Objective. Epiretinal stimulation, which uses an array of electrodes implanted on the inner retinal surface to relay a representation of the visual scene to the neuronal elements of the retina, has seen considerable success. The objective of the present study was to quantify the effects of multi-electrode stimulation on the patterns of neural excitation in a computational model of epiretinal stimulation. Approach. A computational model of retinal ganglion cells was modified to represent the morphology of human retinal ganglion cells and validated against published experimental data. The ganglion cell model was then combined with a model of an axon of the nerve fiber layer to produce a population model of the inner retina. The response of the population of model neurons to epiretinal stimulation with a multi-electrode array was quantified across a range of electrode geometries using a novel means to quantify the model response - the minimum radius circle bounding the activated model neurons as a proxy for the evoked phosphene. Main results. Multi-electrode stimulation created unique phosphenes, such that the number of potential phosphenes can far exceed the number of electrode contacts. Significance. The ability to exploit the spatial and temporal interactions of stimulation may be critical to improvements in the performance of epiretinal prostheses.