DIFFERENCES AMONG TOTAL,
PHASE-LOCKED, AND NON-PHASED
LOCKED POWER AND INTERTRIAL
PHASE
Chapter 20
Tony Ye
Total Power
• Comes from:
• Taking time-frequency decomposition (any method) of each trial
• Averaging the time-frequency power from all trials
• Most commonly used time-frequency approach in
cognitive electrophysiology.
Non-Phase-Locked Power
• Time-frequency representation of the data after the
phase-locked components of the EEG signal are
Non-Phase-Lock
Single Trials
removed.
1
• Is time-locked
• BUT, not phase-locked to
time = 0 event.
• A.K.A. “induced power”
2
3
4
5
6
Example: Non-Phase-Locked Power
Phase-Locked Power
• Subtracting non-phase-locked power from the total power
= phase-locked power!
• Phase-aligned with the time = 0 event
• Time-domain averaging (the ERP)
• Time-frequency-domain averaging
• A.K.A. “evoked power”
Jitters
Example: Phase-Locked Power
Phase-Locked vs. Non-Phase-Locked
• Both are task related:
• Time and/or frequency characteristics change as a function of
engagement in task events.
• Both are unclear in physiological dynamics
(David et al.,
2006)
•
Adjusted Power - unaffected by trial-to-trial variations
ERP Time-Frequency Power
• Compute:
1. The ERP
2. The ERP’s time-frequency representation with any method.
• Raw power values from this analysis are likely to be an
order of magnitude smaller than those from total
power/non-phase locked analyses.
• Why?
• Voltage values of the ERP are generally an order of magnitude
smaller than those of single trials
ERP Time-Frequency Power
All trials and trial-average (ERP)
ERP Time-Frequency Power
• Warning!
• Do not use the baseline period power from total/non-phase-locked
analysis to normalize the ERP T-FP!
• Solution!
• Use ERP baseline to normalize the ERP T-FP
Intertrial Phase Clustering (ITPC)
• The extent to which phase values become clustered over
trials.
• Also referred to as:
• Phase-locking value/factor
• Intertrial coherence
• Phase-based connectivity.
When to Use Total Power?
• Use total power unless you have a good reason not to.
• Most common
• Does not alter data by removing parts of signal
• Interpretation does not rely on assumptions of
neurophysiological events related to phase-locked vs.
non-phase-locked activity.
• GABA interneurons + Excitatory pyramidal cells = Oscillations
When to Use ERP TFP?
• Almost never compared to ERP
• All information can equally be found in ERP
• ERP has higher temporal precision
• ERPs can be easier to interpret
• HOWEVER!
• ERPs are unlikely to have a clean time-frequency representation
• So, use ERP TFP when…
• Attempting to dissociate components of the ERP by their frequency
representation instead of latency and polarity.
When to Use ERP TFP?
Fig 20.1A
Fig 20.1D
When to Use Non-Phase-Locked Power?
1. When examining the relationship between the
information available in time-frequency power and the
ERP.
Fig 20.1A
Fig 20.1B
When to Use Non-Phase-Locked Power?
• Interpreting differences in neurophysiological mechanisms
relies on assumptions:
• Neurophysiological processes that generate the ERP are distinct
from those that generate non-phase-locked activity.
When to Use Non-Phase-Locked Power?
2. When attempting to make single-trial EEG data more
likely to remain stationary
• Sometimes successful but no guarantees
When to Use ITPC?
• When there are hypotheses regarding the timing of band-
specific activity over trials w/ respect to an experiment
event.
• Stimulus onset/response
• Better than the ERP
• ITPCs have more frequency-band specificity
• But not really better than the ERP
• ITPCs have lower temporal precision