[Eeglablist] Question about AVG referencing

Joseph Dien jdien07 at mac.com
Thu Jul 23 12:25:36 PDT 2026


Reference choice influences the waveforms so it influences the 
measurements, both the signal and the noise (Dien, 1998; 2017). A1 and 
A2 are on the underside of the head, which is undersampled, so 
inclusion/exclusion would have more effect.  It also depends on the 
orientation of the EEG fields that you are studying.  If they are 
vertically oriented, then A1/A2 would have more influence. Having some 
sessions different from others is especially a concern for a clinical 
study where you are likely examining individual differences, as it is 
potentially a source of confounds.  You should either drop A1 and A2  
from the rest of them, or add A1/A2 if possible to the missing ones if 
possible, although with a sparse montage channel replacement via 
interpolation or regression probably would not work well.

Dien, J. (1998). Issues in the application of the average reference: 
Review, critiques, and recommendations. /Behavior Research Methods, 
Instruments, and Computers/, /30/(1), 34–43. 
https://urldefense.com/v3/__https://doi.org/10.3758/BF03209414__;!!Mih3wA!H6p5LX0QTQlRXqRhvAdUJlcZtd2fjQYMm3zRwO09Br6Cd3sA8EOfxHzwwUI59dicSgNEw8DVuC_Nep4yDZc$ 
Dien, J. (2017). Best practices for repeated measures ANOVAs of ERP 
data: Reference, regional channels, and robust ANOVAs. /International 
Journal of Psychophysiology/, /111/(1), 42–56. 
https://urldefense.com/v3/__https://doi.org/10.1016/j.ijpsycho.2016.09.006__;!!Mih3wA!H6p5LX0QTQlRXqRhvAdUJlcZtd2fjQYMm3zRwO09Br6Cd3sA8EOfxHzwwUI59dicSgNEw8DVuC_N_5Fm0zU$ 

Cheers!

