Docstoc

Reinhard Blutner

Document Sample
Reinhard Blutner Powered By Docstoc
					    universiteit van
    amsterdam

               9th Szklarska Poreba Workshop On the Roots of Pragmasemantics. February 21-25, 2008   |1




a note on the simulationist
framework and motor theories of
perception

Reinhard Blutner
http://www.blutner.de
                                                                         Institute for Logic,
blutner@uva.nl                                                       Language and Computation
    universiteit van
    amsterdam
                                           February 21-25, 2008   |2




Outline
› Motor theories of perception and the simulationist
  framework
› Caramazza & Mahon's criticism
› Experiments on asymmetries of interpretation/
  production
› Conclusions & challenges
  universiteit van
  amsterdam

             9th Szklarska Poreba Workshop On the Roots of Pragmasemantics. February 21-25, 2008   |3




1 Motor Theories of Perception
and the Simulationist framework
   universiteit van
   amsterdam
                                          February 21-25, 2008   |4




James 1890
 An educated man . . . on entering his house one day
 he received a shock from crushing the finger of one of
 his little children in the door. At the moment of his
 fright he felt a violent pain in the corresponding
 finger of his own body, and this pain abode with him
 three days. (James, 1890, p. 712)
   universiteit van
   amsterdam
                                             February 21-25, 2008   |5




Lotze 1852
 The spectator accompanies the throwing of a billiard
 ball, or the thrust of a swordsman, with slight
 movements of his arm; the untaught narrator tells his
 story with many gesticulations; the reader, while
 absorbed in the perusal of a battle scene, feels a slight
 tension run through his muscular system, keeping
 time as it were with the actions he is reading of.
 (Lotze, 1852; as quoted in James, 1890, p. 525)
        universiteit van
        amsterdam
                                 February 21-25, 2008   |6




   Mirror Neurons

 A mirror neuron is a neuron
 which    fires   both    when
 performing an action and when
 observing the same action
 performed by another creature


› Monkey grasp a nut
› Monkey sees how
  another creature
  grasp a nut
        universiteit van
        amsterdam
                                 February 21-25, 2008   |7




   Mirror Neurons

 A mirror neuron is a neuron
 which    fires   both    when
 performing an action and when
 observing the same action
 performed by another creature


› Monkey grasp a nut
› Monkey sees how
  another creature
  grasp a nut
  universiteit van
  amsterdam

             9th Szklarska Poreba Workshop On the Roots of Pragmasemantics. February 21-25, 2008   |8




2 Caramazza& Mahon's Critizism
    universiteit van
    amsterdam
                                                 February 21-25, 2008   |9




Thesis 1 (weak motor theory of perception)
The perception of actions is constrained by the implicit
knowledge that the central nervous system has
concerning the movements that it is capable of producing
itself.
- The mechanism by which humans perceive others differs
  greatly from how humans perceive inanimate objects.
- Unlike inanimate objects, humans have the distinct property
  of being “like me” in the eyes of the observer.
- This allows us to use the same systems that process knowledge
  about self-performed actions, self-conceived thoughts, and
  self-experienced emotions to understand actions, thoughts,
  and emotions in others.
    universiteit van
    amsterdam
                                                  February 21-25, 2008   | 10




Thesis 2 (strong thesis of simulation)
The perception of actions/emotions is realized by
performing simulations of the relevant movements/
emotions (analysis-by-synthesis mechanism).

Proving thesis 2 requires demonstrating that
 - production programs are necessarily run in the course of
   perception/recognition/understanding
 - production programs are sufficient to ground the conceptual
   content of perception/recognition/ understanding
      universiteit van
      amsterdam
                                                     February 21-25, 2008   | 11



Are Production Programs Needed for Perception?

› 3- to 5-month-old infants can discriminate between a
  point light walker and similar figures with scrambled
  spatial relationships between the moving dots.
   - Given that 3- to 5-month-old babies do not have experience
     with walking, what are they simulating in order that they can
     recognize walking?

