Lecture2-MechanismAndMethods

Lecture2-MechanismAndMethods - Psych 215L: Language...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Psych 215L: Language Acquisition Lecture 2 The Mechanism of Acquisition and Some Child Language Research Methods Describing vs. Explaining “…it gradually became clear that something important was missing that was not present in either of the disciplines of neurophysiology or psychophysics. The key observation is that neurophysiology and psychophysics have as their business to describe the behavior of cells or of subjects but not to explain such behavior….What are the problems in doing it that need explaining, and what level of description should such explanations be sought?” - Marr (1982) Levels of Representation Marr (1982) On Explaining (Marr 1982) “…[need] a clear understanding of what is to be computed, how it is to be done, the physical assumptions on which the method is based, and some kind of analysis of the algorithms that are capable of carrying it out.” “This was what was missing - the analysis of the problem as an information-processing task. Such analysis does not usurp an understanding at the other levels - of neurons or of computer programs - but it is a necessary complement to them, since without it there can be no real understanding of the function of all those neurons.” On Explaining (Marr 1982) On Explaining (Marr 1982) “But the important point is that if the notion of different types of understanding is taken very seriously, it allows the study of the information-processing basis of perception to be made rigorous. It becomes possible, by separating explanations into different levels, to make explicit statements about what is being computed and why and to construct theories stating that what is being computed is optimal in some sense or is guaranteed to function correctly. The ad hoc element is removed…” “But the important point is that if the notion of different types of understanding is taken very seriously, it allows the study of the information-processing basis of perception to be made rigorous. It becomes possible, by separating explanations into different levels, to make explicit statements about what is being computed and why and to construct theories stating that what is being computed is optimal in some sense or is guaranteed to function correctly. The ad hoc element is removed…” Our goal: Substitute “language acquisition” for “perception”. The three levels Computational What is the goal of the computation? What is the logic of the strategy by which it can be carried out? Algorithmic How can this computational theory be implemented in a procedure? What is the representation for the input and output, and what is the algorithm for the transformation? Implementational How can the representation and algorithm be realized physically? The three levels: An example with the cash register Computational What does this device do? Arithmetic (ex: addition). Addition: Mapping a pair of numbers to another number. (3,4) 7 (often written (3+4=7)) Properties: (3+4) = (4+3) [commutative], (3+4)+5 = 3+(4+5) [associative], (3+0) = 3 [identity element], (3+ -3) = 0 [inverse element] True no matter how numbers are represented: this is what is being computed The three levels: An example with the cash register Computational What does this device do? Arithmetic (ex: addition). Addition: Mapping a pair of numbers to another number. Algorithmic What is the input, output, and method of transformation? Input: arabic numerals (0,1,2,3,4…) Output: arabic numerals (0,1,2,3,4…) Method of transformation: rules of addition, where least significant digits are added first and sums over 9 have their next digit carried over to the next column 99 + 5 The three levels: An example with the cash register Computational What does this device do? Arithmetic (ex: addition). Addition: Mapping a pair of numbers to another number. Algorithmic What is the input, output, and method of transformation? Input: arabic numerals (0,1,2,3,4…) Output: arabic numerals (0,1,2,3,4…) Method of transformation: rules of addition, where least significant digits are added first and sums over 9 have their next digit carried over to the next column 1 99 + 5 4 The three levels: An example with the cash register Computational What does this device do? Arithmetic (ex: addition). Addition: Mapping a pair of numbers to another number. Algorithmic What is the input, output, and method of transformation? Input: arabic numerals (0,1,2,3,4…) Output: arabic numerals (0,1,2,3,4…) Method of transformation: rules of addition, where least significant digits are added first and sums over 9 have their next digit carried over to the next column 99 + 5 14 The three levels: An example with the cash register Computational What does this device do? Arithmetic (ex: addition). Addition: Mapping a pair of numbers to another number. Algorithmic What is the input, output, and method of transformation? Input: arabic numerals (0,1,2,3,4…) Output: arabic numerals (0,1,2,3,4…) Method of transformation: rules of addition, where least significant digits are added first and sums over 9 have their next digit carried over to the next column 1 99 + 5 104 The three levels: An example with the cash register Computational What does this device do? Arithmetic (ex: addition). Addition: Mapping a pair of numbers to another number. The three levels Marr (1982) Algorithmic What is the