Psych 240 Exam 3
Statistical language learning (Saffran et al., 1996)
-2 minutes: continuous syllable stream of repeated or non repeated sequences (no gaps) -later light that if looked at blinks, play different sounds -look longer for sound sequences they didn't hear in 2 minutes -starting to sort out syllable sequences
Definition/structure of a problem
-A Problem consists of some initial state in which a person begins and a goal state that is to be attained, plus a non-obvious way of getting from the first to the second. -initial state-> methods-> goal state -often like a maze- moving through possible paths trying to get to goal state
Normative inductive reasoning: Base rate and current evidence
-A key insight is that not only do you have to consider the current evidence (the voices, gums, etc.) You also normatively need to consider the base rate - the overall likelihood of the hypothesis being true. -Different hypotheses have different likelihoods
Critical Period Effects
-Age of Acquisition and Proficiency r≈-0.7) -People who learn language after age 10-12 hardly ever acquire native ability People raised in social isolation 2nd language learners American sign language
Dual route reading model: nonlexical route (grapheme-to-phoneme conversion), lexical route (for irregular words)
-Alphabetic reading is mapping from graphemes to phonemes (the sounds in word). -So we need to know the sounds in words -letter to sound conversion if reading the right way -if english regular language you could pronounce every word this way -lexical and non lexical route= lexical more difficult (overwrite and beat non lexical)
Heuristics: Difference reduction
-At any point, select the operator that moves you closer to the goal state: is new state more similar to goal? (never choose an operator that moves you away) -many directions you could take -if I need to get to California what should I do?= head west? Maybe not best solution- doesn't solve problem -always choose operator that gets you closer to goal state
Bayes theorem--understand the 3 pieces of information you need to determine p(H/E): Base rate or p(H), p(E/H), p(E/not H)
-Bayes theorem tells you the right way, the normative way, of doing probabilistic reasoning -p(H)=overall probability that hypothesis is true= BASE RATE -H=hypothesis -E=evidence -p(H/E)= p(breast cancer/ positive mammogram) o Current Evidence -1. P(E / H): probability of current evidence if hypothesis is true -2. P(E / H): probability of current evidence if hypothesis is false o Base Rate -3. P(H): probability of hypothesis independent of current evidence
Emphasis on similarity
-Judge whether A has some characteristic by relying on the similarity of A to other things with that characteristic; e.g., judge whether your patient is schizophrenic by relying on how similar he or she is to schizophrenics. -Problem: ignores base rates
Bilingualism and SLA
-Bilingualism: having the ability to speak 2 languages. ¥ L1: first language learned. ¥ L2: second language learned. -Simultaneous bilingualism: L1 & L2 learned about the same time (= Bilingual 1st Language Acquisition) -Late bilingualism: Learned in adolescence +. -cognitive control enhanced in bilinguals -bilingual= inhibiting one whole language at a time; exercising cognitive control -ability to speak multiple languages might affect cognitive control (central executive smarts from knowing 2 languages) -may help slow down dementia -Second Language Acquisition (SLA) ¥ people learn a language other than their mother tongue, inside or outside of a classroom ¥ L2s any language other than the learner's 'native language' or 'mother tongue'. They encompass both languages of wider communication, encountered within the local region or community (e.g. at the workplace, or in the media), and truly foreign languages (FLA), which have no immediately local uses or speakers.
Skill learning: Three stages
-Cognitive (declarative, attention, teaching): ¥ Declarative knowledge. ¥ Commit facts to memory. ¥ Rehearse as you try to perform. ¥ Requires attention--can't do second task. ¥ May be independent of skill: the best teacher may not be the most skillful, but rather someone who knows how to describe the cognitive stage well -Associative (strengthen connections, eliminate errors): ¥Strengthen connections that lead to desired result -Feedback is important: See which actions lead to desired result ¥Get rid of actions that lead to errors -Autonomous/Automatic: fast, less attention, less verbalization, feedback less important, need for consistent practice: ¥ Fast ¥ Executed with less attention/consciousness ¥ Less verbalization -Less dependent on verbalization -Declarative knowledge less available ¥ Feedback -Less important -Lower level (proprioceptive not visual) ¥ Need for consistent practice -where a stimulus always gets the same response -Example: Golf -learn the basics by verbal instruction -practice components and see which ones work -components are strung together continued practice produces automatic stroke production
Two kinds of reasoning tasks: Deterministic/Deductive vs. probabilistic/inductive
-Deterministic/Deductive vs. probabilistic/inductive o Deterministic/deductive: trying to determine conclusions that must necessarily follow from premises - IF A implies B and A is true, you can determine that B is true; Distinguish that kind of argument from types of reasoning that is more common in the world. Walk outside, see it's wet, conclude it's rained; Deductive general to specific o Probabilistic/inductive: specific to general; Observe some evidence, now want to draw hypotheses about what's likely true; ex Client reports hearing voices - schizophrenic?