Joe

On 7/23/26 15:09, Joseph Dien wrote:
>
> You should be consistent across your dataset.
>
> Joe
>
> On 7/23/26 04:52, Enrico Fratto wrote:
>> Thank you all for the precious feedbacks. This really helps me 
>> clarifying the assumptions behind the set of electrodes making the 
>> average.
>> As another question: let's say I have eegs which have been recorded 
>> with slightly different electrode sets (i.e. I have A1 A2 in some but 
>> not in others), so that also the AVG defining electrodes slightly 
>> differ. Can still I make pooled analysis on the shared electrodes 
>> (e.g. O2) in the AVG montage?
>> My guess would still be cautiously yes, but I wonder how this can 
>> affect calculations in the setting of a low electrode density, as in 
>> clinically recorded EEGs
>> Best,
>>
>>
>> Enrico Fratto, MD
>> Institute of Neurology,
>> Department of Medical and Surgical sciences
>> Università Magna Graecia,
>> Catanzaro
>>
>> On Thu, Jul 23, 2026, 01:26 Joseph Dien via eeglablist 
>> <eeglablist at sccn.ucsd.edu> wrote:
>>
>>     The reasoning behind the average reference is the physics
>>     principle that
>>     if you have a voltage source in a conducting sphere
>>     (approximation of
>>     head shape), then the voltages over its surface will sum to zero
>>     (as you
>>     allude to).  The average of the entire surface is therefore the
>>     "true"
>>     zero.  Of course, we can't actually measure the entire surface of
>>     the
>>     head, but the more measurement sites we have, the better the
>>     estimate of
>>     the voltage surface.  The argument is that although it is a
>>     necessarily
>>     imperfect estimate, it is more accurate than just arbitrarily
>>     declaring
>>     that an electrode is zero voltage (which is what a conventional
>>     reference does).  We include the reference site because we do have a
>>     voltage measurement for it (by definition it is zero).  In the
>>     case of
>>     an extra-cephalic reference, all that reasoning goes out the
>>     window.  So
>>     typically you would want to not include it in the average and you
>>     would
>>     not want to generate a rereferenced waveform for that site.  That
>>     said,
>>     it has been reported (I don't have the citation on hand) that
>>     basically
>>     the entire body is equipotential with the neck (since
>>     electrically we're
>>     a bag of salt water, which is an excellent conductor), so one could
>>     argue for it being included as an estimate of the underside of the
>>     head.  Potentially, it could be especially helpful for average
>>     reference
>>     schemes like PARE or REST, which seek to correct for the
>>     undersampling
>>     of the bottom part of the head.  You'd need the appropriate
>>     equivalent
>>     coordinates in that case, so it'd take some thinking.  If you have a
>>     reference at another cephalic site (even if not 10-20 or 10-05 or
>>     whatever), it would be nonsensical to drop it.  It'd just be losing
>>     information.  Just treat it like any other EEG site.
>>
>>     The key to thinking about EEG voltage information is that the
>>     fundamental information is not the waveforms from an electrode (they
>>     change with the reference scheme).  The fundamental information
>>     is the
>>     difference, at a given time point, between electrodes. Voltages are
>>     relative measures.  That's why a reference site is of equal value
>>     to the
>>     other EEG sites even though it is a flat zero voltage line prior to
>>     rereferencing.  The real information there is its relative
>>     difference
>>     from each and every other electrode at a given moment in time.  That
>>     does not change with reference site choice or with decisions
>>     about which
>>     electrodes to keep.  My 1998 paper is a helpful reference for
>>     thinking
>>     about all this.
>>
>>     Dien, J. (1998). Issues in the application of the average reference:
>>     Review, critiques, and recommendations. /Behavior Research Methods,
>>     Instruments, and Computers/, /30/(1), 34–43.
>>     https://urldefense.com/v3/__https://doi.org/10.3758/BF03209414__;!!Mih3wA!F_zQRsBFYq5PQs849J1FaIb_YpNn30vDpQS9f86dtATUz2VnGCt_VFDvQmFLIE0a_jUVr24CDqQwWhR6DGI$
>>
>>
>>     Cheers!
>>
>>     Joe
>>
>>
>>
>>     On 7/22/26 09:58, Enrico Fratto via eeglablist wrote:
>>     > Greetings to all the community.
>>     >
>>     > I am reaching out to ask about something which is often
>>     overlooked but
>>     > sometimes clicks hard to understand for me as the average
>>     reference process.
>>     > Let's say I have 22 channels + Fz which is used as the reference.
>>     >
>>     > According to Makoto's famous explanation (
>>     >
>>     https://eeglab.ucsd.edu/wiki/Makoto%27s_preprocessing_pipeline?source=post_page---------------------------&utm_source=chatgpt.com#Re-reference_the_data_to_average_(10/31/2025_Updated) 
>>     <https://eeglab.ucsd.edu/wiki/Makoto%27s_preprocessing_pipeline?source=post_page---------------------------&utm_source=chatgpt.com#Re-reference_the_data_to_average_(10/31/2025_Updated) >
>>
>>     > )
>>     > we are supposed to add a 0 line in the AVG reference
>>     calculation to account
>>     > for the reference channel itself (corresponding to dividing the
>>     sum at the
>>     > n electrodes by n+1 to encompass the reference) when the flat
>>     channel is
>>     > not already present in the data.
>>     > In my case, it would be the sum at the 22 recording electrodes
>>     divided by
>>     > 23 to encompass Fz.
>>     > My question now is: let's say I acquired EEG with a reference I
>>     am not
>>     > interested in (say, extra-cephalic). Would I still have to
>>     encompass it in
>>     > the denominator while calculating the AVG in this case? Or,
>>     because it is
>>     > not part of the head, I treat it as if it does not belong to the
>>     > "zero-potential" assumption and I just estimate the average
>>     using the
>>     > remaining electrodes?
>>     > And what should I do instead if the reference is at another
>>     cephalic custom
>>     > location (not belonging to the standard electrodes)?
>>     >
>>     > Thank you in advance for your help,
>>     >
>>     > Best
>>     >
>>     >
>>     -- 
>>     --------------------------------------------------------------------------------
>>
>>     Joseph Dien, PhD
>>     Senior Research Scientist
>>     Department of Human Development and Quantitative Methodology
>>     University of Maryland, College Park
>>     https://urldefense.com/v3/__http://joedien.com__;!!Mih3wA!F_zQRsBFYq5PQs849J1FaIb_YpNn30vDpQS9f86dtATUz2VnGCt_VFDvQmFLIE0a_jUVr24CDqQw3JjvVI8$
>>
>>     _______________________________________________
>>     To unsubscribe, send an empty email to
>>     eeglablist-unsubscribe at sccn.ucsd.edu or visit
>>     https://sccn.ucsd.edu/mailman/listinfo/eeglablist .
>>
> -- 
> --------------------------------------------------------------------------------
>
> Joseph Dien, PhD
> Senior Research Scientist
> Department of Human Development and Quantitative Methodology
> University of Maryland, College Park
> https://urldefense.com/v3/__http://joedien.com__;!!Mih3wA!H6p5LX0QTQlRXqRhvAdUJlcZtd2fjQYMm3zRwO09Br6Cd3sA8EOfxHzwwUI59dicSgNEw8DVuC_Nd958CkM$ 

-- 
--------------------------------------------------------------------------------

Joseph Dien, PhD
Senior Research Scientist
Department of Human Development and Quantitative Methodology
University of Maryland, College Park
https://urldefense.com/v3/__http://joedien.com__;!!Mih3wA!H6p5LX0QTQlRXqRhvAdUJlcZtd2fjQYMm3zRwO09Br6Cd3sA8EOfxHzwwUI59dicSgNEw8DVuC_Nd958CkM$ 


More information about the eeglablist mailing list