Aging and Nativelikeness in Second Language Acquisition

1
Aging and Nativelikeness in Second Language
Acquisition

• TuBBS
• June 28, 2004
• David Birdsong
• University of Texas
• birdsong_at_ccwf.cc.utexas.edu

2
Session 2Nativelikeness in L2A research
Conceptual, theoretical, and pragmatic
considerations
3
Determining the upper limits of L2A

• Recognize limits of current theories practices
• Fill descriptive voids w/r/t outcomes
• Fill descriptive voids w/r/t factors
• Relate outcomes and factors to models of
  acquisition and representation

4
The native speaker criterion

• Cook (1997)
• Grosjean (1998)
• Flege (2002)
• modulo level of analysis nativelikeness is an
  impossibility methodologically /
  epistemologically inappropriate construct

5
Justifying Nativelikeness in L2A Research

• GOOD CRITERION The potential, or upper limits
  of the learner can be meaningfully measured
  against the yardstick of the native speaker.
• GOOD METHODOLOGY Experimental method calls for
  control subjects gt group comparisons but also to
  identify individuals who perform like natives
  (central tendency/SD, range).
• GOOD METHODOLOGY avoid presumptions of native
  norms.

6
Why Nativelikeness Matters in L2A Theory An
Illustration

• Indeterminacy in end-state L2 grammars

7
Types of non-nativelike L2 end state grammars

• Incompleteness
• Divergence
• UG non-compliance
• Non-native optionality
• Indeterminacy

8
Types of non-nativelike L2 end state grammars

• Incomplete grammar
• lacks some property P of the target grammar,
  e.g., absence of wh-movement in L2 English (wh-
  in situ only John did what?).
• Divergent grammar
• property P is instantiated, but its
  representation is not consistent with the
  representation of that property in the target
  grammar, e.g., divergent parameter settings for
  passivization.
• (Languages vary w/r/t constraints on
  passivization of NP complements, e.g., indirect
  object in L2 French (Marc a été offert le
  cadeau) cf. English Mark was given the present.)

9
Types of non-nativelike L2 end state grammars

• UG non-compliant grammar
• diverges from UG principles or possible
  parameter settings, e.g., structure independence
  (add affix Q to third orthographic word of
  utterance for emphasis Smith Tsimplis
  Epun).

10
Types of non-nativelike L2 end state grammars

• Optionality coexistence of multiple variants of
  construction K with 1) same lexis and 2) same
  meaning, e.g., English grammar
• I expect that I will win / I expect I will win
• I believe that I will win / I believe I will win
• Non-native optionality
• I expect that I will win / I expect I will win /
  
• I expect to win
• I believe that I will win / I believe I will win
  /
• I believe to win.

11
Types of non-nativelike L2 end state grammars

• Indeterminacy Operationalized as intuitions for
  grammaticality that are not clear (Bley-Vroman)
  or that are unstable across Time 1-Time 2
  comparisons (Johnson et al.) or that reflect
  guesswork rather than knowledge (Johnson et al.).
  
• Indeterminacy Probabilistic versus rule-governed
  representations (Bley-Vroman, Johnson et al.)

12
Indeterminate L2 grammars

• Bley-Vroman (1989, p. 45) L1 grammars are
  deterministic (late) L2 grammars are
  indeterminate.
• The knowledge underlying non-native speaker
  performance may be () a different sort of formal
  object from the systems thought to underlie
  native speaker performance.
• Such fundamental differences between the
  knowledge systems produced in first and foreign
  language acquisition suggest that the same
  cognitive learning system does not give rise to
  them both.

13
Johnson et al. (1996)

• Native speakers tend to give highly consistent
  judgments, which suggest that the underlying
  system itself is rule governed () this
  consistency does not appear to be a
  characteristic of adult learners. (p. 337)
• The grammar underlying unstable judgments is
  not the same kind of formal object as that formed
  by child (L1) learners. (p. 335)

14
Johnson et al. (1996)

• 10 native Chinese speakers 10 native English
  controls
• Age of first exposure (formal) to L2 English
  11-16 years mean length of formal training 9.5
  years
• Age of arrival (AoA) in US 18-32 years mean
  24.75 years
• Chronological age 25-40 years mean 31 years
• Immersion in L2 English 5-11 years mean 6.45
  years

15
Johnson et al. (1996)

• Time 1 - Time 2 discrepancies in binary judgments
  (14-day interval)
• Sample items
• Can ride Annie a bicycle?
• Last night the books falled off the shelf.
• Josh lets his kids to watch TV.
• Ryan called Krissy for a date up.
• Instability is observed for L2ers but not among
  native controls.

