Sensorimotor Transformations and Spatial Reference Frames - Lecture 15 Module 3 (Sensory and Motor Neuro)

What is LIP, and why is there ambiguity in terms of why its neurons fire prior to making a saccade (give 3 potential reasons)?

LIP - lateral intraparietal area

Neurons fire before making a saccade, but why is this firing taking place to begin with? We have 3 reasons:

• attention: “something looks interesting over there, even if I don’t move yet - but I’m currently focusing on it”

• intention: “im going to make a saccade to this target/stimulus soon”

• salience/priority: “this stimulus looks interesting even if I’m not actively paying attention to it right now - it might be important later” 

What is the parietal reach reason, and how does it do something similar to LIP?

The parietal reach region or PRR does the same thing as LIP (neurons firing prior to making a movement), but it’s just for arm movements instead of eye movements.

• So the PRR fires before REACHING occurs (instead of saccades)

LIP and PRR do the same thing, just for different kinds of movements, so it naturally means that they have to use different kinds of reference maps depending on the effector (what is actually carrying out the movement)

• in summary, different brain regions signal intention/attention/priority (or salience) of spatial targets for different movements

What is the spatial reference frame in which these neurons represent target locations for future action? In other words, what is the “important stuff” or target stimulus being represented relative to?

For planning different actions, different reference frames might be useful (or different actions benefit from different coordinate systems:

• eye-centered reference frame: useful for making saccades, because the eye pivots around the fovea (whatever you focus on, that’s within the fovea - so the reference frame should be centered on that)

• so LIP represents space in coordinates tied to the fovea

• head-centered frame: useful for planning arm movements, because where you reach depends on where your head is pointing

• body-centered frame: useful for planning reaches

Each brain region picks the map that makes the most sense for the kind of movement it’s planning

Why would it be useful to give someone’s address in Earth-centered coordinates (rather than sun-centered coordinates)?

• Remember that the Earth revolves around the sun, and rotates on its axis

• So if we make the sun the center of the coordinate system (represent the address relative to the sun), it would never be constantly in the same place (from day to day, or from season to season)

• If we made the Earth the center of the coordinate system (represented the address in terms of the Earth), the address would stay in the same place because we are just using Earth coordinates (lines of longitude and latitude)

So we have to think about using the coordinate system that makes the most sense for the action we are performing, or the target stimuli we are interested in.

Consider a monkey performing a memory saccade task. The monkey begins each trial with his eyes pointing at the gray circle (fixation point) and his hand touching the black circle. A target is then flashed at another location.

If the monkey will make a saccade, do neurons in area LIP represent the target location (X) relative to the current position of the eye, hand, or head?

Usually eye-centered - this makes the most sense, since whatever you are fixating on will be within the fovea

• this is relevant to saccades, whose size is just the difference between what you are currently looking at and what you want to look at (desired target) —> your fixation point changes

• an eye-centered reference frame centers the fovea, so we use this for saccades

FROM SLIDES: “responses of retinal ganglion cells clearly depend on the location of a visual stimulus RELATIVE to the fovea of the retina - so as the eyes move, the location of an object in world-centered coordinates ALSO needs to move to maintain the same stimulus on the retina” - i.e. in an eye-centered reference frame, target location remains fixed even when eye position varies

Consider a monkey performing a memory saccade task. The monkey begins each trial with his eyes pointing at the gray circle (fixation point) and his hand touching the black circle. A target is then flashed at another location.

If the monkey will make a reach to the target location X, do neurons in PRR (parietal reach region) represent the target location (X) relative to the current position of the eye, hand, or head?

Usually hand-centered, since the size of a reach would be similarly defined as a saccade (the difference between where your hand currently is and where you want it to be)

List out the experimental set-up of keeping head location constant, but varying eye position. Explain horizontal shifts of curves on a head-centered reference frame graph (head-centered reference frame is the x-axis, firing rate is the y-axis) - what do these tell us? What about when the curves overlap?

NOTE: we record from a SINGLE neuron (so these are tuning curves)

Experiment:

• keep target in the same physical location on the screen

• change the initial eye position (the monkey does NOT move its head, but looks left, center or right)

• record from one neuron to see whether its movement field stays “tied to the eyes” or “tied to the head”

• then they plot target location (for eyes left, eyes center, eyes right - with one curve corresponding to each) VERSUS neural firing for the y-axis

If the curves shift (horizontally): the frame of this neuron is eye-centered, because the neuron cares about where the target is relative to your gaze. 

