V. Prenatal Language Experience and Related Learning Mechanisms Flashcards
prosody
the rhythm and melody of a language; the changes in pitch, stress, duration, and intensity within speech
speech envelope
the overall amplitude fluctuations of speech; the rhythm and melody of speech
Why are bootstrapping theories part of the Nativist tradition?
Bootstrapping theories are those which suggest that children can learn something about one domain of their language by using knowledge in a different domain. These are Nativist theories because they rely on the idea of infants having knowledge of an underlying hierarchical structure of language, which infants would not be able to acquire through the linguistic input they receive alone.
prosodic bootstrapping
using prosodic elements of speech to infer something about the meaning of words in a given language
- e.g. function words tend to be shorter and less emphasized in speech
When do infants gain the ability to hear in the womb? What do they hear?
Infants gain their hearing in the third trimester. They have access to low-pass filtered speech and even then, may have difficulty perceiving speech over 400 Hz.
- infants have access to prosodic elements of speech, but cannot hear words, sentences, phonemes, etc.
- the womb filters out speech
low-pass filtered speech
in prenatal language development, this is speech from 0-1000 Hz
- infants have access to LPF speech in the womb, but even so, they cannot always perceive speech above 500 Hz
- phonemic distinctions not present in LPF speech
newborn speech preferences
- newborns prefer human speech over other sounds
- newborns prefer their mother’s voice over a stranger’s
- newborns generally prefer listening to their native language, however they don’t show a preference if they hear a language that is prosodically/rhythmically similar to their native language
- infants exposed to two languages prenatally will prefer those two languages over others
When does word segmentation begin?
Infants begin word segmentation around 6-12 months.
Why might speakers of a different language seem as though they’re speaking fast? How does this relate to the problem of word segmentation?
Speakers of a language we do not speak may sound as though they’re speaking fast because we lack information about the word boundaries in their language.
- Without word segmentation, we don’t know where words end and begin, which can make it difficult to isolate the phonological forms of specific words
How is word segmentation related to the mapping problem?
The mapping problem requires isolation of the phonological form of a word; it requires being able to identify that word in different environments (different speakers, w/ background noise, etc.) and from a continuous speech stream.
Word segmentation is the process of identifying word boundaries—where words begin and end. This helps us isolate words in continuous speech streams.
cues used to achieve word segmentation
- spaces
- pauses
- words you already know
- stress
- transitional probabilities
- phonological phrases
lexical stress
that different parts of words can have stress or emphasis
- languages can have dominant lexical stress patterns: for example, English is typically stress-initial for bisyllabic words
—in English, stress in 2 syllable words typically falls on the first syllable
- can be used to achieve word segmentation
—if you notice dominant patterns of stress in your language, you may segment words by following those dominant patterns; you won’t always be successful, but you can make pretty good guesses
transitional probability
the probability of transitioning from one element of speech to another in an ordered string of elements
- stress patterns in a language occur a certain amount of times
- transitional probabilities can occur between syllables, words, and sentence elements
- conditional probability: TP(A -> B) = Pr(AB) / Pr(A)
—The transitional probability of element B coming after element A would be the probability of seeing A and B occur together, divided by the probability of seeing A occurring alone
- transitional probabilities have a value of 0-1:
—if B never occurs after A, TP = 0
—if B always occurs after A, TP =1
- you can use the TP between syllables to understand when words begin and end: words are internally consistent and externally differentiated from each other
—we might put word boundaries between syllables that have low transitional probabilities
—two syllables that occur sequentially within words likely have a higher probability than two syllables that occur sequentially between words
—there doesn’t have to always be high TP within a word, however TP can be useful
statistical learning
noticing and registering patterns in order to develop knowledge
- part of the Constructivist perspective of language development
- Using transitional probabilities for learning words and word boundaries is a kind of statistical learning
- Infants have access to visual transitional probabilities as statistical learning as early as 2 months old
How do researchers assess infants’ use of transitional probability?
Researchers used an artificial language to understand how infants may use transitional probability.
- advantages: researchers can control for infants’ prior experience with the language, and they can control statistical regularities in a language
- Infants were played a speech stream consisting of only 3 made-up words:
—only thing giving away that these were individual words were the internal consistency within the words
- Infants were tested using a head turn preference set-up: they were played different sequences of 3-syllables (“bidaku,” “padoti,”)
—Found that infants preferred to listen to non-words longer than words: preference for novelty, for words that are unfamiliar to them
—By 8 months, infants can use transitional probabilities to distinguish between words and non words
