Growing up Bilingual: One System or Two

1
Growing up BilingualOne System or Two?

• Language differentiation and speech perception in
  infancy

2
Extending the models to Phonology

• ULS (Unitary Language System)
• A common, undifferentiated storage system for all
  languages (Volterra Taeschner)
• In Phonology, Flege (1987) suggests a single
  system with category boundaries influenced by
  each language
• DLS (Differentiated Language Systems)
• Language systems differentiated from the
  beginning (Genessee)
• In phonology, Grosjean (1997). One phonology for
  each language.
• One language dominant
• Research with adult bilinguals showing faster
  phonetic perception of L1 contrasts
  (Sebastian-Galles, et. al).

3
Organization of the Lecture

• Examining the Language Differentiation hypothesis
  with research on language discrimination
• Examining the Language Differentiation hypothesis
  with research on phonetic perception

4
Language Discrimination

• At birth, monolingual infants prefer native over
  non-native continuous speech (Moon, et. al, 1994)
  can discriminate the filtered speech of
  rhythmically distinct languages (Mehler, et .al.,
  1988 Nazzi, et. al., 1998)
• By 4-5 months they not only discriminate
  rhythmically distinct languages, but can
  discriminate their own language from rhythmically
  similar languages, including dialectal variations
  of the same language (Nazzi, et. al., 2000
  Ramus, et. al., 2000)

5
Extending to Bilinguals

• Bosch Sebastian-Galles (1997) tested
  Monolingual Spanish and Monolingual Catalan
  infants on their ability to discriminate their
  native language from the unfamiliar one
• Interesting because both from the same rhythmical
  class
• Still some differences (more vowel reduction in
  Catalan)

6
The Orientation Latency Procedure
Picture of a womans face over each loud speaker.
7
Monolingual 4-month infants orient faster to
native language
Bosch Sebastian-Galles, Cognition, 1997
8
Bilingual infants orient more SLOWLY to one of
their native languages over an unfamiliar language
9
Bilingual infants show no difference in
orientation latency to either of their languages
10
Summary

• Infants of 4 months can discriminate their
  language from a rhythmically similar language,
  and show this by orienting faster to the native
  language
• Spanish-Catalan bilingual infants aged 4-5 months
  discriminate both of their native languages from
  an unfamiliar language.
• But, they show this discrimination of native vs.
  unfamiliar language by responding slower to the
  native language, the opposite pattern as
  monoliguals
• This suggests different organization for speech
  processing even at this early age
• No evidence in the orientation latency procedure
  that bilinguals can discriminate their two
  languages

11
Implications

• On the basis of the Bosch Sebastian-Galles 1997
  work, would suggest that bilinguals might be
  listening to language differently from
  monolinguals
• But no evidence of differentiation of their two
  language

12
Can Bilingual Infants discriminate their two
languages?

• Lack of discrimination of the two languages,
  particularly when they are discriminable by
  monolinguals, is supportive of the ULS
• B SG decided to use a more appropriate task to
  see if bilinguals can discriminate
• Used the HTPP, with a familiarization phase

13
Procedure Famliarization Phase

• First familiarized infants to sentences from one
  of their languages
• Present a flashing light at the centre
• Once the infant looked, the image disappeared,
  and a slightly different image appeared on the R
  or L screen
• After the infant looked toward that, sentence
  presentation began and lasted 28 sec or until the
  infant looked away
• Required the infants to accumulate 2 minutes of
  sustained attention to the sentences (1 min to
  each)

14
Procedure Test Phase

• New sentences
• Half in familiar language, half in the other
  language
• Same procedure as before
• Did the infants choose to look/listen longer to
  the sentences in the new language?

15
Sequence of Experiments

• 1st wanted to make sure the results from the
  Orientation Latency procedure would replicate
  with the Familairization/Switch task
• So first tested monolingual Spanish and Catalan
  infants on their ability to discriminate their
  native language from the other

16
Familiarization/Switch Monolingual 4-monthers
discriminate Spanish vs. Catalan
Bosch, et. al., 2001
17
Control Study

• Worried that recovery in Study 1 could have been
  due to the fact that the sentences were new
  (although seems that was controlled for in the
  last experiment as well)
• And to ensure that there wasnt just spontaneous
  recovery
• Tested infants with only materials from one
  language
• Familiarized to one set of sentences from one
  language, and then tested them with a new set
  from the same language

18
Control Experiment rules out simple recovery to
new sentences
19
Can Bilinguals discriminate their two languages?

• Finally ready to conduct the key experiment
• Familiarized bilinguals to sentences in one
  language
• Tested on new sentences from that language vs.
  new sentences from the other language

20
Bilingual 4 monthers DO discriminate their two
languages!
21
Implications

• These results provide very strong evidence for
  the DLS
• If bilingual infants as young as 2 months can
  discriminate their two languages, hard to argue
  for a ULS
• How about even earlier?

22
ProcedureHigh Amplitude Sucking (HAS)Werker
Burns, in prep.
23
Design

• Used alternating (Cowan) version of HAS
  procedure, but as a test of preference
• Baseline minute to set HA suck value
• Presentation of speech contingent on HA suck
• Collected 10 minutes of sucking, 5 minutes for
  Tagalog and 5 minutes for English, with the
  languages alternating by minute
• Order counterbalanced
• Preference DV Do infants select to listen, i.e.
  deliver more HA sucks to Tagalog or to English?

