A neuron can only do two things: Fire. Don’t fire. Firing is the neuron’s output, and most theories of neural computation treat that output as the brain's core information signal. 🧵

Yet most of our macroscopic recordings — EEG, MEG, ECoG, LFPs — mainly reflect postsynaptic transmembrane currents, i.e., the inputs to neurons. How do we reconcile measuring inputs while caring about outputs?

Frustrated that I couldn’t measure firing directly in LFPs, I reviewed the studies that recorded both. To my surprise, a consistent pattern emerged: More firing → more broadband power.

In our new paper, we separated oscillatory and aperiodic components and found that observed broadband shifts are mainly aperiodic – driven by flatter 1/f slopes and larger offsets. Therefore, we termed it “aperiodic broadband power”.

Even simpler, band-averaged gamma power closely tracked aperiodic broadband power (ρ = 0.93). Taken together, the evidence suggests that invasively recorded gamma power can serve as a practical proxy for population firing. Open-access paper: doi.org/10.1016/j.eb...

Full slide deck: