Language Acquisition Finals

Learning words involves separate two tasks by a child

Segmenting and Associating

• Segmenting → In continuous speech, there are rarely pauses between words. A sentence looks like a sound wave, so a child must figure out where one word ends and the next begins.

  • The "Elemeno" example: In the alphabet song, the rapid sequence "L-M-N-O" is often misheard by children as a single word, "Elemeno." Children eventually resolve this by hearing the individual letters used in other contexts.

Associating → Associating meanings with word forms (The Mapping Problem). This is the process of connecting a specific sound (word form) to the correct concept in the real world.

• The Challenge: A single word could theoretically refer to an object, its color, its movement, or its parts.

• Refinement: Children use repeated exposure to narrow down meaning.

  • Rabbit Case: Determining that "Rabbit" refers to the animal itself, not "white" or "running."

  • Apple Case: Correcting the error of calling a tomato an "Apple" (overextension) by learning that the word refers to the fruit, not just anything round and red.

Bootstrapping

The idea that children use the small amount of linguistic knowledge they already possess (sound, meaning, or grammar) to "crack the code" of the parts they don't know yet.

• Semantic Bootstrapping (Meaning → Grammar)

  • Theory (Steven Pinker): Children use their understanding of real-world objects and actions to figure out grammatical rules.

  • The Logic: They observe the world (e.g., a dog chasing a cat) and identify the "do-er" (Agent = a dog) and the "receiver" (Patient = acat).

  • The Learning: By noticing that the Agent ("Dog") comes first in the sentence, they learn the Subject-Verb-Object (SVO) grammatical rule.

• Syntactic Bootstrapping (Grammar → Meaning)

  • Theory: Children use the grammatical structure of a sentence to guess the meaning of a new word.

  • The Logic: The structure tells them if an action is done to someone (Transitive) or done independently (Intransitive).

  • The "Kradding" Experiment:

    • Scenario A (Transitive): "The duck is kradding the rabbit." → Structure implies interaction. Child interprets "kradding" as pushing.

    • Scenario B (Intransitive): "The duck and rabbit are kradding."→ Structure implies parallel action. Child interprets "kradding" as waving arms.

• Prosodic Bootstrapping (Sound → Structure)

  • Theory: Children use the "music" of language (rhythm, stress, pauses) to find word boundaries and grammatical units.

  • Word Boundaries: In English, strong stress often marks the start of a word (e.g., DOC-tor).

  • Clause Boundaries: Pauses and pitch drops help the child hear where one phrase ends and another begins (e.g., "The boy [pause] who ran fast [pause] fell down."").

  • The Bootstrapping: The child hears the pauses or drops in pitch and realizes that "who ran fast" is a separate "chunk" (a clause) that belongs together.

The problem of finding word forms arises because:

1. Speech is continuous (there are no consistent breaks between words)

2. Lack of Isolation (Segmentation Ambiguity): Words are rarely taught one by one; they appear in complex streams. Input: "We are going together." "Together" creates a sequence that sounds exactly like three other words: "To get her." The child must realize that in this context, "Together" is one adverb, not a prepositional phrase about "getting her."

3. Infants are not born with a lexicon

What is Phonological Bootstrapping of Lexical Acquisition?

This theory suggests that infants use the sound patterns (prosody and phonology) of speech to figure out where words begin and end. Since infants do not have a mental dictionary yet, they rely on the "music" and "rules" of the sound stream to chop continuous speech into word-sized chunks.

The Three Main Strategies:

1. Distributional Regularities (Statistical Learning)

Infants act as "little statisticians." They track how often certain sounds stick together.

• They calculate Transitional Probability. If Syllable A is always followed by Syllable B, they likely form a word. If Syllable B is rarely followed by Syllable C, there is likely a boundary between them.

  • The Experiment (8 Months):

    • Input: Babies heard a continuous stream like "bidakupadotigolabubidaku..." (actually four nonsense words repeated: bidaku, padoti, golabu, tupiro).

    • The Math: Inside a word (bi → da), probability was 100%. Between words (ku → pa), probability dropped to 33%.

    • Result: Babies recognized the "words" and were surprised by "part-words" (like kupado), proving they had segmented the stream based on math alone.

2. Typical Word Shapes (Metrical Segmentation)

Infants learn the dominant rhythm of their native language and use it as a cookie-cutter to slice up speech.

• The Trochaic Bias (English): About 90% of English content words follow a Strong-Weak stress pattern (e.g., DOC-tor, CAN-dle, BA-by).