› Subjects who had bilateral paralysis of the face from
  infancy (Mobius syndrome) were not impaired on a
  test of facial affect recognition.
   - What are they simulating?
      universiteit van
      amsterdam
                                                     February 21-25, 2008   | 12



Are Production Programs Sufficient for Perception?

› Perceiving a person that drinks a glass of water
   - Depending on the context you can infer that the person was
     thirsty
   - In another context you may infer the person was nervous
   - In both cases the motor programs (for drinking water) are the
     same!

› Hence, intentions       cannot     be   inferred    by       using
  simulations
    universiteit van
    amsterdam
                                            February 21-25, 2008   | 13




Preliminary Conclusions
› Caramazza concludes from such observations that
  simulations are epiphenomenal. In contrast, I think
  the play a real role in forming conceptual knowledge
› Simulations may contribute in an important way to
  the full “meaning” of object and event concepts (such
  as hammer and drinking)
› However, I doubt that perception/understanding is
  realized by actually running simulations. Rather, their
  role consist in restricting perception/understanding
  via the forming of the relevant knowledge structures.
  universiteit van
  amsterdam

             9th Szklarska Poreba Workshop On the Roots of Pragmasemantics. February 21-25, 2008   | 14




3 Experiments on Asymmetries
of Interpretation/Production
    universiteit van
    amsterdam
                                                February 21-25, 2008   | 15




Zeevat„s Metaphor is Misleading but True

“The situation can be fruitfully compared to the habit of
hiding easter eggs for one's children. The parents
engaged in hiding the eggs balance the amount of effort
with the desired amount of difficulty in finding the egg.
(They also picture the child looking for it and try to keep
it possible for the child of finding the egg, without
spoiling the fun.) For the child it is another matter. They
just have to throw in the effort required for finding the
eggs. Not more of course, but definitely not less. It is not
a complicated balancing act.” (Zeevat 2000: 245)
    universiteit van
    amsterdam
                                         February 21-25, 2008   | 16




A Production Problem for R-Expressions

› Hendriks, Englert, Wubs & Hoeks (to appear): Age
  differences in adults’ use of referring expression
› Children and very old adults produce pronouns where
  R-expressions are more appropriate
     universiteit van
     amsterdam
                                                    February 21-25, 2008   | 17




Sentence Elicitation Study

 A woman hold-ing
 an ice cream cone      The woman          She gives the girl
 is walking past a      comes across a     an ice cream cone.
 road sign.             girl.
     Topic shift          Target Picture




 The girl is eating             she
                        Well, the woman    The woman buys
 from the ice           passes again an    another ice cream
 cream cone.            ice cream van.     come.
     universiteit van
     amsterdam
                                            February 21-25, 2008   | 18



Results
› Elderly adults produce (non-recoverable) pronouns
  significantly more often than young adults when refer-
  ring to the old topic in the presence of a new topic.
› With respect to the comprehension task, no significant
  differences were found between elderly and young
  adults.
     universiteit van
     amsterdam
                                          February 21-25, 2008   | 19




Discussion of the Results
› In principle, I see two possibilities to overcome the
  production problem:
   - Fossilization (via bidirectional learning)
   - The producer learns to reason bidirectionally
› The observation that elderly people have a likewise
  production problem excludes the fossilization idea
› Defossilization doesn„t make any sense
› Hence, the speaker seems to take the hearer into
  account!
      universiteit van
      amsterdam
                                                February 21-25, 2008   | 20




The Pronoun Interpretation Problem
› Young children have problems in interpreting pronouns
  but not in interpreting reflexives.
› Hendriks and Spenader (2004) give a new interpretation of
  children„s delay of the comprehension of pronouns.
› I discuss the validity of this interpretation and present an
  alternative account in terms of iterated learning
› The conclusion will be that there is no empirical evidence
  that he hearer takes the speaker into account (in
  incremental natural language understanding)
     universiteit van
     amsterdam
                                                      February 21-25, 2008   | 21




Children‟s Comprehension of Reflexives*
› Here is an elephant and an alligator. The elephant is
  hitting himself.




› Question: Does the sentence match the picture?
› Children from age 4 on: Yes

 * I thank Petra Hendriks for allowing me to use her slides
     universiteit van
     amsterdam
                                           February 21-25, 2008   | 22




Children‟s Interpretation of Pronouns
› Here is an elephant and an alligator. The elephant is
  hitting him.

