Neurobiology MCDB 320/720

Visual Cortex

In primates there are 5 principal topographically organized visuotopic map regions,

numbered V1 through V5, each with its characteristic inputs and functions. The five major
visual maps serve as the source of information for nearly two dozen topographically organized
maps that are found throughout primate occipital, parietal, and temporal cortex. Each map is
specialized for a set of specific image processing tasks. This example of the divergent
dissection of sensory information is not unique to the visual system. A similar dissection of
sensory information occurs for somatotopic representations. Different forms of tactile and
proprioceptive information are topographically mapped to distinct cortical locations.

Humans are supremely visual mammals, with 1/3 of the cortex devoted to processing visual
information. Here we examine the first destination in cortex, known as area V1. We begin with
a consideration of the information derived from the retina.

Projections and Cell types

The output neurons of the retina
project to three destinations in the CNS. The
primary output from the mammalian retina
consists of ganglion cells that project to the
thalamus, to ultimately drive activity in the
visual cortex. The retinal ganglion cells can be
divided into two major classes, the P and M Temporal Nasal
neurons, and a third minor class, the K
neurons. The destination of these ganglion
cells is a multilayered structure in the LGN
thalamus known as the lateral geniculate
nucleus (LGN). The LGN neurons then LGN: 6 layers total, 3
from each eye; each
project to the primary visual cortex, in a layer has a contalateral
region known as Brodmann area 17 or V1. visual field
The second group of retinal ganglion
cells project to the superior colliculus, a target Visual cortex :Inputs
in the dorsal midbrain. In most vertebrates, from LGN layers
converge to give
from fish to birds, this region is the main binocular vision
visually responsive and image forming region
of the CNS, and is known as the optic tectum. In mammals many of its visual functions have
been supplanted by the neocortex, and the superior colliculus serves to control eye movements.
The final group of retinal ganglion cells consists of the intrinsically photosensitive retinal
ganglion cells (ipRGCs). These ganglion cells express the pigment melanopsin, and are capable
of transducing light. The cells project to both the superchiasmatic nucleus, a region in the
hypothalamus that is involved in regulating circadian rhythmicity, and to the pretectal nucleus of
the midbrain, to control pupillary reflexes (see handout on adaptive responses).
Ganglion cell classes: The ganglion cells involved in visual image processing have
circular center-surround antagonism and fire action potentials. Three major types are recognized.
1. “Midget” ganglion cells – these neurons constitute the vast majority of the P group,
and in excess of three quarters of the entire population of retinal ganglion cells. These cells have
small receptive fields and small dendritic arbors. They are predominant in the macular region of
Neurobiology MCDB 320/720

the retina, are cone-driven, and are involved in providing the highest level of visual resolution.
Both ON and OFF-center responses are found within this subset of ganglion cells, but the larger
representation is for OFF-center cells. These neurons also have single opponent color responses.
The inner plexiform layer of the retina where these neurons have their dendrites is
organized into two principal layers, corresponding to ON and OFF center ganglion cell dendrites.
Thus the corresponding ON or OFF bipolar cells innervate the corresponding dendritic layers.
The firing properties of the P type ganglion cells tend to be relatively sustained, and the action
potential propagation speeds in the optic nerve are slower than for M type cells.
2. “Parasol” ganglion cells are the major type for the M group. These neurons have large
dendritic arbors, and larger receptive fields than the P type ganglion cells. With their large
dendritic expanse, they can sample a wider number of bipolar cells. They are also cone-driven,
and can be classified into ON- or OFF-center types. However, these neurons differ from the P
ganglion cells in that they receive input from multiple cone cell types, so they have broader
spectral responses, and are not involved in color discrimination. By contrast, with their large
receptive fields they are well suited to respond to motion and directionality. These neurons
include the directionally selective ganglion cells that are innervated by starburst amacrines.
The firing properties of the M class ganglion cells are transient and phasic, as opposed to
the more sustained bursts seen with the P type cells. The propagation speed of the M type cells
within the optic nerve is also faster than the P type cells. This is consistent with their importance
in motion detection, where fast propagation and precise tinimg is critical.
3. Koniocellular ganglion cells, (konio in Greek = “dust”) have small receptive fields and
also small dendrites. Relativedly few in number, the K neurons provide information distinct from
the M and P groups. Their functions are less well understood, but they have counterparts in other
mammals, where they are known as W type ganglion cells. The K cells have been identified by
virtue of their unique expression of several kinases and Ca responsive proteins, they have been
proposed to participate, in part in color vision.
Ganglion cell projections: The retinal ganglion cells involved in imaging project to the
lateral geniculate nucleus (LGN) of the thalamus. Ganglion cells from different sides of the
retina project differently, and this relates to optics. We refer to the side of the retina that is
medial and closest to the nose as “nasal” retina, and the side that is lateral and closest to the
temple as “temporal” retina. Projections from the nasal ganglion cells as a rule cross the midline
in the optic chiasm, and project contralaterally. Ganglion cells from temporal side of the retina
remain ipsilateral.
There are six principal LGN layers, each a topographic map driven by either the P or M
ganglion cells. The eye-source alternates between the layers. The first four layers (going from
dorsal to ventral are driven by P ganglion cells, alternating by eye source as ipsi– contra– ipsi-
and contralateral. The P ganglion cell driven layers are known as the parvocellular layers. The
remaining two layers are driven by M ganglion cells, alternating contra- and ipsilateral eye
source, and are referred to as the magnocellular leyers.
There are in addition six thin K cell driven layers: wherever there is an P or M layer,
there is a K layer ventral to it. The eye source for the K layer cells matches the adjacent layer
dorsal to it.
Each layer provides a topographic map of one-half of the visual world - the half that is on
the contralateral side. The 6 layers remain separate in the LGN, and do not interconnect or
communicate extensively with each other. The parvo- and magnocellular layers respectively
project to two different sublayers of layer 4 of primary visual cortex, V1. The K layers also have
Neurobiology MCDB 320/720

