rithm for assembling this sparse network; animals with much larger brains and higher cognitive abilities (such as apes) do not have proportionately larger swaths of their genome devoted to neural function than animals with more modest brains and abilities (such as rodents). Perhaps it is unsurprising that neuroanatomists have not found a hopelessly tangled, arbitrarily connected network, completely idiosyncratic to the brain of each individual, but instead a great deal of repeating structure within an individual brain and a great deal of homology across species.

At the surface of the brains of all
mammals is a sheet of tissue a few
millimeters thick called the cerebral
cortex, or simply cortex, thought to
be the seat of higher cognition. It is
organized at multiple scales, includ-
ing layers running horizontally, verti-
cally integrated columns spanning its
depth, large functionally defined ar-
eas consisting of millions of columns,
and ultimately networks of multiple
areas that are tightly linked. Within
this system, neurons are connected
locally through gray-matter connec-
tions, as well as through long-range
white-matter connections that leave
the cortex to travel to distant cortical
regions or sub-cortical targets (see
Figures 1 and 2).

figure 3. a circuit diagram of the thalamocortical system simulated on the C2 cortical simulator.

L2/3

Cortical
Layers
To other
cortical
areas

L4

L5

From other cortical areas

L6

Cortical neurons in our
circuit are organized
horizontally into
four cortical layers
and vertically into
hypercolumns (dashed
circles). only strong
connections within and
between these popula-
tions are shown for
various neuron types.
These connections were
derived from dozens of
sources, including the
most detailed measure-
ment to date of cortical
gray matter. 6

Thalamus Reticular Nucleus

Cell types

Excitatory
Pyramidal
Excitatory non-pyramidal
Inhibitory
Double-bouquet
Basket/thalamic inhibitory

collected by measurements at the columnar scale has been instrumental in creating our large-scale brain models, as in Figure 3.

Cortical columns organize into cortical areas that are often several millimeters across and appear to be responsible for specific functions, including motor control, vision, and planning. Suggesting the possibility of a specific cortical circuit for each function, the famous Brodmann atlas, Localization in the Cerebral Cortex, offers a segmentation of the brain into cortical areas based on cellular density variations within the six cortical layers. 18 For example, Brodmann area 17 has been definitively linked to core visual-process-ing functionality. Decades of work by hundreds of scientists have focused on understanding the role each cortical area plays in brain function and how anatomy and connectivity of the area serve that function.

While overwhelming evidence in the 20th century supports the functional specialization of cortical areas, the brain also demonstrates a remarkable degree of structural plasticity. For example, it has been demonstrated that an area normally specialized for audition can function as one specialized for vision, and vice versa, by rewiring the visual pathways in the white matter to auditory cortex and the auditory pathways to visual cortex in the developing ferret brain. This astonishing natural reconfigurability gives hope that the core algorithms of neurocomputation are independent of the specific sensory or motor modalities and that much of the observed variation in cortical structure across areas represents a refinement of a canonical circuit; it is indeed this canonical circuit we wish to reverse engineer. The existence of such a canonical microcircuit is a prominent hypothesis, 29 and while a great deal about the local cortical wiring has been measured, 6 the exact form of this microcircuit remains unknown and its role in neurocomputation undemonstrated. Even if a base canonical circuit can be found, to unlock its potential we must also identify and implement the accompanying plasticity mechanisms responsible for tailoring, refining, and elaborating the canonical circuit to its specific function during development and in adult