E.g., Chien & Wexler, 1990; Grimshaw
& Rosen, 1990; Jakubowicz, 1984;
Koster, 1993; McDaniel, Smith Cairns,
& Hsu, 1990; McDaniel & Maxfield,
1992; McKee, 1992.; Spenader, Smits &
Hendriks, subm.

› Question: Does the sentence match the picture?
› Children until at least the age of 6 or 7: Yes
    universiteit van
    amsterdam
                                                   February 21-25, 2008   | 23




Children‟s Production of Pronouns
› Task: Describe what you see on the picture.


                       Cf. De Villiers, Cahillane, & Altreuter, 2006;
                       Spenader, Smits, & Hendriks, subm.



                               The elephant is hitting him.

› Children between 4;6 and 7:
    universiteit van
    amsterdam
                                              February 21-25, 2008   | 24




Children‟s Production of Reflexives
› Task: Describe what you see on the picture.




                            The elephant is hitting himself
› Children between 4;6 and 7:
                             *The elephant is hitting him
    universiteit van
    amsterdam
                                             February 21-25, 2008   | 25




H & S „s explanation: Delayed Bidirection
› The proposal is that children begin with unidirectional
  optimization, and only later acquire the ability to
  optimize bidirectionally.
› Optimizing bidirectionally inherently involves reasoning
  about alternatives not present in the current situation,
  which may be a skill acquired very late, thus explaining
  the lag in acquisition.
› Optimizing bidirectionally requires more processing
  resources than unidirectional optimization
    universiteit van
    amsterdam
                                            February 21-25, 2008   | 26




Alternative Explanation: Fossilization
› The fossilization view applies OT learning theory as a
  mechanism of conventionalization/automatization/
  routinization. It is related to an instance theory of
  automatization (Logan 1988)
› Conceptual advantage of the fossilization account: no
  mind reading capacities are required for the
  processing tasks
› Fossilization accounts for the fact hat elderly people
  (or stressed preople) have no problems with pronoun
  interpretation.
  universiteit van
  amsterdam

             9th Szklarska Poreba Workshop On the Roots of Pragmasemantics. February 21-25, 2008   | 27




4 Conclusions and Challenges
     universiteit van
     amsterdam
                                                      February 21-25, 2008   | 28



Online Processing, Offline processing & Learning
› In NNs & OT, online processing is basically unidirectional
  optimization: expressive vs. interpretive optimization.
› Under special conditions the speaker can take the hearer into
  account (however, normal incremental generation processes
  produce garden path sentences; e.g. “The coach knew you
  missed practice” (Feirerra & Dell 2000)
› I don„t see any empirical evidence that the hearer takes the
  speaker into account (in online processing)
› Offline processing can be bidirectional (grammaticality
  judgments)
› Iterated learning and individual/cultural fossilization is a really
  existing phenomenon.
    universiteit van
    amsterdam
                                                               February 21-25, 2008   | 29



Invited Inferencing Theory of Semantic Change
(Traugott & Dasher 2005; picture taken from Benz 2007)

                          Conventionalization
                        (Generalised Implicature)


                                                          Semantic
                                                          Reanalysis
Global Interpretation Mechanism
  (Particularised Implicature)




    Coded meaning                                   New coded meaning
     (source stage)                                   (target stage)
     universiteit van
     amsterdam
                                                  February 21-25, 2008   | 30



A Small List With Challenges
› What is the best solution to the rad/rat problem?
› The treatment of garden path effects in incremental
  interpretation (the listener„s view taken in OT semantics is
  basically motivated by such examples)
   - The coach knew you missed practice
› Accounting for the findings in experimental pragmatics (scalar
  implicatures, interpretation of reflexives, pronouns, R-
  expressions)
› Taking over the sentence production in mid sentence by the
  hearer. Why does this phenomenon exclude other models of
  incremental processing?
› Reinhard Blutner & Anatoli Strigin: Asymmetric OT, Bidirection
  and Fossilization