Section Summary

Our visual sense allows us to reconstruct the world around us, which seems effortless and automatic, but in fact requires our eyes and brain to engage in extensive neural processing to allow us to perceive our surroundings. Light reflected from surfaces is imaged by the eye’s optical system, the cornea and lens, onto the retina at the back of the eye, where photoreceptors detect photons of light energy. This begins a process that leads to the visual cortex and adjacent visual areas in the occipital and temporal lobes of the brain.

In the first steps in vision, rods and cones capture photons to create an electrical signal that is proportional to the number of photons captured. This process, called "phototransduction," is unusual in that the receptors are depolarized in the dark but become less depolarized (hyperpolarized) in the light. When visual pigment molecules capture photons, it leads to a reduction in the concentration of a signaling molecule, cGMP, which when present activates Na+ channels to allow Na+ ions to enter the photoreceptors and depolarize them. Light reduces cGMP levels, closing Na+ channels and ending depolarization. Our color vision depends on having three different visual pigments in different classes of cones, since a single pigment cannot distinguish between a dim colored light that is absorbed well and a bright light of a different color that is absorbed poorly. Our color vision is trichromatic, but most mammals have only two cone visual pigments and see colors differently from people.

Rods and cones respond to the brightness of the light hitting them, but neurons in the other layers of the retina discard brightness information and convert the information that leaves the eye into the location of edges between light and dark or borders between different colors. Edge-detection begins with bipolar cells, which have circular receptive fields that respond to dark or light spots in the center of the receptive field, with contrasting illumination in the receptive field surround. Uniform illumination that covers both center and surround does not excite bipolar cells. Ganglion cells receive synapses from bipolar cells and inherit their center-surround receptive fields. Ganglion cells are the first neurons in the processing chain that make action potentials, which travel to the brain in the optic nerve. Ganglion cells responsive to colors have opponent-color receptive fields, excited by one color and inhibited by its complement. The complementary, opponent-color pairs are red/green and blue/yellow.

The receptive fields of LGN neurons are circular, but neurons in primary visual cortex V1 have elongated receptive fields with a straight-line edge between the subdivisions. "Simple" cortical cells can be mapped with small spots, but "complex" cells require elongated stimuli such as light or dark bars. The elongated edge can be placed anywhere in a complex cell’s receptive field as long as its orientation angle matches the cell’s preferred orientation. Some cortical cells are responsive to stimuli moving in one direction but not the reverse. Some cells decrease their response if the stimulus bar extends beyond the receptive field ("endstopping"). Spatial frequency gratings at an appropriate angle are more effective stimuli than bars of equivalent width. The cortex’s functional architecture reveals that neurons with the same orientation preference are grouped in vertical orientation columns that when viewed from the cortical surface appear like pinwheels. Neurons driven most strongly by one eye or the other are grouped in broad elongated ocular dominance columns. Examination of cortical layers shows that simple cells are found principally in middle layer 4 where LGN axons arrive, while complex cells are found in the upper and lower layers. Finally, some cortical neurons are selective for colors in addition to having a preferred location, angle and direction of movement.

Neurons in V1 begin cortical visual processing by detecting edges in small regions of the image. The next steps occur in adjoining cortical areas, the extrastriate cortex, where two pathways have been identified: the dorsal stream detecting location and movement, and the ventral stream leading to the inferotemporal cortex (IT), where images of objects elicit responses. Neurons in IT that respond selectively to faces are grouped in face patches, which provide an example of visual processing for a well-defined category of images.

Although the first stages of visual perception seem to be well understood, beginning in the retina and proceeding in steps though V1 and extrastriate cortex, and continuing into IT, the final stages of perception remain a mystery. One problem is how responses to a particular object, activating neurons in many different visual areas, are grouped to identify that object and not other objects in the scene. One theory involves the timing of action potentials, where rhythmic firing or synchrony may establish connections for each object, but this idea remains controversial. Another problem is how small components within a scene are organized into the perception of the full object, a process that seems to require prior knowledge of the object's shape. This suggests a role for top-down processing. One clear example of top-down modulation of earlier neural activity is visual attention, which organizes bottom-up components into a perceived scene. A bottom-up issue is converging activity, where many neurons with small receptive fields activate the next level of neurons with larger receptive fields, raising the question of how much convergence ultimately occurs. Could there be single neurons responsible for perceiving a particular object, so-called "grandmother cells" responsible for perceiving your grandmother? It seems much more likely that any particular object activates an ensemble of neurons, with different but possibly overlapping ensembles for other objects. These many open questions indicate that the final stages of visual processing are still an unsolved puzzle.