Boroditsky examples: Language and space; space and time; effects of grammatical gender
-English speakers speak of time using front-back metaphors -Mandarin Chinese speakers uses vertical metaphors (up and down): next month is down the calendar, last month is up the calendar -Mandarin speakers are faster to confirm that March precedes April when they had just seen a vertically arranged calendar compared to a horizontal format -Boroditsky, Schmidt & Phillips (2003) on object descriptions in Spanish-English and German-English bilinguals: bridge: feminine in German, masculine in Spanish German L1 descriptions: beautiful, elegant, peaceful, slender Spanish L1 descriptions: big, dangerous, strong, sturdy key: masculine in German, feminine in Spanish German L1 descriptions: hard, heavy, jagged, useful Spanish L1 descriptions: little, lovely, shiny, tiny -people's thinking about objects is influenced by the grammatical gender of these objects in their native language; support for SWH
Motor program: Evidence: not response chaining (too fast)
-Evidence: not response chaining (too fast). -Representation of the plan for movement and movement sequences -Abstract program, not specific muscles -Hierarchical. -Composed of subprograms: Less abstract representations of movement sub-parts
Expertise: Chess study (domain specific)
-Expertise usually helps ability to solve problems. Ð More experience Ð Better representations Ð More practice solving problems -But can sometimes harm: Ð Functional Fixedness Ð Water jug 'mental set' -In chess study, experts memory was no better than beginners. -Memory for meaningful configurations much better. -Memory for random configurations slightly worse (probably hindered by schemas) -Chess masters know 50,000 chess patterns. -Chess masters intentionally study these patterns. -For any non-trivial domain, true expertise develops after 10,000 hours of practice (or about ten years
Production system for mean-end analysis
-Flowchart 1: Transform current state into goal state -Flowchart 2: Eliminate a difference
Language processing: bottom-up steps
-Hear sounds -Identify phonemes from sounds -Identify morphemes & words from phonemes -Lexical access (get information about words including meaning from memory) -Put words into phrases and figure out phrase structure of sentence -Compute the meaning of the sentence based on words and phrase structure -Figure out how the sentence fits into context of overall conversation
abstract program, not specific muscles
-Hierarchical -Composed of subprograms
Means-end analysis: subgoaling, difference reduction
-Identify the largest difference between current state and goal state -Set as a subgoal reducing that difference. -Find & apply an operator to reduce the difference -(If operator can't be applied, new subgoal = remove obstacle that prevents applying the operator)
Importance of representation
-Importance of representation: For many problems, the representation may make it easier or harder to solve: Ð Algebra problems easier as equations Ð Geometry problems easier graphically Ð Decision problems easier when relevant information is laid out in a grid
Power law
-Improvement follows a power law: -Improvement diminishes with time. -It takes VERY long to gain the small amount of improvement that separates really good from great -ACT-R predicts power law of practice because of: Ð Proceduralization: Take declarative knowledge and turn into productions Ð Composition: Take several productions and join them together into one
Transparent and opaque orthographies
-In transparent languages (e.g. Italian, Albanian, Maori) there is a one-to-one grapheme-phoneme correspondence. Every grapheme is realized by only 1 phoneme & every phoneme is realized by only 1 grapheme. -In opaque languages such as English this relation can be one-to-many in both directions. A phoneme can be realized by different graphemes & a grapheme can be realized by many different phonemes
Language levels: phrases, sentences, and syntax
-Phrase: organized grouping of words -Syntax: rules that determine word order Syntax is the name for the system that arranges words together into phrases and sentences. Syntactic rules that determine word order to convey meaning. Phrases are the building blocks of sentences, words alone not enough. A phrase is an organized grouping of words. Phrases organize words into meaningful units:eg, (The French bottle) (smells) vs. (The French) (bottle) (smells). Mean different things: In one case bottle is a noun, and in 2nd case bottle is a verb. Sentence is a group of phrases that conveys a meaning. By using these rules, we can help extract meaning Ways of putting words together which are conventionalized in the english language Can still be ambiguity when you specify language
representation of the problem: Problem space
-Problem space: Whole range of possible states and operators, only some of which will lead to goal state
Duncker's ray problem: hint vs. no hint; noticing vs. using analogy