If the curves do NOT shift (horizontally): this means that the frame of reference is tied to whatever is on the x-axis (so the frame of reference for this neuron is head-centered)

List out the experimental set-up of keeping eye position constant, but varying head position. Explain horizontal shifts of curves on an eye-centered reference frame graph (eye-centered reference frame is the x-axis, firing rate is the y-axis) - what do these tell us? What about when the curves overlap?

NOTE: we record from a SINGLE neuron (so these are tuning curves)

Experiment:

• monkey fixates on one point in space (so eye position is constant), but varies head position (head left, head center, head right) 

• so ONLY head position varies

• record from one neuron to see whether its movement field is “tied to the head” or “tied to the eyes”

• then they plot target location (for HEAD left, center, and right) versus neural firing on the y-axis.

If the movement field is eye-centered: curves will not shift, because the neuron only cares about position of the target relative to the eyes (which has a stable, FIXED position)

• the eyes are only fixating on one point

If the movement field is head-centered: curves WILL shift, because the neuron cares about the target position relative to the head (which has a varying position)

What happens when the neural representation of a single neuron is an intermediate between eye-centered and head-centered?

This is where we will see a PARTIAL (not full) horizontal shift.

Note that the curves for eye-centered neurons (when plotted on a head-centered graph) or head-centered neurons (when plotted on an eye-centered graph) AND intermediate neurons both have shifted tuning curves that OVERLAP - it’s just about the DEGREE of the overlap

• the intermediate reference frame neurons have curves that overlap a lot more

The curves shift for the intermediate reference frame neuron, but NOT enough to represent either:

• location of the target relative to the eye

• location of the target relative to the head

We don’t see full alignment for any single coordinate system, because these neurons care about the target relative to BOTH head and eye position.

What happens when we have a single neuron that has an eye-centered neural representation in which response strength varies with eye position?

The tuning curves will remain eye-centered, so:

• we won’t see horizontal shifts in case we plot its responses for the target location when eye position changes

• we WILL see horizontal shifts in case we plot its responses for the target location when HEAD position changes

BUT the strength of the response is MODULATED by the eye position - so:

• some curves are taller

• some curves are shorter

The tallest curve corresponds to that neuron’s preferred (eye) fixation point - where does it like the target to be positioned MOST within its movement field?

If we plot this on an eye-centered graph, the curves overlap, but since the curves all have different amplitudes, the DEGREE of overlap will still be imperfect.

A target flashes at 1 of 12 locations, with the target array being fixed relative to a monkey’s head. After a BRIEF delay, the monkey makes a reach to that target. The monkey is trained to LOOK at a red button and TOUCH a green button while waiting for the target to flash. Talk about the variations in experimental setup, and what that means for the data:

• when we have different initial eye positions

• when we have different initial hand positions

• we are only recording from PRR neurons - are they head-centered, eye-centered, or limb/body-centered

For PRR neurons, we ALWAYS see the largest response for the target right below the eye (indicating that they are eye centered)

• This changes when the initial eye position varies (so the highest response will be for different target locations)

• When initial eye position doesn’t change (only the initial head positions change), the PRR neuron doesn’t care. It will have the largest response for the target right below the eye, which is always in the same place

So the cell codes reaches in an eye-centered reference frame or coordinate system (it cares about potential reach locations relative to where the eye is pointing)

Why might the PRR code potential reach locations relative to where the eye is pointing (rather than having head-centered or limb-centered coordinate systems?)

PRR mostly gets input from visual cortices, it would be kind of stupid to change coordinate systems given that all prior/upstream areas use an eye-centered reference frame

• changing reference frames = introducing unnecessary noise

• PRR is basically just given the job of telling us where to reach, given what we SEE

• the target can give us signals that come through the auditory or somatosensory (touch) streams, but this isn’t relevant to PRR which doesn’t receive those inputs (so they wouldn’t influence the reference frame)

• also, in neural behaviors, saccades and movements are often coordinated, so it would make sense to link these two through an eye-centered reference frame

Many brain areas contain multisensory neurons (neurons that respond to stimuli delivered through multiple sensory modalities). The most EXTENSIVELY studied multisensory area of the brain is the superior colliculus, or SC. 

