Final Flashcards
phonology
study of sounds of language - consonants , vowels, variations among language
Morphology
construction of words out of units that carry meaning (morphemes)
roots, prefixes, suffixes, affixes
Syntax (grammar)
ordering of words to form sentences
The dog bit the man vs. the man bit the dog
Semantics
meaning and logical form
Language involves
reducing thoughts (which may not be linguistic) to an ordering of sounds ...and decoding the order to infer the original thoughts
Multidimensional scaling (MDS) is used
to reconstruct the corresponding distances in the mental space
parse
“assign a tree structure”
prescriptive
how to do it :right” - Don’t split infinitives! Don’t use “ain’t”! Don’t end up a sentence with a preposition
descriptive
systematic, scientific characterizations of how it is actually done
Descriptives attempt to describe language scientifically
deduction
logically certain reasoning
induction
probable reasoning
reasoning
going from premises (existing beliefs) to conclusions (new beliefs)
modus ponens
affirming the antecedent a -> b (if a then b)
parallel
many nodes running at the same time
distributed
knowledge is represented as weights on the connections
oracle
compares the actual output to the target output (supervised learning)
fuzzy boundaries
Objects within them have a family resemblance but no clear definition
classical view of categories
In the classical view, categories have clear definitions
A bachelor is an unmarried adult male person
antecedent
in the implication, the thing that if it’s true then the other is true
consequent
in an implication, this is true if the antecedent is true
The “prototype” (aka fuzzy aka family resemblance) view of human concepts
Mental concepts exhibit degrees of membership, called typicality
And are defined by their central tendencies, called prototypes
Eg birds usually have feathers, fly, lay eggs, sing, make nests, live in trees - but not necessarily
modus tollens
A->b if a then b ~b (false) ------- ~a is false if b is false, and b follows a, a must be false
exemplar model
individuals make category judgments by comparing new stimuli with instances already stored in memory. The instance stored in memory is the “exemplar”.
subjective expected utility theory
goal of human action is to seek pleasure and avoid pain
subjective utility
a calculation based on the individual’s judged weightings of utility (value) rather than on objective criteria
subjective probability
a calculation based on the individual’s estimates of likelihood, rather than on objective statistical computations
heuristics
mental shortcuts that lighten the cognitive load of making decision
bounded rationality
we are rational, but within limits
satisficing
we consider options one by one, and then select one that is satisfactory or just good enough to meet our minimum level of acceptibility
elimination by aspects
we eliminate alternatives by focusing on aspects of each alternative, one at a time
temporal discounting
subjective devaluation of future value (above and beyond interest and uncertainty)
risk aversion
preference for a certain gain over an uncertain loss
certain gain
most people prefer the sure thing
expected value
the long-run value of something
rational price
expected value
availability heuristic
we make judgments on the basis of how easily we can call to mind what we perceive as relevant instances of a phenomenon
illusory correlation
we are predisposed to see particular events or attributes and categories as going together, even when they do not
conjunction fallacy
individual gives a higher estimate for a subset events than for the larger set of events containing the given subset
sunk-cost fallacy
represents the decision to continue to invest in something simply because one has invested in it before and hopes to recover one’s investment
opportunity costs
the price paid for availing oneself of certain opportunities
deductive reasoning
process of reasoning from one or more general statements regarding what is known to reach a logically certain conclusion
proposition
an assertion, which may be either true or false
premises
propositions about which arguments are made
conditional reasoning
the reasoner must draw a conclusion based on an if-then proposition
deductive validity
logical soundness
modus tollens
denying the consequent
syllogisms
deductive arguments that involve drawing conclusions from two premises
categorical syllogism
the premises state something about the category memberships of the terms
inductive reasoning
the process of reasoning from specific facts or observations to reach a likely conclusion that may explain the facts
causal inferences
how people make judgments about whether something causes something else
symmetry
distance(a,b) = d(b,a)
Triangle inequality
shortest distance between two points on a straight line
d(a,b) + d(b,c) >/= d(a,c)
does similarity obey the distance axioms?
No (symmetry and triangle inequality)
featural similarity can be __________ and can ______ the triangle inequality, like
asymmetric, violate - like human similarity judgments
language is
a way of communicating
Chomsky’s review of Skinner’s Verbal Behavior argued
that the infinite compositionality and productivity of language makes this explanation inadequate
competence
the abstract knowledge held by a “competent” speaker of the language: what you must know to distinguish sentences from non-sentences
performance
how it works in practice: the details that deviate from the ideal competence, and the actual mechanisms for carrying it out (speaking and understanding)
pragmatics
practical aspects of conversation (what am I supposed to say next?)
generative grammar
system for producing all and only the legal structure in the given language
____ ____________ experiments seemed to confirm the “psychological reality” of syntax boundaries
tone localization
rewrite rules
describe ways in which certain symbols can be rewritten as other symbols
parse (syntax)
assign a tree structure
principle of minimal attachment
for each new phrase, attach it to the existing tree in the simplest way possible
Garden path sentences
a grammatically correct sentence that starts in such a way that a reader’s most likely interpretation will be incorrect; the reader is lured into a parse that turns out to be a dead end or yields a clearly unintended meaning
phonemes
individual sound classes (tooth = /t/ /oo/ /th/
Phonemes are distinguished by a number of parameters
manner of articulation, place of articulation, voicing characteristics
place of articulation
bilabial (p,b) vs labiodental (f,v) vs various other types
manner of articulation
stop (p,b,t,d) vs fricative (f,s,th) vs various other types
voicing characteristics
voiced/voiceless: f/v. s/z, th/th
voice onset time (vot)
voice onset time (vot)
when the voicing starts relative to the onset of the articulation
after the critical period for learning, speakers are sensitive to distinctions ________ categories, but “deaf” to distinctions ______ their native categories
between, within
Early network models
Pandemonium and Perceptron
Parallel
Many nodes that run at the same time
distributed
knowledge is represented as weights on the connections
Back-propagation algorithm (connectionism)
- Feed an input through the network, obtain an output
- an oracle compare the actual output to the target output (supervised learning)
- using the “error” (discrepancy), update the weights on all the connections
Parallel distributed processing
input layer - > hidden layer -> output layer
ALCOVE model of categorization
Nodes represent exemplars
Changing the weights (learning) means changing which stimulus features tend to activate which exemplars and thus which categories
Stimulus dimension nodes -> learned attention strengths - > exemplar nodes -> learned association weights -> category nodes
Past tense neural network
Learns the correct input-output by backpropagation
in a parallel distributed fashion
without rules
without morphemes
without any distinction between regulars and irregulars
McClelland and Rumelhart proposed to explain the most rule-like and symbolic phenomenon _______ _____
without rules, thus past tense learning would be an application of general learning mechanisms not specific to language
Symbol systems side (connectionism wars, rationalist/nativist)
the brain uses rules operating on symbols to understand the world
different learning mechanisms in different domains
some knowledge is innate
connectionist systems can’t represent the full infinite productivity of human thought
Connectionist side (connectionism wars, empiricist, associationist)
rules and symbols are just epiphenomenal (side effects)
all knowledge is implicit in the connections between neurons
one general mechanism explains learning in all domains
only connectionism systems are biologically plausible
reasoning
going from premises (existing beliefs) to conclusions (new beliefs)
interference
competition between items, decreasing the likelihood of consolidation
rehearsal
repetition of an item to facilitate consolidation
encoding
representation of information in order to facilitate storage
consolidation
movement of information from working memory to long-term memory
primacy effect
suggests that rehearsal is required to consolidate items into long term memory
early items are recalled better (because of less interference)
recency effect
last few items are recalled better (because they are still in working memory)
suggests that items are temporarily held in a small, short-term buffer
dunning-kruger effect
overconfidence bias
library metaphor
access cues are like indexes in the card catalog
what is forgetting?
failure to consolidate vs. failure to retrieve
interference among access cues - generally not “decay”
Declarative memory
knowledge of “facts”
episodic memory
personal experiences; subjective point of view
procedural memory
how to do things; especially motor skills
mental imagery
memory for visual appearance
What is a “chunk”?
a link to long-term memory
a pattern or group
Capacity of visual short-term memory
7+/- 2 items, or 3+/- 1 chunks
imagistic representation
stores the sensory experience
propositional representation
stores the abstract relation (under(cat,chair)
Mental rotation
involves a mental analog of physical rotation
a spatially organized analog of a real picture is progressively transformed
it is different from propositional or declarative info because it encodes info that hasn’t been verbalized
mental image
representation of a visual scene
stored in ltm
retrieved and placed in a short-term visual buffer
examined by the visual system
mental scanning
mental distance is an analog is actual distance
mental imagery (kosslyn)
representations of mental images are quasi-pictorial analogs of real images
evaluation is performed by part of the visual system
mental imagery (pylyshyn)
representations of images are propositional or descriptive
scanning time results are due to cognitive expectations of subjects
association networks
knowledge is stored in the associations (locke, freud)
hierarchical representation of knowledge
Collins and Quillian semantic network: superordinate to subordinate - living thing -> is_a animal -> is_a bird -> robin -> has_property red, sings
