Transduction of the Geomagnetic Field as Evidenced from Alpha-band Activity in the Human Brain

Abstract

Magnetoreception, the perception of the geomagnetic field, is a sensory modality well-established across all major groups of vertebrates and some invertebrates, but its presence in humans has been tested rarely, yielding inconclusive results. We report here a strong, specific human brain response to ecologically-relevant rotations of Earth-strength magnetic fields. Following geomagnetic stimulation, a drop in amplitude of EEG alpha oscillations (8-13 Hz) occurred in a repeatable manner. Termed alpha event-related desynchronization (alpha-ERD), such a response is associated with sensory and cognitive processing of external stimuli. Biophysical tests showed that the neural response was sensitive to the dynamic components and axial alignment of the field but also to the static components and polarity of the field. This pattern of results implicates ferromagnetism as the biophysical basis for the sensory transduction and provides a basis to start the behavioral exploration of human magnetoreception.

Full text

!
1!
Transduction of the Geomagnetic Field as Evidenced from 1!
Alpha-band Activity in the Human Brain!2!
Connie X. Wang1, Isaac A. Hilburn2, Daw-An Wu1,3, Yuki Mizuhara4, Christopher P. Cousté2, 3!
Jacob N. H. Abrahams2, Sam E. Bernstein5, Ayumu Matani4, Shinsuke Shimojo1,3*, & Joseph L. 4!
Kirschvink2,6* 5!
1Computation & Neural Systems, California Institute of Technology, Pasadena, CA, USA. 2Division of Geological 6!
& Planetary Sciences, California Institute of Technology, Pasadena, CA, USA. 3Division of Biology & Biological 7!
Engineering, California Institute of Technology, Pasadena, CA, USA. 4Graduate School of Information Science and 8!
Technology, the University of Tokyo, Bunkyo-ku, Tokyo, Japan. 5Department of Computer Science, Princeton 9!
University, Princeton NJ, USA. 6Earth-Life Science Institute, Tokyo Institute of Technology, Ookayama, Meguro, 10!
Tokyo, Japan. * Corresponding Authors: pmag.contact@caltech.edu 11!
12!
13!
14!
Abstract !15!
Magnetoreception, the perception of the geomagnetic field, is a sensory modality well-16!
established across all major groups of vertebrates and some invertebrates, but its presence in 17!
humans has been tested rarely, yielding inconclusive results. We report here a strong, specific 18!
human brain response to ecologically-relevant rotations of Earth-strength magnetic fields. 19!
Following geomagnetic stimulation, a drop in amplitude of EEG alpha oscillations (8-13 Hz) 20!
occurred in a repeatable manner. Termed alpha event-related desynchronization (alpha-ERD), 21!
such a response is associated with sensory and cognitive processing of external stimuli. 22!
Biophysical tests showed that the neural response was sensitive to the dynamic components and 23!
axial alignment of the field but also to the static components and polarity of the field. This 24!
pattern of results implicates ferromagnetism as the biophysical basis for the sensory transduction 25!
and provides a basis to start the behavioral exploration of human magnetoreception.!26!
27!
Introduction 28!
Magnetoreception is a well-known sensory modality in bacteria (Frankel & Blakemore, 29!
1980), protozoans (Bazylinski, Schlezinger, Howes, Frankel, & Epstein, 2000) and a variety of 30!
animals (Johnsen & Lohmann, 2008; Walker, Dennis, & Kirschvink, 2002; R. Wiltschko & W. 31!
Wiltschko, 1995), but whether humans have this ancient sensory system has never been 32!
conclusively established. Behavioral results suggesting that geomagnetic fields influence human 33!
orientation during displacement experiments (Baker, 1980, 1982, 1987) were not replicated 34!
(Able & Gergits, 1985; Gould & Able, 1981; Westby & Partridge, 1986). Attempts to detect 35!
human brain responses using electroencephalography (EEG) were limited by computational 36!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
2!
methods of the time (Sastre, Graham, Cook, Gerkovich, & Gailey, 2002). Twenty to thirty years 37!
after these previous flurries of research, the question of human magnetoreception remains 38!
unanswered. 39!
In the meantime, there have been major advances in our understanding of animal 40!
geomagnetic sensory systems. An ever-expanding list of experiments on magnetically-sensitive 41!
organisms has revealed physiologically-relevant stimuli as well as environmental factors that 42!
may interfere with magnetosensory processing (Lohmann, Cain, Dodge, & Lohmann, 2001; 43!
Walker et al., 2002; R. Wiltschko & W. Wiltschko, 1995). Animal findings provide a potential 44!
feature space for exploring human magnetoreception – the physical parameters and coordinate 45!
frames to be manipulated in human testing (J. Kirschvink, Padmanabha, Boyce, & Oglesby, 46!
1997; W. Wiltschko, 1972). In animals, geomagnetic navigation is thought to involve both a 47!
compass and map response (Kramer, 1953). The compass response simply uses the geomagnetic 48!
field as an indicator to orient the animal relative to the local magnetic north/south direction 49!
(Lohmann et al., 2001; R. Wiltschko & W. Wiltschko, 1995). The magnetic map is a more 50!
complex response involving various components of field intensity and direction; direction is 51!
further subdivided into inclination (vertical angle from the horizontal plane; the North-seeking 52!
vector of the geomagnetic field dips downwards in the Northern Hemisphere) and declination 53!
(clockwise angle of the horizontal component from Geographic North, as in a man-made 54!
compass). Notably, magnetosensory responses tend to shut down altogether in the presence of 55!
anomalies (e.g. sunspot activity or local geomagnetic irregularities) that cause the local magnetic 56!
field to deviate significantly from typical ambient values (Martin & Lindauer, 1977; W. 57!
Wiltschko, 1972), an adaptation that is thought to guard against navigational errors. These 58!
results indicate that geomagnetic cues are subject to complex neural processing, as in most other 59!
sensory systems. 60!
Physiological studies have flagged the ophthalmic branch of the trigeminal system (and 61!
equivalents) in fish (Walker et al., 1997), birds (Beason & Semm, 1996; Elbers, Bulte, Bairlein, 62!
Mouritsen, & Heyers, 2017; Mora, Davison, Wild, & Walker, 2004; Semm & Beason, 1990) and 63!
rodents (Wegner, Begall, & Burda, 2006) as a conduit of magnetic sensory information to the 64!
brain. In humans, the trigeminal system includes many autonomic, visceral and proprioceptive 65!
functions that lie outside conscious awareness (Fillmore & Seifert, 2015; Saper, 2002). For 66!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
3!
example, the ophthalmic branch contains parasympathetic nerve fibers and carries signals of 67!
extraocular proprioception, which do not reach conscious awareness (Liu, 2005). 68!
If the physiological components of a magnetosensory system have been passed from 69!
animals to humans, then their function may be either subconscious or only weakly available to 70!
conscious perception. Behavioral experiments could be easily confounded by cognitive factors 71!
such as attention, memory and volition, making the results weak or difficult to replicate at the 72!
group or individual levels. Since brain activity underlies all behavior, we chose a more direct 73!
electrophysiological approach to test for the transduction of geomagnetic fields in humans. 74!
!75!
Materials and Methods 76!
We constructed an isolated, radiofrequency-shielded chamber wrapped with three nested 77!
sets of orthogonal square coils, using the four-coil design of Merritt et al. (Merritt, Purcell, & 78!
Stroink, 1983) for high central field uniformity (Fig. 1, and in the section on Extended Materials 79!
and Methods below). Each coil contained two matched sets of windings to allow operation in 80!
Active or Sham mode. Current ran in series through the two windings to ensure matched 81!
amplitudes. In Active mode, currents in paired windings were parallel, leading to summation of 82!
generated magnetic fields. In Sham mode, currents ran antiparallel, yielding no measurable 83!
external field, but with similar ohmic heating and magnetomechanical effects as in Active mode 84!
(J.L Kirschvink, 1992). Active and Sham modes were toggled by manual switches in the distant 85!
control room, leaving computer and amplifier settings unchanged. Coils were housed within an 86!
acoustically-attenuated, grounded Faraday cage with aluminum panels forming the walls, floor 87!
and ceiling. Participants sat upright in a wooden chair on a platform electrically isolated from 88!
the coil system with their heads positioned near the center of the uniform field region and their 89!
eyes closed in total darkness. (Light levels within the experimental chamber during experimental 90!
runs were measured using a Konica-Minolta CS-100A luminance meter, which gave readings of 91!
zero, e.g. below 0.01 ± 2% cd/m2.) The magnetic field inside the experimental chamber was 92!
monitored by a three-axis Applied Physics SystemsTM 520A fluxgate magnetometer. EEG was 93!
continuously recorded from 64 electrodes using a BioSemiTM ActiveTwo system with electrode 94!
positions coded in the International 10-20 System (e.g. Fz, CPz, etc.). Inside the cage, the 95!
battery-powered digital conversion unit relayed data over a non-conductive, optical fiber cable to 96!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
4!
a remote control room, ~20 meters away, where all power supplies, computers and monitoring 97!
equipment were located. 98!
99!
Fig. 1. Schematic illustration of the experimental setup. The ~1 mm thick aluminum panels of 100!
the electrically-grounded Faraday shielding provides an electromagnetically “quiet” 101!
environment. Three orthogonal sets of square coils ~2 m on edge, following the design of Merritt 102!
et al. (Merritt et al., 1983), allow the ambient geomagnetic field to be altered around the 103!
participant’s head with high spatial uniformity; double-wrapping provides an active-sham for 104!
blinding of experimental conditions (J.L Kirschvink, 1992). Acoustic panels on the wall help 105!
reduce external noise from the building air ventilation system as well as internal noise due to 106!
echoing. A non-magnetic chair is supported on an elevated wooden base isolated from direct 107!
contact with the magnetic coils. The battery-powered EEG is located on a stool behind the 108!
participant and communicates with the recording computer via an optical fiber cable to a control 109!
room ~20 m away. Additional details are available in the Extended Materials and Methods 110!
section, and Fig. 5 below. This diagram was modified from the figure “Center of attraction”, by 111!
C. Bickel (Hand, 2016), with permission. 112!
113!
114!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
5!
A ~1 hour EEG session consisted of multiple ~7 minute experimental runs. In each run 115!
of 100+ trials, magnetic field direction rotated repeatedly between two preset orientations with 116!
field intensity held nearly constant at the ambient lab value (~35 µT). In SWEEP trials, the 117!
magnetic field started in one orientation then rotated smoothly over 100 milliseconds to the other 118!
orientation. As a control condition, FIXED trials with no magnetic field rotation were 119!
interspersed amongst SWEEP trials according to pseudorandom sequences generated by 120!
software. Trials were separated in time by 2-3 seconds. The experimental chamber was dark, 121!
quiet and isolated from the control room during runs. Participants were blind to Active vs. Sham 122!
mode, trial sequence and trial timing. During sessions, auditory tones signaled the beginning and 123!
end of experiment runs, and experimenters only communicated with participants once or twice 124!
per session between active runs to update the participant on the number of runs remaining. 125!
When time allowed, Sham runs were matched to Active runs using the same software settings. 126!
Active and Sham runs were programmatically identical, differing only in the position of 127!
hardware switches that directed current to run parallel or antiparallel through paired loops. Sham 128!
runs served as an additional control for non-magnetic sensory confounds, such as sub-aural 129!
stimuli or mechanical oscillations from the coil system. (Note that experimental variables 130!
differing between runs are denoted in camel case as in DecDn, DecUp, Active, Sham, etc., 131!
whereas variables that change within runs are designated in all capitals like FIXED, SWEEP, 132!
CCW, CW, UP, DN, etc.). In Active runs, an electromagnetic induction artifact occurred as a 10-133!
20 microvolt fluctuation in the EEG signal during the 100 ms magnetic field rotation. This 134!
induction artifact is similar to that observed in electrophysiological recordings from trout 135!
whenever magnetic field direction or intensity was suddenly changed in a square wave pattern 136!
(Walker et al., 1997). Strong induced artifacts also occur in EEG recordings during transcranial 137!
magnetic stimulation (TMS) (Veniero, Bortoletto, & Miniussi, 2009). In all cases, the artifact 138!
can only be induced in the presence of time-varying magnetic fields and disappears once the 139!
magnetic field stabilizes (∂B/∂t=0). In our experiments, EEG data following the 100 ms field 140!
rotation interval were not subject to effects from the induction artifact. Furthermore, the 141!
induction artifact is phase-locked like an event-related potential and does not appear in analyses 142!
of non-phase-locked power, which we used in all subsequent statistical tests. Further discussion 143!
of electrical induction is in section 4 of Extended Materials and Methods, below. 144!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
6!
Fig. 2 shows the magnetic field rotations used. In inclination (Inc) experiments (Fig. 145!
2A), declination direction was fixed to North (0˚ declination in our coordinate system), and 146!
participants sat facing North. Rotation of the field vector from downwards to upwards was 147!
designated as an ‘Inc.UP.N’ trial and the return sweep as ‘Inc.DN.N’, with UP/DN indicating the 148!
direction of field rotation. In declination (Dec) experiments (Fig 2B, 2C), we held inclination 149!
(and hence the vertical component of the field vector) constant, while rotating the horizontal 150!
component clockwise or counterclockwise to vary the declination. For trials with downwards 151!
inclination (as in the Northern Hemisphere), field rotations swept the horizontal component 90˚ 152!
CW or CCW between Northeast and Northwest, designated as ‘DecDn.CW.N’ or 153!
‘DecDn.CCW.N’, respectively, with ‘.N’ indicating a Northerly direction. To test biophysical 154!
hypotheses of magnetoreception as discussed below, we conducted additional declination 155!
rotation experiments with static, upwards inclination. As shown in Fig. 2B, rotating an upwards-156!
directed field vector between SE and SW (‘DecUp.CW.S’ and ‘DecUp.CCW.S’) antiparallel to 157!
the downwards-directed rotations provides tests of the quantum compass biophysical model, 158!
while sweeping an upwards vector between NE and NW (‘DecUp.CW.N’ and ‘DecUp.CCW.N’) 159!
provides a general test for electrical induction (Fig. 2C). 160!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
7!
Fig. 2. Magnetic field rotations used in these 161!
experiments. In the first ~100 ms of each 162!
experimental trial, the magnetic field vector 163!
was either: 1) rotated from the first preset 164!
orientation to the second (SWEEP), 2) rotated 165!
from the second preset orientation to the first 166!
(also SWEEP), or 3) left unchanged 167!
(FIXED). In all experimental trials, the field 168!
intensity was held constant at the ambient lab 169!
value (~35 uT). For declination rotations, the 170!
horizontal rotation angle was +90 degrees or -171!
90 degrees. For inclination rotations, the 172!
vertical rotation angle was either +120 173!
degrees / -120 degrees, or +150 degrees / -150 174!
degrees, depending on the particular 175!
inclination rotation experiment. (A) 176!
Inclination rotations between ±60˚ or ±75˚. 177!
