Several weeks ago, I bought myself a nice little gadget, the Muse S brain sensing headband, which is a consumer-grade EEG device for sleep tracking and neurofeedback. I use it predominantly for recording my brain waves during unguided meditation so that I can play around with the resulting EEG data.
Being the overzealous data scientist that I am, I quickly (and likely prematurely in terms of data volume) jumped into all sorts of statistical tests and time-series analyses, partly reproducing, partly failing to reproduce some well-known scientific findings on the measurement of meditative states with electroencephalography.1 But I won’t bore you with the details here, the more so as I’m the only subject in all my experiments anyway. During anomaly detection, however, I found something so striking and interesting that I just cannot abstain from writing a blog post about it. So let’s get right to it.
In addition to diligently recording the first thirteen minutes of my daily meditation practice, I also tracked several variables such as amount of sleep, physical exercise, eating protocol (currently ketogenic with intermittent fasting), caffeine intake, stress throughout the day, subjective level of distraction during meditation, and a few more. The last of these variable, labeled “comments,” captures extraordinary events. Over the course of 51 recorded days, I logged 7 such events that I thought were likely to influence my meditation session—and I’m happy I did, because they show something amazing.
To start with a bang, here are the most significant outliers I found in the data:
This graph illustrates, for every meditation session (dates on the x axis), the daily variance of absolute delta band power differences in the right hemisphere (y axis, exponentiated by 6 for dramatic effect). And what we can clearly see are four spikes, which all occured on those days where I had munched on some magic mushrooms.
Ok, so what does this mean? First of all, raw EEG signals are usually decomposed into different frequency bands, and the oscillations with the lowest frequencies are called delta waves. These slowest of all brain waves typically emerge during deep, dreamless sleep and correlate with loss of bodily awareness. Second, my EEG device has four active electrodes: one on each side of the forehead (AF7, AF8) for signals from the frontal lobe, the others behind each ear (TP9, TP10) for signals from the temporoparietal junction. Third, we can calculate the absolute power of the delta band for each of these channels by taking the logarithm of the sum of the spectral density of the EEG data over the delta frequency range (here, 1-4 hertz).
For the graph above, I simply subtracted the delta band power at electrode TP10 from that at electrode AF8 for every second of recording and absolutized the difference before computing the variance of these power differences for each day.2 So what the spikes indicate is that the similarity between the power in the frontal lobe and the power in the temporoparietal junction was highly unstable over the course of the corresponding meditation sessions, at least for delta waves in the right hemisphere. I did not find the same pattern for other frequency bands, nor in the left hemisphere, nor between opposite electrodes across hemispheres.
The following chart provides a clearer picture of what the variability of that difference looks like for a single recording, in particular, for the highest spike in the previous graph. It also suggests that the spike is unlikely to represent a mere artifact and that the delta band power tends to be higher at TP10 compared to AF8, as indicated by the positive means (dashed horizontal lines).
Maybe you find my difference measure to be somewhat contrived? Good, let’s simplify things! We will, in fact, find something even more intriguing.
In the next diagram, we ditch the channel interaction and inspect the right hemisphere channels separately. More precisely, we look at the variance of their delta band powers for every day of recording:
Could this be any more interesting?
Here we have six spikes, each associated with the consumption of some kind of drug. In addition to psilocybin, we now also see the day on which I had a hangover from drinking on the night before and, to my great surprise, even a spike on the day I received my second dose of Pfizer-BioNTech’s vaccine against COVID-19, roughly 17 hours after which I experienced the whole range of common side effects. (Note that outside of these six days I did not drink any alcohol or take any other drugs, medical or recreational, except for the occasional morning coffee.)
Again, we can take a closer look at the power variability during selected meditation sessions:
Remember when I mentioned that I logged an “extraordinary event” on 7 out of 51 days? Six of these were due to drug intake. The seventh was when I jumped out of a flying plane for the first time in my life (and suffered a headache afterwards). And as it so happens, we see a huge spike in delta band power variance on this day as well, namely, in the left anterior frontal lobe:
Isn’t this just awesome? Seven extraordinary days and precisely seven graphical spikes related to delta oscillations.
The averages of the delta band powers don’t differ significantly when comparing ordinary and extraordinary days; their variances, however, likely do.3 So the critical difference seems to lie in the (in)stability of low-frequency brain waves over time. Drugs and extreme activities appear to destabilize the brain’s delta rhythms, while psilocybin in particular seems to also affect the coordination of that dispersive disruption within the right half of the brain.
Yet this blog post is not a scientific paper and my data is just a single-subject snapshot, so I won’t embarrass myself going any deeper into interpretation mode. Here are just a few hints from related findings, produced by full-fledged scientific studies:
- Psilocybin-induced experiences correlate with the lagged phase synchronization of delta oscillations.4
- The peak of a DMT experience coincides with the emergence of delta oscillations.5 (I tried to replicate this finding with analyses based on the IRASA algorithm and could generate somewhat comparable results, but far from being as impressive as the plots above.)
- One study reported that psilocybin reduces coherence in theta, alpha, and beta bands, but unfortunately I can only access the abstract, which doesn’t mention delta waves.6 (I computed coherence, but did not get convincing results for the delta band.)
- Children with severe intellectual disability due to Angelman’s syndrome show more dynamic and diffuse7 delta power in their brains.8
- Children suffering from Dravet Syndrome show diffuse delta activity during epileptic seizures.9
- Schizophrenia, which we know is mimicked symptomatically by the psychotic effects of psilocybin, is associated with diffuse delta rhythms.10
Finally, I should briefly mention some technical details: I recorded my own brain waves with a 4-channel Muse S headband (affiliate link) and the Mind Monitor app. The sampling frequency was 220 hertz, the notch frequency was 50 hertz, the recording interval was 1 second, the recording time was in evenings between 6 and 10 pm, and the duration per recording was ~13 minutes of meditation practice, of which 600 seconds were extracted using the interval [-630:-30] in order to prevent start and end artifacts from tainting the analysis. During signal pre-processing, artifacts like eye blinks and jaw clenches were removed, together with 47 seconds of poor connection quality, resulting in a total of 31612 time points used for analysis.
