Scientific Reports is an open access online journal published by Nature Publishing Group, and that’s a problem. The problem comes from a lot of cranks and quacks seeing dubious science published in Scientific Reports and believing that it’s the equivalent of Nature, one of the most famous, highest impact journals in existence. I first noticed Scientific Reports a year and a half ago, when I noticed a study that was, in essence, torturing mice in the name of pseudoscience published in the journal.
At the time, I made an analogy that was most definitely not flattering to Scientific Reports. Basically, I said that Scientific Reports was Nature’s answer to PLoS but in reality was increasingly striking me as Nature’s answer to Medical Hypotheses, the journal long famous for publishing “speculative” papers on how vaccines cause autism and a number of other scientifically dubious topics. Am I being harsh? At the time, I was increasingly getting the impression that Scientific Reports is the new go-to journal for pseudoscience. I also noted then that, because Scientific Reports is published by Nature Publishing Group (NPG), articles published there are sometimes mistakenly viewed as having been published in Nature or by Nature. Given that Nature is one of the top three or so scientific journals in the world, that mistake can lead articles appearing in Scientific Reports to be viewed as far more prestigious than they should be.
If you want to know the depths to which Scientific Reports will sink, though, all you have to do is to check out this article, which is the sort of article that makes me wonder what the hell the peer reviewers were thinking when they evaluated it. I first learned of this article from a Tweet:
Long-Term Study of Heart Rate Variability Responses to Changes in the Solar and Geomagnetic Environment in @SciReports … @Neuro_Skeptic @robustgar @schneiderleonid … HELP! https://t.co/w9oPDHbfE1
— Guillaume Dumas ??? (@introspection) March 20, 2018
The study is Long-Term Study of Heart Rate Variability Responses to Changes in the Solar and Geomagnetic Environment by Alabdulgader et al. It’s definitely a doozy. Certainly, I can’t recall ever having seen anything quite like it. My skeptical antennae began twitching right away when I did a little Googling on the HeartMath Institute, the institute in California behind this research. For example, here’s its fundraising pitch:
HMI is conducting innovative research on the interconnectivity between human consciousness and Earth’s energetic systems as well as between people and other living systems. There is evidence that Earth’s magnetic fields help to synchronize, energize and support the interconnection of all living systems.
Science is only scratching the surface of the many benefits to come from the next level of conscious awareness of our interconnectivity with others, animals, trees and nature.
My first thought reading this is that this sounds a hell of a lot like something that Deepak Chopra might have come up with. Now get a load of its “reearch,” which it calls Global Coherence Research:
The idea that the earth, sun and other planetary bodies influence human health and behavior and, on a larger scale, social unrest and significant global events, has been discussed among scientists for decades. The study of interconnectedness, which is not yet fully embraced in the mainstream scientific community, is still in infancy.
Global coherence research uses a multidisciplinary approach that includes the geosciences and astrophysics as well as extensive data from human and animal studies that are correlated to social and global events. The Global Coherence Initiative (GCI) employs the Global Coherence Monitoring System to collect a variety of data, information about Earth’s magnetic field and how it affects and is influenced by human emotions and behaviors.
Seriously, this sounds almost as though the Global Coherence Project is trying to validate astrology. In any case, you can see the problem with this sort of research by some of the data that the Global Coherence Institute cites:
Historically, many cultures believed their collective behavior could be affected by the sun and other external cycles and influences. This belief has proven to be true. On a larger societal scale, increased violence, crime rate, social unrest, revolutions and frequency of terrorist attacks have been linked to the solar cycle and the resulting disturbances in the geomagnetic field. The first scientific evidence of this was provided by Alexander Tchijevsky, a Russian scientist who noticed that more severe battles in World War I occurred during peak sunspot periods. Tchijevsky then conducted a thorough study of global human history dating back to 1749 and compared the occurrence of key events with the occurrence of solar cycles over the same time period until 1926.
Can you see the problem? Basically, how you define “key events” will hugely affect the findings in any such correlative study. In any case, based on these ideas, the HMI sizes:
- Human and animal health, cognitive functions, emotions and behavior are affected by solar, geomagnetic and other earth-related magnetic fields.
- The earth’s magnetic field is a carrier of biologically relevant information that connects all living systems.
- Every person affects the global information field.
- Collective human consciousness affects the global information field. Therefore, large numbers of people creating heart-centered states of care, love and compassion will generate a more coherent field environment that can benefit others and help offset the current planetary discord and incoherence.
Yes, this is definitely some Deepak Chopra-level woo. Don’t believe me? Check out this 2008 video from the Global Coherence Initiative:
And here’s a 2016 video:
“Science-based.” You keep using that word. I do not think it means what you think it means.
