week 3 - 3. Learning mechanisms. Flashcards
what are first words often like?
- Simple to produce, often ‘babble-like’
- Relevant to the baby’s world
- Nouns
- Words that can stand alone grammatically
what are the top 5 most common first words in British infants?
daddy mommy no baa baa yes (Frank et al., 2016)
what are an infants first words often determined by?
- The sounds they produce in their babble- an infant who produces lots of bababa will have /b/ words, etc
- Words that are most frequent in their input- an infant who hears a lot about dogs is likely to acquire ‘dog’
- Words that are most salient in their input- words that stand out – nouns, onomatopoeia, isolated words – are often acquired earliest
how does babble, statistical learning and word segmentation support a Childs language acquisition?
Babble supports the development of a phonological inventory
Statistical learning supports infants to learn the linguistic systems of their ambient language
Word segmentation supports the extraction of learnable linguistic units
what did Bergelson and Swingley find about early lexical acquisition?
2012
• 33 American infants aged 0;6-0;9
• Looking-while-listening paradigm using eye-tracking
• Scene stimuli and paired-picture stimuli
- Fixations were measured for each image
- Longer fixations to the target suggests recognition
Test:
- Researchers wanted to find out when baby’s linked words with meaning.
- They used 0;6 as a control group, only to show a developmental change.
- They found that baby’s of 0;6 had the knowledge to link words and meaning.
- Infants require 1 word a week at first
- Once they have 50 words, they acquire 1-2 per day
- At age 2-6, 10+ words a day
- Acquire = learn + remember + retrieve at will + use easily & (relatively) accurately in different contexts
- Between 10,000 and 14,000 words by age 6
what is fast mapping
learning and retaining some knowledge with very little experience
◦ Useful for short term learning
◦ Poor retention
◦ Ability improves with age
◦ E.g. axiom – we might know the word exists, but perhaps not what it means or how to use it
what is slow mapping
developing full knowledge of words through experience ◦ Necessary for long-term knowledge ◦ Good (lifelong?) retention ◦ Might take years to achieve full knowledge ◦ E.g. morpheme – introduced in introductory module in 2017, developed knowledge over time
discuss the hypothesis: that infants begin with item learning
They acquire single words/phrases, slowly, one-by-one. They use the same handful of words over and over again Vocabulary growth happens very slowly
Item learning can take place through statistical learning; tracking words and items that regularly occur together
Later, acquisition becomes systematic They make assumptions about the world based on a set of learning mechanisms.
This supports speedy lexical acquisition, from an initial ‘vocabulary spurt’ and on into childhood.
discuss Quine’s problem
Whole object assumption
Gavagai! (picture of rabbit)
(1960)
We might assume that gavagai denotes the whole object
What leads us to assume that it does not translate as ‘rabbit ears’? Or ‘running’? Or ‘animal’? Or ‘rabbit bounding along’?
- This assumption is called a ‘noun bias’; infants are thought to make the same inferences in word learning
Across most languages, children tend to learn high quantities of nouns first, and then gradually learn other kinds of words. (Cf. Japanese, Chinese, Korean: Tardif, 1996)
Constraints theory
there are constraints on how children interpret words in relation to meanings
what is mutual exclusivity?
◦ One object, one label
Taxonomic assumption
◦ Children must learn that a word (e.g. dog) relates to a set of items that do not look the same (e.g. all dogs)
◦ i.e. a word used to refer to a specific object must be extended to all objects of that kind
Semantic specificity in word learning: Bergelson & Aslin
2017
46 infants age 1;0-1;2, 1;3-1;5, or 1;6-1;8
Eye-tracking experiment –object pairs presented while hearing a sentence
Two trial types: Familiar and novel
Three conditions: Matching, related, nonce
If infants infer relationships between related objects (e.g. foot and sock) we would expect them to behave differently in the related condition
Infants draw on mutual exclusivity from a young age, but appear to develop taxonomic knowledge over time
Matching: audio stimuli matched visual stimuli. Infants reliably looked at the target image
Related: Younger infants matched the word to the related object, older infants did not. Eg. Foot and an apple on the screen, ‘look at the sock’.
Nonce: dog and unknown item on the screen, ‘look at the Infants reliably looked at the unfamiliar object, and this trend increased with age. This is an example of mutual exclusivity.
discuss the wug test
Berko (1958)
Do children imitate the grammar of those around them, or can they create grammar from scratch?
61 children age 5;0 – 7;0
English plural /s/, /z/ and /iz/
Also past tense, adjectives, possessives
Wug: a word never heard by the children before
At 5;0, most children could generate progressive –ing, regular past tense forms, and the possessive
At 7;0, ability had significantly increased, reaching 80%+ correct in all categories except 3rd person singular
How do we ‘measure’ grammar?
Mean Length of Utterance (MLU)
- How many words/morphemes?
- How many utterances?
MLU= number of words/morphemes divided by number of utterances.
Higher MLU → more complex utterance