If neurons in the deep layers of the SC get visual, auditory, and somatosensory inputs, and have spatially co-registered (OVERLAPPING) receptive fields, what does this allow it to do?

Generate commands for COORDINATED (linked!) eye and head movements that orient the gaze to objects of interest 

• if head position is fixed, the SC codes saccade vectors (since only the eye location can change)

• if the head is free to move, it can code coordinated head and eye movements

What two sensory streams have overlapping receptive fields in SC (superior colliculus) neurons?

Visual and auditory

• for the auditory receptive field, the brain combines both ITD (what is the difference in time for a sound reaching one ear, versus the other?) and ILD signals (what is the difference in sound intensity between the two ears?)

NOTE: Some SC neurons get somatosensory input, but this is less common. 

Because of overlapping visual + auditory receptive fields, many neurons in the SC respond to both visual and auditory stimuli when those stimuli will be targets for an upcoming eye movement (saccade).

In what spatial reference frame do SC neurons represent potential saccade targets? How do we test this?

Usually eye-centered, but can be intermediate as well (between head and eye-centered)

We test this by having a subject fixate at different positions (so eye position varies) and testing whether the receptive field of the neuron moves with the eyes (so we would see shifts in a head-centered graph) OR stays fixed relative to the head

In what reference frame (coordinate system) would it make the most sense to code visual targets in? What about auditory targets?

• eye-centered for visual targets

• head-centered for auditory targets (since they are initially encoded relative to the head)

When we plot the auditory responses of a single SC neuron to a stimulus at a constant head-centered location, and just vary the position of the eyes, what do we expect to see if this particular SC neuron’s neural representations are eye-centered?

What about when we plot the auditory receptive field of the SC neuron for each eye, as a function of BOTH target location relative to the head and eye? Talk about how the data looks different for these two graphs.

1. VARYING EYE POSITION, BUT HEAD POSITION (AUDITORY RESPONSES) REMAIN CONSTANT

We see (horizontal) shifts in the curves for a head-centered graph, showing that the neuron is eye-centered - it cares about the position of the target location relative to the EYES.

2. PLOTTING AUDITORY RECEPTIVE FIELDS - FUNCTION OF TARGET LOCATION RELATIVE TO HEAD

We see multiple S-shaped curves for left, center, and right positions of the eye - since these curves don’t overlap, this is more evidence that the neuron is eye-centered

3. PLOTTING AUDITORY RECEPTIVE FIELDS - FUNCTION OF TARGET LOCATION RELATIVE TO EYE

We see multiple S-shaped curves just like the case above, but since these curves do overlap a lot more (collapse onto a single curve), we say that this is ALSO consistent with an eye-centered receptive field

What do the S-shaped curves (resulting from plotting auditory responses of the neuron, or target position relative to head versus firing rate) look like when we see an intermediate reference frame (neither head NOR eye-centered)?

If these curves operate with an intermediate reference frame, this means that we don’t see FULL horizontal shifts - just partial ones - between the curves

• so the neuron’s response depends on BOTH head and eye position

• this is useful for planning linked head and eye movements - the SC can now smoothly translate between head-centered and eye-centered reference frames (the brain doesn’t need all the neurons to purely represent one reference frame or another - MORE FLEXIBLE)

• NOW, downstream motor areas can extract the separate head and eye vectors - depending on the task, you can read out one variable or another

In the Jay and Sparks experiment, they measured the horizontal shift of the receptive field, for head-centered coordinates, for each pair of eye positions (0 degrees versus -24, 0 versus +24).

What does the data look like when we make a graph where horizontal field shift (in degrees) is the x-axis and the number of cells is the y-axis (so we are going from individual tuning curves to population distribution)? In this case, we have auditory responses representing positive y-values, and visual responses representing negative y-values.

peak at 0 - indicates that the shift (difference in response for the degree pairs) is near 0, so in a head-centered graph, this shows evidence for a head-centered neuron.

peak at 24 degrees - shows a horizontal shift of the curves, and that the neuron is eye-centered

Both distributions have scatter though (which refers to variability between individual neurons) 

NOTE: population responses ≠ population distributions.

• population response: averaging firing rates across neurons for a SINGLE target

• population distribution: summarizes a property of many neurons (their reference frame) - think of it like a histogram for reference frames