• The Strategy: When an English-learning infant hears a strong syllable, they assume a new word has started. For instance, DOC-tor saw the BA-by.

• The "Guitar" Error: Because gui-TAR has a Weak-Strong pattern, infants often missegment it. They hear the stress on TAR and assume that is the start of the word, leaving gui attached to the previous word. For instance, Baby played thegui TAR.

3. Phonotactic Constraints (The "Traffic Rules")

Phonotactics are the rules governing which sounds are allowed to sit next to each other within a word.

• Illegal Clusters (/vt/): The sequence /vt/ is rare inside English words. If a child hears "I love two birds," the /vt/ cluster signals a boundary: /lʌv/ [CUT] /tu/.

• Position Constraints (/ng/): The /ŋ/ sound (as in sing) never starts a word in English. If a child hears "Sing a song," they know /ŋ/ must be the end of the first word, not the start of the second.

How We Know Phonological Bootstrapping Works (The Evidence)

The theory is plausible because speech physically changes at boundaries, and infants are proven to detect these changes.

A. The Acoustic Cues

1. Pauses: Silence is usually inserted between clauses, acting as a "stop sign."

2. Syllable Lengthening: We naturally stretch the last syllable before a boundary (e.g., "When the baby sleeeeps, the house is quiet").

3. F0 Resetting (Pitch): Pitch naturally drops at the end of a clause and jumps up (resets) at the start of a new one. I like him (F0). He is a good boy.

B. Proof of Infant Sensitivity

• The "Coincident" Study (6-7 Months): Babies listened longer to sentences with natural pauses ("Cinderella lived in a big house... [pause] ...with her stepmother") than unnatural ones ("Cinderella lived in a big... [pause] ...house").

• The Newborn Study (3 Days Old): Newborns could distinguish the sound sequence /ma-ti/ when it was inside a word (mathematicien) versus across a boundary (panorama typique). This proves humans are born able to hear boundary cues like lengthening.

• The 9-Month Shift (Phonetic Detail):

  • At 6 Months: Babies discriminate languages based only on Rhythm. They can tell English (Trochaic) from Norwegian (Pitch Accent) but fail to distinguish English from Dutch (same rhythm).

  • But at 9 Months (The Filter Test): They begin to hear Phonotactics constraints. They can finally distinguish English from Dutch because they now detect specific sound rules (like the fact that Dutch allows words to start with /kn/, but English does not). For instance, knight.

Phonological Bootstrapping & Universal Grammar (UG)

While statistical learning (finding word boundaries) is powerful, it is insufficient for learning complex grammatical rules. The child must combine Prosody + Statistics + Phonotactics with an innate Universal Grammar (UG).

• The Limitation of Statistics (The "Ever" Problem):

  • Rule: The word "ever" (Negative Polarity Item) requires a negative word like "No" to license it (e.g., "No person will ever come").

  • Failure: A simple statistical counter fails because the distance between "No" and "ever" can be vast.

  • Conclusion: This dependency suggests humans have a built-in grammatical structure (UG) to handle long-distance rules that simple math cannot solve.

The "Two-Step Model" of Lexical Acquisition

Learning a word is not instant; it happens in two phases.

• Step 1: Form Extraction (The Empty Container):

  • Around 8 months, the child identifies a sound sequence (e.g., /kæ-t/) and stores it in long-term memory. At this stage, it has no meaning.

• Step 2: Meaning Attachment (Filling the Container):

  • Around 10-12 months, the child maps the concept (furry animal) to the stored form.

• Evidence (The Story Experiment): 8-month-olds who heard stories with nonsense words (e.g., "Python") preferred those words two weeks later over new ones, proving they stored the sound form without knowing the meaning.

The Core Challenges of Word Learning

A. Hypothesis Formation (The Child as Scientist)

The child guesses a word's meaning and tests it.

• Input: "Ball" (for a red ball).

• Hypothesis: "Ball" = Red? Round? Toy?

• Test: Child points to a Red Apple and says "Ball." Parent says "No."

• Child rejects "Red" and strengthens "Round" after seeing a blue ball.

B. Word-to-World Mapping Problems

Looking at the world is not enough because the world is messy.

• Scene Ambiguity: A parent says "Look at that!" in a cluttered room. The child doesn't know if "that" is the cat, the truck, or the floor.

• Abstract Nouns: You cannot point to "Idea" or "Tomorrow." These require sentence context (Syntactic Bootstrapping) to learn.

• Time-Locked Verbs: Nouns are usually present (child sees Dog → hears "Dog"). Verbs are often mismatched in time (child sees empty table → hears "We will eat").

C. The Problem of Induction (Perspective Ambiguity)

Even if the event is clear, the perspective is not.

• Give vs. Receive: The physical action of handing over a book is identical for both. Without grammar, the child doesn't know if "Gave" refers to the donor's act or the receiver's.

• Chase vs. Flee: A dog running after a cat is both "chasing" (dog's view) and "fleeing" (cat's view).

Cues and Biases (How Children Solve the Problems)

Children use social cues and mental shortcuts (biases) to narrow down meanings.

A. Social Cues (18 Months)

• Eye Gaze (The Bucket Experiment): Children realize words refer to what the speaker is looking at, not what the child is looking at.

  • Scenario: Researcher looks into a box and says "Modi."

  • Result: Child ignores the toy in their own hand and assumes "Modi" is the thing in the box.

B. The Whole Object Bias

• The Assumption: A new word refers to the entire object, not its parts or properties.

  • Benefit: When hearing "Rabbit," the child doesn't waste time guessing if it means "ears," "white," or "fur."

C. The Mutual Exclusivity Bias

• The Assumption: Every object has only one name.

  • The "Handle" Logic: If a child already knows "Cup," and the parent points to the cup saying "Handle," the child concludes "Handle" must be a part of the cup, because the whole object is already named as “cup”

D. The Taxonomic Bias

• The Assumption: Words label categories (Taxonomy), not themes.

• Markman & Hutchinson Experiment:

  • No Word: Show a Cow → Child picks Milk (Thematic/Story relation).

  • With Word ("Find the Dax"): Show a Cow → Child picks Pig (Taxonomic/Category relation). They are both animals.

  • Conclusion: Words force the child to think in categories (Animals), not stories (Farm life).

Syntactic Cueing of Verb Meaning (Syntactic Bootstrapping).

The Core Concept

Since verbs are difficult to learn just by observing the world (e.g., distinguishing "giving" from "receiving" or "chasing" from "fleeing" just by looking), children rely on sentence structure (syntax) to crack the meaning of the verb.

A. Noun Syntax vs. Verb Syntax ("The Sib" Experiment)

This experiment proves that grammatical markers alone tell the child if a word is an Object or an Action.

• The Setup: A picture shows hands kneading (action) a bowl of dough (substance).

• Condition 1 (Noun): "Show me a sib."

  • Result: Child points to the dough. The article "a" signals a Noun/Thing.

• Condition 2 (Verb): "Show me sibbing."

  • Result: Child points to the kneading hands. The suffix "-ing" signals a Verb/Action.

B. Transitive vs. Intransitive ("The Gorp" Experiment)

This study (Naigles' Preferential Looking Paradigm) proves children use sentence structure to determine who does what to whom.

• The Setup:

  • Screen A (Causative): A Duck pushes a Rabbit down.

  • Screen B (Non-Causative): A Duck and Rabbit wave their arms together.

• The Input: "The duck is gorping the bunny."

• The Logic: The sentence structure is Transitive (Subject + Verb + Object). The child knows this structure usually means "X does something TO Y."

• The Result: The child looks longer at Screen A (Pushing). They conclude "gorp" must mean "to force/push."

Note on Double Objects: If a child hears "John gorped Mary a cake," the structure (Double Object) implies a transfer of possession. The child assumes "gorp" means something like "give," not just "carry," because you cannot say "Carry Mary a cake."

C. Using "Multiple Frames" (The Advanced Naigles Experiment)

Children can compare two sentences to figure out subtle differences between Causative Verbs (change of state, e.g., "break") and Contact Verbs (activity, e.g., "hit").

• The Scenario: Children watched a duck doing something to a frog.

• Group 1: Causative Alternation (Change of State)

  • Input: "The duck sibbed the frog" AND "The frog sibbed."

  • Logic: The object (frog) became the subject in the second sentence. This pattern (like "The glass broke") implies the object underwent a change.

  • Result: Children assumed "sib" meant bending the frog (changing its posture).

• Group 2: Object Omission (Activity/Contact)

  • Input: "The duck sibbed the frog" AND "The duck sibbed."

  • Logic: The subject (duck) stayed the same, just dropping the object. This pattern (like "John painted") implies a continuous activity.

  • Result: Children assumed "sib" meant merely touching the frog's head.

