ERP Boot Camp Lecture #10

Report
The ERP Boot Camp
Setting Up and Running
an ERP Lab
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
Recording Chamber
• Do you really need one?
• Probably not if:
- You’re looking at slow components and can low-pass filter with a
50% cutoff at 30 Hz
- And you’re not near any major source of electrical noise
•
Elevators, centrifuges, power transformers, ventilation fans
- You don’t care about gamma oscillations
• They’re good for keeping subjects focused
• They tend to get warm, so it may actually be better not to
•
have one if skin potentials are a major source of noise
You can build one from 2x4’s and copper screen
Courtesy of Lynne Reder
Seating
• Key points:
- Comfortable to avoid muscle noise
- Don’t want subjects to fall asleep
- Don’t want electrodes to rest on anything
•
Recliners were once common
- Not good if you have electrodes over the back
of the head
• I recommend high-quality office chair
- Glides rather than wheels
- Mark the floor
• I haven’t had much luck with chin rests
Response Devices
• Need to be held in a comfortable position
- Don’t want subject holding arms up
- Standard computer keyboards are bad
• Game controllers work well
- Mass-produced -> reliable
•
Constant and variable timing errors are possible
- RT is so variable that a bit of timing variability will usually have
virtually no impact (unless you are looking at response-locked
averages)
- EMG for best response timing
- Can measure timing errors by putting a mic next to device and
recording “click” along with event code
Hints for Running Subjects
• ~60 minutes of “run time” per session
- More for interesting experiments
- Whole session is about 3 hours
• Runs of 4-6 minutes with 2-3 20-second breaks
- Less makes it inefficient to deal with electrodes, etc.
- More leads to fatigue
- Some labs do all-day sessions with lot of breaks
• Dilution Rule: Don’t dilute good data with bad data
- Adding noisy trials doesn’t improve the S/N ratio
•
Watch the EEG throughout the session
- Look for artifacts, bad connections, etc.
•
Watch the subject with a video camera
Hints for Running Subjects
•
Happiness Rule: A happy subject is a good subject
- Compliance with task
- Compliance with artifact control instructions
- Less noise
•
Talking to subjects
- Treat subject like a person, not like a piece of meat
- Chat while putting on electrodes (or video)
•
Tell them exactly what will happen -- this reduces stress
- Chat during breaks
- Note: Some subjects don’t want to talk -- that’s OK
•
Keeping subjects happy
- Food and drink (is caffeine a confound?)
- Eye drops (single-use)
- Music
Looking at the Data
•
Do a fairly complete analysis of the first subject’s data
before running anyone else
-
•
All the main comparisons among ERP waveforms
Accuracy (and RT if recorded) for each main condition
There may be a serious problem with event codes, etc.
“Nothing focuses the mind quite as much as real data”
Take a look at the individual subjects and the grand
averages every 3-4 subjects
- Grand averages will give you more power to see something funky
in the data
- But don’t get too freaked out if the results look a little funny or
aren’t conforming to your predications
- Be especially concerned about “impossible” results (e.g., effects
that consistently begin before time zero)
- Look at calibration data for each subject if you can
Ethical Issues
•
Everything that applies to behavioral experiments plus…
- Risk of disease transmission
•
•
High impedance helps
Need thoughtful disinfection (even for high impedance)
- Risk of electrical shock
•
-
•
Optical isolation and/or battery power
Headache from electrode cap
Gel in hair
Long duration of experiment
Claustrophobia
Concerns about privacy of EEG data
Providing clear information in advance is the best way to
prevent problems
Stimulus Presentation
• Testing the timing of event codes
- Digitize at ~1000 Hz (higher for auditory)
- Present stimuli along with event codes
- For auditory stimuli, connect auditory output (or a microphone) to
the digitization system
•
You might want to use a square-wave tone or a 50-Hz sine wave
- For visual stimuli, point some kind of light pen to the video
monitor and connect to digitization system
- See when the stimuli are actually presented relative to the event
codes
•
Auditory artifacts
- Speaker in headphones may induce a current
- Post-auricular muscle twitch
CRT Basics
When you draw something, nothing happens until the frame buffer is updated
AND the raster beam reaches the right part of the monitor
LCDs operate similarly, but often there is an additional delay of several
milliseconds before the stimuli actually appear
Stimulus Timing Jitter
•
•
•
•
What does a constant delay between event code and
stimulus do to the averaged ERP?
- Time shift
- Can be fixed with a filter
What does a variable delay between event code and
stimulus do to the averaged ERP?
