Grammar Learning B Flashcards
Types of evidence in language acquisition
Positive
Negative
Inference From Absence
Positive Evidence
Hear an adult use a form
But kids must produce sentences/words they’ve never heard before
Valuable because how else are you supposed to learn the irregular verbs in the first place?
Words that sound like this get a +ed - constrained knowledge
Negative Evidence
An adult tells child that a particular form is not allowed
Parents don’t always correct a kid
Kids won’t make every possible error
Inference From Absence
A child never hears an adult use a particularform
Limits on usefulness: there are lots of things that you are allowed to say that you have never heard - does not allow for novelty
Might be useful if you expect something but don’t see it: a common verb never heard with -ed
No negative evidence problem
Kids don’t usually hear negative reinforcement
Superset hypothesis
Superset Hypothesis
subset
Child’s language is a superset of the target language
Child’s experience is a subset of the target language
Three circles:
What kids hears, Target language, Incorrect/Ocergeneralizations
Target language is “we broke it” and “we went”, but kid also says
“we breaked it” and “we goed” - includes all of target language as
well as incorrect generalizations
Positive evidence alone is insufficient to correct the child’s
hypothesis, because their language would then just be a tiny subset of the target (what they hear is a small subset of the language)
Positive evidence will add to incorrect generalizations superset
They must generalize
How do they correct that?
Negative evidence will decrease from incorrect generalizations superset, but
there’s not enough
It has been claimed that since children do not hear enough negative
evidence, this problem cannot be solved by learning
Used to argue for an innate grammatical knowledge
Brown and Hanlon
Studied mothers’ speech in transcripts of three children
The mothers corrected factual evidence and naughty words
However, most grammatical errors passed without comment
Not enough to solve the superset problem
Non-overt correction
Parents were more likely to repeat verbatim the kids’ well formed
sentences than ill-formed sentences
Parents reformulate their children’s error full utterances much
more than repeat their error-free utterances
Sometimes they reformulate to correct, but sometimes they add
words as well to request more information - how does the child
know which is correction?
Might positive evidence be enough?
The problem with positive evidence is that kids will never hear all of the word forms that are permitted
If they are able to arrive at the correct forms via phonological or semantic analogy, this might be enough
Construction Learning and the no negative evidence problem
• The problem: there are generalizations available (transitive construction).
How do kids learn when to use this generalization and when they cannot use this form?
Positive evidence: utterances with the transitive construction, corrections (after being told that you are wrong)
Negative evidence: being told that you cannot use that construction
Semantic verb classes solution
Verbs that have similar meaning display similar grammatical behaviors
But what about idioms and fixed phrases?
Semantic verb classes solution
Verbs that have similar meaning display similar grammatical behaviors
Children do not need to know how individual verbs behave. They just need to know how classes of verbs behave R pushed S R shoved S R disappeared S* R vanished S*
Do children actually use the semantic verb classes solution?
Brooks and Tomasello
Children taught novel words using the intransitive construction (the
ball meeked)
One set of verbs had meaning similar to verbs that occur in transitive
constructions (roll, bounce)
Another set of verbs had meaning similar to verbs that cannot occur
in transitive constructions (come, go)
An attempt was then made to elicit transitive uses of the verbs
4 year olds were more likely than 2 year olds to generalize the verbs
whose meaning was similar to other verbs that can be used transitively
So it seems like kids to engage in something like generalizing on the
basis on semantic categories - to constrain the generalization
Entrenchment
Repeated exposure of the use of a verb in one construction leads children to infer that the verb cannot be used in other constructions
Do kids use entrenchment to figure out constructions?
Theakston
5 year olds, 8 year olds, undergrads
Children presented with sentences containing argument structure
violations (he giggled her, I’m going to disappear it)
High and low frequency verbs
Provided acceptability judgements
Give them (seven) circles from happy to sad - judge "how right" - really, really right to really, really wrong - kids will point or put a coin on the circle
All three groups were more tolerant of violations for low frequency
verbs
Inference through absence
They only hear it in one way; since it’s high frequency, you would
expect to hear the other construction if it’s valid
This thing is predictable, so it’s absence is conspicuous
No one has an umbrella, so it must not be raining
Supports entrenchment
If it’s a high frequency verb, they assume the verb can only be used the
way they’ve heard it used
Preemption
When children encounter a verb used with one construction with a
particular meaning, they infer that it cannot be used with other
constructions to produce the same meaning
If a child hears a verb with the periphrastic causative construction (she
made him cry, he made her giggle) then they infer that it cannot be used in a straight forward transitive construction
Mutual exclusivity principle
Child knows that only one construction will be used to produce a given
meaning