16
Adams Ross-Feldman (2003)
17
Adams Ross-Feldman (2003)

• Time 1 - Time 2 (7-10 day interval) discrepancies
  in scalar judgments
• Ss at end state (min LOR 10 years)
• Sample items Items from Johnson et al. (1996),
  PLUS
• He was one of those lucky people who loved their
  job.
• Danny wanted Becky and I to go to the game with
  him.
• Which do you wanna film? which of the actors

18
Adams Ross-Feldman (2003)

• Time 1 - Time 2 discrepancies are observed in
  scalar judgments for natives, early bilinguals,
  and late bilinguals.
• Late bilinguals show the least instability
  (across all items).

19
Adams Ross-Feldman (2003)

• Instability is observed on all types of items.
• Late bilinguals differ from early bilinguals and
  natives in having more stable judgments for items
  not used by Johnson et al. (1996), i.e.
• He was one of those lucky people who loved their
  job.
• Danny wanted Becky and I to go to the game with
  him.
• Which do you wanna film? which of the actors

20
Native learner grammars are both unstable . . .

• Adams Ross-Feldman (2003)
• - Overall, rate of indeterminacy is comparable
  for natives and learners.

21
Cumulative T1-T2 differences mean NS 43 mean
NNS 47
22
but NS ? NNS in terms of loci of instability

• Adams Ross-Feldman (2003)
• - Overall, rate of indeterminacy is comparable
  for natives and learners.
• - BUT Stability is greater in squishy
  (peripheral, normative, prescriptive) areas of
  grammar among late L2ers than among natives or
  early L2ers.

23
Instability and Nativelikenessdeserving of
(further) study

• Provisional results gt future (replication)
  studies
• General/Theoretical Instability is a
  characteristic of both native and learner
  grammars gt L1 and L2 not different formal
  objects in this respect.
• - Specific/Descriptive/Methodological
  Instability is a local phenomenon, its incidence
  and degree varying by sentence type token.
  Diagnosis of (non)nativelikeness follows
  accordingly.

24
Instability and Nativelikenessdeserving of
(further) study

• Provisional results gt future (replication)
  studies
• Specific/Theoretical Loci of instability vary
  with AoA.
• - General/Theoretical Sorace, 1996 (385-386)
  Observed indeterminacy may correlate with
  sophistication in L2 (L1 as well?).

25
Age-related determinacy?Hudson Newport (200x
pc)

• W/R/T determiner distribution in MAL
• Children Probabilistic input gt probabilistic
  output
• Adults Probabilistic input gt
• deterministic output

26
Where NOT to go with nativelikeness in L2A
research
27
Where NOT to go with nativelikeness in L2A
research

• (1) IL-logic of causality Observed
  nativelikeness in L2A implies access to Universal
  Grammar
• gt In general, similar ends (linguistic behaviors)
  do not imply similar means.
• gt Reliance on UG must be motivated (minimally) by
  a logical problem impoverished input must be
  demonstrated.
• gt Not all features of a grammar can be subsumed
  under UG umbrella.

28
Where NOT to go with nativelikeness in L2A
research

• (2) Abuse of native criterion
• Late L2 learners must resemble L1ers in every
  respect (to falsify CPH/L2A)
• Failure to attain nativelikeness in all respects
  implies
• a) different learning mechanisms in L2A
• b) inadequacy of L2A learning mechanism
• c) confirmatory evidence of CP

29
Nativelike L2 Nativelike bird song
What is of interest isthe development ofthe
exact species-specific behavior. an
individual birdmust sing exactly in the way that
other birds of that specific species
sing. Hyltenstam Abrahamsson (in press)
30
Is the learning mechanism flawed if these types
of behaviors are observed among L2ers?

