The logical problem of language acquisition

1
The logical problemof language acquisition

• Brian MacWhinney
• CMU

2
Three views

• Socialization Theory
• Language is learned from interactions
• Connectionist Theory
• Language is learned from cues
• Nativist Theory
• Language is innate

3
The facts

• Child Nobody dont like me.
• Mother No, say Nobody likes me.
• Child Nobody dont like me.
• Mother No, say Nobody likes me.
• Child Nobody dont like me.
• Mother No, say Nobody likes me.
• Child Nobody dont like me.
• dialogue repeated five more times
• Mother Now listen carefully, say Nobody likes
  me.
• Child Oh! Nobody dont likeS me.
• (McNeill, 1966)

4
Brown and Hanlon (1970)

• parents correct for meaning not form
• when present, correction was not picked up

5
The intuitive problem

• The child makes an error.
• The adult may correct or identify the error.
• But the child ignores these corrections.
• So, how does the child learn to stop making the
  error?

6
Recovery from Overgeneralization

• u-shaped curve went - goed - went
• child must stop saying
• goed
• unsqueeze
• deliver the library the book

7
LPLA 1 Argument from Poverty of Stimulus

• Noisy input
• Incomplete input
• Ignoring correction
• Not enough feedback
• Unclear referentially (Quines Problem)

8
The Gold Proof

• Text Presentation Utterance Feedback Result
• Child says went. none none
• Child says goed. none none
• Adult says went. --- positive data
• Informant Presentation Utterance Feedback Result
  
• Child says went. good positive data
• Child says goed. bad corrective
• Adult says went. good positive data
• Adult says goed. bad corrective

9
An overly general grammar
10
(No Transcript)
11
How to get the wrong grammar
(1) Start with this finite grammar FG1

• FG1 generates ABD, AC
• (2) Provide ACD as positive data

12
3. Adding an arc yields FG2
13
If we only have plain and simpleFINITE grammars
like FG1 and FG2, information presentation is
enough.
14
But, if the possible human languages include
NONFINITE grammars like NFG1
S -gt AP BPAP -gt A (C)BP -gt (B) D
15

• NFG1 also generates ACBD
• And ACBD is ungrammatical.
• But Informant presentation doesnt tell us this.
• So we will never be able to give up NFG1 and go
  back to FG1 or FG2.
• SoWe will never learn the correct language
  without corrective feedback.

16
Input must be really consistent
17
d-prime maximizes this ratio____hits____false
alarms p (ErrorSignal) must be close to
1.0.p (CorrectSignal) must be close to 0.0.
18
But, sometimes adults say no when a sentence
is correct. This means that p(CorrectSignal)
is not close enough to 0.0to learn on a few
examples.SO, child will need LOTS of examples
if he tries to learn through Signal Detection.
19
But .If a parent were to provide true negative
evidence of the type specified by Gold,
interactions would look like this

• Child me want more.
• Father ungrammatical.
• Child want more milk.
• Father ungrammatical.
• Child more milk !
• Father ungrammatical.
• Child cries
• Father ungrammatical

20
Contrast with this interaction

• Child me want more.
• Father You want more? More what?
• Child want more milk.
• Father You want more milk?
• Child more milk !
• Father Sure, honey, Ill get you some more.
• Child cries
• Father Now, dont cry, daddy is getting you
  some.

21
Snow, Bohannon, Farrar, Hirsh-Pasek, Cross,
Sokolov, MacWhinney, Keith Nelson, and many
others

• 1. Correction is targeted.
• Only simple, clear errors are corrected.
• 2. Input is targeted by developmental level.
• Fine-tuning
• 3. Pickup may be on the next page of the
  transcript.
• 4. Child can process
• overt correction, recasts, restatements,
• clarification questions, What? huh?
• continent queries,
• Some combination of these cues?
• 5. Experiments have shown that correction
  works.

