HOW THE BRAIN MAKES POTENTIAL 177
now, in the brain, they extend across many more variables—in this case,
light points—at much greater statistical depth.
Take almost any pair of points in fi gure 6.1 and follow them across the
fi ve frames. You may soon be aware that their movements vary together,
or covary, over space and time. (More formal mea sures confi rm that this
is the case.) Th e association may be rather weak, but we can say that
this is a primary association. And the association— call it pa ram e ter 1—
is information for predicting the position of one from the other in its
absence (e.g., when it is obscured behind another object).
Th e predictability can be strengthened, however, if the association de-
pends on (i.e., interacts with) levels of another variable. We say this is a
second- level association (or fi rst- level interaction, or de pen dency). Th is is
further information for predictability. Call it pa ram e ter 2: an association
being informed by the value of another variable.
Indeed, we can carry this analy sis through to many other levels or
depths. Th e interaction just mentioned can itself be conditioned by one
or more other variables— and so on, across the whole picture of moving
dots to reveal still further accumulations of useful information. So the
whole experience exhibits par ameters 3, 4, and so on, embedded in
one another at increasing levels or depths. Th ose par ameters are a rich
source of predictability for a system that can register them.
Indeed, estimation of such par ameters in the point- light walker (or
point- light repre sen ta tions of other objects) shows them to be statistically
signifi cant up to the fourth and even higher levels of de pen dency. And
this is only in the nine spots of light, compared with the thousands or
more that would radiate from a whole object.^9 Th e upshot is that an ob-
ject, presenting whole “clouds” of moving points, will proj ect abundant
information for creation of a recognizable image. Th is is impor tant for
its implications about the nature of brain pro cesses and cognition.
So the information captured in the nerve networks is not in the form
of photo graphs or movies. What is stored are those par ameters coding
statistical relations among variables. It is recordings of those par ameters
(by means to be described below) that form a visual attractor to which
visual images bearing them will tend to be pulled. A fl eeting pattern in
the corner of your eye makes you think you have seen a familiar object; a
faint pattern across the street looks like a familiar face; a smudged word
This content downloaded from 139.184.14.159 on Tue, 17 Oct 2017 13:54:00 UTC