WEEK 7- WORD LEARNING Flashcards
Why are words unique to some extent?
they are symbols
how are words different from facts?
they refer or stand for information rather than being the information - eg knowing a word isnt knowing what a dog is it is know the relation between the word and the dog - words convery information
when is there a really steep increase in word learning in children
18 months
what is the most contemporary method of knowing how much vocab children have?
track where infants are lookingo infants more likely to udnerstand words before they say them- machine automated eye tracker- present them with two pictures on the screen- can see how much time was spent looking at each picture- using prerecorded sentence asking them to look at each item
what problem did Quine 1960 come up with?
the mapping problem- when a person points out something in a scene and says ‘gavagai’ introduces a new name- this name could refer to an infinite number of aspects of the rabbit, or just to part of the the rabbit - like the ear or the way the light goes through the ears or where the scene is taking place. but kids are remarkably good at solving this problem. what kind of information do they have at their disposal to solve this problem? as adults we dont think of this as being an ambiguous situation- however it has now been shown that we tend to want to map new words onto concrete objects- hence we guess it is the rabbit being talked about
what is one of the suggestions as to how we learn new words?
word learning is special
Markamn 1990 suggest children are innately constrained to what?
constrained to only consider some possible word meanings - the proposed constaints are: mutual exclusivity, whole-object constraint and taxonomic constraint- all of these allow fast, efficient learning or fast mapping
How did Markson and Bloom 1997 show fast- mapping is specific to words?
an experimenter introduced participant (3-4 yea rolds and adults) to a new object in one of three conditions. one condition was a word ‘koba’ one was a fact ‘my uncle gave me this’ one was a location ‘the sticker goes here’. both the children and the adults learnt the new word quickly. after one month they were asked to find the koba/ one from uncle/ one where the sticker goes. they showed identical performance in the fatc condition compared to the word condition. interestingly they were a bit worse in the location condition
what did Markson and Bloom conclude about fast mapping?
fast mapping is not specific to words. no evidence that words are learned differently from facts
how did Kaminski, call and fisher 2004 test if fast efficient word learning is specific to humans?
they tested dogs (Rico)- found that they were particularly good at learning names for things. they used 2 familiar items and used sirikid for the novel item the white bunny, he was able to pick out the novel object found he has an approximate word vocab of 200 words but mutual exclusivity and can learn the names of new toys by being told just once. but needed a real test make sure its not just owners giving hints about which object to pick. however this is only one dog and would need more to prove this was the case- also might be because dog have been bred to be social creatures
What did Waxman and Booth 2000 suggest about fast mapping
maybe fast mapping is not specific to words and word learning is special for other reasons- just because facts and words are similar in soe ways does not make them the same in all ways. they suggest its like saying cats and dogs arent different because they both have fur. words, because they are symbols can be freely extended to new instances of the same kind (ie taxonomic constraint) but facts cannot be
what paper did waxman and booth 2000 published about word learning
‘principles that are invoked int he acquisition of words but not facts’- an expeirmenter introduced 4 year old participants to a new object in one of two conditions- fact or word ie ‘my uncle gave it to me’ or ‘koba’ they would say ‘look at this one, it is so special to me and you know what?’ and then they would either say ‘it is called a koba’ or ‘my uncle gave it to me’- then ask if there was any other kobas or objects my uncle gave me- tested straight away and then after a week. results: all of the children fast - mapped the word/ facts- they chose the right object wen asked immediately or after a delay. children generalized words and facts differently. 100% of the time they extended words to those of the same kind, never extended to objects or a different kind. didnt do the same thing with facts- did it to also objects of a different kind AND the same kind- ie we are biased to map words to objects - words are special. they conclude words and facts are fast mapped but words and facts are not learned in the same way. we are biased to map words to objects (whole object constraint) and extend words but kind (the taxonomic constraint)
how did smith et al 2002 test where these biases come from?
