# Background

It is known that all sensory information is input to the brain as neural spike sequences. Now, to distinguish between the spike sequences generated by retinal red/blue/green cone cells from each other, and these from the cochlear (inner ear) hair cells, and so on, some sort of encoding scheme must be used.

To further clarify, in the case of retinal cone cells, the rate of incident light quanta in a given energy range, is the primary information that is coded. But, if all 3 types of cone cells generated identical responses for a given rate-of-incidence, like the CCD pixels in our digital cameras do, then there would be no way for upstream neurons to tell what type of cone cell a spike sequence came from. Instead, my guess is that, each type of cone cell encodes the rate-of-incidence in its own characteristic way, similar to how different types of musical instruments sound differently, even when playing the same pitch at the same intensity, via timbre.

An oversimplified illustration of Timbre

# Question

Is there evidence that each sensory neuron type has a characteristic spike sequence pattern?

# Why message type must be encoded in the message itself

During the development of the visual system, the retina, the LGN and the visual cortex develop separately initially and sometime later, axons from the retinal ganglions grow into LGN, and optic radiations from the LGN grow and reach into the cortex. As far as we can tell it is not guaranteed that a specific ganglion will project its axon to a specific neuron in the LGN. All that is guaranteed by the growth process is that ganglions close together will project to LGN neurons that are also close together.

Given this development process, when a higher region say in the V1 receives a spike stream from a neuron somewhere lower, the question arises: how does it know that this spike stream means, red, blue or green? A simple idea that occurred to me from information theory is that the message type could somehow be encoded in the message itself.

# Motivation

Evidence for characteristic spike patterns for each sensory neuron type would take us one more step towards understanding qualia, the hard problem of consciousness. My speculation is that qualia are the neuronal analogs of timbre in musical instruments.

Erwin Schrödinger thought we'd never get there. He said, "The sensation of color cannot be accounted for by the physicist's objective picture of light [as] waves [or as quanta]. Could the physiologist account for it, if he had fuller knowledge than he has of the processes in the retina and the nervous processes set up by them in the optical nerve bundles and in the brain? I do not think so."

I guess he's right in the sense that we will never be able to fully wrap our minds around the mysterious and ineffable nature of qualia.

However, evidence of characteristic spike patterns would offer resolutions to qualia related thought experiments, such as Is there something about Mary?, that philosophers seem to be pulling their hair out over.

-
if all 3 types of cone cells generated identical responses for a given rate-of-incidence ... then there would be no way for upstream neurons to tell what type of cone cell a spike sequence came from. The previous statement is false: Neurons don't just dump all their information onto one channel, they each have an axon that runs into further regions, thus up-stream neurons have information on which cone (and thus what type of cone) is sending simply by the pattern of wiring. –  Artem Kaznatcheev Jul 22 '12 at 8:44
I agree with @ChuckSherrington I think your motivation/background is a red-herring that leaves plenty of room for questioning your question. I would remove it and try to focus on the technical part of your query. Leave the possible connections to qualia for later. Also, to avoid the comment thread swelling like it has on previous questions, feel free to join us in chat. –  Artem Kaznatcheev Jul 22 '12 at 8:46
See the Retinal Ganglion article on Wiki as well, that will tell you about some of the intermediate processing going on in the retina before the information is sent to the thalamus. In general, this is a rate code, but it's more specific to whether the stimulus is on or off center in the field or the degree to which it is red or green (blue gets a bit more complicated). –  Chuck Sherrington Jul 22 '12 at 8:58
@ArtemKaznatcheev Regarding getting meta information simply from the pattern of wiring, this is a good point and an easy mistake to make, which a few of my friends also raised. The answer is somewhat subtle, deserving of a full blog post. I'll put a link to it as soon as I've finished it. –  bfrs Jul 22 '12 at 22:07
@bfrs I'm not going to continue this discussion with you until you've read a bit more about the actual, established, scientific research in this area and you come to the table with a literature reference. I think you've got a lot of passion about this, and that's fantastic (honest) but you need to have the basic physiological facts before you challenge the establishment with an idea. As of right now, you're only using smoke and mirrors to do so. –  Chuck Sherrington Jul 23 '12 at 7:25

This is an interesting idea, but I do not think it's correct. One piece of information that goes against the idea is this: auditory information is encoded by both frequency and amplitude modulation of neural spiking. The idea of spiking rate directly correlating with frequency is at odds with the idea of spiking rate containing this sort of "meta" information about the source of the activity.

We know quite a lot about how sensory information is represented in the nervous system. The simple fact of myelinated axons that don't interfere with each other already accounts for the problem you've proposed a solution for.

if all 3 types of cone cells generated identical responses for a given rate-of-incidence... then there would be no way for upstream neurons to tell what type of cone cell a spike sequence came from

Here I believe is where you've made an error. The transmission of activity from the sensory neuron itself to an upstream neuron inherently conveys information about the sensation. The firing of the sensory neuron is a translation of the external phenomena to an internal neural code. No extra information is needed to represent the source of the activity.

