Truth Value Judgment vs' Picture Selection

1
Truth Value Judgment vs. Picture Selection

• Sergey Avrutin Sergio Baauw
• Utrecht University

EMLAR II Workshop, Utrecht University, 16
November
2
Why use Truth Value Judgment/Picture Selection?

• Investigate childrens understanding of specific
  properties of their language.
• Specifically are children able to assign the
  correct interpretation to a particular linguistic
  unit (sentence, construction)?

3
(Spontaneous) production?

• Information about interpretation is difficult to
  get from spontaneous production data
• Limited control over context (meaning) of
  utterance.
• Spontaneous data might tell us something about
  what a sentence can mean, but not what it cannot
  mean.
• John touched him
• meaning 1 John touched Peter (adult ok)
• meaning 2 John touched John (adult )

4
Solution

• Grammaticality judgments? can not be asked from
  young children.
• Linguistic experiment give to the child the
  possible and impossible (in adult language)
  interpretation, and see which one they allow
• Interpretations are represented by pictures or
  stories carried out by toys. These are matched
  with input sentence.

5
Truth Value Judgment Picture Selection

• We will show that depending on what you are
  looking for, one of them can be more informative.
• As we will see, TVJT is supposed to show which
  interpretation is possible for the child, and PST
  which one is preferred.

6
Truth Value Judgment Task

• Sentence is matched with an image/scene
  representing the correct (adultlike)
  interpretation and the incorrect interpretation
  in different trials.
• Child has to decide whether the sentence
  describes the action represented in the
  picture/by the scene, responding with yes or
  no.
• Example, Principle A
• (2) John touched himself
• a. John touched John
• b. John touched Peter

7
Varieties TVJT

• Asking direct questions about pictures to the
  child (Picture Verification Task I)

Here we see a mother and a girlIs the girl
drying herself?Adult response NO
Here we see a mother and a girl.Is the girl
drying herself?Adult response YES
8
Varieties TVJT

• Advantages
• - simple
• - only need one experimenter (who does the
  questioning and writes down the responses)
• - quick
• Disadvantages
• - child is interrogated, and might feel under
  pressure (could lead to indiscriminate yes
  responses)
• - condition of plausible denial not satisfied
  (Crain Thornton 1998)

9
Varieties TVJT
2. Guessing game. Asking direct questions about
pictures to a second experimenter or puppet
(Picture Verification Task II) Experimenter 1
helper of the child, shows the pictures, tells
the child about the characters involved in the
action depicted (here we see a girl and a mum),
and then gives the guesser a hint (its about
drying) Experimenter 2 guesser, sits
opposite to the child and helper. Does not see
the picture. After hearing the hint he makes
his guess, reading the question from the back
of the picture.
10
Varieties TVJT

• Helper here we see a mother and a girl, and
  it is about drying. Lets see if Sergey can to
  guess what is happening exactly in the picture.

Mmm. a mother and a girlIs the girl drying
herself?Adult response NO
Mmm a mother and a girl.Is the girl drying
herself?Adult response YES
11
Varieties TVJT

• Advantages
• - Child feels more at ease. It is the guesser
  who is being interrogated, not the child
• - More fun its a game, not a test
• - Even more fun when you use a puppet (Kermit,
  Bert) to do the guessing.
• Disadvantages
• - Takes more time
• - Requires two experimenters
• - Condition of plausible denial not satisfied

12
Varieties TVJT

• Asking a child to judge statements uttered by a
  puppet. Two picture sequence.

Two boys, John and Peter, were invited to a
party. They wanted to go disguised as clowns.
Peter has some paint, but it is not enough for
both. He was wondering, what shall I do, paint
Johns face or my own? Lets see what Peter
decided to do.
13
Varieties TVJT

• Asking a child to judge statements uttered by a
  puppet. Two picture sequence.

This is what Peter did Let us ask Kermit
whether he can guess what Peter decided to
do. Kermit MmmJohn and Peter. I know what
happened. Peter painted himself. Adult response
YES
14
Varieties TVJT

• Advantages
• - Might be more interesting because of the
  story.
• - Kermit is a familiar person, might be less
  intimidating than adult experimenter.
• - Context provides both potential meanings of
  the reflexive (paint oneself vs. paint
  somebody else) ? plausible denial.
• Disadvantages
• - More time consuming
• - Two experimenters required

15
Varieties TVJT

• 4. Use of toys to act out the situation instead
  of pictures.
• Advantages
• - Gives the child more the sense of a game.
• - Provides better ways to comply with plausible
  denial
• Disadvantages
• - Very time consuming
• - Requires a lot of practice for the
  experimenters
• - Two experimenters required

16
Yes bias

• Rationale behind TVJT subject will exploit their
  linguistic competence to the limit in order to
  say YES.
• If subjects say NO, this is because their grammar
  prevents them from doing so.
• Therefore, make sure that the test conditions are
  the No-conditions (Yes-conditions are control
  conditions)

17
Butis it true?

• Maybe subjects do not exploit their grammatical
  options to the limit.
• Interpretation of pronouns as complements of
  locative preposition (Baauw 2000)
• (3) La niñai puso la maleta detrás de ellai/k
• The girl put the suitcase behind her.
• Prediction adults 100 yes-responses on
  picture displaying girl putting a suitcase
  behind herself.
• Actual outcome 50 yes-responses!

