48
Patrick Sturt ([email protected]) [email protected] Psychology of Thinking & Language

L2 Thinking

Embed Size (px)

DESCRIPTION

 

Citation preview

Page 1: L2 Thinking

Patrick Sturt ([email protected])

[email protected]

Psychology of Thinking & Language

Page 2: L2 Thinking

• Review of categorical perception• Introduce to the mental lexicon, and the problem of

lexical access.• Introduce some factors that affect lexical access.• Introduce models of context effects in recognising

ambiguous words.• Introduce experimental techniques:

– Lexical decision

– Cross modal priming

– Eye-movement recording

Aims of the lecture

Page 3: L2 Thinking

VOT5 ms

VOT15 ms

VOT35 ms

VOT25 ms

VOT45 ms

VOT55 ms

do

VOT0 ms

to

VOT60 ms

‘perceptual boundary’ of 30ms

• Eg. consonants can be classified as :voiced (eg. b/d/g VOT0 ms) voiceless (eg. p/t/k VOT60 ms)

Review: categorical perception

perceiving things that lie along a continuum as belonging to one distinct category or another.

all sound voiced all sound voiceless

Page 4: L2 Thinking

categorical perception

• methodology reminder– subjects hear pairs of consonants . Eg.

VOT0 ms VOT10 ms

VOT25 ms VOT35 ms

VOT50 ms VOT60 ms

– say whether pairs sound the same or different

same

different

same

crosses category threshold (30ms)

Page 5: L2 Thinking

found that small differences between sounds within a category can still be perceived

categorical perception

Pisoni & Nash (1974):

VOT0 ms VOT10 ms

VOT10 ms VOT10 ms

same

same FASTER

VOT5

VOT15

VOT35

VOT25

VOT45

VOT55

do

VOT0

to

VOT60

all sound voiced all sound voiceless

Nonetheless…

… the category boundary still exists!

Page 6: L2 Thinking

• approx 75,000 words in memory– just 250 msec to find each one

• stored in mental lexicon – mental storehouse of words (in LTM)

• mental lexicon contains lexical entries – one per word– contains all info about the word

• spelling/ pronunciation/ meaning etc.‘cat’

[kæt] noun

Understanding words

Page 7: L2 Thinking

• lexical access : retrieving a word from the mental lexicon– accessing its lexical entry

Lexical access

HOUSE

‘house’[haws]

noun

• seeing/hearing a word starts to activate its lexical entry• when the activation is high enough, lexical access takes place• threshold : the level of activation needed for lexical access

‘toast’[towst]

noun

Page 8: L2 Thinking

cloth

snookermarathon

chair

hillsidetelephone

chop

monkeyhorse

cheeseballroom

cake rain

clifffloor slide

bomb

concertcleanerpullover television piano

Random organisation?

Page 9: L2 Thinking

• We can find out how words are organised by looking at things that make lexical access easy or hard

• How do we know whether a word is easy or hard to access?– Lexical decision task

word organisation

Page 10: L2 Thinking

• Non-words (BRUKE) are ‘fillers’– to check the subject is paying attention– we only look at real words

HOUSENOIKSLEEPNURSEBRUKE

(400 msec)(450 msec)

lexical decision task

• Press YES or NO for whether the following is a real word in English:

• FAST response = easy to access• SLOW response = hard to access

• task variant: lexical naming– measures time taken to pronounce the word aloud

Page 11: L2 Thinking

1.Word Length• short words are faster to access than long words

– E.g. chaos vs. confusion

– in lexical decision (Chumbley & Balota, 1984) – in word naming (Weekes, 1997)

» strongest for 5-12 letter words

• also when saying numbers (Klapp, 1974)

– faster to start saying 91 vs. 77 (syllables: 3 vs. 5)

what affects lexical access time?

Page 12: L2 Thinking

2. Word Frequency• High frequency words = common words (cat, mother, house)

• Low frequency words = uncommon words (czech, 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)

what affects lexical access time?

Page 13: L2 Thinking

– the prior context lowers the threshold for the word’s activation

what affects lexical access time?