D. Is Syntactic Cueing Real? (The Blind Child Study)

This provides powerful evidence that visual observation is not necessary for learning visual verbs; syntax is enough.

• The Problem: If word learning required "looking at the world," blind children should never understand words like "Look" vs. "See."

• The Experiment:

  • Command A: "Touch the chair, but don't look at it." → Child taps/rests hand (passive).

  • Command B: "Now, look at the chair." → Child explores shape/texture with hands (active).

• The Conclusion: Since they cannot see, they must have learned the meaning of "Look" (Active exploration) from the grammar (syntax) of sentences they heard, not from visual observation. They used structural info to figure out the meaning.

Syntactic Bootstrapping

This is the process of utilizing the known "blueprint" of sentence structure to deduce the meaning of new words or fill in gaps when speech is incomplete.

• The Logic: The child thinks, "I know the sentence building rules, so I can guess what comes next."

• The Example (Incomplete Input):

  • Input: "The dog is..." (Speech cuts off).

  • The Inference: The child identifies "The dog" (Subject) and "is" (Auxiliary). Their internal grammar predicts that a verb/action must follow. Using context (e.g., a sound), they mentally complete it: "The dog is barking."

• Why It Matters: This allows children to learn effectively even from degenerated input, which is common in everyday conversation.

Semantic Bootstrapping (Meaning → Grammar)

This is the reverse of syntactic bootstrapping. The child uses their knowledge of the physical world (Objects and Actions) to figure out the abstract rules of grammar (Nouns, Verbs, and Word Order).

• The Logic: The child assumes that words referring to "Things" (like apples) are Nouns, and words referring to "Actions" (like eating) are Verbs. They then look at where these words fit in a sentence to learn grammatical rules.

• Example 1 (Learning Object Position):

  • Known Concept: The child knows an "Apple" is a physical thing.

  • Input: "She's eating an apple."

  • Grammar Learned: Since the "Thing" appears at the end, the child learns the rule that Objects (Nouns) can go at the end of the sentence in English.

• Example 2 (Learning Subject Position):

  • Known Concept: The child knows a "Car" is a physical thing.

  • Input: "The car moves fast."

  • Grammar Learned: Since the "Thing" appears at the start, the child learns the rule that Subjects (Nouns) come before the Verb in English.

Comparison: Semantic vs. Syntactic Bootstrapping

Feature	Semantic Bootstrapping	Syntactic Bootstrapping
Direction	Meaning → Grammar	Grammar ->Meaning
Starting Point	The child knows concepts (Apples, Cars, Eating).	The child knows sentence structures (Subject-Verb-Object).
Goal	To learn grammatical rules (Where do Nouns go?).	To learn word meanings (What does "Gorp" mean?).
The "Apple" Example	"I know 'Apple' is a thing. It's at the end of the sentence. So objects go at the end."	N/A
The "Blind Child" Example	NO	"I hear 'Look' used in active sentences. So 'Look' must mean active exploration."
The "Dog" Example	NO	"The dog is..." ->Grammar predicts a verb is coming (e.g., barking).

Final Summary for Exam

• Phonological Bootstrapping: Using sound/rhythm to find word boundaries (e.g., Trochaic stress, Pauses). English babies expect words to start with a strong beat (Strong-Weak) BA-by, no beginning with ING-. The baby correctly identifies it as one word because it follows the rule. Example (Pauses/Pitch):

  • The Input: "The baby [pause] cried." The baby cried. (F0 resetting) His mum came quickly.

• Semantic Bootstrapping: Using world knowledge to learn grammar rules (Thing = Noun). The child sees a Dog (a physical thing) barking. "I know 'Dog' is a Thing (Agent). It came at the start of the sentence. Therefore, in this language, Subjects/Nouns come first."

• Syntactic Bootstrapping: Using grammar rules to learn word meanings (Transitive verb = Causative meaning).

  • "The Duck is kradding the Rabbit." (Transitive)

    • The Logic: The structure is X verbing Y. The child guesses "Kradding" means pushing (doing something to someone).

    • "The Duck and Rabbit are kradding." (Intransitive)

      • The Logic: The structure is X and Y are verbing. The child guesses "Kradding" means waving (doing an action together/independently)

  • Example (The Blind Child):

    • The Input: "Look at the chair."

    • The Logic: The child hears "Look" used as a command (an active verb). Even without sight, they infer "Look" means active exploration (touching/feeling), not just passive reception.