The magnetic field vector rotates from 178!
downwards to upwards (Inc.UP.N, red) and 179!
vice versa (Inc.DN.N, green), with declination 180!
steady at North (0˚). (B) Declination 181!
rotations used in main assay (solid arrows) 182!
and vector opposite rotations used to test the 183!
quantum compass hypothesis (dashed 184!
arrows). In the main assay, the magnetic field 185!
rotated between NE (45˚) and NW (315˚) with 186!
inclination held downwards (+60˚ or +75˚) as 187!
in the Northern Hemisphere (DecDn.CW.N 188!
and DecDn.CCW.N); vector opposites with 189!
upwards inclination (−60˚ or −75˚) and 190!
declination rotations between SE (135˚) and 191!
SW (225˚) are shown with dashed arrows 192!
(DecUp.CW.S and DecUp.CCW.S). (C) 193!
Identical declination rotations, with static but 194!
opposite vertical components, used to test the 195!
electrical induction hypothesis. The magnetic 196!
field was shifted in the Northerly direction 197!
between NE (45˚) and NW (315˚) with 198!
inclination held downwards (+75˚, 199!
DecDn.CW.N and DecDn.CCW.N) or 200!
upwards (−75˚, DecUp.CW.S and 201!
DecUp.CCW.S). The two dotted vertical 202!
lines indicate that the rotations started at the 203!
same declination values. In both (B) and (C), 204!
counterclockwise rotations (viewed from 205!
above) are shown in red, clockwise in green. 206!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
8!
During magnetic field rotations, EEG was recorded from participants in the eyes-closed 207!
resting state. Auditory cues marked the beginning and end of each ~7 minute run, but 208!
participants were not informed of run mode, trial sequence or stimulus timing. EEG was 209!
sampled at 512 Hz from 64 electrodes arrayed in the standard International 10-20 positions using 210!
a Biosemi™ ActiveTwo system. The experimental protocol was approved by the Caltech 211!
Institutional Review Board (IRB), and all participants gave written informed consent. 212!
We used conventional methods of time/frequency decomposition (Morlet wavelet 213!
convolution) to compute post-stimulus power changes relative to a pre-stimulus baseline interval 214!
(−500 to −250 ms) over a 1-100 Hz frequency range. We focused on non-phase-locked power by 215!
subtracting the event-related potential in each condition from each trial of that condition prior to 216!
time/frequency decomposition. This is a well-known procedure for isolating non-phase-locked 217!
power and is useful for excluding the artifact from subsequent analyses (Cohen, 2014). 218!
Following the identification of alpha band activity as a point of interest (detailed in Results), the 219!
following procedure was adopted to isolate alpha activity in individuals. To compensate for 220!
known individual differences in peak resting alpha frequency (8 to 12 Hz in our participant pool) 221!
and in the timing of alpha wave responses following sensory stimulation, we identified 222!
individualized power change profiles using an automated search over an extended alpha band of 223!
6-14 Hz, 0-2 s post-stimulus. For each participant, power changes at electrode Fz were averaged 224!
over all trials, regardless of condition, to produce a single time/frequency map. In this cross-225!
conditional average, the most negative time-frequency point was set as the location of the 226!
participant’s characteristic alpha-ERD. A window of 250 ms and 5 Hz bandwidth was 227!
automatically centered as nearly as possible on that point within the constraints of the overall 228!
search range. These time/frequency parameters were chosen based on typical alpha-ERD 229!
durations and bandwidths. Alpha power activity in each individualized window was used to test 230!
for significant differences between conditions. For each condition, power changes were 231!
averaged separately within the window, with trials subsampled and bootstrapped to equalize trial 232!
numbers across conditions. Outlier trials with extreme values of alpha power (typically caused 233!
by movement artifacts or brief bursts of alpha activity in an otherwise low-amplitude signal) in 234!
either the pre- or post-stimulus intervals were removed by an automated algorithm prior to 235!
averaging, according to a threshold of 1.5X the interquartile range of log alpha power across all 236!
trials. Further details are provided in sections 1-5 of Extended Materials and Methods, below. 237!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
9!
Results 238!
In initial observations, several participants (residing in the Northern Hemisphere) 239!
displayed striking patterns of neural activity following magnetic stimulation, with strong 240!
decreases in EEG alpha power in response to two particular field rotations: (1) Inclination 241!
SWEEP trials (Inc.UP.N and Inc.DN.N), in which the magnetic vector rotated either down or up 242!
(e.g. rotating a downwards pointed field vector up to an upwards pointed vector, or vice versa; 243!
Fig. 2A red and green arrows), and (2) DecDn.CCW.N trials, in which magnetic field declination 244!
rotated counterclockwise while inclination was held downwards (as in the Northern Hemisphere; 245!
Fig 2B, solid red arrow). Alpha power began to drop from pre-stimulus baseline levels as early 246!
as ~100 ms after magnetic stimulation, decreasing by as much as ~50% over several hundred 247!
milliseconds, then recovering to baseline by ~1 s post-stimulus; this is visualized by the deep 248!
blue color on the time-frequency power maps (Fig. 3). Scalp topography was bilateral and 249!
widespread, centered over frontal/central electrodes, including midline frontal electrode Fz when 250!
referenced to CPz. Fig. 3A shows the whole-brain response pattern to inclination sweeps and 251!
control trials (Inc.SWEEP.N and Inc.FIXED.N) of one of the responsive participants, with the 252!
alpha-ERD exhibited in the SWEEP but not FIXED trials. Similarly, Fig. 3B and 3C show the 253!
declination responses of a different participant on two separate runs (labeled Runs #1 and #2) six 254!
months apart. Response timing, bandwidth and topography of the alpha-ERD in the CCW 255!
sweeps, with negative FIXED controls, were replicated across runs, indicating a repeatable 256!
signature of magnetosensory processing in humans. After experimental sessions, participants 257!
reported that they could not discern when or if any magnetic field changes had occurred. 258!
259!
260!
261!
262!
263!
264!
265!
266!
267!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
10!
Fig. 3. Alpha-ERD as a neural response to 268!
magnetic field rotation. Post-stimulus power 269!
changes (dB) from a pre-stimulus baseline (−500 270!
to −250 ms) plotted according to the ±4 dB color 271!
bar at bottom. (A) Scalp topography of the 272!
alpha-ERD response in an inclination 273!
experiment, showing alpha power at select time 274!
points before and after field rotation at 0 s. 275!
Alpha-ERD (deep blue) was observed in 276!
SWEEP (top row), but not FIXED (bottom row), 277!
trials. (B) Scalp topography of the alpha-ERD 278!
response for two runs of the declination 279!
experiment, tested 6 months apart in a different 280!
strongly-responding participant. DecDn.CCW.N 281!
condition is shown. In both runs, the response 282!
peaked around +500 ms post-stimulus and was 283!
widespread over frontal/central electrodes, 284!
demonstrating a stable and reproducible 285!
response pattern. (C) Time-frequency maps at 286!
electrode Fz for the same runs shown in (B). 287!
Pink vertical lines indicate the 0-100 ms field 288!
rotation interval. Pink/white outlines indicate 289!
significant alpha-ERD at the p<0.05 and p<0.01 290!
statistical thresholds, respectively. Separate runs 291!
shown side by side. Significant alpha-ERD was 292!
observed following downwards-directed 293!
counterclockwise rotations (outlines in top row), 294!
with no other power changes reaching 295!
significance. Significant power changes appear 296!
with similar timing and bandwidth, while 297!
activity outside the alpha-ERD response, and 298!
activity in other conditions is inconsistent across 299!
runs. 300!
301!
302!
303!
304!
305!
306!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
11!
The alpha rhythm is the dominant human brain oscillation in the resting state when a 307!
person is not processing any specific stimulus or performing any specific task (Klimesch, 1999). 308!
Neurons engaged in this internal rhythm produce 8-13 Hz alpha waves that are measurable by 309!
EEG. Individuals vary widely in the amplitude of the resting alpha rhythm. When an external 310!
stimulus is suddenly introduced and processed by the brain, the alpha rhythm generally decreases 311!
in amplitude compared with a pre-stimulus baseline. (Hartmann, Schlee, & Weisz, 2012; 312!
Klimesch, 1999; Pfurtscheller, Neuper, & Mohl, 1994). This EEG phenomenon, termed alpha 313!
event-related desynchronization (alpha-ERD), has been widely observed during perceptual and 314!
cognitive processing across visual, auditory and somatosensory modalities (Peng, Hu, Zhang, & 315!
Hu, 2012). Alpha-ERD may reflect the recruitment of neurons for processing incoming sensory 316!
information and is thus a generalized signature for a shift of neuronal activity from the internal 317!
resting rhythm to external engagement with sensory or task-related processing (Pfurtscheller & 318!
Lopes da Silva, 1999). Individuals also vary in the strength of alpha-ERD; those with high 319!
resting-state or pre-stimulus alpha power tend to show strong alpha-ERDs following sensory 320!
stimulation, while those with low alpha power have little or no response in the alpha band 321!
(Klimesch, 1999). 322!
Based on early observations, we formed the hypothesis that sensory transduction of 323!
geomagnetic stimuli could be detectable as alpha–ERD in response to field rotations – e.g. the 324!
magnetic field rotation would be the external stimulus, and the alpha-ERD would be the 325!
signature of the brain beginning to process sensory data from this stimulus. This hypothesis was 326!
tested at the group level in data collected from 29 participants in the inclination rotation 327!
conditions (Fig 2A) and 26 participants in the declination rotation conditions (Fig. 2B, solid 328!
arrows). 329!
For inclination experiments, we collected data from matched Active and Sham runs 330!
(N=29 of 34; 5 participants were excluded due to time limits that prevented the collection of 331!
sham data). We tested for the effects of inclination rotation (SWEEP vs. FIXED) and magnetic 332!
stimulation (Active vs. Sham) using a two-way repeated-measures ANOVA. We found a 333!
significant interaction of inclination rotation and magnetic stimulation (p<0.05). Post-hoc 334!
comparison of the four experimental conditions (Active-SWEEP, Active-FIXED, Sham-SWEEP, 335!
Sham-FIXED) revealed significant differences between Active-SWEEP and all other conditions 336!
(p<0.05). Downwards/upwards rotations of magnetic field inclination produced an alpha-ERD 337!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
12!
~2X greater than background fluctuations in the FIXED control condition and all the Sham 338!
conditions. Results are summarized in Table 1 and Fig. 4A. 339!
In declination experiments (Fig. 4B), we observed a strikingly asymmetric response to 340!
the clockwise (DecDn.CW.N) and counterclockwise (DecDn.CCW.N) rotations of a downwards-341!
directed field sweeping between Northeast and Northwest. Alpha-ERD was ~3X greater after 342!
counterclockwise than after clockwise rotations, the latter producing alpha power changes 343!
indistinguishable from background fluctuations in the FIXED control condition. Over the 344!
participant pool (N=26 of 26 who were run in this experiment), we ran a one-way repeated-345!
measures ANOVA with three conditions (DecDn.CCW.N, DecDn.CW.N, and DecDn.FIXED.N) 346!
to find a highly significant effect of declination rotation (p<0.001) (Table 1). As indicated in 347!
Fig. 4B, the counterclockwise rotation elicited a significantly different response from both the 348!
clockwise rotation (p<0.001) and FIXED control (p<0.001). Fig. 4D shows the stimulus-locked 349!
grand average across all participants for each condition; an alpha-ERD is observed only for 350!
counterclockwise rotations of a downwards-directed field (left panel). Sham data were available 351!
for 18 of 26 participants in the declination experiments; no major changes in post-stimulus power 352!
were observed in any of the sham conditions (Fig. 4E). 353!
354!
355!
356!
357!
358!
359!
360!
361!
362!
363!
364!
365!
366!
367!
368!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
13!
369!
Fig. 4. Group results from repeated-measures ANOVA for the effects of geomagnetic 370!
stimulation on post-stimulus alpha power. (A) Average alpha-ERD (dB) at electrode Fz in the 371!
SWEEP and FIXED conditions of inclination experiments run in Active or Sham mode. Two-372!
way ANOVA showed an interaction (p<0.05, N=29) of inclination rotation (SWEEP vs. FIXED) 373!
and magnetic stimulation (Active vs. Sham). According to post-hoc testing, only inclination 374!
sweeps in Active mode produced alpha-ERD above background fluctuations in FIXED trials 375!
(p<0.01) or Sham mode (p<0.05). (B) Average alpha-ERD (dB) at electrode Fz in the 376!
declination experiment with inclination held downwards (DecDn). One-way ANOVA showed a 377!
significant main effect of declination rotation (p<0.001, N=26). The downwards-directed 378!
counterclockwise rotation (DecDn.CCW.N) produced significantly different effects from both 379!
the corresponding clockwise rotation (DecDn.CW.N, p<0.001) and the FIXED control condition 380!
(DecDn.FIXED.N, p<0.001). (C) Comparison of the declination rotations with inclination held 381!
downwards (DecDn) or upwards (DecUp) in a subset (N=16 of 26) of participants run in both 382!
experiments. Two-way ANOVA showed a significant interaction (p<0.01) of declination 383!
rotation (CCW vs. CW vs. FIXED) and inclination direction (Dn vs. Up). Post-hoc testing 384!
showed significant differences (p<0.01) between the DecDn.CCW.N condition and every other 385!
condition, none of which were distinct from any other. This is a direct test and rejection of the 386!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
14!
quantum compass hypothesis. (D) Grand average of time-frequency power changes across the 387!
26 participants in the DecDn experiment from (B). Pink vertical lines indicate the 0-100 ms field 388!
rotation interval. A post-stimulus drop in alpha power was observed only following the 389!
downwards-directed counterclockwise rotation (left panel). Wider spread of desychronization 390!
reflects inter-individual variation. Convolution involved in time/frequency analyses causes the 391!
early responses of a few participants to appear spread into the pre-stimulus interval. (E) Grand 392!
average of time-frequency power changes across the 18 participants with sham data in the 393!
declination experiments; no significant power changes were observed. 394!
395!
396!
397!
398!
399!
400!
401!
402!
403!
404!
405!
406!
407!
408!
409!
410!
411!
412!
413!
414!
415!
416!
417!
418!
419!
420!
421!
422!
423!
424!
425!
426!
427!
428!
429!
430!
431!
!432!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
15!
Group&Results&for&Effects&of&Magnetic&Field&Rotation&on&Post-Stimulus&Alpha&Power&
ANOVA&#1:&Inclination&Rotation&x&Magnetic&Stimulation&(N=29)&
Two-Way!Repeated!Measures!ANOVA!
F!
p!
ηp
2!
Main!Effect!of!Inclination!Rotation!
(SWEEP!vs.!FIXED)!
3.26!
0.08!
0.19!
Main!Effect!of!Magnetic!Stimulation!
(Active!vs.!Sham)!
2.47!
0.13!
0.09!
Inclination!Rotation!x!Magnetic!Stimulation!
(Interaction)!
5.67!
0.02*&
0.17!
!
ANOVA&#2:&Declination&Rotation&(N=26)&
One-Way!Repeated!Measures!ANOVA!
F!
p!
ηp
2!
Main!Effect!of!Declination!Rotation!
(CCW!vs.!CW!vs.!FIXED)!
13.09!
0.00003***&
0.34!
!
ANOVA&#3:&Declination&Rotation&x&Inclination&Direction&(N=16)&
Two-Way!Repeated!Measures!ANOVA!
