Language Acquisition Study Guide PDF

1: Setting the stage Language as a cognitive ability ○ Language is the scientific study of the mind and human intelligence including Reasoning Problem solving Abstract thinking Planning Learning Language ○ Focused on understanding co...

1: Setting the stage Language as a cognitive ability ○ Language is the scientific study of the mind and human intelligence including Reasoning Problem solving Abstract thinking Planning Learning Language ○ Focused on understanding cognitive abilities instead of language as a social construct ○ What is it inside of our heads that enables us to speak language and understand language. ○ so our perspective in this class is going to be to looking at the mental structures that generate behavior. ○ Langauge is not just a set of behaviors that we demonstrate First language acquisition vs. second language acquisition ○ First language acquisition: Human infants and children Learning the language(s) of their environment (aka native languages) Unconscious and effortless process ○ Second language acquisition: Humans Learning additional language(s) after first language(s) Requires conscious effort Difficult to achieve native-like skills ○ Hierarchy of linguistic representations; physical vs. abstract characteristics of language (see “Mushrooms are an edible fungus” example) ○ you produce some sound waves. The sound waves go and they hit another person. And something about that person's mind enables them to go from those sound waves to arriving at a similar thought to the one that you had. ○ Mushrooms are inedible fungus. Talk about that as a sentence or an utterance depending on how we're talking about it. There's different ways you can characterize that. spectrogram. Which is representing differences in the amplitude of sound across different frequencies. This is a very physical way of representing this. We can talk about the continuous speech signal and acoustic properties that it has. But we as humans from that kind of continuous speech signal, we extract sound categories. From those sound categories, we can group them together to develop group them into morphine units or parts of words. We can combine those morphines together to form whole words.. Words combine into syntactic structures and those are associated with meanings.  When we were talking about what the language and people were saying, form, meaning, sound, these different sub components, that's what we're talking about. ○ Hierarchical and constrain one another ○ All that children arguably have access to is these sound waves hitting their head. And from the sound waves, they have to abstract to higher and higher more complex, more abstract levels of representation Logical problem of language acquisition ○ We have a finite set of experiences with language, but we have an infinite ability to generate language. ○ How are we able to go beyond what we’ve gotten as input? ○ Developmental problem :How do you get from (what looks like) nothing to something? What stages does the learner move through and which learning mechanisms do they employ? ○ Linguistic knowledge goes beyond experience Audio recording vs. note taking Generalizations across multiple experiences Sixty-two polka-dotted elephants flew to the moon. *polka-dotted the flew sixty-two to elephants moon ○ Not all linguistic properties are revealed through natural conversations ○ What we could say vs. what we have a reason to say ○ Positive and negative evidence Poverty of the stimulus Nature vs. nurture in language development ○ Nature vs. nurture Human uniqueness Demonstrates importance of our biological nature ○ Language deprivation Demonstrates importance of our environments in expressing our capacity for language (think Genie) ○ Biological bases ○ Regions of the brain the specialize in language (Broca’s, Wernicke’s) ○ Physical articulators (vocal tract, hands) ○ Genes linked to language (FOXP2) linked to development of language disorders. Environmental influences ○ Quantity of linguistic input (total # of words) ○ Variability of linguistic input (# of diff. words) ○ Family composition (# of caretakers, # of siblings) ○ Parental responsiveness (vs. phone use) ○ Parental stress (work load, mental health) ○ Socioeconomic status (SES) Cross-sectional methods ○ Cross sectional  So here, you're just going to have one point in time that you measure for each child, so you can measure many more children. And here you can still look at age related questions, right? You can have one group of children that are 12 months and another group of children that are 20 months. Longitudinal methods ○ what longitudinal means is you're going to follow a child across a time frame. If you want to know how a particular structure or aspect of speech develops from the age of 12 months to 24 months, longitudinal. So all of Roger Brown's work was based on that kind of study. He followed a couple of kids with pseudonyms that do grants and like Adam and Eve across time to conclude things about the early production. 