bjones@uhunix1.uhcc.Hawaii.Edu (Bradley R. Jones) (06/22/91)
In article <SLEHAR.91Jun18084520@park.bu.edu> slehar@park.bu.edu (Steve Lehar) writes: >The reason why the brain uses neural spiking, and encodes signal >magnitude as spiking frequency is exactly to avoid the degredation >with distance that is experienced by the alternative method of neural >signaling, i.e. the density of ions of a particular charge. > >The ions, injected at the site of neural input must diffuse passively >along the neuron, which works ok as long as they don't have to diffuse >too far. When you get one of those neurons with an extremely long >axon however, there may be little or no charge left by the time the >signal gets to the end, so the signal decays with distance. While correct in a general sense, this description is not technically accurate. Passive electrical conduction in neurons does not occur at the diffusion rate of the ions in the cell. Electrical conduction in neurons occurs just as in wires: at the speed of light. The charged partical doesn't have to move the entire distance of current flow, it bumps a neighboring ion which bumps another and so on. The reason there is spatial decrement of electrical signals in neurons is because current leaks out across the membrane capacitor as the current flows along the cell. >again to recover. That gulp of ions diffuses outward, and what >happens next depends critically on the density of electrically gated >channels in the local viscinity. If the next one is too far away, >then the charge will not be strong enough to trigger it, and the >charge diffuses slowly in space and time. ^^^^^^ ^^^^^^^^ ^^^^^^ ^^ ^^^^^ ^^^ ^^^^ As I said, the charge travels at the speed of light. The amplitude decrement is due to the current loss at the membrane capacitance. The temporal slowing is also due to the distributed capacitance. But, _and_this_is_important_, when current is injected at a point in a neuron the initial deflection will be measured at all points of the cell nearly instantaneously. The _rise_time_ is what is slowed by the membrane capactitance, so more distant sites show a more slowly rising voltage change than sites near the current injection. The reason this is important is that non-spiking neurons are able to conduct electrical signals much faster than spiking ones (but not as far). Conduction velocity in non-spiking neurons is near-instantaneous. Conduction velocity in spiking neurons depends on the rise time of the voltage change at the sodium channels, and the sodium channel density and distribution. -- -------- Brad Jones -- bjones@uhunix.uhcc.hawaii.edu - bjones@uhunix.bitnet Bekesy Laboratory of Neurobiology, Pacific Biomedical Research Center University of Hawaii, Honolulu, HI 96822
brp@dino.berkeley.edu (Bruce Raoul Parnas) (06/25/91)
In article <13584@uhccux.uhcc.Hawaii.Edu> bjones@uhunix1.uhcc.hawaii.edu (Brad Jones) writes: >In article <SLEHAR.91Jun18084520@park.bu.edu> slehar@park.bu.edu (Steve >voltage change than sites near the current injection. The reason this >is important is that non-spiking neurons are able to conduct electrical >signals much faster than spiking ones (but not as far). Conduction >velocity in non-spiking neurons is near-instantaneous. Conduction >velocity in spiking neurons depends on the rise time of the voltage >change at the sodium channels, and the sodium channel density and >distribution. In the strictly mathematical sense, this conduction is nearly instantaneous- any set of equations used to model this system will immediately show a non-zero response at all points in the cell. However, is this practically useful? While the potential at some distant point will "instantaneously" be several femtovolts, a system which works on millivolt signals will not detect this. In fact, it will be far smaller than intrinsic noise, which sets a lower bound on detectability. The voltage at these points must rise above some value before it may be detected, and this takes time. Conduction velocity should refer to the conduction of signals, and not to femtovolts. In this sense, the regenerative phase of spiking neurons makes them conduct SIGNALS faster than their non-spiking counterparts which (as was mentioned in Brad's message) are at the mercy of membrane capacitance. So, it seems to me that it all really depends on how we decide to define conduction in nerves. > > Brad Jones -- bjones@uhunix.uhcc.hawaii.edu - bjones@uhunix.bitnet bruce (brp@bandit.berkeley.edu)
pluto@koko.ucsd.edu (Mark Plutowski) (06/25/91)
bjones@uhunix1.uhcc.Hawaii.Edu (Bradley R. Jones) writes: >While correct in a general sense, this description is not technically >accurate. Passive electrical conduction in neurons does not occur at >the diffusion rate of the ions in the cell. Electrical conduction in >neurons occurs just as in wires: at the speed of light. ^^^^^^^^^^^^^^^^^^^^^^ Not to pick nits, as I get the point you are making, but is the portion marked with "^^^" approximately true in practice? Even in wires, since current occurs by movement of charge associated with massive particles, (for example, as carried by electrons,) it will be very fast, and, may be taken to be close to the speed of light in vacuum, but will not equal it due to the mass on the charge carrying particles. My question is, is the electrical conduction speed in neurons fast enough that assuming it to be close to the speed of light causes negligible error in the analysis?
