Sensory systems provide the input from which we
form our internal representations of our enviroment, by transducing energy into
neural signals, and then coding this information into a train of digital pulses
which are conveyed along neural pathways.
�A code is a set of rules whereby information is is
transformed from one set of symbols to another�[1],
independent of medium. For instance, in the case of vision, the retinal image
pattern (corresponding to the input symbols) is coded as the firing of neurons
(the brain�s sensory code), transferred electrically as all-or-nothing action
potentials and chemically by reception of neurotransmitter releases across
synapses.
In the nervous system, information
is frequency-coded as action potentials. An action potential is a rapid
all-or-nothing electrical pulse which travels along neural pathways, i.e. down
axons and chemically over synapses through to other neurons. The rapid depolarisation
and hyper-polarisation of an action potential spike is of almost uniform
amplitude, lasting last less than two milliseconds, and firing at a rate of up
to five hundred per second for very intense stimuli.
Like a radio signal, an electrical
signal can be either frequency-coded (FM) or amplitude-coded (AM). The nervous
system has evolved to use digital, frequency-coded signalling.
The frequency-coding works by relating the
intensity of stimulus to the rate of firing. A neuron only fires off an action
potential if the input (whether from a sensory receptor or another neuron) is
above a threshold value. If an action potential is fired, then there is a 1ms
�absolute� period afterward during which the neuron is unable to fire again.
For 5ms following that, the neuron�s threshold is above normal, so that only an
intense stimulus will cause it to fire. By this time, the potassium and sodium
channels have both closed, the electrochemical gradient has been restored and
the vesicles are ready with neurotransmitter to be released, and so the resting
potential has returned to normal (usually about �70mV). This delay sets a cap
of about 500 spikes/second at which the action potentials can be fired, for a
very intense stimulus. Less intense stimuli will only cause the neuron to fire
when the threshold value has had time to lower and approach its normal level,
and at correspondingly larger intervals.
In this way, an analogue sensory input signal is
coded, in terms of its intensity, into a digital frequency code. However, this
does not of course relate anything about the nature of the sensory stimulus
(see below). Frequency-coding has the advantage of degrading less over
distance. In fact, there is enough redudancy for the action potential spike to
pass three nodes of Ranvier before being reinforced by the sodium channels
without the signal being lost, as opposed to an amplitude-coded signal which
would rapidly attenuate and degrade unless propagated at inefficiently frequent
intervals. Consequently, for inter-neural pathways of less than a couple of
millimetres, axon propagation is unnecessary, but a signal can accurately
travel the length of the body if necessary.
In the 18th century, Weber was
trying to relate physical stimulus intensity to psychological sensory magnitude,
requiring a quantification of our subjective internalised measures of sensory
stimulus intensity. He posited that the just noticeable difference (JND) in the
intensity of a stimulus was proportional to the intensity of the original
stimulus, i.e. that we might be just able to perceive the addition of two
candles to two hundred, or four candles to four hundred.
|
DI |
� = c |
|
I |
where I
is the intensity of the standard stimulus, i.e. the one to which comparisons
are being made, DI is
the amount by which this intensity must be altered in order to induce a JND,
and c is a constant.� Even though
the JNDs are in the units of the respective physical stimulus, Weber�s law made
it possible to compare sensory sensitivity between senses.
Weber�s law was found to work reasonably well for
stimuli in the middle range of sensory input, but less for extremes, because it
assumed a direct proportionality between stimulus intensity and neural response
(measurable as the spike rate).
|
Representative (middle-range) values for the
Weber fraction for the different senses |
|
|
Vision (brightness) |
1/60 |
|
Pain |
1/30 |
|
Audition |
1/10 |
|
Taste |
1/3 |
However, our senses have to give us
information about our environment in a very wide range of conditions. For
instance, the variation between pitch blackness on a moonless night and bright
sunshine is a factor of about 1010, yet the maximum spike rate of
action potentials is about 500/second. This is a far greater range than can be
compensated for by dilations and contractions of the pupil whose movement the
brain then subtracts, just as it does not perceive changes in the retinal image
due to movements of the head.
A clue as to how the neural system copes with this
comes from the way in which the threshold value returns to its normal level
over the course of 5ms after a 1ms absolute period of infinite threshold,
implying a non-proportional relationship, with high spike rates indicating even
more extreme stimuli. In order to preserve the level of detail of information
provided by our highly-sensitive eyes and other senses, we have evolved to
logarithmically code sensory information, compressing an enormous range of
inputs into a maximum spike rate. This means that a unitary change in the
frequency signals a tenfold change in the intensity of the stimulus, which is
why Fechner�s law works better than Weber�s proportional law at extreme
intensities:
S = k log10
I
where S =
the subjective magnitude, I for the
physical intensity of the stimulus, k
is a constant whose value depends on the value of the Weber fraction.
Thus, the coding requires only a
single variable, frequency, with the exact relationship between stimulus
intensity and spike rate being non-proportional and modulated according to
trigger region threshold values and inhibitory/excitatory neurons. However, in
order to distinguish a strong smell from a bright light, for example, the
nervous system takes advantage of the enormous numbers of neural pathways
formed. Two theories explain how this network is used to flesh out the code
into our rich internal representations.
Specificity theory explains different sensory
qualities as being tagged to individual neurons. Thus, �redness� or �sourness�
would be signalled by the spike rate of the �red� neuron or the �sour� neuron.
In contrast, the pattern theory defines sensory qualities in terms of a
collection of pathways, so that the sensation of �sourness� might be elicited
by the parallel workings of a group of connected neurons. Certainly, many
regions of the brain may use patterns of firing for coding, such as language
and memory, which explains the lack of succes scientists have had explaining
the brain�s mechanics in spatially localised terms.
In conclusion, the nervous system
uses an extremely efficient form of transmission, in the form of frequency-coded
action potentials. The sensory resolution (i.e. the size of the JND) varies
from sense to sense, probably according to the survival value of the sense,
hence high sensitivity for vision and pain, but lower sensitivity for taste or
pressure. To take into account the enormous range of intensity the senses have
to cope with, the signals are logarithmically-coded, with the pathways of
firing being important in determining the nature of the sensory quality.