Date of Award

Summer 8-15-2021

Author's School

Graduate School of Arts and Sciences

Author's Department

Biology & Biomedical Sciences (Neurosciences)

Degree Name

Doctor of Philosophy (PhD)

Degree Type



In early sensory systems, a copious amount of information needs to be simultaneously processed to extract salient stimulus features and compress representations for efficient transmission. My thesis addresses how the retina processes visual information to detect specific motion patterns and efficiently encode our environment for perception. Around 40 retinal ganglion cell (RGC) types in mammals send specific feature representations through their axons from the retina to the brain. RGCs are the sole source of visual information to the brain. Recent studies suggest that amacrine cells (ACs) generate feature-selective responses of RGCs, but little is known about their neuronal computations. We explored the visual processing of one AC type, the VGLUT3-expressing (VG3-) AC. I found that different dendrites of VG3-ACs encode different information and transmit this information to different RGCs. Dendritic processing of VG3-ACs differentiates complex motion patterns, and a specific layer of the VG3-AC dendrite arbor drives innate defensive responses to approaching stimuli. In addition, I found that VG3-AC dendrites enhance the excitatory input to RGCs for locally sparse motion patterns, for which bipolar cell inputs alone are insufficient, permitting a robust detection of motion in the visual field. Our rich visual perceptions rely on information from the retina, which is at its most compressed in the spike trains of RGCs. How the spike trains of human RGCs encode visual information to support our perception is unknown. We performed large-scale recordings from the human retina, reconstructed RGC response functions, and revealed the visual signals carried by the four types of RGCs (ON and OFF midget and ON and OFF parasol), which account for the majority (~80%) of visual input to the brain. I found that the spatiotemporal receptive fields of these four RGC types cover adjacent areas with equal power in the frequency landscape of natural movies and that asymmetric response functions between individual ON-OFF pairs maximize information transmission and minimize metabolic cost – evidence of the efficient coding in the human retina.


English (en)

Chair and Committee

Daniel Kerschensteiner

Committee Members

Baranidharan Raman, Timothy Holy, Josh Morgan, Carlos Ponce,