-Real-world solution: use sub-lethal doses coming in from different directions. -10-20% of subject came up with this spontaneously= Would Analogy help? -Gick & Holyoak (1980) first presented subjects with story about a general -solving radiation problem: -No analogy: 10% -Analogy + hint: 80% -Analogy, no hint: 30% -tells us people can use analogies if pointed out that two things are analogous -don't realize they can share same deep structure
Use of analogy
-Retrieve a representation of a problem from memory that is similar to the problem you currently face. -People tend to miss deep similarities between problems, because they tend to focus on surface similarities. -we tend to focus on surface level similarities rather than deep structure similarities= harder to spot analogy
Expert problem solving
-Rich organized schemas -Lots of well-organized declarative and procedural knowledge -more sophisticated representations -Spend more time on representation: experts take longer to start solution, but less time to complete it -Less means-end analysis: -pre-stored solutions in long term memory -fewer demands on working memory -More reliance on long term memory, less on working memory -Moving forward, not backward
Sequential reaction time evidence: hierarchical representation (finger-tapping experiment)
-Sequential reaction time evidence: hierarchical representation (finger-tapping experiment) -Make movement sequences: Lm, Rm, Lm, Rm, Li, Ri, Li, Ri and measure: time between responses -are individual components, but bound together in sub components -further you have to go in hierarchy= predict based on distance in the tree vs reaction times of participants
insufficient weight to current data: 50/50 base rate poker chip examples
-Suppose there are two bags with poker chips: 1) 70 red and 30 blue 2) 70 blue and 30 red -a person chooses one bag. What is the probability that it is the predominantly red one? -base rate is .5 for each bag: -i.e., P (red bag) = .5 and P (blue bag) = .5 -Suppose we choose one chip from that bag and the chip is red: By Bayes theorem, what is the probability it is from the red bag? P (red bag / R) = P (R / red bag) P (red bag)= .7 -But People say .6 (too much attention to base rate, not enough weight to current evidence)
Removing constraints: Nine dots
-There are 9 dots arranged in a square below. Using no more than 4 lines, connect all the dots without lifting your pencil - if you don't think outside of the box you don't figure out solution -extend representational space
Hindrances in representation: Top-down preconceptions
-Top-down preconceptions: When we look at a new problem, we tend to encode it in a way consistent with Long-term memory
Working backward
-Transform goal state so it is more similar to the initial state. -Useful if too many paths leading from initial state. -useful for mazes
Two kinds of theories: Normative vs. Descriptive
-Two kinds of theories: Normative vs. Descriptive. o Normative: how one ought to reason; e.g., rules of logic o Descriptive: how people actually reason; e.g., biases, heuristics
Usage-based approaches (Formulas -> Pivot schemas; Learning and generalization, U- shaped learning)
-We learn language like everything else -Language is cut of the same cloth as the rest of cognition -We learn constructions using the usual cognition- our perceptual, learning, analogical problem solving, inductive, categorization, memory ... skills as applied in social interaction -easier to find evidence that children learning specificities of grammar -1/3 of language repetitions with small changes (using schema and subbing something in; where's the butter, wheres the elephant) U-shaped curve for irregular past (went): -Initially, use appropriate form (went) -Learn rule (add -ed) and overgeneralize (goed) -Relearn correct past tense (went)
Types of problems: Well-structured/defined vs. Ill- structured/defined
-Well-structured/defined vs. Ill- structured/defined: -well-structured: Completely specified starting conditions, goal state, and methods for achieving the goal; there's a "right" solution; ex: Geometry proofs -ill-structured: Some aspects are not completely specified; possibly not find a solution -Finding the perfect mate -Choosing a career -Writing the best novel
Algorithms vs. Heuristics
-algorithms: Completely specified sequence of steps that is guaranteed to produce an answer -Usually guaranteed to produce the correct answer -But may be slow or laborious -heuristics: Short cut / "Rule of thumb" -Never guaranteed to produce correct answer -But usually quick and easy
Language and the Brain: Symptoms of Broca's and Wernicke's aphasia
