Getting The Position Right For EEG

By Neuroskeptic | October 7, 2012 10:24 am

In science, it’s often the most ‘boring’, easily overlooked factors that determine whether an experiment succeeds or fails.

A new paper reveals strong effects of body posture on brain electrical activity: Subject position affects EEG magnitudes. Just lying face-up as opposed to face-down can powerfully affect the signal measured using electroencephalography (EEG), according to Justin Rice and colleagues of New York.

Here’s why: EEG uses electrodes, placed on the scalp, to measure the electrical potentials produced by brain firing.

The signal recorded depends, however, not just on the brain activity but also on the quality of the electrical conduction between the brain and the scalp: the signals have to travel through the fluids surrounding the brain, then the skull, and finally the skin, before they’re detected.

EEG users sometimes think of brain-scalp conductivity as a fixed factor, that they can’t control and don’t have to worry about. However, Rice et al point out that the position of the brain shifts within the skull depending upon your posture.

This is because there’s a bit of extra room in there, leaving space for the brain to “bounce around” a little within its fluid cavity. If you’re lying on your back, the brain will lie closer to the back of the skull; if you’re on your front, it’ll be further forward, and so on. So the fluid layer between brain and skull will be corresponding thinner, or thicker.

In a healthy brain the change is only about 1 mm, but the fluid layer’s only 3 mm at most, so that’s a big change.

Others have recognized this problem before, but Rice et al’s data are the clearest evidence yet that posture is a major factor. They showed that subjects lying on their back (supine) showed significantly stronger activity over the back of the brain – which makes sense, as it brings the brain closer to the electrodes. Lying face down (prone) made activity weaker and sitting was in between.

Interestingly – and worryingly – the effect was stronger depending upon the kind of activity being measured. For most kinds of brain activity it was about 40% higher but for gamma waves – the hottest thing in EEG right now – it was almost 80%.

So gamma band activity is especially sensitive to posture, and that raises the worrying possibility that even slight differences in head position between individuals could account for ‘differences’ in gamma power recorded, for example in studies comparing neurological patients and healthy controls; if the controls are sitting up straight while the patients are slouching back… the patients would seem to have more gamma.

ResearchBlogging.orgRice JK, Rorden C, Little JS, and Parra LC (2012). Subject position affects EEG magnitudes. NeuroImage PMID: 23006805

CATEGORIZED UNDER: bad neuroscience, EEG, methods, papers
  • Ivana Fulli MD

    Anyway, I was always told that EEG had a very good time resolution and poor spatial one…

    Is not MEG the future for serious, well thought upon and financed experiments, that is.

  • Anonymous

    I struggle to understand why, despite all the potential for misinterpretation, research likes so much machine mediated data and so little human feedback. Is there anybody trying to use both, for instance getting EEG readings while the subject reports brain zappings (or whatever you call that strong conscious feeling of electrical currents crossing your brain that some people experience)?

  • Bradley Voytek

    Very cool. The gamma findings aren't actually much of a surprise, however. Physiological electro- (and magneto-) encephalography displays a 1/f relationship between frequency and the power at that frequency. So low frequencies have much greater power than the high.

    A few years ago when I published the hemicraniectomy paper I showed that without the skull power is higher across all bands, including gamma. The idea there was that, without the skull in the way to act as a low-pass spatial filter, the signal would be stronger and more focal.

    Here, as the visual cortex moves closer to the skull (by laying down), the absolute change in power will be greater across all frequencies, but for the lower frequencies the proportional change will be relatively lower because the recorded signal is already closer to the “true” signal. In contrast, relatively weak signals such as in the gamma range will be proportionally much higher because the original signal-to-noise was so terrible to begin with.

    As a general rule of thumb, unless I have a very specific hypothesis for a gamma change I don't even bother with it in scalp EEG because there's so much gamma noise (saccade artifacts, muscle activity, etc.)

  • pulnimar

    Are there any internal controls which can be used as baselines to adjust for this?

    Could the average amplitude of the background noise be used as such a baseline?


    So, to beat a dead horse, one of the calibrations used for testing the results of fMRI is iffy making fMRI even more so.

  • Neuroskeptic

    Bradley: Good points. Actually, this paper confirms that a good chunk of the occipital gamma is neuronal (otherwise it wouldn't depend on posture… well… maybe some muscles would be different? But if you look at the time course, it looks kosher).

    And in terms of 'cleaning up' gamma there have been some interesting papers recently.

  • Matt Craddock

    With regard to cleaning up gamma, I've had some success with ICA based methods – specifically I've used Keren et al's method, and there's a similar one by Hassler et al. I'm working on a paper right now showing the application of this method in detail. I'd note, though, that the gamma in the paradigm in the paper above seems very different to the eye movement gamma, which is a broadband spike 200-300ms after stim onset and has quite a different topography. Incidentally, there's some evidence that MEG has problems with the same artefact.

    Anyway, in summary, gamma artefact are srs bsns.

    (Did Blogger eat my earlier post grumbling about the stats in this paper?)

  • LokaSamasta

    To clarify, are we saying that posture affects gamma activity, or that we cannot tell whether or not it does because posture affects measuring accuracy and calibration?

