Topic 10: Early Speech Perception

Question 1

What needs learning?

Answer 1

Phones, Phonemes, Prosody

Question 2

What are phonemes?

Answer 2

The smallest unit of sound that is recognizable as speech rather than as random noise, for example ‘ba’ and ‘pa’.

Question 3

What is prosody?

Answer 3

The rhythm of speech

Question 4

What are phones?

Answer 4

serves as the basic unit of phonetic speech analysis. Phones are generally either vowels or consonants.

Question 5

How do we measure phones, phonemes, prosody?

Answer 5

24 week in womb auditory system is functional
startle response, ultrasound, electrocardiogram, magnetocardiography

ultrasound = less precise
magnetocardiography = more precise

Question 6

What speech qualities are learned in utero?

Answer 6

Foetus able to receive auditory signals
Prosody of language and voice qualities, not phonemes though.
Sounds in womb is muffled, maternal voice transmits well

Question 7

Can frequent exposure to speech sounds lead to familiarity?

Answer 7

Habituation - leads to familiarity
There is evidence of maternal voice familiarity from fetal movement
Evidence of maternal speech characteristics (rhythm, intonation) from fetal heart rate

Question 8

How do we test if neonates are responsive to acoustic noise that they hear in the womb?

Answer 8

By looking at contingency learning
We can look at non-nutritive sucking rate to monitor adaptive changes
(De caspar) saw adaptive changes with familiar passage which suggests familiarity to rhythm of native language in utero.

Question 9

When do speech sounds in the womb change from familiarity to long lasting learning?

Answer 9

When there is perceptual learning
This is long lasting changes in auditory perception due to experience/practice.

Question 10

What is dishabituation?

Answer 10

The process by which a neonate switches from habituation to perceptual learning.

Question 11

What is a study on dishabituation?

Answer 11

Study on English vs Japanese vs Dutch language (Nazzi et al)

English neonates heart rate increased when they heard Japanese tell us they dishabituated as they can tell apart English and Japanese as there’s a different rhythmic structure to English.
Heart rate stayed the same when hearing Dutch as it has the same timing properties.

Question 12

Why is it really hard for a baby to segment speech and find the fundamental units (i.e. phonemes, syllables, words) in speech streams?

Answer 12

Boundaries between sounds and words are hard to spot as phonemes are acoustically merged
Speech sounds are continuous and sequential
Unfamiliar speech is very hard to separate into units of sound (words)

Question 13

How can we solve the problem of speech segmentation in babies?

Answer 13

We can use categorical perception - sounds that lie along a continuum, perceived as belonging to one category or another.
- there’s either voiced and voiceless sounds e.g. ‘s’ and ‘z’
- or a difference in VOT (voice onset time) helps us differentiate phonemes e.g. b and p. (Eimas et al.)

Question 14

Re infants hardwired for categorical perception? - does it support nativism?

Answer 14

Infants are not hardwired for categorical perception. They are not able to learn categories in utero as conditions in utero make it hard to categorise phonemes so not possible. Therefore, does not support nativism.

Question 15

How can we test infant’s ability to detect speech discrimination?

Answer 15

Headturn Preference procedure
This detects sounds, words, sentences

Question 16

How can Infant attention can be measured?

Answer 16

looking time, head turning, heart rate and the rate of sucking on a dummy.

Question 17

What does the headturn preference procedure show?

Answer 17

• shows preference for one audio vs another
• shows discrimination of change in a continuous sound stream (ba ba, da, da)
• used to test dishabituation after habituation

Question 18

What study shows that infants eventually dishabituate to foreign languages?

Answer 18

Detecting native speech experiment: Werker and Tees
- used the Head Turn Procedure
- Found that younger infants perceived both contrasts
- Fewer older infants perceived non-native contrasts
- So overtime become less sensitive to sounds from languages they are not routinely exposed to.

Question 19

When can infants “tune in” to native language? - What’s a study to support this?

Answer 19

Can tune into phonemes by 10-12 months
Kuhl et al - Test infants on ‘r’ vs ‘l’ contrast
Americans use this contrast
But this distinction is very ambiguous with Japanese.
Found that younger infants of both countries can discriminate this contrast
But, older American infants can decipher, not Japanese.
Shows that they are becoming aware of distributional frequency of speech. (how often sounds show up)

Question 20

What are the two types of statistics in native speech?

Answer 20

Distributional probability - infants capture how often some phonemes and syllables ‘show up’ in their native language.

Transitional probability of syllables that co-occur together - infants can capture what speech sounds/syllables are likely to be heard together (in combination). e.g. pretty baby ‘tee and bay’ is the transition of syllables.

Question 21

Are statistics enough for a child to learn a language?

Answer 21

Adults also play an important role in language learning.
Especially through the use of motherese, communication such as eye gaze and joint attention.

Question 22

Do infants actually need people to learn speech contrasts?

Answer 22

Kuhl et al. Tested older American infants lean non-native Mandarin.
They used the head turn procedure to test Mandarin speech.
They were not be able to tell the difference as they were being taught through audio/TV.
However, when taught by a real person they were able to discriminate the Mandarin contrasts.
This shows how infants learn language in a social world with lots of cues to help pick up frequency and combination (statistics) of speech.