Transcript CHILD LANGUAGE

Social and cognitive
development
Language evolution
• special adaptations unique to human
language? (Pinker & Bloom 1990)
• more general social and cognitive abilities
allowing cultural learning? (Tomasello 2008,
Csibra & Gergely 2009)
Evidence for cultural learning
• fixation on faces (Haith et al 1977, Farroni et al
2005)
– increases over the first few weeks of life
– triggered by child-directed speech
– eye contact in response to caregiver’s vocalisations
(Crown et al 2002)
– neonates discriminate faces with eyes oriented
towards them vs. with eyes looking away from them
– the still-face effect:
http://www.youtube.com/watch?v=apzXGEbZht0
(Tronick et al, Nagy 2008)
• imitating facial expressions (Meltzoff & Moore
1977)
– neonates can imitate facial expressions
– and sounds accompanying facial expressions (/m/
versus /a/) (Chen et al 2004)
• turn-taking behaviour before speech
– vocalisations in response to non-human puppet
vocalisations, taking turns
Moving on
• from dyadic interaction to triadic interaction
• dyadic: child + caregiver
• triadic: child + caregiver + the environment
– referential communication: communications with
others about the world
Prerequisites to referential communication
• gaze following: direction of caregivers gaze as
a cue
– 10 month-olds are more likely to follow head
movements if the eyes are open (Brooks &
Meltzoff 2005)
– they pay more attention to an object
• if another person is looking at them (Striano & Rochat
2000) (noticing communicative intentions)
• if another person is also looking at the object (Striano
et al 2006) (joint attention)
Joint attention
• infants’ ability to establish joint attention (by
following adult gaze) predicts vocabulary size
at 14-18 months (Carpenter et al 1998, Brooks
& Meltzoff 2005)
• for children with developmental disorders
(Down Syndrome, ASD), joint engagement is a
strong predictor of language development
(Adamson et al 2009)
• interpreting pointing
– understanding intentions
– using shared knowledge
– by 14 months infants can interpret pointing (Liebal
et al 2009) (but this is later than the first words!!!)
• using pointing to reference objects
– at 8-11 months (Carpendale & Carpendale 2010)
– they point to draw attention to something, to seek
help, to give directions, to request objects
(Tomasello et al 2007)
Advanced imitation
• 12 month-olds copy even nonsense actions if
that’s the intention of the adult (Nielsen 2006)
• imitation of a nonsense action is more likely if
the outcome is known in advance (Southgate
et al 2009)
social learning, following convention
language is also arbitrary, a matter of convention
CONCEPTS, REPRESENTATION
A representation of the world in memory
• Deferred imitation task
– adult performs an action, child is allowed to imitate the action
after a period of delay
– if they can do so, they must some kind of representation of the
action
– 9 month-olds can do the task (with 24 hour delay) (Meltzoff
1988)
• Generalised imitation
– adult performs an action using an object (e.g. gives a drink to an
animal)
– infant is asked to imitate the action using another object
(another animal vs. a toy car)
– at the 14 months infants appropriately generalise (Mandler &
McDonough 1998)
• Playing with toys of different categories
(animals versus vehicles)
– 7 month-olds distinguish the categories (look at a
toy from a new category for longer) (Mandler
2004)
• Symbolic play
– at 18 months they can pretend that one object is
another
The origin of conceptual categories
• statistical learning from perceptual
information? (Mareschal & Quinn 2001)
• re-description of perceptual information?
(Karmiloff-Smith 1992)
• the role of language?
WORD LEARNING
Fenson et al 1994
Fenson et al 1994
• 6-8 months: learn pairings of novel words with
novel objects if the object is moved around as
it is being named (Gogate et al 2010)
• at 14 months they can learn labels without
special attention seeking movement (Werker
et al 1998)
• toddlers associate novel words with novel
objects or actions (Akhtar & Tomasello 1996,
Tomasello et al 1996, Tomasello 2003)
Errors
• under-extension
– word used for a specific objects rather than for
the category (Bloom 1973)
• over-extension
– word used for broader category (Clark 1982)
Vocabulary spurt
• about one new word a day between 18 and 24
months
• about 4 new words a day between 2 and 6
years
• how do they do it?
Mutual Exclusivity and the Principle of
Contrast (Markman 1989, Clark 1993)
• Children assume that a new word is the name
of a new object or action
– built-in constraint (Markman)
– or general-purpose learning strategy (Regier 2003)
• dogs appear to use this strategy (Kaminski et al 2004)
• but children can learn multiple labels!!!
• Ramscar (2011): distributed learning (ever
finer grade differentiation)
Whole object bias (Markman 1989, Hollich
et al 2007)
• we see the world in terms of whole objects
(Spelke 194)
– a connected and bounded region of solid matter
• infants expect novel words to refer to a whole
object category (Waxman & Markow 1995)
• it’s not unique to humans (Winters et al 2008)
The shape bias: dax (Landau et al 1988)
The noun bias
• noun meanings are easier to extend than verb
meanings (Imai et al 2005)
• language (and culture?) dependent
– English speaking children tend to pick objects rather
than actions as referents of novel words
– Mandarin Chinese learners do not show this bias
silent videos of Chinese and American mothers
playing with their toddlers (Snedeker et al 2003)
• Chinese and American adult viewers had to guess what the
mothers were saying
• viewers correctly identified more nouns than verbs in the
American videos, but equal numbers of nouns and verbs in
the Chinese videos
Bayesian model of word learning (Xu &
Tenenbaum 2007, Perfors et al 2011)
• task: learn 24 object names (from 1 vs. 3
examples)
• no word learning biases were built in
• 45 objects were organised into a hierarchy of
clusters based on their similarity ratings (based
on anything: shape, size, colour, animacy, etc)
• names were taught (eg dalmatian for OBJECT8
and OBJECT13)
• results: the model generalised the names the way
3-4 year-olds did.
From syntax to word meaning
• children use syntax to distinguish count nouns
(a dax) from mass nouns (some dax) (Landau
et al 1992)
• argument structure can help identify verb
meaning (The duck is gorping the bunny. vs.
The duck and the bunny are gorping.) (Naigles
1990)