  • Bradley Voytek

    Neuroskeptic: I have few doubts that most gamma is “real” in this case, though there are still some major caveats about looking at gamma without eye-tracking, e.g., even microsaccades cause relatively massive parallax shifts in the visual environment, which changes the representation of the visual scene within the visual cortex, thus resulting in a huge spiking response which will show up as “real” gamma.

    Matt & Neuroskeptic: yeah, the Keren paper–who is in Leon Deouell's lab, who also published the now-infamous 2008 Neuron Yuval-Greenberg EEG/gamma/saccades paper–is excellent. But as I said, I have the luxury of often not having to bother because I have access to so much ECoG data that is far less susceptible to these artifacts, and has much greater power.

  • Brian Blackwell

    Some of the convo in the comments here is a bit over my head, but I wonder if you could use an accelerometer to control for changes in head position. You could have each subject move her head around and be able to precisely match changes in head orientation with changes in EEG readings.

  • Anonymous

    A number of studies have found cerebral blood flow differences related to body position, I wonder whether these effects are also related to the observed changes. It is perhaps too simplistic to suggest that more blood might get to the occipital cortex when supine due to gravity, though I guess it might be something to consider.

  • Neuroskeptic

    Anonymous: Hang on… that's actually a very good point that I hadn't considered. All other things being equal you'd expect more occipital blood in the supine position.

    Whether that would have any effect on EEG I don't know; but it might affect the fMRI signal…

  • Ivana Fulli MD

    Bradley Voytek,

    Can you tell the laypeople what neuroscientists are expecting from EEG research (when neurosurgery is not involved)?

    ( A clinician should be allowed to hope that some of the autisms -other than the few known genetic deseases associated with autism syndrome- might find “a labtest” someday from a good “time resolution tool” whatever the size of the signal because even the more able autistic persons seem to want a little more time than a neurotypical person in “taking in a sentence” – but will have other time related qualities like being able to look longer at things without fatigue or whatever)

  • Matt Craddock

    Ok, i'm a bit of a pedant, but…

    They never tested supine or prone against sitting; from their Figure 4 it looks like supine is similar to sitting in all conditions (or at least, with the error bars they have, isn't significantly different from sitting). I'd say the difference between supine and prone is probably a lot larger than the difference between sitting up straight and slouching back, which you highlight as a potential confound. Muscle noise would probably be a bigger problem there – people sitting up straight would probably be a lot tenser. And it's harder to think of a situation where one group would be on their face and the other on their back.

    supine vs prone isn't significant in any individual condition other than gamma, but that could also be down to stat power, since there are 5 extra participants in the gamma experiment. They never test any of the conditions against each other, so the stats aren't there to support the claim that the difference between prone and supine differs across different experiments/measures. I mean it looks plausible enough, like Bradley says, but…

    Loka: more that posture affects our ability to record gamma activity rather than it affecting gamma activity itself.

  • Richard

    @Bradley: cool paper (yours) – just skimmed it for the moment. I was recently at an epilepsy meeting where one of the speakers mentioned that many people (particularly clinicians) consider the skull to be a low-pass *frequency* filter, whereas actually it isn't (for frequencies of interest to EEG, even VFOs). However, the *spatial* filtering will make it appear to have a low-pass frequency filtering effect, as low frequency activity tends to engage larger areas of cortex than higher frequency activity, which is more focal and so more susceptible to spatial smearing. I guess your results, with broad-band power increase but a more focal signal, supports that, or am I misinterpreting?

    @Neuroskeptic nice summary – I guess the %age change in gamma power will depend on the generating mechanism behind the gamma and therefore the alignment of the dipoles with regard to the electrodes, so may be worse for some types of gamma than others (as described in their model)… though certainly looks like another factor worth controlling for in EEG when comparing frequency-power changes.

    Damn biology.

  • Bradley Voytek
  • DS

    Some of you have expressed interest in this CSF-positional dependence in MEG work.

    Yes, MEG would be expected to suffer as well. But unlike with EEG the sensors in MEG (SQUIDS) can move a lot with respect to the brain. MEG sensors are mounted on a helmet in which there is considerable room for subject movement – even for my big head. So this CSF-positional effect would probably be small compared to all other positioning problems of MEG sensors relative to brain.

  • DS

    … which is of course not good news.

  • Ivana Fulli MD


    I thought that in a MEG testing,the subject of study was always placed in a sitting position. Right or wrong?

    As a psychiatrist interested in aspies, I am a MEG hopeful since I am told it has a good time resolution and also because I feel bad to advise someone to volunteer without personal benefits to be investigated with a little radioactive material (PET) and for the clients I know, although they may not always like to be touched to place electrodes (NB: Do not touch autistic persons lightly but firmly and it is less unpleasant for many of them) they suffer more from the noise of RMI testing- the latter probably affect their brain images…

  • Richard

    @Bradley I think that's what I was trying to say, but said much more clearly, plus some evidence. Thanks :)

  • DS

    Ivana asked: “I thought that in a MEG testing,the subject of study was always placed in a sitting position. Right or wrong?'

