracle that we

manage to speak and understand as
well as we do.
 What do we do?
We understand speech stream which
includes no discrete boundaries to
indicate where one word ends and
another begins
 To understand speech...

 Vibrations arrive at eardrum

 discharged in auditory nerve;
 brain translates nerve signals into sounds;
 separated from background noise
 separated into individual words (segmented)
 Words are accessed in brain to find meanings;
 Words and grammatical structures are
interpreted
 Link with prior knowledge…… NO EFFORT?
 Further aggravating things
 different accents
 different speech rates
 stammering
 incomplete sentences
 ambiguity

 vocabulary of between 50.000-100.000 words

 2-4 words per second
Our language awareness is raised
under special conditions:

 language impairments,
 talk to children learning the language,
 when we are not sure what was said,
 when we cannot find the words,
 learning languages.
 A nyelv szerkezete
language

sounds meaning
grammar

phonetics semantics
morphology
phonology pragmatics
syntax

articulation suffixes
meaning

Sound patterns
structure intention
Sounds
 Categorical perception of sounds
 Continuous vs categorical (Bird vs big)
 The role of Voice Onset Time –
 Studies of Alvin Liberman (pa/ba)
 Yet: a/u are less so – motor theory of perception?
 Kuhl: Infants and chinchillas can do it – neither speaks
 Develops in infancy – 6-9 months
 Bilinguals- there is a debate on their categories
 Top-down construction – the Ganong effect
 Dash/tash or Dask/task
 problems with computer speech perception
Categorical perception in bilinguals
Voices modified with
the Klatt synthesizer
 A nyelv szerkezete
language

sounds meaning
grammar

phonetics semantics
morphology
phonology pragmatics
syntax

articulation suffixes
meaning

Sound patterns
structure intention
Morphosyntax
– remember typologies!

 Morphology
 The forms of words
 Particularly important in languages using
cases
 3 main types of languages
 Isolating
 Agglutinative
 Fusional
 Often studied with priming paradigms
The Great Rule Debate
 Remember yesterdays discussion about
rules and memory (chunks)?
 Language has the same problem with
regulars
 You think English is an easy language?
Have a thought about irregulars!
 The rules seem flexible at best. If 2 mouses are mice, then
why aren't 2 blouses blice? Or or 2 houses hice?
 If it's one ox and 2 oxen, why shouldn't it be one fox and 2
foxen? "Henry, grab the shotgun, there's foxen in the henhice!"
Ambivalence is there!
 What do you call
 A radius and another radius?
 A nucleus and another nucleus?
 A focus and another focus?
 An octopus and another octopus?
 A virus and another virus?
 A chorus and another chorus?
 A campus and another campus?
 A bacterium and another bacterium?
 A medium and another medium?
 An album and another album?
 Irregulars tend to get lost over time-forgotten
 Is Hungarian an easy language?
 Difficulties
 agglutinating system
 Lots of irregulars
 Lots of subrules
 Direct and indirect object marked on the verb

 BUT the good news: you can make

yourself understood even if you get all
these wrong
Hungarian

 Altaic language –
Finno-Ugoric
 Not Indo-European
language
 http://linguistics.buffalo.edu/people/faculty/dryer/dryer/fa
mily.maps
The Hungarian noun declination system
–18 cases

 Nominative – default case Exercise: try

conjugating these:
 The cat is on the mat.
•mázli (fluke)
 Accusative /Vt/
•ribizli (blackcurrent)
 The girl hit the boy.
•tojás (egg)
 Plural /Vk/
•lakás (apartment)
 The cats
•virág (flower)
 Dative /nVk/
•house (ház)
 I gave the flowers to the lady.
 Comitative /vVl/
 I went to the market with Jane.
A double division line
This would require tons of subrules!

Meaning NOM ACC PLUR DAT COM

[-] [-ØT] [-ØK] [-NAK] [-VAL]
[dog] kutya kutyát kutyák kutyának kutyával
[luck] mázli mázlit mázlik mázlinak mázlival
[giraffe] virág virágot virágok virágnak virággal
[egg] tojás tojást tojások tojásnak tojással
[house] ház házat házak háznak házzal
[monkey] majom majmot majmok majomna majomma
k l
[mouse] egér egeret egerek egérnek egérrel
[horse] ló lovat lovak lónak lóval
The Great Rule Debate

 The formation of different

morphological forms of words:
 Memory? – is there a limit to memorized
forms (in acquisition time)
 Computational load?
 declarative memory

manipulation - procedural

Rule-based theories
memory

rules dual route rote memory

 Symbolic and abstract units of language

 V+-ed
 Full regularity hypothesis
 the transformation remains, but underlying forms
predict surface forms
 Run ->rin (run, ran,run - cling, clang, clung)
 The Wug Test