22
BUT For the sake of analysis, let us grant
thatCorrective feedback is not available,
and if available is not used, and if used
is not effectiveThere are 5 potential
solutions to LPLA1
23
1. Simple Blocking -- Baker
General Rule ordered after specific
rule.Specific rule bleeds context for general
rule.
1. produce went (bleeds condition)2. add
-ed (cant apply if went already fired)
Benign cases permit blocking solution.Malignan
t cases dont.My Thesis All cases are benign.
24
2. Conservatism

• Conservative child learners only use forms they
  have heard adults use.
• Logically, the constraint of conservatism would
  work, but overgeneralizations prove that learners
  are not conservative.
• If children waited until each form were
  confirmed, they would never say
  goed unsqueeze deliver the library the
  book
• They would never overgeneralize. But they do
  and so do L2 learners.
• Conservatism can explain obedience to principles

25
(No Transcript)
26
(No Transcript)
27
3. Indirect negative evidenceLasnik, Chomsky,
Braine, Berwick, Siskindaverage frequency of V
frequency of goaverage frequency
of V-ed frequency of goed x
x y y
28
If x/y gt x/y by a large amountand if y is
frequentThen Y is blocked. do undo tie
untie zip unzipsqueeze (unsqueeze)
29
Using Indirect Neg Evidence

• N in relative N in complement
• N extracted N extracted
• Bill thought the thieves were carrying the loot.
• What did Bill think the thieves were carrying.
• The police arrested the thieves who were carrying
  the loot.
• What did the police arrest the thieves who were
  carrying?

30
4. ProbabilismHorning (1969) shows that Golds
Proof fails for probabilistic grammars.These
can be identified on positive evidence
alone.Labovs variable rules are a good example
of probabilistic grammars.
31
5. Competition generator --
rules blocker -- constraints generator --
analogic pressure blocker -- episodic support
32
(No Transcript)
33
(No Transcript)
34
Single trial learning to criterion will not occur
when analogic pressure is strong.
35
(No Transcript)
36
EPISODES are specific encounters with particular
form-function relationsEXTENSIONAL PRESSURE is
based on patterns involving multiple exemplars.
Morphological extension is to a new
stem.Semantic extension is to a new referent.
37
Modeling analogic pressure
38
Recovery from Overgeneralization

• 1. Rote learning through episodic support
• Emergence of went
• 2. Growth of generalization through extensional
  pressure
• Occasional use of goed
• 3. Competition between pathways
• went competes with goed
• 4. Processing
• went is slower than goed (Kawamoto, 1993)
• expressive monitoring (MacWhinney, 1978)
• adaptive resonant connections strengthen went
  (Grossberg, 1987)

39
LPLA 2 Non-occurring errors

• Chomsky recent advances make the logical
  problem trivial, since there occuring is so
  little left to learn
• Problems
• No system of triggers has been identified
• No rules for the interaction of triggers with
  data is available
• No agreement on parameter interactions has been
  reached
• For these reasons, few have accepted Chomskys
  analysis.

40
Structural Dependency

• The man who is first in line is coming.
• Is the man who __ first in line is coming?
• Is the man who is first in line ___ coming?
• This only applies to non-parameterized aspects of
  language.

41
No need for positive evidence

• Chomsky A person might go through much or all
  of his life without ever having been exposed to
  relevant evidence, but he will nevertheless
  unerringly employ the structure-dependent
  generalization, on the first relevant occasion.
• Hornstein and Lightfoot People attain knowledge
  of the structure of their language for which no
  evidence is available in the data to which they
  are exposed as children.

42
Emergentist solution

• Item-based learning for aux
• Movement formulated in terms of relations, not
  position (this is the crucial step)
• Competition yields construction (not needed
  initially, but part of general solution)
• As a result (3) is produced instead of (2)

43
More cases

• Who did John believe the man that kissed ___
  arrived?
• Who did John believe __ kissed his buddy?
• What did you stand between the wall and __?
• What did you see a happy ___?

44
General Issue

• Conservatism
• Parsing
• Competition from Alternative (in situ)
• Impossible Meaning
• Universal Constraints