input, output, and method of transformation? Input: arabic numerals (0,1,2,3,4…) Output: arabic numerals (0,1,2,3,4…) Method of transformation: rules of addition Implementational How can the representation and algorithm be realized physically? A series of electrical and mechanical components inside the cash register. Mapping the Framework: Algorithmic Theory of Language Learning “Although algorithms and mechanisms are empirically more accessible, it is the top level, the level of computational theory, which is critically important from an information-processing point of view. The reason for this is that the nature of the computations that underlie perception depends more upon the computational problems that have to be solved than upon the particular hardware in which their solutions are implemented. To phrase the matter another way, an algorithm is likely to be understood more readily by understanding the nature of the problem being solved than by examining the mechanism (and the hardware) in which it is embodied.” Mapping the Framework: Algorithmic Theory of Language Learning Goal: Understanding the “how” of language learning Goal: Understanding the “how” of language learning First, we need a computational-level description of the learning problem. First, we need a computational-level description of the learning problem. Computational Problem: Divide sounds into contrastive categories (Speech perception, phoneme identification) Computational Problem: Divide spoken speech into words (Word segmentation) hu@wz´f®e@jd´vD´bI@gbQ@dw´@lf x x x x x x x x x x x x x x x x x x x x x x x x x x x C1 x x x x x x x C4 x x x x x x C2 x C3 x x x x x hu@wz ´f®e@jd ´v D´ bI@g bQ@d w´@lf who‘s afraid of the big bad wolf Mapping the Framework: Algorithmic Theory of Language Learning Mapping the Framework: Algorithmic Theory of Language Learning Goal: Understanding the “how” of language learning Goal: Understanding the “how” of language learning First, we need a computational-level description of the learning problem. First, we need a computational-level description of the learning problem. Computational Problem: Identify word classes that behave similarly (Grammatical categorization) Computational Problem: Identify the concept a word is associated with (Word-meaning mapping) “This is a DAX.” “I love my daxes.” Dax = that specific toy, teddy bear, stuffed animal, toy, object, …? DAX = noun Mapping the Framework: Algorithmic Theory of Language Learning Mapping the Framework: Algorithmic Theory of Language Learning Goal: Understanding the “how” of language learning Goal: Understanding the “how” of language learning First, we need a computational-level description of the learning problem. Second, we need to be able to identify the algorithmic-level description: Computational Problem: Identify the rules of word order for sentences. (Syntax: grammatical rules of the language) Jareth juggles crystals Subject Verb Object German Kannada Subject tObject Verb Object Subject Verb t Subject English Subject Verb Object Object tVerb Input = sounds, syllables, words, phrases, … Output = sound categories, words, grammatical categories, sentences, … Method = statistical learning, algebraic learning, prior knowledge about how human languages work, … Framework for language learning (algorithmic-level) What are the hypotheses available (for generating the output from the input)? Ex: general word order patterns Input: words (adjective and noun) Output: ordered pair Adjective before noun (ex: English) red apple Noun before adjective (ex: Spanish) manzana roja apple red Framework for language learning (algorithmic-level) What are the hypotheses available (for generating the output from the input)? Ex: general word order patterns What data are available, and should the learner use all of them? Ex: exceptions to general word order patterns Ignore special use of adjective before noun in Spanish Special use: If the adjective is naturally associated with the noun: la blanca nieve the white snow Why not usual order? Snow is naturally white. Framework for language learning (algorithmic-level) What are the hypotheses available (for generating the output from the input)? Ex: general word order patterns What data are available, and should the learner use all of them? Ex: exceptions to general word order patterns How will the learner update beliefs in the competing hypotheses? Ex: shifting belief in what the regular word order of adjectives and nouns should be This usually will involve some kind of probabilistic updating function. Experimental Methods: What, When, and Where Another useful indirect measurement Head Turn Preference Procedure Another useful indirect measurement Head Turn Preference Procedure Infant sits on caretaker’s lap. The wall in front of the infant has a green light mounted in the center of it. The walls on the sides of the infant have red lights mounted in the center of them, and there are speakers hidden behind the red lights. Another useful indirect measurement Sounds are played from the two speakers mounted at eye-level to the left and right of the infant. The sounds start when the infant looks towards the blinking side light, and end when the infant looks away for more than two seconds. Note on infant attention: Familiarity vs. Novelty Effects Head Turn Preference Procedure Thus, the infant essentially controls how long he or she hears the sounds. Differential preference for