This document discusses the heterogeneity of 10Hz rhythms seen in EEG data and proposes guidelines for their proper measurement and analysis. It presents a tentative dictionary of various 10Hz rhythms distinguished by their spatial distribution, frequency localization, and functional significance. It also puts forth a theory relating EEG spectral peaks to instantaneous brain oscillation patterns, and how the time scale of analysis impacts which patterns appear as peaks. Analyzing 10Hz rhythms at a fine spectral resolution and temporal scale can provide insights into distinct brain processes and functions.

1. Emmanuelle Tognoli & J. A. Scott Kelso
Center for Complex Systems and Brain Sciences, Florida Atlantic University, Boca Raton, FL-33431 – USA
tognoli@ccs.fau.edu ; kelso@ccs.fau.edu

2. AIM
Historically, “alpha” band has either been (1) treated as a coherent
whole, (2) divided into “low-” and “high-alpha” or (3) subdivided into a
limited set of anatomo-functional processes such as alpha, rolandic mu,
tau and sometimes parietal mu.
High-resolution spectral analysis (Tognoli et al., 2007) opened up the
discovery of numerous rhythms in the 10Hz band, i.e. stable,
reproducible peaks in induced EEG spectra that emerge as properties
of a diversity of brain functional networks. A tentative functional
dictionary of these rhythms is presented below.
The main goals of this communication are to stress the heterogeneity of
10Hz rhythms, to discuss strategies for their proper measurements and
to propose a theory of rhythms and temporal scales in which inter-
individual and intra-individual variability may be explained.

3. A non-exhaustive list of 10Hz rhythms in human
waking EEG
Induced EEG rhythms were obtained with high-resolution
FFT of multiple EEG epochs each lasting between 1 and
16 sec. Epochs were multiplied with a Tukey window to
prevent leakage, and were padded when necessary to
achieve high-spectral resolution. EEG spectra are
visualized with a 4D colorimetric mapping to identify
spatial organization. Rhythms are distinguished on the
basis of three criteria: spatial distribution, spectral
localization and reactivity/functional significance.

4. mAL
Seen during behavioral
arrest following right
hand or finger
movement.
10-12Hz typical.
mM
Seen during rest states, and
modulated during movement
steady-states.
8-10Hz typical
mCR
Seen during imitation tasks
(data from Bernier, Dawson,
Webb & Murias, 2007). Is
associated with mCL.
10-12Hz typical.
e
Seen during covert attention
tasks. Is associated with x.
10-12Hz typical.
x
Seen during covert
movements of spatial
attention. Sometimes
associated with e.
10-12Hz typical.
mCL
Seen during imitation
tasks (data from
Bernier et al., 2007).
Is generally
associated with mCR.
10-12Hz typical.
f1,f2
Seen during social
coordination and possibly
during multisensory
integration. Composed of 2
components that are
independently modulated.
10-12Hz typical.
n
Seen during imitation
tasks. Signification
unknown.
8-10Hz typical.
Bilateral a variant
Asymmetrical variant seen with
eyes open. Possibly related to
asymmetrical selective attention
to left and right hemifields.
9-10Hz typical.
a
Seen during eye
closed and drowsy
states.
9-10Hz typical.

5. Example of a spectrum with a large number of 10Hz
rhythms. Power is best measured at peak electrode,
where confound from other rhythms is minimal.

6. STRATEGIES FOR PROPER MEASUREMENT
RHYTHMS SEPARATION AND ACCURATE
POWER ESTIMATES:
When distinct rhythms are not separated, power is
distorted by confound in respective changes from
each peak (see example below). Condition-
dependent change may be missed if one rhythm
decreases and another increases nearby. The
change may be ascribed to a location that reflect
neither of the rhythms...etc. Peaks should be
separated either by careful identification of each
peak boundary or by spectral decomposition.

7. Example of bias in power
measurement due to improper
separation of 2 distinct rhythms.

8. SPECTRAL OVERLAP AND SPECTRAL-SPATIAL
DISTORTION:
Because of the crowding in the 10Hz band, there is
potential for overlap: 2 rhythms close enough in frequency
to be embedded in one another. The consequence is
merging of spatial and spectral features: apparent peak is
displaced at the intersection domain of both originating
components (well known examples are bilateral rhythms
that appear medially –e.g. alpha–). Spatial and spectral
distortion are harmful to the proper interpretation of
components’ functional properties. Frequency analysis with
a fine spectral resolution (in the order of 0.1-0.2Hz) greatly
minimizes this problem. Choice of the time scale (principle
4, right) also helps, with shorter temporal windows most
likely to reveal the true properties of each rhythm.

9. FREQUENCY BOUNDARIES AND POWER
ESTIMATION:
In several subjects, the same rhythm may appear
at different frequencies and peaks may exhibit
different width (see example of n in 4 subjects on
the left). Rhythm identification and power
estimation should be performed on a subject-by
subject basis.

