The retina is much more than a transducer of photons to action potentials. Significant processing occurs before visual information exits the eye and enters the brain. Photoreceptors reduce the amount of neurotransmitter they release in response to photons of certain wavelengths striking and activating them. From this basic input, significant divergence can be seen in the firing of action potentials by different subtypes of ganglion cells which form the output of the retina (see Figure 2).
One illustrative example, described by Barlow and Levick, is the phenomenon of direction selectivity (DS). This form of neural computation manifests in a population of ganglion cells (DSGCs) which respond preferentially to stimuli moving across their receptive field in one of the four cardinal directions. For reference, a receptive field is the location in which light shone on the retina can stimulate the cell. Typically, this area corresponds closely to the spread of a cell's dendritic arbor (the small thin processes which resemble branches of a tree). Movement in the opposite to the preferred direction produces little to no response.
On-Off DSGCs respond to both the leading and trailing edge of a stimuli (either light or dark stimuli). You can see this in the animated image below (Figure 3). The vertical lines represents the number of action potentials. The neural circuit responsible for this response is primarily composed of bipolar cells (excitatory), starburst amacrine cells (inhibitory), and the direction selective ganglion cells themselves.
Previous studies have found that the excitatory drive from bipolar cells is not direction whereas the inhibitory input from starbursts is seen only for movement in the null direction. The manner in which starbursts compute direction is the focus of my research. Look to my list of publications to see my work.