Whether hippocampal cells in non-human primates can encode schemas that can help to generalize between different environments is not clear. Here, the authors trained macaques to use a joystick to navigate a virtual maze to find a hidden reward in a location defined by the relative position of visual landmarks. Once proficient in the first maze, the monkeys were much faster at learning to find the reward in an isomorphic maze with different landmarks. By aligning the firing maps of different hippocampal neurons in the two mazes, the authors found that a subset of neurons encoded the monkey’s position or task state relative to the reward similarly in both mazes. These ‘schema cells’ thus generalize across the two mazes.