References

Li Liang, L., Lu, T., & Wang, X. (2002) Neural Representations of Sinusoidal Amplitude and Frequency Modulations in the Primary Auditory Cortex of Awake Primates. J. Neurophysiol 87:2237-2261.

-
This is surely more effective than the 5 page tome on the visual system I was thinking about writing! Nice work. –  Chuck Sherrington Jul 25 '12 at 8:49
Your description of the auditory system seems to be at odds with what I've read elsewhere, von Bekesy's Place Theory, which essentially says that the cochlea does a Fourier transform of the input pressure wave recorded by the eardrum. Also, I feel you have missed the problem I'm describing, here's another attempt: If red/blue/green are all converted to the same neural code, a rate coded spike stream, how do upper regions know which is which? –  bfrs Jul 26 '12 at 7:39
@bfrs Auditory physiology has advanced significantly since Bekesy. The paper I linked shows this, and you can find many more on the same subject. I do not believe I have misunderstood you. You're not considering that the individuality of neurons is itself a type of encoding. The upper regions know which is which because different retinal ganglion cells (simplification here) respond to different colors. When a green/red ganglion cell fires, it is informing the upper regions that there is green/red within that place on the retina. No extra information needs to be encoded in the firing rate. –  Preece Jul 26 '12 at 8:02
@Preece Thanks for responding. Ok, here's where I keep getting confused: What if right after the green/red ganglion's firing, a blue/yellow ganglion right next to it fires...how does the upper region know its not another (or the previous) green/red one firing a second time? –  bfrs Jul 26 '12 at 8:17
@bfrs Because each ganglion cell has it's own axonal projection to the LGN. The LGN is more or less a topographical map of ganglion cells, so the information is preserved. –  Preece Jul 26 '12 at 8:19

After an eye opening experience rereading Hubel's book, Eye, Brain & Vision and a bunch of papers, I found that the answer to my question is a superficial yes and a strong no.

It is unsurprising that neuroscientists have investigated if there are analog characteristics to neural spike streams:

Neurons communicate with each other using stereotyped electric pulses, called spikes. It is believed that neurons convey no information other than the frequency of the transmitted spikes, called the firing rate. However, it is possible that neurons may communicate some information through the finer temporal patterns of the spikes because biological, as well as electrical/mechanical signals generally reveal internal conditions of the signal generator...

... First, we found that neurons exhibit stable firing patterns that can be characterized as “regular”, “random”, and “bursty”. Second, we observed a strong correlation between the type of signaling pattern exhibited by neurons in a given area and the function of that area.

So, in a superficial sense the answer is yes, sensory neurons might have characteristic firing patterns. However, when it comes to transmitting generic analog signals, neurons have one big handicap: any analog features in the signal are lost at synaptic junctions, where the electrical signal is transformed into a chemical message.

One way to preserve at least part of the analog information, say message type encoded in a characteristic pattern, might be to have the message transmitted over a sequence of neurons of the same size and type which all fire in a similar way. However, in the visual system, there is no evidence for this. The retinal ganglion cells have their axons end in the LGN whose neurons are very different, and these in turn send their spikes to V1 whose neurons are again different. So, any characteristic patterns generated by the ganglions are lost in the LGN synapses.

The bigger question still remains: how do upper regions figure out what type of ganglion is sending the spike stream? (I'll formulate this more properly as a separate question).

-
to your final question. Did you spend time thinking about my comment. The key part of neural-networks is their wiring pattern, I recommend reading a bit of wikipedia before follow up questions, though. –  Artem Kaznatcheev Jul 26 '12 at 13:29
to whoever downvoted, could you comment on what you found wrong with my answer. –  bfrs Jul 26 '12 at 22:24
@ArtemKaznatcheev instead of the blog post promised, I added a new section to the question. In essence what I'm trying to convey is that unlike a human designed NN or chip where the wiring is completely pre-specified, in the brain, the genetic code does not seem to do this. –  bfrs Jul 26 '12 at 22:30
there are plenty of NN where the wiring is not specified initially. Why don't you look up the cascade correlation algorithm or some of the questions I asked. Making significant edits to a question after it has been answered is considered poor form. –  Artem Kaznatcheev Jul 26 '12 at 22:39
How do upper regions figure out what type of ganglion is sending the spike stream? Receptive fields. Neurons with one type of receptive field in the retina are mapped onto those with a similar receptive field in the thalamus (the notion of topography that Preece mentions) which map onto neurons with a similar receptive field in higher centers. I reaffirm that a bout with a good textbook will open these worlds up to you. –  Chuck Sherrington Jul 26 '12 at 23:14