18
Picture Selection Task

• Example of interpretative preference
• First the boy scratched the man, and then
• a. the woman scratched him.
• b. the woman scratched HIM
• This tendency can easily be overridden by
  semantic factors
• (5) John hit Bill, and then Mary punished him.
• Picture Selection Task Subjects choose the
  picture that (best) matches the input sentence.

19
Picture Selection Task

First the boy scratched the man, and then the
woman scratched him.
20
Picture Selection Task

• It is necessary to include a filler (so,
  minimally three pictures to choose from).
• Fillers are good indicators of whether the child
  is paying attention.
• Introductory picture is sometimes required,
  sometimes desirable, in order to set a context
  (introduction of characters)

21
Comparison Truth Value Judgment/Picture Selection

• 3-6 year old children allow pronouns to corefer
  with local c-commanding antecedents (Chien
  Wexler 1990 Koster 1993).
• (6) a. John touched him. (50 adultlike)
• b. John touched himself. (almost 100
  adultlike)
• Do children lack the linguistic rules for
  pronoun interpretation or do they have problems
  with the application of these rules?

22

• It is logically possible that children dont know
  something thus they guess (e.g. 50)
• In this case, however, there should be no
  difference in what task they perform.
• But if it is ability to implement certain
  knowledge, then making a task less demanding may
  result in a better performance.

23
Processing approach Grodzinsky and Reinhart 1993

• Theoretical background (based on Reinhart 1983
  etc.)
• Binding vs. Coreference
• Traditionally
• Binding only with QP Coreference only with
  referring elements
• (7) Every boy believes he is smart ? ?x (boy (x)
  x believes x is smart)
• (8) The boy believed he was smart ? xboy yhe
  xy
• If we view DPs like the boy as generalized
  quantifiers, Reinhart claims, we should allow two
  readings for (8)
• ?x (boy(x) x believes x is smart) --------?bound
  variable anaphora
• xboy yhe xy -------------------------?
  coreference

24

• Sometimes the two readings will give us exactly
  the same interpretation, as in this case
• (9) The boy believed he was smart
• But sometimes not
• (10) Only Lucie loves her husband
• Bound variable Lucie (x) is the only person
  such that x loves xs husband
• Coreference Lucie loves the person who no one
  else loves

25
Back to Principle B

• Father Bear pinched him
• TWO possibilities
• Bound variable ?x (x pinched x)
• Coreference FBx himy, xy
• Claim only bound variable is subject to Pr. B
• But what prohibits coreference?
• Answer Rule I

26
Rule I

• Coreference is impossible if it gives you exactly
  the same interpretation as bound variable
  anaphora
• To check Rule I one has to compute both bound
  variable and coreference, and compare them are
  they the same?

27
Claim (Grodzinsky and Reinhart 1993)

• Such comparison is too difficult for children and
  aphasics because it requires keeping in short
  term memory two representations and comparing
  them.
• The task overloads their processing capacity
  thus they guess. Hence, a chance performance.

28

• In other words, there is no problem with
  knowledge, but limitation of processing
  resources.
• If this is true, different methodologies can give
  us different results.

29
Conflicting results?

• Example
• TVJT Philip and Coopmans 1996 (Dutch data)
• Children are at chance in Principle B
• PST --- Zuckerman et al 2002 (Dutch data)
• Children show almost 80 correct performance

30
Why such improvement?

• TVJT Children are forced to carry out a certain
  computation to compare two representations
• PST Children are free to choose it is only
  natural that they will try to avoid what is
  difficult that is to avoid making comparison
  between two representations

31

• Thus, the difference between experimental results
  becomes informative because it shows that the
  problem is not in knowledge, but in processing
  capacity.
• And it also shows that children are rational
  subjects they avoid operations that are
  resource consuming if there is an option to give
  a cheaper answer.

32
Prediction

• If you do the same experiment with the same
  children using the two methodologies, you should
  get a different results for the same group of
  children.
• Baauw and Zuckerman (2003) showed that this is
  the case.

33
PST vs.TVJT(same children!)
B Het jongetje raakte hem aan The boy
touched him.ECM Het jongetje zag hem dansen
The boy saw him dance.
34
Some suggestions

• Time limit not more than 30 min session
• Make sure children know/understand the
  toys/pictures and know the names of the
  characters (especially in cross-linguistic
  research!)
• In TVJT, ask children, whenever possible, why
  they say NO
• It is better to have two experimenters

35
Some suggestions

• Always have a familiarization session (so that
  children feel comfortable being tested they
  should believe it is all a game).
• Include practice items.
• Dont push the child! If he/she is no longer
  interested in playing, stop the session.

36
Some suggestions

• Beware of carry-over effect! Have enough fillers
  in-between.
• Always be positive to the child never say the
  child was wrong.
• Always test an adult control group.

37
The Main Conclusion

• It is not the case that there are good and
  bad methodologies.
• In many cases, it matters what you are looking
  for (e.g. acceptance vs. preference)
• This is the decision you need to make before you
  start designing a test.