Safe navigation of autonomous agents in human centric environments requires\nthe ability to understand and predict motion of neighboring pedestrians.\nHowever, predicting pedestrian intent is a complex problem. Pedestrian motion\nis governed by complex social navigation norms, is dependent on neighbors'\ntrajectories, and is multimodal in nature. In this work, we propose SCAN, a\nSpatial Context Attentive Network that can jointly predict socially-acceptable\nmultiple future trajectories for all pedestrians in a scene. SCAN encodes the\ninfluence of spatially close neighbors using a novel spatial attention\nmechanism in a manner that relies on fewer assumptions, is parameter efficient,\nand is more interpretable compared to state-of-the-art spatial attention\napproaches. Through experiments on several datasets we demonstrate that our\napproach can also quantitatively outperform state of the art trajectory\nprediction methods in terms of accuracy of predicted intent.\n