The apparent motion of features in a visual scene — optic flow — provides rich information about the animal’s self-motion and its environment [Gibson 1950]. An animal estimates the optic flow with motion sensitive neurons. In flies, the T4/T5 neurons encode small-field directional motion signals by sampling neighboring eye facets. The signals are then integrated by large-field neurons, which are tuned to complex global patterns of optic flow. Are these patterns simply inherited from their T4/5 inputs or synthesized by a more complex computation? Near the center of the eye, the 4 T4 subtypes have been shown to encode motion along the 4 cardinal directions — forward, backward, up and down [Maisak, et al 2013; Takemura, et al 2013]. In this case, a simple summation would generate a translational flow pattern. However, it’s not possible for all T4’s locally preferred direction (LPD) to follow the cardinal directions across the eye while maintaining a uniform sampling of the eye’s hexagonal grid. We set out to systematically describe, for the first time, the global organization of the local directionally selective neurons using computational neuroanatomy of the FAFB EM data set [Zhang, et al 2017]. We reconstructed hundreds of T4 neurons and determined their dendritic orientation (proxy for LPD). To compare the organization of LPDs to the eye structure and to global optic flow fields, we developed an “eye map” using uCT of an entire fly head to register the neuronal coordinates of the EM reconstructions into eye coordinates. We found that T4 neurons are mostly aligned to the local hexagonal grid, but this grid maps onto the eye with systematic spatial variations. This mapping has pronounced effects on the local motion sensitivity and subsequently the global motion pattern encoding. Our results demonstrate that the organization of the sensory apparatus substantially informs neural computations.