A Lecture about…A Lecture about…
Phonetic AcquisitionPhonetic Acquisition
Veronica Weiner
May, 2006
What is phonetic acquisition?What is phonetic acquisition?
Learning Learning what soundswhat sounds are part of a language are part of a language Learning to Learning to categorizecategorize sounds across different speakers and contexts sounds across different speakers and contexts Learning what groups of sounds constitute Learning what groups of sounds constitute wordswords Learning what Learning what prosodic cuesprosodic cues (pitch, intonation, etc.) are part of a (pitch, intonation, etc.) are part of a
languagelanguage
There are There are developmental stagesdevelopmental stages during the first year of life during the first year of life Phonetic acquisition continues until Phonetic acquisition continues until adulthoodadulthood
Available Available toolstools to language learners to language learners Statistical regularities in the auditory inputStatistical regularities in the auditory input Motherese, social interactionMotherese, social interaction Multimodal statistics (lip reading)Multimodal statistics (lip reading) ““Learning by doing” (babbling)Learning by doing” (babbling) Specialized neural apparatus? Specialized neural apparatus?
Some research questionsSome research questions
How different sound features (prosody, transition How different sound features (prosody, transition probabilities, multimodal correlations) interact at probabilities, multimodal correlations) interact at different stages of developmentdifferent stages of development
The relationship between production and The relationship between production and perceptionperception
Brain activity related to phonetic acquisitionBrain activity related to phonetic acquisition
Differences between L1 and L2Differences between L1 and L2
and more…and more…
What I will cover (What I will cover (Agenda!Agenda!))
Part I: Part I: “Phacts about Phonemes”“Phacts about Phonemes” Some properties of auditory input that I think are Some properties of auditory input that I think are
interesting / useful / good to knowinteresting / useful / good to know
Part II: Part II: Some key results in the fieldSome key results in the field Infant timelineInfant timeline 5 important papers5 important papers
Part III: Part III: Other interesting papersOther interesting papers The role of social interactionThe role of social interaction Babbling in sign languageBabbling in sign language Phonetic acquisition in blind childrenPhonetic acquisition in blind children
Part I: Part I:
“Phacts” about “Phacts” about PhonemesPhonemes
a.k.a.a.k.a.
“Phun” with Phonemes“Phun” with Phonemes
The components of languageThe components of language
PhonemesPhonemes
Phonetic unitsPhonetic units are combined into are combined into phonemic phonemic
categories categories or justor just phonemesphonemes
Example - Example - /r//r/ and and /l//l/ are are phonetic unitsphonetic units that that compose :compose :
different phonemesdifferent phonemes in in EnglishEnglish
thethe same phonemesame phoneme in in Japanese.Japanese.
Each language uses about Each language uses about 4040 phonemes phonemes
Properties of phonemes: across Properties of phonemes: across languageslanguages
““General auditory perceptual General auditory perceptual abilities provided the ‘basic abilities provided the ‘basic cuts’ that influence the choice cuts’ that influence the choice of sounds for the phonetic of sounds for the phonetic repertoire of the world’s repertoire of the world’s languages”languages”
(Kuhl, NRN, (Kuhl, NRN, 2003) 2003)
Infants are more sensitive to phonemic boundaries than other sounds at an “equivalent distance,” even in languages they have not heard.
(Eimas, 1988) (Eimas, 1988)
Image source: http://www.eskimo.com
Phonemes, phonetic units, who Phonemes, phonetic units, who cares?cares?
To distinguish all the words in a language, a person To distinguish all the words in a language, a person must be able to distinguish all the phonemes of must be able to distinguish all the phonemes of that languagethat language
Why? Phonemes are defined as the minimum Why? Phonemes are defined as the minimum element of contrast between wordselement of contrast between words
‘‘rake’ vs. ‘lake’ rake’ vs. ‘lake’
Coding the phonemes of speech data is a difficult Coding the phonemes of speech data is a difficult problem for computersproblem for computers
Telling phonemes apartTelling phonemes apart
Image source: Wikipedia entry for Formant
Telling phonemes apartTelling phonemes apartAre auditory properties ambiguous?Are auditory properties ambiguous?
Figure source: Kuhl P. Nat Rev Neuro, 2004.Original figure: Peterson & Barney, 1952
Telling phonemes apartTelling phonemes apartAuditory properties: maybe less ambiguous than we Auditory properties: maybe less ambiguous than we
thought thought (at least for vowels).(at least for vowels).
Figure source: Hillenbrand et al 1995
Hillenbrand’s group added sound duration and formant contours.
Higher dimensional space of auditory features.
Telling phonemes apartTelling phonemes apartDistinguishing most phonemes across speakers and Distinguishing most phonemes across speakers and
contexts is still a very difficult problem, though.contexts is still a very difficult problem, though.
100 200 300 400 500 600
100
200
300
400
500
600
700
800
Source: Tony Ezzat, CBCL
50 100 150 200 250 300 350 400 450 500 550
100
200
300
400
500
600
700
800
Male (Adler) Female, high voice, a bit slower (Mary Pat)
Open research questionOpen research question
What What areare the the relevant featuresrelevant features, then, for , then, for acquiring phonemes (or distinguishing them with acquiring phonemes (or distinguishing them with a computer)?a computer)? AuditoryAuditory Articulatory (production-related)Articulatory (production-related)
Part II:Part II:
Some Key Results Some Key Results in Phonetic Acquisitionin Phonetic Acquisition
since 1983since 1983
1. Acquisition over the 11. Acquisition over the 1stst year: an overview year: an overview
Figure source: P Kuhl. Nat Rev Neurosci. 04.
2. Acquisition in older children2. Acquisition in older children
Hazan and Barrett (2000) showed that phonetic Hazan and Barrett (2000) showed that phonetic acquisition occurs even after the age of 12.acquisition occurs even after the age of 12.
They used a They used a minimal pairminimal pair procedure: procedure:
2. Acquisition in older children2. Acquisition in older childrenHazan and Barrett (2000)Hazan and Barrett (2000)
A sound is synthesized that “morphs” in equal steps between Sue and Shoe. A sound is synthesized that “morphs” in equal steps between Sue and Shoe. The The cuescues that distinguish the sounds can be varied independently or together. that distinguish the sounds can be varied independently or together. Subjects are tested on their choice of Sue or Shoe at each step.Subjects are tested on their choice of Sue or Shoe at each step.
Sue vs. shoe
- Friction frequency
- F2 Transition
% S
ue
Bottom figure: P Kuhl. NRN. 04.
2. Acquisition in older children2. Acquisition in older childrenHazan and Barrett (2000)Hazan and Barrett (2000)
Slope of psychometric function
Children 6-12 vs. Adults
Adult perception is significantly more
“categorical” than 12 y.o.’s
Different 50% equivalence point in children and adults
3. Babies can distinguish 3. Babies can distinguish fewerfewer sounds as sounds as they agethey age
From: Werker and Tees (1983)From: Werker and Tees (1983)
Thompson (Salish) is a Native Indian language spoken in British Columbia.
Thompson phonemes are not distinguishable by English speaking adults.
They are distinguishable by babies.
Experiment 1
3. Babies can distinguish 3. Babies can distinguish fewer fewer sounds as they sounds as they ageage
From: Werker and Tees (1983)From: Werker and Tees (1983)
Experimental procedure: Conditioned Head Turn
Babies are conditioned to look at the speaker when there is a sound change by getting visual reinforcement for correct head turns
Procedure works only in around 25% of babies
Photo source: Werker lab website
3. Babies can distinguish 3. Babies can distinguish fewerfewer sounds as sounds as they agethey age
From: Werker and Tees (1983)From: Werker and Tees (1983)
Cross sectional data: 10-12 subjects, M&F approx split, for each bar in figure.
Longitudinal data: 3 M, 3 F
Experiment 2
•Tested more ages
•Tried Salish and Hindi
•Cross sectional data
•Longitudinal data
4. Babies sensitive to frequency distributions in 4. Babies sensitive to frequency distributions in soundssounds
From: Maye, Werker, and Gerken (2002)From: Maye, Werker, and Gerken (2002)
•Synthetic sounds ranging from [da] to [ta].
