Adjunct control and the poverty of the stimulus: availability vs. evidence
Juliana Gerard
LAGB 10 September 2019 . julianagerard.com/lagb.pdf [email protected] Queen Mary University of London . julianagerard.com/dgfs.pdf introduction input: direct input: generalize UG implications 2
Acquisition
• What? • When?
Adjunct control
(1) John1 called Mary before PRO1 running to the shop
• Non-adultlike as late as 6-7
• How? • Evidence in the input?
Goodluck (1981), Hsu et al (1985), McDaniel et al (1991), Cairns et al (1994), Broihier & julianagerard.com/dgfs.pdf Wexler (1995), Goodluck (1998, 2001), Adler (2006), Gerard et al (2017, 2018) introduction input: direct input: generalize UG implications 3
Why?
• how? • adjunct control
julianagerard.com/dgfs.pdf introduction input: direct input: generalize UG implications 4
How?
Complex structure behavior grammar prediction: behavior 2
grammar grammar
julianagerard.com/dgfs.pdf introduction input: direct input: generalize UG implications 5
Adjunct control
(1) John1 called Mary2 before PRO1/*2 running to the shop
julianagerard.com/dgfs.pdf introduction input: direct input: generalize UG implications 6
Adjunct control
(1) John1 called Mary2 before PRO1/*2 running to the shop
julianagerard.com/dgfs.pdf introduction input: direct input: generalize UG implications 7
Adjunct control
before after (1) John1 called Mary2 beforewhile PRO1/*2 running to the shop without …
julianagerard.com/dgfs.pdf introduction input: direct input: generalize UG implications 8
Adjunct control
(1) John1 called Mary2 before PRO1/*2 running to the shop
Available interpretation(s): Adults: Subject control 4-6 yo: Subject control Object control Sentence internal behavior Free reference
Goodluck (1981), Hsu et al (1985), McDaniel et al (1991), Cairns et al (1994), Broihier & julianagerard.com/dgfs.pdf Wexler (1995), Goodluck (1998, 2001), Adler (2006), Gerard et al (2017, 2018) introduction input: direct input: generalize UG implications 9
Adjunct control
Adult grammar - Attachment height - Closest c-commanding NP
called
running to julianagerard.com/dgfs.pdf the storeshop introduction input: direct input: generalize UG implications 10
Adjunct control
Adult grammar - Attachment height - Closest c-commanding NP
→ broader implications for dependencies
julianagerard.com/dgfs.pdf introduction input: direct input: generalize UG implications 11
Case study: how? Adult grammar Attachment height Closest c-commanding NP H1. From the input a. Inferred from direct observation (direct input) b. Generalization from a similar structure (generalize) domain specific or domain general
julianagerard.com/dgfs.pdf introduction input: direct input: generalize UG implications 12
Case study: how? Adult grammar Attachment height Closest c-commanding NP H1. From the input a. Inferred from direct observation (direct input) b. Generalization from a similar structure (generalize) domain specific or domain general H2. Not from the input (UG) poverty of the stimulus
• Implications for • Input • Other dependencies julianagerard.com/dgfs.pdf introduction input: direct input: generalize UG implications 13
H1. Evidence in the input
a. Direct evidence: • attachment height • closest c-commanding NP
b. Generalization • Complement control (2) a. John wanted PRO to run to the store b. John told Mary PRO to run to the store
• Finite adjuncts he (3) John called Mary before she ran to the store Bill introduction input: direct input: generalize UG implications 14
a. Direct evidence
(4) John1 called Mary2 before PRO1/*2 running to the shop
? ? input input Available interpretation(s): grammar grammar Adults: Subject control 4-6 yo: Subject control grammar: Object control attachment height Sentence internal Free reference grammar: closest c-commanding NP
Goodluck & Behne (1992), Goodluck (1998) Goodluck (1981), Hsu et al (1985), McDaniel et al (1991), Cairns et al (1994), Adler (2006) introduction input: direct input: generalize UG implications 15
Attachment height
• Non-adultlike grammar: Variable attachment • Object control
Goodluck (1981), Hsu et al (1985), McDaniel et al (1991), Cairns et al (1994), Adler (2006) introduction input: direct input: generalize UG implications 16
Evidence: binding across clauses
(5) John called her1 before PRO meeting Mary1 at the shop
If evidence, assumes: • binding principles already learned (or don’t need to be learned)