one type of sound over the other is used as evidence that infants can detect a difference between the types of sounds. For procedures that involve measuring where children prefer to look (such as head turn preference), sometimes children seem to have a “familiarity preference” where they prefer to look at something similar to what they habituated to. Other times, children seem to have a “novelty” preference where they prefer to look at something different to what they habituated to. Kidd, Piantadosi, & Aslin (2010) provide some evidence that this may have to do with the informational content of the test stimulus. There may be a “Goldilocks” effect where children prefer to look at stimuli that are neither to boring nor too surprising, but are instead “just right” for learning, given the child’s current knowledge state. Computational Methods Why use computational modeling? Computational Methods: How “Given a model of some aspect of language acquisition, implementing it as a computational system and evaluating it on naturally occurring corpora has a number of compelling advantages. First of all by implementing the system, we can be sure that the algorithm is fully specified, and the acquisition model does not resort to hand-waving at crucial points. Secondly, by evaluating it on real linguistic data, we can see whether naturally occurring distributions of examples in corpora provide sufficient information to support the studied claims across a divergent range of acquisition theories. Thirdly, study of the system can identify the mechanisms that cause changes in the algorithm’s hypotheses during the course of acquisition. Finally, the computational resources required of the model can be concretely assessed and (not so concretely) compared against the resources that might be available to a human language learner.” - Clark & Sakas 2011 Computational Methods Control over the entire learning mechanism: - what hypotheses the (digital) child considers - what data the child learns from - how the child updates beliefs in different hypotheses Ground with empirical data available - want to make this as realistic as possible (ex: use actual data distributions, cognitively plausible update procedures) - a good source of empirical data: CHILDES database http://childes.psy.cmu.edu/ Download annotated transcripts from the database. Download the program to search these transcripts, and its manual. Back to modeling Sample learning models Gauges of modeling success & contributions to science Formal sufficiency: does the model learn what it’s supposed to learn when it’s supposed to learn it from the data it’s supposed to learn it from? Developmental compatibility: Does it learn in a psychologically plausible way? Is this something children could feasibly do? Explanatory power: what’s the crucial part of the model that makes it work? How does this impact the larger language acquisition story? Sample learning models Morphology (Rumelhart & McClelland 1986, Yang 2002, Albright & Hayes 2002, Yang 2005, Chan & Lignos 2011): learning to identify word affixes from segmented speech Learning the interpretation of referential elements (Regier & Gahl 2004, Foraker et al. 2007, 2009, Pearl & Lidz 2009, Pearl & Mis 2011): learning to identify syntactic category and semantic referent of one from segmented speech and referents in the world Syntactic acquisition (Yang 2004, Reali & Christiansen 2005, Kam et al. 2008, Pearl & Weinberg 2007, Perfors, Tenenbaum, & Regier 2011, Pearl & Sprouse 2011): learning to identify correct word order (rules) from speech segmented into words Stress (Pearl 2008, Pearl 2011, Legate & Yang 2011 forthcoming): learning to identify correct stress patterns (and rules behind them) from words with stress contours Phoneme acquisition (Vallabha et al . 2007, Feldman, Griffiths, & Morgan 2009, Dillon et al. 2011 Ms., Feldman et al. 2011): learning contrastive sounds from acoustic data Word segmentation (Swingley 2005, Gambell & Yang 2006, Goldwater et al. 2009, Johnson & Goldwater 2009, Blanchard et al. 2010, Jones et al. 2010, Pearl et al. 2011): learning to identify words in fluent speech from streams of syllables Categorization (Mintz 2003, Wang & Mintz 2008, Chemla et al. 2009, Liebbrandt & Powers 2010): learning to identify what category a word is (noun, verb) from segmented speech General Modeling Process (1) Decide what kind of learner the model represents (ex: normally developing 6-month-old child learning first language) (2) Decide what data the child learns from (ex: Bernstein corpus from CHILDES) and how the child processes that data (ex: data divided into syllables) (3) Decide what hypotheses the child has (ex: what the words are) and what information is being tracked in the input (ex: transitional probability between syllables) (4) Decide how belief in different hypotheses is updated (ex: based on transitional probability minima between syllables) General Modeling Process (5) Decide what the measure of success is - precision and recall (ex: finding the right words in a word segmentation task) - matching an observed performance trajectory (ex: English past tense acquisition often has a U-shaped curve) - achieving a certain knowledge state by the end of the learning period (ex: knowing there are 4 vowel categories at the end of a phoneme identification task) - making correct generalizations (ex: preferring a correctly formed sentence over an