•Training set 1: There are more sounds on the ends near [da] and [ta] (bimodal distribution).
•Training set 2: Most sounds are in the middle (unimodal distribution).
4. Babies sensitive to frequency distributions in 4. Babies sensitive to frequency distributions in soundssounds
From: Maye, Werker, and Gerken (2002)From: Maye, Werker, and Gerken (2002)
Experimental procedure: Experimental procedure: Looking time experimentLooking time experiment comparing alternating and non-alternating stimulicomparing alternating and non-alternating stimuli
Babies are 6-8 months old.Babies are 6-8 months old.
Results:Results:
Babies look longer at [ta] and [da] pairs after they have been Babies look longer at [ta] and [da] pairs after they have been trained on the trained on the bimodal distributionbimodal distribution..
4. Babies sensitive to frequency distributions in 4. Babies sensitive to frequency distributions in soundssounds
From: Maye, Werker, and Gerken (2002)From: Maye, Werker, and Gerken (2002)
Interesting corollary:Interesting corollary:
After hearing After hearing unimodalunimodal stimuli, babies actually stimuli, babies actually discriminate [ta] and [da] discriminate [ta] and [da] worse worse than most infants their than most infants their ageage
Does this mimic what happens when infants hear a Does this mimic what happens when infants hear a unimodal distribution in their unimodal distribution in their ambient languageambient language??
Only 6 blocks of 16 training sounds were used hereOnly 6 blocks of 16 training sounds were used here
5. Using statistical information to learn word 5. Using statistical information to learn word boundariesboundaries
From: Saffran, Aslin, and Newport (1996)From: Saffran, Aslin, and Newport (1996)
One key problem in word segmentation:One key problem in word segmentation:
Strategies:Strategies:
- Statistical ‘pret- Statistical ‘pretty baty baby’ transitional probabilitiesby’ transitional probabilities
- Prosodic (word stress)- Prosodic (word stress)
Image source: P Kuhl. NRN. 04.
5. Using statistical information to learn word 5. Using statistical information to learn word boundariesboundaries
From: Saffran, Aslin, and Newport (1996)From: Saffran, Aslin, and Newport (1996)
Can Can 8 month old babies8 month old babies extract transitional probabilities from extract transitional probabilities from synthesized speech that contains no breaks, pauses, stress synthesized speech that contains no breaks, pauses, stress differences, or intonations?differences, or intonations?
Experimental procedure: a Experimental procedure: a Looking TimeLooking Time experiment experiment
Babies hear 2 minutes of speechBabies hear 2 minutes of speech Looking at a light elicits a ‘word,’ ‘part word,’ or ‘non word’ on Looking at a light elicits a ‘word,’ ‘part word,’ or ‘non word’ on
repeatrepeat Looking time is measuredLooking time is measured
5. Using statistical information to learn word 5. Using statistical information to learn word boundariesboundaries
From: Saffran, Aslin, and Newport (1996)From: Saffran, Aslin, and Newport (1996)
Training:
pabikugolatupabikudaropi
Testing:
word part-word
pabiku kudaro
Figure source: P Kuhl. NRN. 04.
6. Prosody vs. statistics to segregate 6. Prosody vs. statistics to segregate words words From: Johnson and Jusczyk (2001)From: Johnson and Jusczyk (2001)
Replicated Saffran et al.’s resultsReplicated Saffran et al.’s results Added intonation and stress cues Added intonation and stress cues
90% of English words stress the first syllable90% of English words stress the first syllable
When statistics says, “isn’t a word,” but prosody says “is a When statistics says, “isn’t a word,” but prosody says “is a word,” who wins?? word,” who wins??
Experiment 1: Replication of Saffran et. alExperiment 1: Replication of Saffran et. al Experiment 2: Part-words had a stressed first syllableExperiment 2: Part-words had a stressed first syllable Experiment 3: Part-words were coarticulated (spoken together), Experiment 3: Part-words were coarticulated (spoken together),
while statistical words were not.while statistical words were not. Experiment 4: Control: statistical words were coarticulated, part Experiment 4: Control: statistical words were coarticulated, part
words were not.words were not.
