投影片 1 - National Sun Yat

Download Report

Transcript 投影片 1 - National Sun Yat

Ch 10 Linguistic Universal and
SLA
Ellis, R. (1994). Second
Language Acquisition.
Shu-ing Shyu
Introduction
• Two approaches toward the study of the
language
– Externalized (E) approach
– Internalized (I) approach
• Linguistic universals
– Typological universals (Greenberg 1966;
Hawkins 1983; Comrie 1984; Croft 1990)
– Universal Grammar (Chomsky)
• Interlanguage Theory: Another perspective
– Adjemian (1976): interlanguages are ‘natural
language’, subject to all the same constraints
 rule governed
• ‘the grammatical nature of a learner’s IL’ rather than
‘strategies’ (1976: 306)
– Interlanguage Structural Conformity Hypothesis
(Eckman 1991: 24)
• “The universal generalization that hold for the
primary languages also hold for interlanguages.”
• 11 adult Asians learning English
–  final stop + stop > fricative + stop (1991: 24)
Typological universals
and SLA-1
• Absolute universals (implicational
universals)
• E.g. If a lg has a noun before a demonstrative,
then it has a N before a RC (Hawkins 1983: 84).
• If a lg is SOV, then if the adjective precedes the N,
then the genitive precedes the N.’
•  in one direction only
• Universal tendencies (non-implicational
universals)
– Exceptions? How many can be tolerated?
Typological universals
and SLA-2
• Implicational Universals:
– NP Accessibility Hierarchy (NPAH)
– Subj < DO < IO < oblique < genitive < object of comparative
(Comrie & Keenan (1979))
– See table 10.1 (Jones 1991)
– The grammatical function of the relative clauses in the matrix
clause has various degree of difficulity.
– 2. Markedness as relative phenomenon. Implicational
universals presuppose a markedness relationship.
• If a lg has property X also has property Y 
Y is unmarked in relation to X.
Typological universals
and SLA-3
• Evidence to determine the level of
markedness of specific linguistic features:
– Structure:
– Behavior: grammatical versatility (gender
pronoun, active vs. passive)
– Frequency:
Typologically motivated
studies of SLA-1
• Qs:
– What influence do typological universals have
on the order of acquisition of grammatical
features?
– What effect does markedness have on
learning difficulity?
– How does the typological status of
grammatical features in the native and target
lgs affect L1 transfer?
Typologically motivated
studies of SLA-2
• Negative Placement
– Preverbal negation < postverbal negation (Dahl
(1979))
• Wode (1984) preverbal negation appears in interlg, but not
present in either L1 or L2.
•  absolute universal?
• Preposition stranding and pied piping
– Pied-piping < prep stranding (implicational universal)
(Mazurkewich 1984; Bardovi-Harlig 1987)
• Bardovi-Harlig (1987): markedness hypothesis is not tenable.
Typologically motivated
studies of SLA-3
• Relativization
– Does the Accessibility Hierarchy (AH) account for avoidance
behavior?
• Schachter (1974): left-branching lg speakers avoid using RCs in
English
• Gass (1980: 138): avoid relativizing on low positions in the AH
He saw the woman who was kissed by the man. (OS)
< He saw the woman that the man kissed. (OO)
• Akagawa (1990): no support for AH in L1 Japanese speaking
English as L2  mixed evidence
– Does the AH explain the order of acquisition of relativizable NP
positions?
• Supporters: Gass (1980), Pavesi (1986), Eckman, Bell & Nelson
(1988).
Typologically motivated
studies of SLA-4
• Relativization
– Does the AH explain the use of pronominal copies?
• Support: Gass (1979; 1980), Hyltenstam (1984): markedness
interacts with L1 transfer: whether or not pronominal copies
occurred in the learners’ L1
• Against: Tarallo & Myhill (1983):
– Does the AH explain the acquisition of the forms of
different relativizers.
• Hawkins (1989): predicted by the AH
– Qui (S) < que (O)< don’t (Genetive)
• Conclusion:
– Markedness correlates with acquisition order
– But: Tarallo & Myhill (1983), Hawkins (1989):
Learners construct rules for RC on the basis of the
adjacency of categories in the surface configuration.
Typologically motivated
studies of SLA-5
• Limitations
– Descriptive observations, rather than
explanations
– Levels other than syntax (e.g. discourse) are
neglected.
Universal Grammar
and SLA-1
• To address 3 Qs:
– 1. What does UG consist of?
– 2. What role does UG play in L1 acquisition?
– 3. What should the domain of a theory of SLA
be?
The Theory of UG-1
• Principles: underlines the grammatical rules of all lgs:
– Configurationality, Subjacency, Case theory, etc.
• Parameters: variations among lgs
• two or more settings
– Word order, pro-drop, etc.
– Pro-drop properties:
Null subjects
No null subjects
No expletives
Have expletives
permit variable w.o.
Fixed w.o.
No that-trace effect
show that-trace effect
Chai hai ditto che e venuto?
*Who have you said that is come?
