The Competition Model
Brian MacWhinney- CMU
Elizabeth Bates
Janet McDonald
Kerry Kilborn
Judit Osman-Sági
Vera Kempe
Yoshinori Sasaki
Csaba Pléh
Antonella Devescovi
Takehiro Ito
Jeffrey Sokolov
Arturo Hernandez
Michèle Kail
Klaus-Michael Köpcke
Ovid Tzeng
Beverly Wulfeck
Ping Li
Empirical Results Published in:
MacWhinney, B., & Bates, E. (Eds.) The crosslinguistic study of sentence processing. New
York: Cambridge University Press, 1989.
15 articles since then
1
1. The Input

A. Lexical Functionalism -- constructions

B. Input-driven Learning -- cues, frequencies

Cue validity predicts cue strength
[p(function)|form] - comprehension
[p(form)|function] - production
2
2. The Learner
Distributed representations -> transfer
Emergent modularity



Neuronal commitment, automaticity
Capacity



Functional neural circuits
Perspective-taking
3
3. The Context

Classroom context




Negative feedback is positive feedback
Instructional format interacts with learner
characteristics
Role of computerized instruction
Setting up input contexts


Role of lexical richness
Learner must learn how to learn
4
1A. Lexical Functionalism
Form
(cue, device)
Function
(role, meaning)
5
Competition between devices
Competition between interpretations
Agent
Marking
competition
Patient
Marking
hidden
Agent
Function
competition
Patient
Function
6
Cue validity -> cue strength
Cues -> Interpretations
Comprehension
Meanings -> Devices Production
pre
agr
init
nom
the
giv
def
hidden
act
top
per
7
Some cues
The tiger pushes the bear.
The bear the tiger pushes.
Pushes the tiger the bear.
The dogs the eraser push.
The dogs the eraser pushes.
The cat push the dogs.
Il gatto spingono i cani.
8
The dog was chased by the cat.

Comprehension - Interpretations compete
Agent: The dog vs. the cat
Patient: The dog vs. the cat

Production - Devices compete
Dog placement: preverbal, postverbal, by-clause
Cat placement: preverbal, postverbal, by-clause
9
Cue interactions
•
•
•
•
•
Peaceful coexistence
Cue coalitions
Competition between interpretations during
comprehension
Competition between devices during
production
Change from category leakage and
reinterpretation
10
Cues vary across languages
English:
The pig loves the farmer
SV > VO > Agreement
German:
Das Schwein liebt den Bauer.
Den Bauer liebt das Schwein
Case > Agreement > Animacy>Word Order
Spanish:
El cerdo quiere al campesino.
Al campesino le quiere el cerdo.
"Case" > Agreement > Clitic > Animacy > Word Order
11
Exotic Patterns
Navajo:
*Yas
lééchaa’í yi-stin.
snow
dog
him-frooze.
Lééchaaa’ yas
bi-stin
dog
snow
him-frooze
7-level hierachy of Animacy -- switch reference
12
Basic results



Reliable Cues Dominate
Cue Strengths Summate
Competition Cells show most variability
13
Ungrammaticality

Continuity for pockets of grammaticality





Hungarian possessive for accusative
Croatian neutralized case in masculine
Japanese “wa” marking
Slowdown for grammatical sentences in
Russian, Hungarian, Spanish without the
“preferred cue”
Cue summation for pronominal processing
14
English Word Order
Language by Word Order
Cognition (1982)
10 0
P e r c e nta g e c h oic e of fir s t n oun
80
German
60
Italian
40
20
English
0
NVN
VNN
NNV
15
Italian Agreement
Language by Agreement
JVLVB 1984
10 0
P e r c e nta g e c h oic e of fir s t n oun
80
60
German
40
English
20
Italian
0
Ag 0
Ag 1
Ag 2
16
English Children
English
60
50
W ord Order
40
30
20
10
Anima cy
Ag re eme nt
0
2
3
4
5
Ad ult
Age
17
Hungarian Children
Hungarian
50
Case
% ch o i ce first n o u n
40
30
20
Wo rd Ord er
10
Anim acy
0
3
4
5
6
Age
18
Italian Children
Italian
60
50
Agr eeme nt
40
Anima cy
30
20
W ord Order
10
0
2
3
4
5
Ad ult
Age
19
Cue validity (availability)

