Lecture 8: Motion- perception & processing
Every visually guided organism detects motion, but not color, stereo (depth), and form.
Motion- blindness is rare compared to other visual deficits the reason that we have that is
because there are so many areas in the brain that deal with motion processing, many areas
are motion sensitive. In that sense it’s a very robust system.
Rabbits, frogs, turtle’s birds, and squirrels can process motion on the level of the retina
already. In humans we don’t have this capability at all (non- retinal). Motion sensitivity is a
very common system, it’s robust, there are different mechanisms in different animals.
Motion is often talked about as a change in position through time and space of an object.
Objects can be based by motion. Motion is not necessarily based on objects. Illusory
rotation.
Bilocal correlator is based on the compound eyes of the fly. It’s a conceptual model to help
you deal with motion processing. The idea is that you have a light source moving from left to
right, it passes over the two input zones, it stimulates the one on the left first, then it sends
its signal towards the delay where it’s maintained for a fraction. Since the stimulus is still
moving, the second input zone is stimulated and that also sends a signal. So, because there’s
a time difference, the two signals end up at the X and the same time (pp). The Reichart
detector is the two of them connected. (pp) Signal cancels out due to inhibition when the
stimulus doesn’t move. Bilocal correlator will, next to motion with a specific direction and
speed, respond to flicker and to static stimuli.
The easiest way to change speed selectivity is to mess around with the delay. A longer delay
means more time for the signal to reach the summation point. This also mean the model is
tuned to lower speeds. Or change the span, it’s the same for span. Delays need to ‘line up’ to
get a signal.
Detector failure. The bilocal correlator is based on luminance. Reversed phi motion: due to
the dependency on luminance correlations we cannot see the motion correctly. The signals
can’t be matched based on luminance/ color anymore. This fits well to the bilocal correlator
model. The failure to see motion (to the same degree) suggests we use a mechanism like the
Reichart detector. The ability to see this kind of motion (motion not based on simple
luminance changes) suggests we use additional mechanisms beyond the Reichart detector.
First order motion is based on luminance (and/ or color), luminance change in one position,
then in other. Second order motion is motion NOT based on luminance. First and second
order motion are separable. There’s a double dissociation between first and second order
motion. That’s when a deficit (e.g., from brain damage) selectively impairs the ability to
perceive second-order motion while leaving first-order motion perception intact. Conversely,
another deficit selectively impairs first-order motion perception while sparing second-order
motion perception. There’s more interocular transfer of second (later brain structure)
compared to first order. It refers to a phenomenon in visual perception where information
processed by one eye produces an effect observable in the other eye.
è Conclusion: different mechanisms with second order processing originating later in
the visual hierarchy. First order is more eye specific. It precedes 2nd order processing.
, 1st and 2nd order motion stimuli showed us we process multiple motion types in different
brain areas. Many different types of motion and they all require different types of receptive
fields.
Local motion is usually attributed to V1. Because local motion requires small receptive fields,
a small receptive field means that you gather information locally. V1 receives input from
LGN, V2, MT. It’s known for its orientation selectivity, and it shows about 20% direction
selectivity. There’s also local motion in V2 and V3. They have many direction selective
responses, they receive much of their input from V1. They have direction selective
properties like V1. Because the properties remain the same, it appears that no new motion
processing steps occur. So V1 seems to be doing a decent amount of motion processing and
the areas subsequent on it seem to be reliant on it.
There’re some things that you can’t do with small receptive fields. That’s referred to the
aperture problem and it’s usually used for motion processing. The aperture problem is used
to illustrate that local information can’t always give you a correct answer, sometimes you
need global information. The idea is that small receptive fields provide local information,
but the local motion signal differs from the global motion signal. Large receptive fields
capture global motion signal. So, it’s local vs. global.
Global motion is usually attributed to MT. Global motion requires large receptive fields.
Local to global convergence. These large receptive fields have high degree of convergence.
MTL is also known as hV5. MT gets input from the superior colliculus, pulvinar and V1-4.
90% of the cells (motion processing cells) are direction selective. They have large receptive
fields, meaning that they converge small receptive fields. It’s associated with the perception
of motion.
Complex motion (a form of global motion) is usually attributed to MST (large receptive
fields). This area is all about rotation, expansion, and contradiction. It integrates both local
and global motion signals. With different directions and speeds.
Humans are very good at biological motion. It helps with object recognitions, it helps with
identifying activities, species, gender, emotional state, and identify persons. Biological
motion perception from Point- Lights is a spontaneous and universal perceptual ability,
occurring both inside and outside traditional laboratory environments. In chickens there is
these innate preferences for biological motion. In humans, objects need to be learned
sensitivity to motion- defined form and biological motion may take years.
MT & global motion: combine local motion signals to code the global direction of motion.
MT & motion perception: no global motion perception after MT lesions.
Multiple models to explain motion after effect -> ratio & disinhibition models, same but
different. Ratio model says that the two cells integrate later, they’re opponent cells, they
combine to form perception, and it’s the relative activity that determines perception. If
they’re equally strong, there’s no motion. If one is stronger, that’s the motion that you
perceive. In the disinhibition model the two cells inhibit each other. You get an absolute
increase over the baseline, that is what motion perception would be.
Every visually guided organism detects motion, but not color, stereo (depth), and form.
Motion- blindness is rare compared to other visual deficits the reason that we have that is
because there are so many areas in the brain that deal with motion processing, many areas
are motion sensitive. In that sense it’s a very robust system.
Rabbits, frogs, turtle’s birds, and squirrels can process motion on the level of the retina
already. In humans we don’t have this capability at all (non- retinal). Motion sensitivity is a
very common system, it’s robust, there are different mechanisms in different animals.
Motion is often talked about as a change in position through time and space of an object.
Objects can be based by motion. Motion is not necessarily based on objects. Illusory
rotation.
Bilocal correlator is based on the compound eyes of the fly. It’s a conceptual model to help
you deal with motion processing. The idea is that you have a light source moving from left to
right, it passes over the two input zones, it stimulates the one on the left first, then it sends
its signal towards the delay where it’s maintained for a fraction. Since the stimulus is still
moving, the second input zone is stimulated and that also sends a signal. So, because there’s
a time difference, the two signals end up at the X and the same time (pp). The Reichart
detector is the two of them connected. (pp) Signal cancels out due to inhibition when the
stimulus doesn’t move. Bilocal correlator will, next to motion with a specific direction and
speed, respond to flicker and to static stimuli.
The easiest way to change speed selectivity is to mess around with the delay. A longer delay
means more time for the signal to reach the summation point. This also mean the model is
tuned to lower speeds. Or change the span, it’s the same for span. Delays need to ‘line up’ to
get a signal.
Detector failure. The bilocal correlator is based on luminance. Reversed phi motion: due to
the dependency on luminance correlations we cannot see the motion correctly. The signals
can’t be matched based on luminance/ color anymore. This fits well to the bilocal correlator
model. The failure to see motion (to the same degree) suggests we use a mechanism like the
Reichart detector. The ability to see this kind of motion (motion not based on simple
luminance changes) suggests we use additional mechanisms beyond the Reichart detector.
First order motion is based on luminance (and/ or color), luminance change in one position,
then in other. Second order motion is motion NOT based on luminance. First and second
order motion are separable. There’s a double dissociation between first and second order
motion. That’s when a deficit (e.g., from brain damage) selectively impairs the ability to
perceive second-order motion while leaving first-order motion perception intact. Conversely,
another deficit selectively impairs first-order motion perception while sparing second-order
motion perception. There’s more interocular transfer of second (later brain structure)
compared to first order. It refers to a phenomenon in visual perception where information
processed by one eye produces an effect observable in the other eye.
è Conclusion: different mechanisms with second order processing originating later in
the visual hierarchy. First order is more eye specific. It precedes 2nd order processing.
, 1st and 2nd order motion stimuli showed us we process multiple motion types in different
brain areas. Many different types of motion and they all require different types of receptive
fields.
Local motion is usually attributed to V1. Because local motion requires small receptive fields,
a small receptive field means that you gather information locally. V1 receives input from
LGN, V2, MT. It’s known for its orientation selectivity, and it shows about 20% direction
selectivity. There’s also local motion in V2 and V3. They have many direction selective
responses, they receive much of their input from V1. They have direction selective
properties like V1. Because the properties remain the same, it appears that no new motion
processing steps occur. So V1 seems to be doing a decent amount of motion processing and
the areas subsequent on it seem to be reliant on it.
There’re some things that you can’t do with small receptive fields. That’s referred to the
aperture problem and it’s usually used for motion processing. The aperture problem is used
to illustrate that local information can’t always give you a correct answer, sometimes you
need global information. The idea is that small receptive fields provide local information,
but the local motion signal differs from the global motion signal. Large receptive fields
capture global motion signal. So, it’s local vs. global.
Global motion is usually attributed to MT. Global motion requires large receptive fields.
Local to global convergence. These large receptive fields have high degree of convergence.
MTL is also known as hV5. MT gets input from the superior colliculus, pulvinar and V1-4.
90% of the cells (motion processing cells) are direction selective. They have large receptive
fields, meaning that they converge small receptive fields. It’s associated with the perception
of motion.
Complex motion (a form of global motion) is usually attributed to MST (large receptive
fields). This area is all about rotation, expansion, and contradiction. It integrates both local
and global motion signals. With different directions and speeds.
Humans are very good at biological motion. It helps with object recognitions, it helps with
identifying activities, species, gender, emotional state, and identify persons. Biological
motion perception from Point- Lights is a spontaneous and universal perceptual ability,
occurring both inside and outside traditional laboratory environments. In chickens there is
these innate preferences for biological motion. In humans, objects need to be learned
sensitivity to motion- defined form and biological motion may take years.
MT & global motion: combine local motion signals to code the global direction of motion.
MT & motion perception: no global motion perception after MT lesions.
Multiple models to explain motion after effect -> ratio & disinhibition models, same but
different. Ratio model says that the two cells integrate later, they’re opponent cells, they
combine to form perception, and it’s the relative activity that determines perception. If
they’re equally strong, there’s no motion. If one is stronger, that’s the motion that you
perceive. In the disinhibition model the two cells inhibit each other. You get an absolute
increase over the baseline, that is what motion perception would be.