6 stages of early language production ○ Stage 1: Preverbal Roughly 0-6 months Main vocalizations: crying, cooing Begins with involuntary responses to stimuli Infant vocal tract develops to become more adult-like (4 months) infants begin to gain more control range of sounds/segments infants can produce increases Larynx drops, allowing for a greater range of consonants Jaw control develops, allowing for a greater range of vowels No intention to communicate Beginning of language-like sounds ○ Stage 2: Babbling Roughly 6-12 months Infants start to produce syllable-like sounds Initially vowels [V] Consonants introduced [CVCV] Then more variability Infants acquiring both signed and spoken languages Infants starting to manipulate their articulators Vocal track, mouth or hands Lots of stops in early babbling, regardless of language environment Infants produce things that sound like words, but without meaning Mama, dada, papa, baba ○ Stage 3: One-word utterances Roughly 12-20 months Toddlers start to produce meaningful words Entire utterances are made up of 1 word Words referent to concrete aspects of their environment and routine activities Names of objects or food items: ball, milk Names of people: Mommy  Words associated with social routines: hi, no Words may not sound adult-like (difficult to judge when babble becomes a word) ○ Stage 4: Two-word utterances Roughly 20-30 months More and more words (50+) More concrete verbs, modifiers Biggest utterances are made up of 2 words Mommy sock More drink Daddy home I hungry Go play Easier to see communicative intentions Beginning of structure, but limited by small vocabulary ○ Stage 5: Telegraphic utterances Roughly 30-40 months Named after writing conventions for telegrams Cost/word, functional words are removed Utterances can be made up of more than 2 words, but still are not adult-like That my book Big doggie eating Adam need one Why no functional words? Kids don’t know them? They aren’t as important to convey meaning (and talking is hard work, so kids conserve effort) ○ Stage 6: Beyond telegraphic Roughly 40 months and beyond Utterances get longer and complex Functional categories are more prevalent Conversations go beyond present moment Storytelling, discussions of past and future Conversational abilities are still not adult-like Kids mention irrelevant information Kids do not take others’ perspective or knowledge into account II. Lexical development Mapping problem ○ First step - You have to isolate the phonological form, the sound that is the word that you're trying to learn. You have to isolate this because you have to find something that has a certain amount of consistency across contexts. ○ The second step is to identify a concept or a meaning that you want to relate it to. ○ Third step - You have to know how to use it, basically. ○ Understanding the range of contexts that can be used in and determining the syntactic category. So that means you've got phonetic and phonological knowledge about it, you have syntactic knowledge about it. You also have semantic knowledge about it and pragmatic knowledge about it. ○ How you isolate the phonological form, identify a concept to relate it to, and how you link these two things together. The problem of how you do that, that's called the mapping problem, how you map from a sound to a meaning Induction and the hypothesis space ○ How do we know which thing on the left side, linked to which thing on the right side of the screen, right? This is a problem of induction. It requires inductive inference. Because children are only ever exposed to specific examples of a word being used, but from these specific examples, they must derive general principles about that. ○ infer what the word means based on how it is used and their own biases  ○ Constrain the set of options or narrow the hypothesis space The indeterminacy of reference and the Gavagai problem ○ So what makes the mapping problem, in particular so hard, has to do with the evidence that we get from any specific example of using the word. This has to do also with the indeterminacy of reference. ○ The idea here is that in any one situation, the evidence that you're going to get is compatible with multiple hypotheses ○ But also because there are different ways of looking at the same situation. This is also expanding the number of ways we have of talking about something: gavagai problem WVO Quine ○ It is a thought experiment and to think about what the situation would be like, if we were trying to learn a new word in a language that we didn't know. He poses this example where you should think about yourself as an anthropologist who's visiting and observing a new tribe. You hear someone in the tribe say Gavagai, and this is what you see in front of you. Let's put ourselves in the position of this anthropologist. What could gavagai mean? What's the concept that we want to link this phonological form to? ○ Logically, the child has nothing to go off of to figure out, or you the anthropologist has nothing to go off to figure out. ○ Is it talking about that kind of object, a part of that kind of object, all of the parts of that kind of object together, but never separated. You don't know what's relevant, exactly. Okay. This is the indeterminacy of reference. Word learning biases: whole object bias, mutual exclusivity ○ Whole object bias Also known as: Whole Object Constraint Kids assume that new nouns will refer to an object (in its entirety, instead of some aspect of it, part of it, or state that it is in) Recall the Gavagai problem: YES Rabbit NOT Ears, furry, brown, alive, running.... Why? Partly referential bias + syntactic category (nouns refer