Role of language in conceptual development Combinatorial Nature of Language Case Study #1: Reorientation.

Role of language in conceptual developmentCombinatorial Nature of Language

Case Study #1: Reorientation

Hide a sticker hereChildren search here ~50% of the time

And here ~50% of the time

Even when one wall is red!

Hermer & Spelke, 1994*Lecture notes adapted from slides provided by A. Shusterman

1. Creatures with little or no spatial language (like rats, or children before the age of 5) fail to search correctly in a disorientation task.

The age at which children acquire adult-like search patterns correlates with the age at which they can produce the phrases “left of X” and “right of X.”

Evidence that language might support spatial representation

39

39

12

10

1.5 - 2 YEAR OLDS (Hermer & Spelke, 1996)

No red wall

49

31

12

8Red wall

1. Creatures with little or no spatial language (like rats, or children before the age of 5) fail to search correctly in a disorientation task.

The age at which children acquire adult-like search patterns correlates with the age at which they can produce the phrases “left of X” and “right of X.”

2. Only verbal shadowing impairs search in human adults.

Evidence that language might support spatial representation

5

83

8

5

VERBAL SHADOWING (Hermer, Spelke, & Katsnelson, 1999)

No shadow

33

36

15

17Shadow

v.s. RHYTHM SHADOWING

Evidence that language might support spatial representation

5

83

8

5

VERBAL SHADOWING (Hermer, Spelke, & Katsnelson, 1999)

No shadow

33

36

15

17Shadow

RHYTHM SHADOWING

0

95

3

2No Shadow

11

73

9

6Shadow

1. Creatures with little or no spatial language (like rats, or children before the age of 5) fail to search correctly in a disorientation task.

The age at which children acquire adult-like search patterns correlates with the age at which they can produce the phrases “left of X” and “right of X.”

2. Only verbal shadowing impairs search in human adults.

3. Acquisition of spatial language (“left of X”) correlated with subsequent success on a search task 1-2 years earlier than typically observed.

Evidence that language might support spatial representation

22 4-year-oldsSUBJECTS

SESSION ONE

Language training in Body Parts GameLanguage training on Object-Body Game

SESSION TWO

Body Parts Post TestObject-Body Post TestSearch Task (up to 8 trials)

Acquisition of left and right language on reorientation performance

Body Parts Game

Objects Game

Kick your right leg.

Raise your left arm.

Wiggle the fingers on your left hand.

Show me the one on your right.

Point to the toy on your left.

Give me the one to the right of you.

Learners: Performed at levels above chance

in both tasks in Session 2 post-tests.

Non-learners: Performed at levels below chance

in one or both tasks in Session 2 post-tests.

SESSION 2 LANGUAGE LEARNING

chance

Non -

Learners (n=11)

Learners

(n=8)

00102030

60708090

100

Body Parts Objects

4050

% c

orr

ect

49

Controls

32

75

Learners

18

0

7

41

Non-learners

39

9

11 7

12

1. Creatures with little or no spatial language (like rats, or children before the age of 5) fail to search correctly in a disorientation task.

The age at which children acquire adult-like search patterns correlates with the age at which they can produce the phrases “left of X” and “right of X.”

2. Only verbal shadowing impairs search in human adults.

3. Acquisition of spatial language (“left of X”) correlated with subsequent success on a search task 1-2 years earlier than typically observed.

Evidence that language might support spatial representation

How does language create thought?Ways in which language could create a

thought:create link between innate modulesfacilitate an analogy or an insightenable greater cognitive complexity

How does language create thought? The linguistic combination hypothesis.

Geometry system Object system

Left vs. right.

(Can do “left of a short wall”)

Red wall.

(Tracks objects & features)

Spelke, 2003.

Notice: No way to represent “left of the red wall.”

How does language create thought? The linguistic combination hypothesisGeometry system Object system

Left vs. right.

(Can do “left of a short wall”)

Red wall.

(Tracks objects & features)

“left of the red wall”

Spelke, 2003.

Revisiting the evidence

1. Creatures with little or no spatial language (like rats, or children before the age of 5) fail to search correctly in a disorientation task.

The age at which children acquire adult-like search patterns correlates with the age at which they can produce the phrases “left of X” and “right of X.”

2. Only verbal shadowing impairs search in human adults.

3. Acquisition of spatial language (“left of X”) correlated with subsequent success on a search task 1-2 years earlier than typically observed.

Evidence that language might support spatial representation

Some researchers’ opposition:1.Is there a geometric module? If there is, it is “leaky” and not encapsulated.2.Acquisition of “left of X” or “right of X” unnecessary for the integration of geometric and object information.

Creatures with no such spatial languages do make use of landmark/object features such as color to reorient.

Newcombe & Ratliffe (in press).Leaky module, language not necessary:Animals make use of wall color

Success w/o languageForaging vs. Escape tasks

Pre-linguistic children make use of wall colorRoom size: small fail, large succeed

Alternative: Adaptive Combination of Information View

Newcombe, N.S. & Ratliff, K.R. (2007) Explaining the development of spatial reorientation: Modularity-plus-language versus the emergence of adaptive combination. In J. Plumert & J. Spencer (Eds.), The emerging spatial mind (pp. 53-76). Oxford University Press.

Shusterman & Spelke (2005)Shusterman, Lee, & Spelke (unpublished data)

Language acquisition is not necessaryLearning expressions “left of X” or “right of

X” is not necessary for incorporating colored wall.

Subjects: 4-year-olds, 16 per conditionTask: Reorientation, 2 blocks

Influence of Language CuesShusterman, Lee, & Spelke (unpublished data)

Block 1 (baseline: 4

trials)

Block 2(test: 4 trials)

Control “I am hiding it over here”

No Cue No Cue

Salient“Look at the pretty red wall!”

No Cue Cue

Spatial“I’m hiding it at the red wall.”

No Cue Cue

Relevant“The red wall can help you get the sticker.”

No Cue Cue

Block 1

Control

Block 2

20

45

17

170.25

33

45

11

110.12

Influence of Language CuesShusterman, Lee, & Spelke (unpublished data)

20

45

17

17Control

0.25

“I’m hiding it over here.”

Spatial

8

77

9

6

0.69“I’m hiding it at the red wall.”

27

55

11

8Salient

0.28

“Look at the pretty red wall!”

“The red wall can help you get the sticker.”14

72

11

3Relevant

0.58

Influence of Language CuesShusterman, Lee, & Spelke (unpublished data)

Influence of Language CuesShusterman, Lee, & Spelke (unpublished data)

Language acquisition is not necessaryLearning expressions “left of X” or “right of X”

is not necessary for incorporating colored wall.

Language is usefulLanguage may be a means to “unlock” latent

knowledge.Colored wall needs to be construed as relevant

to search task. Language can be used to point this out to children who do not make use of such cues.

Shusterman & Spelke’s response? “[H]uman and animal minds are endowed with domain-general,

central systems that orchestrate the information delivered by core knowledge systems. One such system, associative learning, is common to human adults, infants, and nonhuman animals; it allows organisms to adapt their behavior to long-term regularities in the environment. A second system, however, is unique to human children and adults: the language faculty and the specific natural languages whose acquisition the language faculty supports. The latter system provides a medium that human children and adults use to combine information rapidly and flexibly, both within and across core domains.”

“We suspect that the majority of children end up learning via some version of linguistic combination process, but there may be different paths to the same end. We would not be surprised if the occasional child found an alternative way to solve the reorientation game, as did trained animals and children in our first experiment [i.e., cued children].”

Unitary vs. Two step: “learning a particular linguistic structure (left of X) enables children to construct a unitary representation of a concept like left of the red wall.”

1. Creatures with little or no spatial language (like rats, or children before the age of 5) fail to search correctly in a disorientation task.

The age at which children acquire adult-like search patterns correlates with the age at which they can produce the phrases “left of X” and “right of X.”

2. Only verbal shadowing impairs search in human adults.

3. Acquisition of spatial language (“left of X”) correlated with subsequent success on a search task 1-2 years earlier than typically observed.

Evidence that language might support spatial representation

Newcombe & Ratliff (2008) in Cognitive Psychology.1.Dual task impairs adults in problem-solving.

It is not just verbal shadowing. Choice of other non-verbal concurrent task impairs search in human adults.

2.Again, language is unnecessary for integration of geometric and object information.

Instruction given prior to verbal shadowing could influence incorporation of wall color. Adaptive Combination of Information View better explains this than Language Overcoming Encapsulation View.

Newcombe & Ratcliff used Brooks (1968)’s spatial interference task

TASK: Given a letter (such as below), starting from the asterisked vertex and going with the direction of the arrow, indicate if the vertex is the most “top/bottom”.

YES, YES, YES, NO, NO, NO, NO, NO, NO, YES

Performance: No secondary task blue wall > No secondary task white room.No secondary task white room = Spatial visualizing task blue wall.

Shusterman & Spelke’s response?

Participants: • Nicaraguan Sign

Language users in Managua, exposed to NSL community before age of 6

• 8 1st cohort (entered before 1986) and

• 9 2nd cohort (entered after 1986)

Does spatial language guide spatial representation? Evidence from NSL.

Pyers, J., Shusterman, A., Senghas, A., Emmorey, K., & Spelke, E. (2007). SRCD.

About NSL: • NSL is a young

language (~30 yrs old).

• Children introduce more linguistic complexity than adults can acquire.

• Older signers are inconsistent in marking left-right relations.

Digression: More data

0

2

4

6

8

0 1 2 3 4 5 6 7 8 9 10

RT in seconds

Num

ber

corr

ect

out

of 8

r=-0.77, p<.02

10

68

10

12Cohort 1

2

95

3

0Cohort 2

0

1

2

3

4

5

6

7

8

COHORT 1 COHORT 2

Num

ber

corr

ect

out

of

8

SEARCH ACCURACY

LANGUAGE EFFECT: Accuracy significantly correlated with consistency in marking left-right relationships (r=0.67, p=.02).

Digression

Role of linguistic label for adults?In the absence of stable frame-of-reference

language, representation is noisy and inefficient.

Language serve to enhance memory by providing a verbal label.

Digression

1. Creatures with little or no spatial language (like rats, or children before the age of 5) fail to search correctly in a disorientation task.

The age at which children acquire adult-like search patterns correlates with the age at which they can produce the phrases “left of X” and “right of X.”

2. Only verbal shadowing impairs search in human adults.

3. Acquisition of spatial language (“left of X”) correlated with subsequent success on a search task 1-2 years earlier than typically observed.

Evidence that language might support spatial representation

Re #4: “These findings provide the strongest evidence to date that the acquisition of spatial language closely mirrors the development of reorientation abilities within an individual child.”

Importance of Shusterman & Spelke’s language training data:To the Combinatorial View of LanguageAcquiring spatial language (that does not

specify anything specific with regards to features of objects s.a. wall color) in one fell swoop allows children to integrate object information with geometric information.

… but evidence potentially problematic… (STILL CORRELATIONAL)

49

Controls

32

75

Learners

18

0

7

41

Non-learners

39

9

11 7

12

54

Learners & Non-Learners

30

6

9 “[P]erhaps these children [the learners] were simply better problem-solvers, and therefore succeeded at the language games and the reorientation game independently.

Where does this leave us?Inconclusive: Acquisition of spatial language (“left of X”) is responsible

for subsequent integration of geometric and object information in search task.

Conclusive: Language is not necessary. Language is beneficial.1. Explicitness of instruction to highlight the incorporation of

landmark/wall color improves search (Shusterman, Lee, & Spelke, in prep.; Shusterman & Spelke, 2005)

2. Removal of use of language via shadowing could impair search (Hermer, Spelke, & Katsnelson, 1999)

3. Impoverished linguistic means/lack of linguistic expression impairs search (Pyers et al. 2007)

Where does this leave us?DISCUSSION QUESTIONS

From these studies, what can we say about role of language?

What other studies do you want to see?

Role of language in conceptual developmentCombinatorial Nature of Language

Case Study #2: Number

數字san1 si4 wu3 liu4 qi1 三 四 五 六 七

Zhe4li3 you3 _____ ge pin2guo3.

san1 si4 wu3liu4 qi1

Stages of Children’s Number Word Comprehension

“Will you give me _______ apple/s?”

“one” “two” “three” “four” “five” “six”

Pos

t-in

duct

ion

counters

Pre

-In

du

ctio

n

Number of objects requested by experimenter

1-knowers > > > > >

2-knowers > > > >

3-knowers > > >

1. Memorize the count list. 2. Become 1-knower (learn “one”),

then 2-knower, then 3-knower.3. Learn how count list represents

integers.

27 37 38 42

1-knower 2-knower 3-knower

18 months

24

Memorize listCountersCP-knower

PuzzlesWhy does it take so ?*!@ long for children to

learn the meaning of number words?

In contrast, why did it take you so little time to learn the meanings of: san1 si4 wu3 liu4 qi1?

IS THIS A CASE OF LANGUAGE CONSTRUCTING NEW CONCEPTS?

Core KnowledgeAt the foundations of human knowledge is a set of

core systems for representing significant aspects of the environment:

Objects and their motions Agents and their goal-directed actions Places and their geometric relations Sets and their approximate numerical relations

Core systems have signature limits: domain-specific, task-specific, encapsulated.

Core systems are shared by other animals, persist in adults, and show little variation by culture, language, or sex.

Content of this slide is from Liz Spelke.

Combinatorial Nature of LanguageSpelke (2003)Two core systems combined via languageObject File SystemAnalog Magnitude System

Object File System(A System of Core Knowledge)

Wynn (1992): 1+1 = 1

Object File System(A System of Core Knowledge)

Wynn (1992): 2-1 = 1

Object File System(A System of Core Knowledge)Feigenson, Carey, & Hauser (2002), etc.

Infants watched as crackers are placed into 2 buckets. Then infant is given a choice of one of two buckets.

Object File System(A System of Core Knowledge)Feigenson, Carey, & Hauser (2002), etc.

Infants watched as crackers are placed into 2 buckets. Then infant is given a choice of one of two buckets.

2 vs. 42 vs. 4

3 vs. 63 vs. 6

2 vs. 32 vs. 31 vs. 41 vs. 41 vs. 31 vs. 31 vs. 21 vs. 2

Success?Success?ConditionCondition

2 vs. 42 vs. 4

3 vs. 63 vs. 6

2 vs. 32 vs. 31 vs. 41 vs. 41 vs. 31 vs. 31 vs. 21 vs. 2

Success?Success?ConditionCondition

SET SIZE LIMIT:3!!!

Object File System(A System of Core Knowledge)Feigenson & Carey (2003), etc.

Infants watched as experimenter placed ball(s) into a box. Then infant is allowed to search for the balls. Infant’s search

time is measured

Infant retrieves 1 ball.

1-in-1-out (Box Expected Empty)Measure infant’s searching for more balls.

Infant retrieves 1 ball.

4-in-1-out (Box Expected Non-empty)Measure infant’s searching for more balls.

1 vs. 4 condition depicted:

Experimenter secretly hides the other ball.

step 1 step 2 step 3

Object File System(A System of Core Knowledge)

Set size limit: infants (3) adults (4-5)

Linked to object tracking ability: http://ruccs.rutgers.edu/faculty/pylyshyn/DemoPage.html

animals same.

Applies for not just objects, but also abstract individuals (e.g., events, sounds)

BOTTOM LINE: represent EXACT small numbers.(e.g., Exactly 3 objects)

Analog Magnitude System(A System of Core Knowledge) Approximate number system -- Cannot represent

EXACT numerosities!

Analog Magnitude System(A System of Core Knowledge) Approximate number system -- Cannot represent

EXACT numerosities!

System is usually engaged only for large numbers and not small numbers. Evidence from failures of 4+ objects in earlier

mentioned studies

When comparing two numbers, the comparison depends on the ratio of the numbers (subjected to Weber fraction).

Abstract representation Cross-modal comparisons

Shared with other animals

Analog Magnitude System(A System of Core Knowledge)

Xu & Spelke (2006)Depicting 8 vs. 16 condition

Habituationphase

or

Etc. Etc.

Test Phase

and

Analog Magnitude System(A System of Core Knowledge)Xu & Spelke (2006),

etc.

1. Success, and Ratio 6-months – 1:2 ratio

(8 vs. 16, 16 vs. 32) 9-months – 2:3 ratio

(8 vs. 12; 16 vs. 24) Adults – 7:9 ratio

2. Modality Invariance Dots, jumps, beeps,

syllables3. Addition &

Subtraction

Habituationphase

or

Etc. Etc.

Test Phase

and

Content of this slide is from Liz Spelke.

Analog Magnitude System(A System of Core Knowledge)

Limits of the core number systemsNeither system sufficient for representing

natural numbers (positive integers)Object file:

Set Limit ObjectA, ObjectB, ObjectC (not necessarily number

representation)

Analog magnitude: Lack precision & doesn’t represent indiv. (Can’t

represent diff. betw 2 & 3 is the same as diff. betw 7 & 8)

Content of this slide is from Liz Spelke.