Transcript The Emergence of Grammar in Systems of Iconic and

How students acquire things you
never teach them
Robert Kluender
Department of Linguistics, UCSD
UCCLLT Workshop on
Grammar and Language Teaching
June 20, 2004
Is L2 acquisition like L1 acquisition?
The results of critical period research:
 there is a definite decline in ultimate
attainment with age across childhood
 it affects L1A more than L2A
 unclear how much of it is biological
 phonology correlates better with AoA
than morphosyntax (but accentless
non-native speakers seem to exist)
Is L2 acquisition like L1 acquisition?
 Often it is assumed that L2A differs
from L1A most in terms of implicit
learning
 However, there is also very clear
evidence of implicit learning in L2A
What would constitute proof?
 The best evidence for implicit learning
in L1A is reorganization
 We identify L1 reorganization by a
temporary increase in systematic
errors, the “U-shaped” learning curve
 Is there any evidence for a U-shaped
learning curve in L2 acquisition?
The strange case of unaccusatives
Why unaccusative verbs are a good test
case for implicit learning:
 they are found in every language, i.e.
are a well-attested linguistic universal
 they behave systematically
 they are not theory-dependent
 nobody knows about them, so they
can’t possibly be explicitly taught
An aside/exhortation from my
hobbyhorse soapbox bully pulpit
An aside/exhortation from my
hobbyhorse soapbox bully pulpit
 “Stop drilling!” (BVP)
An aside/exhortation from my
hobbyhorse soapbox bully pulpit
 “Stop drilling for UG!” (RK)
An aside/exhortation from my
hobbyhorse soapbox bully pulpit
 “Stop drilling for Universal Grammar!”
An aside/exhortation from my
hobbyhorse soapbox bully pulpit
 “Stop drilling for L2 evidence of
wh-movement constraints !”
An aside/exhortation from my
hobbyhorse soapbox bully pulpit
 Even assuming that they are part of
Universal Grammar, wh-movement
constraints are a moving target as to
– their overall status in the theory
– their current theoretical formulation
 Accumulating evidence that they are
instead a processing phenomenon
The strange case of unaccusatives
 native speakers are naturally unaware
of this phenomenon
in their
1st language
 it’s never taught to them in school
 2nd language learners are never
exposed to it, because 2nd language
teachers don’t know about it, either
 2nd language learners acquire it
nonetheless
What are “unaccusative” verbs?
 the unfortunate name stems from
Perlmutter (1977),
who first discussed the phenomenon
 unaccusatives are intransitive verbs
 whose subject is the undergoer
(also called “patient” or “theme”)
rather than the agent of the action
Two types of intransitive verbs
unergative verbs
subject is
AGENT
unaccusative verbs
subject is
UNDERGOER
 She left.
 She lay down.
 She hid.
 She arrived.
 She fell.
 She disappeared.
Some unaccusative verbs have
transitive counterparts
 transitive form:
The heat melted the butter.
 unaccusative form:
The butter melted.
 transitive form:
The children broke the vase.
 unaccusative form:
The vase broke.
Some unaccusative verbs have
transitive counterparts
 transitive form:
The heat melted the butter.
 unaccusative form:
The butter melted.
 transitive form:
The children broke the vase.
 unaccusative form:
The vase broke.
Tests for unaccusativity:
agentive -er suffixation in English
 She arrived.
 She fell.
 She disappeared.
 *arriver
 *faller
 *disappearer
Tests for unaccusativity
 Italian:
auxiliary selection in
passato prossimo
(Lei) è arrivata / caduta / sparita.
she is arrived
fallen
disappeared
*ha arrivata / caduta / sparita.
has arrived fallen
disappeared
Tests for unaccusativity
 Italian passive and reflexive verbs
also take essere (‘to be’)
as auxiliary in passato prossimo
 This means that all undergoer
subjects take essere as auxiliary
in Italian passato prossimo
Tests for unaccusativity
Of all the Romance languages,
 Italian has best retained the Latin
distinction between esse and habere,
 French has retained it to some degree
but lost other parts of it,
 while the other Romance languages
have lost it altogether
Tests for unaccusativity
 German and Dutch make very similar
distinctions in the perfect tenses
(e.g. Sie ist hingefallen in German)
 The distinction used to exist
in English, but now is found only in
archaic usage (e.g. Christian hymns)
– “Joy to the world, the Lord is come”
– “Alleluia, He is risen”
Another appeal for the inclusion of
linguistic knowledge in L2 teaching
 Consider how torturous it is using
traditional grammar to explain which
verbs take ‘be’ as auxiliary in perfect
tenses of European languages,
 and then consider how much easier
your life might be in this regard if you
referred to the L2 literature on
unaccusative verbs (Sorace 1993a)
Tests for unaccusativity
Italian: ne-cliticization
 transitive verbs:
Mario ha letto molte lettere
Mario has read many letters
Mario ne
ha letto molte
Mario of=them has read many
Tests for unaccusativity
Italian: ne-cliticization
 intransitive (unergative) verbs:
Hanno lavorato molte persone
have worked many persons
*Ne
hanno lavorato molte
of=them have worked many
Tests for unaccusativity
Italian: ne-cliticization
 intransitive (unaccusative) verbs:
Sono arrivate molte persone
are arrived many persons
Ne
sono arrivate molte
of=them are arrived many
Tests for unaccusativity
Italian: ne-cliticization
 transitive verbs:
Mario ha letto molte lettere
Mario has read many letters
Mario ne
ha letto molte
Mario of=them has read many
Generalizations from our tests
 English unaccusatives do not allow
agentive -er suffixation because
they do not take agent arguments
 only verbs with undergoer subjects
(unaccusative, passive, and reflexive)
take essere as auxiliary in Italian
 only verbs with undergoer arguments
(i.e. transitive objects & unaccusative
subjects) allow Italian ne-cliticization
Preliminary conclusions
 Unaccusative verbs have undergoer
subjects
 Remarkably enough, L2 learners
unconsciously seem to pick up on this
The unaccusative hierarchy
(Sorace 1993a/2000)