distinct projections: the majority project to layer 4 of the V1 region of visual cortex, and the
balance to the superior colliculus. In conclusion at the thalamus there are 12 distinct retinotopic
maps driven by the P, M and K ganglion cells, each superimposed, layer by layer, in the LGN.
Each map is driven by neurons that respond to light from the contralateral side.

LGN and Cortical Neurons

The parvocellular and magnocellular
neurons of the LGN have response properties
similar to those of retinal ganglion cells:
namely circular receptive fields with center-
surround antagonism. Steven Kuffler first
observed this circular receptive field property
in rabbit ganglion cells, and it was
subsequently found by David Hubel and
Torsten Wiesel for LGN neurons. Circles Lines
When Hubel and Wiesel did their first
recordings from V1, they imagined a
hierarchical arrangement of neurons, where
LGN inputs would converge to generate a
class of “simple cells”, where the cell’s Fig. 2. The left edge of this crescent shaped object can
be represented by a large number of cells with circular
properties would emerge from the combined receptive fields, or by a smaller number of “simple”
properties of its inputs. This would serve to cells with elongated, linear receptive fields.
generate antagonistic surrounds, orientation
specificity, and binocular properties. They then imagined convergence from simple cells to give
other neurons in V1 more specialized properties, and they named these neurons complex and
hypercomplex cells. The basic idea is that the cell’s receptive fields and response properties
were shaped by a feed-forward connectivity, where each cell would function based on the
convergent properties of its cortical inputs.
Simple cells have a receptive field composed of flanking elongated ON and OFF
subfields of a specific orientation. Simple cells respond to bar-shaped stimuli as opposed to the
circular receptive fields of LGN neurons. Multiple LGN neurons make direct excitatory
glutamatergic synaptic connections onto simple cells. Hubel and Wiesel proposed (and this was
later shown by David Ferster) that the simple cell receptive fields are built up by convergence
of LGN neurons with adjacent, partially overlapping receptive fields, aligned in a specific
orientation. Thus LGN neurons whose ON-center receptive fields line up in a row would
coexcite a simple cell, generating a preference for a bright bar that matched the orientation of
the row. Similarly, LGN neurons with center-OFF receptive fields would converge onto a
different simple cell, to generate a cell that favored a dark bar of a given orientation. Ferster
also observed GABAergic IPSPs in simple cells. These potentials could not be coming from
the LGN neurons, which are excitatory, but from within the cortex by local inhibitory
interneurons (see below). The interneuron-derived IPSPs were evoked in opposite fashion to
the LGN-derived EPSPs. Thus a visual stimulus that increases excitation of a simple cell (for
example, a bright bar of light on the cell’s ON subfield, decreases the frequency of IPSPs,
while a stimulus that decreases simple cell excitation (such as a bright bar on the cell’s OFF
subfield), increases the IPSPs. This opponent ‘push-pull” signaling between EPSPs and IPSPs
ultimately depended on the interneurons that communicate between adjacent simple cells of
similar orientation. Since the several subfields of a simple cell’s receptive field all have the
Neurobiology MCDB 320/720

