[Eeglablist] re-referencing

Nicholas Rosseinsky rosseinsky.nicholas.m at gmail.com
Sat Oct 4 12:51:27 PDT 2014

Oops - sorry: hit send waaaaaay too early ...

1. Eric asked about references concerning re-referencing.
Here are some, not specifically concerned with IC decomposition though:

Nunez, P. L. (2010). REST: a good idea but not the gold standard. *Clinical
neurophysiology: official journal of the International Federation of
Clinical Neurophysiology*, *121*(12), 2177

Qin, Y., Xu, P., & Yao, D. (2010). A comparative study of different
references for EEG default mode network: the use of the infinity
reference. *Clinical
neurophysiology*, *121*(12), 1981-1991.

Hu, S., Stead, M., Dai, Q., & Worrell, G. A. (2010). On the recording
reference contribution to EEG correlation, phase synchorony, and
coherence. *Systems,
Man, and Cybernetics, Part B: Cybernetics, IEEE Transactions on*, *40*(5),

2. I may well be wrong, but it seems to me that the reference question
could be discussed a little more, generally. Notably, the average reference
seems to be though of as "harmless" or "default", but obviously it
subtracts from the data the average fluctuation of brain activity relative
to e.g. mastoid potential! In an idealized case in which e.g. mastoid
electrodes are already "at" some idealized "nominal ground" (Nunez 2010),
this means that average referencing is subtracting out the common
electrophysiologically-relevant activity captured in
channel-potentials-relative-to-mastoid. (I'm aware of the unrealizability
of this idealized case, and of the problems concerned with the potential of
the body-as-battery ... I'm just sayin').

Notably, try *visualizing* (making  a topoplot movie of) alpha-band
continuous EEG before and after average referencing. There are
stereotypical and large spatiotemporal travelling patterns present in
non-averaged data that disappear after averaging (disclaimer: *in the data
I am currently looking at,* which has significant quality and
scalp-coverage limitations). IC decomposition of non-averaged data
certainly seems to have a spatially "ugly" component that captures e.g.
common alpha activity - but to throw this out because it doesn't "look
dipolar" is surely the wrong criterion. The question is whether this kind
of activity is relevant to analyses. If alpha travelling waves affect task
performance (Patten et al., 2012), and if you are interested in analysing
or identifying such effects ... take care with choosing average reference.

As I understand it (please, educate me, listers!) average reference is
theoretically innocuous only if electrodes completely cover the spherical
head; the further away one is from this (again, practically-unrealizable)
ideal, the more *potentially* "impactful" average-referencing *might* be .
And in other cases (my data may be particularly poor  in this regard), it's
going to be a judgment call best guided by:

1. visualizing your data before and after preprocessing to know what
effects your pre-processing steps are having; and,
2. making an intelligent, context-dependent, analysis of how any
preprocessing effects are going to affect your subsequent analyses.

I don't think there are general principles here, and:* that doesn't mean
that the choice doesn't matter, *i.e. "no general principles" doesn't mean
"choosing average reference won't affect my analyses".

Patten, T. M., Rennie, C. J., Robinson, P. A., & Gong, P. (2012). Human
cortical traveling waves: dynamical properties and correlations with
responses.*PloS one*, *7*(6), e38392.

Hope that helps some, and please, if musings under 2 are nonsense - help me!

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20141004/ad4c3080/attachment.html>

More information about the eeglablist mailing list