 Jean Berko
Gleason’s test
 He administered it
to children to see
how much they
know about the
rules.
  Developmental data in Hungarian
Age and inflectional paradigms (Pléh, Palotás
& Lőrik, 1994)

100
80
oroszlán
Correct %

60 hal
40 róka
20 madár
viziló
0
majom
4 5 6 7 8
Age
 declarative memory

manipulation - procedural

Dual route
memory

rules dual route rote memory

 Pinker & Ullman

 Race model –
rather unfair:
irregular always
wins
 Doublets are
stored – in both
Stems forms
Phrases, sentences
Idioms
Regulars
Irregulars
Regulars (frequency)
 Acquisition dissociation

 Children start off

knowing both regulars
100

90
and irregulars
80

70
 then somehow they
60 tend to forget about
50

40 irregulars
30

20
 Goed
10

0
 doed
2 years 3-4 years 5-6 years
Single Pattern Associator

 Tried to exploit
phonological
similarities of
irregulars
 Back-progagation
 Using sounds as
input and other
sounds as output
 declarative memory

manipulation - procedural

Rote memory
memory

rules dual route rote memory

 Rummelhart & McClelland

Symbolic Sub-
symbolic
deterministic Rules
Probabilistic WPM SPA
Morphological families
 Harald Baayen
 Transitions are not clear-cut
between regulars and
irregulars
 „The whole takes
precedence over the parts”
 Lexical Decision Task
 Press YES or NO for whether the following is
a real word in English:

HOUSE
SLEEP
NOIK
BRUKE
NURSE

 Non-words (BRUKE) are ‘fillers’

 Just to check the subject is paying attention
 We only look at real words
 FAST response = easy to access (450 msec)
 SLOW response = hard to access (500 msec)
 What affects lexical access
time?

 1. Word Frequency
 High frequency words = common words (cat, mother,
house)
 Low frequency words = uncommon words (accordion,
compass)

 High frequency are faster to access than

Low frequency
 even when they’re balanced on other features (e.g.
length)
 E.g. Pen vs. Pun
 Rubenstein et al. (1970)
 The Logogen Model
Morton (1969)

 Accounts for the frequency effect

 The lexical entry for each word comes

with a logogen
 The lexical entry only becomes
available once the logogen ‘fires’

 When does a logogen fire?

 When you read/hear the word
Think of a logogen as being like a
‘strength-o-meter’ at a fairground

When the bell rings, the

logogen has ‘fired’
 ‘cat’
[kæt]

• what makes the logogen fire?

– seeing/hearing the word

• what happens once the logogen has fired?

– access to lexical entry!
 ‘cat’ ‘cot’ Low
[kæt] [kot] freq
takes
longer

• So how does this

help us to explain the
frequency effect?
– High frequency
words have a lower
threshold for firing
–E.g. cat vs. cot
 What affects lexical access time?
 2. Semantic Priming Effects
 (Meyer & Schvandeveldt, 1971)
 Subject sees 2 words
 Must say YES or NO whether both are real
words SLOW
 doctor grass
FAST … because nurse
 doctor nurse is already
‘warmed up’ by
having just
activated doctor
 Spreading Activation Model
cradle

baby bed hospital nurse animal

dentist doctor
bird mammal

rain canary
fever
heat
delirium
sun ostrich
green
grass
yellow
 Spreading Activation Model
cradle

baby bed hospital nurse animal

dentist doctor
bird mammal

rain canary
fever
heat
delirium
sun ostrich
green
grass
yellow
 Semantic Network
cradle

baby bed hospital nurse animal

dentist doctor
bird mammal

rain canary
fever
heat
delirium
sun ostrich
green
grass
yellow
 Fits nicely with Logogen
Model

 Each of the nodes in the network has a

logogen with it

 When we read doctor, its logogen

fires
= doctor gets ‘activated’
 The activation from doctor spreads to
nurse, this lowers the threshold for
nurse
 so make nurse faster to access
 ‘doctor’ ‘nurse’
[doktə] [nə:s]

• spreading activation
from doctor lowers the Spreading
threshold for nurse to activation
fire network

– So nurse take less doctor nurse

time to fire

docto nurse
Previous experiments
 Pinker és Prince (1998)
 Lukács, 2001

 Acoustic visual priming

 Lexical decision task

 Plural form - stem

 Priming effect
• ++ regulars
• + irregulars
• 0 phonological
Experimental design
Prime Target
Word suffixed root
Do you see an existing
Modality aud visual word or not?