F!
p!
ηp
2!
Main!Effect!of!Declination!Rotation!
(CCW!vs.!CW!vs.!FIXED)!
3.77!
0.03*&
0.24!
Main!Effect!of!Inclination!Direction!
(Dn!vs.!Up)!
0.89!
0.36!
0.06!
Declination!Rotation!x!Inclination!Direction!
(Interaction)!
6.49!
0.004***&
0.30!
433!
Table 1. Group results from repeated-measures ANOVA for the effects of magnetic field 434!
rotation on post-stimulus alpha power. ANOVA #1 shows a significant interaction of 435!
inclination rotation (SWEEP vs. FIXED) and magnetic stimulation (Active vs. Sham) in the 436!
inclination experiments. Based on post-hoc testing, alpha-ERD was significantly greater in 437!
SWEEP trials in Active mode, compared with all other conditions (p<0.05). ANOVA #2 shows a 438!
significant main effect of declination rotation when inclination is static and downwards as in the 439!
Northern Hemisphere. Alpha-ERD was significantly greater following counterclockwise 440!
rotations (p<0.001). ANOVA #3 shows a significant interaction of declination rotation and 441!
inclination direction in declination experiments designed to test the “Quantum Compass” 442!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
16!
mechanism of magnetoreception. A significant alpha-ERD difference (p<0.01) between 443!
counterclockwise down (DecDn.CCW.N) and counterclockwise up (DecUp.CCW.S) argues 444!
against this hypothesis in humans. !445!
446!
447!
The asymmetric declination response provided a starting point for evaluating potential 448!
mechanisms of magnetosensory transduction, particularly the quantum compass hypothesis, 449!
which has received much attention in recent years (Hore & Mouritsen, 2016; Ritz, Adem, & 450!
Schulten, 2000). Because the quantum compass cannot distinguish polarity, we conducted 451!
additional declination rotation experiments in which the fields were axially identical to those in 452!
the preceding DecDn experiments, except with reversed polarity (Fig. 2B; reversed polarity 453!
rotations shown as dashed arrows). In the additional DecUp conditions, Magnetic North pointed 454!
to Geographic South and up rather than Geographic North and down, and the upwards-directed 455!
field rotated clockwise (DecUp.CW.S) or counterclockwise (DecUp.CCW.S) between SE and 456!
SW. In later testing, we ran 16 participants in both the DecDn and DecUp experiments to 457!
determine the effects of declination rotation and inclination direction in a two-way repeated 458!
measures ANOVA with six conditions (DecDn.CCW.N, DecDn.CW.N, DecDn.FIXED.N, 459!
DecUp.CCW.S, DecUp.CW.S, and DecUp.FIXED.S). A significant interaction of declination 460!
rotation and inclination direction (p<0.01) was found (Fig. 4C and Table 1). DecDn.CCW.N 461!
was significantly different from all other conditions (p<0.01), none of which differed from any 462!
other. Thus, counterclockwise rotations of a downwards-directed field were processed 463!
differently in the human brain from the same rotations of a field of opposite polarity. These 464!
results contradict the quantum compass hypothesis, as explained below in Biophysical 465!
Mechanisms. 466!
From previous EEG studies of alpha oscillations in human cognition, the strength of 467!
alpha-ERD is known to vary substantially across individuals (Klimesch, 1999; Klimesch, 468!
Doppelmayr, Russegger, Pachinger, & Schwaiger, 1998; Pfurtscheller et al., 1994). In 469!
agreement with this, we observed a wide range of alpha-ERD responses in our participants as 470!
well. Some participants showed large drops in alpha power up to ~60% from pre-stimulus 471!
baseline, while others were unresponsive with little change in post-stimulus power at any 472!
frequency. Histograms of these responses are provided in Fig. 6-8 of Extended Materials and 473!
Methods below. 474!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
17!
To confirm that the variability across the dataset was due to characteristic differences 475!
between individuals rather than general variability in the measurement or the phenomenon, we 476!
retested the strongly-responding participants to see if their responses were stable across sessions. 477!
Using permutation testing with false discovery rate (FDR) correction at the p<0.05 and p<0.01 478!
statistical thresholds, we identified participants who exhibited alpha-ERD that reached 479!
significance at the individual level and tested them (N=4) again weeks or months later. An 480!
example of separate runs on the same participant is shown in Figs. 3B and 3C, and further data 481!
series are shown in the Fig 9 of Extended Materials and Methods. Each participant replicated 482!
their results with similar response tuning, timing and topography, providing greater confidence 483!
that the observed effect was specific for the magnetic stimulus in the brain of that individual. 484!
While the functional significance of these inter-individual differences is unclear, the 485!
identification of strongly responding individuals gives us the opportunity to conduct more 486!
focused tests directed at deriving the biophysical characteristics of the transduction mechanism. 487!
488!
Biophysical Mechanisms 489!
Three major biophysical transduction hypotheses have been considered extensively for 490!
magnetoreception in animals: (1) various forms of electrical induction (Kalmijn, 1981; 491!
Rosenblum, Jungerman, & Longfellow, 1985; Yeagley, 1947), (2) a chemical/quantum compass 492!
involving hyperfine interactions with a photoactive pigment (Schulten, 1982) like cryptochrome 493!
(Hore & Mouritsen, 2016; Ritz et al., 2000), and (3) specialized organelles based on biologically-494!
precipitated magnetite similar to those in magnetotactic microorganisms (J.L. Kirschvink & 495!
Gould, 1981). We designed the declination experiments described above to test these 496!
hypotheses. 497!
Electrical Induction. According to the Maxwell-Faraday law (∇ × E = -∂B/∂t), electrical 498!
induction depends only on the component of the magnetic field that is changing with time 499!
(∂B/∂t). In our declination experiments, this corresponds to the horizontal component that is 500!
being rotated. The vertical component is held constant and therefore does not contribute to 501!
electrical induction. Thus, we compared brain responses to two matched conditions, where the 502!
declination rotations were identical, but the static vertical components were opposite (Fig 2C). A 503!
transduction mechanism based in electrical induction would respond identically to these two 504!
conditions. Video 1 shows the alpha-ERD magnetosensory response of one strongly-responding 505!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
18!
individual to these two stimulus types. In the top row, the static component was pointing 506!
upwards, and in the bottom row, the static field was pointing downwards. In the DecDn.CCW.N 507!
condition (lower left panel), the alpha-ERD (deep blue patch) starts in the right parietal region 508!
almost immediately after magnetic stimulation and spreads over the scalp to most recording sites. 509!
This large, prolonged and significant bilateral desynchronization (p<0.01 at Fz) occurs only in 510!
this condition with only shorter, weaker and more localized background fluctuations in the other 511!
conditions (n.s. at Fz). No alpha-ERD was observed following any upwards-directed field 512!
rotation (DecUp.CCW.N and DecUp.CW.N, top left and middle panels), in contrast to the strong 513!
response in the DecDn.CCW.N condition. Looking across all of our data, none of our 514!
experiments (on participants from the Northern Hemisphere) produced alpha-ERD responses to 515!
rotations with a static vertical-upwards magnetic field (found naturally in the Southern 516!
Hemisphere). 517!
These tests indicate that electrical induction mechanisms cannot account for the neural 518!
response. This analysis also rules out an electrical artifact of induced current loops from the 519!
scalp electrodes, as any current induced in the loops would also be identical across the matched 520!
runs. Our results are also consistent with many previous biophysical analyses, which argue that 521!
electrical induction would be a poor transduction mechanism for terrestrial animals, as the 522!
induced fields are too low to work reliably without large, specialized anatomical structures that 523!
would have been identified long ago (Rosenblum et al., 1985; Yeagley, 1947). Other potential 524!
confounding artifacts are discussed in sections 6 and 7 of the Extended Materials and Methods, 525!
below. 526!
Quantum Compass. From basic physical principles, a transduction mechanism based on 527!
quantum effects can be sensitive to the axis of the geomagnetic field but not the polarity (Ritz et 528!
al., 2000; Schulten, 1982). In the most popular version of this theory, a photosensitive molecule 529!
like cryptochrome absorbs a blue photon, producing a pair of free radicals that can transition 530!
between a singlet and triplet state, with the transition frequency depending on the local magnetic 531!
field. The axis of the magnetic field – but not the polarity – could then be monitored by 532!
differential reaction rates from the singlet vs. triplet products. This polarity insensitivity, shared 533!
by all quantum-based magnetotransduction theories, is inconsistent with the group level test of 534!
the quantum compass presented above. The data (Table 1 and Figure 4C, dark blue bars) showed 535!
clearly distinct responses depending on polarity. We additionally verified this pattern of results 536!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
19!
at the individual level. Video 2 shows the alpha-ERD magnetosensory response in another 537!
strongly-responding individual. Only the DecDn.CCW.N rotation (lower left panel) yields a 538!
significant alpha-ERD (p<0.01 at Fz). Lack of a significant response in the axially identical 539!
DecUp.CCW.S condition indicates that the human magnetosensory response is sensitive to 540!
polarity. This means that a quantum compass-based mechanism cannot account for the alpha-541!
ERD response we observe in humans. 542!
543!
Response'Selectivity'544!
The selectivity of brain responses for specific magnetic field directions and rotations may 545!
be explained by tuning of neural activity to ecologically relevant values. Such tuning is well 546!
known in marine turtles in the central Atlantic Ocean, where small increases in the local 547!
geomagnetic inclination or intensity (that indicate the animals are drifting too far North and are 548!
approaching the Gulf Stream currents) trigger abrupt shifts in swimming direction, thereby 549!
preventing them from being washed away from their home in the Sargasso Sea (Light, Salmon, 550!
& Lohmann, 1993; Lohmann et al., 2001; Lohmann & Lohmann, 1996). Some migratory birds 551!
are also known to stop responding to the magnetic direction if the ambient field intensity is 552!
shifted more than ~ 25% away from local ambient values (W. Wiltschko, 1972), which would 553!
stop them from using this compass over geomagnetic anomalies. From our human experiments 554!
to date, we suspect that alpha-ERD occurs in our participants mainly in response to geomagnetic 555!
fields that reflect something close to "normal" in the Northern Hemisphere where the North-556!
seeking field vector tilts downwards. This would explain why field rotations with a static 557!
upwards component produced little response in Northern Hemisphere participants. Conducting 558!
similar experiments on participants born and raised in other geographic regions (such as in the 559!
Southern Hemisphere or on the Geomagnetic Equator) could test this hypothesis. 560!
Another question vis-à-vis response selectivity is why downwards-directed CCW 561!
(DecDn.CCW.N), but not CW (DecDn.CW.N), rotations elicited alpha-ERD. The bias could 562!
arise either at the receptor level or at higher processing levels. The structure and function of the 563!
magnetoreceptor cells are unknown, but biological structures exhibit chirality (right- or left-564!
handedness) at many spatial scales – from individual amino acids to folded protein assemblies to 565!
multicellular structures. If such mirror asymmetries exist in the macromolecular complex 566!
interfacing with magnetite, they could favor the transduction of one stimulus over its opposite. 567!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
20!
Alternatively, higher-level cognitive processes could tune the neural response towards 568!
counterclockwise rotations without any bias at the receptor level. As of this writing, we cannot 569!
rule out the possibility that some fraction of humans may have a CW response under this or other 570!
experimental paradigms, just as some humans are left- instead of right-handed. We also cannot 571!
rule out the existence of a separate neural response to CW rotations that is not reflected in the 572!
alpha-ERD signature that we assay here. 573!
The functional significance of the divergent responses to CW and CCW is also unclear. It 574!
may simply arise as a byproduct during the evolution and development of other mirror 575!
asymmetries (such as north-up vs. north-down), which serve a clearer functional, ecologically 576!
relevant purpose with a lower biological cost. It may also be that the alpha-ERD response 577!
reflects non-directional information, such as a warning of geomagnetic anomalies, which can 578!
expose a navigating animal to sudden shifts of the magnetic field comparable to those used in our 579!
experiments. For example, volcanic or igneous terranes are prone to remagnetization by 580!
lightning strikes, which produce magnetic fields powerful enough to leave local, 1-10 m scale 581!
remnant (permanent) magnetizations strong enough to warp the otherwise uniform local 582!
geomagnetic field. A large-scale example is the Vredefort Dome area in South Africa 583!
(Carporzen, Weiss, Gilder, Pommier, & Hart, 2012) where lightning remagnetization has been 584!
studied extensively. Such anomalies are common in areas with volcanic or igneous basement 585!
rock and can be located by simply wandering around with a hand compass held level at waist 586!
height and observing abnormal swings of the compass needle from magnetic north. An animal 587!
moving through isolated features of this sort would experience paired shifts; the magnetic field 588!
direction and intensity would change as the anomaly is entered and then return to normal upon 589!
exiting. If the magnetosensory system evolved in the brain as a warning signal against using the 590!
magnetic field for long-range navigation while passing through local field anomalies, sensitivity 591!
to only one directional excursion is needed. Future experiments could test this speculation by 592!
sweeping field intensity through values matching those of lightning-strike and other anomalies to 593!
check for asymmetric patterns of alpha desynchronization. 594!
A further question is whether the response asymmetry occurs only in passive experiments 595!
when participants experience magnetic stimulation without making use of the information or also 596!
in active experiments with a behavioral task, such as judging the direction or rotation of the 597!
magnetic field. Behavioral tasks with EEG recording could be used to explore the 598!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
21!
magnetosensory system in more detail and may uncover the unknown function of the observed 599!
response and its asymmetry. 600!
!601!
General Discussion 602!
As noted above, many past attempts have been made to test for the presence of human 603!
magnetoreception using behavioral assays, but the results were inconclusive. To avoid the 604!
cognitive and behavioral artifacts inherent in testing weak or subliminal sensory responses, we 605!
decided to use EEG techniques to see directly whether or not the human brain has passive 606!
responses to magnetic field changes. Our results indicate that human brains are indeed collecting 607!
and selectively processing directional input from magnetic field receptors. These give rise to a 608!
brain response that is selective for field direction and rotation with a pattern of neural activity 609!
that is measurable at the group level and repeatable in strongly-responding individuals. Such 610!
neural activity is a necessary prerequisite for any subsequent behavioral expression of 611!
magnetoreception, but such magnetically-triggered neural activity does not demand that the 612!
magnetic sense be expressed behaviorally or enter an individual’s conscious awareness. 613!
The fact that alpha-ERD is elicited in a specific and sharply delineated pattern allows us 614!
to make inferences regarding the biophysical mechanisms of signal transduction. Notably, the 615!
alpha-ERD response differentiated clearly between sets of stimuli differing only by their static or 616!
polar components. Electrical induction, electrical artifacts and quantum compass mechanisms 617!
are totally insensitive to these components and cannot account for the selectivity of brain 618!
responses. Indeed, while birds have evolved a method of navigation that would allow them to 619!
navigate by combining a non-polar magnetic sense with gravity, that strategy would not be able 620!
to distinguish our test stimuli (see section 8 of the Extended Materials and Methods). In 621!
contrast, ferromagnetic mechanisms can be highly sensitive to both static and polar field 622!
components, and could distinguish our test stimuli with differing responses. Finally, magnetite-623!
based mechanisms for navigation have been characterized in animals through neurophysiological 624!
