Language Flashcards
Infant speech and speech perception: what changes in the first year of life?
Speech - 2 month olds “coo”, 6 month olds “babble”, 10 month olds babble in their native language, and 12 month olds have their first words.
Speech perception - newborns: the prosody (speech contours) of their native language, prefer speech to non-speech and show some level of categorical perception. By the first 6 months, infants show categorical perception across the world’s languages (eg: Japanese kids can discriminate R/L. By 10 months categorical perception is limited to native language.
What are the basic milestones/trajectories of word learning, morphosyntactic development and pragmatics?
12 months - first word
2-6 years = word explosion 100-2000 words compared to 5000 to 20000 words at 7
Morpho-syntactic = plurals stuff. 2-3 year olds show mastery of basic inflectional morphology by going from wug to wugs. They then find out about irregulars (eg: tooths and mouses) and over-apply them to regular situations.
Pragmatics typically develop throughout middle childhood along with social congition (ToM) and metacognition
How do deaf children respond to varying levels of input? And at what age are they exposed to it?
Many deaf children are born to speaking parents. If the parents don’t sign, kids will make up their own basic gestural system (homesign). If they don’t learn properly by puberty, they will likely never go beyond their homesign.
If parents do learn to sign, kids will learn and easily go beyond them. eg: parents use correct inflectional and derivational morphemes 65% of the time, while kids use the correct ones 85+% of the time.
Kids with native signing parents show no delays in language development compared to hearing kids
NSL: What changes across cohorts of signers?
Firstly, language splits meaning up into compositional parts, so they can be recombined to generate more kinds of meaning. For example, “the ball rolled down the hill”: the motion (rolled) and path of motion (down the hill) occur simultaneously, but are expressed independently in language.
There were 2 conditions: A - simultaneous sign expression of balling rolling down the hill (Spanish signer) and B - componential expression of the ball rolling down the hill (NSL signer. Results showed usage went down across cohorts for A and up across cohorts for B, indicating a preference for componential expression and an ineffectiveness of simultaneous.
What are the arguments and research for nativism for the complexity of language?
Nativists like Noam Chomsky strongly argued that behaviourist and Piagetian learning theories simply do not hold up. Hierachical structure, phrase trees and recursion are simply too complex to be explained by these mechanisms. Therefore, we must have some sort of genetic endowment for the ability to at least learn language, if not endowment for language itself. This endowment is called Universal Grammar (UG) and is built around the VP = NP + V model.
Consider the sentence “the boy is crazy”. To turn this into a question, we simply take the “is” and put it at the front of the sentence: “Is the boy crazy?” It could be then claimed that the rule for turning these types of sentences into questions is to take the “is” and put it at the front, and a constructivist would argue children can learn this through statistical tracking etc. However, this rule does not always apply. Consider, “The boy who is smoking is crazy”. As a question, it becomes “Is the boy who is smoking crazy?” not, “Is the boy who smoking crazy?”
Thus, the hierachical structure always trumps individual words/phrases learnt through statitsical tracking. Children have never been reported to complete this task incorrectly, which suggests UG and innate language complexity. Rules of syntax operate over abstract syntactic structure. Chomsky concluded that structure dependence is an innate schematism. This is the “parade case of innate constraint”.
How do constructivists apply general learning mechanisms to explain language development?
They accept we do have some genetic endowment, but it’s not very specific. Rather, the input is rich and kids can quite easily learn from. More powerful learning mechanisms have emerged since Chomksy refuted Piaget and Skinner eg: statistical learning, analogical/relational abstraction, social reasoning (ToM).
Statistical learning - hypothesised that children will learn the transitional probabilities between sounds, words and phrases and how they are distributed more globally to learn grammatical categories and phrase structure rules. Transitional probability = chance that after a particular syllable, another will occur. For example, in the phrase pretty baby, the TP between pre and tty and ba and by would be high, but the TP between tty and ba would be low. The beginning of words predict the end of words, but whole words don’t predict other whole wordsSo, can infants do this? Saffran, Aslin and Newport (1996) tested this with 8 month olds by using continuous streams of artificial grammar with no gaps. The only cue for word boundaries was transitional probability information: high within words (bi > da) and low between words (ku > pa). The infants listened for 2 minutes and the Es measured how long they listened to repetition of words and non-words. They showed a novelty preference for non-words, which means they track transitional probabilities and perhaps this is used to build up from sounds to words etc.
How does this re-explain UG from constructivist perspective? Both constructivists and nativists agree kids never make mistakes in “the boy who is smoking is crazy”. One study proposed that “who smoke” is a phrase often used by humans (high statistical regularity) while “who smoking isn’t. The study found that “is” questions produced 0 structural dependence errors while “can” questions produced about 7% and up to 43% structural dependence errors. UG would not predict such specific word-difference effects.
Gradual abstraction - refers to the build up of verbs and nouns by applying rules about verbs and nouns you already know. Children are actually quite conservative, according to constructivists, and only use words in ways they’ve heard them used before. They break down utterances into schemas that can be generalised to new situations. Novel-verb studies investigate whether infants can use made-up verbs in new constructions to see if they have a ‘verb-general rule’. eg: the bird is tamming the sock. No 2 year olds could apply it to new construction, half of 2.5 year olds could apply it and all 4 year olds had no problem. It sggests that kids who can’t do it don’t have a categorical representation of verbs, but just know a bunch of words that happen to be verbs that exist as independent ISLANDS.
Outline the back and forth between early abstraction accounts and constructivist accounts.
1