6. Prosody vs. statistics to segregate 6. Prosody vs. statistics to segregate words words From: Johnson and Jusczyk (2001)From: Johnson and Jusczyk (2001)
Experiment 1: Replication of Saffran et. alExperiment 1: Replication of Saffran et. alExperiment 2: Part-words had a stressed first Experiment 2: Part-words had a stressed first syllablesyllableExperiment 3: Part-words were coarticulated Experiment 3: Part-words were coarticulated (spoken together), while statistical words were (spoken together), while statistical words were not.not.Experiment 4: Control: statistical words were Experiment 4: Control: statistical words were coarticulated, part words were not.coarticulated, part words were not.
In 8 month olds, prosody wins. In 8 month olds, prosody wins.
6. Prosody vs. statistics to segregate 6. Prosody vs. statistics to segregate words words From: Johnson and Jusczyk (2001)From: Johnson and Jusczyk (2001)
To what extent do these results generalize toTo what extent do these results generalize to
Different Different developmental stagesdevelopmental stages?.?.
Different Different input conditionsinput conditions??
Part III: Odds and Part III: Odds and EndsEnds
(Other interesting (Other interesting results)results)
Social effects on phonetic acquisitionSocial effects on phonetic acquisition
Kuhl et al. (03) show that Kuhl et al. (03) show that English speaking 9 month English speaking 9 month old babies can learn old babies can learn Mandarin phonetic Mandarin phonetic contrasts from a live person contrasts from a live person but not a video of that but not a video of that person.person.
Goldstein et al. (03) show that Goldstein et al. (03) show that 8 month old babies 8 month old babies receiving positive feedback receiving positive feedback from their mothers from their mothers vocalized more than yoked vocalized more than yoked controls.controls.
Figure source: P Kuhl. NRN. 04.
Phonetic acquisition in the blindPhonetic acquisition in the blind
Evidence of multimodal phonetic acquisition.Evidence of multimodal phonetic acquisition.
Rowland (83) shows that blind children make fewer vocalizations but Rowland (83) shows that blind children make fewer vocalizations but begin babbling at the same time as sighted children.begin babbling at the same time as sighted children.
Slide source: N Pitchford lecturenotes, Nottingham University
Babbling in sign language learnersBabbling in sign language learners
Pettito et al. (91) Pettito et al. (91) show that deaf show that deaf babies exposed to babies exposed to sign language sign language make ten times make ten times more “babbling” more “babbling” hand gestures than hand gestures than controls.controls.
SummarySummary
What we have learnedWhat we have learned
1.1. Languages are distinguished in part by their sets of phonemesLanguages are distinguished in part by their sets of phonemes
2.2. Discriminating phonemes is a challenging computational problemDiscriminating phonemes is a challenging computational problem
3.3. Phonetic learning and performance proceeds in stages, beginning in early infancyPhonetic learning and performance proceeds in stages, beginning in early infancy
4.4. Phonetic learning continues into adulthoodPhonetic learning continues into adulthood
5.5. Babies can distinguish fewer sounds as they ageBabies can distinguish fewer sounds as they ageFrom: Werker and Tees (1983)From: Werker and Tees (1983)
6.6. Babies are sensitive to frequency distributions in soundsBabies are sensitive to frequency distributions in sounds From: Maye, Werker, and Gerken (2002) From: Maye, Werker, and Gerken (2002)
7.7. Statistical information can be used to learn word boundariesStatistical information can be used to learn word boundaries From: Saffran, Aslin, and Newport (1996) From: Saffran, Aslin, and Newport (1996)
8.8. Prosody can be compared with statistics to segregate words Prosody can be compared with statistics to segregate words From: Johnson and Jusczyk (2001)From: Johnson and Jusczyk (2001)
9.9. Phonetic learning is affected by social contextPhonetic learning is affected by social context
10.10. Some facts about phonetic acquisition in blind and deaf learnersSome facts about phonetic acquisition in blind and deaf learners
Thanks!Thanks!