Universal statistics of hippocampal place fields across species and dimensionalities

Hippocampal place cells form a spatial map by selectively firing at specific locations in an animal’s environment ¹ . Until recently the hippocampus appeared to implement a simple coding scheme for position, in which each neuron is assigned to a single region of space in which it is active ¹ . Recently, new experiments revealed that the tuning of hippocampal neurons to space is much less stereotyped than previously thought: in large environments, place cells are active in multiple locations and their fields vary in shape and size across locations, with distributions that differ substantially in different experiments ^2–7 . It is unknown whether these seemingly diverse observations can be explained in a unified manner, and whether the heterogeneous statistics can reveal the mechanisms that determine the tuning of neural activity to position. Here we show that a surprisingly simple mathematical model, in which firing fields are generated by thresholding a realization of a random Gaussian process, explains the statistical properties of neural activity in quantitative detail, in bats and rodents, and in one-, two-, and three-dimensional environments of varying sizes. The model captures the statistics of field arrangements, and further yields quantitative predictions on the statistics of field shapes and topologies, which we verify. Thus, the seemingly diverse statistics arise from mathematical principles that are common to different species and behavioral conditions. The underlying Gaussian statistics are compatible with a picture in which the synaptic connections between place cells and their inputs are random and highly unstructured.