bjones@uhunix1.uhcc.hawaii.edu (Brad Jones) (06/25/91)
In article <1991Jun24.175406.24957@agate.berkeley.edu> brp@dino.berkeley.edu (Bruce Raoul Parnas) writes: >In the strictly mathematical sense, this conduction is nearly instantaneous- >any set of equations used to model this system will immediately show a >non-zero response at all points in the cell. However, is this practically >useful? ....[section deleted]... >In this sense, the regenerative phase of spiking neurons makes them conduct >SIGNALS faster than their non-spiking counterparts which (as was mentioned in >Brad's message) are at the mercy of membrane capacitance. So, it seems to >me that it all really depends on how we decide to define conduction in nerves. Yes, this is a good point. _Information_ transfer in a non-spiking neuron is not necessarily faster than it would be in a spiking neuron. Many variables affect how fast the input signal reaches a level sufficient to release transmitter at a distant site. In order to maximize the length constant, a neuron may have a very high membrane resistance which will increase the time constant and lead to a slower rise time of the passive signal at a distant site. Still, for a short neuron having a membrane potential near or above the activation threshold for Ca channels in the presynaptic membrane (e.g. retinal bipolar cells), information transfer must be faster without spikes than it would be with them. This is because there is no inter-spike refractory period associated with the passive conduction. Also, the speed of passive conduction is very important in myelinated nerve where discrete spiking nodes are separated by non-spiking regions a millimeter or so in length. Of course the specialization that makes this useful is the decreased membrane capacitance and increased membrane resistance afforded by the Schwann cell wrapping. -- -------- Brad Jones -- bjones@uhunix.uhcc.hawaii.edu - bjones@uhunix.bitnet Bekesy Laboratory of Neurobiology, Pacific Biomedical Research Center University of Hawaii, Honolulu, HI 96822
bjones@uhunix1.uhcc.hawaii.edu (Brad Jones) (06/25/91)
In article <pluto.677811956@koko> pluto@koko.ucsd.edu (Mark Plutowski) writes: ....[deleted section]... >Even in wires, since current occurs by movement of charge associated >with massive particles, (for example, as carried by electrons,) it will >be very fast, and, may be taken to be close to the speed of light in >vacuum, but will not equal it due to the mass on the charge carrying >particles. I suspect you are right. In fact, it wouldn't surprise me if the speed of electrical conduction in an ion containing solution depended on the concentration of the ions, etc. and was even slower than that in a wire. Still, it must be better than .8 of the speed of light even under the worst circumstances. Any physicists out there with a real number? >My question is, is the electrical conduction speed in neurons fast >enough that assuming it to be close to the speed of light causes >negligible error in the analysis? This reminds me of a joke Jeff Wine used to tell in his introductory Physiological Psychology classes at Stanford: Physicists try to get their results correct to the 8th decimal place, biologists try to get their decimal point in the right place, and psychologists try to get the sign right. ;-) The more serious answer is that the time increment in physiological models of neuronal activity is so much longer than the time scale of electrical conduction that it is assumed to be instantaneous. As far as physiological measurements are concerned, I don't think the sweep speed on my oscilloscope is fast enough to even come close... -- -------- Brad Jones -- bjones@uhunix.uhcc.hawaii.edu - bjones@uhunix.bitnet Bekesy Laboratory of Neurobiology, Pacific Biomedical Research Center University of Hawaii, Honolulu, HI 96822
brp@bandit.berkeley.edu (Bruce Raoul Parnas) (06/28/91)
In article <13628@uhccux.uhcc.Hawaii.Edu> bjones@uhunix1.uhcc.hawaii.edu (Brad Jones) writes: >rise time of the passive signal at a distant site. Still, for a short >neuron having a membrane potential near or above the activation >threshold for Ca channels in the presynaptic membrane (e.g. retinal >bipolar cells), information transfer must be faster without spikes >than it would be with them. This is because there is no inter-spike >refractory period associated with the passive conduction. The information that gets lost in this scheme is timing. The result of a continuous modulation of Ca channels is a smearing of the temporal information present at the soma as the signal traverses the axon. In some systems this is ok, but in sensory systems this won't work. Spikes provide a means for having precise synchrony to temporal events (not single neurons, of course, but the ensemble response of neural populations). Again, it all depends on the "information" that one wishes to transfer. If time is relevant, spikes are very important. Many neural network models forego spikes in favor of analog (continuous) signals that represent something like mean spike rate. In this scheme the temporal information is lost and this is, i believe, simply not an adequate model for neurons in sensory systems. Long live spikes! > Brad Jones -- bjones@uhunix.uhcc.hawaii.edu - bjones@uhunix.bitnet bruce (brp@bandit.berkeley.edu)