Task frequency
F(task T) / F(all tasks)

Simple availability
F(cue A present) / F (all cases of task)

Contrast availability
F(cue A present ^ cue A contrasts)
20
Cue validity (reliability)
 Simple reliability
F(cue A present ^ cue A correct) /
F (cue A present)
 Contrast reliability
F(cue A present ^ cue A contrasts ^ cue A correct) /
F (cue A present^cue A contrasts)

Conflict reliability
F(cue A conflicts with other cue ^ cue A wins) /
F(cue A conflicts with any cue)

SA -> CA -> SR -> CR -> Conflict transition
21
Cue validity vs. cue strength



Cue validity is based on (tedious) counts of
texts
Cue strength is first assessed through ANOVA
analyses in Competition Model experiments
Cue strength is then modeled using MLE
22
MLE models of cue strength
P (first noun) = ∏ S i (first) / ∏ S j (others)
 Two choice case
P (first noun) =

∏ S i (first) /∏ S i (first) + ∏ S j (second)
Models vary number of parameters and can be additive
or multiplicative
23
Pronouns - an online example
MacDonald and MacWhinney (1989)
Just before dawn, Lisa was fishing with Ron in the boat,
and she caught a big trout right away.
and lots of big trout were biting.
 Priming of referent at 500 msec for unambiguous
gender.
 Slowdown in processing of probes right at 0msec delay
when there is a gender contrast only.
24
Pronouns - implicit causality
McDonald and MacWhinney (1994)
Probes presented at 4 Delay Times:
D1
* 100
*
D2
D3
pro *
200
D4
*
end
*
Gary amazed Ellen time after time, because he was so talented.
N1
Probes:
V
N2
filler
referent
non-referent
distractor
verb
, because PRO predicate.
Gary
Ellen
Frank
amazed
25
Results and Competition
1. Slowdown in processing of probes at pronoun when there is a
contrast.
2. Facilitation from pronoun onwards when first noun
advantage agrees with implicit causality.
3. Activation of N2 right at the pronoun for E-S verbs!
4. Standard Competition Model cue summations and
competitions, all right when they should occur.
26
2. The Learner


Distributed representations -> transfer
Emergent modularity


Neuronal commitment, automaticity
Capacity


Functional neural circuits
Perspective-taking
27
Parasitic Learning -- Kroll
“turtle”
Translation route
“tortuga”
28
Transfer







Principle: Everything that “can transfer” will.
Connectionism predicts transfer
Word order can transfer
Phonology can transfer
Meaning can transfer
Morphological markings cannot
Early bilinguals as mixed
29
Transfer beyond the word








I want to go to school.
Yo querer ir a escuela.
I would like to go to school.
(I) would-like to-go to the-school.
xx quer-rí-a ir a la-escuela.
Do you want to eat at my house?
You want not want at me eat, huh?
Translation with feedback may not be so bad.
30
Emergent modularity

Growing modules



Farah and McClelland
Jacobs, Jordan, Barto
Kim et al. fMRI study
31
Capacity restrictions




Detectability
Complexity (for production)
Assignability (memory load)
Online load minimization



