LEARNING BUILDS

ON LEARNING
A critique by Joseph David Brady
ABOUT THE PAPER

Title: "Learning builds on learning: Infants’ use of native language

sound patterns to learn words"

Author: Katharine Graf Estes

Year: 2014
ABOUT THE AUTHOR
KATHARINE GRAF ESTES
• Ph.D. in Psychology
• Faculty at UC
• Member of the Society for Research in
Child Development and the
International Society for Infant Studies.
OTHER PAPERS BY THE SAME AUTHOR

Infants generalize Learning about sounds

representations of contributes to learning
statistically segmented about words: Effects of
words. prosody and phonotactics
on infant word learning.

Statistically coherent
labels facilitate
categorization in 8-month-
olds.
THE SIGN (SAUSSURE ET AL., 1916)

THE SIGNIFIER: The sound

sequence on the level of
expression (eg: /kæt/).
THE SIGNIFIED: The concept on
the level of meaning (eg: The
concept of a cat).
PROBLEM ONE: RANGE
PROBLEM
When a child learns a new signifier, how do they know how restrictive the
signified concept is?
POSSIBLE SOLUTIONS
• ‘‘____s are ____-shaped.’’ (Smith et al., 2002)
• The Shape Bias: a strategy whereby an infant extends the name of
an object to new objects of the same shape rather than other
characteristics such as color or texture. (Diesendruck et al., 2003)
PROBLEM TWO: SEGMENTATION
PROBLEM
How can a child know where one word starts and another begins?
POSSIBLE SOLUTIONS
• Adults identify words in fluent speech more rapidly when phonotactic
word boundary cues are present than when they are absent.
(McQueen, 1998)
• In an experiment, 9-month-olds were able to identify target words in
a supportive phonotactic context but not when the phonotactic
information was unavailable. (Mattys & Jusczyk, 2001)
PHONOTACTIC LEGALITY
ILLEGAL FEATURE CONCLUSION

Positional illegality: The phoneme The /ŋ/ is in the middle or at the

/ŋ/ is illegal in an onset (initial) end of a word.
position.

The cluster must be either side of a

Complete illegality: The boundary. (eg. 'Twelfth Night'.)
cluster /fθn/ is an illegal
combination.
IS SEGMENTATION IMPORTANT?
YES IT IS!

Toddlers who showed successful segmentation as infants have

larger vocabularies than those who didn't. Infant word
segmentation performance was also related to vocabulary and
syntax at 4 to 6 years of age. (Newman et al., 2006)
WHAT IS ESTE'S NICHE?
WE KNOW... WE DON’T KNOW...
• Phonotactic information is a • Precisely how infants use
valuable environmental cue phonotactic information.
for finding words.
• Whether phonotactic
• Infants can detect this cue. information help infants to
• Early segmentation ability is associate sounds with
meaning to acquire new
a fundamental skill for
words.
language acquisition
HOW WILL SHE OCCUPY THE NICHE?

1 - Establishing a set of object labels that 14-month-olds did

not readily learn in the absence of supporting information
about the sounds of the labels.

2 - Examining whether infants would learn those object labels

when they first had the opportunity to use phonotactic word
boundary cues to segment the target words.
THE HYPOTHESIS

"Infants would use prior knowledge to

support new learning. We predicted that
infants would use knowledge of native
language sound patterns to detect the new
words in fluent speech."
EXPERIMENT ONE: THE SETUP
PARTICIPANTS POSITIONING
• 10 girls and 8 boys. • Each infant sat on a parent’s
lap.
• 13.7-15.3 months old.
• To prevent bias, the parent
• Mean age: 14.3 months.
listened to music over
• All from English-speaking headphones and the
homes. experimenter controlled the
experiment remotely.
EXPERIMENT ONE: THE PAIRS
The two utterances /gæf/ and /tov/ were
chosen to be the signifiers used in the
experiment, and the objects shown on the
left of this slide were chosen to be the
signifieds.
The utterances were the ones used the
experiment carried out by Mattys &
Jusczyk in 2001.
EXPERIMENT ONE: THE PROCEDURE

First, the familiarization trial: A computerized image of a small rotating grey screen
appeared on the television while repetitions of the nonword /niːm/ played.

Second, the habituation trials: The label-object pairings were presented one at a time,
randomized by blocks.