A class of modal Vs, distinct from
main Vs
The Theory of UG-2
• Markedness: UG provides a basis for determining
markedness:
– “core” and “periphery” features
core features
periphery features
governed by UG
Not governed by UG
Unmarked ∵ require minimal
evidence for acquisition
marked ∵ require more substantial
evidence for acquisition
– White (1989a): markedness as internal to the L, vs. sth external,
evident only in extant lgs (markedness in lg typology).
– Zobl (1983): projection capacity
– Markedness is understood in relation to the amount of primary
linguistic evidence needed to acquire a given property
– Evidence of one feature in a cluster may enable learners to
acquire the other features associated with it, irrespective of
whether they have experienced these features in the input.
UG and L1 Acquisition1
• What is the nature of the child’s
experience of the target lg?
– The poverty of the stimulus
• What does the LAD (Lg Acquisition Device)
consist of?
The poverty of the
stimulus
Miller & Chomsky (1963); Chomsky
Caretaker talk: far less degenerate
1965; McNeill 1966: inadequate source than was claimed (Ch. 6)
of information for language acquisition
Miller & Chomsky (1963); Chomsky
Caretaker talk: far less degenerate
1965; McNeill 1966: inadequate source than was claimed (Ch. 6)
of information for language acquisition
White (1989a): positive evidence will
not suffice. How can children
distinguish ungrammatical Ss from
grammatical ones?
 rely on innate knowledge
White (1990:124): (1) input alone
cannot explain L1 acquisition (2) the
child use be equipped w/ knowledge
that enables the deficiencies of the
input to be overcome.
Children act in accordance with
indirect negative evidence. Avoid
certain errors ∵ not heard before (This
argument cannot replace those based
on innateness)  do produce errors
that they could never heard in the input
The language faculty-1
• --UG ensures that relatively little evidence is
needed for the child to determine that a given
principle is operative in the TL or to decide
which setting of a parameter is the right one
• --UG prevents children from construction wild
grammars (Goodluck 1986)
• --errors can be unlearned on the basis of
positive evidence
•  a lg is learnable because the child needs to
entertain only a small subset of the hypotheses
that are consistent w/ the input data.
The methodology of
UG-based studies-1
• (1) how to ensure that the subjects have the
requisite level of L2 proficiency to demonstrate
whether or not a particular principle is operating
in their interlanguage grammar.
• (2) the need to rule out the effects of the L1
– direct access to UG
– indirect access through their L1
• E.g. L1 (Chinese, Japanese, Korean) -Subjacency
L2+Subjacency
•  in accordance w/ UG constraints provided by not through
formal instruction
The methodology of
UG-based studies-2
• (3) parameter setting: to establish which setting
out of those possible is reflected in the learner’s
interlanguage
– bi-directional studies w/ control groups
• L1 non-pro-drop-L2 pro-dropL1:
– (English)-L2 (Spanish
• pro-drop-L2 non-pro-dropL1) L1
– (Spanish)-L2 (English)
• (4) Data collection:
• elicited data have been preferred:
• act-out tasks, picture identification task, sentence-joining,
card-sorting tasks,
• grammaticality judgment task-making metalingual
assessment.
The methodology of
UG-based studies-3
• UG-based SLA research:
– experimental in nature: control and experiment groups, elicited
data
– have precise hypotheses about the nature of SLA
– explanatory in nature (not the description of learner lg)
• Problems:
–
–
–
–
Grammaticality judgment
Implicit or explicit knowledge?
Whether UG is alive or dead in the L2 learner
 no clear evidence to support the hypotheses that learners
acquire unmarked ‘core’ features before marked ‘peripheral’
features (e.g. dative alternation)
some empirical
research-1
•
Subjacency
–
Studies for the availability of UG
•
Richie (1978b)
–
–
–
•
Bley-Vroman, Felix, and Ioup (1988)
–
–
–
–
–
–
•
rightward movement of PP
‘*That a man has just passed by was not noticed in a car.’
Japanese graduate students can judge it less grammatical  linguistic universals
are intact in the adult (1978b:43)
92 advanced Korean Ls of L2 English
grammaticality judgment
--response bias: reject Ss irrespective of grammatical of ungrammatical Ss
a reluctance to make use of ‘not sure’
similar response pattern to NSs’
 UG is accessible to adult learners
White, Travis, and Maclachlan
–
–
–
–
–
Malagasy Learners of English
(subject extraction in wh- interroatives ok in Malagasy)
judgment test, a written elicited production task
nearly all high-intermediate Ls, 1/2 low-intermediate Ls reject Subjacency vio.
that Ls who accepted Ss violate Subjacency is because they had not reached a
stage of syntactic development for the principle to become active (not acquired
yet)
some empirical
research-2
•
Subjacency
– Studies against the availability of UG
•
Schachter (1989)
– L1: Korean (no Subjacency), Chinese (weak subjacency),
Indonesian (has subjacency, but not in wh-movement)
– English NS as controls; judgment test
– different from NSs’ judgment
– NSs passed both the syntax and the subjacency tests.
– NNSs only passed the syntax tests.