to objects) Partly an extra-linguistic fascination with objects (consider ball vs. lava example in the text) What counts as a whole object? Something that determines what kind of object it is Types of overextensions are informative: Usually shape. Not texture, size or function. ○ Mutual exclusivity Also known as: the Principle of Mutual Exclusivity Markman (1989) ○ Every object has only one name/label Do you see the dax! Give me the dax! Limitations? Bilingual or multilingual language acquisition (dog vs. perro) Synonyms (bunny vs. rabbit)  Parts of objects (paw vs. rabbit) This word learning bias emerges between 14 to 17 months So first words (around 12 months) cannot be learned using this bias Theories of word learning: ostension, fast-mapping, cross-situational word learning, bootstrapping ○ How do they map a phonological form to a concept given the indeterminacy of reference? several different theories of this. ○ Ostension John Locke (1690) Teaching by intentional demonstrating or showing Requires: Child can observe the object of reference We draw attention to it They pay attention to it We use the word to label it Sometimes supported by joint attention Triangulating attention on an object Likely supports early concrete vocabulary: semantic seed Limitations? ○ Fast mapping Susan Carey (1978) Establishing link from sound to meaning in one exposure (not really how) Similar to ostension, works in limited contexts Doesn’t fill in all the details, but crucial support to get a link started Chromium experiment – Carey & Bartlett (1978) w/ 3yo Can you bring me the chromium tray? Chromium = olive green Constraints/limitations? Narrow range of possibilities (tray is a known word) They differ in one dimension (color) One already has a known label (red) ○ Cross-situational word learning Siskind 2006, Yu & Smith 2007 Generalization across experiences with a word Hypothesis space narrows with each experience Can work together with fast- mapping Called: Propose but verify Initial weak link is strengthened across experience  ○ ○ Bootstrapping Gleitman 1990 Referent isn’t always easy to find in the physical world Linguistic context can reveal a word’s meaning I love eating gorps gorp = edible object He daxed the ball dax = action verb I blick that it’s going to rain today Joint attention ○ the developmental ability of a child to share focus with another person on a common object or event, which is considered a crucial foundation for language acquisition, as it allows children to link words with specific things and understand the meaning of language through shared attention with a caregiver; essentially, it's the skill of looking at something while also acknowledging that another person is looking at the same thing, enabling effective communication and learning. ○ Infant pointing increases when experimenter responds to their points with new information (Wow! or Yuck!) Infant pointing decreases when experimenter responds to their points by establishing joint attention ○ your mom points at something, you make joint attention with her, you attend to this object together, she labels it for you. Parental report of vocabulary (CDI) ○ MacArthur Bates Communicative Development Inventories (CDI) Establishes norms in receptive and expressive/productive vocab for 0-3y ○ Comprehension tasks Peabody Picture Vocabulary Test (PPVT) Measures receptive vocab for 2.5-99y via point Looking while listening tasks (LWL) Measures receptive vocab under 2y via eye gaze CDI PPVT LWL Early bias for noun learning ○ Referential bias Also known as: Principle of reference  Words refer to things (words have meanings) Most basic & fundamental bit of knowledge we have about words Hard to see how to learn word meanings without this Recall: relating phonological form to word meaning Baldwin & Markman (1989) Specific formulation: nouns refer to objects (in the physical world) Finding: providing a new word as a label increases infants’ interest in an object. Alternative formulations: Verbs refer to actions or events (kick, swim, give, hug) Adjectives refer to properties of objects (yellow, fuzzy, striped, tall) ○ Overextension and underextension ○ Underextension Applying a word in fewer cases than you should (based on adult knowledge) Cup = infant’s specific sippy cup (and not any other cup) Doggie = the family dog (and not any other dog ○ Overextension Applying a word in more cases than you should (based on adult knowledge) Cup = cups + bowls when someone is drinking out of it Doggie = dogs + cats + other furry animals Based on shape, texture, color, natural kinds, functions III. Digging beneath the surface Measuring naturalistic productions vs. designing experimental manipulations ○ Development of pointing abilities ○ Through pointing, I can offer information. I can request information or I can tell people to do things. So by looking at pointing capacities in kids under 18 months, we'll start to understand a bit of what you saw in the reading are called proto speech checks. (So proto declarative and proto imperative and proto interrogatives) Types of communicative intentions; force vs. content of utterance ○ Force of an utterance Communicative effect of an utterance What the speaker’s intention or purpose is in making the utterance For example: declarative, interrogative, imperative ○ Content of an utterance What the utterance is about For example: about our dinner plans ○ Conventional ways of achieving different kinds of speech acts Declarative force associated with declarative clauses (statements) 10. I am making chicken for dinner. 