One good cue is enough (Russian, Spanish)
Waiting for a reliable cue: Russian, Hungarian
No use waiting for cue that will not be reliable,
German die Frau küßt der ...
32
DutchL1 EnglishL2
M e a n P e rc e n t firs t n o u n ch o ic e
10 0
90
80
70
60
Gro up 1
50
Gro up 2
Mono lingu als in
40
Eng lish
30
20
10
0
NVN
VNN
NNV
Fig 8.5 Gro up x wo rd o rd er in teraction in Englis h
33
JapaneseL1 EnglishL2
100
Novic e JE 2
90
M ean perc ent f ir s t noun c hoi c e
80
Monolingual
J apanes e
Advanced
J E2
Monolingual
Englis h
70
60
50
40
30
20
10
0
NVN
VNN
NNV
Fig 8. 7 Group x word order interact ion i n Engli sh
34
EnglishL1 DutchL2
10 0
P e rc e n t a g e va ria n ce a cc o u n te d fo r
80
60
Noun an ima cy
Case inflection
Word o rde r
40
20
0
Eng lish
E/D 1
E/D 2
E/D 3
Dutch
35
DutchL1 EnglishL2
100
P er c e nt ag e v a r ia nc e a c co un te d fo r
80
60
No un Animacy
Ca se Infl ection
Word Order
40
20
0
Du tc h
D/E 1
D/E 2
D/E 3
Engli sh
Gr oup
36
Aphasics - Word Order
Language x Group x Word Order
Brain and Language (1986)
100
Pe rce n t a g e C h o i ce o f F Irs t N o u n
80
Italians
60
Englis h Broc a
40
Englis h Nor mals
20
0
NVN
VNN
NNV
Wo rd Ord er
37
Aphasics - Agreement
La nguage x Group x Agree me nt
Bra in and La nguage (198 6)
100
P er c e nt ag e C h oi ce o f Fir s t N ou n
80
60
Eng lish Broca
40
Italian Bro ca
Eng lish Normal
20
Italian No rma l
0
Ag0
Ag1
Ag2
Agre ement
38
Case in Croatian Normals
Cas e x Animac y in Se rbo-Cr oatian Normals
120
100
80
P er c e nt ag e C h oi ce o f Fir s t N ou n
AA
60
AI
IA
40
20
0
No m-Acc
Acc-No m
Ambigu ous
Ca se
39
Case in Croatian Aphasics
Cas e x Animac y in Se rbo-Cr oatian Broc a's
100
80
P er c e nt ag e C h oi ce o f Fir s t N ou n
AI
60
AA
40
IA
20
0
No m-Acc
Acc-No m
Ambigu ous
Ca se
40
Word Order in Production
Wo rd Ord er in Broca's Ap hasics
100
G
E
I
T
80
P ercen tag e U se
60
Turki sh
It ali an
German
English
40
20
E
T
I
T
G
I
E
G
0
SVO
SOV
Other
41
Some generalizations
Children learn the most valid cues first.
 Aphasics preserve the most valid cues.
They also rigidify on the strongest devices
 L2 learners attempt transfer, but then learn
cues. They gradually reach L1 levels of cue
strength.
 Connectionism predicts transfer.

42
3. The Context
Providing negative evidence
go + PAST
meaning
competition
word
episodic
support
word
analogic
pressure
went
episodic
support
competition
go + ed
analogic
pressure
43
Word learning - Merriman
demitassse
episodic
support
competition
cup
extensional
pressure
44
Recovery in syntax
Dative Role
competition
to ___
“to”
episodic
support
verb
episodic
support
V + NP + NP
extensional
pressure
45
Complex cases
"pour arg1 arg2 arg3 "
competition
"1 pours 2 into 3"
lexical frame
group frame:
1 verbs 2 into 3
"1 pours 3 with 2"
group frame:
1 verbs 3 with 2
46
MacDonald et al.
47
MacDonald et al.
48
Open issues




Neuronal Commitment
Social Identification
Resonance
Setting up Input Contexts
49
Conclusions



Models of Input, Learner, and Context must
interlock
Competition Model is properly accounts for
what we know about language learning, but
The model must be developed still further.
50
Descargar

The Competition Model Brian MacWhinney