language - complexity of speech perception Flashcards

1
Q

lecture 2 objectives

A

Understand why speech perception should be difficult

◦ The continuous flow of speech
◦ Coarticulation

Be able to describe how some features of language might aid speech s segmentation (e.g., word frequency, syllable stress)

Be able to describe how the categorical perception of phonemes helps us to recognise phonemes

Be able to describe factors that influence categorical perception of phonemes (e.g., speech rate, “lexical bias”)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

why should speech perception be difficult

A

-because you cant find breaks between words, its like a long stream of speech, between individual words in a sentence
-this isn’t a problem in your first language

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

what is the segmentation problem

A

-how do we segment speech up into its individual words to understand what has been said to us
-we feel like we hear pauses between words to mark boundaries but this is an illusion

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

oronyms
what do they show us

A

a string of words or phrase that sounds the same as another string of words or phrase (a passage of speech you can slice up in more than one way
prove the segmentation problem
eg four candles fork handles , sounds the same
show us that there arent gaps between words, if there were they wouldn’t be difficult to splice up

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

coarticulation - problem

A

the articulation of two or more speech sounds together, so that one influences the other.
-vocal apparatus is constantly preparing for the next phoneme
eg “b” phoneme in “bill” “ball” “able” rob they are all different

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

what are the adv and disadv of coarticulation

A

adv
-allows speech to be faster and more fluent
-gives listener clues as to upcoming phonemes (our brain uses coarticulation to predict what it will hear next

complication
-no definitive acoustical pattern from any phoneme
-speech perception system has to see past these differences to recognise phonemes

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

how can we segment speech
-evidence from the speech perception system

A

-when we co articulate and blend all the sounds in the words , we tend to do it more within words than between words
-so where we have a boundary between two words, phonemes blend together less
-speech perception system sensitive to this as if it hears more coarticulation between phonemes, it will assume that’s within a word not between words

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

50 most commonly used words accounts for more than……..
how does our perception system use this to make intelligent guesses

A

more than half the speech we hear (Miller 1951)
-uses this info to make intelligent guesses around what its just heard

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

how does our speech perception system use word frequency to guess what we hear

A

the more common words (most common words) are used- eg if you hear something that sounds like thing or have it probably has heard it- and it can be used to make predictions, eg four candles , four is more frequent than fork so you would guess four is being said instead of fork candles

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

what is a syllable

A

rhythmic unit of speech

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

what is syllable stress
give examples

A

A stressed syllable has a longer, louder, and higher sound than the other syllables in the word
syllable stress can be strong or weak
REcord noun (a vinyl) stress is on the first syllable
reCORD (verb) recording eg stress is on the last syllable

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

-content words and function words

A

-content words-(open class words) -one of a huge number of words that convey meaning in a sentence i.e nouns, verbs,adjectives,adverbs
-function words(closed class words)-one of a limited number of words that do the grammatical work of language eg determiners and prepositions -the,a,to,in,and, and because

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

content and function words are stressed differently
-how

A

-content words tend to start with a stressed syllable
-function words tend to be unstressed

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

if you hear a _____ syllable it is likely to be at the start of a ______ word
-if you hear a weak syllable it is likely to be either towards the end of a _____ word , or the start of a _____ word
-how does speech perception make use of this

A

-if you hear a strong syllable it is likely to be at the start of a content word
- if you hear a weak syllable it is likely to be either towards the end of a content word, or at the start of a function word
-makes use of this -its called the metrical segmentation startegy

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

evidence for syllable stress being used to segment speech

A

-cutler and butterfield (1992) looked to see whether we used syllable stress to segment speech
-they got people in a lab and played them sequences of words that violated this rule (presented items like conDUCT aSCENTS upHILL), -so has no meaningful context to aid segmentation and has a stress pattern that violates our expectations
-they found that these items reliably induced mishearing in their participants eg participants heard “the doctor sends the pill” “a duck descends some hill”

eg doctor (they heard duct (the stressed syllable) and thought this must be the start of the word (proving syllable stress))

-this wouldn’t happen unless their speech perception system was using information to guide them to segment up speech stream

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

explain categorical perception

A

-humans are better at hearing the differences between categories of sound than variations within a category
-we are sensitive to the differences between b and p phonemes
-but insensitive to the differences between variations of b

17
Q

categorical perception
developmental work from infants -how infants develop their knowledge about how the language they speak works

A

-werker and tees 1984
-english and hindi
hindi has phonetic distinctions english doesnt
-at 6 months old, infants with English speaking parents could make phonetic distinctions present in hindi but not english
-but the ability to do this declined rapidly over the next 2 months
-proves by around 8 months they are starting to ignore the differences in phonemes that aren’t relevant to their language

-kikuyu
-at 2 months infants with kikuyu speaking parents could distinguish between p and b phonemes (kikuyu doesn’t make this distinction) ability to do this lost by around 1 year

  • essentially we are just trying to see at what age to babies stop making phonetic distinctions
18
Q

in the first year of life babies start to understand ____ ______ important to them

A

phonetic differences
-they no longer acknowledge phoneme difference they don’t need to know

19
Q

voice onset time (vibration starts)
b is a _____ phoneme
p is an ______ phoneme

A

voiced
unvoiced

20
Q

in english in adults
[ba] has a __ms voice onset time
[pa] has a __ms voice onset time

A

0
60 (delay vibration of vocal cords by 60 millisceonds)
shows what phoneme it is

21
Q

-ordinary b sounds starts at
0ms and
-ordinary p sounds has a 60ms Voice onset time
-between b and p is anywhere between 0-60ms
how do people react when presented with examples between ba and pa?

A

(eg lisker and abramson 1970)
-the graph shows theres quite a sharp cut where people say its ba or pa
-shows we have sharp categorical perception between two categories

22
Q

discrimination task of ba and pa
-present pairs of stimuli and participants say whether the sounds are same or different on trials
-what are the results of this experiment

A

-10ms vot and 30ms voice onset time - participants tend to say the perceive the sound as Ba
-40 and 60 vot- participants say its pa
-20 ms VOT and 40 ms VOT participants tend to perceive them as different so 20 is ba and 40 pa
-participants find dicriminations across the boundary

23
Q

factors that influence categorical perception

A

-individual differences: different people will categorise voice onset time slightly differently

-speech rate (Summerfield 1981):perceptual system adjusts to fast rates of speech (eg like speeding lecture up) a sound with a short VOT that should be perceived as b will be perceived as p when somebody talks quickly

-lexical identification shift
-if e hear a phoneme that is ambiguous between {b] and [p] we tend to report a phoneme that creates a word
eg a 30ms VOT phoneme will be reported as [b] if its followed by -ook (book)
but the same VOT phoneme will be reported as [p] if it is followed by -ick (pick)

24
Q

categorical perception will vary according to the _____ you speak

A

language
-eg languages that dont distinguish between b and p wont show categorical perception from those phonemes
-categories are something that you have to learn as an infant

25
Q

trained non human animals can show similar effects
explain

A

-kuhl and miller (1978) found similar patterns in chinchillas that had been trained to distinguish 0ms VOT and 80ms VOT phonemes