hypothesised that word learning- biases are built up from experience with initial word- meaning associations. a ‘dumb attentional mechanism’ or DAM is born. they say whole object in taxonomic bias could be built up by experience. more dictated by shape- when they hear a new word they match it to an objects shape. tend to learn words for concrete objects first. they taught 17 month olds new words and then asked them to generalise and find another one of the same thing. or they were not trained to generalise. children that learned a lot of new connections picked out objects of the same shape. whatever they learned appeared to promote learning outside the lab, infants trained over shape bias added quite a large number more object names over the course of the experiment- suggesting this is a mechanism of word learning that is helpful for children and not innate ie word learning mechanisms are special but not innate
how did smith et al 2002 suggest children learn words
infants use correlations between words and the features of the referents to develop word learning biases- these come more specialised with further learning
what did waxman and booth say about the smith et al 2002 account
they said this is not a good account of how children figure out and extend the meaning of words- children dont generlize words based on looks (perceptual information). they generalise based on what somethign IS (to things of the same kind)- they call this ‘conceptual’ information in constrast to perceptual information. often things that look alike are of the same kind so it is a good cue but it is not diagnostic. texture also tends to be important. waxman and booth say word learning is special
what study did Waxman and Booth conduct to show children do not use perceptual cues alone?
3 year olds taught the novel word dax by an experimenter in two conditions- animate and artifact- animate had biological entities such as a mum and a dad - asked ‘is this another dax?’ with shape size and texture change- found that changing the shape had a big impact meaning it is very important same with the texture change but nobody cared if there was a size change. so went one step further and gave gave strong perceputal cues such as eyes- leads them to reject texture changes - smith and colleagues said it pushes children to generalize by shape and texture but booth and waxman asked whether conceptual information could override that? they found that adding eyes did not lead children to reject a texture change - so the way in which children are generliazing isnt affected by perceptual information in fact they are using deeper information (waxman and booth)
what did this lead to waxman and booth thinking about word biases?
they think these biases are innate and specific to words- bias is to map and extend based object kind
what was smith et al’s repsonses to the DAM aacount?
they said the context of naming pushes children’s attention to the relevant feature ie animacy cues push attention to texture in the same way verb syntax pushes attention to events. these are all features of learning context. also these are three year olds so this mapping bias cold be learned. said it is also not a good task - an object with eyes still doesn’t count as animacy
if these biases develop from experience, what did Smith and Yu (2008) say about how infants learn their first words?
infants rapidly learn word-referent mapping via cross-situational statistics- environment provides ample opportunities to learn but each individual event provides ambiguous information. across time the probability of seeing a ball when hearing ball is higher than seeing a dog (cross sectional word learning)
How did Smith and Yu test how children learn words cross situationally
12 and 14 month old infants see 6 new objects hear 6 new labels, both objects labelled on each trial, each event is ambiguous on its own but not across time/ occurances. the probability of of an object appearing with a particular object was higher. researchers wanted to know if they could work out which object went with which word using preferntial looking test- found children are pretty good at handling a bit of ambiguity but isnt clear if they can figure it out
how did medina et al argue that words can and cannot be learned by observation
medina et al 2011 said it sounds great but this paradigm is too simple- not fair presenting them with only two objects because the real world is a lot more messy - usually have to track a lot more objects in the environment
what study by medina et al showed even adults fail at learning new words observationally
adults shown muted videos of naturalistic interactions between children and caregivers. hear a beep where a noun occured and asked to guess what was said. when shown ambiguous clips they dont get better over time. the only times they learn is when they are shown an unambiguous clip first. the way they tested this was video the way children interacted with their caregivers. if adults cant track this information then children unlikely. you really need these unambiguous instances for learning to happen so potentially people would be biased to only learn when they were really sure that they knew what a word means
what did Yurovsky et al 2013 say was another solution for word learning
the child’s view is different. head mounted a camera- they are not on baord with general learning mechanisms to figure out what words mean. what they point out is that childs view of situation very different from third persons perspective. if mum is presenting word for her child there is a lot of stuff present she could be talking about. if child wears camera that has first person view- you can see there is less ambiguity- visually dominant objects standing out more- adults learn across situations when given children’s view. if you showed children third person view they didnt really improve over time whereas when shown first person perspective there was learning higher than chance- less ambiguity
what reason did Yu and Smith 2012 give for children having less ambiguity
embodies attention- pointed out childrens bodies are different from adults- have shorter arms ect- if you have short arms and you are holding an object of interest to you you cannot hold it that far away from your face- takes up a huge amount of space so object is visually dominant- taught children three different objects at a time. also objects more likely to be dominant when children are holding them- most likely to learn words in times of object dominance as they are less ambiguous- the ones that were learned were the same size as they ones not dominant