24
(No Transcript)
25
Preliminary Conclusions Tag/Eng Preference Study

• English infants do show a preference for English
  over Tagalog
• BFLA Tag/Eng newborns do NOT prefer one of their
  languages over the other
• Instead, they choose to listen equally to both
  Tagalog and English, and do so significantly more
  than the English babies listen to Tagalog
• This suggests that both languages are equally
  dominant at birth in the BFLA infant

26
Unanswered questions future research

• Do BFLA infants differentiate their two
  languages, or is there initially one
  undifferentiated language space?
• To test, need to see if BFLA newborns can
  discriminate their two first languages
• Do BFLA infants encode enough detail about
  language to prefer a familiar language over an
  unfamliar one?
• To test, need to see if BFLA newborns will show a
  preference for one of their native languages over
  an unfamiliar language

27
Part IIConsonant Discrimination in BFLA Infants
(with Tracey Burns)

• In previous work we have shown there to be a
  reorganization in consonant discrimination in the
  first year of life
• Look at voicing, in a BFLA population and compare
  to French-only and English-only groups

28
Age changes in non-native discrimination(Werker
Tees, 1984)
29
Vowel Perception in Bilinguals

• Bosch Sebastian-Galles (2003) tested Catalan
  and Spanish-Catalan bilingual infants on their
  ability to discriminate the Catalan only /e/-/E/
  contrast
• At 4 months, both groups discriminated
• At 8 months, only the Catalan infants did
• By 12 months, both groups discriminated again
• Does the data at 8 months indicate confusion
  (ULS) or language dominance?
• DLS by 12 months

30
Extensions to consonants

• Test monolingual and bilingual infants on their
  ability to discriminate a b-p difference that is
  instantiated differently in each of their
  languages
• Do they confuse all three sounds? (ULS)
• Do they choose the category distinction for one
  of their languages (language dominance)
• Or do they use all three categories and maintain
  both boundaries (DLS)

31
VOT in French English

• English has two categories of VOT, short lag, and
  long lag with aspiration
• French, also has two categories of VOT, but they
  are long lead and short lag
• -------1------------------1------------------1----
  ----- Fr. /ba/ Fr./pa/-Eng /ba/
  Eng /pa/

32
Design

• -------1------------------1------------------1----
  ----- Fr. /ba/ Fr./pa/-Eng /ba/
  Eng /pa/
• Habit infants to middle stimulus, test on both
  others with order counterbalanced
• Test English, French, and Fren/Eng BFLA infants

33
Procedure

• Language exposure assessed using an adapted
  version of the questionnaire developed by Bosch
  and Sebastián-Gallés (1997)
• Infants tested in a visual habituation procedure
• Auditory stimuli presented with a checkboard
  display
• Pre and post tests

34
Visual Habituation Procedure
35
Results 6-8 Month Olds
Looking Time (seconds)
36
Results 68 Month Olds

• Both groups of infants dishabituate to ba but
  not significantly to pha
• 6 8 month olds show the same pattern of
  response regardless of home language environment

37
Experiment 2

• Stimuli
• same as Experiment 1
• Procedure
• same as Experiment 1
• Participants
• 1012 month old infants being raised in
  monolingual English or bilingual English/French
  households

38
Results 10-12 Month Olds
Looking Time (seconds)
39
Results 1012 Month Olds

• Infants being raised in English speaking homes
  dishabituate to pha but not to ba
• This is the pattern expected given their native
  boundary and previous research (Werker and Tees,
  1984)
• Infants being raised in English/French homes do
  not show a significant change in looking times to
  either of the two test stimuli
• Data are not uniform and suggest within-group
  differences
• These results suggest that the timecourse, and
  possibly nature, of bilingual phonetic
  representation is distinctly different than that
  of monolinguals

40
Experiment 3

• Stimuli
• same
• Procedure
• same
• Participants
• 14-21 month old infants being raised in
  monolingual English, monolingual French, or
  bilingual English/French households

41
Results 14 Month Olds
Looking Time (seconds)
42
Results 14 Month Olds

• Infants being raised in monolingual homes
  dishabituate to stimuli that cross the category
  boundary in their native language
• Infants being raised in bilingual homes do not
  dishabituate to either stimuli
• Bilingual data are not normally distributed
  some infants appear to be responding as
  monolinguals, while others dishabituate equally
  to both stimuli
• Age does not predict pattern of response
• Suggest that it is not the timecourse that
  differs but rather the nature of the
  representation for those infants dishabituating
  to both stimuli

43
Conclusions

• 6 month olds separate the stimuli into two
  discrete phonetic categories
• Category boundary is the same regardless of
  language input
• 10 month olds from monolingual homes place the
  category boundary in the appropriate location for
  their native language
• 10 month olds from bilingual homes do not appear
  to categorize the stimuli
• This suggests that the timecourse, and possibly
  the nature, of bilingual phonetic representation
  is distinctly different than that of monolinguals

44
Conclusions

• 14 month olds from monolingual homes place the
  category boundary in the appropriate location for
  their native language
• 14 month olds from bilingual homes appear to
  divide into two groups
• Infants who categorize the stimuli as
  monolinguals in one of their two languages
• Infants who maintain both English and French
  category boundaries equally

45

• This suggests that for some bilinguals, it is the
  nature, not the developmental time course, of
  phonetic representation that is distinctly
  different than that of monolinguals

46
Bilingual Infants Only
47
Theoretical Implications

• Is it language mixing at 10-12 months for
  bilinguals? Maybe
• But by 14 months there are 3 distinct patterns
• Some infants show language dominance
• Others show two differentiated systems
• No evidence of language mixing

48
Conclusions

• Taken together, these studies provide pretty
  convincing support against the ULS hypothesis
• There is little evidence of sustained language
  mixing
• Stronger evidence for either DLS or one dominant
  language