The local field potential (LFP) is extracted by placing a extracellular microelectrode in the middle of a group of neurons without being too close to any particular cell. When ion channels open and close, charged molecules diffuse around. This changes the electrical potential of the medium surrounding the cells, which the electrode detects and transduces.
Researchers then can reduce the amplitude of any signals with higher than ~ 3oo Hz (oscillations / second) via a low-pass filter, which should remove rapid fluctuations in the electrical potential like action potentials. Instead of APs, the local field potential (LFP) is meant to measure relatively slower moving electrical currents in the surrounding area. Mainly, LFPs should reflect changes in the membrane potential of postsynaptic neurons following the binding of a neurotransmitter to a receptor at the postsynaptic density.
This is an empirical question though, and one that has been investigated more of late. With this in mind, David et al (here) correlated activity between spiking events in the high frequency filtered band (greater than ~ 600 Hz) with activity in the LFP band. They recorded from the primary auditory cortex of passively listening ferrets with tungsten microelectrodes with resistances of 1–5 MΩ.
The following figure indicates that there is a correlation between single unit (i.e. neuron) activity (SUA; denoted via green circles and blue x’s) and raw LFP activity (black curve in middle). The researchers “cleaned” the LFP signal by removing the parts predicted by the spiking activity, this is shown via the blue and green curves and the differences between the raw and cleaned curves is shown at bottom:
So, there is sometimes a correlation between single unit activity and LFP recordings, which introduces a bias into the LFP activity and should make researchers a bit wary. Nevertheless, LFP studies have yielded some interesting insight. For example:
1) Groups of neurons in spatially distinct areas of the cortex coordinate their activity via oscillatory synchronization. In their highly cited paper, Gray and Singer (here) recorded from the primary visual cortex (BA17) of kittens. When they passed a light bar stimulus of a particular orientation and direction through the receptive field of their recorded regions, the amplitude of multi-neuron activity and LFPs tracked the optimal light bar orientation in 14 of 25 trials*. The following chart shows recordings when the light bar is passed through the receptive field (open squares / above) and baseline recordings at the start of each trial (filled squares / below):
Because the multi unit activity and LFP responses are so similar, the authors conclude that neurons involved in the generation of oscillations are restricted to a small volume of cortical tissue and are probably organized in a single cortical column.
* These were the 14 trials in which the MUA response matched the LFP response, the other ones were passed off as anomalies.
2) Propagation through cortical networks can occur via neuronal avalanches. Beggs and Plenz (here) recorded from cell cultures prepared from rat sensory cortex and grown on multi-electrode arrays with 60 channels. They used the LFP definition as negative population spikes. The LFPs were at least 24 ms apart at any given electrode. They then organized these LFPs into bin widths of 4 ms and showed how the neural activity propagated in the network:
Especially interesting is that the avalanche is a form of correlated population activity distinct from oscillations, synchrony, and waves.
3) Inhibition of motor cortices via beta region activity in congruent action observation. Tkach et al (here) trained macaque monkeys to move a cursor to a target while recording LFPs in cortical motor regions (BA4 and BA6). They also recorded LFPs while the monkeys were passively watching themselves repeat the cursor movement. The lines on left show the mean LFP power in the 10 – 25 Hz (beta) range averaged over several electrodes. The bars on right show the integrated mean power, indicate that beta activity is consistently higher when the monkeys are passively watching:
The authors suggest that there is a continuum between inhibition and activation in the motor system. These results would fit that contention in that motor activity in this beta range is higher when the monkeys are watching the cursor, which they have control over during active trials and which they presumably would have to inhibit the desire to move more strongly.
Inspired by CalTech’s Question #16 for cognitive scientists: “What is a local field potential? What are its underlying causes? What can we learn from such data?”
Gray CM, Singer W. 1989 Stimulus-specific neuronal oscillations in orientation columns of cat visual cortex. PNAS link here.
Beggs JM, et al. 2003 Neuronal Avalanches in Neocortical Circuits. J Neuro PMID: 14657176
Tkach D, et al. 2007 Congruent Activity during Action and Action Observation in Motor Cortex. J Neuro doi:10.1523/JNEUROSCI.2895-07.2007
David SV, et al. 2010 Decoupling Action Potential Bias from Cortical Local Field Potentials. doi:10.1155/2010/393019