- Distribution of delays is convolved with jitter-free “real” waveform
- Modest low-pass filter
For most cognitive paradigms, effects are minimal
- But not always
You need to understand exactly what the jitter is doing,
and this usually requires measuring it
Stimulus Timing Jitter
1
Distribution of
Stimulus Delays
0.8
Example: Stimulus appears 0, 10,
or 20 ms after event code with
even distribution
0.6
0.4
0.2
Waveform appears at 0, 10, or 20
ms with equal likelihood
0
0 ms
10 ms
20 ms
1
Averaging these together is equal
to replacing each point in the
distribution of delays with a scaled
and shifted version of the ERP
waveform
0.8
0.6
0.4
0.2
0
-100
0
-0.2
-0.4
-0.6
100
200
300
400
500
600
700
0-ms delay
10-ms delay
20-ms delay
Avg Across Delays
The result is slightly low-pass
filtered and shifted to the right in
time
What is frequency response of
filter produced by jitter?
Writing an ERP Paper
•
Rule #1: Write with a specific audience in mind
- But keep in mind that the reviewers are the first and most
important audience
•
Rule #2: Intro must end with a set of competing
hypotheses about a general issue and then a set of
corresponding predictions
- May want to explicitly address reason for using ERPs
•
Rule #3: Results should be organized to lead reader to a
conclusion (logical flow of ideas)
- Descriptive statistics first, then inferential statistics
•
Rule #4: Discussion should recap major results and
conclusions that can be drawn
- Often followed by possible objections that can be discarded (and
perhaps some that cannot)
Example Decision Letter
Dear Dr. XXXXX:
I gave your manuscript a quick reading so that I could choose appropriate
reviewers. After this quick reading, it was clear to me that the manuscript
cannot be accepted in its present form. Thus, to save everyone time and
effort, I am rejecting this version of the manuscript, but I will allow you to
submit a revised manuscript if you are certain you can overcome the
problems with the current version.
Example Decision Letter
Dear Dr. XXXXX:
In most ERP papers that make a significant contribution to broad questio ns
in cognitive neuroscience, the Introduction ends with a set of specific
predictions about the pattern of ERP results that will be obtained. That is,
competing hypotheses about broad issues in cognitive neuroscience are
raised in the first part of the Introduction, and then specific predictions
about the ERP results are given that will distinguish between the competing
hypotheses. This is both an indicator of the likely importance of the study
and an important aid to readers, who can much better underst and the
methods and results if they know what general patterns of results can be
expected and how these results are related to the broad issues addressed by
the study. The present manuscript does not contain such predictions. If you
submit a revision, you should rewrite the Introduction in this manner.
I gave your manuscript a quick reading so that I could choose appropriate
reviewers. After this quick reading, it was clear to me that the manuscript
cannot be accepted in
present
form.
Thus,
torecap
save
It is its
also useful
for the Discussion
section to begin
with a brief
of the everyone time and
major findings and the conclusions that can be drawn from them. Again,
helps make
it clear that the
represents
a s ignificant advance inbut I will allow you to
effort, I am rejectingthis
this
version
ofstudy
the
manuscript,
knowledge about the broad issues that were initially raised in the
Introduction. It is also helpful for the casual reader who does not want to
submit a revised manuscript
if you are certain you can overcome the
read the details of your methods and results, but instead wishes to quickly
see your “bottom line.” It was quite difficult for me to determine what
problems with the current
version.
conclusions could
be drawn from this manuscript . If you submit a revision,
you should rewrite the beginning portion of the Discussion in this manner.
[Some details about this particular study]
In most ERP papers that
contribution
to broad questio ns
ACTION: Imake
cannot accepta
thesignificant
manuscript for publication
in its current form.
However, I invite you to submit a revision if you feel that you can fully
address the concerns
in this letter. Please note,
however,
that a a set of specific
in cognitive neuroscience,
the raised
Introduction
ends
with
substantial amount of time and effort is required to review a manuscript, so
should not submit
you are certain
that thewill
revised be obtained.
predictions about theyou
pattern
ofa revision
ERPunless
results
that
That is,
manuscript draws solid and important conclusions about the cognitive and/or
neural processes involved in [topic of study]. That is, the paper should be of
competing hypotheses
about broad issues in cognitive neuroscience are
interest to [topic of study] researchers who are not interested in ERPs per
se. If you believe that your revised manuscript makes this sort of strong
raised in the first part
the
Introduction,
specific
predictions
and of
important
contribution,
I will be very happy toand
send thethen
revision out
for
review.
about the ERP results are given that will distinguish between the competing
hypotheses. This is both an indicator of the likely importance of the study
and an important aid to readers, who can much better underst and the
Method Section Should Include…
• Number of trials per condition (explicitly)
• Recording sites, electrode type, amplifier gain, filters,
sampling rate and resolution, impedance, reference, and
offline re-referencing
•
•
- Include impulse response function details for offline filters
Artifact rejection procedures
- Include observed mean and range of % rejected trials
- Include # of subjects rejected and standard for rejection
- Rejection of trials with behavioral errors
ERP measurement procedures
- Measurement windows and perhaps justification
• Greenhouse-Geisser epsilon adjustment
• See Picton et al. (2000)

similar documents