hofmeister slides

Upload: lobna

Post on 08-Jul-2018

217 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/19/2019 Hofmeister Slides

    1/60

    PHILIP HOFMEISTER UNIVERSITY OF ESSEX

  • 8/19/2019 Hofmeister Slides

    2/60

    Average speech rate is around 150 wpm; reading rate tendsto be higher 180-200 wpm

  • 8/19/2019 Hofmeister Slides

    3/60

    COMPUTATIONALPROBLEM

    In any given sentence, the listener may need to

    identify words, e.g. dictionary-style look-up

    identify lexical categories (noun, verb, etc.)

    resolve syntactic ambiguities

    combine words with previous words (potentially over longspans)

    integrate visual information

    take into account speaker’s social status

    remember prior sentences & topic

    plan next utterance

  • 8/19/2019 Hofmeister Slides

    4/60

    PSYCHOLINGUISTICS

    Computational problem: howcan humans complete thecognitive tasks necessary tocommunicate with one anothergiven rapid, incremental natureof language?

  • 8/19/2019 Hofmeister Slides

    5/60

    BASIC FACTS

    Language processing is incremental

    You don’t wait to process a word or sentence

  • 8/19/2019 Hofmeister Slides

    6/60

    COMPUTATIONALPROBLEM

    Computational problem is compounded by incrementality& uncertainty

    That desert trains . . .

  • 8/19/2019 Hofmeister Slides

    7/60

    COMPUTATIONALPROBLEM

    Computational problem is compounded by incrementalityThat desert trains . . .

     [  NP  That desert] trains young people to be tough.

     [  S That desert trains come irregularly] is well-known.

  • 8/19/2019 Hofmeister Slides

    8/60

     WORDPROCESSING

  • 8/19/2019 Hofmeister Slides

    9/60

    How do weperceive sounds &words?

    How do weperceive soundaccurately given anoisy input?

  • 8/19/2019 Hofmeister Slides

    10/60

    PHONEMERESTORATION

    EFFECT

    Context plays an early role in perceptualprocesses

  • 8/19/2019 Hofmeister Slides

    11/60

    PHONEMICRESTORATION

    The state governors met with the respectivelegislatures convening in the capital city 

  • 8/19/2019 Hofmeister Slides

    12/60

    PHONEMERESTORATION

    Even when people know the phoneme is missing,they still hear it

    Seems to be a very fast-acting & strong effect ofcontext

  • 8/19/2019 Hofmeister Slides

    13/60

    PHONEMERESTORATION

    Or is it?

    maybe you just think you heard it after the factto make sense of the input

  • 8/19/2019 Hofmeister Slides

    14/60

    PHONEMERESTORATION

    It was found that the *eal was on the TABLEIt was found that the *eel was on SHOE

    Participants restored a phoneme based onevidence that came later!

  • 8/19/2019 Hofmeister Slides

    15/60

    PHONEMERESTORATION

    What to make of these conflicting results?Sentence contexts may have post-lexicaleffects

    Word contexts may have earlier, even pre-lexical effects

  • 8/19/2019 Hofmeister Slides

    16/60

     WORDPROCESSING

    How are words stored & accessed in the brain?

  • 8/19/2019 Hofmeister Slides

    17/60

     WORDPROCESSING

    All words are not processed the sameSome take a long time to process; others ashort time

    If the mind just has a dictionary, why would ittake longer to look up any word?

  • 8/19/2019 Hofmeister Slides

    18/60

    VISUAL WORDRECOGNITION

    Several factors have been identified as being

    critical in the speed of word recognitionfrequency: how often has the word beenexperienced?

    age of acquisition: when was the word firstlearned?

  • 8/19/2019 Hofmeister Slides

    19/60

    FREQUENCYEFFECTS

    Whaley (1978): frequency is the most important

    factor in word recognitione.g. “abhor” named & recognized slower than“sleep”

    effects are measurable for very frequent vs.very infrequent, frequent vs. infrequent

  • 8/19/2019 Hofmeister Slides

    20/60

    FREQUENCYEFFECTS

    predictability of frequency breaks down withextremely infrequent words

    individuals differ in their experience

    what’s common for me may be uncommon foryou

  • 8/19/2019 Hofmeister Slides

    21/60

    AGE OFACQUISITION

    frequency is highly correlated with age of

    acquisitionmore frequent words are typically learnedearlier, e.g. “go”, “see”, not “abhor”

    words learned earlier named more quickly andaccurately

  • 8/19/2019 Hofmeister Slides

    22/60

    EXPERIENCE

    In short, both factors suggest that personalexperience plays a huge role in how we processwords

    much of our experience is shared

    both AOA & frequency likely have independenteffects (Morrison & Ellis, 2000)

