Abstract The overarching goal of this proposal is to obtain general principles of dendritic computation. The starburst amacrine cell (SAC) of the mammalian retina is an excellent model to study dendritic function because it has a rich set of generic dendritic processing machineries and a well-defined computation – direction selectivity. Its radially oriented dendritic sectors are tuned to different directions of visual motion stimuli, and thereby confer direction selectivity to the output neurons of the retina. Previous studies have generated substantial knowledge on the spatiotemporal patterns of synaptic inputs onto the SAC dendritic arbor. However, how motion-evoked inputs are transformed by SAC dendrites to generate robust outward direction selectivity is not well understood. This application therefore focuses on the dendritic mechanism and how it is influenced by the synaptic input. We will obtain the objective of the proposed research using a combination of patch clamp recording, two-photon calcium, glutamate and voltage imaging during visual stimulation, and behavioral assay. Insights from this proposal will challenge the conventional view that synaptic inputs are processed by dendrites equipped with a stable set of intrinsic biophysical properties. Our study will advance the field of dendritic computation by establishing a more dynamic relationship between synaptic activity and the algorithm of dendritic integration.