-broca's aphasia: o Difficulty with production: Slow, halting speech. o Agrammatical: no function words (be, of, the). o Gist maintained but expression telegraphic o Comprehension largely intact. o Broca's area important for speech production (inc. imagined or actual production of sign language o Ex: "Stroke... Sunday, ... Arm, talking - bad" o comprehension is fine o telegraphic (used to charge for each word, so leave out not as important)- content words not function words o broca's area important in production of language -wernicke's aphasia: o Difficulty with comprehension. o Fluent grammatical speech, but makes little sense (word salad). o Made-up words (neologisms), word substitutions o Ex: Boy, I'm sweating, I'm awful nervous, you know, once in a while I get up, I can't mention the tarripoi, a month ago -different lesions in the brain leads to different language problems
Properties of language: communicative, arbitrarily symbolic, regularly structured, structured at multiple levels, generative, productive, dynamic
-communicative: Language permits us to communicate with one or more people who share our language. -arbitrarily symbolic: Language creates an arbitrary relationship between a symbol and what it represents: an idea, a thing, a process, a relationship, or a description. -regularly structured: Language has a structure; only particularly patterned arrangements of symbols have meaning, and different arrangements yield different meanings. -structured at multiple levels: The structure of language can be analyzed at more than one level (e.g., in sounds, meaning units, words, and phrases). -generative, productive: Within the limits of a linguistic structure, language users can produce novel utterances. The possibilities for creating new utterances are virtually limitless. -dynamic: Languages constantly evolve
what are the other differences between child and adult learners?
-differences in social contexts, attitudinal view
Think-aloud protocol
-doesn't tell us what type of heuristic -what are people thinking about while trying to solve problem? -conscious protocols might not reflect all cognitive processes
Representations fit problems
-example: Buddhist monk -Is there a spot along the path that the monk will occupy on both trips at precisely the same time of day? ¥ Text-based representation difficult ¥ Imagining a single monk walking up and down the mountain misleading. ¥ Modify representation to an isomorphic problem: two monks. -Both leave at the same time -One from the top of the mountain -One from the bottom of the mountain -At some point, they must meet
Less means-end analysis
-pre-stored solutions in long term memory -fewer demands on working memory
Phrase structures
-same words can sometimes be grouped into phrases differently which changes meaning
Functional fixedness: Candle problem, two-rope problem
-see an object as having only a fixed, familiar function -Duncker's Candle problem demonstrates this effect: how can you use the following items to support a candle on a wall? -Initial problem state affects the types of solutions people will entertain. -Improvements when tacks are not initially inside the box -Improvements when the box is referred to explicitly -two-rope problem: see plyers as something that pulls pins out, not as pendulum -Demonstrates functional fixedness -People think of the pliers as a grabbing device, not a pendulum
Motherese
-slower rate -higher average pitch -exaggerated pitch contours
How speech slips show the reality of syntactic, morphemic, phonemic, phonotactic and phonetic levels
-syntactic level: -morphemic level: a hole full of floors [a floor full of holes]; a language needer learns [a language learner needs]; slicely thinned [exchange error - thinly sliced]= In these examples, morphemes have undergone exchange, and thus these too must be part of our mental organization of the speech wave -phonemic level: o The phonological frame for the word dog indicates that it has one syllable, that the onset is a consonant and that the rhyme portion of the syllable is a vowel-consonant sequence. When slots in this frame are filled by activated phonemes the intended word is ready for articulation. o In order to move a sound, the speaker must think of it as a separable unit. So the speaker is imposing a structure on the speech signal in his or her mind, even though this structure does not exist physically. This is why we say that the sound unit is psychologically real. -phonotactic level: -phonetic level: glear plue sky [clear blue sky] o Here it is phonetic features, not whole sounds, which are being moved from one sound to another. o the vibration of the vocal folds is moving from the [b] in blue to the [k] in clear
Speech production and speech errors: Fromkin's model of speech production
1. Identification of meaning 2. Selection of synactic structure 3. Generation of intonation contour 4. Insertion of content words 5. Formation of affixes and function words 6. Specification of phonetic segments -Ex: she's already trunked two packs (packed two trunks)= stage 4 error
the thirty million word gap predicts vocabulary and grammar development
30 million world gap between child in "professional" SES and children in "welfare" SES
How many languages in the world?
