Spike Train Correlations: Measuring Neural Interactions in the Brain

How can we tell whether neurons are coordinating their activity, rather than simply firing at similar rates or responding to the same event?

In this hands-on session, we will explore statistical methods for finding correlations in neuronal spike trains. We will work with simulated data, where we know the underlying correlation structure, as well as multi-trial recordings from macaque motor cortex.

We will begin by visualising and describing spike trains, then use cross-correlation histograms to look for pairwise relationships. A correlation peak is not automatically evidence of a meaningful interaction, so we will compare different surrogate methods and discuss how the choice of null model affects our conclusions.

Later, we will move beyond pairs of neurons and look at approaches for detecting synchronous activity and recurring patterns across larger populations. Along the way, we will experiment with analysis choices such as bin size, time window and firing-rate stationarity.

The aim is not to treat these methods as black boxes. You will change parameters, compare results and build an intuition for what each analysis can and cannot tell us.

Tools

• Elephant: analysis tools for electrophysiological data in Python
• Neo: data structures for representing electrophysiology recordings