Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

26
Relational Language Supports Relational Cognition Dedre Gentner Northwestern University http://www.psych.northwestern.edu/psych/people/faculty/gentner/ Research supported NSF SLC Grant SBE-0541957, the Spatial Intelligence and Learning Center (SILC) and by NSF – ROLE award number 21002/REC-

Transcript of Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Page 1: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Relational Language Supports Relational Cognition

  

Dedre GentnerNorthwestern University

  http://www.psych.northwestern.edu/psych/people/faculty/gentner/

 

Research supported NSF SLC Grant SBE-0541957, the Spatial Intelligence and Learning Center (SILC) and by NSF – ROLE award number 21002/REC-0087516

Page 2: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Language as Tool Kit

Relational Language and Relational Cognition

• Relational reasoning is central in higher-order cognition

• But relations are not obvious in the world

• So how do children acquire the stock of relations needed to understand and reason about the world?

Hypothesis: Relational language fosters the development of accessible relational structure and is thus central

to the development of analogy (Gentner, 2003)

Page 3: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Does the language we speak influence the way we think?

Two traditions in Cognitive Science

Sapir-Whorf hypothesis"We dissect nature along lines laid down by our native language... (Whorf, 1956: 213)

Vygotsky: Inner Speech“…learning to direct one's own mental processes with the aid of words or signs is an integral part of the process of concept formation." (Vygotsky, 1962, p. 59)

Page 4: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Development of Analogy: Relational Shift

• Young children have difficulty perceiving relational similarities– Initial focus on object similarities (Gentner, Halford)

Relational Correspondenc

e

Object Match

Page 5: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Relational Correspondenc

e

Object Match

“on”

“in”

“under”

Prior studies: English-speaking children who are given spatial relational language perform better on spatial analogy task than matched controls (Loewenstein & Gentner (Cognitive Psychology, 2005)

Page 6: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Plan

• Studies that vary presence of overt spatial language with English-speaking children (Loewenstein & Gentner, 2005)

• Studies of homesigners and hearing children in Turkey using the same task (Gentner, Ozurek, Goldin-Meadow & Gurcanli, in preparation)

Page 7: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Loewenstein & Gentner, 2005

Claim: Spatial Language Relational Encoding Analogical Mapping

Predictions:1. Overt spatial language will facilitate spatial encoding and

mapping for young children

2. Older children will no longer need overt language, because they already have internalized spatial language

3. However, language benefits will again appear for older children if the task is made more difficult

4. Language benefits derive from the semantics of the terms

5. Language influences the spatial representation; it is not merely a transient attentional effect

so benefits will be retained over time

Page 8: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Spatial Mapping Task: Neutral Version

For 6 trials (2/location):

Baseline: “I’m putting it here”

Language: “I’m putting it on/in/under the box”

Loewenstein & Gentner, 2005

Hiding Box

Finding Box

“the winner”

Page 9: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Search Trials

0

0.2

0.4

0.6

0.8

1.0On, In, Under

Baseline

3;8N=20

4;1N=20

*

Pro

port

ion

Corr

ect

Results • an early advantage of relational language

• which disappears with age

On, In, Under

Page 10: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Prediction: Language benefits will be seen for older children on this difficult task

For 6 trials (2/location):

Baseline: “I’m putting it here”

Language: “I’m putting it on/in/under the box”

Hiding Box

Finding Box

“the winner”

Loewenstein & Gentner, 2005

Spatial Mapping: Cross-mapped task

Page 11: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Pro

port

ion

Corr

ect

0

0.2

0.4

0.6

0.8

1.0

3;8 yrs 4;1 yrsN=20

4;7 yrsN=20

5;2 yrsN=20

On, In, Under

Baseline

*

*

On, In, Under

• Young children fail on a difficult task • Older children show benefits of overt relational language

Page 12: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Language benefits depend on the semantics of the terms: top middle bottom

ON

IN

UNDER

TOP

MIDDLE

BOTTOM

Monotonic

Height Increase

On, In, Under

Local Figure-ground Relations

Top, Middle, Bottom

Connected System of Relations

Page 13: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

DifficultCross-Mapping

Relational Match with Competing Object Match

Neutral Objects

Relational Match Only

Spatial Language andSpatial Mapping

Pro

port

ion

Corr

ect

Pro

port

ion

Corr

ect

Pro

port

ion

Corr

ect

0

0.2

0.4

0.6

0.8

1

On, In, Under

Baseline

*

3;8 4;1 4;7 5;2

*

*

3;0

0

0.2

0.4

0.6

0.8

1

On, In, UnderBaseline

*

3;8 4;1 4;7 5;2

*

*

3;0

3;8 4;1 4;7 5;23;0

1

0

0.2

0.4

0.6

0.8Top, Middle, Bottom

*

*

*

Page 14: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Language benefits are retained over time

0.0

0.2

0.4

0.6

0.8

1.0

3;8N=30

4;1N=30

Pro

port

ion

Corr

ect

Search Trials

Baseline

TMB*

2dayslater

Retention test:Day 1. Standard task

Day 2. Two days later, “same game” -- No spatial language --

Both groups getsame procedure

Page 15: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

What about children who lack spatial language?

• Deaf children of hearing parents in Turkey who have not been taught sign language

• Turkish hearing children (matched on cognitive task)

Gentner, Ozyurek, Goldin-Meadow & Gurcanli

Background: Informally, the deaf children appear to use few if any spatial terms.

In contrast, hearing Turkish children appear to use both spatial nouns and spatial case markers (postpositions)

Page 16: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Plan

1. Spatial language elicitation task given to deaf children and hearing children

2. Spatial analogy task•Deaf and hearing children matched on a separate cognitive task

•Spatial analogy task given to both groups 1. neutral task 2. cross-mapping task

Gentner, Ozyurek, Goldin-Meadow & Gurcanli

Page 17: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Spatial Language Elicitation Task

Children were shown short videotaped events• 24 simple motion events with toys.• In each event, a figure moves toward a goal object

and ends {near, behind on top of, or in} the goal obj• Transitive and intransitive:

a boy moves next to a girla baby crawls to a cata man gives a flower to a woman

a motorcyclist gets on top of a motorcycle

After each clip, the child was asked to describe what happenedeither in speech (hearing chidren)or in their own homesign (deaf children)

Children were given a page showing the characters in action to help them describe the event

Gentner, Ozyurek, Goldin-Meadow & Gurcanli

Page 18: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Ozyurek, Gentner, Goldin-Meadow & GurcanliCODING

HEARING CHILDREN:Case Markers: e.g., from / to / towardsSpatial Nouns: e.g., near /on top/ over/ behind/ inside

ottobus kiz-a gittiBUS GIRL-TO WENT

ottobus kizin yanin-a gittiBUS GIRL NEXT-TO WENT

DEAF CHILDREN:points and actionspoints to goalsactions (traces of paths)two-handed gestures

Case Markers:

a point made NEXT TO, ON, UNDER, etc., an object Spatial Nouns:

two-handed gesture,: e.g., one hand ON TOP OF the other, one hand IN the other

Page 19: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Children's Descriptions of Spatial Vignettes

Gentner, Ozyurek, Goldin-Meadow & Gurcanli, in preparation

0.920.96Proportion of vignettes with one or more action verbs

2.02.6Sentences per vignette

8.09.1Words per vignette

DeafHearing

0.050.56Proportion of sentences with case markers(over all sentences)

*

0.00.26Proportion of sentences with spatial nouns(over all sentences)

Page 20: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Summary and Predictions

– The homesigners match hearing children in amount of action and object talk.

– But they appear far less likely to develop and use terms that directly denote spatial relations.

– If relational language is important in achiving a clear relational representation, then the homesigners may be less likely to form such delineated spatial representations

– And therefore they may be less able to carry out spatial analogies as in the box-mapping task.

Page 21: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Gentner, Ozyurek, Goldin-Meadow & Gurcanli

Relational Match Only

Neutral

Relational Match with Competing Object Match

X-Map

Participants:• 13 Deaf Turkish children with invented Homesign systems• 13 Hearing children (native speakers of Turkish)

Deaf and hearing matched for performance on separate spatial cognitive task

All children receive Neutral task followed by Cross-mapped task

Page 22: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Deaf and hearing children were matched on Levine et al. (1999) Spatial Task

Example of a spatial manipulation task: Whichfigure can you make with the two top figures?

Page 23: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Performance on Neutral and Cross-mapped Tasks

0.00

0.20

0.40

0.60

0.80

1.00

Deaf Hearing

Pro

po

rtio

n c

orr

ect

.

Neutral

Xmap

N=13 N=13

**t(24) = 2.785, p=.01* t(24) = 2.115, p-.05

***

Page 24: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

DifficultCross-Mapping

Relational Match with Competing Object Match

Neutral Objects

Relational Match Only

Spatial Language andSpatial Mapping

Pro

port

ion

Corr

ect

Pro

port

ion

Corr

ect

Pro

port

ion

Corr

ect

0

0.2

0.4

0.6

0.8

1

On, In, Under

Baseline

*

3;8 4;1 4;7 5;2

*

*

3;0

0

0.2

0.4

0.6

0.8

1

On, In, UnderBaseline

*

3;8 4;1 4;7 5;2

*

*

3;0

3;8 4;1 4;7 5;23;0

1

0

0.2

0.4

0.6

0.8Top, Middle, Bottom

*

*

*

Page 25: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

Summary– Using language for spatial relations invites a correspondingly

precise encoding of spatial relational structure

– This relational encoding – promotes analogical mapping in young children– is semantically specific– is retained over time

• Deaf homesigners who lack clear spatial terms are disadvantaged in the spatial mapping task

• Language may not be required for the task; deaf children are able to learn the task, although at an older age. But language is an important facilitator

• These findings are consistent with the idea that relational language provides tools for encoding and using relations in the world

Page 26: Relational Language Supports Relational Cognition Dedre Gentner Northwestern University

 

The End

http://www.psych.northwestern.edu/psych/people/faculty/gentner/