Third, The test trials: During 'same' trials, the infant viewed the original label–object
pairings. During 'switch' trials, the original label–object pairings were switched.
EXPERIMENT ONE: THE RESULTS
• 10 infants displayed longer looking to 'switch' trials.
• 8 infants displayed longer looking to 'same' trials
• There was no reliable correlation between difference scores and
attention spans.
"In Experiment 1, 14-month-olds exhibited no evidence of learning
the label–object pairings for gaffe and tove."
EXPERIMENT TWO: THE SETUP
PARTICIPANTS POSITIONING
• 19 girls and 19 boys. • Each infant sat on a parent’s
lap.
• 13.7-15.3 months old.
• The infants were randomly
• Mean age: 14.3 months.
assigned to one of two
• All from English-speaking groups.
homes.
EXPERIMENT TWO: THE GROUPS
GROUP ONE: The army is trying a new GROUP TWO: The army is trying a new fang
bean gaffe hold next week. The old pine gaffe tine next week. The old tong gaffe
gaffe house tends to break too often. Of tends to break too often. Of course,
course, everyone knows the main gaffe hoist everyone knows the king gaffe tool is worn.
A wrung gaffe tops the list of new inventions.
is worn. A spun gaffe heads the list of new Still, some think that a strong gaffe tin is
inventions. Still, some think that a stone gaffe better. A hang gaffe tote is being used for
hod is better. A lean gaffe hall is being used the moment. A gruff tove knows most forest
for now. A brave tove trusts most forest animals. Not many people know that a calf
animals. Few people know that a live tove tove needs so much care. Some believe that
takes so much care. Some think that an eve a skiff tove nibbles old wheat stalks. The
tove twists old wheat stalks. The mauve tove roof tove nest is hidden in the jungle. Many
tree is hidden in the jungle. Spies are scientists are interested in buff tove naval
stories. On most sunny days, a leaf tove naps
interested in the dove tove territories. On quickly.
sunny days, a groove tove tires very quickly.
EXPERIMENT TWO: PHONOTACTIC PROBABILITY
GROUP ONE: GROUP TWO:
bean gaffe hold fang gaffe tine
/ng/ /fh/ /ŋg/ /ft/
The sequences /ng/ and /fh/have The sequences /ŋg/ and /fh/have
low probability of occurring within low probability of occurring
words and high probability of between words and high probability
occurring between words. of occurring within words.
GOOD CUES! BAD CUES!
EXPERIMENT TWO: THE PROCEDURE

First, infants were played one of the two speeches,

depending on the group to which they were assigned.

Next, having listening to their speech, infants

participated in the final task described in experiment one.
EXPERIMENT TWO: THE RESULTS
LOOK
HERE!
EXPERIMENT TWO: THE RESULTS

BLOCK 1 BLOCK 1 BLOCK 2 BLOCK 2

(SAME) (SWITCH) (SAME) (SWITCH)

No cues: 7.39 seconds 7.35 seconds 8.04 seconds 6.14 seconds

Good cues: 6.22 seconds 8.54 seconds 6.67 seconds 6.13 seconds

Bad cues: 6.86 seconds 6.53 seconds 5.52 seconds 6.40 seconds
WAS THE HYPOTHESIS CORRECT?
• "Infants would use prior knowledge to support new
learning. We predicted that infants would use
knowledge of native language sound patterns to
detect the new words in fluent speech." Yes!
DID SHE OCCUPY THE NICHE?
•1 - Establishing a set of object labels that 14-month-
olds did not readily learn in the absence of supporting
information about the sounds of the labels. Yes!
•2 - Examining whether infants would learn those object
labels when they first had the opportunity to use
phonotactic word boundary cues to segment the target
words. Yes!
CRITIQUE: HOW UNIQUE IS THIS RESEARCH?
"The present study investigates whether 9-month-olds make use of
phonotactic cues to segment words from fluent speech. Using the
Headturn Preference Procedure, we found that infants listened to
a CVC stimulus longer when the stimulus previously appeared in a
sentential context with good phonotactic cues than when it
appeared in one without such cues." (Mattys & Jusczyk, 2001)
CRITIQUE: WHAT'S HAPPENING HERE?

LOOK
HERE!
REFERENCES
Diesendruck G., & Bloom P. (2003). How specific is the shape bias? Child Dev. 74 168–17

Erikson, L., et al. (2014). Statistically coherent labels facilitate categorization in 8-month-olds. Journal of Memory and Language, 72, 49-58.

Estes, K. G. (2012). Infants generalize representations of statistically segmented words. Frontiers in Psychology: Language Science, 3.

Estes, K. G., & Bowen, S. (2013). Learning about sounds contributes to learning about words: Effects of prosody and phonotactics on infant word learning. Jo urnal of
Experimental Child Psychology, 114, 405-417

Gonzalez-Gomez, N., & Nazzi, T. (2013). Effects of prior phonotactic knowledge on infant word segmentation: The case of nonadjacent dependencies. Journal of Speech,
Language, and Hearing Research: JSLHR, 56, 840–849.

Mattys, S. L., & Jusczyk, P. W. (2001). Phonotactic cues for segmentation of fluent speech by infants. Cognition, 78, 91-121.

Mattys, S. L., et al. (1999). Phonotactic and prosodic effects on word segmentation in infants. Cognitive psychology, 38, 465–494.

McQueen, J. M. (1998) Segmentation of continuous speech using phonotactics. Journal of Memory and Language, 39, 21-46.

Newman, R., et al. (2006). Infants' early ability to segment the conversational speech signal predicts later language development: a retrospective analysis. Developmental
psychology, 42, 643-655.

Saussure, F., et al. (1916). Cours de linguistique générale. Lausanne; Paris: Payot.

Smith, L. B., et al. (2002). Object name learning provides on-the-job training for attention. Psychological Science, 13.