Cognition Exam 3 Class Notes

the importance of concepts

inferring about similar objects and how it will be to experience and identify things

important bc w/o categorization, cognition would be chaotic

categorization

process of identifying an object and putting it in a category

concept

your mental representation that allows you to put objects into a category

category

all the things in the world

ex: all the cherries in the world

classical view

a concept is a list of necessary and sufficient features you need to have

eg to be classified as a grandmother you need to be female and parent of a parent

-by the classical view, concepts are definitions. this works for some terms (like legal terms) but not others

typical categorization

more typical exemplars come to mind first, you are also faster at confirming that they are a member of a category

eg furniture - sofa, its faster to verify more typical instances

probabilistic models

category membership is a matter of probability, not all or none

similarity model

the likelihood of being a member of a category is calculated by computing similarity of the exemplar to the concept

similarity model 1: prototypes

experimenters give card w dot pattern to categorize into A or B, give feedback

-start w perfect prototype, then more dots around

study exemplars until can categorize correctly, subjects as good on prototype (thing they havent seen before) as old items

how a prototype model works

Ps confident because the matching process is based on similarity in prototypes

-feels like the prototype model is right but its not

prototype model problem: ad hoc categories

temporary, goal-driven, and context-dependent conceptual groupings created spontaneously, rather than stored in long-term memory. seem like typicality but feels made up, weakens prototype model

examples "things to take from a house on fire" or "items to sell at a garage sale"

Second similarity model: exemplar model

People classify new objects by comparing them to stored memories of specific, previously encountered examples

exemplar problem 1: feature selection

selecting features to make them similiar or different, but similarity can be different depending on what features we select

-ex: comparing Joe Biden and a pack of gum, they arent similar but they could be depending on what features you pick

exemplar problem 2: similarity and context

similarity changes depending on context, this is a problem because context wasnt mentioned as being important for similarity

exemplar problem 3: categorization based on rules

-similarity: object 3 in diameter, is it more similar to a pizza or quarter

-categorization: the object is 3 in diameter, is the object a pizza or quarter

is vs what it could be

low similarity high diagnosticity

a feature or piece of information is highly effective for identifying an object, even if it does not share many common features with other members of its category

-ex: if a person jumped into a pool with their clothes on is the person drunk?

multiple systems view — dealing with problems of models

theres multiple ways categories work

-Allen and Brooks digger vs builder experiment: memorize or rule is at least 2 features

difficulty with phonemes

individual speech sounds

-french vs english phonemes are different, thats why sounding like a native speaker is hard

-”beads on a string”

written communication

-most alphabets are phonetic

-some alphabets are syllabic (written symbol system)

-phonemes → graphemes → words

grapheme

a group of letters

-could correspond to multiple words/sounds

depth of orthography

complexity of english is higher than finnish

the closeness of the relationship between orthography and phonology

mental representation of words

-most words are very infrequent

-sound, spelling, and meaning are separate but limited

sentences

assembled from phonemes → words, word order is important

-connect different ideas and concepts temporarily, so ideas don’t interfere with long-term memory and doesnt change concept of self

-helps in “what-if” scenarios

structured representations

create new temporary relationships quickly

-specify complex relationships and allow them to be expressed

grammar

set of rules that describes the legal sentences that can be constructed

-its not what you find in a grammar book exactly, it depends on what people carry in their head, usual from a general consensus

incorrect theory: associations

Grammatical sentences are understood word by word based on the association of the rest of the words in the sentence

ex: “the boy took his baseball bat and hit the ___” (probably ball but could be window)

→ associations cant drive comprehension

incorrect theory: word chain theory

moves left to right to generate many sentences

-looked promising but didnt work because it had issues with dependencies

dependencies

focuses on direct, directed links between words (head-dependent relations) rather than phrase structure constituents

-eg verbs must agree “either” implies “or” “if” implies “then”

ex: either the girl eats candy or the girl eats candy

solution to dependency problem

phrase structure grammars

-phrase structure specifies a limited number of sentence parts and a limited number of ways the parts can be combined, how you extract meaning

sentence = noun phrase + verb phrase

language as open-ended

-open-endedness is accounted for bc definitions can be recursive (meaning that a definition has that definition embedded in it)

textual representations

relating sentences to each other

-they are important bc thats what sticks, like a summary of the text as a whole

shared knowledge

-more common points of reference

-eg talking to a friend vs talking to a stranger

social context

shows power dynamics

eg: a command by a boss = “you might want to clarify that”, but that is a suggestion to a boss by employee

physical context

obvious

“hand me that would you?”

conversational norm — violations lead to inferences

1. quantity: say all that is needed and no more “when is class” “9:30”

2. quality (truth): “this paper is good”

3. relevance: “are you coming” → “I have a conflict” is relevant. Violation = “how was the toast” → “his wife is pretty”, irrelevant

why is phoneme perception hard?

