PSYC 212 Lecture Notes - Lecture 7: Parahippocampal Gyrus, Fusiform Face Area, Extrastriate Cortex
6 Nov 2018
School
Department
Course
Professor
Object recognition
Mind reading
Retinotopy: visual information coming from adjacent locations
in the retina will project to adjacent locations in the primary
visual cortex (V1)
○
Therefore it should be possible to reconstruct the retinal image
based on the brain activity
○
Show hundreds of random images and shapes to a participant
(10 x 10 patches)
1.
Use patterns of activity across thousands of voxels in V1 to
predict the luminosity of smaller rectangles within the 10 x 10
patch
2.
Use the predictive patterns to predict which shape the
participant is seeing based solely on brain activity
3.
•
Dream reading
Record activity in Higher Visual Cortex (HVC; lateral occipital
complex (LOC), fusiform face area (FFA), and parahippocampal
place area (PPA), and lower visual cortex (LVC; V1 to V3) while
participants are sleeping
1.
Wake them up in sleep stages 1 or 2, and ask them what they
were draming about
2.
Use the patterns of activity in the ~20 seconds before
awakening to predict what participant is dreaming of
3.
Use the predictive patterns to predict dream content solely on
brain activity
4.
•
Beyond V1
Extrastriate cortex: brain region bordering primary visual cortex
that contains other areas involved in visual processing is split
into a dorsal "what" pathway and a ventral "where" pathway
"where" pathways is concerned with the locations and
shapes of objects but not their names or functions
§
"what" pathway is concerned with the names and
functions of objects regardless of location
§
○
Early areas (V1, V2, V3, V4 and MT) maintain retinotopic
organization
○
Receptive field sizes increase from V1 to higher level areas
○
Receptive field complexity increases from V1 to higher level
areas
For instance, "boundary ownership." neurons in V1 would
respond equally to the identical edge, but V2 neurons
would respond more in "a" because the black edge is
owned by the square
§
○
Neurons in V4 respond to concave, convex, or straight edges
○
The contour has to have a precise relationship with the rest of
the figure (e.g. convex contour pointing up from the figure)
○
Preferred location covers a large portion of the retinal image
○
•
What and where
Ungerleider and Mishkin (1982)
Monkeys are trained to perform a landmarks task (the
food is the bin closes to the landmark) and an object task
(the food is in the bin under the square)
1.
Lesions to the parietal cortex (where pathway) will create
deficits in the landmark task, but not object task
2.
Lesions to the inferotemporal cortex (where pathway)
will create deficits in the object task, but not landmark
task
3.
○
•
Inferotemporal (IT) cortex
Inferotemporal (IT) cortex: part of the cerebral cortex in the
lower portion of the temporal lobe, important for object
recognition
○
Visual agnosia: failure to recognize objects in spite of the ability
to see them
○
Apperceptive agnosia: inability to form a percept despite
normal vision. The basic elements of the object are seen (lines,
etc) but cannot be integrated into a stable percept
Lesion closer to V1
§
Cannot copy
§
○
Associative agnosia: there is a perceptual representation of the
object, but the patient doesn't know what the object is
Lesion further away from V1
§
Can copy, but cannot identify
§
Can recognize object is presented in another modality
§
Prosopagnosia
An inability to recognize the identity of faces□
Face inversion effect□
§
○
Receptive field properties of iT neurons
Very large - some cover half the visual field
§
Don't respond well to spots or lines
§
Do respond well to stimuli such as hands, faces, or
objects
§
○
Object recognition can be fast (~150 ms)
Not a lot of time for feedback from higher brain area
§
Feed-forward process: a process that carries out a
computation (e.g. object recognition) one neural step
after another, without the need for feedback from a later
stage to an earlier stage
§
Grandmother cells: could a single neuron be responsible
for recognizing your grandmother?
§
○
The temporal lobe also comprises the hippocampus, a structure
that is important for storing and retrieving memories
○
Grandmother cells might therefore be imporant for our
recognition of certain learned objects
○
However, our experience of an object can't only just depend on
these types of cells
○
For instance, when we look at an object (say a toaster), we
know exactly where it is, and whether or not we are looking at
an actual toaster or just the word "toaster"
○
A truly invariant neuron wouldn’t give us information about
these important details, which are essential in guiding action
○
•
Object recognition
The problem of object recognition (slide 36)
The pictures are just a bunch of pixels on a screen, but in
each case you perceive a house
§
How do you recognize that the first and third images are
depicting the same house but from different viewpoints?
§
How does your visual system move from points of light,
like pixels, to whole entities in the world, like houses?
§
Structuralism: dots make lines -> lines make corners ->
corners make shapes -> shapes are connected to one
another and make objects
§
○
•
Gestalt grouping rules
Our perceptual system needs to make inferences based on
rules
Gestalt: in German, 'form' or 'whole'
§
Gestalt psychology: "the whole is different than the sum
of its parts"
§
Gestalt grouping rules: a set of rules that describe when
elements in an image will appear to group together
§
Good continuation: a Gestalt grouping rule stating that
two elements will tend to group together if they lie on
the same contour
§
Closure: a closed contour is preferred to an open contour
§
Occlusion: the perception that a form is occluding
another form can be understood using the principles of
good continuation and closure
§
Similarity: similar looking items tend to group (colour,
size, orientation, etc…)
But items have to be similar on only one dimension,
not a complex conjunction of dimensions
□
§
Proximity: items that are near each other tend to group
But proximity can be overruled by grouping or
connecting the dots
□
§
Common region: items will group if they appear to be
part of the same larger region
§
Connectedness: items will tend to group if they are
connected
§
Texture segmentation: carving an image into regions of
common texture properties
Depends on the statistics of textures in one region
vs another
□
§
○
•
Perceptual committees
Perception involves figuring out which interpretation of the
reality is the most likely
○
The pandemonium model: demons represent neurons, and
each level is like a brain area
○
Committees must integrate conflicting opinions and reach a
consensus
○
•
Lecture 7
Tuesday, January 30, 2018
12:58 PM
Document Summary
Retinotopy: visual information coming from adjacent locations in the retina will project to adjacent locations in the primary visual cortex (v1) Therefore it should be possible to reconstruct the retinal image based on the brain activity. Show hundreds of random images and shapes to a participant (10 x 10 patches) Use patterns of activity across thousands of voxels in v1 to predict the luminosity of smaller rectangles within the 10 x 10 patch. Use the predictive patterns to predict which shape the participant is seeing based solely on brain activity. Record activity in higher visual cortex (hvc; lateral occipital complex (loc), fusiform face area (ffa), and parahippocampal place area (ppa), and lower visual cortex (lvc; v1 to v3) while participants are sleeping. Wake them up in sleep stages 1 or 2, and ask them what they were draming about. Use the patterns of activity in the ~20 seconds before awakening to predict what participant is dreaming of.