6000-7000
McGurk effect
Close your eyes, and listen. The McGurk effect - YouTube Open your eyes, replay and listen again. The McGurk effect - YouTube Is he saying "ba ba" or "da da"? It's called the McGurk effect. Lips saying ga. Sound playing ba. You hear a compromise. You synthesize auditory and visual information. In most conversations, people around the world look at some part of the person's face Better but longer The phoneme you hear depends on visual input
Lexical Access: Context effects and Swinney priming experiment
Accessing word meaning (Lexicon = Mental Dictionary) When does context have its effect? Hypothesis 1: Get all word meanings from the lexicon, context operates later. Hypothesis 2: Context allows you to get only the correct word meaning from the lexicon. All meanings that are appropriate for a word are retrieved from memory (mental dictionary). A selection process uses context to eliminate the meanings that are not appropriate to the context. Context exerts its effect so quickly that you never access the inappropriate word meanings. Thus, only the correct meaning is retrieved. Swinney (1979) evidence shows that all word meanings are accessed initially, but after a short lag, there is a context effect in selecting the meaning that is most appropriate to the context. Many words lexically ambiguous
Ambiguity: lexical, syntactic
Ambiguity occurs when the same surface structure can correspond to one or more deep structures. Ambiguity can take place in a couple of ways. -Lexical ambiguity: word has two different meanings' ex) Time flies like an arrow, but fruit flies like a banana -Syntactic ambiguity: words can be grouped together into more than one phrase structure; ex) "Pat talked about partying with Jerry Springer," One morning I shot an elephant in my pajamas (How it got in my pajamas, I'll never know.)
Isomorphs
An Isomorph is changing the representation of a problem. E.g. hearing or reading a story about the monk who walks up a mountain one morning and down the next. You can better address the question of whether he crosses a point at the same time if you were to create an isomorph of the problem (in a different representation) by drawing out the mountain and the crossing lines of his path up and down. So they look different on the surface but have the same underlying problem.
Matthew effects
Children who hear more speech are faster at processing the words they know, and faster rates of processing speech give children an advantage in learning new words from the speech they hear Matthew effects: Unto those who have shall be given. The number of different words mothers use, the number of different grammatical structures they use, and the grammatical complexity of their speech are all positive predictors of their children's vocabulary, grammatical development, or both
Effects of experience upon language development (Hart & Risley)
Children whose mothers address a great deal of speech to them develop vocabulary more rapidly than children whose mothers talk to them less (Hart & Risley, 1995)
Developmental dyslexia and difficulties in phonological processing
Difficulty in learning to reading abilities. -Child's reading below what would be expected from their age, IQ, home background, & level of education. Difficulty reading Difficulty spelling Short term memory impairments Poor phonological awareness Impairments in: -Phonological awareness -Goat without the -t -n sounds in fish -Odd one out rhyme hat fat man cat -Phonologiocal reading -Word attack skills - words out of context -Non-word reading -Phonological coding in Working Memory -Low digit span -Lexical access from LTM -Color naming -RAN (Rapid Automatised naming) -difficulty reading words out of context -poor short term memory span
Lexical Access: Swinney priming experiment
Experiment: Subjects heard a paragraph over earphones. "he found several spiders, roaches, and bugs (x) in the corner of his room" On the computer screen at time (x), individual words were presented for lexical decision. lexical decision task: faster to decide if something is a word or a non-word if preceded by semantically related item. Two hypotheses bug will prime ANT more than SPY, indicating people have accessed the appropriate meaning of bug right away. bug will prime both of them equally, indicating that they retrieve both meanings from the lexicon. Result: SEW: 960 ANT: 900 SPY: 900 ANT = SPY < SEW Conclusion: All the meanings are primed. You retrieve both meanings of BUG from the lexicon. Suggests that you must narrow down the meanings later. But when? When looking up "bugs" meanings, do they look up all or just 1? Both meanings of bugs are accessed immediately after hearing a word (ant and spy) even though sentience is reasonably constraining Experiment 2: When does meaning get narrowed down? "Rumor had it that, for years, the government building had been plagued with problems. The man was not surprised when he found several spiders, roaches, and other bugs in the corner (X) of his room." Reaction Times to word presented at time X: Conclusion: After time passes (4 syllables; 750-1000 ms), "bug" primes only the relevant meaning
Stages in problem-solving: Representation, planning, execution, checking
Form a representation Construct a plan Execute plan Checking/Evaluation
Holophrastic stage, telegraphic stage, learning syntax
