Pitfalls When Scanning Two Brains In Synchrony

By Neuroskeptic | January 12, 2014 12:43 pm

The past few years have seen many neuroscientists becoming interested in ‘hyperscanning‘. Rather that contenting themselves to scan just one brain at a time, hyperscanners simultaneously measure activity from two (or even more) people, using techniques such as fMRI and EEG.

This technically demanding method is said to provide information about what happens in the brain(s) during social interactions. In particular, many people are excited by the ability of hyperscanning to measure synchronization (or ‘hyperconnectivity’) of activity between interacting brains.

However, British neuroscientist Adrian P Burgess raises a note of caution in a new paper: On the interpretation of synchronization in EEG hyperscanning studies

The problem stems from how we choose to define ‘synchrony’ mathematically. One popular metric is called the phase-locking value (PLV) over time; this is a measure of the degree to which two waves (in this case, two frequency bands from the activity of two different brains) have a consistent relative phase.

The trouble is that two waves of the same frequency will tend to keep the same relative phase, even if they’re not interacting with each other. So the PLV will tend to systematically overstate synchrony.

It gets worse. The PLV could also ‘detect’ changes in synchrony in response to external stimuli. Burgess illustrates this tendency to produce spurious effects by taking some standard EEG data from a series of individuals, and then pairing them up, treating them as if they were part of a hyperscanning study.

He shows that statistically significant changes in  PLV ‘synchrony’ occur in response to simple stimuli, such as the individuals closing their eyes, or seeing a picture on a screen. See below: in this case the spurious synchronization is in the theta band:


Burgess concludes:

Existing measures of hyper-connectivity are biased and prone to detect coupling where none exists. In particular, spurious hyper-connections are likely to be found whenever any difference between experimental conditions induces systematic changes in the rhythmicity of the EEG.

He explains that had these dubious effects been observed in a genuine hyperscanning experiment, instead of a mock-up, they might have been seen as real (and interesting):

This strong and systematic pattern of findings using PLV in these very different paradigms is troubling because, in the absence our knowledge that these hyper-connections must be spurious, they might easily have been accepted as real… The real problem is that, although the PLV is widely used as a measure of phase synchronization, a high value of PLV does not necessarily mean there is any true phase synchronization at all.

Burgess recommends other synchrony measures, such as the circular correlation coefficient (CCorr), which, as he shows in various simulations, are much more robust. So this paper ought to be read closely by anyone involved in this kind of research. Fortunately, not all hyper-synchrony studies have used the PLV.

I have to confess that I’ve never been all that convinced by the idea of brain synchrony in terms of the phase of brainwaves. Taking the alpha rhythm as an example, everyone has alpha waves buzzing away at a frequency of roughly 10 Hz but we never notice them (except perhaps when looking at this optical illusion). If the phase of my alpha waves suddenly reset to be in synch with yours… neither of us would notice. So what would it mean?

Phase synchrony is probably the simplest measure of hyperconnectivity… but that doesn’t mean it’s the best one.

ResearchBlogging.orgAdrian P. Burgess (2013). On the interpretation of synchronization in EEG hyperscanning studies: a cautionary note Front Hum Neurosci DOI: 10.3389/fnhum.2013.00881

  • Guillaume

    While it is clear that any future hyperscanning study must choose Circular Correlation (CCOR) or Kraskov Mutual Information (KMI) instead of Phase Locking Value (PLV), previous results still indicate at least a joint “systematic changes in the rhythmicity of the EEG.”

    Moreover, Burgess precises in the discussion that under conditions that can be “matched in terms of stimulus presentation and movement, and if appropriate control conditions are used, then much of these problem would be resolved.”

    Overall, measuring simultaneously the activity in multiple brains reveals different types of information. First, the fact that there are indeed two brains—from the same species usually—and not two random number generators. Second, how these two brains are immersed in the same task and perceptual context, and how any change affects them in a similar fashion. Third, to what extend communication at the social level can increase the similarity of their dynamics. Since the phase of brainwaves has been associated with behavior in both motor and perceptual processes, it seems reasonable to study how sharing information at the sensorimotor level could impact the phase dynamics at the collective level. In the end, the big issue concerns the functional role of brainwaves, which is still an open question.

  • Jon Brock

    I’ve never really understood the hype around hyperscanning. Even if my brain is synchronized to your brain it’s not actually responding to your brain – it’s responding to your behaviour (which may itself be contingent on my behaviour, which is itself contingent on my brain).

    Just looking at synchronization of brain responses seems to be ignoring what’s actually interesting and important, which is how the brain (of one person) coordinates perception and action during social interactions.

    • Laith Alexander

      I don’t think hyperscanning was ever meant to measure a brain’s response to another brain. It’s just meant to measure how two brains change their activity in response to cues from another individual – which is what we do when we interact with another person.

      No-one claimed that we were measuring psychic ability :-)

      • Jon Brock

        If I want to know “how brains change their activity in response to cues from another individual”, why would I try and do that by taking behaviour out of the equation?

        • Guillaume

          Technically, the behavior is still there since it is used as a control parameter. This allows to study interindividual variability at the dyadic scale and in real-time, which is important in the case of spontaneous social interaction.

          • Jon Brock

            What do you mean by control parameter in this context?

          • http://www.twitter.com/introspection Introspection

            I mean the parameter which differentiates the experimental conditions that are compared. For instance, interactional synchrony or imitation role. Then, you can look at how certain behavior will change the similarity/dissimilarity of brain activity between the two subjects in interaction.

    • teddybowties

      and what generates that behaviour? your brain.


  • Enrico Glerean

    As an advocate of inter-subject correlation and similar methods, I find it interesting to see when brain activity synchronises across participants, not only as a strong measure of reliability but also as a predictor of behaviour/stimulus processing. I think the main take home message of this paper is not about avoiding PLV in hyperscanning, but more about performing robust statistical tests that take into account external factors (e.g. stimulus presentation) by using bootstrap permutations or bayesian methods.



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