(Walker et al., 1997), histological (Diebel, Proksch, Green, Neilson, & Walker, 2000) and pulse-625!
remagnetization studies (Beason, Wiltschko, & Wiltschko, 1997; Ernst & Lohmann, 2016; R. A. 626!
Holland, 2010; R.A. Holland & Helm, 2013; R. A. Holland, Kirschvink, Doak, & Wikelski, 627!
2008; Irwin & Lohmann, 2005; J.L. Kirschvink & Kobayashi-Kirschvink, 1991; Munro, Munro, 628!
Phillips, Wiltschko, & Wiltschko, 1997; Munro, Munro, Phillips, & Wiltschko, 1997; W. 629!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
22!
Wiltschko, Ford, Munro, Winklhofer, & Wiltschko, 2007; W. Wiltschko, Munro, Beason, Ford, 630!
& Wiltschko, 1994; W. Wiltschko, Munro, Ford, & Wiltschko, 1998, 2009; W. Wiltschko, 631!
Munro, Wiltschko, & Kirschvink, 2002; W. Wiltschko & R. Wiltschko, 1995), and biogenic 632!
magnetite has been found in human tissues (Dunn et al., 1995; Gilder et al., 2018; J. L. 633!
Kirschvink, Kobayashi-Kirschvink, & Woodford, 1992; Kobayashi & Kirschvink, 1995; Maher 634!
et al., 2016; Schultheiss-Grassi, Wessiken, & Dobson, 1999). 635!
These data argue strongly for the presence of geomagnetic transduction in humans, 636!
similar to those in numerous migratory and homing animals. Single-domain ferromagnetic 637!
particles such as magnetite are directly responsive to both time-varying and static magnetic fields 638!
and are sensitive to field polarity. At the cellular level, the magnetomechanical interaction 639!
between ferromagnetic particles and the geomagnetic field is well above thermal noise (J.L. 640!
Kirschvink & Gould, 1981; J. L. Kirschvink, Winklhofer, & Walker, 2010), stronger by several 641!
orders of magnitude in some cases (Eder et al., 2012). In many animals, magnetite-based 642!
transduction mechanisms have been found and shown to be necessary for navigational behaviors, 643!
through neurophysiological and histological studies (Diebel et al., 2000; Walker et al., 1997). A 644!
natural extension of this study would be to apply the pulse-remagnetization methods used in 645!
animals to directly test for a ferromagnetic transduction element in humans. In these 646!
experiments, a brief magnetic pulse causes the magnetic polarity of the single-domain magnetite 647!
crystals to flip. Following this treatment, the physiological and behavioral responses to the 648!
geomagnetic field are expected to switch polarity. These experiments could provide 649!
measurements of the microscopic coercivity of the magnetite crystals involved and hence make 650!
predictions about the physical size and shape of the crystals involved (Diaz Ricci & Kirschvink, 651!
1992 203), and perhaps their physiological location. 652!
Previous attempts to detect human magnetoreception may have been confounded by a 653!
number of factors. Response specificity and neural tuning to the local environment (Block, 654!
1992) make it likely that tests using stimuli outside the environmental range would likely fail, 655!
and past computational methods were not as good at isolating the neural activity studied here 656!
(Boorman et al., 1999; Sastre et al., 2002), further discussed in section 9 of the Extended 657!
Materials and Methods, below. Other experiments were conducted in unshielded conditions and 658!
may have been subject to radio-frequency noise which has been shown to shut down 659!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
23!
magnetoreceptivity in birds and other animals (Engels et al., 2014; Landler, Painter, Youmans, 660!
Hopkins, & Phillips, 2015; Tomanova & Vacha, 2016; R. Wiltschko et al., 2015). 661!
At this point, our observed reduction in alpha-band power is a clear neural signature for 662!
cortical processing of the geomagnetic stimulus, but its functional significance is unknown. In 663!
form, the activity is an alpha-ERD response resembling those found in other EEG investigations 664!
of sensory and cognitive processing. However, the alpha-ERD responses found in literature take 665!
on a range of different spatiotemporal forms and are associated with a variety of functions. It is 666!
likely that the alpha-ERD seen here reflects the sudden recruitment of neural processing 667!
resources, as this is a finding common across studies. But more research will be needed to see if 668!
and how it relates more specifically to previously studied processes such as memory access or 669!
recruitment of attentional resources. 670!
Further, alpha-ERD probably represents only the most obvious signature of neural 671!
processing arising from geomagnetic input. A host of upstream and downstream processes need 672!
to be investigated to reveal the network of responses and the information they encode. 673!
Responses independent from the alpha-ERD signature will likely emerge, and those might show 674!
different selectivity patterns and reflect stimulus features not revealed in this study. Does human 675!
magnetoreceptive processing reflect a full representation of navigational space? Does it contain 676!
certain warning signals regarding magnetic abnormalities? Or have some aspects degenerated 677!
from the ancestral system? For now, alpha-ERD remains a blank signature for a wider, 678!
unexplored range of magnetoreceptive processing. 679!
Future experiments should examine how magnetoreceptive processing interacts with 680!
other sensory modalities in order to determine field orientation. Our experimental results suggest 681!
the combination of a magnetic and a positional cue (e.g. reacting differently to North-up and 682!
North-down fields). However, we cannot tell if this positional cue uses a reference frame set by 683!
gravity sensation or is aligned with respect to the human body. In birds, orientation behavior 684!
reflects a magnetic inclination compass that identifies the steepest angle of magnetic field dip 685!
with respect to gravity (R. Wiltschko & W. Wiltschko, 1995; W. Wiltschko, 1972), and this 686!
compass can operate at dips as shallow as 5˚ from horizontal (Schwarze et al., 2016). Because 687!
magnetism and gravity are distinct, non-interacting forces of nature, the observed behavior must 688!
arise from processing of neural information from separate sensory systems (J. L. Kirschvink et 689!
al., 2010). Evolution has driven many of the known sensory systems down to their physical 690!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
24!
detection limits with astounding specificity (Block, 1992). Gravitational information is known to 691!
arise in the utricle and saccule of the vertebrate vestibular system due to the motions of dense 692!
biominerals activating hair cells (Lopez & Blanke, 2011), and a magnetite-based 693!
magnetosensory organ has been localized at the cellular level in fish (Diebel et al., 2000; Eder et 694!
al., 2012; Walker et al., 1997). The neural processing of magnetic with gravitational sensory 695!
cues could perhaps be addressed by modifying the test chamber to allow the participant to rest in 696!
different orientations with respect to gravity or by running the experiment in the zero-gravity 697!
environment of the international space station. 698!
In the participant pool, we found several highly responsive individuals whose alpha-ERD 699!
proved to be stable across time: 4 participants responded strongly at the p<0.01 level in repeated 700!
testing over weeks or months. Repeatability in individual participants suggests that the alpha-701!
ERD did not arise due to chance fluctuations in a single run, but instead reflects a consistent 702!
individual characteristic, measurable across multiple runs. A wider survey of individuals could 703!
reveal genetic/developmental or other systematic differences underlying these individual 704!
differences. 705!
The range of individual responses may be partially attributed to variation in basic alpha-706!
ERD mechanisms, rather than to underlying magnetoreceptive processing. However, some 707!
participants with high resting alpha power showed very little alpha-ERD to the magnetic field 708!
rotations, suggesting that the extent of magnetoreceptive processing itself varies across 709!
individuals. If so, distinct human populations may be good targets for future investigation. For 710!
example, studies of comparative linguistics have identified a surprising number of human 711!
languages that rely on a cardinal system of environmental reference cues (e.g. North, South, 712!
East, West) and lack egocentric terms like front, back, left, and right (Haviland, 1998; Levinson, 713!
2003; Meakins, 2011; Meakins & Algy, 2016; Meakins, Jones, & Algy, 2016). Native speakers 714!
of such languages would (e.g.) refer to a nearby tree as being to their North rather than being in 715!
front of them; they would refer to their own body parts in the same way. Individuals who have 716!
been raised from an early age within a linguistic, social and spatial framework using cardinal 717!
reference cues might have made associative links with geomagnetic sensory cues to aid in daily 718!
life; indeed, linguists have suggested a human magnetic compass might be involved (Levinson, 719!
2003). It would be interesting to test such individuals using our newly-developed methods to see 720!
if such geomagnetic cues might already be within their conscious awareness, aiding their use of 721!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
25!
the cardinal reference system. In turn, such experiments might guide the development of training 722!
procedures to enhance geomagnetic sensitivity in individuals raised in other language and 723!
cultural groups, advancing more rapidly studies on the nature of a human magnetic sense. 724!
In the 198 years since Danish physicist Hans Christian Ørsted discovered 725!
electromagnetism (March 1820), human technology has made ever-increasing use of it. Most 726!
humans no longer need to rely on an internal navigational sense for survival. To the extent that 727!
we employ a sense of absolute heading in our daily lives, external cues such as landmarks and 728!
street grids can provide guidance. Even if an individual possesses an implicit magnetoreceptive 729!
response, it is likely to be confounded by disuse and interference from our modern environment. 730!
A particularly pointed example is the use of strong permanent magnets in both consumer and 731!
aviation headsets, most of which produce static fields through the head several times stronger 732!
than the ambient geomagnetic field. If there is a functional significance to the magnetoreceptive 733!
response, it would have the most influence in situations where other cues are impoverished, such 734!
as marine and aerial navigation, where spatial disorientation is a surprisingly persistent event 735!
(Poisson & Miller, 2014). The current alpha-ERD evidence provides a starting point to explore 736!
functional aspects of magnetoreception, by employing various behavioral tasks in variety of 737!
sensory settings. 738!
739!
Conclusion 740!
741!
We conclude that at least some modern humans transduce changes in Earth-strength 742!
magnetic fields into an active neural response. Given the known presence of highly-evolved 743!
geomagnetic navigation systems in species across the animal kingdom, it is perhaps not 744!
surprising that we might retain at least some functioning neural components, especially given the 745!
nomadic hunter/gatherer lifestyle of our not-too-distant ancestors. The full extent of this 746!
inheritance remains to be discovered. 747!
748!
749!
750!
751!
752!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
26!
Extended Materials and Methods. 753!
Detailed additional instructions concerning the custom-built equipment and instrumentation are 754!
provided below. All experiments were performed in accordance with relevant guidelines and 755!
regulations following NIH protocols for human experimentation, as reviewed and approved 756!
periodically by the Caltech Administrative Committee for the Protection of Human Subjects 757!
(Caltech IRB, protocols 13-0420, 17-0706, and 17-0734). All methods were carried out in 758!
accordance with relevant guidelines and regulations. Informed consent using forms approved by 759!
the Institutional Review Board was obtained from all subjects. No subjects under the age of 18 760!
were used in these experiments. 761!
762!
1. Magnetic Exposure Facility. We constructed a six-sided Faraday cage shown in Figs. 1 and 763!
5 out of aluminum, chosen because of: (1) its high electrical conductivity, (2) low cost and (3) 764!
lack of ferromagnetism. The basic structure of the cage is a rectangular 2.44 m x 2.44 m x 2.03 765!
m frame made of aluminum rods, 1.3 cm by 1.3 cm square in cross-section, shown in Fig. 5A. 766!
Each of the cage surfaces (walls, floor and ceiling) have four rods (two vertical and two 767!
horizontal) bounding the perimeter of each sheet. On the cage walls three vertical rods are 768!
spaced equally along the inside back of each surface, and on the floor and ceiling three 769!
horizontal rods are similarly spaced, forming an inwards-facing support frame. This frame 770!
provides a conductive chassis on which overlapping, 1 mm thick aluminum sheets (2.44 m long 771!
and 0.91 m wide) were attached using self-threading aluminum screws at ~0.60 m intervals with 772!
large overlaps between each sheet. In addition, we sealed the seams between separate aluminum 773!
panels with conductive aluminum tape. The access door for the cage is a sheet of aluminum that 774!
is fastened with a 2.4 m long aluminum hinge on the East-facing wall such that it can swing 775!
away from the cage and provide an entrance/exit. Aluminum wool has been affixed around the 776!
perimeter of this entrance flap to provide a conductive seal when the flap is lowered (e.g. the 777!
cage is closed). Ventilation is provided via a ~3 m long, 15 cm diameter flexible aluminum tube 778!
(Fig. 5E) that enters an upper corner of the room and is connected to a variable-speed ceiling-779!
mounted fan set for a comfortable but quiet level of airflow. The end of the tube in contact with 780!
the Faraday cage is packed loosely with aluminum wool that allows air to pass and provides 781!
electrical screening. LED light strips (Fig. 5H) provide illumination for entrance and exit. These 782!
lights are powered by a contained lithium ion battery housed in an aluminum container attached 783!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
27!
at the top end of the Faraday cage, adjacent to the entrance of the ventilation air duct (seen as the 784!
red battery in Fig. 5E). 785!
In all experiment sessions, power to the lights was switched off. A small USB-powered 786!
infrared camera and microphone assembly (Fig. 5G) mounted just inside the cage on the North 787!
wall allows audiovisual monitoring of participants inside the room. Instructions to the 788!
participants are given from a pair of speakers mounted outside the Faraday cage (Fig. 5F), 789!
controlled remotely by experimenters and electrically shorted by a computer-controlled TTL 790!
relay when not in use. Acoustic foam panels are attached to the vertical walls to dampen echoes 791!
within the chamber as well as to reduce the amplitude of external sound entering the chamber. 792!
To complete the Faraday shielding, we grounded the cage permanently at one corner with a 2.6 793!
mm diameter (10 AWG) copper wire connected to the copper plumbing in the sub-basement of 794!
the building. RMS noise measurements from the cage interior using a Schwarzbeck Mess™ 795!
Elektronik FMZB 1513 B-component active loop Rf antenna, a RIGOL™ DSA815/E-TG 796!
spectrum analyzer, and a Tektronix™ RSA503A signal analyzer indicated residual noise 797!
interference below 0.01 nT, in the frequency range from 9 kHz to 10 MHz. 798!
Electrical cables entering the Faraday cage pass through a side gap in the aluminum 799!
ventilation duct and then through the aluminum wool. Rf interference is blocked further on all 800!
electrical cables entering the room using pairs of clip-on ferrite chokes (Fair-Rite™ material #75, 801!
composed of MnZn ferrite, designed for low-frequency EMI suppression, referred from here-on 802!
as ferrite chokes) and configured where possible using the paired, multiple-loop “pretty-good 803!
choke” configuration described by Counselman (Counselman, 2013) (Fig. 5I). Inside the 804!
shielded space are located a three-axis set of square coils approximately 2 m on edge following 805!
the Merritt et al. four-coil design (Merritt et al., 1983) (using the 59/25/25/59 coil winding ratio) 806!
that provides remarkably good spatial uniformity in the applied magnetic field (12 coils total, 807!
four each in the North/South, East/West, and Up/Down orientations as seen in Fig. 5A). The 808!
coils are double-wrapped inside grounded aluminum U-channels following a design protocol that 809!
allows for full active-field and sham exposures (J.L Kirschvink, 1992); they were constructed by 810!
