Cognitive Science Midterm 2 notes 11/3/2013 2:08:00 PM
Symbolic versus Imagistic representations
Symbolic: language-like, prepositional; ie Frege, Chomsky, Newell
Imagistic: experience-like, associative
o Shepard & Metzler 1971: is this picture a rotation of the
figure time to confirm “same figure” is a linear function of
angle of rotation interpretation= people mentally rotate
one figure onto the other= imagistic mental representations
o Kosslyn 1973: focus/memorize picture focus on one end
asked if picture contains certain items time to confirm
dependent on how close/far the item is from the point of
focus interpretation = same as Shepard & Metzler =
imagistic mental representations (“scanning” picture)
Classic challenge to behaviorism (whitehead)
Defining feature of human cognition
ELIZA- chatterbot, applied pattern-matching to reply to human
SHRLDU-natural language understanding in micro-world (Links
language “understanding”, symbolically represented, with
observable action in the world)
o Language linked to action
o “All language use can be thought of as a way of activating
procedures within the hearer.” Terry Winograd, 1973
o 1. Syntactic analysis: Grammatical analysis of input
o 2. Semantic analysis: Determine what the sentence
o 3. Perception and inference: Consult the “world” for
answers to questions posed.
o **All implemented as procedures
Language understanding as dependent on multiple interacting
Marr’s 3 levels framework Contribution to cognitive science generally
Marr Prize given annually to best student paper at the Cognitive
Science Society conference
o Computational theory: the FUNCTIONAL GOAL of the visual
system is to determine the shape of objects in the world
(goal: to fly)
o Representation and algorithm: ??? (represented through:
curved wings, aerodynamics)
o Hardware implementation: neurons somehow (how exactly:
feathers, steel, balsa wood, etc)
o Afterimages can be easily explained at the
implementational neural level: photoreceptor adaptation.
o The ambiguity of a Necker cube suggests an explanation
involving competing 3D interpretations.
Children as rational scientists: goal is to correctly understand the
world around them
Marr’s model of vision
Contribution to vision science
Sees image: 1) primal sketch 2) 2.5D sketch 3) 3D sketch
Continuum from imagistic to symbolic
Hierarchal structure in perception
Says what representations and processes the visual system uses
but not the functional goal nor how it is implemented physically
Theories of the mind often driven by new technology
Many theories assume the mind is a machine
Possibility of non-human understanders
o Solipsism (no mind exists except mine) ------
anthropomorphism (attribute human abilities to anything)
Metaphors for the mind have historically reflected the technology of
o Water technology of antiquity is reflected in the Greek
pneumatic concept of the soul, e.g. Hippocrates. (ie Humors: substances that balance each other out in a healthy human’s
o Clockwork of enlightenment is reflected in the concept of
“mechanical man”. (ie Mechanical Turk, digesting duck)
o “The reception of light, sounds, odors, tastes, warmth, and
other like qualities into the exterior organs of sensation; the
impression of the corresponding ideas upon a common
sensorium and on the imagination; the retention or imprint of
these ideas in the memory; ... and finally, the external
motions of all the members of the body ... I wish that you
would consider all of these as following altogether naturally in
this machine from the disposition of its organs alone, neither
more nor less than do the movements of a clock or other
automaton from that of its counterweight and wheels...”
o Hodgkin & Huxley’s (1952) model of neuron action potential
propagation drew on “telegrapher’s equation”, e.g. the
transatlantic undersea cable.
o Tabula Rasa- blank slate; Blank Paper- paper; brain as
Physical Symbol System hypothesis
Newell and Simon (1975): A physical symbol system has the
necessary and sufficient means for general intelligent action.
o ie problem-solving!
Physical Symbol Systems
o 1. Symbols are physical patterns.
o 2. Symbols can be combined to form complex symbol
o 3. Contains processes for manipulating complex symbol
o 4. The processes for generating and transforming complex
symbol structures can themselves be represented by
symbols and symbol structures.
Syntax: the identification and manipulation of such symbols based
purely on their shape. Semantics: the meaning and changes in meaning that these
syntactic manipulations are meant to correspond to.
Search space: Space of possible problem solutions – must search
through this space to find optimal solution.