Sutton (2015) introduction input: direct input: generalize UG implications 17
Evidence: binding across clauses
(5) John called her1 before PRO meeting Mary1 at the shop
If evidence, assumes: • binding principles already learned (or don’t need to be learned) • processing resources to maintain both clauses in memory
Sutton (2015) introduction input: direct input: generalize UG implications 18
Evidence: binding across clauses
(5) John called her1 before PRO meeting Mary1 at the shop
If evidence, assumes: • binding principles already learned (or don’t need to be learned) • processing resources to maintain both clauses in memory • frequent enough in the input → children’s perception (intake vs input)
Lidz & Gagliardi (2015), Omaki & Lidz (2015), Sutton (2015) introduction input: direct input: generalize UG implications 19
Evidence: binding across clauses Domain specificity
(5) John called her1 before PRO meeting Mary1 at the shop
If evidence, assumes: • binding principles already learned (or don’t need to be learned) • processing resources to maintain both clauses in memory • frequent enough in the input → children’s perception (intake vs input)
Lidz & Gagliardi (2015), Omaki & Lidz (2015), Sutton (2015) introduction input: direct input: generalize UG implications 20
a. Direct evidence
(4) John1 called Mary2 before PRO1/*2 running to the store
? ? input input Available interpretation(s): grammar grammar Adults: Subject control 4-6 yo: Subject control grammar: Object control attachment height Sentence internal Free reference grammar: closest c-commanding NP
Goodluck & Behne (1992), Goodluck (1998) Goodluck (1981), Hsu et al (1985), McDaniel et al (1991), Cairns et al (1994), Adler (2006) introduction input: direct input: generalize UG implications 21
Evidence for adultlike PRO?
• Not direct • Single instance likely consistent with many possibilities (1) John called Mary before PRO running to the shop
→ subject, sentence-internal, agent, discourse
1. Adjunct control: subject 2. Closest c-commanding NP introduction input: direct input: generalize UG implications 22
Evidence for adultlike PRO?
• Not direct • Single instance likely consistent with many possibilities (1) John called Mary before PRO running to the store
→ subject, sentence-internal, agent, discourse
1. Adjunct control: subject 2. Closest c-commanding NP introduction input: direct input: generalize UG implications 23
Evidence that PRO = subject?
• Previous studies: non-adultlike (4-6 year olds) • Subject control • Object control • Sentence internal • Free reference adult grammar
x x
Goodluck (1981), Hsu et al (1985), McDaniel et al (1991), Cairns et al (1994), Broihier & Wexler (1995), Goodluck (1998, 2001), Adler (2006), Gerard et al (2017, 2018) introduction input: direct input: generalize UG implications 24
Evidence that PRO = subject?
• Previous studies: non-adultlike (4-6 year olds) • Subject control • Object control non-adult grammar • Sentence internal • Free reference adult grammar
evidence for non-adult grammar x x x
Goodluck (1981), Hsu et al (1985), McDaniel et al (1991), Cairns et al (1994), Broihier & Wexler (1995), Goodluck (1998, 2001), Adler (2006), Gerard et al (2017, 2018) introduction input: direct input: generalize UG implications 25
H1. Evidence in the input
Adult grammar Attachment height Closest c-commanding NP a. Inferred from direct observation b. Generalization from a similar structure introduction input: direct input: generalize UG implications 26
b. Similar structures
Adult grammar Attachment height Closest c-commanding NP • Complement control (6) a. subject: John wanted PRO to run to the shop b. object: John told Mary PRO to run to the shop
• Finite adjuncts he (7) John called Mary before she ran to the shop Bill introduction input: direct input: generalize UG implications 27
Complement control
(6) a. subject: John wanted __ to run to the shop b. object: John told Mary __ to run to the shop (8) John called Mary before PRO running to the shop
→ Form (null) → Closest c-commanding NP
• Adultlike behavior observed before adjunct control → infer antecedent from complement control, then generalize?
Goodluck (1981), Hsu et al (1985), McDaniel et al (1991), Goodluck & Behne (1992), Cairns et al (1994) introduction input: direct input: generalize UG implications 28
Complement control
→ infer antecedent from complement control, then generalize?