Magnetic Measurements, Ltd., of Lancashire, U.K. (http://www.magnetic-measurements.com). 811!
This double-wrapped design gives a total coil winding count of 118/50/50/118 for all three-axes 812!
coil sets. 813!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
28!
To provide a working floor isolated from direct contact with the coils, we suspended a 814!
layer of ~2 cm thick plywood sheets on a grid work of ~ 10 x 10 cm thick wooden beams that 815!
rested on the basal aluminum plate of the Faraday shield that are held together with brass screws. 816!
We covered this with a layer of polyester carpeting on top of which we placed a wooden 817!
platform chair for the participants (Fig. 5B). Non-magnetic bolts and screws were used to fasten 818!
the chair together, and a padded foam cushion was added for comfort. The chair is situated such 819!
that the head and upper torso of most participants fit well within the ~1 m3 volume of highly 820!
uniform magnetic fields produced by the coil system (J.L Kirschvink, 1992) while keeping the 821!
participants a comfortable distance away from direct contact with the Merritt coils. 822!
We suspended the three-axis probe of a fluxgate magnetometer (Applied Physics 823!
Systems™ model 520A) on a non-magnetic, carbon-fiber, telescoping camera rod suspended 824!
from the ceiling of the Faraday cage (Fig. 5D). This was lowered into the center of the coil 825!
system for initial calibration of field settings prior to experiments and then raised to the edge of 826!
the uniform field region to provide continuous recording of the magnetic field during 827!
experiments. Power cables for the coils and a data cable for the fluxgate sensor pass out of the 828!
Faraday cage through the ventilation shaft, through a series of large Rf chokes (Counselman, 829!
2013), a ceiling utility chase in the adjacent hallway, along the wall of the control room, and 830!
finally down to the control hardware. The control hardware and computer are located ~20 m 831!
away from the Faraday cage through two heavy wooden doors and across a hallway that serve as 832!
effective sound dampeners such that participants are unable to directly hear the experimenters or 833!
control equipment and the experimenters are unable to directly hear the participant. 834!
In the remote-control room, three bipolar power amplifiers (Kepco™ model BOP-100-835!
1MD) control the electric power to the coil systems (Fig. 5J) and operate in a mode where the 836!
output current is regulated proportional to the control voltage, thereby avoiding a drop in current 837!
(and magnetic field) should the coil resistance increase due to heating. Voltage levels for these 838!
are generated using a 10k samples per channel per second, 16-bit resolution, USB-controlled, 839!
analog output DAQ device (Measurement Computing™ Model USB-3101FS), controlled by the 840!
desktop PC. This same PC controls the DC power supply output levels, monitors and records the 841!
Cartesian orthogonal components from the fluxgate magnetometer, displays video of the 842!
participant (recordings of which are not preserved per IRB requirements), and is activated or 843!
shorted, via TTL lines, to the microphone/speaker communication system from the control room 844!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
29!
to the experimental chamber. As the experimenters cannot directly hear the participant and the 845!
participant cannot directly hear the experimenters, the microphone and speaker system are 846!
required (as per Caltech Institute Review Board guidelines) to ensure the safety and comfort of 847!
the participant as well as to pass instructions to the participant and answer participants’ questions 848!
before the start of a block of experiments. The three-axis magnet coil system can produce a 849!
magnetic vector of up to 100 µT intensity (roughly 2-3X the background strength in the lab) in 850!
any desired direction with a characteristic RL relaxation constant of 79-84 ms (inductance and 851!
resistance of the four coils in each axis vary slightly depending on the different coil-diameters 852!
for each of the three nested, double-wrapped coil-set axes). Active/Sham mode was selected 853!
prior to each run via a set of double-pole-double-throw (DPDT) switches located near the DC 854!
power supplies. These DPDT switches are configured to swap the current direction flowing in 855!
one strand of the bifilar wire with respect to the other strand in each of the coil sets (J.L 856!
Kirschvink, 1992) (Fig. 5C). Fluxgate magnetometer analog voltage levels were digitized and 857!
streamed to file via either a Measurement Computing™ USB 1608GX 8-channel (differential 858!
mode) analog input DAQ device, or a Measurement Computing™ USB 1616HS-2 multifunction 859!
A/D, D/A, DIO DAQ device connected to the controller desktop PC. Fluxgate analog voltage 860!
signal levels were sampled at 1024 or 512 Hz. Although the experimenter monitors the 861!
audio/video webcam stream of the participants continuously, as per Caltech IRB safety 862!
requirements, while they are in the shielded room the control software disconnects the external 863!
speakers (in the room that houses the experimental Faraday cage and coils) and shorts them to 864!
electrical ground during all runs to prevent extraneous auditory cues from reaching the 865!
participants. Light levels within the experimental chamber during experimental runs were 866!
measured using a Konica-Minolta CS-100A luminance meter, which gave readings of zero 867!
(below 0.01 ± 2% cd/m2). 868!
869!
870!
871!
872!
873!
874!
875!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
30!
876!
Fig. 5. Additional images of critical aspects of the human magnetic exposure facility at 877!
Caltech. A. Partially complete assembly of the Faraday cage (summer of 2014) showing the 878!
nested set of orthogonal, Merritt square four-coils (Merritt et al., 1983) with all but two 879!
aluminum walls of the Faraday cage complete. B. Image of a participant in the facility seated in 880!
a comfortable, non-magnetic wooden chair and wearing the 64-lead BioSimTM EEG head cap. 881!
The EEG sensor leads are carefully braided together to minimize electrical artifacts. The chair is 882!
on a raised wooden platform that is isolated mechanically from the magnet coils and covered 883!
with a layer of synthetic carpeting; the height is such that the participant’s head is in the central 884!
area of highest magnetic field uniformity. C. Schematic of the double-wrapped control circuits 885!
that allow active-sham experiments (J.L Kirschvink, 1992). In each axis of the coils, the four 886!
square frames are wrapped in series with two discrete strands of insulated copper magnet wire 887!
and with the number of turns and coil spacing chosen to produce a high-volume, uniform applied 888!
magnetic field (Merritt et al., 1983). Reversing the current flow in one of the wire strands via a 889!
double-pole-double-throw (DPDT) switch results in cancellation of the external field with 890!
virtually all other parameters being the same. This scheme is implemented on all three 891!
independently controlled coil axes (Up/Down, East/West and North/South). D. Fluxgate 892!
magnetometer (Applied Physics Systems 520A) three-axis magnetic field sensor attached to a 893!
collapsing carbon-fiber camera stand mount. At the start of each session the fluxgate is lowered 894!
to the center of the chamber for an initial current / control calibration of the ambient geomagnetic 895!
field. It is then raised to a position about 30 cm above the participant’s head during the 896!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
31!
following experimental trials, and the three-axis magnetic field readings are recorded 897!
continuously in the same fashion as the EEG voltage signals. E. Air duct. A 15 cm diameter 898!
aluminum air duct ~2 meters long connects a variable-speed (100 W) electric fan to the upper SE 899!
corner of the experimental chamber; this is also the conduit used for the major electrical cables 900!
(power for the magnetic coils, sensor leads for the fluxgate, etc.). F. & G. An intercom / video 901!
monitoring system was devised by mounting a computer-controlled loudspeaker (F) outside the 902!
Faraday shield on the ceiling North of the chamber coupled with (G) a USB-linked IR video 903!
camera / microphone system mounted just inside the shield. Note the conductive aluminum tape 904!
shielding around the camera to reduce Rf interference. During all experimental trials a small 905!
DPDT relay located in the control room disconnects the speaker from computer and directly 906!
shorts the speaker connections. A second microphone in the control room can be switched on to 907!
communicate with the participant in the experimental chamber, as needed. An experimenter 908!
monitors the audio and video of participants at all times, as per Caltech IRB safety requirements. 909!
H. LED lights, 12 VDC array, arranged to illuminate from the top surface of the magnetic coils 910!
near the ceiling of the chamber. These are powered by rechargeable 11.1 V lithium battery packs 911!
(visible in E) and controlled by an external switch. I. Ferrite chokes. Whenever possible, these 912!
are mounted in a multiple-turn figure-eight fashion (Counselman, 2013) on all conductive wires 913!
and cables entering the shielded area and supplemented with grounded aluminum wool when 914!
needed. J. Image of the remote control area including (from left to right): the PC for controlling 915!
the coils, the DPDT switches for changing between active and sham modes, the fluxgate control 916!
unit, the three power amplifiers that control the current in the remote coil room, and the separate 917!
PC that records the EEG data. Participants seated in the experimental chamber do not report 918!
being able to hear sounds from the control room and vice versa. 919!
920!
921!
922!
923!
924!
925!
926!
927!
928!
929!
930!
931!
932!
933!
934!
935!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
32!
2. Participants. Participants were 34 adult volunteers (24 male, 12 female) recruited from the 936!
Caltech population. This participant pool included persons of European, Asian, African and 937!
Native American descent. Ages ranged from 18 to 68 years. Each participant gave written 938!
informed consent of study procedures approved by the Caltech Institutional Review Board 939!
(Protocols 13-0420, 17-0706, and 17-0734). 940!
941!
3. Experimental Protocol. In the experiment, participants sat upright in the chair with their 942!
eyes closed and faced North (defined as 0° declination in our magnetic field coordinate reference 943!
frame). The experimental chamber was dark, quiet and isolated from the control room during 944!
runs. Each run was ~7 minutes long with up to eight runs in a ~1 hour session. The magnetic 945!
field was rotated over 100 milliseconds every 2-3 seconds, with constant 2 or 3 s inter-trial 946!
intervals in early experiments and pseudo-randomly varying 2-3 s intervals in later experiments. 947!
Participants were blind to Active vs. Sham mode, trial sequence and trial timing. During 948!
sessions, auditory tones signaled the beginning and end of experiments and experimenters only 949!
communicated with participants once or twice per session to update the number of runs 950!
remaining. When time allowed, Sham runs were matched to Active runs using the same 951!
software settings. Sham runs are identical to Active runs but are executed with the current 952!
direction switches set to anti-parallel. This resulted in no observable magnetic field changes 953!
throughout the duration of a Sham run with the local, uniform, static field produced by the 954!
double-wrapped coil system in cancellation mode (J.L Kirschvink, 1992). 955!
Two types of trial sequences were used: (1) a 127-trial Gold Sequence with 63 FIXED 956!
trials and 64 SWEEP trials evenly split between two rotations (32 each), and (2) various 150-trial 957!
pseudorandom sequences with 50 trials of each rotation interspersed with 50 FIXED trials to 958!
balance the number of trials in each of three conditions. All magnetic field parameters were held 959!
constant during FIXED trials, while magnetic field intensity was held constant during inclination 960!
or declination rotations. In inclination experiments (Fig. 2A of the main text), the vertical 961!
component of the magnetic field was rotated upwards and downwards between ±55°, ±60°, or 962!
±75° (Inc.UP and Inc.DN, respectively); data collected from runs with each of these inclination 963!
values were collapsed into a single set representative of inclination rotations between steep 964!
angles. In each case, the horizontal component was steady at 0° declination (North; Inc.UP.N 965!
and Inc.DN.N). Two types of declination experiments were conducted, designed to test the 966!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
33!
quantum compass and electrical induction hypotheses. As the quantum compass can only 967!
determine the axis of the field and not polarity, we compared a pair of declination experiments in 968!
which the rotating vectors were swept down to the North (DecDn.N) and up to the South 969!
(DecUp.S), providing two symmetrical antiparallel data sets (Fig. 2B of the main text). In the 970!
DecDn.N experiments, the vertical component was held constant and downwards at +60° or 971!
+75°, while the horizontal component was rotated between NE (45°) and NW (315°), along a 972!
Northerly arc (DecDn.CW.N and DecDn.CCW.N). In DecUp.S experiments, the vertical 973!
component was held upwards at −60° or −75°, while the horizontal component was rotated 974!
between SW (225°) and SE (135°) along a Southerly arc (DecUp.CW.S and DecUp.CCW.S). 975!
Again, runs with differing inclination values were grouped together as datasets with steep 976!
downwards or steep upwards inclination. To test the induction hypothesis, we paired the 977!
DecDn.N sweeps with a similar set, DecUp.N, as shown on Fig. 2C. These two conditions only 978!
differ in the direction of the vertical field component; rotations were between NE and NW in 979!
both experiments (DecDn.CW.N, DecDn.CCW.N, DecUp.CW.N and DecUp.CCW.N). Hence, 980!
any significant difference in the magnetosensory response eliminates induction as a mechanism. 981!
982!
4. EEG Recording. EEG was recorded using a BioSemi™ ActiveTwo system with 64 983!
electrodes following the International 10-20 System (Nuwer et al., 1998). Signals were sampled 984!
at 512 Hz with respect to CMS/DRL reference at low impedance <1 ohm and bandpass-filtered 985!
from 0.16-100 Hz. To reduce electrical artifacts induced by the time-varying magnetic field, 986!
EEG cables were bundled and twisted 5 times before plugging into a battery-powered BioSemi™ 987!
analog/digital conversion box. Digitized signals were transmitted over a 30 m, non-conductive, 988!
optical fiber cable to a BioSemi™ USB2 box located in the control room ~20 m away where a 989!
desktop PC (separate from the experiment control system) acquired continuous EEG data using 990!
commercial ActiViewTM software. EEG triggers signaling the onset of magnetic stimulation 991!
were inserted by the experiment control system by connecting a voltage timing signal (0 to 5 V) 992!
from its USB-3101FS analog output DAQ device. The timing signal was sent both to the 993!
Measurement Computing USB-1608GX (or USB-1616HS-2) analog input DAQ device, used to 994!
sample the magnetic field on the experiment control PC and a spare DIO voltage input channel 995!
on the EEG system’s USB2 DAQ input box, which synchronized the EEG data from the optical 996!
cable with the triggers cued by the controlling desktop PC. This provided: (1) a precise 997!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
34!
timestamp in continuous EEG whenever electric currents were altered (or in the case of FIXED 998!
trials, when the electric currents could have been altered to sweep the magnetic field direction, 999!
but were instead held constant) in the experimental chamber, and (2) a precise correlation (±2 1000!
ms, precision determined by the 512 samples per second digital input rate of the BioSemiTM 1001!
USB2 box) between fluxgate and EEG data. 1002!
1003!
5. EEG Analysis. Raw EEG data were extracted using EEGLAB™ toolbox for MATLAB™ 1004!
and analyzed using custom MATLAB™ scripts. Trials were defined as 2- or 3-s epochs from 1005!
−0.75 s pre-stimulus to +1.25 or +2.25 s post-stimulus, with a baseline interval from −0.5 s to 1006!
−0.25 s pre-stimulus. Time/frequency decomposition was performed for each trial using Fast 1007!
Fourier Transform (MATLAB™ function fft) and Morlet wavelet convolution on 100 linearly-1008!
spaced frequencies between 1 and 100 Hz. Average power in an extended alpha band of 6-14 Hz 1009!
was computed for the pre-stimulus and post-stimulus intervals of all trials, and a threshold of 1010!
