214 A CREATIVE COGNITIONIn vertebrates, from reptiles to fi sh to apes, the same story of previ-
ously unrecognized cognitive abilities has been unfolding. And, again, it
is being realized that these are abilities based on deeper structures than
simple cue- response associations. Th e standard computational model
cannot explain how we deal with real, dynamic environments. So let us
see what such structures are and how they are connected with the brain.
Being more clear about this is crucial to an understanding of what varies
in cognitive systems, even at a rudimentary level.
FROM BRAIN TO COGNITIONTh e distinctive demand on cognitive systems is that they fi rst need to be
very good at assimilating those structures. Th is is not only because envi-
ronmental structures are deep in the statistical sense, but also because
they change frequently— more quickly than can be dealt with by epige ne-
tic, developmental, or physiological pro cessing. Cognitive functions must
be able to use such structure to predict immediate and distant futures
from inevitably partial current information.
As seen in chapter 6, the brain is superbly evolved for supporting just
such functions. Incoming stimuli are highly variable and noisy, and so
are the responses of single neurons to them. Yet cognitive experience
of the environment is much more stable and consistent. How is this
achieved?
We have already seen that the brain is a dynamical system, quite dis-
tinct from the linear pro cessor in the computational model of cognition.
It assimilates the structure of experience at vari ous levels, but not in the
form of direct sensory rec ords, such as snapshots or movies. What must
be internalized are the par ameters describing the statistical relationships
in experience. Th ese comprise the grammars of experience. Th is is what
the vast numbers of connections in extensive neural networks are for.
By virtue of such grammars, activity in the network becomes both
constrained and generative. As with the whirl pool in the river (or Bénard
cells, or a million other natu ral structures), they are attractors or basins
of attraction. Inputs consistent with the cognitive grammar— like the
This content downloaded from 139.184.14.159 on Tue, 17 Oct 2017 13:54:24 UTChttp://www.ebook3000.com