suppose that the inhibitory path involves a delay, perhaps produced by still another intermediate cell. Then, if the stimulus moves in one direction,
say, right to left, as in the illustration of Barlow and Levick's model,
the intermediate cell is excited by one of its inputs just as the inhibition
arrives from the other, whose field has just been crossed. The two effects cancel,
and the cell does not fire. For left-to-right movement, the inhibition arrives
too late to prevent firing. If many such intermediate cells converge on a
that cell will have the properties of a directionally selective complex
We have little direct evidence for any schemes that try to explain the
behavior of cells in terms of a hierarchy of complexity, in which cells at
each successive level are constructed of building blocks from the previous level.
Nevertheless, we have strong reasons for believing that the nervous system
is organized in a hierarchical series. The strongest evidence is anatomical:
for example, in the cat, simple cells are aggregated in the fourth layer of
the striate cortex, the same layer that receives geniculate input, whereas the complex
cells are located in the layers above and below, one or two synapses further
Thus although we may not know the exact circuit diagram at each stage,
we have good reasons to suppose the existence of some circuit.
The main reason for thinking that complex cells may be built up from center-surround cells, with a step in between, is the seeming necessity
of doing the job in two logical steps. I should emphasize the word logical because
the whole transformation presumably could be accomplished in one physical
step by having center-surround inputs sum on separate dendritic branches of
complex cells, with each branch doing the job of a simple cell, signaling
electrotonically (by passive electrical spread) to the cell body, and hence to the
whenever a line falls in some particular part of the receptive field.
The cell itself would then be complex. But the very existence of simple cells suggests
that we do not have to imagine anything as complicated as this.
SOME COMMENTS ON
HOW WE SEE
Why are movement-sensitive cells so common? An obvious first guess is that they tell us if the visual landscape contains a moving
object. To animals, ourselves included, changes in the outside world are far more
important than static conditions, for the survival of predator and prey alike.
It is therefore no wonder that most cortical cells respond better to a moving
object than to a stationary one. Having carried the logic this far, you may
now begin to wonder how we analyze a stationary landscape at all if, in the interests
of having high movement sensitivity, so many oriented cells are insensitive
to stationary contours. The answer requires a short digression, which takes
us to some basic, seemingly counterintuitive facts about how we see.
First, you might expect that in exploring our visual surroundings, we
let our eyes freely rove around in smooth, continuous movement. What our two
eyes in fact do is fixate on an object: we first adjust the positions of
our eyes so that the images of the object fall on the two foveas; then we hold that position
for a brief period, say, half a second; then our eyes suddenly jump to a new
position by fixating on a new target whose presence somewhere out in the visual
field has asserted itself, either by moving slightly, by contrasting with
the background, or by presenting an interesting shape. During the jump, or saccade,
which is French for "jolt", or "jerk" (the verb),
the eyes move so rapidly that our visual system does not even respond to the resulting movement of
the scene across the retina; we are altogether unaware of the violent change.
(Vision may also in some sense be turned off during saccades by a complex
circuit linking eye-movement centers with the visual path.) Exploring
a visual scene, in reading or just looking around, is thus a process
of having our eyes jump in rapid succession from one place to another.
Detailed monitoring of eye movements shows vividly how unaware we are of any of this. To monitor eye movements we first attach a tiny mirror
to a contact lens, at the side, where it does not block vision; we then reflect
a spot of light off the mirror onto a screen. Or, using a more modern version
invented by David Robinson at the Wilmer Institute at Johns Hopkins, we
can mount a tiny coil of wire around the rim of a contact lens, with the
subject seated between two orthogonal pairs of bicycle-wheel size hoops containing coils of wire; currents in these coils induce currents in the contact-lens
which can be calibrated to give precise monitoring of eye movements.
Neither method is what you would call a picnic for the poor subject.
In 1957, Russian psychophysicist A. L. Yarbus recorded eye movements
of subjects as they explored various images, such as a woods or female
faces (see the illustrations below), by showing the stopping places
of a subject's gaze as dots joined by lines indicating the eyes' trajectory
during the jumps. A glance at these amazing pictures gives us a world of information about our vision—even about the objects and details that interest
us in our environment.
So the first counterintuitive fact is that in visual exploration our
eyes jump around from one point of interest to another: we cannot explore a stationary scene by swinging our eyes past it in continuous movements. The visual
system seems intent instead on keeping the image of a scene anchored on
our retinas, on preventing it from sliding around. If the whole scene moves
occurs when we look out a train window, we follow it by fixating on
an object and maintaining fixation by moving our eyes until the object gets out
whereupon we make a saccade to a new object. This whole sequence—
following with smooth pursuit, say, to the right, then making a saccade
to the left—is called nystagmus. You can observe the sequence—perhaps
next time you are in a moving train or streetcar—by looking at your neighbor's
eyes as he or she looks out a window at the passing scene—taking care
not to have