1.5X the interquartile range was applied to identify trials with extreme values of log alpha 1011!
power. These trials were excluded from further analysis but retained in the data. After 1012!
automated trial rejection, event-related potentials (ERPs) were computed for each condition and 1013!
then subtracted from each trial of that condition to reduce the electrical induction artifact that 1014!
appeared only during the 100 ms magnetic stimulation interval. This is an established procedure 1015!
to remove phase-locked components such as sensory-evoked potentials from an EEG signal for 1016!
subsequent analysis of non-phase-locked, time/frequency power representations. Non-phase-1017!
locked power was computed at midline frontal electrode Fz for each trial and then averaged and 1018!
baseline-normalized for each condition to generate a time/frequency map from −0.25 s pre-1019!
stimulus to +1 s or +2 s post-stimulus and 1-100 Hz. To provide an estimate of overall alpha 1020!
power for each participant, power spectral density was computed using Welch’s method 1021!
(MATLAB™ function pwelch) at 0.5 Hz frequency resolution (Welch, 1967). 1022!
From individual datasets, we extracted post-stimulus alpha power to test for statistically 1023!
significant differences amongst conditions at the group level. Because alpha oscillations vary 1024!
substantially across individuals in amplitude, frequency and stimulus-induced changes, an 1025!
invariant time/frequency window would not capture stimulus-induced power changes in many 1026!
participants. In our dataset, individual alpha oscillations ranged in frequency (8 to 12 Hz peak 1027!
frequency), and individual alpha-ERD responses started around +0.25 to +0.75 s post-stimulus. 1028!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
35!
Thus, we quantified post-stimulus alpha power within an automatically-adjusted time/frequency 1029!
window for each dataset. First, non-phase-locked alpha power between 6-14 Hz was averaged 1030!
over all trials regardless of condition. Then, the most negative time/frequency point was 1031!
automatically selected from the post-stimulus interval between 0 s and +1 or +2 s in this cross-1032!
conditional average. The selected point represented the maximum alpha-ERD in the average over 1033!
all trials with no bias for any condition. A time/frequency window of 0.25 s and 5 Hz was 1034!
centered (as nearly as possible given the limits of the search range) over this point to define an 1035!
individualized timing and frequency of alpha-ERD for each dataset. Within the window, non-1036!
phase-locked alpha power was averaged across trials and baseline-normalized for each condition, 1037!
generating a value of alpha-ERD for each condition to be compared in statistical testing. 1038!
In early experiments, trial sequences were balanced with nearly equal numbers of FIXED 1039!
(63) and SWEEP (64) trials, with an equal number of trials for each rotation (e.g. 32 Inc.DN and 1040!
32 Inc.UP trials). Later, trial sequences were designed to balance the number of FIXED trials 1041!
with the number of trials of each rotation (e.g. 50 DecDn.FIXED, 50 DecDn.CCW, and 50 1042!
DecDn.CW trials). Alpha-ERD was computed over similar numbers of trials for each condition. 1043!
For example, when comparing alpha-ERD in the FIXED vs. CCW vs. CW conditions of a 1044!
declination experiment with 63 FIXED (32 CCW and 32 CW trials) 100 samplings of 32 trials 1045!
were drawn from the pool of FIXED trials, alpha-ERD was averaged over the subset of trials in 1046!
each sampling, and the average over all samplings was taken as the alpha-ERD of the FIXED 1047!
condition. When comparing FIXED vs. SWEEP conditions of an inclination experiment with 50 1048!
FIXED, 50 DN, and 50 UP trials, 200 samplings of 25 trials were drawn from each of the DN 1049!
and UP conditions and the average alpha-ERD over all samplings taken as the alpha-ERD of the 1050!
SWEEP condition. Using this method, differences in experimental design were reduced, 1051!
allowing statistical comparison of similar numbers of trials in each condition. The alpha-ERD 1052!
values for each participant in each condition are shown as histograms for the DecDn (Fig. 6), 1053!
DecUp (Fig. 7), and Sham declination (Fig. 8) experiments. These values were used in statistical 1054!
testing at the group level. 1055!
1056!
1057!
1058!
1059!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
36!
1060!
Fig. 6. Histogram of alpha-ERD responses over all participants (N=26) in the DecDn 1061!
experiment. The panels show the histogram of individual responses for each condition. 1062!
Frequency is given in number of participants. Because we looked for a drop in alpha power 1063!
following magnetic stimulation, the histograms are shifted towards negative values in all 1064!
conditions. The CCW condition shows the most negative average in a continuous distribution of 1065!
participant responses, with the most participants having a >2 dB response. 1066!
1067!
1068!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
37!
1069!
Fig. 7. Histogram of alpha-ERD responses over all participants (N=16) in the DecUp 1070!
experiment. The panels show the histogram of individual responses for each condition. No 1071!
significant magnetosensory response was observed in any condition, and no clear difference is 1072!
apparent between the three distributions. 1073!
1074!
1075!
1076!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
38!
1077!
Fig. 8. Histogram of alpha-ERD responses over all participants (N=18) in the Sham Declination 1078!
experiment. The panels show the histogram of individual responses for each condition. No 1079!
significant magnetosensory response was observed in any condition, and no clear difference is 1080!
apparent between the three distributions. 1081!
1082!
1083!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
39!
Three statistical tests were performed using average alpha-ERD: (1) Inc ANOVA 1084!
(N=29), (2) DecDn ANOVA (N=26), (3) DecDn/DecUp ANOVA (N=16). For the inclination 1085!
experiment, data were collected in Active and Sham modes for 29 of 34 participants. Due to 1086!
time limitations within EEG sessions, sham data could not be collected for every participant, so 1087!
those participants without inclination sham data were excluded. A two-way repeated-measures 1088!
ANOVA tested for the effects of inclination rotation (SWEEP vs. FIXED) and magnetic 1089!
stimulation (Active vs. Sham) on alpha-ERD. Post-hoc testing using the Tukey-Kramer method 1090!
compared four conditions (Active-SWEEP, Active-FIXED, Sham-SWEEP and Sham-FIXED) 1091!
for significant differences (Tukey, 1949). 1092!
For the DecDn experiment, data were collected from 26 participants in Active mode. A 1093!
one-way repeated-measures ANOVA tested for the effect of declination rotation (DecDn.CCW 1094!
vs. DecDn.CW vs. DecDn.FIXED) with post-hoc testing to compare these three conditions. For 1095!
a subset of participants (N=16 of 26), data was collected from both DecDn and DecUp 1096!
experiments. The DecUp experiments were introduced in a later group to evaluate the quantum 1097!
compass mechanism of magnetosensory transduction, as well as in a strongly-responding 1098!
individual to test the less probable induction hypothesis, as shown in Video 1. For tests of the 1099!
quantum compass hypothesis, we used the DecDn/DecUp dataset. A two-way repeated-measures 1100!
ANOVA tested for the effects of declination rotation (DecDn.CCW.N vs. DecDn.CW.N vs. 1101!
DecUp.CCW.S vs. DecUp.CW.S vs. DecDn.FIXED.N vs. DecUp.FIXED.S) and inclination 1102!
direction (Inc.DN.N vs Inc.UP.S) on alpha-ERD; data from another strongly-responding 1103!
individual is shown in Video 2. Post-hoc testing compared six conditions (DecDn.CCW.N, 1104!
DecDn.CW.N, DecDn.FIXED.N, DecUp.CCW.S, DecUp.CW.S and DecUp.FIXED.S). 1105!
Within each group, certain participants responded strongly with large alpha-ERD while 1106!
others lacked any response to the same rotations. To establish whether a response was consistent 1107!
and repeatable, we tested individual datasets for significant post-stimulus power changes in 1108!
time/frequency maps between 0 to +2 or +3 s post-stimulus and 1-100 Hz. For each dataset, 1109!
1000 permutations of condition labels over trials created a null distribution of post-stimulus 1110!
power changes at each time/frequency point. The original time/frequency maps were compared 1111!
with the null distributions to compute a p-value at each point. False discovery rate correction for 1112!
multiple comparisons was applied to highlight significant post-stimulus power changes at the 1113!
p<0.05 and p<0.01 statistical thresholds (Benjamini & Hochberg, 1995). Fig. 9 shows repeated 1114!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
40!
runs (Run #1 and Run #2) of two different participants (A and B) in the DecDn experiment. The 1115!
outlined clusters indicate significant power changes following magnetic field rotation. In each 1116!
case, the significant clusters are similar in timing and bandwidth across runs up to six months 1117!
apart. 1118!
1119!
1120!
1121!
1122!
1123!
1124!
1125!
1126!
1127!
1128!
1129!
1130!
1131!
1132!
1133!
1134!
1135!
1136!
1137!
1138!
1139!
1140!
1141!
1142!
1143!
1144!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
41!
1145!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
42!
Fig. 9. Repeated results from two strongly-responding participants. In both (A) and (B), 1146!
participants were tested weeks or months apart under the same conditions (Run #1 and Run #2). 1147!
Time/frequency maps show similar timing and bandwidth of significant alpha power changes 1148!
(blue clusters in outlines) after counterclockwise rotation, while activity outside the alpha-ERD 1149!
response, and activity in other conditions is inconsistent across runs. Black/white outlines 1150!
indicate significance at the p<0.05 and p<0.01 thresholds. The participant in (A) had an alpha 1151!
peak frequency at <9 Hz and a lower-frequency alpha-ERD response. The participant in (B) had 1152!
an alpha peak frequency >11 Hz and a higher-frequency alpha-ERD response. Minor power 1153!
fluctuations in the other conditions or in different frequency bands were not repeated across runs, 1154!
indicating that only the alpha-ERD was a repeatable signature of magnetosensory processing. 1155!
1156!
Video 1. Test of the electrical induction mechanism of magnetoreception using data from a 1157!
participant with a strong, repeatable alpha-ERD magnetosensory response. Bottom row shows 1158!
the DecDn.CCW.N, DecDn.CW.N and DecDn.FIXED.N conditions (64 trials per condition) of 1159!
the DecDn.N experiment; top row shows the corresponding conditions for the DecUp.N 1160!
experiment. Scalp topography changes from –0.25 s pre-stimulus to +1 s post-stimulus. The 1161!
CCW rotation of a downwards-directed field (DecDn.CCW.N) caused a strong, repeatable alpha-1162!
ERD (lower left panel, p<0.01 at Fz); weak alpha power fluctuations observed in other 1163!
conditions (DecDn.CW.N, DecDn.FIXED.N, DecUp.CW.N, DecUp.CCW.N and 1164!
DecUp.FIXED.N) were not consistent across multiple runs of the same experiment. If the 1165!
magnetoreception mechanism is based on electrical induction, the same response should occur in 1166!
conditions with identical ∂B/∂t (DecDn.CCW.N and DecUp.CCW.N), but the response was 1167!
observed only in one of these conditions: a result that contradicts the predictions of the electrical 1168!
induction hypothesis. 1169!
1170!
Video 2. Test of the quantum compass mechanism of magnetoreception using data from 1171!
another strongly-responding participant. Bottom vs. top rows compare the DecDn.N and 1172!
DecUp.S experiments in the CCW, CW and FIXED conditions (DecDn.CCW.N, DecDn.CW.N, 1173!
DecDn.FIXED.N, DecUp.CW.S, DecUp.CCW.S and DecUp.FIXED.S with 100 trials per 1174!
condition). The quantum compass is not sensitive to magnetic field polarity, so magnetosensory 1175!
responses should be identical for the DecDn.CCW.N and DecUp.CCW.S rotations sharing the 1176!
same axis. Our results contradict this prediction. A significant, repeatable alpha-ERD is only 1177!
observed in the DecDn.CCW.N condition (lower left panel, p<0.01 at Fz), with no strong, 1178!
consistent effects in the DecUp.CCW.S condition (top left panel) or any other condition. 1179!
1180!
1181!
1182!
1183!
Extended Discussion 1184!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
43!
1185!
6. Controlling for Magnetomechanical Artifacts. A question that arises in all studies of 1186!
human perception is whether confounding artifacts in the experimental system produced the 1187!
observed effects. The Sham experiments using double-wrapped, bonded coil systems controlled 1188!
by remote computers and power supplies indicate that obvious artifacts such as resistive 1189!
warming of the wires or magnetomechanical vibrations between adjacent wires are not 1190!
responsible. In Active mode, however, magnetic fields produced by the coils interact with each 1191!
other with maximum torques occurring when the moment u of one coil set is orthogonal to the 1192!
field B of another (torque = u x B). Hence, small torques on the coils might produce transient, 1193!
sub-aural motion cues. Participants might detect these cues subconsciously even though the coils 1194!
are anchored to the Faraday cage at many points; the chair and floor assemblies are mechanically 1195!
isolated from the coils; the experiments are run in total darkness, and the effective frequencies of 1196!
change are all below 5 Hz and acting for only 0.1 second. No experimenters or participants ever 1197!
claimed to perceive field rotations consciously even when the cage was illuminated and efforts 1198!
were made to consciously detect the field rotations. Furthermore, the symmetry of the field 1199!
rotations and the asymmetric nature of the results both argue strongly against this type of artifact. 1200!
During the declination experiments, for example, the vertical component of the magnetic field is 1201!
held constant while a constant-magnitude horizontal component is rotated 90˚ via the N/S and 1202!
E/W coil axes. Hence, the torque pattern produced by DecDn.CCW.N rotations should be 1203!
identical to that of the DecUp.CW.S rotations, yet these conditions yielded dramatically different 1204!
results. We conclude that magnetomechanical artifacts are not responsible for the observed 1205!
responses. 1206!
1207!
7. Testing for Artifacts or Perception from Electrical Induction. Another source of artifacts 1208!
might be electrical eddy currents induced during field sweeps that might stimulate subsequent 1209!
EEG brain activity in the head or perhaps in the skin or scalp adjacent to EEG sensors. Such 1210!
artifacts would be hard to distinguish from a magnetoreceptive structure based on electrical 1211!
induction. For example, the alpha-ERD effects might arise via some form of voltage-sensitive 1212!
receptor in the scalp subconsciously activating sensory neurons and transmitting information to 1213!
the brain for further processing. However, for any such electrical induction mechanism the 1214!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
44!
Maxwell-Faraday law holds that the induced electric field E is related to the magnetic field 1215!
vector, B(t), by: 1216!
1217!
∇ × E = -∂B(t)/∂t (1). 1218!
1219!
During a declination rotation, the field vector B(t) is given by: B(t) = BV + BH(t), where BV is 1220!
the constant vertical field component, t is time, BH(t) is the rotating horizontal component, and 1221!
the quantities in bold are vectors. Because the derivative of a constant is zero, the static vertical 1222!
vector BV has no effect, and the induced electrical effect depends only on the horizontally-1223!
rotating vector, BH(t): 1224!
1225!
∇ × E = -∂BV/∂t - ∂BH(t)/∂t = - ∂BH(t)/∂t (2). 1226!
1227!
As noted in the main text, Video 1 shows results for the induction test shown in Fig. 2C 1228!
for which the sweeps of the horizontal component are identical, going along a 90˚ arc between 1229!
NE and NW (DecDn.CCW.N and DecUp.CCW.N). The two trials differ only by the direction of 1230!
the static vertical vector, BV, which is held in the downwards orientation for the bottom row of 1231!
