The ERP Boot Camp
Baselines, Difference Waves,
and the MONSTER Paradigm
All slides © S. J. Luck, except as indicated in the notes sections of individual slides
Slides may be used for nonprofit educational purposes if this copyright notice is included, except as noted
Permission must be obtained from the copyright holder(s) for any other use
Segmenting & Baselining
•
Prior to averaging, we must extract segments (epochs) of
the EEG surrounding the relevant events
Voltage
Time
•
Baseline correction is usually performed at this point
- This is important for some types of artifact rejection procedures
(e.g., absolute voltage thresholds)
- For most purposes, baseline correction can be performed at any
time
Reason 1: DC Offset
115 µV
Reason 2: Baseline Drift
How to Correct Baseline
Goal: Subtract estimate of DC offset from the waveform
Mean prestimulus voltage is usually a reasonable estimate
Subtract this value from each point in the waveform
2
cond1
1.5
cond1-baseline
1
0.5
0
-200
0
200
400
600
800
-0.5
-1
Note: Anything that messes up the baseline (e.g.,
noise, overlap) will be propagated to your amplitude
measurements
Baseline Distortion Example 1
2
cond1+noise
cond1+noise-baseline
1.5
1
0.5
0
-200
0
200
400
600
800
-0.5
-1
Entire waveform shifted down (negative)
because of positive noise blip
Baseline Distortion Example 2
1.5
1
0.5
0
-200
0
200
400
600
800
-0.5
cond1
-1
cond1+overlap
-1.5
cond1+overlap-baseline
How to Correct Baseline
•
•
•
What to use for response-locked averages?
- A period that is equivalent across conditions
- Often, only the prestimulus period is guaranteed to be equivalent
Simple option- Average using a long pre-response interval and use a time range
that is prior to the stimulus for every response
Complex option- Use the prestimulus period for each individual trial
Difference Waves
w1,1
C1
w2,1
E1
w3,1
w2,2
C2
w1,2
E2
w3,2
w1,3
w2,3
C3
w3,3
E3
E1
E2
If a single component varies across
conditions, it can be isolated by means
of a difference wave
C3
E3
C1
C2
For this to work, the conditions must be
so similar that only one component
varies across conditions
Example: N170
This voltage reflects face-related activity
plus everything else that is active at 170
ms
Rossion & Jacques (2009)
Example: N170
This difference reflects only brain activity
that differentiates between faces and cars
Rossion & Jacques (2009)
N170 and Development
Faces
Scrambled Faces
Cars
Scrambled Cars
Kuefner et al (2010, Frontiers in Human Neuroscience)
It looks like the N170 scalp distribution changes over development
But this could be due to other components in this time range
Kuefner et al (2010, Frontiers in Human Neuroscience)
Subtracting scrambled faces removes nonspecific activity
The face-specific activity has the same distribution over development
Kuefner et al (2010, Frontiers in Human Neuroscience)
Shortcomings of Difference Waves
• May not isolate a single component
- Example: Rare-Frequent yields P2, N2, P3
•
If Cond1–Cond2 difference wave is reduced, cannot
distinguish between smaller Cond1 and larger Cond2
- For most components, need to look at original waveforms as well
as difference waves (not for N2pc/LRP)
- Ask Emily how we solved this problem for LRP
•
Changes in latency between Cond1 and Cond2 produce
an apparent difference in amplitude
- Not a problem if you view difference wave as simply meaning a
difference in the time course of amplitude
The MONSTER Approach
•
•
•
•
Manipulation of Orthogonal Neural Systems Together in
Electrophysiological Recordings
General approach in which multiple components are
simultaneously isolated with high efficiency by means of
a factorial design and difference waves
Each factor used to create a difference wave that isolates
a specific component
Our version of MONSTER isolates 4 components
-
C1- Sensory processing in primary visual cortex
N2pc- Shift of visual attention
P3- Task-governed stimulus categorization
Lateralized Readiness Potential (LRP)- Response activation in
primary motor cortex
Isolating the C1 Wave
•
•
•
•
•
Onset 40-60 ms; peak ca. 80-100 ms
Thought to arise from area V1
Negative for upper-field stimuli; positive for lower-field
stimuli
When positive, merges together with P1
Largest at midline occipital-parietal sites
C1 Wave Elicited
by Stimulus at
Location Y
+1mV
-1mV
C1 Wave Elicited
by Stimulus at
Location X
100
ms
200
ms
Mangun, Hillyard, & Luck (1993)
Isolating the C1 Wave
C1 Wave Elicited
by Stimulus at
Location Y
+1mV
-1mV
C1 Wave Elicited
by Stimulus at
Location X
100
ms
200
ms
Mangun, Hillyard, & Luck (1993)
Isolating the C1 Wave
Isolating the LRP
Left Hemisphere
(Ipsi)
Right Hemisphere
(Contra)
Right Hemisphere
(Ipsi)
Left Hemisphere
(Contra)
LRP = Contra minus ipsi, averaged over left & right hands
Smulders & Miller (2010)
Isolating the LRP
LRP = Contra minus ipsi, averaged over left & right hands
Smulders & Miller (2010)
Isolating the LRP
Left Hemisphere
(Ipsi)
Right Hemisphere
(Contra)
Right Hemisphere
(Ipsi)
Left Hemisphere
(Contra)
Nonlateralized components eliminated by subtraction
Smulders & Miller (2010)
Isolating the LRP
Left Hemisphere
(Ipsi)
Right Hemisphere
(Contra)
Right Hemisphere
(Ipsi)
Left Hemisphere
(Contra)
Overall hemisphere differences eliminated by subtraction
(RHem more positive than LHem for both hands)
Isolating the LRP
Left Hemisphere
(Ipsi)
Right Hemisphere
(Contra)
Right Hemisphere
(Ipsi)
Left Hemisphere
(Contra)
Overall hand differences eliminated by subtraction
(RHand more positive than LHand for both hemispheres)
Smulders & Miller (2010)
Isolating the LRP
LRP provides a pure measure of the relative activation of
correct vs incorrect response at each moment in time
Smulders & Miller (2010)
Isolating the N2pc Component
Isolating the P3 Wave
• P3 amplitude depends on the probability of a task-defined stimulus
•
•
•
category, not the probability of a physical stimulus
P3 probability effect cannot begin until after categorization occurs
Rare-minus-frequent difference wave isolates processes that occur after
perception and categorization
P3 latency indexes “stimulus evaluation time”
P3 Difference Waves
The MONSTER Approach
• Manipulation of Orthogonal Neural Systems Together in
•
Electrophysiological Recordings
Goal- Efficiently measure the speed and integrity of a
broad set of specific neurocognitive processes
- First step in determining the nature of deficits in a given patient
group
- Possibly useful for diagnosis of disease subtypes (e.g., in multiple
sclerosis)
•
Use orthogonal subtractions to isolate specific ERP
components that reflect well-characterized
neurocognitive processes
- Same data divided different ways to isolate different components
Stimuli and Task
Task:
Attend to black or white
Press left for A (p = .80 or p = .20)
Press right for B (p = .20 or p = .80)
Left/Right and .80/.20 are counterbalanced
Stim:
Duration = 200 ms
SOA = 1500 ± 150 ms
2 blocks of 512 stimuli
Stimuli and Task
C1: Upper Checks minus Lower Checks
N2pc: Contra target minus Ipsi target
P3: Rare minus Frequent
LRP: Contra to response hand minus Ipsi to response hand
These subtractions are orthogonal -- same 1024 trials divided into
different pairs of subsets and then subtracted
Electrode Sites
LRP
P3
N2pc
C1