11. We exercise every day.  Interrogative force associated with interrogative clauses (questions) 12. Are you making chicken for dinner? 13. Do you exercise everyday? Imperative force associated with imperative clauses (commands) 14. Make chicken for dinner! 15. Exercise every day! Nativism & Constructivism ○ Early 1900s: behaviorism Focus on studying observable behavior Denial of internal or mental processes Perspective on learning: Pavlovian conditioning, reinforcement learning B.F. Skinner’s (1957) Verbal Behavior ○ Starting in the 1950s: the cognitive revolution Study of the human mind, including cognitive abilities and mental processes Noam Chomsky’s (1957) Syntactic Structures & (1959) Review of B.F. Skinner’s Verbal Behavior Divisions with the modern field (Empiricists, Nativists & Constructivists) ○ Tensions: measurement vs. inference ○  ○ ○ Language acquisition device ○ A hypothesized set of mental structures and processes that allows the language learning child to learn language based on the input they receive ○ By hypothesis, the language acquisition device should allow any typically-developing child to learn any of the world’s languages ○ The LAD is typically depicted by a box and arrow model that distinguishes different stages of the process and how they are related to each other ○ Types of evidence: positive evidence, direct negative evidence, indirect negative evidence ○ Positive evidence  The learner needs to observe features that are grammatical when they are produced Universally accepted as evidence that is: Available to the learner Utilized by the learner ○ Direct negative evidence The learner needs to compare and conclude which features are ungrammatical when they are corrected Clearly available to the learner, but disagreement about whether it is actually utilized. ○ Indirect negative evidence The learner needs to reason about what features could be grammatical and infer that they are not if they do not get to observe them. ○ A lack of evidence (Poverty of the Stimulus) The learner cannot learn about feature because there’s no evidence available. They must already know it. Poverty of the Stimulus arguments  ○ ○ We know a fact about our language (as adults) It seems like we never had the opportunity to learn it from experience ○ Invitation for future research: Perhaps we haven’t considered the right kind of evidence (and constructivists will search for this kind of evidence) Otherwise, we have to conclude that we didn’t learn this fact about our language and we already knew it, as part of our innate gift (supporting the nativist approach) Kovács et al (2014) ○ Title: Pointing as Epistemic Request ○ Main Focus: The study investigates the role of pointing in infants as a means of seeking information, contrasting two conditions: Sharing and Informing. ○ Key Concepts: 1. **Pointing as an Epistemic Request**: Infants use pointing not just to share interest but also to request information about objects. 2. **Social Context**: The interaction between infants and responsive adults influences infants' knowledge acquisition. #### ○ Experimental Design: - **Participants**: 32 healthy, full-term, monolingual 12-month-olds. - **Conditions**: - **Sharing Condition**: The experimenter labels an object with a familiar word while sharing attention (e.g., calling a cat a "kitty"). - **Informing Condition**: The experimenter labels the same object with a novel word (e.g., calling the cat a "dax"), which may indicate unreliability in labeling. Procedure: - Infants were exposed to eight trials where they could point to a puppet displayed by the experimenter. - The experimenter interacted with the infant before the trials to ensure engagement. - Infants who did not point or were distracted were excluded from the analysis. ○ Findings: 1. **Pointing Behavior**: Infants pointed similarly in both conditions initially but showed increased pointing in the Informing condition in later trials, suggesting they preferred responses that provided new information. 2. **Expectations from Pointing**: Infants may expect to gain generalizable information about the referent, such as its kind, function, or properties. 3. **Valence of Responses**: The emotional response of the experimenter (positive or negative) influenced infants' subsequent pointing behavior. ○ Theoretical Implications: - The study supports the idea that infants are not just passive recipients of information but active seekers, using pointing as a tool to elicit knowledge. - It aligns with the theory of natural pedagogy, suggesting that infants are motivated to learn from their interactions. ○ Conclusion: The research highlights the dual role of pointing in infants as both a communicative act and a request for information, emphasizing the importance of the social context in knowledge acquisition. ○ IV. Phonetic and phonological development Categorical perception of phonemes and perceptual narrowing ○ ○ phonemic contrasts are a minimal difference in speech sounds or phonemes. ○ When something has a negative voice onset time or a voice onset time around zero, basically, it's a B in English. If it has