As mentioned in footnote 3, everything discussed so far has resulted from explorations to find hypotheses, not from tests to confirm them. Now, one week later, I conducted an n=1 experiment, recording my brain waves both prior to taking a medium dose (2 grams) of magic mushrooms as well as during the ensuing psychedelic experience, each while meditating for 20 minutes in a sitting position with my eyes closed.
My initial hypotheses suggested the following predictions:
- The variance of absolute delta band powers will be increased during the psilocybin-induced experience.
- The variance of absolute delta band power differences in the right hemisphere will also be increased.
- The mean of absolute delta band powers will not be affected significantly.
- High-frequency bands will remain unaffected.
While the hypothesis underlying my second prediction was falsified, the three other events occurred largely as predicted, as you can see in the following plots:
Thus, the effects of psilocybin on my brain appear to be minor in terms of the means of absolute band powers (at best a small reduction in mean delta band power), but large when looking at their variances, which more than doubled for the delta frequency range of 1-4 hertz. This confirms my hypothesis that psilocybin disrupts the meditative brain by causing instability in slow brain waves.
- For a great review, see: Deolindo CS, Ribeiro MW, Aratanha MA, Afonso RF, Irrmischer M and Kozasa EH (2020) A Critical Analysis on Characterizing the Meditation Experience Through the Electroencephalogram. Front. Syst. Neurosci. 14:53. doi: 10.3389/fnsys.2020.00053.
- Formulaically, my difference metric is comparable to a well-researched measure called frontal alpha asymmetry (FAA); what I did was switching from interhemispheric to intrahemispheric and changing the frequency band. I also experimented with some common measures of neuronal signal synchrony, but to little avail in terms of graphical impressiveness.
- Welch’s t-tests yielded 3.26 (p<0.05) for A8, 4.32 (p<0.01) for TP9, 3.04 (p<0.5) for TP10, and p>0.5 for A7, after Bartlett’s and Levine’s tests for equal variances had confirmed that the assumption of homoscedasticity is violated, especially at channels TP9 and TP10. Note also that the assumption of normality was rejected for the variance of delta band powers at the frontal channels (AF7, AF8), but not at the temporoparietal channels (TP9, TP10), by Shapiro-Wilk tests. Furthermore, my observations are time-dependent, which in the present case is not as bad as it could be (because a Dickey-Fuller test showed that the data is stationary and because I have been practicing meditation for many years), but still. Finally, what I’m presenting here are selected findings from an extensive exploratory data analysis; in order to run scientifically sound tests, I would first have to record additional data before testing any hypotheses. In brief, the statistics here should be taken with an ample dose of salt. For me, all this is more about having fun with data than about doing legit science.
- Kometer M, Pokorny T, Seifritz E, Volleinweider FX. Psilocybin-induced spiritual experiences and insightfulness are associated with synchronization of neuronal oscillations. Psychopharmacology (Berl). 2015 Oct;232(19):3663-76. doi:10.1007/s00213-015-4026-7. Epub 2015 Aug 1. PMID: 26231498.
- Timmermann, C., Roseman, L., Schartner, M., Milliere, R., Williams, L., Erritzoe, D., Muthukumaraswamy, S., Ashton, M., Bendrioua, A., Kaur, O., Turton, S., Nour, M. M., Day, C. M., Leech, R., Nutt, D. J., & Carhart-Harris, R. L. (2019). Neural correlates of the DMT experience assessed with multivariate EEG. Scientific reports, 9(1), 16324. https://doi.org/10.1038/s41598-019-51974-4.
- Palenicek, T., Tyls, F., Viktorinova, M., Prokopcova, D., Korcak, J., Androvicova, R., Brunovsky, M., & Horacek, J. (2016). PM502. Effect of psilocybin on EEG brain connectivity in healthy volunteers – preliminary report. International Journal of Neuropsychopharmacology, 19(Suppl 1), 83. https://doi.org/10.1093/ijnp/pyw041.502.
- Note that “diffuse” primarily means that an EEG activity is spread over large areas of both hemispheres, not that it is necessarily abnormal.
- (1) Joel Frohlich, Lynne M Bird, John Dell’Italia, Micah A Johnson, Joerg F Hipp, Martin M Monti, High-voltage, diffuse delta rhythms coincide with wakeful consciousness and complexity in Angelman syndrome, Neuroscience of Consciousness, Volume 2020, Issue 1, 2020, niaa005, https://doi.org/10.1093/nc/niaa005; (2) Sidorov, M.S., Deck, G.M., Dolatshahi, M. et al. Delta rhythmicity is a reliable EEG biomarker in Angelman syndrome: a parallel mouse and human analysis. J Neurodevelop Disord 9, 17 (2017). https://doi.org/10.1186/s11689-017-9195-8.
- Kim SH, Nordli DR Jr, Berg AT, Koh S, Laux L. Ictal ontogeny in Dravet syndrome. Clin Neurophysiol. 2015 Mar;126(3):446-55. doi:10.1016/j.clinph.2014.06.024. Epub 2014 Jun 30. PMID: 25046982.
- Matsuura M, Yoshino M, Ohta K, Onda H, Nakajima K, Kojima T. Clinical Significance of Diffuse Delta EEG Activity in Chronic Schizophrenia. Clinical Electroencephalography. 1994;25(3):115-121. doi:10.1177/155005949402500309.