    AOA particularly effects reading rate

  • 8/19/2019 Hofmeister Slides

    23/60

    SENTENCEPROCESSING

  • 8/19/2019 Hofmeister Slides

    24/60

    AMBIGUITYRESOLUTION

    Many strings contain some ambiguity ofinterpretation (although we typically don’texperience confusion)

    The boy saw the girl with the telescope

     I heard Liam say he saw the movie yesterday 

  • 8/19/2019 Hofmeister Slides

    25/60

    AMBIGUITYRESOLUTION

    Syntactic category ambiguityThat . . . 

    That is weird. = [deictic noun]

    That show is weird. [=determiner]

    That people like pole-vaulting is weird.[=complementizer]

  • 8/19/2019 Hofmeister Slides

    26/60

    AMBIGUITYRESOLUTION

    Why is ambiguity so important?

    you don’t know how to interpret “that” immediately,and may have to wait a fairly long time before receivingdisambiguating info

    ambiguity makes the computational problem harder

  • 8/19/2019 Hofmeister Slides

    27/60

    AMBIGUITYRESOLUTION

     How do people deal with ambiguity?

    Option #1: Select a default analysis based onsyntactic principles and go with that

  • 8/19/2019 Hofmeister Slides

    28/60

    AMBIGUITYRESOLUTION

    . . . that . . .Analyze as determiner

    Sets expectations for upcoming noun phrase

    Upside: parser always knows what to do

    Downside: it may be wrong!

  • 8/19/2019 Hofmeister Slides

    29/60

    MINIMALATTACHMENT

    The man the woman

    NP NP

    NP

    The man

    the woman

    NP S

    NP

    NP

    Choose the simpler analysis

  • 8/19/2019 Hofmeister Slides

    30/60

    AMBIGUITYRESOLUTION

     How do people deal with ambiguity?

    Option #2: The short-sightedness of thelanguage processing system determines howambiguity is dealt with

  • 8/19/2019 Hofmeister Slides

    31/60

    BACK TO THE DATA

    Tom said that Bill had taken the cleaning out yesterday 

  • 8/19/2019 Hofmeister Slides

    32/60

    John said that he heard Karen wrecked her car yesterday.

    Sentences get harder to process as the dependenciesbetween arguments increase in length (Gibson 1998)

    memory representations decay

    discourse processing interferes with past discourse

    processing

  • 8/19/2019 Hofmeister Slides

    33/60

    AMBIGUITYRESOLUTION

     How do people deal with ambiguity?

    Option #3: The language system strategicallyuses multiple constraints, including context &probabilistic information to quickly resolveambiguity

  • 8/19/2019 Hofmeister Slides

    34/60

    Brown corpus of English77.5% of “that” are complementizers

    11.1% are determiners

    11.5% are demonstrative pronouns

    = context-independent lexical frequencies

    GIBSON (2006)

  • 8/19/2019 Hofmeister Slides

    35/60

    Sentence-initially, however, that is more likely to be adeterminer than a complementizer

    In other words, your analysis of the ambiguous wordthat depends on where you see it

    GIBSON (2006)

  • 8/19/2019 Hofmeister Slides

    36/60

    CONSTRAINT-BASED THEORIES

    On constraint-based views of language processing, humanssolve the computational problem of language by utilizinga number of sources of information to make sense of theinput

  • 8/19/2019 Hofmeister Slides

    37/60

    PRODUCTION

  • 8/19/2019 Hofmeister Slides

    38/60

    SPEECHERRORS

    Errors @ different levels of language processing

    phonological, syntactic, and semantic

    Slips of the tongue

  • 8/19/2019 Hofmeister Slides

    39/60

    SPEECHERRORS

    anticipations: substitutions of upcoming units

    sidewalk➜ widewalk

    table of contents➜ 

    cable of contentsperseverations: repetition of preceding unit

    walk the beach ➜ walk the beak

    addition

    spic and splan; TARGET: spic and span

    deletion

     his immoral soul; TARGET: his immortal soul

  • 8/19/2019 Hofmeister Slides

    40/60

    SPEECHERRORS

    metathesis (aka exchanges / spoonerisms)fill the pool➜  fool the pill

    chimichangas➜ chichimangas

    slippery crags ➜ crippery slags

     Are my keys in the door➜ Are my doors in the key?