same orientation, the interneurons that help generate the subdomains also share the same
orientation specificity.
The elongated receptive fields of simple cells are in principle well-suited for detecting
the location and orientation of edges and for following contours. Consider the crescent being
represented in Fig. 2 above. Instead of devoting a separate neuron with a circular receptive
field for each point along an edge (shown on the left), the CNS can devote a smaller number of
cortical cells with elongated receptive fields (as shown on the right).
Simple cells converge onto “complex” cells, whose receptive fields are larger, and do
not usually show the single flanking antagonistic subfields of simple cells. New properties
emerge in the complex cells. The neurons show preferences for movement and direction. They
also show properties such as “end-stopping”, where the cell responds to line stimuli of a
specific length.
Two important features of V1 were not immediately apparent in the original studies
based on extracellular recordings in the 1960s. First, there is a substantial corticothalamic
projection from V1 back to the LGN. This output from cortex to thalamus originates with
pyramidal cells, mostly in layer 6, and projects in an orderly fashion within the LGN. This
retrograde cortex-to-LGN projection suggests a modulatory feedback function, where primary
visual cortex would control the activity of the inputs it receives. It is reported that this helps
sharpen the boundaries of receptive fields, and may also synchronize the activity of neurons
responding to similar visual features. A similar feedback circuitry is known to exist for other
afferent systems, such as for auditory and somatosensory information. A more speculative idea
is that other regions of cortex may also modulate the activity of the visual cortex, by
influencing the activity of the LGN. For example, animal locomotion has been found,
unexpectedly, to modulate the activity of visual neurons.
Second, the processing of information within the visual cortex involves local inhibitory
interneurons that innervate pyramidal cells (such as the simple cells discussed above) as well as
each other. The inhibitory neurons of the cortex can be grouped into three major classes by
virtue of molecular markers that distinguish them, and are known as PV, VIP, and SST local
circuit interneurons. The interneurons form a local circuit with pyramidal cells, and in addition
to expressing GABA, they release a variety of neuromodulatory peptides. The inhibitory
interneurons also are subject to modulation by neurotransmitters such as serotonin. How these
interneuronal circuits control local processing within the cortex is an important area of current
research. Each type of interneuron projects preferentially to distinct domains of the pyramidal
cell, such as the basal (somatic) vs apical dendrites, where they would exert different effects.
Also as theinhibitory neurons also innervate each other (e.g. VIP neurons often strongly inhibit
SST cells, that in turn inhibit pyramidal cells). Thus their network interactions can lead not
only to pyramidal cell inhibition but also a release from inhibition, known as disinhibition,
which can is excitatory. As discussed above, the simple cell receptive fields with their
antagonistic flanking subdomains are due to these inhibitory interactions, as are the responses
of more complex cells with properties such as directionality, as proposed by Barlow and
Levick.

Cortical organization
Another specialization of primate visual cortex concerns the parvocellular blobs. Blobs
are cylindrical patches of neurons in layers 2 and 3 that are concerned with color perception, and
were discovered by Margaret Wong-Riley. The functions of blobs were discerned by David
Hubel and Margaret Livingstone. Among the blob inputs are color-responsive neurons in the P
Neurobiology MCDB 320/720

and K pathways. In the blobs, and also in color stripes found in area V2, we find color sensitive
“double- opponent” neurons , that respond to two colors in opponent fashion in their center. Thus
the cell might be excited in the center by red and inhibited by green, and respond in a reverse
fashion in its surround, excited by green and inhibited by red. Yellow-blue relationships also are
found. In between the blobs, the so-called interblob regions, are parvocellular cells with
preference for contrast and edge detection.
Another category of binocular cells in V1 responds to retinal “disparity”, a strategy to
determine the distance of objects, conferring the sense of 3 dimensionality to objects
(stereopsis). Ideally, we position our eyes so that the focused image of an attended object falls
on homologous retinal surfaces, with the retinal output from those points driving binocular
neurons in V1. However, in addition to the binocular cells that respond to the homologously
matched retinal locations (referred to as zero retinal disparity), there are binocular neurons in V1
that are driven by excitation of retinal sites on the two eyes that are slightly medial, or
alternatively slightly laterally displaced from the homologous positions (usually by less than 2
degrees). This retinal disparity occurs optically when light arrives from locations behind or in
front of the object observed. Retinal disparity is the basis for local depth perception.
The hypercolumn: For each location in visual space there is a corresponding local group
of neurons in V1 that share overlapping receptive fields, where each neuron is specialized for
one or more kinds of visual information, such as contours, orientation, color, motion,
binocularity, and/or stereopsis. This local ensemble of cells, with its characteristic internal
circuitry, would constitute an image processing unit or “module”. Thus for each location in
visual space there is a corresponding cortical module. Hubel and Wiesel called the module the
hypercolumn, and proposed that it would be the fundamental processing unit for orientation
and binocularity responses. The orientation and ocular dominance columns they described in
cortex are thus made up of adjacent hypercolumns, arrayed across the surface of visual cortex
to form the retinotopic map.
Not all of the needed visual processing can be crammed into a single visual map. The
solution mammals use is to have many parallel maps in cortex, where each map has a
specialized function. There are some two dozen visual maps in primate cortex, which means
that there is also an extensive fragmentation of visual information. Thus, color information is
in one place, movement in another, object identity in a third map, and so forth. How the
information in these multiple maps is kept synchronized temporally and combined “mentally”
to generate a unified perception remains the major challenge of visual neuroscience.

HK May 2022