Informativeness, Relevance and Scalar Implicature

1
Informativeness, Relevance and Scalar Implicature

• Author Roybn Carston
• Presenter Ovidiu Fortu

2
Outline

• Introduction
• Pragmatics
• Principles underlying the implicature phenomenon
• Types of implicatures
• Examples

3
Introduction

• Pragmatics
• Implicature anything that is inferred from an
  utterance but that is not a condition for the
  truth of the utterance. (Gazdar, 1979)
• Presupposition anything that is presupposed to
  be true given an utterance

4
Presupposition

• Possible criterion given an utterance U, the
  proposition p that is inferred by listener from
  both U and not U is a presupposition
• Example
• The king of France is bold.
• The king of France is not bold.
• From both sentences, the affirmation and the
  negation, we infer that there is a king of France

5
Implicature

• Implicatures are inferred based on the assumption
  that the speaker observes or flouts some
  principles of cooperation (different authors have
  identified different principles)
• Grice 4 principles (so called maxims)
• Levinson (1981), Horn (1984) 2 principles

6
Grice Principles

• Quantity maxim
• the communication must be adequately but not
  overly informative
• Quality maxim
• the speaker does not believe it to be false and
  has adequate evidence for his statement
• Maxim of relation or relevance
• the communication must be relevant
• Maxim of manner
• the communication must be clear, unambiguous,
  brief, and orderly

7
Grice Principles, reduced form

• The Q-principle
• Say as much as you can (given I)
• The I-principle
• Say no more than you must (given Q)

8
Types of Implicatures

• Standard implicature based on the assumption
  that the speaker observes the cooperation
  principles
• A Ive just run out of petrol.
• B There is a garage just around the corner.
• B infers that he can find oil at the garage.

9
Types of Implicatures

• Flouting implicatures based on the assumption
  that the speaker deliberately flouts one of the
  communication principles
• A The capital of Morocco is Casablanca
• B Yes, and the capital of U.K. is Moskow
• A infers that his statement was wrong.

10
Types of Implicatures, another classification

• Generalized implicatures inferred without a
  special reference to context
• John walked into a house yesterday.
• Infer that the house was not Johns house
• Particularized implicatures inferred only due
  to a special context
• A Can you tell me the time?
• B Well, the milkman is here.
• It must be the time when the milkman comes.

11
Properties of Implicatures

• Strong dependency on context (see the complex
  implicature example)
• Defeasibility (they are not entailments, and
  addition of new facts can cancel them)

12
Why is the problem of implicature hard?

• Deals with the logic defying aspects of
  communication
• The cooperation principles are hard to formulate
  (work is still done in this area, and no author
  claims he has a final form of the principles)
• Implicatures are hidden, i.e. they do not
  appear in text, which makes a statistical
  approach less accessible

13
Scalar Implicature

• Lexical (and logical) scales
• all, most, many, some
• numbers
• subset, set
• According to the cooperation principles, the
  speaker must use the right member of the scale

14
Scalar Implicature, Examples

• Bill has got some of Chomskys papers
• Infer that Bill does not have all the Chomskys
  papers
• There will be five of us for dinner tonight
• Infer that there will not be more than five of us
  for dinner tonight
• A I like Mary. She is intelligent and good
  hearted.
• B Yes, she is intelligent.
• Infer that B thinks Mary is not good-hearted

15
Complex Scalar Implicatures

• Scenario Kais parents promise him rewards for
  things he does not like to do a small reward for
  washing his hair, a medium reward for eating
  broccoli and peas, and a high reward for cleaning
  up his room.
• Kais mother says
• Kai had broccoli and peas.
• We infer that Kai did not clean up his room

16
Scalar Implicatures

• Based on the Q-principle
• The speaker must not make a weaker claim (i.e.,
  he must say as much as he can, as long as this
  does not increase the effort)
• It takes the same amount of effort to say
• John walked into his house yesterday.
• John walked into a house yesterday.

17
Other Types of Scales

• Ranked entities
• A Is Jill a professor yet?
• B Shes a senior lecturer.
• Infer that Jill is not a professor.
• Whole/part relation
• A Did you manage to read that chapter I gave
  you?
• B I read the first couple of pages.
• Infer that B didnt read the whole chapter.