Holophrastic (one-word) stage -No syntax, need context (gestures, affect) to disambiguate -Undergeneralization & overgeneralization for first ~75 words -undergeneralization= doggie refers to their dog not all dogs (specific instance); looks like they know word for dog, but really just their dog -overgeneralization= the cat is also doggie- one word for all 4 legged creatures -beginning of use of symbols -Do understand some phrases Telegraphic (two-word) stage -Correct use of word order: -subject-action -action-object -Can convey a lot of information succinctly (like a telegraph) -"telegraphic speech": only include important words: "give cup" Learning syntax/rules (syntactic overgeneralization) -Start learning syntactic/grammar rules -Examples: -Morphosyntax (plural -s, past tense -ed) -Pivot schemas -U-shaped learning
Phonological STM predicts acquisition of vocabulary and syntax
Later, Individual differences in Phonological STM predict vocabulary and syntax acquisition -unique prediction of phonological short term memory store is a predictor of long term vocab and grammar Learning -PV had deficit of phonological working memory- cant learn new words Learning Rules: -Must learn how words are combined Grammar -Must generalize to novel sentences So can't just memorize wordings (formulaic phrases) -Need to acquire rules that can be applied to new sentences
Learning system not mere imitation - Wug tests
Imitation "Parrot-fashion learning" or Rules that apply to new cases? -Wug test - Berko-Gleason video -Learn general rules that apply to new cases -4-5 yr-olds know Plural of "wug" is "wugs" Past tense of "rick" is "ricked" -applying correct verbal morpheme to nonsense words
Using analogy: Luchins task (sometimes a hindrance), stuck in set
Improper application of analogous solutions
Infants discriminate all phonemes but gradually lose irrelevant distinctions
In 1st year, infants can discriminate all phonemes from all languages (Kuhl - 'citizens of the world') Gradually lose discriminations that are not important in their own language so that by 8-10 months they become what Kuhl calls 'language-bound infants of their community') English /r/ vs. /l/ Japanese / r / At the interval, will see Kuhl describe this research on infants 'taking statistics' -children taking statistics on language they're exposed to
Rich organized schemas, sophisticated representations
Lots of well-organized declarative and procedural knowledge More sophisticated representations
Language learning begins before birth: infant's preference for mother's voice, language, stories
Newborn babies prefer: -Their own mother's voice over other women's voices -Their mother's language spoken by another woman over another language -A nursery rhyme their mother read before they were born to an unfamiliar nursery rhyme
Generative approach (No feedback, Common developmental sequences, Poverty of the Stimulus, therefore: Language Universals, innate Language Acquisition Device)
No Feedback: -Adults usually do not usually explicitly correct children's grammar or pronunciation Only correct meaning -Furthermore, correcting grammar/pronunciation does not even help! -Kids correct themselves despite little negative feedback. Linguistic Universals: -Language is very complex -Acquisition task is not well constrained Irregular mappings from sounds to meanings Little negative feedback -All normal humans acquire language in pretty much the same order of stages of acquisition -The Poverty of the Stimulus -Suggests linguistic universals General language principals Embodied in every language (innate) -Language Acquisition Device -all humans roughly go through same stages at the same time -like going through puberty -never found a LAD though
Hobbits and Orcs
On one side of a river are three hobbits and three orcs. They have a boat on their side that is capable of carrying two creatures at a time across the river. The goal is to transport all six creatures across to the other side of the river. At no point on either side of the river can orcs outnumber hobbits (or the orcs would eat the outnumbered hobbits). The problem, then, is to find a method of transporting all six creatures across the river without the hobbits ever being outnumbered. -cant be solved with difference reduction -must be wrong strategy- this isn't possible -people don't like taking steps that move closer to start state -sometimes you have to step back first
Reading: Ideographic, syllabic, alphabetic and consonantal orthographies; graphemes, phonemes
Orthography: sounds are mapped onto a set of written symbols -Ideographic languages: words/morphemes are mapped with a pictorial symbol (character); ex) Chinese -Syllabic: syllables are associated with visual representation; ex) Japanese syllabic symbols (kana) -Alphabetic: each letter (grapheme) is supposed to represent a phoneme -Consonantal scripts: not all sounds are represented, vowels are not written down at all; ex) Hebrew Different languages and different people have changed how we conventionalize symbols in different ways Different communities came up with different solutions to solving task of getting language down- crucial implications for how people learn to read in those languages Orthography: writing system of a culture -cultures have interpreted in different ways
Does bilingualism delay decline?