10. n, small amplitude, 9Hz
n, large amplitude,
11Hz
n, broad band, 9Hz
n, partially buried, 9Hz

11. A THEORY OF EEG OSCILLATIONS AND
TEMPORAL SCALES
Atoms of EEG spectra are instantaneous
spatio-temporal patterns (Tognoli & Kelso,
2009; see left picture below). Those patterns
arise from self-organized neural activity in
two basic types of cortical structures: sulci
(e.g. first pattern) and gyri (e.g. last pattern).

13. To circumvent the issue of Signal-to-Noise Ratio
and obtain “stable” frequency estimates,
investigators usually compute the power
spectrum over a cumulative time of minutes.
In the 10Hz range, patterns typically last one or
two cycles. Five minutes of EEG may contain
~2000 patterns.
Which of these patterns contribute to spectral
peaks and how?

14. PRINCIPLE 1: ONLY SUSTAINED PATTERNS SURVIVE
INTO SPECTRAL PEAKS
If a pattern can be sustained over a significant period of
time over the time-scale of spectral analysis (duration and
recurrence), it may be seen as a spectral peak. If a pattern
cannot be sustained, it will blend in the floor of EEG spectra
with other transient oscillations, irrespective of its
importance for brain dynamics and function.
PRINCIPLE 2: SPECTRAL PEAKS ARE SPATIALLY
REORGANIZED
Spectral peaks never show spatial organization of sulcal
patterns (two spatially distinct maxima). It suggests that
cumulative power maps are composed of a non-uniform
group of spatio-temporal patterns sharing some spatial and
spectral properties.

15. Slowly
changing
drowsy EEG

16. Fast changing
active EEG

17. PRINCIPLE 3: ACTIVATION AFFECTS
PATTERNS’ RATE OF CHANGE
In drowsy EEG, patterns’ rate of change is slow.
The longer the time window over which EEG
spectra are computed, the most likely rhythm will
appear.
In active EEG, many patterns occur, none of
which may stand out from the crowd. The longer
the time window over which EEG spectra are
computed, the least likely rhythms will appear :
transient peaks will blend in the “spectral floor”.

19. PRINCIPLE 4: SHORTER (WELL-DEFINED)
TEMPORAL WINDOWS ARE BETTER
Most mental processes may not have the ability to
be sustained over seconds to minutes (either at
once or in iterations). In such cases, long spectral
windows favor task-unrelated brain rhythms (fillers)
and are detrimental to understanding task-related
brain rhythms.
To emphasize task-related transient rhythms, task
should be shorter, and dependent variables should
be collected to pinpoint the temporal onset of
mental processes.

20. CONCLUSION
Ten Hz oscillations are often conceived as resting EEG states of little
cognitive or behavioral relevance. It has further been shown that this
band had no correlation with the bold signal (Niessing et al., 2005). Yet
the time scale of this oscillation (~100-300 msec) is of prime
importance for cognition and behavior and an excess peak at 10Hz
over the 1/f distribution of EEG power spectra suggests that it is a
preferred frequency in the human brain. We present a tentative
dictionary of 10Hz rhythms, advance guidelines for their accurate
measurement and analysis, and propose a theory that relates EEG
rhythms in spectral average with instantaneous patterns in continuous
EEG. Distinguishing the variety of 10Hz rhythms, understanding their
functional significance and properly estimating their individual
contributions in spectral studies are challenges that will be determinant
for advance in non-invasive electrophysiology: a lexicon of brain
functional processes is sketched that could be useful for basic and
clinical science, neuro-feedback, brain computer interface and neuro-
pharmacological studies.

21. REFERENCES
Bernier, R., Dawson, G., Webb, S. & Murias, M. (2007). EEG mu rhythm
and imitation impairments in individuals with autism spectrum disorder.
Brain and Cognition, 64: 228-237.
Kelso, J. A. S. 1995 Dynamic Patterns: the Self-Organization of Brain and
Behavior. MIT Press: Cambridge, MA.
Niessing, J., Ebisch, B., Schmidt, K.E., Niessing, M., Singer W., Galuske,
R.A. (2005). Hemodynamic signals correlate tightly with synchronized
gamma oscillations. Science, 309: 948-951.
Tognoli, E., Lagarde, J., De Guzman, G. C., Kelso, J. A. S. 2007 The phi
complex as a neuromarker of human social coordination. Proc. Natl. Acad.
Sci. U.S.A. 104, 8190-8195.
Tognoli E., (2008). EEG coordination dynamics: neuromarkers of social
coordination. In Fuchs A, Jirsa VK (eds.) Coordination: Neural, Behavioral
and Social Dynamics. Springer.
Tognoli E., Kelso J.A.S (2009). Brain Coordination Dynamics: True and
False Faces of Phase Synchrony and Metastability. Progress in
Neurobiology, 87(1): 31-40.