a positive voice onset time. If the valve starts before the voicing comes in, then we perceive it as a p in English. ○ A minimal difference in speech sounds that has meaningful consequences in a language ○ Different languages have different sets of phonemic contrasts ○ Minimal pair in English: “bat” vs. “pat” ○ Voice Onset Time Voicing = vibration of the vocal chords Time between release of a stop and when voicing starts on a following vowel Figure from Traube & D’Alessdandro (2005)  ○ Adults respond to a continuum of sound as if it is divided into categories ○ Eimas et al. (1971) Infants noticed the change across the same category that adults did, but they did not notice changes within the categories this study was again using high amplitude sucking procedure, but there was a innovation that they made for the study off of what we've already seen. And that was that instead of just tracking the rate across time, and comparing the rate for listening to speech versus non speech, instead, what they did is they wanted to see within the speech stimulus, when infants recognize the difference between P and B. ○ But by the time that they are 12 months, they start to ignore distinctions ○ High-amplitude sucking procedure ○ Eimas et al. (1971) ○ Infants cannot hold up head until 3-6 months ○ Sucking reflex: Develops in the womb Triggered when roof of mouth is contacted ○ Measure understanding indirectly through physiological or neural functions ○ Vouloumanos & Werker (2007) High-amplitude sucking procedure Newborns (104 days old) listen to stimulus in 2 4-minute blocks ○ Boredom reflex: New things are more exciting Head-turn paradigm ○ And 3-6 months, infants can lift their head and control can hold their neck up. So we don't have to rely on sucking rate anymore. Their visual acuity is also better. So instead, we're going to rely on where they look, and in particular, how they turn their head as evidence of where they're looking. Habituation methods ○ Habituation is when the nervous system's response to a stimulus decreases with repeated exposure  ○ If any change matter, then you wouldn't see this de habituation sorry, you wouldn't see this continued habituation. You would see that both conditions would have this high level, change, they should recover interest. Werker & Tees (1984) ○ **Phonetic Categories and Discrimination**: The study focuses on infants' ability to discriminate between different phonetic categories, specifically alveolar stops like /k/ and /d/. - Infants were tested using a head turn (HT) paradigm, where they were conditioned to turn their heads in response to changes in speech sounds. ○ 2. **Stimulus Characteristics**: - Four exemplars of each sound were recorded, with variations in duration, fundamental frequency, and intonation randomized. - The average duration of a stimulus was 500 ms, with a 1500 ms interstimulus interval. - Key acoustic cues for differentiating sounds included amplitude of the burst and the slope of the second and third formant transitions. ○ 3. **Experimental Design**: - The study employed a longitudinal design to examine developmental changes in phonetic discrimination across different ages (6-8 months, 8-10 months, and 10-12 months). - Six subjects (3 males and 3 females) were tested at these ages to observe changes in their ability to discriminate nonnative contrasts. ○ 4. **Results**: - Most infants aged 6 to 8 months reached the criterion for discrimination, while fewer infants aged 10 to 12 months did. A significant decline in the ability to discriminate nonnative contrasts was observed as infants aged, particularly between 8-10 months and 10-12 months. ○ 5. **Comparison with Adults**: - Adult English speakers performed worse than both Adult-Thompson and Infant-English groups, indicating a decline in cross-language speech perception from infancy to adulthood. - The results suggest that specific linguistic experience is necessary to maintain phonetic discrimination ability, as infants lose this ability without such experience by 10 to 12 months of age. ○ 6. **Methodology**: - The HT paradigm involved infants turning their heads towards a loudspeaker when a change in speech sound was detected, with correct responses reinforced by a visual reward (toy animal). - The experimenters monitored the infants' responses and recorded data on their performance. ○ 7. **Acoustic Analysis**:  - The study included acoustic analyses to ensure that the consonant sounds in the stimuli were consistent and to identify differences between velar and uvular sounds. Infant-directed speech: characteristics and purpose ○ Speech that is addressed to infants, contrasted with adult-directed speech (ADS) ○ What differences do we notice between IDS and ADS? Higher pitch Repetitive Repeating words Repeating types of sentences Slower Pauses Simpler types of sentences Back-and-forth dynamic More emotion Facial expressions ○ Overall, IDS is more: Attention-grabbing Soothing Exaggerates speech properties (hyperarticulation for acoustic clarity) Reinforces exchange between speaker and infant (contingency) Adults’ production of IDS Automatic, upon addressing an infant Does not involve conscious intention ○ Hyperspeech hypothesis – Fernald (2000) IDS facilitates comprehension - maximizes predictability & provides perceptual support for infant intake ○ Why change pitch