How brain activity gives rise to our thoughts, feelings and conscious awareness lies at the heart of understanding what it means to be human. Neurons are the fundamental building blocks of the brain and understanding how the brain controls behavior ultimately involves understanding the activity of individual neurons. The problem is that our brains contain many neurons connected in countless ways, and our ability to probe each neuron individually in the working brain is woefully limited. Progress in understanding the brain depends on finding the right way to simplify. In particular, we need to effectively reduce the complexity of the brain by understanding how neurons work together in groups and ensembles.
In order to make a seemingly intractable problem tractable, we need to simplify by exploiting regularities, correlations, which exist in the activity of groups of neurons. Correlations reflect the presence of patterns in the firing of ensembles of neurons.
One of my earliest contributions was to show how individual neurons contribute to neural ensembles. I did this by measuring how well the time of each action potential can be predicted from local field potential (LFP) activity. LFP activity is an electrical potential generated by current flow in the vicinity of a recording electrode. When I started my work on LFP activity in the monkey, the LFP received relatively-little serious attention because it is unclear why it is important. Neurons are the building blocks of the brain, not LFPs. My contribution has been to show that by estimating spike-field coherence – the coherence between the spike train of a neuron and the local field potential (LFP) – I can reveal whether a given neuron does or does not display correlated activity with groups of other neurons. Thus, coherence simplifies because it provides a definition for how neurons can be grouped into ensembles: Group neurons according to whether or not they show significant coherence with LFP activity. I can also identify neurons in one part of the brain that show coherence with LFP activity in another part of the brain. As I will describe below, coherent ensembles perform particular computations that guide behavior, and ensembles of neurons that do not fire coherent patterns of spikes do not perform these computations. As a result, I propose that LFP activity is important because it can reveal, and even help us understand, how neurons form ensembles.