Our ongoing research suggests that it does: lifelong bilinguals diagnosed with dementia on average show symptoms of the disease four years later than comparable monolinguals, with a similar rate of decline for both groups after the diagnosis
Phonological awareness
Phonological awareness Goat without the -t n sounds in fish Odd one out rhyme hat fat man cat -Phonologiocal reading Word attack skills - words out of context Non-word reading -we need to know sounds of words: not obvious (cant segment out all individual sounds) -phonological awareness important: good predictor of their ability to read -Developing phonological awareness improves reading skills. -As children learn to read, phonological awareness increases. -Training on phonological awareness can lead to an improvement in segmenting & reading skills in general. -Phonological awareness = a central concept in reading & is impoverished in non-skilled readers.
Knowledge compilation: proceduralization, composition
Proceduralization: Take declarative knowledge and turn into productions Composition: Take several productions and join them together into one -early stage on proceduralization= verbal declarative knowledge; make if-then production rules -want to compose these together
Language levels: phonemes- different in different languages
Smallest unit of speech sound used to distinguish one word from another Vowels Consonants Around 40 in English Range from 11 to around 60 across languages (Out of ~100 total) Different phonemes in different languages L vs. R in Japanese Tonal differences (Chinese) clicking sound (Xhosa - South Africa) -Phonotactic Rules of order: There are rules for sequencing phonemes Phonemes are building blocks of language
Language levels: morphemes
Smallest unit that signals meaning prefixes, suffixes, roots (-ed, -s, syn-) Can be entire words 50-80,000 morphemes Location matters (teams / steam) Rules of a language say how they can be combined. "Atoms" of meaning -content vs function morpheme: "book"- free morpheme that can stand on its own "s" at the end of the word means something, but not at beginning
Cognitive advantages in bilinguals: executive control in aging adults
This group of studies examines the hypothesis that lifelong bilingualism protects individuals against this decline by reducing the rate of the natural slowing that comes with age. These studies have shown that the bilingual advantage found in adulthood increases in magnitude with age as bilinguals maintain higher levels of cognitive control beyond 60 years old
Trapped by easy or familiar perspective: Trains meeting problem
Trains meeting problem -If both trains travel at the rate of twenty-five miles per hour, and the bird flies at one hundred miles per hour, how many miles will the bird have flown before the trains meet? -Trapped by a familiar perspective Ð Most people take perspective of bird. Ð Some might use advanced mathematics of infinite series and limits. Ð BUT, just notice how long the trains will travel (1 hour) Ð Bird, flying at 100 miles/hour, travels 100 miles
Language processing is predictive: Cloze probability affects (1) fixation time (eye- tracking in reading), (2) N400 (measuring ERP)
We're constantly predicting what is coming "Today it's hot, but yesterday it was ..." "To fix this I need to find a hammer and some ..." Your brain uses these predictions to process language more efficiently -Cloze probability: Likelihood of that word showing up in that particular situation; once upon a _= the cloze probability of "time" is very high; judged by people who speak that language -If you hook someone up to an EEG, you can see this extra processing. Let's just orient you to this graph to start. First, because neuroscientists love to be confusing, negative voltages are plotted up. At the left side, you can see where it says target word onset. That is showing when the final word of the sentence appeared. The blue line is the brain activity for a normal sentence (i.e., "A grape is a fruit.) Cloze probability for that sentence is relatively high. The red line shows the activity for a sentence like "A grape is a flashlight." You can see that peak around 400 ms. That's the N400, and it's thought to represent extra processing of the word that you didn't expect to be there -N400 is an ERP response to unexpected words Typically is a negative voltage peaking around 400ms after the word (hence N400) -Reacts to the statistical likelihood of words, not their semantic content "A grape is a flashlight" results in N400 "A grape is a fruit" doesn't give one
