MEG is reference free but EEG has been using different types of reference for the past few decades. Below are some of the usual EEG reference selections:
- 1 electrode on top of the head (Cz)
- Average between electrodes on the two ears
- Average of all connected electrodes (average reference)
- Average of two mastoid references
- Single mastoid (left or right) reference
- Nose reference

Using one type of reference compared to another seems to depend on personal preference. Choosing a noisy reference means that the noise will be containing on all data channels. So what is the best reference?
Thanks,

A. Delorme

I have seen many references been used to compute gamma in EEG
(mastoid, Cz, average reference or ears). The nose should not be used
for gamma as it is contaminated by eye tremors (see Fries, 2009 in
Neuron). Qin et al (2010, Clinical Neurophsyiology) propose a reference
free mecanism called REST and that is based on source localization. It
thus depend on the forward model you are selecting. Nunez in the same
issue provides a nice review of the ideal reference pointing out that
average reference is still considered a good compromised.

There
is also independent component analysis (ICA). ICA applied to EEG data
should be reference free. However, I have not seen it used to computing
reference free EEG. It is probably because a large number of authors
using ICA work on the component themselves rather than the data
channels.

Christian

ps: I wrote a very
long and detailed response and this web site lost it because I stayed
loged in too long (above is the summary I retyped). Beware other users.

Thanks Christian,

 

following up on my own thread here.
There was a related discussion of choice of reference on the popular EEGLAB
mailing list. Some more opinions below

 

First an introduction

 

* * *[url=http://www.mrc-cbu.cam.ac.uk/research/eeg/eeg_intro.html*]http://www.mrc-cbu.cam.ac.uk/research/eeg/eeg_intro.html*[/url]
Electric potentials are only defined with respect to a reference, i.e. an
arbitrarily chosen "zero level". The choice of the reference may
differ depending on the purpose of the recording. This is similar to measures
of height, where the zero level can be at sea level for the height of
mountains, or at ground level for the height of a building, for example. For
each EEG recording, a "reference electrode" has to be selected in
advance. Ideally, this electrode would be affected by global voltage changes in
the same manner as all the other electrodes, such that brain unspecific
activity is subtracted out by the referencing (e.g. slow voltage shifts due to
sweating). Also, the reference should not pick up signals which are not
intended to be recorded, like heart activity, which would be "subtracted
in" by the referencing. In most studies, a reference on the head but at
some distance from the other recording electrodes is chosen. Such a reference
can be the ear-lobes, the nose, or the mastoids (i.e. the bone behind the
ears). With multi-channel recordings (e.g. >32 channels), it is common to
compute the "average reference", i.e. to subtract the average over
all electrodes from each electrodes for each time point. This distributes the
"responsibility" over all electrodes, rather than assigning it to
only one of them. If a single reference electrode was used during the
recording, it is always possible to re-reference the data to any of the
recording electrodes (or combinations of them, like their average) at a later
stage of processing. In some cases "bipolar" recordings are carried
out, where electrode pairs are applied and referenced against each other for
each pair (e.g. left-right symmetrical electrodes).

 

From Tarik Bel Abar

 

in my opinion, average reference is
a good bet when you have denser coverage of the head with enough electrode density,
something like 60 and above, preferably closer to 100 and above. But it also
depends on what reference was active during recording, depending on the make of
your nets. For basic introductions you can check out the links to various
resources for your education regarding re-referencing. 

0. http://dionysus.psych.wisc.edu/lit/Topics/Psychophysiology/CacioppoText/Ch.2.pdf    

1. [PDF] from uni-konstanz.de   

2. http://www.springerlink.com/content/6785124141026861/fulltext.pdf   

3. http://sccn.ucsd.edu/pipermail/eeglablist/2008/002510.html   

4. [PDF] from tu-ilmenau.de   

5. http://www.springerlink.com/content/6785124141026861/   

6. http://onlinelibrary.wiley.com/doi/10.1111/1469-8986.3850847/abstract 

 

From Stephen Politzer-Ahles

 

Steve Luck's book *An Introduction
to the Event-related potentials method *has a lot of information on references
and the consequences of  the reference
you choose. A common issue with average reference is that it's best when you
have a high-density cap with many channels, so it may not be a good reference
to use with a 19-channel dataset. (I don't know if the amount of noise matters
particularly for reference, but I don't have any experience myself using
average reference--since I've only worked with low-density caps, I've never had
a chance to use average reference).

 

From Yin Tian

 

In recent years, Dr Yao designed a method to translate the
actual recordings with an actual physical reference, such as Cz, uni-mastoid,  to the idea recordings with reference at
Infinity (zero theoretically) (REST-- reference electrode standardization
technique, 2001). It has been evaluated and compared with the other common
references thoroughly. According to their detailed numeric evaluation, in all
cases, the best one is REST (zero), then the average reference.  The linked-ears, or uni-mastord is not
recommended. With my practice, REST does improve the data quality in ERP and
EEG analysis. The followings are the papers showing the details

 

Yao D.A method to standardize a reference of scalp EEG
recordings to a point at infinity. Physiol Meas 2001;22:693–711.         

 

From Joe Dien

 

19-channels is a bit sparse for
average reference but it also depends on how they are distributed on the head,
not just the number.  It also depends on
the ERPs you're interested.  The mastoids
are definitely not electrically silent (they are located very close to the
lateral temporal language areas).  There
are also implications for statistical power of ANOVAs depending on the geometry
of the ERP component dipolar field and the reference site.  It also depends on what kind of analyses you
are planning on and also on how well known the ERP component is and whether you
plan on making inferences about functional neuroanatomy, among other
things.  Regarding noise, it depends on
what kind of noise you're talking about. 
Do you mean electrode noise (bad contacts that would only affect the one
electrode) or background noise (coherent electrical fields influencing multiple
electrodes that just happen to not be of interest, including EOG) or mechanical
noise (movement artifacts affecting multiple electrodes caused by physical
disturbances of the electrodes)?

 

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.