change of location
[selects BE ]
change of state/condition
continuation of a pre-existing state
existence of state/condition
change of state-transitive counterpart
uncontrolled process
controlled process (motional)
controlled process (non-motional)
[selects HAVE ]
The unaccusative hierarchy
(Sorace 2000)








arrive, fall
[selects BE ]
change of state/condition
continuation of a pre-existing state
existence of state/condition
change of state-transitive counterpart
uncontrolled process
controlled process (motional)
controlled process (non-motional)
[selects HAVE ]
The unaccusative hierarchy
(Sorace 2000)








arrive, fall
[selects BE ]
become, disappear, die
continuation of a pre-existing state
existence of state/condition
change of state-transitive counterpart
uncontrolled process
controlled process (motional)
controlled process (non-motional)
[selects HAVE ]
The unaccusative hierarchy
(Sorace 2000)








arrive, fall
[selects BE ]
become, disappear, die
stay, remain
existence of state/condition
change of state-transitive counterpart
uncontrolled process
controlled process (motional)
controlled process (non-motional)
[selects HAVE ]
The unaccusative hierarchy
(Sorace 2000)








arrive, fall
[selects BE ]
become, disappear, die
stay, remain
be, seem
change of state-transitive counterpart
uncontrolled process
controlled process (motional)
controlled process (non-motional)
[selects HAVE ]
The unaccusative hierarchy
(Sorace 2000)








arrive, fall
[selects BE ]
become, disappear, die
stay, remain
be, seem
break, melt, sink
uncontrolled process
controlled process (motional)
controlled process (non-motional)
[selects HAVE ]
The unaccusative hierarchy
(Sorace 2000)