•
Bley-Vroman et al.(1988:8)
– ease of processing rather than UG
•
Schachter & Yip (1990)
– processing effects
some empirical
research-3
•
Subjacency
– Studies against the availability of UG
•
Schachter (1989)
•
Bley-Vroman et al.(1988:8)
•
Schachter & Yip (1990)
– processing effects
•
 “relative acceptability’ rather than
‘absolute obedience’ to UG
– more types of data elicited, not just
grammaticality judgment tasks
some empirical
research-4
•
The pro-drop parameter
–
parameter resetting
•
White (1985; 1986)
–
–
•
parameter resetting
 L2 Ls opt for the L1 setting of the pro-drop P, as their proficiency
increases they switch to the L2 setting
White (1985) --pro-drop features cluster in IL grammars?
–
–
–
–
–
–
–
–
–
L1: Spanish, French (1985,6)L1: Italian (1986)L2: English
Ss judgment task (missing expletives and pronouns subjects), some w/
ungrammatical S-V inversion, and a that-trace effect
Ellis’ critics: more reliable and valid data needed, not just judgment
data
Spanish and Italian Ls inclined to accept subjectless Ss, different from
the French.
Spanish Ls likely to accept that-trace than the French
NO difference in judgments of Ss w/ ungrammatical S-V inversion
 L2 Ls don’t interact directly w/ L2 data, but initially transfer L1
setting
given time, reset L2 value
TL features may cluster in IL G
some empirical
research-5
•
The pro-drop parameter
–
parameter resetting
•
Hilles (1986)
–
–
–
–
–
•
Longitudinal data from Cancino et al. (1978)
L1: Spanish (only one 12-yr old Colombian)
L2: acquisition of English negatives and aux
Hypothesis 1: Jorge L1 (pro-drop) switch to L2 (null pro-drop)
Hypothesis 2: the switch co-occurs w/ the emergence of aux, triggered
by the acquisition of expletive subject  Support her two hypotheses
Hilles (1991)
–
–
–
–
–
Longitudinal data from Cancino et al. (1978) L1: 6 Spanish (2 children,
2 adolescents, 2 adults)
Q: to what extent the use of pronominal Ss and V inflection were
correlated overtime?
3 of the Ls (2 children, 1 adolescent) manifested a strong correlation
b/w the emergence of pronominal Ss and V inflection
 their acquisition guided by UG (mirror L1 acquisition)
other 3 Ls: no such correlation  lack access to UG
some empirical
research-6
•
The pro-drop parameter
–
parameter resetting
•
Lakshmanan (1991)
–
–
–
•
Longitudinal data of 3 children (Marta in Cancino ea al.’s study,
Muriel a French child in Gerbault 1978, Uguisu Japanese by
Hakuta 1974)
 support neither a ‘transfer’ nor a ‘developmental’ explanation
 no unequivocal evidence in favor of the clustering effect
Hilles reported
Phinney (1987)
–
–
–
–
Bidirectional study: English-speaking Ls of L2 Spanish,
Spanish-speaking Ls of L2 English
written composition
 L1 value transfer
 no evidence of transfer
some empirical
research-Summary
•  no real support for a parameter-setting
model of SLA., no clear evidence of any
clustering effects (Lakshmanan’s and
Phinney’s)
– Learning principles: (clearer results)
– L2 learners do not follow the Subset
Principle
• Summary:
• White: L2 Ls do not have access to learning principles
like the Subset principle. They build a superset L2 G,
often influenced by L1. They restructure this G, creating
a subset G by negative E.
• Testable hypotheses:
– (1) L1 Ls will not make certain kinds of error whereas L2 Ls will
 Subset Principle operative in L1A
– (2) L2 Ls with no access to negative evidence will fail to
eliminate superset errors (not investigated)
– (3) L2 Ls who receive formal instruction will eliminate errors.
(received support: Ch. 14)
The logical problem of
SLA
• (i) SLA is essentially the same as for L1A
• (ii) SLA is different because L2 learners
achieve variable success
• (iii) L2 competence is qualitatively
different from L1 competence.
–  different views regarding the role of UG in
SLA
Access to UG in SLA
(Table 10.3)
• The complete (direct) access view:
Parameter-setting Model (Flynn 1984; 1987)
• (2) No-access view: Clahsen and Muysken
(1986; Meisel 1991) (the Fundamental
Difference Hypothesis)
• (3) Partial-access (indirect) view: Schachter
(1988) (i.e. via L1)
• (4) Dual access position: Felix (1985):
Competition Model
The role of negative
evidence
• L1 Ls: negative evidence not available
• L2 Ls: have access to both corrective feedback & explicit
grammatical information
• Negative evidence is beneficial to L2 learning
– White (1991): Adv placement can be successfully learnt through
formal instruction:
•  negative evidence triggers the resetting of a parameter to its L2
value
– Schwartz (1986): negative evidence can result in the A of
grammatical knowledge
• Ellis: if UG exists to enable children to acquire
grammatical competence solely on the basis of positive E,
it is hardly felicitous to propose that L2 Ls can access
parts of it with the help of negative evidence. (p. 457)
Thank you.