in IDS? Grab attention, convey emotion (perceptual support) ○ Changes to speech rate in IDS slower ○ Why change speed in IDS? Acoustic clarity More time to process less input ○ Changes to facial expressions in IDS More smiling Raised eyebrows Widened eyes ○ Why change facial expressions in IDS? Grab attention, convey emotion (perceptual support) Another cue to articulation ○ Changes in IDS: expanded F1 and F2 range vowels ○ Simpler structures A cluster of features lead each utterance to be more simple  Shorter units of speech More pauses between speech units Longer pauses between speech units Lots of repetition across speech units V. Prenatal language experience and related learning mechanisms Prenatal language exposure ○ Language exposure before birth Sound carries through the womb, but not all properties of speech are preserved ○ At birth Preference for human speech over other sounds ○ Between 0-6 months Ability to discriminate a wide range of segments or speech sounds ○ Between 6-12 months Specialization in discriminating segments that give rise to meaningful differences in native language ○ Hearing comes online in 3 rd trimester Womb filters out some aspects of speech but not all Under 400 Hz (Low-pass filtered speech) Perceptible to the fetus Relevant for prosodic properties of speech 500-1000 Hz Sometimes perceptible to the fetus 1000+ Hz Not perceptible to the fetus Relevant for phonemic distinctions Low-pass filtered speech ○ Under 400 Hz (Low-pass filtered speech) Perceptible to the fetus Relevant for prosodic properties of speech Prosodic aspects of speech vs. phonemic aspects of speech ○ prosody ○ Properties of larger units of speech: syllables (/ba/), phrases (“the little boy”) as opposed to single segments (/b/) ○ Rhythm and melody of a language Speech envelope ○ Pitch Duration Intensity Stress ○ Prosodic bootstrapping Prosody is available in typical prenatal experience Prosody is related to higher categories (words, syntactic constituents) Perhaps knowledge of prosody can help unlock those higher layers Word segmentation as a problem in language acquisition ○ Word segmentation The identification of word boundaries in a continuous speech stream  ○ An important step in the process of language acquisition ○ Word segmentation starts in the first year of life (6-12 months) ○ Cues that are used by infants to achieve word segmentation ○ Possible cues Spaces Pauses Words you already know Stress Languages can have dominant lexical stress patterns Lexical stress = stress within a word English is typically stress-initial in bisyllabic words ○ Typical pattern: strong-weak Transitional probabilities Regular statistical pattern within a language At 8 months, with no other cues to use, infants use transitional probability to distinguish “words” as familiar units from “non-words” as less familiar units. High transitional probability suggest the kind of predictable pattern we find inside of words instead of between them Phonological phrases Prosodic units are built out of words. If you find a prosodic break, you know that it must also be a word boundary Phrasal prosody ○ Rhythm and melody of speech Regular units with cross-linguistic similarities ○ Phrase-initial strengthening ○ Phrase-final lengthening ○ Other cues for word segmentation Stress patterns (used by 7 months) Kingdom vs. guitar Known words or transitional probabilities (used by 7 months) Biden, Vietnam, Mommy, the baby’s name Phonotactic constraints (used by 10 months) Languages differ in how sounds can combine: Dutch onset clusters: zwabber, vlug Allophonic cues (used by 10 months) Sounds can change based on the other sounds around them (coarticulation): night rate vs. nitrate Consider:  different cues at different times, multiple cues combine to solve the puzzle Why are some cues used later than other cues? What distinguishes allophonic & phonotactic cues from others? VI. Syntactic development and related learning mechanisms Prosodic cues to syntactic structure ○ Lengthening at end of syntactic constituents/units ○ For any word: it should be long at the end of a unit It should be shorter elsewhere Prosodic bootstrapping ○ Prosodic bootstrapping is a hypothesis that infants use prosodic cues in speech to learn about the grammar and lexicon of their native language. Prosodic features include pitch, rhythm, tempo, and amplitude. Knowledge of function words vs. content words ○ Newborns discriminate between function words and content words on the basis of their acoustic properties ○ Minimality: shorter, lack of stress, simpler syllable structure ○ Some function words are in dependencies with nouns (determiners) ○ Some function words are in dependencies with verbs (pronouns as subjects) ○ Do 14-month-old infants know the difference YES ○ Function words are shorter and have different stress than content words, this distinction is recognized by infants. 