ablak
250 ms
Window (IN)
Window
 Contrasting theories

Exp. Stem Suffixed Paradig

result priming priming m
Theory priming
Rules & subrules ? ? 
Dual route ? ? 
WPM   
 A nyelv szerkezete
language

sounds meaning
grammar

phonetics semantics
morphology
phonology pragmatics
syntax

articulation suffixes
meaning

Sound patterns
structure intention
Pragmatics
 Sperber and Wilson – relevance theory
 most new information
 least amount of effort
 We automatically assume, that
 a) implicit messages are relevant enough to be
worth bothering to process
 b) the speaker will be as economical as they
possibly can be in communicating it.
 Specially important in the understanding of
irony. You are very hard-working.
 The intellectual Gurus – Derrida
Language and thought

Why do we bother to use

euphemisms and politically correct
terms?
of Newspeak in 1984.
 Language (with capital L):
 the human language capacity
 linguistic universals
 languages (with small l):
 individual languages (e.g., English,
Arabic…)
 types of languages (e.g., Indo-European,
Semitic…)
 Slobin, Dan

 If each language is simply an alternative code for the

same underlying cognitive processes and states, the
diversity of languages can be ignored by cognitive
science.
 But if linguistic diversity reflects cognitive diversity,
individual languages are critical independent
variables in cognitive science theory and research.
 Linguistic determinism
 Impossibility to avoid linguistic category
traps
 No understanding can be established
between the cultures
 Linguistic relativity
The Sapir-Whorf Hypothesis
 “We cut up and organize the spread and flow of
events as we do largely because, through our
mother tongue, we are parties to an agreement to
do so, not because nature itself is segmented in
exactly that way for all to see.”

 “From this fact proceeds what I have called the

‘linguistic relativity principle,’ which means, in
informal terms, that users of markedly different
grammars are pointed by their grammars toward
different types of observations … and hence are not
equivalent as observers …”
The language-thought interface

 Vocabulary size debates

 The great snow debate
 Categorization debates
 Focal colours debate
 Grammatical constructions & gender
 Grammatical gender of words
 Spatial language debates
 Description systems
The Snow-word debate
 Franz Boas
•Geographer and physicist originally –
interested in language and anthropology
•Clash between sciences and
humanities (psychophysics)
•Against the orthogonal evolutionary
theory – all cultures are equally developed
•Debate among geographers – is cultural
diversity determined by environmental
factors or „memes”? Language: On
•He mentions that Eskimos have four words: aput ("snow Alternating Sounds
on the ground"), qana ("falling snow"), piqsirpoq ("drifting No inferior languages –
snow"), and qimuqsuq ("snowdrift"), where English has is mispercieved
only one ("snow").
Edward Sapir

 Studied indigenous
languages of the Americas
 More interested in structure
of languages – language drift
and actually – universalisms!
 He fell ill and B. L. Whorf
took over his classes – and
the snow legend started
Benjamin Lee Whorf
 Chemical engineering – later studied
linguistics with Sapir
 His hobby was studying languages –
mainly meso-American ones (hopi,
nahuatl, maya)
 The Hopi language is seen to contain no
words, grammatical forms, construction or
expressions or that refer directly to what we
call “time”, or to past, present, or future…
 He was employed at an insurance
company to explore causes of fire
 Empty gasoline drums
The end of the snow debate
 There is no such language as Eskimo..
 Eskimo people might have more words for snow –
but so do ornitologists for birds! This is true of any
expert…
 What is a word? All inuit languages are polisynthetic
– agglutinating very ardently
 Tusaatsiarunnanngittualuujunga.
 I can't hear very well.
 -tsiaq-well
 -junnaq-be able to
 -nngit-not
 -tualuu-very much
 -junga1st pers. singular present indicative non-specific
The end of the hopi debate
 B.L. Whorf
 The Hopi language is seen to contain no words,
grammatical forms, construction or expressions or that
refer directly to what we call “time”, or to past, present,
or future
 Malotki
 The hopi use a very complicated character and a time
very similar to that of other cultures
 Hindi though
 Has the same word for yesterday as for tomorrow! ‘kal’
 Spanish
 Same problem ‘ya’
The language-thought interface

 Vocabulary size debates

 The great snow debate
 Categorization debates
 Focal colours debate
 Grammatical constructions & gender
 Grammatical gender of words
 Spatial language debates
 Description systems
The language-thought interface

 Vocabulary size debates

 The great snow debate
 Categorization debates
 Focal colours debate
 Grammatical constructions & gender
 Grammatical gender of words
 Spatial language debates
 Description systems
Frames of reference and dead
reckoning
Hypothesis

 Speakers of languages using absolute

frames of reference will be good dead
reckoners – r-statistic close to 1.00
 • Speakers of languages using relative
frames of reference will be poor dead
reckoners – r-statistic approaching
zero.
The pointing task dead-
reckoners
 Today we learned

 How difficult it is to use language,

although it’s easy and how difficult it is
to really know what a language is.
 That by studying different levels of
language we can learn about general
mechanisms of the mind
 That linguistic relativity might have
some truth in it and the wuestion again
is: How much?