incorrectly formed one) Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research Saffran, Aslin, & Newport (1996): groundbreaking study showing experimental support for infant ability to track statistical probability between syllables when trying to segment words from fluent speech. (See Romberg & Saffran (2010) for a review of infant statistical learning abilities.) Saffran et al. proposed that some aspects of acquisition were “best characterized as resulting from innately biased statistical learning mechanisms rather than innate knowledge”. Denison et al. (2011): evidence for probabilistic reasoning abilities in 6-month-olds Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research Statistical learning long considered part of acquisition process (Chomsky 1955, Hayes & Clark 1970, Wolff 1977, Pinker 1984, Goodsitt, Morgan, & Kuhl 1993, among others), but traditionally viewed as playing secondary role rather than primary one. Why? Children were not believed to be capable of tracking statistical information in language input to the extent that they would need to for learning linguistic knowledge (Chomsky 1981, Fodor 1983, Bickerton 1984, Gleitman and Newport (1995), among others). Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research Question 1: What kinds of statistical patterns are human language learners sensitive to? Thiessen & Saffran (2003): 7-month-olds prefer syllable transitional probability cues over language-specific stress cues when segmenting words, while 9-month-olds show the reverse preference. Graf Estes, Evans, Alibali, & Saffran (2007): word-like units that are segmented using transitional probability are viewed by 17month-olds as better candidates for labels of objects. Thompson & Newport (2007): adults can use transitional probability between grammatical categories to identify word sequences that are in the same phrase, a precursor to more complex syntactic knowledge. Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research Question 1: What kinds of statistical patterns are human language learners sensitive to? Other statistics involving relationships of adjacent units: backward transitional probability (Perruchet & Desaulty 2008, Pelucchi, Hay, & Saffran 2009b) and mutual information (Swingley 2005). Non-adjacent dependencies: Newport & Aslin (2004): non-adjacent statistical dependencies between consonants and between vowels, but not between entire syllables Mintz (2002, 2003, 2006): frequent frames used to categorize words. (ex: the___one is a frame that could occur with big, other, pretty, etc.). Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research Question 2: To what extent are these statistical learning abilities specific to the domain of language, or even to humans? Not specific to language: Saffran et al. (1999): both infants and adults can segment nonlinguistic auditory sequences (musical tones) based on the same kind of transitional probability cues that were used in the original syllable-based studies. Similar results have been obtained in the visual domain using both temporally ordered sequences of stimuli (Kirkham et al., 2002) and spatially organized visual “scenes” (Fiser and Aslin, 2002). Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research Question 1: What kinds of statistical patterns are human language learners sensitive to? More sophisticated statistics/inferences: Yu & Smith (2007) and Smith & Yu (2008): Both adults and 12- to 14-month-old infants can track probabilities of word-meaning associations across multiple trials where any specific word within a given trial was ambiguous as to its meaning. Xu & Tenenbaum (2007): investigated how humans learn the appropriate set of referents for basic (cat), subordinate (tabby), and superordinate (animal) words. Both adults and children between the ages of 3 and 5 are capable of integrating the likelihood of an event occurring into their internal models of word-meaning mapping in a way easily predicted by standard Bayesian inference techniques. Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research Question 2: To what extent are these statistical learning abilities specific to the domain of language, or even to humans? Not specific to humans: Hauser et al. (2001): cotton-top tamarins can segment the same kind of artificial speech stimuli used in the original Saffran et al. (1996) segmentation experiments as well as human infants. Saffran et al. (2008): tamarins could also learn some simple grammatical structures based on statistical information, but were unable to learn patterns as complex as those learned by infants. Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research Question 3: What kinds of knowledge can be learned from the statistical information available? Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research The Bayesian approach Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research The Bayesian approach offers a concrete way to examine what knowledge is required for acquisition, and whether that required knowledge is domainspecific or domain-general, without committing to either view a priori . - has led to the investigation of a new set of questions that previous approaches have not considered: whether human language learners can be viewed as being optimal statistical learners (i.e., making optimal use of the statistical information in the data), and in what situations. - Something more easily investigated through computational modeling studies rather than traditional experimental techniques. - can potentially address the question of why they make the generalizations they do, i.e., because these generalizations are statistically optimal given the available data and any learning biases, innate or otherwise. Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research The Bayesian approach - - Also, may be different ways to approximate Bayesian inference that are not so resource-intensive. Bonawitz, Denison, Chen, Gopnik, & Griffiths (2011) discuss a simple sequential algorithm called Win-Stay, Lose-Shift that matches human behavior consistent with Bayesian inference. Makes the space of hypotheses considered by the language learner explicit (doesn’t matter whether they are based on domain-specific or domain-general cognitive constraints) - Encodes the learner's biases by assigning an explicit probability distribution over these hypotheses. - Can operate over the kinds of highly structured representations that many linguists believe are correct (e.g., Regier & Gahl 2004, Foraker et al. 2009, Pearl & Lidz 2009, Pearl & Mis 2011, Perfors et al. 2011). Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research The Bayesian approach likelihood of observed data The Bayesian approach prior belief in hypothesis P(hypothesis | data) = P(data | hypothesis) * P(hypothesis) P(data) posterior likelihood of hypothesis likelihood of data period, no matter what hypothesis “The product of priors and likelihoods often has an intuitive interpretation in terms of balancing between a general sense of plausibility based on background knowledge and the data-driven sense of a “suspicious coincidence.” In other words, it captures the tradeoff between the complexity of an explanation and how well it fits the observed data.” – Perfors et al. 2011, Bayesian tutorial Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research The Bayesian approach Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research Generative framework: observed data are assumed to be generated by some underlying process or mechanism explaining why the data occurs in the patterns it does. Ex: words in a language may be generated by a grammar Bayesian learner evaluates different hypotheses about the underlying nature of the generative process, and makes predictions based on the most likely ones. Probabilistic model = a specification of the generative processes at work, identifying the steps (and associated probabilities) involved in generating data. Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research The Bayesian approach Usual three steps of a Bayesian model: 1) Define hypothesis space – which hypotheses are under consideration? 2) Define prior distribution over hypotheses – which are more/less likely? 3) Define likelihood update – how does data affect learner’s belief? From Perfors et al. 2011, Bayesian Tutorial Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research The Bayesian approach Hypothesis space can contain multiple levels of representation – shows power of bootstrapping (using preliminary or uncertain information in one part of the grammar to help constrain learning in another part of the grammar, and vice versa) Goldwater et al. (2006, 2009): two levels of representation -- words and phonemes -- though only one of these (words) is unobserved in the input and must be learned. Johnson (2008): learning both syllable structure and words from unsegmented phonemic input improved word segmentation in a Bayesian model similar to that of Goldwater et al. Feldman et al. (2009): simultaneously learning phonetic categories and the lexical items containing those categories led to more successful categorization than learning phonetic categories alone. Yuan et al. (2011): simultaneously learning individual word meaning and more abstract features involved in word meaning Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research The Bayesian approach Note: intended to provide a declarative description of what is being learned, not necessarily how the learning is implemented. Instead: only assume that the human mind implements some type of algorithm (perhaps a very heuristic one) that is able to approximately identify the posterior distribution over hypotheses. Some studies looking at how Bayesian inference might be implemented: - Pearl, Goldwater, and Steyvers (2010, 2011): implementing Bayesian inference in constrained learners with limitations on memory and processing - Shi, Griffiths, Feldman, & Sanborn (2010): exemplar models may provide a possible mechanism for implementing Bayesian inference, and have identifiable neural correlates. Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research The Bayesian approach A note on hierarchical Bayesian models: Allow generalizations at multiple levels. (Dewar & Xu (2010): 9-month-olds can do this.) Learner uses observable data to learn about properties of bags in general (ex: uniform vs. mixed distribution), not just properties of individual bags. Analogy: bags = language properties From Kemp, Perfors, & Tenenbaum (2007) Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research The Bayesian approach A main contribution: provide a way to formally evaluate claims about children’s hypothesis space. - Can indicate if certain constraints or restrictions are required in order to learn some aspect of linguistic knowledge (e.g., Regier & Gahl 2004, Perfors, Tenenbaum, & Regier 2011, Foraker et al. 2009, Pearl & Lidz 2009, Pearl & Mis 2011, Perfors et al. 2011). - If a Bayesian learner looking for the optimal hypothesis given the data cannot converge on the correct hypothesis, this suggests that the current conception of the hypothesis space cannot be correct. Required knowledge may take the form of an additional constraint on the hypothesis space that gives preference to certain hypotheses over others, or eliminates some hypotheses entirely. Statistical Learning, Inductive Bias, & Bayesian Inference in Language Acquisition Research The Bayesian approach in many different linguistic domains - Phonetics & perceptual learning: Feldman, Griffiths, & Morgan 2009, Feldman et al. 2011, Dillon et al. 2011 - Word segmentation: Goldwater, Griffiths, & Johnson 2009, Johnson & Goldwater 2009, Pearl, Goldwater, & Steyvers 2010, 2011 - Word-meaning mapping: Xu & Tenenbaum 2007, Frank, Goodman, & Tenenbaum 2009 - Syntax-semantics mapping: Regier & Gahl 2004, Pearl & Lidz 2009, Foraker, Regier, Khetarpal, Perfors, & Tenenbaum 2009, Pearl & Lidz 2011 - Syntactic structure: Perfors, Tenenbaum, Gibson, & Regier 2010, Perfors, Tenenbaum, & Regier 2011 Experimental Methods How do we tell what infants know, or use, or are sensitive to? Extra slides Experimental Methods How do we tell what infants know, or use, or are sensitive to? Researchers use indirect measurement techniques. Researchers use indirect measurement techniques. High Amplitude Sucking (HAS) High Amplitude Sucking (HAS) Infants are awake and in a quietly alert state. They are placed in a comfortable reclined chair and offered a sterilized pacifier that is connected to a pressure transducer and a computer via a piece of rubber tubing. Once the infant has begun sucking, the computer measures the infant’s average sucking amplitude (strength of the sucks). A sound is presented to the infant every time a strong or “high amplitude” suck occurs. Infants quickly learn that their sucking controls the sounds, and they will suck more strongly and more often to hear sounds they like the most. The sucking rate can also be measured to see if an infant notices when new sounds are played. Experimental Methods Experimental Methods How do we tell what infants know, or use, or are sensitive to? How do we tell what infants know, or use, or are sensitive to? Researchers use indirect measurement techniques. High Amplitude Sucking (HAS) Test Condition 1 Test Condition 2 Researchers use indirect measurement techniques. Control (baseline) High Amplitude Sucking (HAS) Test Condition 1 Test Condition 2 Control (baseline) Difference when compared to baseline Experimental Methods Experimental Methods How do we tell what infants know, or use, or are sensitive to? How do we tell what infants know, or use, or are sensitive to? Researchers use indirect measurement techniques. High Amplitude Sucking (HAS) Test Condition 1 Test Condition 2 Researchers use indirect measurement techniques. Control (baseline) High Amplitude Sucking (HAS) Infants have sophisticated discrimination abilities, but they don’t abstract sounds into categories the way that adults do. Adult perception “tQ” “dQ” No difference phonemic category phonemic category Experimental Methods Eyetracking: measures fixations on target picture “Where’s the baby?” How do we tell what infants know, or use, or are sensitive to? Researchers use indirect measurement techniques. High Amplitude Sucking (HAS) Infants have sophisticated discrimination abilities, but they don’t abstract sounds into categories the way that adults do. Infant perception “tQ 1” “dQ 2” “tQ 2” “dQ 1” Eyetracking: measures fixations on target picture Looking at children’s brains “Where’s the baby?” ERPs: Event-related brain potentials, gauged via electrode caps. The location of ERPs associated with different mental activities is taken as a clue to the area of the brain responsible for those activities. Good: non-invasive, relatively undemanding on the subject, provide precise timing on brain events “Where’s the baby? “Where’s the vaby? Bad: poor information on exact location of ERP since just monitoring the scalp Looking at children’s brains Looking at children’s brains Brain-imaging techniques: gauge what part of the brain is active as subjects perform certain tasks Brain-imaging techniques: gauge what part of the brain is active as subjects perform certain tasks PET scans: Positron emission topography scans - subjects inhale low-level radioactive gas or injected with glucose tagged with radioactive substance - experimenters can see which parts of the brain are using more glucose (requiring the most energy) MEG: Magnetoencephalography - subjects have to be very still - experimenters can see which parts of the brain are active fMRI scans: functional magnetic resonance imaging - subjects have to be very still inside MRI machine, which is expensive to operate - experimenters can see which parts of the brain are getting more blood flow or consuming more oxygen Looking at children’s brains Brain-imaging techniques: gauge what part of the brain is active as subjects perform certain tasks Optical Topography: Near-infrared spectroscopy (NIRS) - transmission of light through the tissues of the brain is affected by hemoglobin concentration changes, which can be detected ...
View Full Document

Ask a homework question - tutors are online