issues • non-adultlike behavior for complement control → evidence for non-adult grammar
• assumptions • inference for complement control only • complement control → adjunct control • antecedent, but not e.g. verb form introduction input: direct input: generalize UG implications 29
Finite adjuncts
he (7) John called Mary before she ran to the shop Bill
• Attachment height? • Same learning problem as adjunct control • Finite → non-finite
• Linguistic input? • Non-finite vs finite adjuncts introduction input: direct input: generalize UG implications 30
Linguistic input main clause subject other internal external referent 150
100
50 per million utterances million per 0
finite non-finite CHILDES - Macwhinney (2000) introduction input: direct input: generalize UG implications 31
Linguistic input Antecedent main clause subject John called Mary after __? ran to the shop other internal external referent 150 e.g. Mommy fixed it before she left
100
50 per million utterances million per 0
finite non-finite introduction input: direct input: generalize UG implications 32
Linguistic input Antecedent main clause subject John called Mary after __? ran to the shop other internal before while external referent 150
John called Mary after __ running to the shop 100 before while without 50 instead of
for per million utterances million per 0
finite non-finite introduction input: direct input: generalize UG implications 33
Linguistic input Antecedent ran main clause subject John called Mary after __? to the shop running other internal external referent 150
100
50 per million utterances million per 0
finite non-finite introduction input: direct input: generalize UG implications 34
Generalization?
Antecedents • frequent frequent introduction input: direct input: generalize UG implications 35
Generalization?
Antecedents • frequent frequent • infrequent frequent? introduction input: direct input: generalize UG implications 36
Generalization?
Antecedents • frequent frequent • infrequent frequent? Issues: • Finite: free reference introduction input: direct input: generalize UG implications 37
Generalization?
Antecedents • frequent frequent • infrequent frequent? Issues: • Finite: free reference • Non-finite: non-answer ? introduction input: direct input: generalize UG implications 38
b. Generalization summary Adult grammar Attachment height Closest c-commanding NP • Problematic • Complement control: antecedent • Finite adjuncts: attachment height • Finite adjuncts: antecedent • Non-finite adjuncts: non-answer introduction input: direct input: generalize UG implications 39
H2. Universal Grammar Adult grammar • Universal principles Attachment height Closest c-commanding NP • Poverty of the stimulus • Structure present, but evidence not available
• Implications for • input • acquisition
Chomsky (1965) introduction input: direct input: generalize UG implications 40
Input
• Abstract principles → cross-linguistic variation • finiteness • complementizer • after, before, while,… : finite and non-finite • without, despite, … : non-finite only → English introduction input: direct input: generalize UG implications 41
Input
• Abstract principles → cross-linguistic variation • finiteness • complementizer
• Evidence • overt morphemes • frequency introduction input: direct input: generalize UG implications 42
Acquisition
• Role: mapping overt → abstract • vs. learning abstract
? input ? grammar input grammar
grammar: attachment height
grammar: closest c-commanding NP introduction input: direct input: generalize UG implications 43
Prediction: exceptions
• e.g. ‘for’ (9) a. Active: John1 called Mary2 after PRO1/*2 going to the shop b. Passive: John1 was called by Mary2 after PRO1/*2 going to the shop
(10) a. Active: John1 thanked Mary2 for PRO*1/2 going to the shop b. Passive: John1 was thanked by Mary2 for PRO1/*2 going to the shop
→ salient in the input? introduction input: direct input: generalize UG implications 44
Prediction: exceptions
(10) a. John1 thanked Mary2 for PRO1/*2 going to the shop b. John1 was thanked by Mary2 for PRO1/*2 going to the shop • e.g. ‘for’
→ 70% “thank you for ___ ing” frame main clause subject 100 other internal external referent
50
0 per million utterances million per
non-finite introduction input: direct input: generalize UG implications 45
Implications
• How is adjunct control acquired? • Indirectly • Not inferred from input
• Implications • Other dependencies, constraints • Non-adultlike stages introduction input: direct input: generalize UG implications 46
Other dependencies
NP [PRO ] c-command
X [ Y ] discourse?
evidence c-command introduction input: direct input: generalize UG implications 47
Other dependencies
NP [PRO ] c-command
X [ Y ] cdiscourse?-command
evidence c-command
• Variation? • Long-distance • Logophors • Aspect introduction input: direct input: generalize UG implications 48
Non-adult stages
• Poverty of the stimulus • no evidence not acquired from the input
• Logic for non-adult grammar
input grammar input grammar • no evidence not acquired from the input introduction input: direct input: generalize UG implications 49
Conclusion
• Availability vs evidence
• Implications for other dependencies introduction input: direct input: generalize UG implications 50
Thanks!
• Organisers of LAGB
• Ulster University Research Institute for Modern Languages and Linguistics • University of Maryland Project on Children’s Language Learning • Jeffrey Lidz • NSF DDRI grant BCS-1551662
contact: [email protected]