Speech Perception in Infants Peter D. Eimas et al. (1974)

1
Speech Perception in InfantsPeter D. Eimas et
al. (1974)

• Ricardo Tabone LIN7901

2
Not exactly like this.
3
Purpose

• Test speech signal discrimination in infants
• Age groups 1 month old and 4 months old
• Compare infant perception to adult perception
• Type of stimulus selected VOT continuum

4
Why VOT?

• Universal continuum
• Boundaries are not arbitrary
• There are three universal categorical boundaries
  along VOT
• Intuitively
• Negative VOT Pre-voicing (e.g., Thai)
• Short lag VOT (all languages)
• More Positive VOT (most languages)
• English uses only the last two categories (p)/(b)

5
Figure 1 /ba/ vs. /pa/
6
How do we know what the baby knows?

• Babies have a sucking reflex.
• Babies react to a new stimulus by accelerating
  the sucking reflex
• They attached a pressure sensor to a blind nipple
  to measure sucking responses
• However, after a few minutes, habituation sets in
  and sucking rate decreases
• Important calculate the baseline sucking rate
  such that its possible to notice an acceleration

7
How is it related to ERPs?

• In a nutshell
• Sucking rate increase mimics MMN
• Finding a baseline sucking rate mimics averaging
  out the random brainwaves in EEG recording

8
Experiment

• The babies were placed on their backs
• A speaker behind the babys head delivered sounds
  at variable amplitudes
• Amplitudes was controlled by sucking rate slower
  sucking rate? louder stimulus
• Need to counteract natural habituation
• The first stimulus in a pair was repeated for
  about 5 minutes. Then the second one for 4
  minutes.

9
Stimuli /ba/ to /pa/

• VOTS
• -20, 0, 20, 40, 60, and 80ms
• Perceived as /b/ in English -20, 0, 20
• Perceived as /p/ in English 40, 60, 80
• Eimas had already discovered that babies could
  distinguish differences of 20ms in VOT
• So, these differences were adequate for phonemic
  discrimination

10
Stimuli conditions

• 20D Pair of stimuli with VOT 20ms apart.
• 20ms (/ba/) and 40ms (/pa/)
• Different phonemic categories
• 20S Pair of stimuli with VOT 20ms apart.
• -20ms (/ba/) and 0ms (/ba/)
• 60ms (/pa/) and 80ms (/pa)
• Same phonemic category
• 0 Pair of identical stimuli (Control condition)
• Selected randomly for all subjects

11
Subjects

• 20D 8 subjects of each age group
• 20S 8 subjects of each age group
• 0 10 subjects of each age group
• Total 52 poor little babies!

12
Results

• 20D Significant increase of sucking rate after
  stimulus shift Discrimination of phonemic
  category
• 20S No discrimination
• Non significant decrease of sucking rate after
  stimulus shift (4-month olds)
• Non significant increment of sucking rate after
  stimulus shift (1-month olds)
• 0 Natural habituation progress.

13
Figure 2 Sucking Responses
14
Figure 3 Changes in Response Rate
15
Conclusion

• Infants as young as 1 month have
• Categorical discrimination
• Even with no linguistic experience!
• This ability must be biologically encoded
• However, not capable of discriminating within
  categories.

16
The End
17
The Discrimination of Foreign Speech Contrasts by
Infants and AdultsSandra Trehub et al. (1976)
18
Experiments

• Infants 5-17 weeks old
• Contrasts
• pa x pã (French/Polish) (Experiment I)
• za x ra (Czech) (Experiment II)
• Adults (Experiment III)
• Contrasts
• ri x li (English)
• za x ra (Czech)

19
Experiments

• Experiments I and II
• 20 subjects for each experiment
• No babies from French/Polish homes (Experiment I)
• No babies from Czech/Slavic homes (Experiment II)
• Stimuli
• 4 different recordings of each sound on a tape
• Stimulus length 500ms
• ISI 500ms (1s stimulus length)
• pa x pã recorded by French speaker
  (Experiment I)
• za x ra recorded by Czech speaker (Experiment
  II)

20
Experiments

• Experiments I and II
• Design
• Control Subjects no sound change
• Experimental subjects sound change after x
  minutes
• x time at which sucking rate drops by 1/3 for 2
  mins
• Sounds were played on a continuous loop
• (1 _ 2 _ 3 _ 4 _ 1 _ etc)
• Until the moment of sound change

21
Experiment I and II Results

• Results
• Infants can discriminate foreign language
  contrasts
• Figure 1
• Percentage of maximum sucking rate (y-axis)
• Time, aligned by sound change (x-axis)
• Significant different between control and
  experimental subjects

22
Figure 1 Categorical Discrimination
23
Experiment III

• Subjects
• 10 university psychology students (U of T)
• Native speakers of English
• Some exposure to a foreign language (not Czech)
• Stimuli
• 16 trial series
• Each trial 10 sounds randomly separated by
  1,2,3,4 or 5 seconds.
• li x ri recorded by English speaker
• za x ra recorded by Czech speaker (same as
  Exp. II)

24
Experiment III

• Stimuli
• Sound change at random position (gt 3rd position)
• E.g.
• AAAAAAAAAA (no change)
• AAABBBBBBB (change at 3)
• BBBBAAAAAA (change at 4)
• BBBBBBBBBB (no change)
• Inter-Trial Interval 10 seconds

25
Experiment III

• Procedure
• Indicate whether there was a sound change
• Indicate confidence level (1 to 4)
• 1 guessing 4 very sure
• Results
• d .83 for Czech
• Confidence 2.58 (change) 2.50 (no-change)
• d ? for English
• Confidence 4.00 (change) 3.92 (co-change)

26
Results English and Czech
27
Conclusion

• Infants 5-17 weeks old have
• Categorical discrimination of foreign contrasts.
• Adults
• Confused by foreign contrast.
• Above chance discrimination, but babies are
  better.
• Infants must be using
• low level auditory perception
• Perception and production develop differently

28
The End