Encoding of saccadic scene changes in the mouse retina

The task of the visual system is to extract behaviourally relevant
information from the visual scene. A common strategy for most
animals ranging from insects to humans is to constantly reposition
gaze by making saccades within the scene. This ‘fixate and saccade’
strategy seems to pose a challenge, as it introduces a highly
blurred image on the retina during a saccade, but at the same time
acquires a ‘snapshot’ of the world during every fixation. The
visual signals on the retina are thus segmented into brief image
fixations separated by global motion. What is the response of a
ganglion cell to ‘motion blur’ caused by a saccade, and how does it
influence the response to subsequent fixations? Also, how does the
global motion signal influence the response dynamics of a ganglion
cell? In this thesis, we addressed these questions by two
complementary approaches. First, we analysed the retinal ganglion
cell responses to simulated saccades. We analysed two important
aspects of the response - 1) response during a saccade-like motion,
2) response to fixation images. For about half of the recorded
cells, we found strong spiking activity during the saccade. This
supports the idea that the retina actively encodes the saccade and
may signal the abrupt scene change to downstream brain areas.
Furthermore, we characterized the responses to the newly fixated
image. While there appears to be only little influence of the
preceding motion signal itself on these responses, the responses
depended strongly on the image content during the fixation period
prior to the saccade. Thus, saccadic vision may provide ‘temporal
context’ to each fixation, and ganglion cells encode image
transitions rather than currently fixated images. Based on this
perspective, we classified retinal ganglion cells into five
response types, suggesting that the retina encodes at least five
parallel channels of information under saccadic visual stimulation.
The five response types identified in this study are as follows: 1)
Classical Encoders - Response only to preferred stimuli; 2) Offset
Detectors - Response only to the saccade; 3) Indifferent Encoders -
Response to all fixated images; 4) Change Detectors - Response only
when the new image after the saccade differs from the previous
image; 5) Similarity Detectors - Response only when the new image
after the saccade is similar to the previous image. Second, we
analysed the influence of global motion signals on the response of
a retinal ganglion cell to the stimulus in its receptive field. The
stimulus beyond the receptive field is designated as remote
stimulus. We chose simple stimulus that represent various
configurations used in earlier studies, thus allowing us to compare
our results. We show that the remote stimulus both enhances and
suppresses the mean firing rate, but only suppresses the evoked
activity. Furthermore, we show that the remote stimulus decreases
the contrast sensitivity and modifies the response gain. Thus, the
ganglion cells encode the stimulus in relation to the whole scene,
rather than purely respond to the stimulus in the receptive field.
Our results suggest that the global motion signals provide ‘spatial
context’ to the response of the stimulus within the receptive
field.