Naive Realism: we perceive the world as it is - direct access to external world

Representative Realism: interpretations via intermediary “sense data”

Idealism: the only reality is reality of mind - no external world
Dualism: only two kind of stuff - mental and physical stuff - mind and matter
Monism: there is only one kind of stuff - either materialism/physical or idealism/mental
Functionalism: understanding the “function” of the sensory systems is all that we need to know to
“understand” them

Psychophysics: Relationship between matter (energy) and mind (sensation)

JND: Just Noticeable Difference - smallest magnitude change that can be detected
Weber Law: as stimulus intensity increases, magnitude of change (JND) must increase
proportionately to remain noticeable, dr/r=constant, x stimulus, y sensation

Detection > yes/no: no perfect / absolute threshold since we have noise

Methods of constant stimuli
Method of limits
Method of adjustments
Discrimination > bigger/smaller than
Estimation > report stimulus exactly

Signal detection: y possibility, x intensity, miss hit, correct rejection, false alarm
Above criterion > we’ll accept a signal more
Below criterion > we’ll reject a signal more
Sensitivity = distance between means of the two curves

action potential triggers release of vesicles

transmitter molecules bind to receptor
post-synaptic electrical signal

Invasive methods
- electrophysiology (electrodes)
- imaging (voltage sensitive dyes)
Non-invasive methods
- fMRI (functional magnetic resonance imaging)
- EEG (electroencephalography)
- MEG (magnetoencephalography)

Visible light: electromagnetic radiation within a narrow energy range

Photon: a quantum of visible light
Wave: stream of photon that can be bent by lenses
Light can be
- Absorbed: taken up and not transmitted at all
- Scattered: dispersed in an irregular fashion and unpredictable directions, which is what most light
does
- Refracted: altered as it passes through a somewhat transparent surface - ex: photons are
refracted by
cornea and lens
- Reflected: redirected in the same direction
Cornea: The transparent “window” into the eyeball (carries 2/3 of eye’stotal refractive power)
Aqueous humor: watery fluid behind cornea
Lens: allows changing focus
Pupil: The dark circular opening at the center of the iris in the eye, where light enters the eye
Vitreous humor: transparent fluid that fills main cavity of the eye (gel-like; may contain “floaters”)
Retina: light-sensitive membrane in the back of the eye that contains rods and cones.
Rods specialize for night vision, respond well in low lighting, don’t process color, and
concentrated outside of fovea
Cones specialize for daylight vision, respond well in bright lighting, and have fine visual acuity,
process color, highest concentration in fovea
Optic Disc: where optic nerve connects to retina no light sensitive cells - blind spot > filling in
Photon the hits outer segment of discs, which are filled with rhodopsin (the opsin of rods) at
the chromophore (exact spot that captures light)
Opsin: protein that changes shape when absorbing a photon
When opsin changes shape membrane channel closes, dark current stops flowing in,
membrane becomes more negatively charged (hyperpolarized)
Transduction: Conversion of energy (light) from one form to another (neural)
Receptive Fields in V1
• Neurons in V1 respond more to bars of certain orientations and less to others
Cells in striate cortex respond best to bars of light rather
than to spots of light
• “simple” cells: prefer bars of light, or prefer bars of dark
• “complex” cells: respond to both bars of light and dark
Many cortical cells respond especially well to:
• Moving lines
• Bars
• Edges
• Gratings
• Direction of motion
Perceiving & Recognizing Objects
Moving beyond V1
• Theories for object recognition
• Receptive fields for houses? - view-dependent model that would only recognize particular views
of an
object
• Template matching? - no capacity for an infinite amount of templates for every possible view of
the object in
every orientation in every position
• View-independent model: a model that will recognize an object independent of the particular
viewpoint
• Edges are important! - which edges go together?
• Mid-level (middle) vision: A loosely defined stage of visual processing that comes after basic
features have been
extracted from the image and before object recognition and scene understanding
• Involves the perception of edges and surfaces
• Determines which regions of an image should be grouped together into objects
• In Extrastriate cortex bordering V1, including V2, V3, V4, inferotemporal cortex (IT), etc.
Mid-Level Vision
Extrastriate Cortex
• Receptive fields of extrastriate cells are more sophisticated than those in striate
cortex.
• They respond to visual properties important for perceiving objects such as
“boundary ownership” - for a given boundary, which side is part of the object and
which side is part of the background?
• While the same visual input occurs in (a), (b) and (c), a V1 neuron would respond
the same to all three, whereas a V2 neuron might respond more to (b) than (c)
because the black edge is owned by the square in (b) but not in (c)
Extrastriate Cortex
Finding & making sense of edges
How do you find edges of objects, and know which ones go together?
• Through Mid-Level Vision - following a structuralist model such that
perceptions are the sum of atoms of sensation (bits of color, orientation, etc.)
But one problem - how good the edge information is?
• Illusory Contour - a contour that is perceived even though nothing changes
from one side of it to the other in the image - visual system fills in gap
• Contours seem to go beyond the information available
• Some edges cannot be detected by computer-
based edge detectors, and sometimes these
computers find too many edges
Models for Mid-Level Vision
• Gestalt Psychology - perception is greater than the
sum of its parts
• Gestalt grouping rules: set of rules that describe
when elements in an image will appear to
group together
• A way to incorporate prior knowledge into
perception
• Bayesian approach - a formal, mathematical
system that combines information about the
current stimulus with prior knowledge about the
conditions of the world to form our perception
Accidental Viewpoint
Evidence for view-independent models
• Produces regularity in the visual image that is not present in the world
• We avoid accidental viewpoints - our perceptual system will be hesitant to
adopt interpretations that assume an accidental viewpoint - our brain doesn’t
rely on any one particular view
What & Where Pathways
• After extrastriate cortex, processing of object information
is split into
- “Where” pathway: concerned with locations and
shapes of objects, but not names or functions
- “What” pathway: concerned with names and functions
of objects regardless of location
Example
• Inferotemporal (IT) cortex - part of the what pathway, important for object and
people recognition independent of viewpoint, and large receptive fields
• When lesioned > leads to agnosia - failure to recognize objects in spite of the
ability to see them.
Face Perception
holistic processing relying on the analysis of the
entire object and not on adding together a set of
smaller parts or features
• Grandmother cell - single neuron
responsible for representing some
complex object eg Jennifer
Anniston
• Specialized regions revealed by
fMRI
• FFA responds differentially to
faces
• PPA responds differentially to
images of places (houses)
Perceptual Disorders
• Agnosia - failure to recognize objects despite being able to see them
• Prosopagnosia - inability to recognize faces
• Double Dissociation - more evidence for specialized regions!
Methods in Object Recognition
• Electrophysiology - Neural Recordings
• Neuroimaging - (f)MRI and MEG
• PPA: Parahippocampal place area— responds preferentially to places, such as pictures
of houses.
• FFA: Fusiform face area—responds to faces more than other objects.
• EBA: Extrastriate body area—specifically involved in the perception of body parts.
• Machine Learning - Decoding and DNNs
• Multi-level neural networks that can be trained to recognize objects.
• Many instances of an object are shown to the network, with feedback.
• Over time, the network can recognize new instances of the object that it has never been
trained on.