-phonemes are produced fast

-different speakers produce phonemes differently, eg regional accents

ex: “Bob” pronounced differently, phonemes different but you know it means Bob → problem in language theory

coarticulation problem

a single speaker produces phonemes differently, based on the context of the phoneme

example: lip rounding before saying “tulip” and buzzing sound before saying the letter v

phonemes — restoration effect

top-down processing influences perceptual info

ex: “the state governers met theirh their respective leg*slatures”

→ the * wasnt noticed and the restoration effect seems embedded in processing system

McGurk Effect

auditory system not on its own, you can also use vision like looking at someone’s mouth in a crowded room to understand what they are saying

-illusion that you combine visual and auditory info

ex: ambiguous audio track with conflicting visuals

phonemes — categorical perception

-category of auditory system where people dont perceive slight differences in phonemes

-with machine speech you can vary voice onset time to the millisecond eg bat/bad example

speech stream

no space between words

ex: i scream you scream we all scream for ice cream

lexicon

mental dictionary of all the words you know

-the matching process between input and lexicon are how words are perceived

-to test, you can do cross-modal priming

cross-modal priming

testing across modalities

-hearing “hat” gives lexical access to activate meaning and spelling

-dont need perfect lexical access

→ Gaskell et al showed mispronounced words do get lexical access if they’re mispronounced, fast RTs indicate lexical access

parser

for phrase structures to work, something needs to decide what is a noun phrase and what is a sentence

parser cues

1. key words: like “a” indicates that a noun phrase follows (ex: Fodor and Garrett: the car that the man whom the dog bit drove crashed vs the car the man the dog bit crashed)

2. word order: subject, verb, direct object considered typical/default in engliosh

3. principle of minimal attachment: if new word can be attached to an existing node in phrase structure, go with that interpretation (readers interpret the simplest meaning first)

ambiguity

more than one solution for a sentence structure

ex: they are frying chickens (parser sensitive to context)

-can be solved with background knowledge

sentences after extracting meaning

propositions, not phrase structures, are stored

-relations correspond to verbs and adjectives, relation followed by arguments

propositions

smallest unit of knowledge that can stand as an assertion (can be true/false)

ex: Dan is handsome and bitter

reading

important bc its practical

-when you read you are co-opting different functions

pictographs

using pictures/symbols that look like what they represent, depend on context

difficulty: its hard to draw something complicated, abstractions

logograph

can be abstract symbols, need prior knowledge to know what it is

problems:

1. proper names

2. memorization problem: reading and writing becomes a class issue

grammar problem with logographs

does not differentiate time, counterfactual states or if the reference is general, etc

-knowledge of grammar is implicit so most writing systems use a code that is mostly phonological

decoding

means first identifying letters and differentiate between them, then mapping letter to sound, and hearing sound

decoding challenge: identifying letters

eg b vs d and p vs q

-inconsistency in pronunciation (caked vs naked)

decoding challenge: hearing individual speech sounds

produced differently based on context

-you need to be conscious of hearing the difference between sounds like “b” and “p” because when reading you need to assign sound to letter

problem with dyslexia

not a visual problem but a problem with phonemic awareness

fluency

depends on spelling representation

-kids develop orthographic representations when they become more fluent, and teaching them translation rules and spelling representations increase fluency

self-teaching hypothesis

account of where representations come from → fluency is very important for comprehension bc it involves working memory to get lexical access

in stroke patients: can pronounce normal and hard words like cake or yacht easily but cant pronounce made up words like slint

the simple view of reading

reading = decoding + oral language

-very high correlation of listening comprehension and heading comprehension

the situation model

in reading we remember meaning not the ways things are phrased

-you need to know the meaning of each thing in the sentence and scan long term memory for connection

causal connections

important for comprehension, “therefore” is an example of a connection to draw in a sentence

“trish spilled coffee, therefore dan got a rag”

introspectionism

goal: account for conscious experience

-felt imagery was central to thought but according to behaviorists imagery is a private event and not observable

