Evoked potential alpha interference

mdspre at vmsa.technion.ac.il mdspre at vmsa.technion.ac.il
Mon Dec 13 14:16:42 EST 1993

Dear all,

We are having tremendous problems with alpha-wave interference in our
laser evoked potential (LEP) pain studies.

Briefly, we are using a laser to produce a sharp, short (~20 msec) painful
stimulus.  The laser is rigged so as to provide a time-lock signal for EEG,
which is recorded on a PC-based system (at 256 Hz) and analyzed
off-line to yield EPs.  Normally we average 20-100 trials to produce
an EP.  Pain-related EPs are typically large and easily seen in single
trials, so fewer are usually needed (fortunately!).  There is a 5-8
second (random) inter-stimulus interval.  Our subjects are customarily
given some sort of task, depending on the particular experiment, such as
counting backwards by 7s from a rather large number, or counting stimuli,
or counting "target" stimuli.  Subjects are instructed to keep their eyes
open, and fixate on some point.

Alpha (or waveforms in the alpha band -- what else can I call it?) is
present in most of our subjects under these conditions, throughout most
of the experiments.  The amplitude of the alpha is about 20% of the
signals of interest.  As the alpha survives EP averaging, we were under
the impression we were seeing "alpha-locking" (any references, anyone?
Or is this electrophysiology's deep, dark secret?), except that the alpha is
also present in the averaged *pre*-stimulus interval.  Perhaps the pre-
stimulus alpha is locked to the previous stimulus -- but how does it
survive averaging, considering the inter-stimulus interval is randomly
varied?  (Or perhaps the pre-stimulus alpha is locked to the later stimulus,
and thiotimoline is a neurotransmitter? :-) )

We've started playing about: for example, we've tried rejecting the trials
which appear contaminated with alpha (most of them), and re-doing the
average, but alpha still appears about as strongly as in the fully-averaged
EPs.  I'm considering popping the individual trials into Fourier analysis,
finding the alpha peak, and notch-filtering it out, but this strikes me as
cheating (is it?).  Can anyone suggest any alternatives?  Better still,
can anyone figure out the cause of the alpha, whether some feature of our
procedure, or some sampling artifact of our system, or some defect of

Thanks in advance, as usual.

Elliot Sprecher, Ph.D.
Experimental Psychologist, Research Coordinator
Institute of Clinical Neurophysiology
Rambam Medical Center
Haifa, Israel

More information about the Neur-sci mailing list