Berlin and Kay - Implicational Hierarchy of Color Terms: Universalism
o every language has a small number of basic color terms: -one-morphemic -not contained within another color -not restricted to a small number of objects o if a language has blue, it also green and yellow etc o not random sample of 11 color terms
Language levels: words
Words are either single morphemes or combinations of morphemes. adults know about 50-80,000 morphemes, but several hundred thousand words. Strange, stranger, strangers, strangest each counted as separate words. Impressive memory: "abrasive" or "yodel" is less than once per million, but not forgotten. Specific rules say what morphemes can be combined. ("lifted" is ok; "goed" is not)
The neural pathway for repeating a heard word
auditory cortex to Wernicke's area to Broca's area to motor cortex
Kay and Regier: hemispheric differentiation: right hemisphere shows universalism; left hemisphere shows linguistic determinism
o things in right visual field projected into left hemisphere o odd one out in left visual field or right visual field o reaction time task o sometimes odd one out crosses category boundary o left hemisphere of brain which is mainly carving world for language shows linguistic determinism (early stages) o no linguistic determinism in right side of the brain: Linguistic relativity effects in the left hemisphere but not the right
Spend more time on representation
experts take longer to start solution, but less time to complete it
How to get from declarative to procedural
knowledge compilation
Sapir Whorf - Linguistic determinism, Linguistic relativity
language shapes thought patterns; two parts of the hypothesis: -linguistic determinism: language determines certain non-linguistic cognitive processes; language affects basic cognitive processes -linguistic relativity: across languages, different non-linguistic cognitive processes are affected; different languages code language in different ways; speakers of relative languages think differently; new language=new way of thinking
Moving forward
not backward
Mammogram example: Doctors confuse p(H/E) and p(E/H)
o 80% of all women with breast cancer have a positive mammogram result; Patient A.C. has a positive mammogram. What is the probability that she has cancer? o p (pos mammogram / breast cancer)= 8 / 10 = .80 o p (breast cancer / positive mammogram)= 8 / 107 = .07 o Doctors confuse probabilities
Davidoff and Robison: English, Berinmo, Himba - linguistic determination of categorical perception
o Choose the "odd man out" in a set of 3 color chips o Perceptual distances were the same for each pair in the set o Observers judged colors from the same linguistic category (for their language) to be more similar; they were at chance for decisions relating to other language's color categories o "Our data show that the possession of color terms affects the way colors are organized into categories. Hence, we argue against an account of color categorization that is based on an innately determined neurophysiology. Instead, we propose that color categories are formed from boundary demarcation based predominantly on language. Thus, in a substantial way we present evidence for linguistic relativity."
Rosch color memory: Universalism
o each language draws its basic color terms from this list of 11 names: white, black, red, yellow, green, blue, brown, purple, pink, orange, gray o a language with only two basic color terms will have black and white; when there is third term, it is always red o Rosch (Heider, 1972) tested Dani and U.S. participants memory for focal colors (the most representative example of a basic color: the bluest blue, the reddest red, etc.) -both groups' memory was better for focal than for non-focal colors o not strong evidence of linguistic determinism
ACT-R: Procedural memory as production rules: If-then
¥ Adaptive Control of Thought ¥ Procedural knowledge -Separate from Declarative -Made of Production rules: If-then statements -Novice: Declarative. o Knowledge compilation: -Proceduralization: Take declarative knowledge and turn into productions -Composition: Take several productions and join them together into one