Cortical Variability Dynamics

Neural responses are variable: even when the same stimulus or movement is repeated, neurons do not fire the same spike train every time. This unit introduces cortical variability as a central feature of neural activity, rather than simply as noise to be ignored.

In the homework, you will first explore variability within individual spike trains. Using motor cortex recordings and simulated spike trains, you will examine inter-spike interval distributions, compare real neurons to Poisson and Gamma process models, and quantify firing irregularity with measures such as the coefficient of variation and local variation.

The first in-person session focuses on spike timing variability. You will work with real movement-aligned spike trains, visualize responses with rasters, and learn how different variability measures reveal different aspects of neural firing. A key idea is that local measures such as Lv can distinguish intrinsic spike timing irregularity from slower changes in firing rate.

The second session shifts from variability within a trial to variability across trials. You will analyze spike count distributions, compute the Fano factor, and compare count variability before and after movement onset. You will also examine time-resolved Fano factor estimates to see how variability changes throughout a behavioral event.

The aim of this unit is to connect statistical models of spike generation with real cortical data. By comparing Poisson, Gamma, and doubly stochastic processes, you will learn how different sources of variability can be separated, and why cortical neurons often show more trial-by-trial variability than simple renewal models predict.