– interacts with frequency• low frequency words are facilitated (speeded up) more

• 3. Priming– when word-reading is ‘facilitated’ (speeded up) by prior context

a) repetition priming (Scarborough et al., 1977)

– a repeated word is read faster second time round

• eg. pen … pen

– pun speeds up pun more than pen speeds up pen

– pen1st (‘context’) speeds up pen2nd

Page 14: L2 Thinking

• subject sees 2 words

• must say (YES/NO) whether both are real words

– doctor grass

– doctor nurseSLOW

FAST

b) Semantic Priming (Meyer & Schvandeveldt, 1971)

what affects lexical access time?

• suggests…– reading doctor somehow speeds up nurse

• doctor & nurse are linked in our minds (but NOT doctor & grass)

priming: when language processing (eg. reading nurse) is ‘facilitated’ (speeded up) by prior context (e.g., reading doctor)

Page 15: L2 Thinking

• accounts for priming effects• words in memory represented in a ‘network’• each word is a ‘node’ • nodes (words) are connected to other nodes related in

meaning• accessing a word causes activation• activation spreads to connected nodes

spreading activation model

Page 16: L2 Thinking

canary

bird

animal

ostrich

mammal

Spreading Activation ModelSpreading Activation Model

yellow

doctordentist

fever

green

baby

cradle

bed hospital

sun

rainheat

grass

nurse

delirium

Page 17: L2 Thinking

canary

bird

animal

ostrich

mammal

Spreading Activation ModelSpreading Activation Model

yellow

doctordentist

fever

green

baby

cradle

bed hospital

sun

rainheat

grass

nurse

delirium

Page 18: L2 Thinking

• like semantic priming, but for word related in sound (not meaning)

– trail chute

– shoot chute

SLOW

FAST … because chute is already ‘warmed

up’ by having just activated shoot

c) Phonological (sound) Priming– (Evett & Taylor, 1982)

• What does this tell us about how the lexicon is organised?

– not only a semantic network but also an phonological network

what affects lexical access time?

Page 19: L2 Thinking

canary

bird

animal

ostrich

mammal

Semantic NetworkSemantic Network

yellow

doctordentist

fever

green

baby

cradle

bed hospital

sun

rainheat

grass

nurse

delirium

Page 20: L2 Thinking

poor

pour

shorn

prawn

pawn

Phonological NetworkPhonological Network

shin

shootchute

shore

sharp

door

dirt

shirt short

chin

shipgin

harp

sure

court

Page 21: L2 Thinking

• Bruce (1958) • participants heard words against background noise

• better recognition for words in context vs. out of context

priming: when language processing is facilitated by prior context

what affects lexical access time?

• 4. context

• words are recognised better in sentence contexts– Lieberman (1963)

• participants heard words either in isolation or in sentences

• in isolation, take almost twice as long to recognise

Page 22: L2 Thinking

what affects lexical access time?• 5. lexical ambiguity

– Which we will look at in detail in the rest of the lecture…

– What is the first sentence that comes to mind given the word:• BANK• COACH• BEAM• PORT

Page 23: L2 Thinking

lexical ambiguity

– WRITTEN word: ‘lead’• lead [lid] vs. lead [led]

• homographs: written the same but have different meanings

• lexical ambiguity: when a word has more than one meaning= polysemous words:

• bank; straw; letter

• ambiguity might be only in written or spoken language– SPOKEN word: [najt]

• knight vs. night

• homophones: sound the same but have different meanings

Page 24: L2 Thinking

lexical ambiguity

• many many words are ambiguous

• coach, straw, bank, chick, tea/tee, bug, tie, here/hear, mail, you/ewe, bread, letter, plane, eye/I, pound, score, watch, house, table, phone, plate, trip, branch, shore/sure, bed, two/too/to, top, cite/sight, fork, pan, …………..

• How do people choose the right meaning?

Page 25: L2 Thinking

Effect of context on meaning

• …but how?• is the relevant meaning selected immediately?• or do we first consider all meanings?

• contexts helps select the relevant meaning– The fisherman sat down on the bank– The businessman put his money in the bank

Page 26: L2 Thinking

3 models of context effects

• Autonomous access model• All of a word’s meanings are accessed from the lexicon,

regardless of context.

• Contextually appropriate meaning is selected later in an integration phase.