arrive, fall
[selects BE ]
become, disappear, die
stay, remain
be, seem
break, melt, sink
blush, tremble, shine
controlled process (motional)
controlled process (non-motional)
[selects HAVE ]
The unaccusative hierarchy
(Sorace 2000)








arrive, fall
[selects BE ]
become, disappear, die
stay, remain
be, seem
break, melt, sink
blush, tremble, shine
run, dance, swim
controlled process (non-motional)
[selects HAVE ]
The unaccusative hierarchy
(Sorace 2000)








arrive, fall
[selects BE ]
become, disappear, die
stay, remain
be, seem
break, melt, sink
blush, tremble, shine
run, dance, swim
talk, work
[selects HAVE ]
The unaccusative hierarchy
(Sorace 1993a)
“The hierarchy embodies the fact that
the notion of dynamic change,
whose most concrete manifestation is
change of location, is at the root of
unaccusativity, and identifies verbs of
directed motion as core cases for
essere/être-selection.”
(Sorace 1993a: 81)
L2 sensitivity to semantic aspects
of unaccusativity (Sorace 1993b)
 Subjects
– English/French near-native speakers
of Italian in Italy, no Italian origins
– began learning after age 15 (18-27),
average 9 years of exposure (5-15)
 Materials and Procedure
– acceptability judgements on auxiliary
selection with unaccusative verbs
L2 sensitivity to semantic aspects
of unaccusativity (Sorace 1993b)
L2 sensitivity to semantic aspects
of unaccusativity (Sorace 1993b)
 L2 speakers were sensitive to
unaccusative hierarchy categories
 Only native speakers had significantly
different judgements between the
two auxiliaries in every category
 L2 speakers had significantly different
judgements between auxiliaries only
at the high end of the hierarchy
(two highest categories)
L2 sensitivity to semantic aspects
of unaccusativity (Sorace 1993b)
L2 sensitivity to unaccusativity
 L2 learners are sensitive to the
unaccusative hierarchy and
the semantic distinctions between
verb subtypes that it represents
 Is this only because these are highly
advanced, near-native learners?
Is there any evidence for a Ushaped learning curve in L2A?
 L2 learners passivize unaccusatives
*He was arrived early.
*My mother was died when I was just
a baby.
*This problem is existed for many
years.
*Most of people are fallen in love and
marry with somebody.
Unaccusative passivization errors
(Oshita 1998/2000)
Unaccusative passivization errors
(Oshita 1998/2000)
Is there any evidence for a Ushaped learning curve in L2A?
 Learners are never exposed to these
errors in input from native speakers
 They occur in the output of ESL
students of diverse L1 backgrounds
 They appear only at advanced or high
intermediate levels of L2 instruction
 Even at this level, L2 usage of
unaccusatives is 90% error-free
Why these particular errors?
 Recall that unaccusative verbs
pattern with passive verbs in Italian
with regard to auxiliary selection,
as both have undergoer subjects
 Passive verbs in English also have
undergoer subjects, and require
passive verbal morphology
Why these particular errors?
Sentence
3
Subject Verb phrase
#
3
# Verb Direct object
#
#
#
#
#
#
Robin killed
Kelly
AGENT
PATIENT
Why these particular errors?
Sentence
3
Subject Verb phrase
#
3
# Verb Direct object
#
#
#
#
#
#
 was killed Kelly
PATIENT
Why these particular errors?
Sentence
3
Subject Verb phrase
#
3
# Verb Direct object
#
#
#
#
#
#
 was killed Kelly
PATIENT
Why these particular errors?
Sentence
3
Subject Verb phrase
#
3
# Verb Direct object
#
#
#
#
#
#
Kelly was killed [
]
PATIENT
Why these particular errors?
Sentence
3
Subject Verb phrase
#
3
# Verb Direct object
#
#
#
#
#
#
 died
Kelly
PATIENT
Why these particular errors?
Sentence
3
Subject Verb phrase
#
3
# Verb Direct object
#
#
#
#
#
#
 died
Kelly
PATIENT
Why these particular errors?
Sentence
3
Subject Verb phrase
#
3
# Verb Direct object
#
#
#
#
#
#
Kelly died
[
]
PATIENT
