VI. Syntactic Development and Related Learning Mechanisms Flashcards
How might infants use prosody as a cue for syntactic boundaries?
Languages may use certain prosodic cues to mark syntactic constituents. For example, English typically demonstrates lengthening at the end of syntactic constituents—words are longer at the end of syntactic constituents.
Prosody may also signal to infants the content of a syntactic constituent (e.g. nouns vs. function words), thus marking where the constituent begins and ends.
How might infants use prosody as a cue for the composition of syntactic units?
The words in syntactic units often take on certain prosodic features in a language.
- head-initial vs. head-final languages: languages whose function words appear before content words are head-initial (English)
—Head-initial languages tend to show final lengthening or stress on phrases
—Head-final languages tend to show initial lengthening or stress on phrases
- The phrasal prosody of a syntactic constituent can cue infants on the order of the elements within that constituent
What knowledge do infants have of function words early-on?
-
Shi et al. (1999) found that newborns are able to discriminate between function words and content words on the basis of just their prosodic properties
—function words are typically shorter, less stressed, and have a simpler syllable structure
—In this experiment, infants were habituated to a list of content words. Then, the audio would switch to one of two types depending on the condition: 1) control: infants would begin hearing another list of content words; 2) experimental: infants would begin hearing a list of function words
—Researchers found that infants’ sucking rate increased during the switch in the experimental condition, signaling that newborn infants could recognize the difference between function and content words on the basis of prosody - Shi et al. (2006) found that infants recognize function words as early as 11 months
- Shi et al. (2010) found that 14 mo. old infants can recognize the difference between function words that appear with nouns (the, a) , and function words that appear with verbs (I, he, she)
frequent frames
a proposed learning mechanism for how infants learn to categorise words into syntactic/grammatical categories
- distributional analysis: tracking words and how often they occur amongst other words
- a frequent frame consists of two words that often occur together, but with a word between them: e.g. “she ____ it”
- familiarity with frequent frames can help infants develop “pseudocategories” for words
How have researchers tested the frequent frames theory of learning?
1) Mintz (2003): conducted a corpus study (analysing language texts) to select the 45 most frequent frames used between 6 different children and their parents from the CHILDES database
—Checked how often FF occur across the children; how often FF overlapped between corpora
—Checked the probability of similarity between the (content) words found in specific FF
—frequent frames is both a Nativist and Constructivist strategy
- the frequency and the locality of two words can tell us something about their relationship
2) He & Lidz (2017) + Carvalho (2019): tested 14 and 18 mo. old infants and found that frequent frames are in effect at 18 months. in which they habituated them to one of two sentences with visual stimuli. They found that 14 mo. olds dishabituated to both changes, but 18mo. olds only dishabituated to the verb change.
- infants were shown noun or verb sentences, and then placed into 1/2 test conditions: infants in the noun switch condition would hear the same noun sentence. infants in the verb switch condition hear the same verb sentence; the difference was that the video would change (in the noun switch, this isn’t a problem because the penguin remains the same; in the verb switch, this is a problem because the penguin is not doing the same action as it was in the habituation phase)
How does the frequent frames theory apply to languages besides English? What may be some challenges for frequent frames?
Frequent frames can work for other languages because different languages have different established word-orders (e.g. head initial vs. head final, noun-predicate vs. predicate-noun)
- some languages (e.g. Hungarian) have variable word order.
- some languages (e.g. Romance languages) have different word orders depending on modifiers and pronouns. (“Ella los quiere” vs. “Ella quiere los libros”)
- some languages have changes in noun-modifier agreement depending gender or number
semantic seeds
a proposed learning mechanism in which children develop a strong and stable understanding of a select group of words, and know the conceptual categories to which these words belong
- using semantic seeds, children can develop frequent frames
- semantic seeds which belong to the same conceptual category can appear in the same frequent frame
How might semantic seeds and frequent frames work together as mechanisms of word learning?
Children build knowledge of frequent frames through semantic seeds; children build knowledge of conceptual categories through frequent frames.
- semantic seed and frequent frames methodology works together to form children’s understanding of syntax and syntactic categories
How have researchers tested infants’ use of a semantic seed?
1) Brusini et al. (2019)
2) Babineau et al. (2021)
Babineau et al. (2019): goal, methods, findings, conclusions
goal: can 3-4 year old children use semantic seeds for syntactic bootstrapping? in other words, can 3-4 yr olds use a semantic seed to learn something about how a novel syntactic context can inform the meaning of a novel word in that context?
methods:
training phase
1) kids are exposed to a novel determiner, “ko”, with known words (semantic seeds)
—”ko” either replaces determiners (which pair with known nouns) or pronouns (which pair with known verbs)
2) kids are exposed to a new kind of action and a new kind of object via visual stimuli (man waving arms vs. man holding pink octopus)
test phase
3) kids hear “Wow look! Ko nuve!” (or any variation) while viewing the two videos side by side
4) researchers measure children’s eye gaze
findings: when signalled, children shifted their attention to the matching video depending on what condition they were assigned to; researchers found that children in the noun condition spent less time watching the verb video when hearing “ko nuve!” in the test phase.
—Children still had a preference for the verb video, but this is presumably because it is more visually stimulating
conclusions: 3-4 year old children are able to use a semantic seed to learn new words
WEAKNESSES: 1) done w/ 3-4 mo. old children, but does this work for 12-18 mo. old children (the period for which this learning mechanism is proposed)? 2) this study requires too many steps to be tested on younger children
local dependencies (syntax)
in which elements in a phrase have structural relationships with nearby elements
—e.g. “The man burned the pizza.” burned is a VP that is related to the NP that follows, the pizza
—e.g. “The girl walked the dog.” etc.
non-local dependencies (syntax)
in which elements in a phrase have structural relationships with distant elements
—e.g. “What did the man burn __?”
the NP that would follow burn is replaced with what and moved to the beginning (wh-movement)
—There is a relationship between “what” and the VP, however they do not occur next to each other, making this a non-local dependency
Perkins & Lidz (2021)
goal: can infants aged 14-18 months represent abstract non-local dependencies? in other words, do infants 14-18 months understand that wh-phrases preceding verbs cannot occur with local objects following verbs?
methods: in a look-while-listening task, infants heard audio of either grammatical or ungrammatical passages: e.g. “What did the cat eat?” vs. “*What did the cat eat it?” Infants were shown a screen of abstract moving shapes, and researchers measured children’s interest for the sentences using their looking time towards the screen.
findings: 18-mo. old children had a longer looking time towards the screen when presented with grammatical than ungrammatical sentences –> this result was found for both grammatical + ungrammatical Wh-questions, and both ungrammatical + grammatical declarative statements.
conclusions: 18 mo. old children have knowledge of local-dependencies and how the occurrence of “what” early-on can satisfy a verb’s need for a direct object.