Video 1 and upwards in the top row. As only the DecDn.CCW.N sweep elicits alpha-ERD, and 1232!
the DecUp.CCW.N sweep does not elicit alpha-ERD, electrical induction cannot be the 1233!
mechanism for this effect either via some artifact of the EEG electrodes or an intrinsic 1234!
anatomical structure. 1235!
We also ran additional control experiments on “EEG phantoms,” which allow us to 1236!
isolate the contribution of environmental noise and equipment artifacts. Typical setups range 1237!
from simple resistor circuits to fresh human cadavers. We performed measurements on two 1238!
commonly-used EEG phantoms: a bucket of saline, and a cantaloupe. From these controls, we 1239!
isolated the electrical effects induced by magnetic field rotations. The induced effects were 1240!
similar to the artifact observed in human participants during the 100 ms magnetic stimulation 1241!
interval. In cantaloupe and in the water-bucket controls, no alpha-ERD responses were observed 1242!
in Active or Sham modes suggesting that a brain is required to produce a magnetosensory 1243!
response downstream of any induction artifacts in the EEG signal. 1244!
1245!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
45!
8. Non-polar magnetoreceptivity (attributed to birds) cannot explain the present data. 1246!
Birds and some other animals display a magnetic inclination compass that identifies the 1247!
steepest angle of magnetic field dip with respect to gravity (R. Wiltschko & W. Wiltschko, 1995; 1248!
W. Wiltschko, 1972), and as noted earlier this compass can operate at dips as shallow as 5˚ from 1249!
horizontal (Schwarze et al., 2016). This allows a bird to identify the direction of the closest pole 1250!
(North or South) without knowing the polarity of the magnetic field. If a bird knows it is in the 1251!
(e.g.) Northern Hemisphere, it can use this maximum dip to identify the direction of geographic 1252!
North. However, this mechanism could not distinguish between the antipodal (vector opposite) 1253!
fields used in our biophysical test of polarity sensitivity. If we create a field with magnetic north 1254!
down and to the front, the bird would correctly identify North as forward. However, if we point 1255!
magnetic north up and to the back, the bird would still identify North as forward because that is 1256!
the direction of maximum dip. 1257!
Because magnetism and gravity are distinct, non-interacting forces of nature, the 1258!
observed behavior must arise from neural processing of sensory information from separate 1259!
transduction mechanisms (J. L. Kirschvink et al., 2010). If polarity information is not present 1260!
initially from a magnetic transducer or lost in subsequent neural processing, it cannot be 1261!
recovered by adding information from other sensory modalities. As an illustration, if we gave 1262!
our participants a compass with a needle that did not have its North tip marked, they could not 1263!
distinguish the polarity of an applied magnetic field even if we gave them a gravity pendulum or 1264!
any other non-magnetic sensor. 1265!
At present our experimental results in humans suggest the combination of a magnetic and 1266!
a positional cue. However, we cannot tell if this positional cue is a reference frame using gravity 1267!
or one aligned with respect to the human body. This could perhaps be addressed by modifying 1268!
the test chamber to allow the participant to rest in different orientations with respect to gravity. 1269!
1270!
9. Sastre et al. EEG Study. Our results perhaps shed light on a previous study attempting to 1271!
detect the presence of a low-frequency magnetic stimulus on human brainwaves, which found no 1272!
significant effects. As part of a major initiative to investigate possible electromagnetic effects on 1273!
cancer by the US National Institute of Health and the Department of Energy during the 1990’s, 1274!
Sastre et al. (Sastre et al., 2002) analyzed EEG signals for power changes in several frequency 1275!
bands averaged over 4 s intervals before and after changes in the background magnetic field. 1276!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
46!
However, they did not do the time/frequency analysis that we report here nor averaging of 1277!
repeated rotations over many trials; wavelet methods were not used as frequently at that time. 1278!
To test the impact of these differences in data analysis algorithms, we analyzed our data using 1279!
the techniques in Sastre et al. These analyses did not reveal any significant differences in total or 1280!
band-specific power between any conditions. Thus, our results are consistent with previous 1281!
findings. 1282!
Other differences between our studies lie in the stimulation parameters. In four of seven 1283!
conditions from Sastre et al. (A, B, C and D), the field intensities used (90 µT) were twice as 1284!
strong as the ambient magnetic field in Kansas City (45 µT) and were well above intensity 1285!
alterations known to cause birds to ignore geomagnetic cues (W. Wiltschko, 1972). 1286!
Additionally, Sastre et al. chose to use clockwise but not counterclockwise rotations (conditions 1287!
B and C). In our study, rotating the declination clockwise did not yield statistically significant 1288!
effects although the reasons are not yet understood (Table 1). 1289!
1290!
1291!
Acknowledgements. This work was supported directly by Human Frontiers Science Program 1292!
grant HFSP-RGP0054/2014 to S.S., J.L.K. and A.M., and more recent analysis of data was 1293!
supported by DARPA RadioBio Program grant (D17AC00019) to JLK and SS, and Japan 1294!
Society for the Promotion of Science (JSPS) KAKENHI grant 18H03500 to AM. Previous 1295!
support to J.L.K. from the Fetzer institute allowed construction of an earlier version of the 2 m 1296!
Merritt coil system. C.X.W. and S.S. have been partly supported by JST.CREST. We thank 1297!
Dragos Harabor, James Martin, Kristján Jónsson, Mara Green and Sarah Crucilla for work on 1298!
earlier versions of this project and other members of the Kirschvink, Shimojo, and Matani labs 1299!
for discussions and suggestions. We also thank James Randi, co-founder of the Committee for 1300!
the Scientific Investigation of Claims of the Paranormal (CSICOP), for advice on minimizing 1301!
potential artifacts in the experimental design. Dr. Heinrich Mouritsen of the University of 1302!
Oldenberg gave valuable advice for construction of the Faraday cage and input on an earlier draft 1303!
of the manuscript. 1304!
1305!
Author Contributions. J.L.K. initiated, and with S.S. and A.M., planned and directed the 1306!
research. C.X.W., D.A.W. and I.A.H. largely designed the stimulation protocols and conducted 1307!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
47!
the experiments and data analysis. C.P.C., J.N.H.A., S.E.B. and Y.M. designed and built the 1308!
Faraday cage and implemented the magnetic stimulation protocols. All authors contributed to 1309!
writing and editing the manuscript. 1310!
1311!
Online Content: All digital data are available at https://data.caltech.edu/records/930 and 1312!
https://data.caltech.edu/records/931 , including MatLabTM scripts used for the automatic data 1313!
analysis. 1314!
1315!
References 1316!
1317!
Able,!K.!P.,!&!Gergits,!W.!F.!(1985).!Human!Navigation:!!Attempts!to!Replicate!Baker's!1318!
Displacement!Experiment.!In!J.!L.!Kirschvink,!D.!S.!Jones,!&!B.!J.!MacFadden!(Eds.),!1319!
!"#$%&'&%()'*+'$%,"-'."&'*$("$/(!"#$%&*,%0%1&'*$('$(2,#"$'3+34((5(6%7(1320!
)'*+"#$%&'3+!(pp.!569-572).!New!York:!Plenum!Press.!1321!
Baker,!R.!R.!(1980).!Goal!orientation!by!blindfolded!humans!after!long-distance!1322!
displacement:!possible!involvement!of!a!magnetic!sense.!80'%$0%9(:;<(4469),!555-1323!
557.!!1324!
Baker,!R.!R.!(1982).!=>+"$($"?'#"&'*$("$/(&@%(A&@(3%$3%:!Simon!and!Schuster.!1325!
Baker,!R.!R.!(1987).!Human!Navigation!and!Magnetoreception!-!the!Manchester!1326!
Experiments!Do!Replicate.!5$'+"-()%@"?'*>,9(BC,!691-704.!doi:Doi!10.1016/S0003-1327!
3472(87)80105-7!1328!
Bazylinski,!D.!A.,!Schlezinger,!D.!R.,!Howes,!B.!H.,!Frankel,!R.!B.,!&!Epstein,!S.!S.!(2000).!1329!
Occurrence!and!distribution!of!diverse!populations!of!magnetic!protists!in!a!1330!
chemically!stratified!coastal!salt!pond.!D@%+'0"-(E%*-*#F9(;AG(3-4),!319-328.!doi:Doi!1331!
10.1016/S0009-2541(00)00211-4!1332!
Beason,!R.!C.,!&!Semm,!P.!(1996).!Does!the!avian!ophthalmic!nerve!carry!magnetic!1333!
navigational!information?!H*>,$"-(*I(JK1%,'+%$&"-()'*-*#F9(;GG(5),!1241-1244.!!1334!
Beason,!R.!C.,!Wiltschko,!R.,!&!Wiltschko,!W.!(1997).!Pigeon!homing:!Effects!of!magnetic!1335!
pulses!on!initial!orientation.!5>L9(;;M(3),!405-415.!doi:Doi!10.2307/4089242!1336!
Benjamini,!Y.,!&!Hochberg,!Y.!(1995).!Controlling!the!False!Discovery!Rate!-!a!Practical!and!1337!
Powerful!Approach!to!Multiple!Testing.!H*>,$"-(*I(&@%(N*F"-(8&"&'3&'0"-(8*0'%&F(8%,'%3(1338!
)O!%&@*/*-*#'0"-9(CP(1),!289-300.!!1339!
Block,!S.!M.!(1992).!Biophysical!Aspects!of!Sensory!Transduction.!In!D.!P.!Corey!&!S.!D.!1340!
Roper!(Eds.),!8%$3*,F(Q,"$3/>0&'*$!(Vol.!45,!pp.!424).!Marine!Biological!Laboratory,!1341!
Woods!Hole,!Massachusetts:!Rockfeller!University!Press.!1342!
Boorman,!G.!A.,!Bernheim,!N.!J.,!Galvin,!M.!J.,!Newton,!S.!A.,!Parham,!F.!M.,!Portier,!C.!J.,!&!1343!
Wolfe,!M.!S.!(1999).!6RJ=8(N%1*,&(*$(=%"-&@(JII%0&3(I,*+(JK1*3>,%(&*(S*7%,OT'$%(1344!
U,%V>%$0F(J-%0&,'0("$/(!"#$%&'0(U'%-/3!(NIEHS!Ed.!!Vol.!99-4493).!Research!Triangle!1345!
Park,!NC!27709:!Department!of!Health!&!Humn!Services,!US!Government.!1346!
Carporzen,!L.,!Weiss,!B.!P.,!Gilder,!S.!A.,!Pommier,!A.,!&!Hart,!R.!J.!(2012).!Lightning!1347!
remagnetization!of!the!Vredefort!impact!crater:!No!evidence!for!impact-generated!1348!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
48!
magnetic!fields.!H*>,$"-(*I(E%*1@F3'0"-(N%3%",0@OS-"$%&39(;;P.!doi:Artn!E01007!1349!
10.1029/2011je003919!1350!
Cohen,!M.!X.!(2014).!5$"-F.'$#(6%>,"-(Q'+%(8%,'%3(W"&"(Q@%*,F("$/(S,"0&'0%(S,%I"0%.!1351!
Cambridge,!Massachusetts:!MIT!Press.!1352!
Counselman,!C.!(2013).!Excellent,!Easy,!Cheap!Common-Mode!Chokes.!6"&'*$"-(D*$&%3&(1353!
H*>,$"-(*I(&@%(5+%,'0"$(N"/'*(N%-"F(T%"#>%9(M;(1),!3-5.!!1354!
Diaz!Ricci,!J.!C.,!&!Kirschvink,!J.!L.!(1992).!Magnetic!domain!state!and!coercivity!predictions!1355!
for!biogenic!greigite!(Fe3S4):!A!comparison!of!theory!with!magnetosome!1356!
observations.!HX(E%*1@F3X(N%3X9(GP,!17309-17315.!!1357!
Diebel,!C.!E.,!Proksch,!R.,!Green,!C.!R.,!Neilson,!P.,!&!Walker,!M.!M.!(2000).!Magnetite!defines!1358!
a!vertebrate!magnetoreceptor.!6"&>,%9(M<A(6793),!299-302.!doi:10.1038/35018561!1359!
Dunn,!J.!R.,!Fuller,!M.,!Zoeger,!J.,!Dobson,!J.,!Heller,!F.,!Hammann,!J.,!.!.!.!Moskowitz,!B.!M.!1360!
(1995).!Magnetic!material!in!the!human!hippocampus.!),"'$(N%3()>--9(BA(2),!149-1361!
153.!!1362!
Eder,!S.!H.,!Cadiou,!H.,!Muhamad,!A.,!McNaughton,!P.!A.,!Kirschvink,!J.!L.,!&!Winklhofer,!M.!1363!
(2012).!Magnetic!characterization!of!isolated!candidate!vertebrate!magnetoreceptor!1364!
cells.!S,*0(6"&-(50"/(80'(Y(8(59(;<G(30),!12022-12027.!1365!
doi:10.1073/pnas.1205653109!1366!
Elbers,!D.,!Bulte,!M.,!Bairlein,!F.,!Mouritsen,!H.,!&!Heyers,!D.!(2017).!Magnetic!activation!in!1367!
the!brain!of!the!migratory!northern!wheatear!(Oenanthe!oenanthe).!H(D*+1(S@F3'*-(1368!
5(6%>,*%&@*-(8%$3(6%>,"-()%@"?(S@F3'*-9(:<B(8),!591-600.!doi:10.1007/s00359-017-1369!
1167-7!1370!
Engels,!S.,!Schneider,!N.!L.,!Lefeldt,!N.,!Hein,!C.!M.,!Zapka,!M.,!Michalik,!A.,!.!.!.!Mouritsen,!H.!1371!
(2014).!Anthropogenic!electromagnetic!noise!disrupts!magnetic!compass!1372!
orientation!in!a!migratory!bird.!6"&>,%9(C<G(7500),!353-356.!1373!
doi:10.1038/nature13290!1374!
Ernst,!D.!A.,!&!Lohmann,!K.!J.!(2016).!Effect!of!magnetic!pulses!on!Caribbean!spiny!lobsters:!1375!
implications!for!magnetoreception.!H(JK1()'*-9(:;G(Pt!12),!1827-1832.!1376!
doi:10.1242/jeb.136036!1377!
Fillmore,!E.!P.,!&!Seifert,!M.!F.!(2015).!Anatomy!of!the!Trigeminal!Nerve.!In!R.!S.!Tubbs,!E.!1378!
Rizk,!M.!Shoja,!M.!Loukas,!N.!Barbaro,!&!R.!Spinner!(Eds.),!6%,?%3("$/(6%,?%(R$Z>,'%3!1379!
(Vol.!1,!pp.!319-350):!Academic!Press.!1380!
Frankel,!R.!B.,!&!Blakemore,!R.!P.!(1980).!Navigational!Compass!in!Magnetic!Bacteria.!1381!
H*>,$"-(*I(!"#$%&'3+("$/(!"#$%&'0(!"&%,'"-39(;CO[(Jan-),!1562-1564.!doi:Doi!1382!
10.1016/0304-8853(80)90409-6!1383!