Why these particular errors?
Sentence
3
Subject Verb phrase
#
3
# Verb Direct object
#
#
#
#
#
#
Kelly *was died [
]
PATIENT
Why these particular errors?
Sentence
3
Subject Verb phrase
#
3
# Verb Direct object
#
#
#
#
#
#
Kelly was killed [
]
PATIENT
Why these particular errors?
 By hypothesis, when learners
recognize that there is an undergoer
(patient) in subject position,
 they associate this with passive
morphology on the verb (be),
 and therefore passivize the verb
 even if it is not needed, as is the case
with unaccusative verbs (Oshita 1998/2000)
Why these particular errors?
 Note that this is a perfectly
reasonable mistake to be making:
 it shows unconscious sensitivity to
the presence of undergoer arguments
in subject position,
 and analogizes a known morphosyntactic pattern for such subjects.
 This is pretty sophisticated;
presumably this is why it occurs late.
Is this a U-shaped learning curve?
 Oshita (1998/2000) claims that it is
 But there was no empirical evidence:
 The data show the middle of a slump,
but no early error-free period,
and no subsequent recovery
 So is this a U-shaped learning curve
or just a nose dive that learners
never pull out of?
Follow-up: Klieman & Kluender
Corpus study of writing samples from
advanced ESL students in the
Chinese Learner English Corpus
 6% unaccusative passivization rate
 passivization more than twice as
frequent as other unaccusative errors
 more errors at intermediate levels,
but same % of passivization errors
Percentage of passivization errors
Percentage of passivization errors
Follow-up: Klieman & Kluender
Spoken/written production,
error detection in
Russian L2 speakers of English
 modified ILR OPI
 no differences in spoken elicitation,
but errors only at level 2+ and below
 ability in error detection significantly
different by proficiency level
Error detection by proficiency level
Follow-up: Klieman & Kluender
Spoken/written production,
error detection in
Russian L2 speakers of English
 no differences in spoken elicitation,
but errors only at level 2+ and below
 ability in error detection significantly
different by proficiency level
Follow-up: Klieman & Kluender
Spoken/written production,
error detection in
Russian L2 speakers of English
 no differences in spoken elicitation,
but errors only at level 2+ and below
 ability in error detection significantly
different by proficiency level
 written production errors only at
level 2+ and below
Follow-up: Klieman & Kluender
 Clear evidence for recovery at level 3
 But is this merely circular evidence?
No unaccusative passivization errors
because “no systematic errors”
of grammar at level 3 and above?
 But unaccusativity is not targeted by,
or even on the radar screen of OPI
 In any case, the problem goes away
But is it a U-shaped learning curve?
 It’s at least a J-shaped learning curve
 Still no reliable data from early L2A:
Initial attempts to use the same
procedures on low-proficiency
Russian learners failed
 At a minimum, there is evidence for
implicit learning and reorganization
Summary:
L2 sensitivity to unaccusativity
 L2 learners are sensitive to
semantic verb subtype distinctions
on the unaccusative hierarchy
(auxiliary selection in Italian)
 L2 learners show sensitivity to the
fact that unaccusative verbs
take undergoer subjects by
overgeneralizing passive morphology
(passivization errors in English)
What would constitute proof?
 The best evidence for implicit learning
in L1A is reorganization
 We identify L1 reorganization by a
temporary increase in systematic
errors, the “U-shaped” learning curve
 Is there any evidence for a U-shaped
learning curve in L2 acquisition?
Is L2 acquisition like L1 acquisition?
L2 acquisition of unaccusativity:
 an indisputable language universal
 implicit learning with no explicit input
 overgeneralization
 low error rates
 eventual recovery