Here are my notes on the following papers:
- Laminar differences in receptive field properties of cells in cat primary visual cortex, Gilbert (1977)
- Receptive field structure varies with layer in the primary visual cortex Martinez, Wang, Reid, Pillai, Alonso, Sommer, Hirsch (2005)
I also use some information from:
- The Role of Horizontal Connections in Generating Long Receptive Fields in the Cat Visual Cortex Bolz, Gilbert (1989)
These notes leave out a lot, but I’ve tried to capture some useful takeaways.
What is a “receptive field”?
Here’s an opinion piece from Marcus.
It can mean a lot of things.
Sometimes “receptive field” refers specifically to the map that is formed from flashing a stimulus into different parts of the visual field.
Sometimes these maps are made from simply flashing the stimulus. Other times they’re made by using some tricks, using multiple stimuli and trying to narrow down the effect one of them is having.
But this map can miss a lot of information. Here are some other details of receptive fields that can be enlightening:
- How the cell responds to a moving stimulus, not just a flashing stimulus. Some cells become active only when the stimulus enters the “receptive field”, and then remains quiet as it moves around. Other cells will continuously spike while the stimulus is inside the receptive field.
- How the cell responds to different shapes. Sometimes a cell will respond more and more as you use a longer stimulus. Other times this “summation” will quickly stop. This extended area probably should be considered part of the “receptive field”, but it would likely be missed by flashing a stimulus.
- Is the cell binocular or monocular?
The “receptive field” of a cell is a description of all of this. It’s not a single map.
Simple cells
What makes a cell simple?
- Separate On and Off subregions
- The 1977 paper measured “summation”: the more of a subregion that’s stimulated, the stronger the response
- The 2005 paper measured “push-pull”, finding that simple cells can be classified by the fact that their On subregion is inhibited by a dark stimulus and their Off subregion is inhibited by a light stimulus. This was measurable especially because they used intracellular recording – they didn’t just watch for extracellular spikes.
From the 2005 paper, here are the cells that were deemed “simple”. These fields were made by flashing square stimuli. Red is On, blue is Off.
The 1977 paper measured more properties of the cells: end-inhibition, orientation selectivity, direction selectivity. And they didn’t just use flashing stimuli, so the receptive fields look different.
Both of these papers focused on the fact that the simple cells were in the areas that receive input from the thalamus.
Complex cells
What makes a cell complex?
- The same part of the receptive field responds to On and Off stimuli.
- Respond optimally to a slit that is much narrower than the field.
- i.e. No summation.
- The response increases with slit width up to a certain point, then decreases as it is further widened. This ideal width is usually much narrower than the full receptive field width. And a slit of the proper orientation produces a good response wherever it’s placed within the field.
- They respond continuously as slits were moved across the full receptive field. Simple cells only respond as you enter the subregion.
The 1977 paper classified complex cells into two types: “standard” and “special”.
Standard complex cells
- Gave a stronger response as the length increased up to the full length of the receptive field.
- In other words, they “show summation along the orientation axis”
Special complex cells
- Respond at least as well to very short slits (1/8 º) as to a slit that extended the full length of the receptive field (average 3º).
- In other words, they “don’t show summation along the orientation axis”
- Some of them become inhibited as the slit gets longer, even before the slit leaves the receptive field.
Some other qualities of special complex cells:
- They tend to receive more equal input from the two eyes than the simple and standard complex cells.
- They have the highest spontaneous activity.
- In L3 and L4 had they have larger receptive fields than standard complex cells.
And an observation from Gilbert’s discussion:
- They are in the same areas that receives input from the cat LGN’s C Laminae, which receives similar retinal input as the superior colliculus.
- “The C geniculate laminae projected in two dense bands to the upper and lower borders of layer IV” (source)
- Note that the LGNs of different animals look different. When we talk about “C laminae”, we’re often talking about cat.
Spontaneous activity
It’s interesting that a few specific parts of V1 have high spontaneous activity.
The band of cells in Lower L3 / Upper L4 seems to be related to Upper L5. Maybe one drives the other, or maybe they play similar roles in two different microcircuits.
Different cell groups
The 1977 paper lists a lot of variations in receptive fields. Here I’m attempting to organize these details and use them to divide the layers into functional pieces.
I doubt this is complete. For example, I haven’t divided L6 into L6a and L6b because the paper didn’t make any observations that divided L6 into two parts.
Layer 2
- Standard complex
- Never simple
- Small receptive field
Layer 3
Upper L3
- Standard complex
- Small receptive field
Lower L3
- Standard complex and special complex
- Medium receptive field
- Some simple cells. They also have medium receptive field.
- High spontaneous activity
- High preferred velocity
- Same as Upper L4?
Layer 4
Upper L4
- Medium receptive field
- Standard and special complex
- High spontaneous activity
- Same as Lower L3?
L4ab
- Small receptive fields
- Simple
- More monocular, but also lots of binocular
- Low preferred velocity
Lower L4
- Standard complex
- Medium receptive fields
- Same as upper L5? But it has less spontaneous activity.
Layer 5
Common to both, or didn’t distinguish between them while measuring
- Medium receptive fields
- Mostly binocular
- Higher preferred velocities
Upper L5
- Standard complex, special complex
- High spontaneous activity
Lower L5
- Standard complex
- Very little end-inhibition (but I didn’t think the data was very convincing)
Layer 6
- Simple, standard complex
- Sometimes simple
- Very large receptive fields. Weaker response as you move away from center.
- No end-inhibition
- Higher preferred velocities
More on Layer 6 receptive fields
Long receptive fields
Here I’ll use other people’s words.
The 1989 paper described this part of the 1977 paper as:
Layer 6 has cells which respond optimally to long bars and require that the centre of the receptive field be included in the stimulus.
And Thomson described this 1989 paper as:
When layer 5 was blocked very locally with GABA application, layer 6 cells lost the component of their receptive fields that corresponded in visual space with the inactivated region of layer 5.
So these Layer 6 receptive fields are very long, but they usually require input from near the center of the receptive field. Their receptive field is probably extended by Layer 5 cells in nearby columns.
It’s interesting that the 2005 paper didn’t seem to find significantly larger receptive fields in L6. Possible explanations:
- They only used flashing square stimuli. The 1977 paper specifically said that L6’s extended receptive field can only be seen when you use long slits that start from the center.
- But would this still be required for intracellular recording? Maybe.
- They didn’t measure very many L6 cells, and they failed to get a response from many of the ones they did measure.
No end-inhibition
One interesting note from the 1977 paper: there is essentially no end-inhibition in Layer 6.
The paper points out that the superficial layers have similar amounts of end-inhibition, starting with the simple cells in L4. Meanwhile, Lower L5 has almost no end-inhibition. (Once again, I didn’t think this L5 claim had much supporting data, but it might be correct)
So maybe L6 simple cells are related to Lower L5 in the same way that L4 simple cells are related to the superficial layers.
Does “simple” mean “thalamorecipient”?
It’s pretty easy to believe that cortex connections are, roughly:
- thalamus → simple cell → complex cell → more complex cell → …
These two studies found that simple cells live exclusively in the areas that receive input from the thalamus.
Also, complex cells respond to moving stimuli over their entire field, while simple cells only respond as the moving stimulus enters the receptive field, so complex cells are probably “summing” a bunch of simple cells.
Also, as you move from “simple” to “standard complex” to “special complex”, the receptive field tends to get more and more binocular. You would expect this tendency – a cell will be statistically more likely to be more binocular as it has more and more cells in its input tree.