1 of 28

The neural code�Oscillations

Kenneth D Harris, UCL

2 of 28

Local field potential

  • Oscillations are generated by rhythmic firing of action potentials in populations of neurons

  • But they are often measured by the local field potential (LFP), measured inside the brain

  • Or the electroencephalogram (EEG), measured outside the skull
    • Where they were first discovered

  • These potentials reflect synaptic activity

3 of 28

4 of 28

Physical basis of the LFP signal

  •  

Synaptic input

Intracellular current

Charging current (capacitive)

plus leak current (resistive)

Return current

5 of 28

Power spectrum

  •  

6 of 28

White noise, pink noise, narrow- and broadband oscillations

7 of 28

LFP power spectra

  • Typical “1/f” shape (pink noise). Can multiply by f to normalize
  • Oscillations seen as modulations around this
  • Usually small, broad peaks
  • Ocassionally narrowband

(log scale)

(log scale)

8 of 28

Two types of oscillation

  • “Inactive” oscillations
    • When the brain is resting/idling/sleeping
    • Alpha (eyes closed, visual cortex)
    • Beta (immobility, motor system)
    • Slow, delta, spindle (sleep, thalamus/cortex)

  • Active” oscillations
    • When the brain is in an active state
    • Theta (spatial exploration, rodent hippocampus)
    • Gamma (Cortex/hippocampus, active states)

9 of 28

Two types of oscillation

  • “Inactive” oscillations
    • When the brain is resting/idling/sleeping
    • Alpha (eyes closed, visual cortex)
    • Beta (immobility, motor system)
    • Slow, delta, spindle (sleep, thalamus/cortex)

  • Active” oscillations
    • When the brain is in an active state
    • Theta (spatial exploration, rodent hippocampus)
    • Gamma (Cortex/hippocampus, active states)

10 of 28

Visual cortical Alpha rhythm

  • 8-12 Hz oscillation over visual cortex of humans
  • Particularly when the eyes are closed or during rest (but not sleep)

11 of 28

Beta rhythm in motor system during immobility

Human globus pallidus and substantia nigra

Off L-DOPA

On L-DOPA

12 of 28

Sleep: delta, slow, and spindle oscillations

13 of 28

Two types of oscillation

  • “Inactive” oscillations
    • When the brain is resting/idling/sleeping
    • Alpha (eyes closed, visual cortex)
    • Beta (immobility, motor system)
    • Slow, delta, spindle (sleep, thalamus/cortex)

  • Active” oscillations
    • When the brain is in an active state
    • Theta (spatial exploration, rodent hippocampus)
    • Gamma (Cortex/hippocampus, active states)

14 of 28

Theta oscillation: locomotion

15 of 28

Gamma oscillation: multifarious high-frequency oscillations

16 of 28

High-gamma correlates with firing rate

  • High-frequency broadband power usually correlates with firing rate
  • Sometimes called high-gamma or epsilon
  • Is it an oscillation?

17 of 28

Attention reduces low-frequency power in V1

18 of 28

Different cells and cell types fire on different phases

  • Hypothesis: different theta phases reflect different steps in a computation cycle
  • Inputs from different regions arrive in different phases
  • Different inhibitory classes are active

19 of 28

Phase of firing can encode information

20 of 28

How do you measure phase?

  • Hilbert transform estimates instantaneous phase and amplitude of a narrowband oscillation
  • Doesn’t work for broadband signals! So you might need to filter your data.

21 of 28

Phase histogram

  • Different cells prefer different phases

  • How do we compute each cell’s mean phase?

22 of 28

Linear mean doesn’t work

23 of 28

Circular mean

  •  

 

 

 

R

24 of 28

How does phase depend on another variable?

  • Don’t use linear regression!

25 of 28

Nadaraya-Watson smoothing for phase

  •  

To avoid divide zero errors

26 of 28

Phase field

  • Local likelihood estimation for von Mises

Physical space

Abstract space

27 of 28

Confirmatory statistics:

  • How to test that phase is independent of place?

  • Shuffling method?

  • Test statistic?

28 of 28

Interpretation of phase precession

  • Three types of temporal coding:

1. Following the timing of an external stimulus

2. Sequence of activity reflecting steps in a serial computation

3. Generating new temporal structure that encodes non-temporal information

  • Phase precession is not type 1.
  • Could it be type 2? Or both 2 and 3?