**Intelligence as search through a symbol structure
Alan Turing: The world's first and most influential computational
Imitation game: human or machine?
o Criterion: can convince you it’s a human
o Linguistic in nature
Chinese room argument
Turing test is an inadequate criterion
Searle (who speaks no Chinese) receives input in Chinese and
through syntactic symbol manipulation based on symbol shape,
produces appropriate output in Chinese
Still has no understanding of Chinese
o Systems reply: John doesn’t understand Chinese but the
system as a whole does
Searle: Even with rules memorized and after becoming
the system, still no understanding of Chinese
o Robot reply: true that room doesn’t understand Chinese but
not because it’s symbol-based…instead it is because it has no
embodied experience to link its symbols to
Searle: Even if grounded, they’re still just meaningless
o **symbol grounding problem: problem of how words and
thoughts become meaningful to speakers and thinkers while
the problem of intentionality refers to how words and
thoughts connect up in the world
Parts of brain:
o Midbrain o Hindbrain
Lobes of cerebral hemisphere:
o Frontal- motor control, speech, planning
o Parietal- spatial, sensory integration
o Temporal- objects, auditory
o Occipital- vision
The case of Phineas Gage
o “The equilibrium ... between his intellectual faculties and
animal propensities, seems to have been destroyed. He is ...
capricious and vacillating, devising many plans of future
operations, which are no sooner arranged than they are
abandoned ...” –Doctor report
o Destroyed part of frontal lobe responsible for planning and
o Some cerebral asymmetries
Right handed people have more areas in their left
hemisphere associated with language.
o But most functions are in both hemispheres
o Contralateral organization
Information from left visual field is processed in the
right hemisphere, and vice versa.
Two cortical vision systems:
Two streams hypothesis
o Ungerleider and Mishkin monkey experiments
o Identification versus localization
Interpretation: parietal cortex involved in spatial
recognition and receives most of the information from
visual cortex in same hemisphere
So damage to right parietal cortex leads to left
Dorsal- “where” pathway
Ventral- “what” pathway
Place cells in hippocampus: a neural correlate of cognitive maps.
o From rats running around a triangle track spatial firing of
place cells Words in the brain: serial versus parallel processing models
Serial model of word processing
Visual input auditory form semantic processing (meaning)
o ie passively view words passively listen to words
generate a related verb speak a visually presented word
Image people’s brains while they participate in tasks that tap
into these hypothesized processes
Problem: some tasks tap into more than one “functional box”
o Solution: subtraction method” = subtract the image of one
task from that of another, to zoom in on a functional box.
Interpretation: actually a parallel process in which there is overlap
in between functional boxes -Peterson
Accidents, surgery, imaging as sources of knowledge
Accidents can yield knowledge (although unsystematic and
uncertain) concerning the functional organization of the brain
o ie Phineas Gage case
If damage to area A is linked to a deficit in function F,
what can we conclude?
3. F may be carried out directly by A – or by another
area that is downstream from A.
Systematic surgical intervention can help disambiguate.
o ie Ungerleider and Mishkin’s monkey experiments
o Positron emission tomography (PET)- technique that produces
an image of blood flow in the body, which correlates with
Neural computation, connectionism, as an alternate view to computation in
“Classical” symbolic view where:
o The mind can be thought of in terms of symbol structures,
and syntactic operations on them.
o Physical symbol system hypothesis.
o Strong innate component
o Examples: propositions, logical inference, scripts, schemas,
symbolic cognitive architectures... o Proponents: Fodor, Chomsky, Pylyshyn, Simon & Newell
Non-symbolic representations, massively parallel processing,
multiple soft constraints
o Hypothesis of how one thought leads to another
Spreading activation: Activation spreads along links of a
semantic network, activating other concepts.
Similar to how neural networks work
But in a semantic network, each concept is a
node, whereas in a neural network, a concept is
often a pattern of activation across nodes.
Connectionists argue that:
o continuous distributed representations are the key to
understanding human cognition
o general-purpose learning mechanisms can be used to
acquire these representations
Attractions of Connectionism
o A simple, unified account of how cognition works and where
knowledge comes from.
Powerful, domain-general learning mechanisms
o Ability to explain much of the “fuzziness” of human
reasoning (dealing with almost-dogs).
o The increased neural plausibility of artificial neural
networks (as opposed to logic).
Criticisms of Connectionism
o Although fu