  • 8/19/2019 Hofmeister Slides

    41/60

    SPEECHERRORS

    evidence for the psychological reality of phones,morphemes, and syntactic units

    substitution of words & phrases tells us aboutthe organization of meaning

    substituted words tend to be semanticallyrelated

    turn the lights off ➜ turn the lights on

  • 8/19/2019 Hofmeister Slides

    42/60

    SPEECHERRORS

    exchanges only seem to involve elements at thesame level of processing

    sounds and words don’t exchange

    sounds and morphemes don’t exchange

    fill the bucket➜ bill the fucket

    # fill the bucket ➜ buckill the fet

  • 8/19/2019 Hofmeister Slides

    43/60

    SPEECHERRORS

    exchanges only seem to involve elements at thesame level of processing

    sound exchanges rarely (if ever) happen acrossdifferent word position

    hit the ball➜ bit the hall

    # hit the ball➜ hib the tall

    phonemes in onsets exchange with other onsetphonemes, nuclei exchange with other nuclei,etc.

  • 8/19/2019 Hofmeister Slides

    44/60

    How selfish are we as speakers?

  • 8/19/2019 Hofmeister Slides

    45/60

    COMMONGROUND

    Wardlow Lane &Ferreira (2008)

    Q: Would speakersonly use modifierslike big or smallwhen listenercould see both abig and smallobject?

  • 8/19/2019 Hofmeister Slides

    46/60

    COMMONGROUND

    Wardlow Lane &

    Ferreira (2008)some informationwas privileged

    e.g. only speakercould see twohearts mentioned

  • 8/19/2019 Hofmeister Slides

    47/60

    COMMONGROUND

    Wardlow Lane &Ferreira (2008)

    RESULTS: Even iflistener couldn’tsee one elementin the contrastset, speaker wasmore likely to usea modifier

  • 8/19/2019 Hofmeister Slides

    48/60

    COMMONGROUND

    Wardlow Lane &Ferreira (2008)

    LOW SALIENCECONDITION:experiment pointsto the relevantobject to name

    HIGH SALIENCE:reference tocontrasting item

  • 8/19/2019 Hofmeister Slides

    49/60

    COMMONGROUND

    Wardlow Lane &Ferreira (2008)

    Speakers morelikely to usemodifyingdescriptions whenit’s highly salientto them, but notto listener

  • 8/19/2019 Hofmeister Slides

    50/60

    COMMONGROUND

    Speaker needs and sense of salience outweighdemands for communicative success

    speakers were using terms such as “big heart”when listener only saw one heart

  • 8/19/2019 Hofmeister Slides

    51/60

    EGOCENTRICLANGUAGE

    At least in some circumstances, speakers ignoretheir listener(s) perspective

  • 8/19/2019 Hofmeister Slides

    52/60

    LANGUAGE &

    MIND

  • 8/19/2019 Hofmeister Slides

    53/60

    In many western cultures, we talk of spatialrelations with words like “to the left of”, “to theright of”, etc.

    frame of reference: speaker or listener biased

    In other languages, spatial relations can be basedon absolute (i.e. unchanging) features

    Object-centered coordinates: frame of referencebased on items’ “perspective”

  • 8/19/2019 Hofmeister Slides

    54/60

  • 8/19/2019 Hofmeister Slides

    55/60

    Bowerman, Levinson, and colleagues argue thatmany speakers not only don’t use relative framesof reference, they don’t think in terms of relativeframes of reference

    Guguu Yimithirr

  • 8/19/2019 Hofmeister Slides

    56/60

    g(Australia): onlyuse Absoluteframe ofreference

    “There’s anant on your

    south leg”

    Tzeltal(Mexico):absolute frameof referencebased ongeographical

    landmarks

  • 8/19/2019 Hofmeister Slides

    57/60

  • 8/19/2019 Hofmeister Slides

    58/60

    Halligan (2003): all individuals have anegocentric view of space

    Alternative: individuals recruit different framesof reference and language capitalizes uponthese different available systems

  • 8/19/2019 Hofmeister Slides

    59/60

    Experiment showed participant a path a man traveled on Table 1

    Participant turned around and asked to show how the man traveledout of a maze

    Again, Tzeltal overwhelmingly Absolute FoR

    LANGUAGE &

  • 8/19/2019 Hofmeister Slides

    60/60

    LANGUAGE &MIND

    Answering how language influences cognitionturns out to be a very tricky question

    Very difficult to separate culture & experiencefrom language