[Eeglablist] odd ERP images

[Scott Makeig]

Dorothy -

Your erpimage plots look quite normal to me. Those used to looking at their
data via trial averages, and ignoring trial-to-trial varations (as
highlighted by erpimage plotting) are often surprised at the extendt of
trial-totrial variability in their data.

I would suggest you explore this variability further by:
1. Recomputing the erpimages with trial smoothing set to 1 (i.e. turned off)
- note the single-trial variability underlying these smoothed images.
2. Recomputing erpimages sorted by potential in the time window of some ERP
feature of interest (e.g. N100 or some other peak in the average). Note what
percentage of trials show a feature similar to the average feature.

The trial-to-trial variability may have any or all of four sources: (1)
Changes in brain function (e.g. from attention to inattention, etc), (2)
Presence or absence of activity from 'other' sources (brain or non-brain) in
the channel data, or in some cases, (3) Partial phase locking of activity in
some frequency band produced by the time-locking events. The remaining
possibility (4) is trial to trial size or latency changes in a strictly
stereotyped series of potentials evoked by the stimulus (i.e., in a 'true
ERP').

Distinguishing the effects of these four possible causes is difficult: To
eliminate the first, careful observation of the subject and evaulation of
changes in their performance is useful --To do this most sensitively, I
suggest increasing the bandwidth of behavior produced by subjects and
recorded in the experimental paradigm. To minimize the second, spatial
filtering for source activities is highly useful -- I suggest you explore
using ICA for this purpose. After separating the source contributions, you
can use inter-trial coherence (ITC) to evaluate the third. Distinguishing
between the third and fourth is greatly aided, at least, by source
separation. See Makeig et al., TICS 2004 (
http://sccn.ucsd.edu/Papers/TICS04.html) for further discussion.

Best luck in your explorations!

Scott Makeig



On Jan 18, 2008 7:10 AM, Dorothy Bishop <Dorothy.Bishop at psy.ox.ac.uk> wrote:

> I am a new EEGlab user and have been using EEGlab to explore a fairly
> noisy dataset gathered with 11-12 yr old children.
> The paradigm involves presenting auditory stimuli; there is a train of
> tones (with SOAs of 250-450 ms), then a delay of 800 ms, then a new train.
> We are looking at the response to the first stimulus in a train. Given the
> long SOA prior to the stimulus, we don't expect to see any carryover from
> previous stimulus, and in adults we see a fairly standard auditory ERP, with
> small P1, then N1 (around 100 ms), P2.
> In children of this age one does not always get N1 and P2 ; sometimes
> instead an initial P1 which then descends to a N2 around 250 ms.
>
> The ERPimages look odd to me, and give concerns that there may be some
> kind of problem with how the paradigm was administered. It would be helpful
> to know if others have seen anything similar - if so, is there an
> explanation.
> In short, I'd expect to be able to see vertical stripes in the ERPimage,
> given that this file just contains stimuli all of the same kind.
> Instead, we see that pattern for a block of stimuli, but it then changes.
>  So what one sees are horizontal blocks in the image.
> The ERPimage is not sorted, so the stimuli are just in the order in which
> they occurred.
> (50% of trials used a different tone frequency, and these have been
> removed, so the trials are not actually adjacent, in that there may have
> been intervening trials of a different kind. However, the same is seen
> regardless of which stimulus is used).
> So it looks like the same kind of response occurred for 10-20 stimuli, but
> then it changed for the next 10-20, and so on.
>
> I've put some sample ERPimages up on
> http://psyweb.psy.ox.ac.uk/oscci/Miscellaneous.htm
> Each ERPimage is from a different subject.  The stripey horizontal pattern
> is pretty common.
>
>
>
>
>
>
> Dorothy Bishop
> Professor of Developmental Neuropsychology
> Department of Experimental Psychology
> University of Oxford
> OX1 3UD
>  http://psyweb.psy.ox.ac.uk/oscci/
>
> tel: +44 (0)1865 271369
> fax: +44 (0)1865 281255
>
>
>
>
> _______________________________________________
> eeglablist mailing list eeglablist at sccn.ucsd.edu
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
>



-- 
Scott Makeig, Research Scientist and Director, Swartz Center for
Computational Neuroscience, Institute for Neural Computation, University of
California San Diego, La Jolla CA 92093-0961, http://sccn.ucsd.edu/~scott
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20080124/9d7363da/attachment.html>