• Re-ordered access model• All of a word’s meanings are accessed from the lexicon

• Context can increase the speed with which a meaning becomes available

• Direct access model:• Only contextually appropriate meaning is accessed in the first

place.

Page 27: L2 Thinking

Understanding ambiguous words

• Swinney (1979)– bugs: insects/listening devices

– The man wasn’t surprised when he found several spiders, roaches and other bugs in the corner of the room

– context supports insects• but do people still instantly consider both meanings?

Page 28: L2 Thinking

Cross-modal lexical decision (Swinney, 1979)

– target word appears after subjects hear bugs• ant (related to “insect”)

• spy (related to “listening device”)

• sew (unrelated to either meaning)

• Participants listen to sentences, and look at a screen– words appear on screen

– lexical decision task

…he found several spiders, roaches and other bugs in the corner…

Slow

FAST

FAST

AMBIGUOUS CONDITION

Page 29: L2 Thinking

Cross-modal lexical decision

– target word appears after subjects hear insects• ant (related to “insect”)

• spy (related to “listening device”)

• sew (unrelated to either meaning)

• Participants listen to sentences, and look at a screen– words appear on screen

– lexical decision task

…he found several spiders, roaches and other insects in the corner…

Slow

FAST

SLOW

UNAMBIGUOUS CONDITION

Page 30: L2 Thinking

Cross-modal lexical decision

• So both meanings were available immediately after bugs– context had no immediate effect

• ant (related to “insect”)

• spy (related to “listening device”)

• sew (unrelated to either meaning)

• But Swinney also tested 3 syllables downstream

…several spiders, roaches and other bugs in the corner…

baseline

FAST

SLOW

AMBIGUOUS CONDITION

Page 31: L2 Thinking

• only relevant meaning remains (3 syllables) downstream

Cross-modal lexical decision

• So both meanings were available immediately after bugs– context had no immediate effect

• ant (related to “insect”)

• spy (related to “listening device”)

• sew (unrelated to either meaning)

• But Swinney also tested 3 syllables downstream

…several spiders, roaches and other insects in the corner…

SLOW

FAST

SLOW

UNAMBIGUOUS CONDITION

Page 32: L2 Thinking

Discussion

• suggests lexical access makes available all meanings instantly – independently of context:

• but context rapidly “kicks in”: (“integration”)– selects the appropriate meaning– screens out irrelevant meanings– process happens quickly (at least within 3 syllables)

• Supports Autonomous Access model

• but is this always true?

Page 33: L2 Thinking

• in fact, processing can vary, depending on …– the particular ambiguous word– the particular prior context

• All Swinney’s words were “balanced” (each meaning occurs roughly equally.• What happens with “biased” words (one meaning occurs more often than another)?

Understanding ambiguous words

Page 34: L2 Thinking

biased vs. balanced words

• port– harbour– wine

more common

less common

• bark– part of tree– dog noise

roughly balanced

BIASED

BALANCED

Page 35: L2 Thinking

• biased words:– common meaning is always accessed

– uncommon meaning is only sometimes accessed

• in fact, processing can vary, depending on …– the particular ambiguous word

Understanding ambiguous words

– Depends on the particular prior context

Page 36: L2 Thinking

Effect of context: Rayner & Duffy (1986); Duffy et al. (1988)

• eye-tracking ambiguous words– balanced (eg. bark) vs. biased (eg. port)

Eye-movements in reading:-difficulty of lexical access can be measured from “gaze duration” (time spent first fixating a word before moving to another word).

To their surprise, the bark was unusual 1 2 3 4 5 6 78

Page 37: L2 Thinking

Effect of bias: Basic finding (Rayner & Duffy; 1986)

• eye-tracking ambiguous words– balanced (eg. bark) vs. biased (eg. port)

slower gaze duration than control same as control

• neutral context:

He found the bark was…

control: howl control: soupLast night the port was…

• both meanings accessed– 2 meanings competing

• 1 meaning accessed– no competition

Page 38: L2 Thinking

Effect of context: Duffy et al. (1988)

• Manipulated whether the disambiguating context came before or after the critical word:– CONTEXT BEFORE:

– CONTEXT AFTERBecause they heard it from so far away, the bark/howl wasdifficult to identify

Unfortunately, the bark/howl was difficult to identify, because they heard it from so far away.

Page 39: L2 Thinking

Effect of context: Duffy et al. (1988)

• Also manipulated whether the ambiguous word was balanced (e.g. “bark”) or biased (e.g. “port”):– BALANCED:

– BIASED Because they heard it from so far away, the bark/howl wasdifficult to identify

Even though it had a strange flavour, the port/soup was a great success.

• The context always disambiguated to the less common meaning (e.g. “wine” meaning of port).• Ambiguous words were compared with unambiguous controls

Page 40: L2 Thinking

Duffy et al gaze duration results:

240

245

250

255

260

265

270

275

280

Balanced Biased

Ambiguous Control

245

250

255

260

265

270

275

280

Balanced Biased

Ambiguous Control

CONTEXT BEFORE CONTEXT AFTER

Page 41: L2 Thinking

Summary of Duffy et al

• When context after the critical word, lexical access was difficult for the balanced ambiguous word

• lexical access happens before disambiguating context

• 2 meanings are accessed at the same time, causing competition, and difficulty

• When context comes before the critical word, lexical access was difficult for the biased ambiguous word

• The context promoted the less preferred reading, making it accessed earlier than usual. Caused competition between the two readings.

This pattern of results is called the subordinate bias effect

Page 42: L2 Thinking

The re-ordered access model

– most frequent meaning is always accessed (harbour)– but context can “promote” uncommon meaning (wine)– If so, both meanings become available together

• meanings compete, causing difficulty

Page 43: L2 Thinking

Subordinate bias effect

– most frequent meaning is accessed first (harbour)– this is checked with context

• but it doesn’t make sense! comprehension break-down

– have to re-access for 2nd meaning (wine)• break-down & re-analysis causes slow-down

• slow-down when context biases to uncommon meaning– both meanings are accessed, but when? (2 accounts)

• IN SEQUENCE: Integration account‘Autonomous/ordered access’

Page 44: L2 Thinking

• context strongly biases uncommon meaning– When she served it to her guests, the port … slow-down

Subordinate bias effect

• slowdown could be caused by…

• competition between 2 concurrent meanings (“Re-ordered

access”) • 1st (common) meaning being ruled out (“Integration account”)

Page 45: L2 Thinking

Dopkins et al (1992)

• Tried to distinguish between integration account and re-ordered access account, also in an eye-tracking experiment:

• Positive condition: context highlights a feature of the uncommon meaning, but doesn’t rule out less common:– Having been examined by the King, the page was soon marched off to

bed. (King highlights “servant” meaning)

• Negative condition: context rules out common meaning• Having been hurt by the bee sting, the page was soon marched off

to bed. (Bee-sting rules out “paper” meaning)

• Neutral condition: • Just as Henrietta feared, the page was soon marched off to bed.

Page 46: L2 Thinking

Results of Dopkins et al

840860880900920940960980

1000

Completion times for"disambiguating" region

Positive

Negative

Neutral

Dopkins et al measured time taken to finish reading the final disambiguating region marched off to bed.

Page 47: L2 Thinking

Summary of Dopkins et al• Speed up in disambiguating region, both for positive and for negative

conditions• Shows that the prior context in the positive condition caused selection

of the less common meaning, at least some of the time.• Does not support integration account:

– Having been examined by the king, the page…• Still no reason to reject “paper” reading of page at this point, so

integration model would select “paper” reading.• Therefore “integration” model would predict difficulty at “was

soon marched off to bed”.• Results support the “re-ordered access” account.

• The phrase the king caused the “servant” meaning to be accessed more quickly

• Therefore less difficulty at “was soon marched off to bed”.

Page 48: L2 Thinking

Summary of Lecture

• READING:

– Harley, T. (2001). Psychology of Language. Hove: Psychology Press. Ch 6

• Lexical access: The process of retrieving details of a word from long-term store (lexicon).

• Speed of lexical access is affected by:• Length• Frenquency• Priming

• How do we process ambiguous words?• For balanced words (Swinney 1979), all meanings become

simultaneously available, and context is used to select.• Context can affect the order with which meanings become available

(Duffy et al, 1988; Dopkins et al, 1992).