Pavio — concrete and abstract noun experiment

concrete noun = physical thing like a potato

abstract noun = justice, miracle

-concrete better remembered bc you can visualize it

rotation experient

Ps say whether objects are the same or different when rotated

-RT for same or different related to the angle of rotation

structured relationships

-separate systems

-specify complex relationships

-create new relationships on the fly

propositions

describe relations, they have syntax, truth value, and are abstract, not spatial

images

depictive, just show whats there

-have no syntax, truth value only when described, concrete, and spatial medium

imagery

searching for something in your mind, could be an epiphenomenon

epiphenomenon= feeling of looking

homunculus problem

issue with epiphenomenon

-there isnt someone in your head watching a TV, this would create an infinite regress saying someone is looking at a tv in your head

imagery property: rotation

same (rotated) or different (mirrored)

imagery property: size zooming

far away images are hard to see details

-size of the image matters, reaction time decreases as images size increases

imagery property: scanning

scanning a longer distance takes longer

-ex: Ps scan island map and then asked to mentally travel different distances

→ consistent w idea that visual mental images are inherently spatial

imagery property: brain locus

propositions are linguistic representations → use to answer questions like what color is a bee

-brain informs cognitive theory

how imagery processing works

1. generate

2. maintain

3. inspect

4. transform

generation

takes time, created piece by piece

-fill in the G experiment: show Ps capital letters and memorize them to imagine it on a grid

-generating reflects the what from the where distinction from perception, damage in pathway creates issue in spatial aspect when imagining

ventral damage

impairs visual imagery

damage to dorsal

impairs spatial imagery

maintenance

moving attention away completely removes mental imagery

inspection

problem bc you cant inspect something to figure out what it is without already knowing what it is

-allows you to determine what a visual image is, not always perfect

transformation

spatial medium, manipulates images

-allow you to prep for future actions

-ex: will this bed fit in the room

decision making

2 or more choices and you have to select one

-what are the consistent rules that guide decisions like this?

expected value theory

get the highest value → (probability obtain) x (value)

predicts people choose highest number

problem with expected value theory

different people assign different values to non-monetary and monetary outcomes

expected utility theory

value of outcomes vary depending on the individual and context

→ (probability win) x (value of prize to me), suggests people choose rationally as long as probability and value of prize dont change

irrationality

effects of how you measure preference and effects of problem description

expected utility theory problem: problem description

people engage in more risks to avoid negative outcomes, this shouldnt make a difference but it does

-island disease example

heuristic

shortcuts to calculate probability or value, but can be misleading and lead to strange outcomes

coin toss example: people pick coins that are randomly flipped because they look more random → this is an example of a heuristic

representativeness

as a measure of probability: used when asked to judge the probability of an event

availability in probability

used when trying to call things to mind

ex: the more examples you can think of the more probable you judge an event to be

-estimating causes of death is usually overestimated bc its reported on the news more

sunk costs

time, money, or another investment that is irretrievably spent

-shouldnt affect your decisions but it does

ex: wearing shoes you dont like just because they were expensive to “get money's worth”

anchoring and adjustment

heuristic used to estimate value by using a quick estimate based on memory or info provided, then adjust the estimate

anchor: saying a high price first then adjusting a lot more

bayesian theories

shortcuts are based on experience and we should think about the way that we use prior knowledge to influence current decisions and develop theories

bayesian theories of choice

prior belief combined w new info → probability

-suggest that what look like biases and shortcuts all stem from ways we update/dont update our knowledge

dual process theories

system 1: fast, associative, not demanding of working memory

system 2: slow, uses abstract reasoning, analysis, does demand working memory

ground beef example (fat vs lean): show negative and positive connotation in choice even though they are the same

reasoning

refers to problems in which you are given some facts you are to take as true, then either draw a conclusion or evaluate a suggested conclusion

syllogisms

form of deductive reasoning consisted of 3 statements (2 premises and 1 conclusion), question is whether the conclusion follows from the 2 premises

-can be valid without leading to a true conclusion, hundreds of forms of them

-from a bayesian perspective: we would say people cant ignore prior beliefs in this

conditional syllogism

consists of major premise, minor premise, and conclusion

-structured around “if-then” statements

case-based reasoning

we remember what worked before, maybe we’re just good at scenarios that we are familiar with → cant be the only way we reason bc subjects fail on versions similar to familiar ones and do well on unfamiliar ones

bouncer version: people can likely recall the answer to reasoning problems they’ve seen before

evolutionary view of case-based reasoning

-we are social animals and evolved to do social exchange → we can catch cheaters

ex: its a social rule that to drink alcohol you must be an adult

in experiment: people’s choices changed depending on who they saw as potentially cheating, involves precaution