• Example Tense/Aspect in English (esp. German
  L1)
• I have studied for the test last night
• I have studied for the test
• I studied for the test last night
• Example Case as a function of register in
  English
• Who did you see? (OK in all registers except
  highly formal writing / declamation)
• ? Whom did you see? (precious/affected in all
  registers except highly formal writing /
  declamation)

31
Is the learning mechanism flawed if these types
of behaviors are observed among L2ers?

• Example Gender as a function of register in
  French
• Standard French espèce (species type/kind
  of) feminine
• Popular French espèce (type/kind of)
  feminine or masculine depending on the noun
  it is a determiner for
• Cest une fem espèce de poisson masc Its a
  species of fish (standard and popular)
• Cest un masc espèce de poisson masc Its a
  kind of fish (popular)
• Cest unefen espèce de poisson masc Its a
  kind of fish (standard)

32
Is the learning mechanism flawed if these types
of behaviors are observed among natives?

• Examples of errors Romance gender
• Romance auxiliary choice
• Romance past participle agreement
• Germanic case
• Example English transitivity (syntax of
  complementation)
• LAY (transitive) LIE (intransitive)
• SET (transitive) SIT (intransitive)
• Example In any language, failure to attain
  subtle command of register or rhetoric of
  persuasion.

33
Loss of ability to learn? Vindication of CP? Or
shibboleth (diagnostic of non-nativeness)?

• gt Comparisons with natives shouldnt be carried
  out ad infinitum
• gt Need for reasonable and theoretically-motivated
  parameters of what counts as evidence

34
Where NOT to go with nativelikeness in L2A
research

• (3) Characterizing departures from
  nativelikeness as fossilization
• gt Basic disagreements Is fossilization a
  product or a process? Explanans or explanandum?
  
• gt A protean, catch-all term relating to
  non-nativelikeness at end state A label in
  search of referents.
• gt Does non-native Behavior B qualify as
  fossilization? Does non-native Grammar G
  exemplify fossilization? Referents in search of
  labels.

35
Where NOT to go with nativelikeness in L2A
research

• (3) Characterizing departures from
  nativelikeness as fossilization
• One-size-fits-all label lacks useful granularity
  (e.g., no distinction is made among types of
  non-nativelikeness incompleteness, divergence,
  etc.
• Counterintuitiveness non-native indeterminacy
  permanent instability a type of
  fossilization?

36
Where to go with nativelikeness in L2A research

• Specific Further study of indeterminacy
• Specific Further study of comprehensive
  nativelikeness
• General Fill descriptive gaps in knowledge of
  upper limits

37
Where to go with nativelikeness in L2A research

• What are the upper limits of late L2A?
• Heuristically rich research directions
• 1. Universal Learnability Hypothesis
• 2. Selective Processability Hypothesis

38
Universal Learnability Hypothesis

• The hypothesis In late L2A at end state
• No feature of the L2 grammar is unlearnable.
• (Anything can be learned to nativelike levels by
  someone.)
• The evidence
• Nativelikeness in morphosyntax, lexis,
  phonology
• Nativelikeness in segmental- and global-level
  pronunciation
• Nativelike pragmatics

39
Universal Learnability Hypothesis

• An example
• Demonstrably difficult for Japanese L1 to
  perceive pronounce English /r/ and /l/
• McCandliss McClelland (various) showed that
  the distinction can be taught.

40
Universal Learnability Hypothesis

• Japanese learning English /r/ and /l/
• Initially unable to perceive the difference
• Training regimen required exaggerating and
  drawing attention to the difference
• Successful perception followed by accurate
  pronunciation on test items
• Poor generalization to new items.
• NB subjects were not at end state in English
  L2A.

41
Universal Learnability Hypothesis

• Japanese learning English /r/ and /l/
• Behavioral standard must be demanding (not
  limited to familiar items).
• Specify the role of training (necessary but not
  alone sufficient). Also, training required on
  perception as well as on production.
• SEGUEs

42
Universal Learnability Hypothesis

• SEGUE It is not claimed that the means to this
  end (nativelikeness) are the same as those
  employed in L1A. Nor is it claimed that the
  mental representations are identical for
  monolinguals and L2ers.
• SEGUE It is not claimed that the linguistic
  processing done by L2ers is the same (speed,
  manner) as that of L1ers gt

43
Selective Processability Hypothesis

• The hypothesis In late L2A at end state
• Not all on-line language processing tasks can be
  performed to native levels.
• The evidence On certain language processing
  tasks (e.g., detection of fine acoustic
  distinctions inherent in syllable stress,
  consonant voicing, and vowel duration),
  nativelike performance is not observed among
  sampled late L2 learners (e.g., Dupoux
  Peperkamp, 2002).

44
Selective Processability Hypothesis

• MORE TO COME (sorry!)

45
End of Session 2THANK YOU

• Session 3
• Neurocognition and late L2A