So let’s take a look at the paper itself. The first thing I noticed is that the corresponding author, is a real scientist. He’s the director of the NASA Ames Genome Research Facility (Mountain View, CA), where he has pioneered the development of largescale functional genomics projects, including high resolution tiling arrays for the entire human genome and various model organisms. Other investigators include Abdullah Alabdulgader, who is a cardiologist at the Prince Sultan Cardiac Center in Saudi Arabia. The rest of the investigators seem to be from the HMI or the Lithuanian University of Health Sciences.
So what did the investigators do? sizing that autonomic nervous system (ANS) activity, as reflected by heart rate variation (HRV) measured by EKG, responds to changes in geomagnetic and solar activity. To justify this sis, they write one of the longest introductions I’ve seen in a scientific paper in a long time. Basically, in any paper that is not a Cell paper, I tend to grow a bit suspicious of introductions that are this long and convoluted. In any event, to test this sis, researchers recruited 18 females, all employees of the Prince Sultan Cardiac Center in Hufuf, Saudia Arabia to participate. This included 8 members of the nursing staff, 6 members of the housekeeping staff, and four from the research department). The mean age was 32 ± 8 years, and subjects had no known physical or mental health disorders. None took medications known to affect autonomic function. All participants underwent weekly 24–72 hour ambulatory HRV recordings with Bodyguard HRV recorders. These recorders sample the ECG at 1000 Hz and calculates the inter-beat-interval (IBI), which is the time in milliseconds between consecutive heartbeats. Data were stored locally and periodically uploaded. Participant recordings were generally 72-hours in length and were scheduled once a week over a 5-month period between April and the end of August 2012. A total of 960 twenty-four hour long HRV recordings were obtained.
OK, right away I wondered something. Why just 72 hour recordings? For instance, I once had to undergo a 30 day monitor study back when it was suspected that I had an arrhythmia. That was something like 15 years ago or more. Back then the devices were bulky, and I can’t believe I carried the electronic box around with me all the time except when taking a shower, but I did. (It didn’t show anything.) I’d be willing to bet that the devices are much smaller and less annoying to wear now. Another thought that came to mind was: What about smartwatches? The HMI proclaims itself to be all about embracing technology, but it used technology that was downright primitive. Not only can smart watches detect heartbeats and are being studied to detect, for instance, atrial fibrillation through the characteristically irregularly irregular rhythm, but some are even being developed with EKG leads. Be that as it may, heart rhythm fluctuations are separated into three primary frequency bands: high frequency (HF), low frequency (LF), and very low frequency (VLF)44. The HF range equates to rhythms with periods that occur between 2.5 and 7 seconds and is thought to reflect parasympathetic activity primarily related to the respiratory cycle. The LF range equates to rhythms and modulations with periods that occur between 7 and 25 seconds and has been suggested to reflect sympathetic activity in addition to baroreceptor (blood pressure detector) activity involved in short-term blood pressure regulation.
Environmental parameters were measured as described:
Space weather and environmental measures were obtained from three sources, comprising nine measures. The solar wind speed, Kp index, Ap index, number of sunspots, F10.7 index, and the geomagnetic polar cap index (PCN) were downloaded from NASA Goddard Space Flight Center’s Space Physics Data Facility as part of the Omni 2 data set. Cosmic ray counts were downloaded from Finland’s University of Oulu’s Sodankyla Geophysical Observatory’s website. Power in the time varying magnetic field in two frequency bands, Schumann Resonance Power (SRP), 3.5 to 36 Hz and ULF power, 2 mHz to 3.5 Hz were obtained from a recording site located in Boulder Creek, California. Table 2 summarizes frequency band ranges for the HRV and magnetic field measures used in this study. The HeartMath Institute maintains a network of highly sensitive induction coil magnetometers (Zonge ANT-4; sensitivity 10−12 T) as part of a special project called the Global Coherence Initiative87. Each site includes two magnetometers positioned in the north-south and east-west axis to detect local time varying magnetic field strengths over a relatively wide frequency range (0.001–50 Hz) while maintaining a flat frequency response. The data acquisition infrastructure collects and timestamps all data using GPS time signals before uploading to a common server. Each magnetometer is continuously sampled at a rate of 130 Hz. Figure 1 shows the time domain data for the environmental data activity across the study period.
The authors then used multivariate linear regression analysis to look for correlations between each HRV variable and the environmental variables described above. They applied the Bonferonni correction for multiple comparisons after removing Circadian effects. What they claim to have found is that autonomic nervous system activity was affected by solar wind and other factors. For instance, they claim that increases in solar wind intensity were correlated with increases in heart rate, which the investigators interpreted as a biological stress response. Increases in cosmic rays, solar radio flux, and Schumann resonance power were all associated with increased HRV and parasympathetic activity. You can look at all the graphs, if you like. To me, visually they’re quite noisy and not particularly convincing.
