Transcript On the Path to Verb Learning

Foundations of Verb Learning:
Labels Promote Action Category
Formation
Shannon M. Pruden & Kathy Hirsh-Pasek
Temple University
Early Verb Learning:
The Research
 Verbs are difficult to learn (Gentner,
1982; Gillette, et al., 1999; Imai et al.,
2003; Meyer et al., 2003).
 Paradox: Verbs appear in children’s
earliest vocabularies (Choi, 1998; Choi
& Bowerman, 1991; Fenson, et al.,
1994; Nelson, 1989; Tardif, 1996).
 Why are verbs difficult
to learn?
Prerequisites for Learning Verbs
1.
The conceptualization
of actions and events.
1.
Pay attention to
actions.
2.
The mapping of words
to these actions and
events.
2.
Form categories of
these actions.
3.
Map words to these
actions.
Gentner & Boroditsky, 2001
Golinkoff et al., 2002
Why are Verbs Difficult to
Learn?
A mapping problem?
OR
Lack of conceptual
knowledge?
Most of what has been done
on verbs…
 Early production of relational terms

Choi & Bowerman, 1991
 Choi & Gopnik, 1995
 Tardif, 1996
 Mapping relational terms onto actions and events
Choi, et al., 1999
Maguire, et al., 2003
 Naigles, 1996


Little research addresses the question
of whether infants have the conceptual
knowledge needed to learn verbs.
Prerequisites for Learning Verbs
1.
The conceptualization
of actions and events.
1.
Pay attention to
actions.
2.
The mapping of words
to these actions and
events.
2.
Form categories of
these actions.
3.
Map words to these
actions.
Gentner & Boroditsky, 2001
Golinkoff et al., 2002
Conceptual Prerequisites in Place?
Some speculate that conceptual prerequisites are in
place at an early age.
“relations…are, I suspect, perceived quite early…it
is not perceiving relations but packaging and
lexicalizing them that is difficult” (Gentner & Boroditsky, 2001,
p.326)
“vocabulary acquisition in the real case may reduce
mainly to a mapping problem” (Snedeker & Gleitman, 2004, p.
280)
“the young child’s conceptual
repertoire may be rich and varied
enough from the start…” (Snedeker &
Gleitman, 2004, p. 261).
Semantic Components that
Relational Terms Encode
Spatial Expressions
Motion Verbs
 Containment
 Path
 Support
 Manner
 Degree of Fit
 Result
Languages package these
components in different ways
Slobin, 2001; Talmy, 1985
Semantic Components that
Relational Terms Encode
Spatial Expressions
Motion Verbs
 Containment
 Path
 Support
 Manner
 Degree of Fit
 Result
Path and Manner in Motion Verbs
 Focus on path and manner:
(1) Universally codified in languages across world.
Jackendoff, 1983; Langacker, 1987; Talmy, 1985
(2) They are treated differently across languages.
Slobin, 2001; Talmy, 1985
 English - Manner encoded in verb; path encoded in
preposition.
 Spanish - Path encoded in verb; manner encoded in
adverb (optionally).
(3) Path may be conceptual primitive needed for
learning motion verbs.
Mandler, 2004
Early Event Perception
Are infants able to” decompose
scenes into constituent parts relevant to
linguistic expressions in language?”
(Clark, 2003, p. 168)
Discriminating Path and Manner
 14-month-olds discriminate path and manner.
 Low vocab. infants:more attention to changes in path
 High vocab. infants: more attention to changes in
manner
 7-month-olds discriminate path and manner
Pulverman et al. (2003; 2004)
 10-month-olds discriminate path and manner
 More naturalistic events with humans.
Casasola, Hohenstein, & Naigles (2003)
Categorization of Actions
Can infants form categories of actions?

“words…refer to categories of objects and
events, or properties of these things.”
Oakes & Rakison (2003)

