Many brain functions are thought to rely on collective behaviours of large numbers of interacting neurons. This makes statistical mechanics a valuable toolbox in the attempt to understand neural systems. This talk with focus on memory and the celebrated "attractor neural network" paradigm. After a general introduction, I will explain how these ideas can be applied to the so-called place cells in the hippocampus, a neural system thought to support the memory of space. A model is proposed that can be studied in great details with analytical methods commonly used for disordered systems. The various collective behaviours it exhibits are thought to be related to recent experimental results.