I can’t help but note a number of problems with this study. First, eighteen subjects is a very small sample set for a study like this, and two of them dropped from the study because of discomfort from the electrodes. That’s only sixteen subjects! I mean, seriously. This is just a correlational study. These subjects were not undergoing any intervention other than periodic monitoring of EKG. I’m supposed to take it seriously that the HMI could only recruit 18 women from one hospital in Saudi Arabia, where, I must admit, I’m a bit skeptical of the strength of human subject research protections? Next up, as some of you have been discussing in the comments of this post, we’re looking at effects that, if they exist at all, are likely to be small. How on earth can one separate such small effects from the much larger effects of confounders, the amount of exercise participants were getting (and when), diet, and the like? I don’t see that the investigators even tried.
Another problem, noted on Twitter by Neuroskeptic, is this:
So I took a quick look at the paper. I think the problem is autocorrelation of the timeseries. You can't treat a timeseries as a vector of independent observations.
— Neuroskeptic (@Neuro_Skeptic) March 20, 2018
I think you’re right: Bonferroni assumes independence. That + huge amount of noise -> fatal
— Guillaume Rousselet (@robustgar) March 20, 2018
I think you’re right: Bonferroni assumes independence. That + huge amount of noise -> fatal
— Guillaume Rousselet (@robustgar) March 20, 2018
And they’re right. The Bonferonni correction assumes independence of the variables in the multiple comparisons being corrected for, but clearly the various measures of HRV are not independent both because they are a time series and because some of the measures are calculated from others. And all of this from only sixteen subjects! Basically, this study is not even particularly useful as a pilot project.
It’s also useful to look at the plausibility of the hypothesis here. Is it possible that geomagnetic conditions can influence human biology? Sure. Is it likely to be a large effect? Not really, given how small the magnitude of the energy experienced by an individual from such magnetic and electrical fields is. If such effects occur, was the design of this study such that it was likely to detect them? Definitely not. The number of subjects is too small, and its sampling of HRV seems not to be prolonged enough. Also, the study investigators used what appear to me to have been the incorrect statistical analyses to detect such correlations. Indeed, I strongly suspect that, had the correct statistical modeling been used, no effect would have been detected.
I haven’t seen naturopaths gloating over this study yet, but I do know that they definitely like this project. Indeed, a naturopath named Robert Kachko wrote an article with one of the study authors, Rollin McCraty, touting the sort of research the HMI is doing.
As for Scientific Reports, it is not Nature, its provenance as a Nature Publishing Group journal notwithstanding. Unfortunately, it does seem to be becoming the go-to journal that still has some respectability for publishing dodgy research. It really needs to clean up its act.
21 replies on “The Global Coherence Initiative: Woo on a global scale”
Is that “Medical Hypotheses” in the second paragraph?
Thanks – feel free to delete this!
Not only are the HRV variables not independent, the environmental variables used in the multivariate regression are not independent. Kp is derived from Ap, so those variables will be very highly correlated. Sunspot number and F10.7 are two different ways of measuring solar activity, so those variables will be highly correlated. And obviously there is a causal relationship between the variations of solar wind speed and sunspot number/F10.7 and the variations of Kp/Ap and PCN. To me it looks like the authors of this study made no effort to learn what those various indices are supposed to mean.
Isn’t there a far more fundamental issue with the tiny study group working in the same facility, likely living in the same area and consequently being exposed to essentially identical far more influential environmental factors like workplace stress and weather? (Or is this what’s meant by “autocorrelation of the timeseries”?)
The small sample size and existence of confounders is mentioned in the OP, yes. I’m focusing on the parts of the study about which I actually know something.
Let’s just say it would be shorter and easier to list what they did right than to list what they did wrong.
Science Reports was creepy Kent Anderson’s baby before he definitely did not get fired, just by the by.
^ Whoops, wrong open-access checkbook.
Scientific Reports are actually the reason why Nature Publishing group dumped Frontiers (where such a paper would not raise any eyebrows) after heavily investing in them.
After NPG merged with Springer, Frontiers remained completely outside, and are now owned by NPG’s mother company Holtzbrinck and some investors. Now Scientific Reports robs Frontiers of their bunk, with very competitive pricing and Nature family association!
I blogged about that failed merge here, in this satirical April Fools post: https://forbetterscience.com/2016/06/14/conspiracy-theory-is-npg-being-assimilated-by-frontiers/
Wait until the woo-meisters get a hold of this!
The old goat prevailing over prn.fm is continuously ramping up fear about solar flares’ effects on the grid.