Therefore, motion verbs label categories of
actions and events rather than single
events.
For example, “running”
· “Running” is
considered
the same
action
whether
performed by
Carl Lewis or
Grandma.
Finding the Invariant Path and
Manner in Motion Events
 Can infants abstract the invariant action within
a motion event?
Pruden, et al. (2004)
 Infants (7 - 15 months) familiarized to events from
the same category.
 Same path across multiple exemplars of manner
(Path Study)
 Same manner across multiple exemplars of path
(Manner Study)
Path Study: Familiarization Trials
 Four familiarization trials
 Importantly - no linguistic stimuli accompanied
events
Bend Around
Twist Around
Spin Around
 Vary manner across same path
 Example, “Around”
Toe Touch Around
Path Study: Test Trials
“Flap Around”
“Flap Past”
Novel Manner, Familiar Path
In-category event
Novel Manner, Novel Path
Out-of-category event
Results
Novelty Preference Score (%)
Can infants find the invariant action?
0 .6
0 .5 5
7- to 9-months
0 .5
10- to 12-months
13- to 15-months
0 .4 5
0 .4
Path
Manner
Summary: Finding the Invariant Action
 10-month-olds were able to find invariant path
across varying manners.
 13-month-olds were able to find invariant manner
across varying paths.
 Path to manner developmental progression
 There is a group of infant in both studies who do
not find the invariant action.
· 7- to 9-month-olds
Conceptual prerequisites to learn
verbs may be in place early in life
The next step in learning verbs…
Adding language to non-verbal scene
What effect might language
have on the processing of
these non-linguistic scenes?
Two Possibilities
 Adding language increases complexity of the
task and may hinder category formation
(Stager & Werker, 1997)
 Adding language facilitates category
formation
(Balaban & Waxman, 1997; Waxman & Markow, 1995)
To our knowledge, only one study has added
language to an event categorization task…
Spatial Categorization and
Labels
Quick Time™ and a
TIFF ( LZW) dec ompr ess or
ar e needed to s ee thi s pi cture.
Loose-fit
 10- to 18-month-olds: no
abstract spatial category of
“on” (Casasola & Cohen, 2002)
support
Quic kTime™ and a
TIFF (LZW) decompr ess or
ar e needed to see thi s picture.
“ON”
Quick Time™ an d a
TIFF ( LZW) d ec ompre sso r
ar e n eed ed to s ee this pic ture .
Tight-fit
Quic kTi me™ and a
TIFF (LZW) decompr es sor
ar e needed to see this picture.
support
 Can a linguistic label
facilitate infants’ spatial
categorization of support
relations?(Casasola, 2005)
 Linguistic label helped
infants form an abstract
category of “on”
Expanding on Casasola…
1. Exploring categorization of events
based on path and manner.
2. Testing pre-verbal infants
The Present Studies
Our Paradigm
 Preferential Looking
Paradigm: forced-choice
split-screen

(Hirsh-Pasek & Golinkoff, 1996)
 Dependent Variable:
Looking Time
Novel, easily manipulated
and controlled stimuli
Stimuli Across Studies






6 Paths
Over
Under
Past
Around
Behind
In Front






6 Manners
Flap
Spin
Twist
Side Bend
Bend Forward
Toe-Touch
General Method
 Introduction
 Salience Trials
 Four Familiarization Trials
 Test Trials
 All trials are 12 s
Introduction Trial
· Purpose: To ensure infants look to both sides
Salience Trial
 Purpose
 To show that infants do not have any a
priori preferences for test events.
 What they see
 Two clips simultaneously.
 Same clips they see at test.
 Assumption
 Infants will not have a preference for
either clip.
Familiarization Trials
 Four exemplars of the category are
shown.
 Trials are separated by attention-getter:
 Picture of a baby
 Accompanied by music
QuickTime™ and a
Radius SoftDV ™ - NTSC decompressor
are needed to see this picture.
Test Trials
 Test trials
 Two clips shown simultaneously
 In-category event (familiar exemplar)
 Out-of-category event (novel exemplar)
 Predictions
 Infants who can find the invariant action
will show a preference for one of these
clips.
Predictions
 No salience preference for test clips will be
found
 Infants will show increased attention during
familiarization
 Labels heighten attention (ala. Baldwin &
Markman, 1989 with objects)
 Labeling will help infants abstract the
invariant path or manner
Study 1: Do Labels Help Infants
Abstract the Invariant Path?
Participants
 24 7- to 9-month-olds
 Mono-lingual English-speaking homes.
 All infants full-term births.
 Equal numbers of males and females.
Familiarization Trials
 Four familiarization trials
 Same stimuli/design as Pruden et al. (2004)
 Vary manner across same path
 Example, “Under”
Spin under
Toe touch under
Side bend under
Flap under
 During each familiarization trial, they
hear novel verb “javing” 4 times.
Salience/Test Trials
Starry “Twist Under”
Starry “Twist Over”
Novel Manner, Familiar Path
In-category event
Novel Manner, Novel Path
Out-of-category event
8-month-old infant
Results: Salience Preference
Novelty Preference
Score (%)
Is there a salience preference?
0.6
0.55
P a th
Ma nne r
0.5
0.45
Study
Results: Enhanced Attention
Average looking time during familiarization (sec)
DOES A LABEL INCREASE ATTENTION?
12
10
8
LABEL
6
NO LABEL
4
2
0
PATH STUDY
MANNER STUDY
Results: Finding the Invariant
Path
DOES A LABEL HELP CHILDREN ABSTRACT THE INVARIANT
ACTION?
Novelty Preference Score (%)
0 .6 5
0 .6
LA BE L
N O LA BE L
0 .5 5
0 .5
0 .4 5
P A T H ST U DY
M A N N E R ST U DY
Summary: Path Study
Our Predictions
 No Salience Preference
Our Results
 No Salience Preference
 Increased attention
during familiarization
No increase in attention
during familiarization
 Label facilitates finding
the invariant path
 Label facilitates finding
the invariant path
Study 2: Do Labels Help Infants
Abstract the Invariant Manner?
Participants
 24 7- to 9-month-olds
 Mono-lingual English-speaking homes.
 All infants full-term births.
 Equal numbers of males and females.
Familiarization Trials
 Four familiarization trials
 Same stimuli/design as Pruden et al. (2004)
 Vary path across same manner
 Example, “Twist”
Twist around
Twist in front
Twist over
Twist Past
 During each familiarization trial, they hear
novel verb “javing” 4 times.
Salience/Test Trials
“Toe Touch Under”
“Twist Under”
Novel Manner, Novel Path
Out-of-category event
Familiar Manner, Novel Path
In-category event
Results: Salience Preference
Novelty Preference
Score (%)
Is there a salience preference?
0.6
0.55
P a th
Ma nne r
0.5
0.45
Study
Results: Enhanced Attention
Average looking time during familiarization (sec)
DOES A LABEL INCREASE ATTENTION?
12
10
8
LABEL
6
NO LABEL
4
2
0
PATH STUDY
MANNER STUDY
Results: Finding the Invariant
Manner
DOES A LABEL HELP CHILDREN ABSTRACT THE INVARIANT
ACTION?
Novelty Preference Score (%)
0 .6 5
0 .6
LA BE L
N O LA BE L
0 .5 5
0 .5
0 .4 5
P A T H ST U DY
M A N N E R ST U DY
Summary: Manner Study
Our Predictions
 No Salience Preference
Our Results
 No Salience Preference
 Increased attention
during familiarization
 Increased attention
during familiarization
 Label facilitates finding
the invariant manner
Label provides no
facilitative effect in
finding invariant manner
What does all of this mean?
 These studies show
Infants can abstract the invariant actions
that are encoded in relational terms, like
motion verbs
Labels help infants find invariant actions
 Developmental Progression
Path first, then manner
What are labels doing?
 Auditory stimulation enhances attention to objects
(Baldwin & Markman, 1989; Roberts & Jacob, 1991)
 Labels invite categorization
(Balaban & Waxman, 1997; Waxman & Markow, 1995)
 Labels highlight similarities and promote
comparison
(Gentner & Namy, 1999; Lowenstein & Gentner, 2005)
 Is there anything special about a label?
Tone study
Complex musical melodies study
Future Studies: Role of
Comparison
 Comparison helpful in
categorization, learning new
adjectives and verbs.
 (Childers, in press; Gentner & Namy, 2000;
Oakes & Ribar, 2004; Waxman & Klibanoff,
2000)
 Active comparison of actions promote abstraction of invariant
actions?
 Current Studies
 Sequential familiarization
 Next Study
 Simultaneous familiarization
Future Studies - Trends
 Path then manner developmental trend
 Would 10- to 12-month-olds use a label to find
the invariant manner?
 Would we see cross-linguistic differences in our
studies?
 14- to 17-month-old Spanish-speaking infants show
same developmental pattern as English-speaking infants
in discrimination task.
 What about abstracting the invariant action?
Future Studies: Individual
Differences
 Individual differences
in performance
Path Study with Label: 7- to 9-month-olds
Novelty Preference Score (%)
0.9
0.8
 Some children did not
show a novelty
preference
0.7
0.6
0.5
0.4
0.3
0.2
 Do individual
differences predict
later language
development?
Acknowledgements…
Dr. Roberta
Golinkoff
Meredith Jones
Natalie Sheridan
Thanks to all the parents
and children who
participated in these studies
at the Temple Infant Lab.
Gwen Albertson
QUESTIONS?
For information contact:
Shannon Pruden
[email protected]
Visit my website at:
astro.temple.edu/~spruden