18
Other Types of Scales

• Instance-of
• A Do you have any juice?
• B I have grape, orange and tomato.
• Infer that B does not have any apple, lemon..
• Alternate values (not necessarily ordered)
• A Did you get Paul Newman's autograph?
• B I got Joanne Woodward's.
• B didnt get Newmans autograph

19
Quantity principle, refined

• Welker (1994) shows that the quantity principle,
  as formulated by Grice, is too strong
• A I'm having a dinner party and I need four more
  chairs.
• B John has two chairs.
• Implicature B has at most two chairs
• A I'm having a dinner party and I need four more
  chairs.
• B John has four chairs.
• This time, no implicature

20
Quantity principle, refined

• Communication must be "... as informative as is
  required (for the current purposes of the
  exchange)"
• Idea even the scalar implicatures depend on
  context not only the surrounding text, but also
  the situation

21
False Predictions

• Not all scales generate implicatures all the time
• The relevance principle may cancel some scalar
  implicatures
• A What did you buy for your mother?
• B I bought her flowers.
• Assuming that roses are on top of the flowers
  scale, this leads to the implicature I didnt
  buy her roses.

22
False Predictions, continued

• The implicature is not inferred because the
  statement is relevant enough
• However, a possible implicature in this example
  is I didnt buy her a present. The difference
  is that while it is not relevant which kind of
  flowers he bought, it is relevant whether he
  bought a present or not

23
Matsumotos constraint

• Let ltS, Wgt be a scale (with S stronger than W)
• Then a scalar implicature is inferred if the
  following condition is met
• the choice of W instead of S must not be
  attributable to the observance of the maxims of
  quantity-2, relation or obscurity avoidance
  (manner-1).

24
Matsumotos constraint, contd

• Equivalently, observing the quantity-2,
  non-obscurity and relevance takes precedence over
  observing quantity-1
• Idea the relevance maxim seems to hold the key
  to the process of inferring implicatures

25
Affirmative Implicatures

• So far, the scalar implicatures seem to simply
  negate the stronger claim when the weaker is
  presented
• We can also have implicatures that do not involve
  negation
• If you finish your thesis by September you'll be
  eligible for the job.
• Implicature You'll be eligible for the job if
  and only if you finish your thesis by September.

26
Pragmatic Schemes

• Let S and W be members of a scale, with S
  stronger than W
• Q-based implicature
• S entails W
• "W" implicates "not S"
• R-based implicature
• S entails W
• "W" implicates "S"

27
Pragmatic Schemes, applied

• P and Q entails P or Q
• "P or Q" Q-implicates "not P and Q"
• Thus the implicature is not P or not Q, or only
  one of P and Q can hold
• P iff Q entails if P, Q
• "if P, Q" R-implicates "P iff Q"
• If you finish your thesis by September you will
  be eligible for the job as seen above, the
  implicature is that the condition is necessary

28
Informativeness

• In both previous examples, the implicatures
  enrich the informational contents of the message
• Observation What is conveyed always implies
  logically what is said
• Conclusion the implicature mechanism allows the
  quantity of information in a message to grow

29
RichardsonRichardson critique

• I broke a finger.
• implicates I broke one of my own fingers.
• I found a finger.
• implicates I found someone else's finger.
• Which of the schemes can be applied?
• Q-implicatures or R-implicatures?
• Again, relevance is the key

30
Cardinal numbers

• Problem A and B go to a party. They make a bet,
  A says that there will be 20 people at the party
  when they arrive. When they get to the party,
  there are 25 people. Who wins the bet?

31
Cardinal numbers, ambiguity

• The source of ambiguity is the use of numbers
  the sentence there will be 20 people can be
  used to express
• There will be at most 20 people there.
• There will be exactly 20 people there.
• There will be at least 20 people there.
• The context of the bet supports the second
  interpretation

32
Cardinal numbers, continued

• In Britain you have to be 18 to drive a car.
• The new houses are big enough for families with
  three children.
• A default reasoning (world knowledge is
  essential) decides the interpretation (at most
  at least)

33
Conclusions

• The Q principle and R(I) principle give rise to
  the same result a strengthening of the meaning
  of the utterance
• The relevance principle plays a key role, which
  constrains the Q and R principles
• Cardinal numbers are a special case of scale
  they allow punctual interpretation, but also
  interval interpretation

