Beyond Salience: Interpretation of personal and demonstrative pronouns

1
Beyond Salience Interpretation of personal and
demonstrative pronouns

• Brown-Schmidt, Byron, Tanenhaus, 2005

2
Outline

• Background
• Givenness Hierarchy
• Brown-Schmidt et al. suggestion
• Expt1
• Expt2
• Expt3
• General Discussion

3
Pronoun and Discourse

• Multiple ways to refer to the same object (e.g.
  Mug, it, that)
• First mentioned nouns in a compound subject
  (Donald and Minnie went ...) are more likely to
  be produced as a pronoun than section mention
  nouns
• Pronoun interpretation influenced by
  non-linguistic factors such as visual salience
  (Arnold Lao, 2008 CUNY conference poster)

4
Givenness Hierarchy

• Discourse entities vary in salience (cognitive
  status)
• Discourse entities can be group along a hierarchy
  of salience
• Pronouns can likewise be grouped according to
  this hierarchy
• Focus hierarchy works like Horn scales of
  informativeness

5
Sample Givenness Hierarchy
Status In focusgt Activatedgt Familiargt Uniquely
identifiablegt Referentialgt Type
identifiable Form It/them/they This/that That
(noun)/ The (noun) This (noun, indefinite use)
A (noun) This (noun)
In focusgt Activatedgt It/them/they This/that
Table 1. From Brown-Schmidt et al.
6
Issues with the salience account

• Salience may not be the determining factor
  between usage or it and that
• Demonstrative pronouns can be used to refer to
  things with a similar sense to a previous
  referent (Channon, 1980)
• Example
• Patron 1 Ill have a hamburger and fries.
• Patron 2 Ill have that, too.

Example from Brown-Schmidt et al. 2005
7
Conceptual Composites

• Pronominal reference for conceptual composites
  less clear
• In previous example, hamburger and fries were
  potentially equally focused as entities
• Demonstratives may prefer more complex entities
  as referents

8
Hypotheses

• Givenness Hierarchy
• it prefers most focused entity as referent
  while that prefers less focused entities as
  referents
• Counter proposal
• it prefers salient entities while that
  prefers conceptually complex entities

9
Experiment 1

• Real world object manipulation task
• Set of four instructions to manipulate objects
• Two different types of objects, blocks or
  household items
• Two factors
• Pronoun (it/that)
• Type of instruction (on top/next to)
• Four objects in the display
• Household items were conceptually grouped
  together (e.g. Lamp and table)

10
Example Display

• Put the cup (on/next to) the saucer.
• Now put (it/that) over by the lamp
• Put the table next to the lamp.
• Now put the cup in front of the table.

11
Predictions

• Givenness
• it will prefer the most focus entity, the
  previous theme
• that will prefer the less focused entity, the
  previous goal
• Complexity
• it will prefer the salient entity, the previous
  theme
• that will prefer the conceptual composite if
  available

12
Results

• Main effect of pronoun type
• it preferred theme more than that
• Main effect of object location
• next to has more theme interpretations than on
  top
• Main effect of object type
• More theme interpretations for blocks
• Interaction effect of object type/location
• More theme interpretation for blocks compared to
  other objects when instruction was on top

13
Selections by condition
Fig. 2. Experiment 1 referent selections for
blocks (A) and objects (B) conditions split by
object location (on top/next to), and pronoun
(it/that). Grey portion, theme responses white,
composite black, goal.
14
Expt1 Proportion of fixations
Fig. 3. Experiment 1. Relative proportion of
fixations to theme, goal and destination for next
to/it (A), next to/that (B), on top/it (C), and
on top/that (D), for both blocks and objects.
15
Action contingent analysis
Fig. 4. Experiment 1. Relative proportion of
fixations to theme, goal, and destination for on
top/that (chose composite), and next to/it (chose
theme), respectively, for blocks and objects
together.
16
More contingent analysis
17
Summary Expt 1

• The pronoun it preferred the most focused
  entity
• Occasionally interpreted as a conceptual
  composite (but still showed early theme bias)
• The pronoun that preferred conceptual composite
  interpretation
• When composite not available, that prefers a
  task relevant entity

18
Experiment 2 (prosody)

• The pronoun that tends to get more stress than
  the pronoun it
• Stressed pronouns tend to prefer less salient
  entities
• Example
• Lolita slapped Doris and then she hit Hubert.
• Lolita slapped Doris and then SHE hit Hubert.
• Stress may determine referent when context is
  complementary (Kameyama, 1999)

19
Experiment 2 (goal)

• Test 3 hypotheses
• Stress of a personal pronoun shifts
  interpretation away from focus
• Stressing a personal pronoun results in
  preference of least salient entity
• Stress is the primary difference between
  unstressed it and regular stress that

20
Experiment 2 (method)

• Same task as Expt 1
• Pairs no longer functionally related
• Manipulated stress of it pronoun
• Manipulated the stress of that pronoun
• Two version 2a and 2b
• Version 2a included naturally produced sentences
• Version 2b spliced the critical word into a
  carrier sentence and removed the word Now

21
Experiment 2 Results

• Replicated Experiment 1
• Stress version of it led to fewer theme
  responses
• No effect of stress for that
• Interaction effect of pronoun by stress

22
Responses 2a and 2b
2a
2b
23
Experiment 2 Summary

• Effect of stress for personal pronouns supported
• No support for complementary hypothesis
• Inconclusive results regarding the main
  difference in it and that is stress
• Stress matters for it
• Stress seems less important for that

24
Discourse coherence

• Discourse centering theory
• Cross sentence coherence is a better predictor of
  pronoun preference than within sentence salience
• Backward looking center is the least oblique part
  of the previous sentence that is repeated in the
  next sentence
• SubjectltDOltIndirect OltAdjuncts

25
Experiment 3

• Same method as Experiment 1
• Compared conditions with 2 sentences before
  pronoun to conditions with only 1 sentence per
  pronoun
• Found no effect of the manipulation

26
Experiment 3 Offline Results
Fig. 9. Experiment 3. Proportion of theme, goal,
and composite selections for trials with two
context sentences before the pronoun (A), and
trials with one context sentence before the
pronoun (B), split by pronoun (it/that) and
location (next to/on top). Grey portion, theme
responses white, composite black, goal.
27
Summary

• Interpretation of the pronoun it seems to be
  affected mostly by salience
• Interpretation of it is modulated by stress
• The pronoun that prefers conceptually complex
  referents when available