13 month olds can differentiate between function words they know and invented function words and are fascinated by the real function words they already know. 14 month olds can distinguish where function words go, such as words that go with nouns versus words that go with verbs. Infant’s knowledge of function words helps them to identify frequent frames and syntactic categories and thus supports syntactic development. Frequent frames ○ The semantic seed supports syntactic development by helping infants to identify frequent frames and then identify conceptual categories, thus supporting syntactic development. These early acquired concrete nouns allow language learners to pick up on patterns and connect the syntactic dots of their language. ○ Using function words to find frequent frames ○ proposed learning mechanism Doing distributional analysis in order to achieve grammatical categorization Distributional analysis = Tracking words and how they occur amongst other words Grammatical categorization = Deciding which syntactic category words belong to ○ Frequent frame = Two words that often occur together but have another word in between them  Local context for a (type of) word A ____x____ B ○ ○ Can be disrupted by modifiers and cross-linguistic issues like variable word order ○ Nevertheless, the general frequent frames strategy seems to work for English, French, Spanish, Turkish, ASL and other languages, with some nuances across each language. Semantic seed ○ Proposed learning mechanism Small group of words that are well-understood (via ostension) and their conceptual categories are known Distributional analysis suggests that a subset of these words occur in the same frames (plus knowledge that they have similar conceptual category) Infer that these verbs have the same syntactic category, and build that syntactic category around the conceptual one ○ Choice of participant population based on research question and practical constraints ○ sometimes it's expensive to use the methodology, whatever. In building up to testing that expensive or complex thing, we often start with baby steps in a way of using adult populations or older children, if we're interested in what's going on with infants as a proof of concept that this thing could work at an older age, or maybe it works at a younger age. Another thing we do is use computational models as to model how a computer would learn the language. Eye-tracking and preferential looking paradigms ○ The preferential looking paradigm is an experimental technique used to study how infants and toddlers understand and orient to visual and auditory stimuli. It involves comparing how long an infant looks at different stimuli, such as images  or videos. The paradigm is often used in combination with eye tracking to collect gaze data Babineau et al. (2019) ○ Babineau et al. (2019) Eye-tracking study 3-4 year old kids Training Either ko = determiner (ko apple) Or ko = pronoun (ko walks) Test What does nuve mean if we can say ko nuve? Method Introduce new determiner, paired with several known words Introduce a new type of action and then a new type of object Introduce a new word used with the determiner, and ask whether it refers to the action or object Introduce new determiner, paired with several known words (semantic seed) Introduce new determiner (ko), paired with several known words Introduce a new type of action (waving both arms) and then a new type of object (pink octopus) Introduce a new word (nuve) used with the determiner, and ask whether it refers to the action or object if participants learn that nuve refers to the pink octopus, then we have evidence for the role of the semantic seed in decoding frequent frames Discussion Why is there an overall preference for the event video? Can we still conclude that kids do well in the noun condition? Why did the researchers run this study with 3-4 year olds, if the semantic seed is a learning mechanism used in the range 12-18 months? How many steps are involved in the study? Could we skip any of these steps? Results - The study suggests that children can learn about syntactic categories and word meanings through distributional analysis of language, highlighting the importance of context in language acquisition. - It also indicates that children's ability to pay attention to function words is a significant factor in their vocabulary development. - Children demonstrated the ability to use the context provided by the novel function word "ko" to infer the meanings of novel words.  - They were able to map novel nouns to objects and novel verbs to actions based on their exposure to the syntactic structure during the familiarization phase. ○ Prosodic bootstrapping + frequent frames + semantic seed to build syntactic knowledge Prosodic cues (minimality) help the learner identify function words Frequent frames (frequency + locality) help the learner to track the relationship between function words and content words The semantic seed (ostension) gives the learner a strong foundation to determine the link between word meaning and syntactic category By the end of this process, the learner should have good knowledge of local relationships in syntax VII. Verb learning Syntactic bootstrapping ○ Syntactic bootstrapping is a theory in developmental psycholinguistics that explains how children learn word meanings by analyzing the structure of language and recognizing syntactic categories. It's based on the idea that children use their knowledge of syntax to understand the meaning of sentences and verbs ○ Refinement of the linguistic evidence hypothesis ○ Developmental observation: some words are more difficult than others to learn (because they are learned later in development) ○ Descriptive observation: words that have similar meaning occur in similar kinds of sentences ○ Theoretical hypothesis: there is an underlying link between the meanings of sentences and their syntactic structures ○ Learning hypothesis: knowledge of syntactic structure helps the language learning child to solve the mapping problem for new words Human Simulation Paradigm ○ An existing tradition of using computational models to simulate learning Why not use a human model to simulate learning? ○ In this case: How to model a greater or lesser conceptual repertoire? How to model greater or lesser access to linguistic input? Challenges associated with verb learning ○ Linguistic experience hypothesis ○ Related to the perceptual experience hypothesis: Verbs do not have reliable physical correlates the way a noun might There are many perspectives to take on an event (many different kinds of relations coexist in them) ○ What would the event version of the Whole Object Bias be? Conceptual complexity vs. linguistic experience hypotheses ○ Word learning is dependent on conceptual development Action/event concepts are more complex than object concepts  Object concepts have to develop before action/event concepts, as a result: object concepts are more available for word learning at early stages ○ Conceptual development is not the only thing that holds word learning back! Building syntactic structure is also important. Verbs require local dependencies (a subject and/or object) whereas nouns are more independent The learner has to break into the syntactic structure of their language before they can start to decode verbs Also, sentences provide evidence about words Gillette et al. (1999) ○ Human Simulation Paradigm Participants: undergraduates at UPenn Adults have undergone conceptual development, so we know they will have no difficulty on the conceptual side Task: learn the meaning of a new word, on the basis of differing kinds of evidence We can limit linguistic evidence entirely (silent videos) We can limit evidence of the physical world (sentences w/ no video) We can mix and match, provide some but not all language… ○ Research Questions (RQs) 1. **How do children identify verbs and nouns in language acquisition?** - The study investigates the mechanisms through which children learn to associate verbs and nouns with their meanings, particularly focusing on the role of contextual information and co-occurrence of words. 2. **What is the effect of different types of information (e.g., noun co-occurrence, video context) on verb identification success?** - The research aims to determine how various forms of contextual information influence the ability of subjects (likely children) to correctly identify verbs in relation to nouns. 3. **How does performance in verb identification change across trials?** - The study examines whether subjects improve their identification of verbs and nouns as they are exposed to more examples over time. ○ Methods - **Participants:** The study involved subjects who received credit in a psychology course or were paid for their participation. The specific demographic details of the participants are not provided in the excerpt. - ○ **Design:** A repeated measures ANOVA was used to analyze the data across 48 items, with no significant within-item order effects found. The study also controlled for corpus frequency, indicating that the number of times each test word appeared in maternal samples did not significantly affect the results. - Experiment 1 Input for word learning GORP/DAX/BLICK Silent scene No linguistic input Are there going to be times that this works perfectly? Are there going to be times that this fails miserably?  Experiment 3 6 conditions that vary in what can serve as input for learning the word Information from the physical world/scene Information from specific words (lexical bootstrapping) Information from syntactic structures (syntactic bootstrapping). Condition 1 = same as Experiment 1 Silent scene No linguistic input Condition 2 No scene Only a list of nouns that occur in the same sentence as the word What will work and what won’t? Condition 3 A combo of Cond 1 & Cond 2 Condition 4 No scene Only “jabberwocky” syntax Keep real function words Invent content words Condition 5 No scene Full sentence except verb Condition 6 Full information Scene + real sentence ○ **Scoring:** Responses were scored as correct only if they were morphemically identical to the target word, although some leniency was applied in scoring to account for synonyms and related terms. A more liberal scoring method was also evaluated, where responses reflecting knowledge of the correct referent were considered correct. ○ Results - ○ **Overall Performance:** The results indicated that verb identification was generally low, with only 15% success in one experiment and less than 8% in another condition. However, performance improved significantly across trials, suggesting that subjects were benefiting from accumulating evidence. - ○ **Effect of Information Types:** The study found that noun co-occurrence information was surprisingly useful for verb identification, achieving a mean correct response rate of 16.5%. This suggests that even without knowing the structural position of nouns, subjects could make inferences about the verbs. - ○ **Statistical Analysis:** A repeated measures ANOVA demonstrated significant improvements in performance across trials (F6:74, df5;230, P