34
Scalar implicatures experiments at the
semanticspragmatics interface

• Authors
• Anna Papafragoua, Julien Musolinob
• Presenter Ovidiu Fortu

35
Paper contents

• A study of how scalar implicatures are inferred
  by humans
• Two sets of experiments are performed with a
  group of young children to test their ability to
  infer implicatures
• Only scalar implicatures are considered for the
  tests

36
First Set of Tests

• Subjects30 5-year-old children and 30 adults,
  all native speakers of Greek (all experiments in
  Greek)
• Three scales
• oli, meriki (all, some)
• tris, dio (three, two)
• teliono, arxizo (finish,start)

37
Experiment setup

• Subjects are presented a situation that allows a
  stronger claim
• Subjects (both adults and children) answer
  questions about the situation
• Questions admit yes/no answers (the subjects must
  assess the truth value of a claim in the given
  context

38
Results

• While adults have no problem of inferring
  implicatures, children seem to be less sensitive
  to weak clauses
• Only 10 - 12.5 of the weak claims in case of
  scale (all, some)
• However, for other scales, children have better
  results 65 in case (three, two), which shows
  that different scales are perceived differently

39
Justifications for Answers

• Subjects were also required to provide a brief
  justification for their answers
• Adults overwhelmingly justified their answers by
  stating the stronger claim
• Children gave two types of justifications
• Repeating of the given statement
• The stronger statement
• Even in cases when they gave the right answer,
  the children had wrong justifications (rougly 70
  of the justifications for the scale ltall, somegt
  were of the first type)

40
Experiment 2

• Subjects a set of 30 children (distinct from the
  first set)
• Children were trained to recognize pragmatic
  anomaly
• The stories that described the situations were
  modified to focus on the performance of the
  principal character

41
Experiment 2, Results

• Children could reject the weak statements
  reliably better
• 52 ltall, somegt (previously 12)
• 47.5 ltfinish, startgt (previously 10)
• 90 ltthree, twogt (previously 65)

42
When children are more logical than adults
experimental investigations of scalar implicature

• Author
• Ira A. Noveck
• Presenter Ovidiu Fortu

43
Objectives

• Study the scalar implicatures experimentally
• Establish that scalar implicatures are
  psychologically real and common in reasoning
  scenarios
• Establish how this class of weak scalar terms
  develops

44
Three sets of experiments

• Experiment 1 modal might (when it is
  comparable with must)
• Experiment 2 a follow up of experiment 1
  designed to determine the extent to which the
  scalar implicature can be suspended
• Experiment 3 investigates weak claims based on
  the some quantifier

45
Experiment 1

• Subjects 32 5-year olds, 20 7 year olds, 16 9
  year olds, 20 adults (all native English
  speakers)
• Set-up two open boxes, one with a parrot and one
  with a parrot and a bear
• A puppet then states 8 claims
• (1) There has to be a parrot in the box (true)
• (2) There does not have to be a parrot in the box
  (false)
• (3) There might be a parrot in the box (true)
• (4) There cannot be a parrot in the box (false)
• (5) There has to be a bear in the box (false)
• (6) There might be a bear in the box (true)
• (7) There does not have to be a bear in the box
  (true)
• (8) There cannot be a bear in the box (false)

46
Experiment 1, Results

• Is the statement of the puppet true?
• Necessary conclusion (parrot)
• Has to be a parrot Yes 75 90
  88 100
• Does not have to be a parrot No 72 75 75
  100
• Might be a parrot Yes 72 80
  69 35
• Cannot be a parrot No 66 80
  100 100
• Total
  73 81 83 83
• Possible conclusion (bear)
• Has to be a bear No 47 65
  88 100
• Does not have to be a bear Yes 66 75 81 100
• Might be a bear Yes 53 80
  100 100
• Cannot be a bear No 53 80
  100 100
• Total 55
  75 92 100

47
Experiment 3

• The results and setup were very similar to the
  ones in previous paper
• The tests with older subjects showed better
  results more than 85 for 7-year olds 10 year
  olds had performance comparable to adults
• Children have more problems with this scale (in
  one test only 6 rejected all weak claims)

48
Conclusions

• Ability to communicate using pragmatics is
  developed later in the growth
• Implicatures are difficult to infer, requiring
  more experience