    There are MEG systems that can be used for scanning in the sitting upright position (the usual) or the lying horizontal position. But the point is that there is plenty of room within the helmet thereby permitting the subject to move relative to the sensors.

    In the case of EEG the sensors are attached to the scalp thereby permitting less movement of the subject relative to the sensors. But as the article referenced by Neuroskeptic shows (see the EEG simulations) a little motion of the brain relative to the sensors (postulated by the paper to be the result of movement within the CSF) produces large changes in signal. So on a per-channel basis one would expect the situation to be worse for MEG than EEG.

    Some might argue that if one is localizing the source by “solving” the MEG inverse problem then it should not matter if the signals in the sensors varies as the head moves. That argument would make some sense if the MEG inverse problem were not so horribly ill posed!

  • Ivana Fulli MD


    Thanks for the explanations.

    Still, from a clinical point of view,not touching an autistic scalp and working in silence protected you from very serious bias indeed : studying the brain “signals” of any autistic person with a big noise around or electrodes on the scalp of the subject is a sure recipe to get their brain unable to function properly.

    Plus you make the person suffer without personal benefits.

  • LokaSamasta

    ralifeIvanna I'm asperger's – we're hoping the science will catch up soon and vindicate the medicines we've found useful all along – the controversial ones which are almost impossible to research!

    Celina, I'm checking out your blog, please have a look at Rat Park ( and which led to a novel and highly effective treatment protocol for those who suffer from substance use concerns and financial concerns –

  • LokaSamasta

    “ralife” was the capture code – dunno how it got there, sorry Ivanna!

    I'm a yoga teacher – I want to make sure that when this news gets out, no new yoga woo is generated along the lines of 'science proves yoga poses can be used to modulate gamma brain waves'.

  • omg

    Ivana, the whole thing is bullsh!t. I was reading a report from two different research radiographers based on the same MRI scans the other day. One was one of the best in the country, Yale graduate Ivy League fella. He said it was a bone, the other said it was a growth. Health researchers don't know anything. I'm still trying to figure out what neuroscientists actually do. You need a surgical specialist with practical training to decipher from their experience an appropriate interpretation. The surgeon said it was an airspace and he ended up being right. Perhaps it's not the tools of the trade but the experienced worker.

  • Anonymous

    In MEG, the subject's data is usually co-registered to their structural MRI, and the analysis is typically conducted in a space (the boundary element model) defined by the inner skull surface from that MRI scan. It is worth noting that most of the MEG scans are conducted in the upright position, while the MRI scans are conducted supine. (Although as someone already pointed out, the MEG scan can also be conducted in the supine position).

    The position of the subject's head relative to the MEG sensors is always known; in any modern MEG system this can be monitored continuously throughout the measurement, and older systems measured it at the beginning and end of each scan.

    The big question is, how much does the brain slosh around between the upright and supine position? This would be an interesting application of the new hybrid MEG/ultra-low-field-MRI systems (e.g., Vesanen et al, doi:10.1002/mrm.24413)

  • DS

    The accuracy of the position measurement of the head (not the brain) in MEG is as probably greater than the small amount of sloshing around that the brain does in the CSF. So that does not help much. Given that the inverse problem is so horribly ill-posed and that nobody has ever propagated measurement error and constraint error through any of the MEG inverse problem calculations I am confident when I say that none of this matters anyhow.

    There's a lot of wishful thinking and story telling in the MEG source localization world.

  • Ivana Fulli MD


    Why should not the proud user of any “new hybrid MEG/ultra-low-field-MRI systems” be free to make a little publicity for it without encountering deadly irony?

    And why should researcher use any resource available when it is a noisy procedure instead of looking creatively at working in the least invasive way on the subjects of research?

    I understand that one could feel the need to remain anonymous in commenting on their subject of research on the broad sense for political reasons -or whatever- but I wish more anonymous would use a recognizable pseudo-even when they do not intend to charm NS admirers into reading their own blogs or buy what they spammed for!

    Even if the chosen initials are those of one's grand grand-mother or whatever, it is pleasant to be able to follow the thinking of some anonymous but recognizable commentators-and when lacking time, it makes possible to choose which comments to read instead of reading only the post.

  • omg

    Thank you Anon for the freshman txtbook refresher. The “big question” should be how much data is being lost to interpretation? I swear we live in an era of excessive analysis. Take any hay in a haystack and make it a needle. Everyone does it so excessively that elusive needle gets harder and harder to find. There's a term for it in politics, forget the name… it's polluting the campaign by pulling off political red herrings. Regulating science literature in the way of replication – might need to be more of that. Hi Ivana.

  • Neurofeedback

    This is a really fascinating article. Thank you.

  • Arturo Tozzi cns

    Apart from methodological concerns, this paper raises an important conceptual issue: is the body able to influence and modify brain function?



No brain. No gain.

About Neuroskeptic

Neuroskeptic is a British neuroscientist who takes a skeptical look at his own field, and beyond. His blog offers a look at the latest developments in neuroscience, psychiatry and psychology through a critical lens.


See More

@Neuro_Skeptic on Twitter


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Collapse bottom bar