Now imagine if solar flares at high intensity interfered with our precious bodily fl… I mean heart rhythms
Orac (and Guillaume Rousselet) seems to have misunderstood the problem with using Bonferroni correction for dependent data. It does guarantee that the type I error rate is less than the chosen level alpha, but the power suffers greatly in this case, particularly when there are many dependent variables.
Proof that the false positive rate is below nominal (in LaTeX code):
Let $A_1,A_2, \ldots, A_n$ be events that null hypotheses 1 through n are rejected in error, respectively. Then their union is the event that at least one null hypothesis is rejected in error. The probability of at least one rejected null hypothesis is then $P(A_1 \cup A_2 \cup \ldots \cup A_n) \leq P(A_1)+P(A_2)+\ldots+P(A_n)$ (the right hand side is an overestimate because we are ignoring overlaps between the events).\
If we set $P(A_1)=P(A_2)=\ldots=P(A_n)=\alpha/n$, then the right hand sum becomes $\alpha$, so we have successfully bounded our false positive rate to be no more than $\alpha$.\
I can illustrate the reduction in power as follows:
Suppose we had 100 statistical tests, of which there are only two independent tests and the rest of which correspond perfectly to one of the independent tests. Then the Bonferroni correction (with alpha=0.05) would require us to set each test to have false rejection rate 0.0005, and the overall false positive rate is in fact 2(0.0005)-(0.0005^2), or approximately 0.1%. When we crank up confidence level like that, we must live with an attendant decrease in power.
Now, correlation doesn’t equal causation, which is what I think it going on here. Both humans and solar phenomena exhibit medium term trends (Icorrect me if I’m wrong here), so the two should be correlated. Doesn’t mean one is causing the other any more than my age cohort’s increasing mortality rate over time is caused by the increase in petroleum prices.
WordPress supports LaTeX. Just use ‘$latex’ instead of ‘$’ to invoke it.
Thanks. Let me try that.
Let be events that null hypotheses 1 through n are rejected in error, respectively. Then their union is the event that at least one null hypothesis is rejected in error. The probability of at least one rejected null hypothesis is then (the right hand side is an overestimate because we are ignoring overlaps between the events).
If we set , then the right hand sum becomes , so we have successfully bounded our false positive rate to be no more than . Note that we made no particular assumption about the events, independence or otherwise.
The Global Coherence Initiative was started by a woo-peddler named Gregg Braden, who Dr Stuart Robbins over at the Exposing PseudoAstronomy podcast covered in his 17th episode where he showed that Braden was using ‘cherry picking’ to obtain supporting data for his claims.
Back in the leadup to 2012, Braden was heavily involved in promoting claims that time was slowing down, that the Earth was going to ‘tip over’, etc, etc.
I though that the end of that would have been the end of him, but obviously the group he created is still going.
In fact the HMI goes further back, to a company selling EmWave biofeedback stress-relieving devices / software, which our host encountered back in 2006:
Selling the EmWave2 still seems to be their core activity; the rest of the intellectual superstructure is secondary scams.
This study is partly right. If I go down to a beach around Phuket, my heart rate will definitely change. All though the change will have little to do with the sun but with clothing (or lack there of) by certain people on the beaches.
This study just seems to be a rehash of the idea of Gaia.
But that lack of clothing is correlated with latitude, as it doesn’t occur in Barrow, AK.
The problem of autocorrelation brought up by Neuroskeptic is not directly related to the Bonferroni correction or its assumptions about independence. The Bonferroni correction is applied to fix the problem of inflating type I error by testing many regressions, while autocorrelation in the time series biases the significance WITHIN each specific regression – unrelated issues.
In any case, the assumption of independence in the Bonferroni correction doesn’t mean that in non-independent tests the type I error rate will not be controlled – actually the other way around: when tests are dependent, then in effect you need to correct for LESS multiple comparisons (as an extreme case, running 1000 identical tests is the same as running just one – no correction required). So if the independence assumption is invalid you are actually over-correcting – increasing the probability of type II error rather than type I.
As long as you feel comfortable, clothes matter not.
[…] study published in Scientific Reports that I’ve discussed. In fact, a mere two weeks ago, I discussed a particularly dodgy study published in Scientific Reports that claimed to have found that human heart rate variation (HRV) […]
[…] Nature Publishing Group that has been known for a propensity to publish dubious studies promoting New Age nonsense and even antivaccine pseudoscience. Now, I will admit to some articles that I have submitted to […]
How closely, really, does the arterial pulse correspond to heart electrical activity? Is its response sort of ‘smoothed out’? I’ve had some instances where there was NO arterial pulse for a few beats though the heart was fine. I’ll wait for the watches with wireless EKG electrodes.
The Global Coherence Initiative. Hmm, perhaps that explains this: