What are we trying to explain? Multiple facets of a simple behavior Stephen G. Lisberger Howard Hughes Medical Institute W.M. Keck Center for Integrative.

What are we trying to explain?Multiple facets of a simple behavior

Stephen G. LisbergerHoward Hughes Medical Institute

W.M. Keck Center for Integrative NeuroscienceDepartment of Physiology, UCSF

What we are trying to explain

QuickTime™ and aMotion JPEG OpenDML decompressor

are needed to see this picture.

What does Barry have to do to hit a home run?

• Sense ball motion and determine speed and direction of motion.

• Decide whether or not to swing the bat.

• Program the correct swing to meet the ball just in front of his right foot.

• Make any corrections as the ball “moves”.

• All in 400 ms!

What we are trying to explain, simplified

QuickTime™ and aSorenson Video 3 decompressorare needed to see this picture.

Methods for studying pursuit eye movement

The behavior: pursuit eye movements

QuickTime™ and aYUV420 codec decompressorare needed to see this picture.

Smooth pursuit in the time domain

Target motion versus image motion

Anatomy of pursuit

Why are there so many different parts to the pursuit circuit?

Perhaps each part of the circuit is doing a different computation.

Caudatenucleus

The signals recorded in different areas are more similar than different: consider

motor cortex (FPA), parietal cortex, and the cerebellum during pursuit

• Transient and sustained responses

• Directional tuning

• Firing related to target speed

• Similar latencies

Neurons in the FPA fire nicely during pursuit

Neurons in the cerebellum also fire

nicely during pursuit

Even neurons in the parietal cortex -- in area MST -- fire nicely during pursuit

Neural responses during the behavior don’t distinguish different parts of the pursuit

circuit -- now what?

• It’s especially troubling that all of these areas combine similar blends of signals related to eye velocity, head velocity, and visual motion.

• The same would be true in the basal ganglia, probably.

• The same would probably be true for neural activity related to Barry’s home run swing.

• So, let’s look at the behavior and see if we can use similar signals to support different behavioral features.

What are we trying to explain -- the many features of the pursuit behavior.

• Smooth eye movement persists without image motion.

• Population decoding for visual-motor transformations (probably done at multiple levels)

• On-line gain control

• Target choice in real-world situations

• Long-term adaptive changes (learning)







Natural inputs show that image motion goes away during accurate pursuit.

Visual input consists of a big

pulse of image motion

…followed by a time of very little image motion and big sustained eye motion

Continued tracking

during image stabilization

provides evidence for “eye velocity

memory”

A simple feedback controller won’t show velocity memory, and requires high internal

gain to achieve excellent tracking

+ gT E+-

I

E = g I = g (T – E) = gT – gEE + gE = gT

E (1 + g) = gTGain = E / T = g / (1+g)

Gain is 0.5, if g = 1Gain is 0.9, if g = 9Gain is 0.99, if g = 99

An idea about how to create velocity memory with a positive feedback circuit

What discharge would we expect if we recorded from F? Signals related to eye velocity. MST, FPA, and the cerebellum all potentially qualify. But this is a nuts-and-bolts function and we imagine it is supported by cerebellar circuits.

Lisberger, Exp. Brain Res. Suppl 6: 501-514, 1982.







How do we think about visual-motor transformations in terms of neural circuits?

Population code for image motion

Preliminary command for eye

velocity Motor command

Parieto-ponto-cerebellar circuit

How do we think about visual-motor transformations in terms of neural circuits?

MT


velocity Motor command








So far, we’ve treated pursuit as a negative feedback control system with velocity memory

retina

Feedforward Gain

Next, I’ll show that the feedforward gain of pursuit is variable. Gain control is another feature of pursuit that we need to understand at the level of neural circuits.

Use of perturbations to probe the gain of visual-motor transformations for pursuit

Research of Joshua Schwartz

Target position



Target position



Target position



Target position

Perturbations cause large responses during pursuit: the gain is set to “high”

During fixation During pursuit

Perturbations cause small responses during fixation: the gain is set to “low”

During fixation During pursuit

Pursuit is a negative feedback control system with a variable feedforward gain

retina

VariableGain

How do we think about gain control in terms of neural circuits?

MT

Vectoraveraging


velocityX

Motor command


Site of gain control

Gain control circuit

?

Stimulation in FPA evokes smooth eye movements

Stimulation of FPA enhances the response to a perturbation delivered during fixation

Enhancement is in the direction of the perturbation, not in the direction of the evoked eye movement

Enhancement is in the direction of the perturbation, not in the direction of the evoked eye movement


MT

Vector averaging


velocity

Frontal Pursuit Area

XMotor command

What are the input signals that allow the frontal pursuit area to control the gain

of the visual-motor transformation?

Parieto-ponto-cerebellar circuit Parieto-frontal circuit



MT

Vector averaging


velocity


XMotor command

Vision needs to be one input signal, since image motion has to turn the pursuit system on and initiate smooth tracking.




MT

Vector averaging


velocity


XMotor command

Feedback of motor command

The frontal pursuit area needs to represent eye velocity as well as image velocity. (Note another positive feedback loop, like the one through the cerebellum)



Now we know why the FPA and cerebellum have very similar outputs during pursuit

FPA Cerebellum

Now we know why the FPA and cerebellum have very similar outputs during pursuit

FPA Cerebellum

Uses image motion to initiate pursuit by increasing the internal gain of pursuit.

Uses eye velocity to keep the internal gain of pursuit high when image motion vanishes because of perfect tracking.

Uses image motion to initiate pursuit by driving eye acceleration.

Uses eye velocity to keep a moving eye moving when image motion vanishes because of perfect tracking.

Same basic signals, same basic positive feedback circuit, same basic goal of maintaining excellent tracking when the sensory input to the system goes away. Operating at very different levels of the motor hierarchy, one at a modulatory level, one at a nuts-and-bolts level.







Anatomy of pursuit

How does it work when each part is doing the same neural computation?

Caudatenucleus

Collaborators

Anne ChurchlandMark ChurchlandRich KrauzlisEd MorrisNicholas PriebeJoshua SchwartzLeeland StoneMasaki Tanaka

Albert Fuchs

What are we trying to explain? Multiple facets of a simple behavior Stephen G. Lisberger Howard Hughes Medical Institute W.M. Keck Center for Integrative.

Documents