Gilder,!S.!A.,!Wack,!M.,!Kaub,!L.,!Roud,!S.!C.,!Petersen,!N.,!Heinsen,!H.,!.!.!.!Schmitz,!C.!(2018).!1384!
Distribution!of!magnetic!remanence!carriers!in!the!human!brain.!80'%$&'I'0(N%1*,&39(1385!
[(11363).!doi:DOI:10.1038/s41598-018-29766-z!1!1386!
Gould,!J.!S.,!&!Able,!K.!P.!(1981).!Human!homing:!an!elusive!phenomenon.!80'%$0%9(1387!
:;:(4498),!1061-1063.!!1388!
Hand,!E.!(2016).!Polar!explorer.!80'%$0%9(BC:(6293),!1508-1510,!1512-1503.!1389!
doi:10.1126/science.352.6293.1508!1390!
Hartmann,!T.,!Schlee,!W.,!&!Weisz,!N.!(2012).!It's!only!in!your!head:!expectancy!of!aversive!1391!
auditory!stimulation!modulates!stimulus-induced!auditory!cortical!alpha!1392!
desynchronization.!6%>,*'+"#%9(A<(1),!170-178.!1393!
doi:10.1016/j.neuroimage.2011.12.034!1394!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
49!
Haviland,!J.!B.!(1998).!Guugu!Yimithir!cardinal!directions.!J&@*39(:A(1),!25-47.!doi:DOI!1395!
10.1525/eth.1998.26.1.25!1396!
Holland,!R.!A.!(2010).!Differential!effects!of!magnetic!pulses!on!the!orientation!of!naturally!1397!
migrating!birds.!H(N(8*0(R$&%,I"0%9(P(52),!1617-1625.!doi:10.1098/rsif.2010.0159!1398!
Holland,!R.!A.,!&!Helm,!B.!(2013).!A!strong!magnetic!pulse!affects!the!precision!of!departure!1399!
direction!of!naturally!migrating!adult!but!not!juvenile!birds.!H*>,$"-(*I(&@%(N*F"-(1400!
8*0'%&F(R$&%,I"0%9(;<,!20121047.!!1401!
Holland,!R.!A.,!Kirschvink,!J.!L.,!Doak,!T.!G.,!&!Wikelski,!M.!(2008).!Bats!use!magnetite!to!1402!
detect!the!earth's!magnetic!field.!S-*3(2$%9(B(2),!e1676.!1403!
doi:10.1371/journal.pone.0001676!1404!
Hore,!P.!J.,!&!Mouritsen,!H.!(2016).!The!Radical-Pair!Mechanism!of!Magnetoreception.!In!K.!1405!
A.!Dill!(Ed.),!5$$>"-(N%?'%7(*I()'*1@F3'039(\*-(MC!(Vol.!45,!pp.!299-344).!1406!
Irwin,!W.!P.,!&!Lohmann,!K.!J.!(2005).!Disruption!of!magnetic!orientation!in!hatchling!1407!
loggerhead!sea!turtles!by!pulsed!magnetic!fields.!H(D*+1(S@F3'*-(5(6%>,*%&@*-(8%$3(1408!
6%>,"-()%@"?(S@F3'*-9(;G;(5),!475-480.!doi:10.1007/s00359-005-0609-9!1409!
Johnsen,!S.,!&!Lohmann,!K.!J.!(2008).!Magnetoreception!in!animals.!S@F3'03(Q*/"F9(A;(3),!29-1410!
35.!doi:Doi!10.1063/1.2897947!1411!
Kalmijn,!A.!J.!(1981).!Biophysics!of!Geomagnetic-Field!Detection.!RJJJ(Q,"$3"0&'*$3(*$(1412!
!"#$%&'039(;P(1),!1113-1124.!doi:Doi!10.1109/Tmag.1981.1061156!1413!
Kirschvink,!J.,!Padmanabha,!S.,!Boyce,!C.,!&!Oglesby,!J.!(1997).!Measurement!of!the!1414!
threshold!sensitivity!of!honeybees!to!weak,!extremely!low-frequency!magnetic!1415!
fields.!H(JK1()'*-9(:<<(Pt!9),!1363-1368.!!1416!
Kirschvink,!J.!L.!(1992).!Uniform!magnetic!fields!and!Double-wrapped!coil!systems:!!1417!
Improved!techniques!for!the!design!of!biomagnetic!experiments.!1418!
)'*%-%0&,*+"#$%&'039(;B,!401-411.!!1419!
Kirschvink,!J.!L.,!&!Gould,!J.!L.!(1981).!Biogenic!magnetite!as!a!basis!for!magnetic!field!1420!
sensitivity!in!animals.!)'*(8F3&%+39(;B,!181-201.!!1421!
Kirschvink,!J.!L.,!&!Kobayashi-Kirschvink,!A.!(1991).!Is!geomagnetic!sensitivity!real?!1422!
Replication!of!the!Walker-Bitterman!conditioning!experiment!in!honey!bees.!1423!
5+%,'0"$(]**-*#'3&9(B;,!169-185.!!1424!
Kirschvink,!J.!L.,!Kobayashi-Kirschvink,!A.,!&!Woodford,!B.!J.!(1992).!Magnetite!1425!
biomineralization!in!the!human!brain.!S,*0(6"&-(50"/(80'(Y(8(59([G(16),!7683-7687.!!1426!
Kirschvink,!J.!L.,!Winklhofer,!M.,!&!Walker,!M.!M.!(2010).!Biophysics!of!magnetic!1427!
orientation:!strengthening!the!interface!between!theory!and!experimental!design.!H(1428!
N(8*0(R$&%,I"0%9(P(8>11-(:,!S179-191.!doi:10.1098/rsif.2009.0491.focus!1429!
Klimesch,!W.!(1999).!EEG!alpha!and!theta!oscillations!reflect!cognitive!and!memory!1430!
performance:!a!review!and!analysis.!),"'$(N%3(),"'$(N%3(N%?9(:G(2-3),!169-195.!!1431!
Klimesch,!W.,!Doppelmayr,!M.,!Russegger,!H.,!Pachinger,!T.,!&!Schwaiger,!J.!(1998).!Induced!1432!
alpha!band!power!changes!in!the!human!EEG!and!attention.!6%>,*30'(T%&&9(:MM(2),!1433!
73-76.!doi:10.1016/s0304-3940(98)00122-0!1434!
Kobayashi,!A.,!&!Kirschvink,!J.!L.!(1995).!Magnetoreception!and!EMF!Effects:!!Sensory!1435!
Perception!of!the!geomagnetic!field!in!Animals!&!Humans.!In!M.!Blank!(Ed.),!1436!
J-%0&,*+"#$%&'0(U'%-/34(()'*-*#'0"-(R$&%,"0&'*$3("$/(!%0@"$'3+3!(pp.!367-394).!1437!
Washington,!DC.:!American!Chemical!Society!Books.!1438!
Kramer,!G.!(1953).!Wird!die!Sonnenhohe!bei!der!Heimfindeorientierung!verwertet?!H*>,$"-(1439!
*I(2,$'&@*-#F9(GM,!201-219.!!1440!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
50!
Landler,!L.,!Painter,!M.!S.,!Youmans,!P.!W.,!Hopkins,!W.!A.,!&!Phillips,!J.!B.!(2015).!1441!
Spontaneous!magnetic!alignment!by!yearling!snapping!turtles:!rapid!association!of!1442!
radio!frequency!dependent!pattern!of!magnetic!input!with!novel!surroundings.!S-*3(1443!
2$%9(;<(5),!e0124728.!doi:10.1371/journal.pone.0124728!1444!
Levinson,!S.!C.!(2003).!81"0%('$(T"$#>"#%("$/(D*#$'&'*$.!Cambridge:!Cambridge!University!1445!
Press.!1446!
Light,!P.,!Salmon,!M.,!&!Lohmann,!K.!J.!(1993).!Geomagnetic!Orientation!of!Loggerhead!Sea-1447!
Turtles!-!Evidence!for!an!Inclination!Compass.!H*>,$"-(*I(JK1%,'+%$&"-()'*-*#F9(;[:,!1448!
1-9.!!1449!
Liu,!G.!T.!(2005).!The!trigeminal!nerve!and!its!central!connections.!In!N.!R.!Miller!,!N.!J.!1450!
Newman!,!V.!Biousse,!&!K.!J.!B.!(Eds.),!^"-3@(_(=*F&`3(D-'$'0"-(6%>,*O21@&@"-+"-*#F9(1451!
A&@(%/'&'*$!(6th!ed.,!Vol.!1,!pp.!1233-1268).!Philadelphia:!Lippencott!Williams!&!1452!
Wilkins.!1453!
Lohmann,!K.!J.,!Cain,!S.!D.,!Dodge,!S.!A.,!&!Lohmann,!C.!M.!(2001).!Regional!magnetic!fields!as!1454!
navigational!markers!for!sea!turtles.!80'%$0%9(:GM(5541),!364-366.!1455!
doi:10.1126/science.1064557!1456!
Lohmann,!K.!J.,!&!Lohmann,!C.!M.!F.!(1996).!Detection!of!magnetic!field!intensity!by!sea!1457!
turtles.!6"&>,%9(B[<(6569),!59-61.!doi:DOI!10.1038/380059a0!1458!
Lopez,!C.,!&!Blanke,!O.!(2011).!The!thalamocortical!vestibular!system!in!animals!and!1459!
humans.!),"'$(N%3(N%?9(AP(1-2),!119-146.!doi:10.1016/j.brainresrev.2010.12.002!1460!
Maher,!B.!A.,!Ahmed,!I.!A.,!Karloukovski,!V.,!MacLaren,!D.!A.,!Foulds,!P.!G.,!Allsop,!D.,!.!.!.!1461!
Calderon-Garciduenas,!L.!(2016).!Magnetite!pollution!nanoparticles!in!the!human!1462!
brain.!S,*0(6"&-(50"/(80'(Y(8(59(;;B(39),!10797-10801.!1463!
doi:10.1073/pnas.1605941113!1464!
Martin,!H.,!&!Lindauer,!M.!(1977).!Der!Einfluss!der!Erdmagnetfelds!und!die!1465!
Schwerorientierung!der!Honigbiene.(HX(D*+1X(S@F3'*-X9(;::9(,!145-187.!!1466!
Meakins,!F.!(2011).!Spaced!Out:!Intergenerational!Changes!in!the!Expression!of!Spatial!1467!
Relations!by!Gurindji!People.!5>3&,"-'"$(H*>,$"-(*I(T'$#>'3&'039(B;(1),!43-77.!doi:Pii!1468!
932432693!10.1080/07268602.2011.532857!1469!
Meakins,!F.,!&!Algy,!C.!(2016).!Deadly!Reckoning:!Changes!in!Gurindji!Children's!Knowledge!1470!
of!Cardinals.!5>3&,"-'"$(H*>,$"-(*I(T'$#>'3&'039(BA(4),!479-501.!1471!
doi:10.1080/07268602.2016.1169973!1472!
Meakins,!F.,!Jones,!C.,!&!Algy,!C.!(2016).!Bilingualism,!language!shift!and!the!corresponding!1473!
expansion!of!spatial!cognitive!systems.!T"$#>"#%(80'%$0%39(CM,!1-13.!1474!
doi:10.1016/j.langsci.2015.06.002!1475!
Merritt,!R.,!Purcell,!C.,!&!Stroink,!G.!(1983).!Uniform!Magnetic-Field!Produced!by!3-Square,!1476!
4-Square,!and!5-Square!Coils.!N%?'%7(*I(80'%$&'I'0(R$3&,>+%$&39(CM(7),!879-882.!1477!
doi:Doi!10.1063/1.1137480!1478!
Mora,!C.!V.,!Davison,!M.,!Wild,!J.!M.,!&!Walker,!M.!M.!(2004).!Magnetoreception!and!its!1479!
trigeminal!mediation!in!the!homing!pigeon.!6"&>,%9(MB:(7016),!508-511.!1480!
doi:10.1038/nature03077!1481!
Munro,!U.,!Munro,!J.!A.,!Phillips,!J.!B.,!Wiltschko,!R.,!&!Wiltschko,!W.!(1997).!Evidence!for!a!1482!
magnetite-based!navigational!''map''!in!birds.!6"&>,7'33%$30@"I&%$9([M(1),!26-28.!1483!
doi:DOI!10.1007/s001140050343!1484!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
51!
Munro,!U.,!Munro,!J.!A.,!Phillips,!J.!B.,!&!Wiltschko,!W.!(1997).!Effect!of!wavelength!of!light!1485!
and!pulse!magnetisation!on!different!magnetoreception!systems!in!a!migratory!bird.!1486!
5>3&,"-'"$(H*>,$"-(*I(]**-*#F9(MC(2),!189-198.!doi:Doi!10.1071/Zo96066!1487!
Nuwer,!M.!R.,!Comi,!G.,!Emerson,!R.,!Fuglsang-Frederiksen,!A.,!Guerit,!J.!M.,!Hinrichs,!H.,!.!.!.!1488!
Rappelsburger,!P.!(1998).!IFCN!standards!for!digital!recording!of!clinical!EEG.!1489!
International!Federation!of!Clinical!Neurophysiology.!J-%0&,*%$0%1@"-*#,(D-'$(1490!
6%>,*1@F3'*-9(;<A(3),!259-261.!doi:10.1016/s0013-4694(97)00106-5!1491!
Peng,!W.,!Hu,!L.,!Zhang,!Z.,!&!Hu,!Y.!(2012).!Causality!in!the!association!between!P300!and!1492!
alpha!event-related!desynchronization.!S-*3(2$%9(P(4),!e34163.!1493!
doi:10.1371/journal.pone.0034163!1494!
Pfurtscheller,!G.,!&!Lopes!da!Silva,!F.!H.!(1999).!Event-related!EEG/MEG!synchronization!1495!
and!desynchronization:!basic!principles.!D-'$(6%>,*1@F3'*-9(;;<(11),!1842-1857.!1496!
doi:10.1016/s1388-2457(99)00141-8!1497!
Pfurtscheller,!G.,!Neuper,!C.,!&!Mohl,!W.!(1994).!Event-related!desynchronization!(ERD)!1498!
during!visual!processing.!R$&(H(S3F0@*1@F3'*-9(;A(2-3),!147-153.!doi:10.1016/0167-1499!
8760(89)90041-x!1500!
Poisson,!R.!J.,!&!Miller,!M.!E.!(2014).!Spatial!disorientation!mishap!trends!in!the!U.S.!Air!1501!
force!1993-2013.!5?'"&(81"0%(J$?',*$(!%/9([C(9),!919-924.!1502!
doi:10.3357/ASEM.3971.2014!1503!
Ritz,!T.,!Adem,!S.,!&!Schulten,!K.!(2000).!A!model!for!photoreceptor-based!1504!
magnetoreception!in!birds.!)'*1@F3(H9(P[(2),!707-718.!doi:10.1016/S0006-1505!
3495(00)76629-X!1506!
Rosenblum,!B.,!Jungerman,!R.!L.,!&!Longfellow,!L.!(1985).!Limits!to!induction-based!1507!
magnetoreception.!In!J.!L.!Kirschvink,!D.!S.!Jones,!&!B.!J.!MacFadden!(Eds.),!!"#$%&'&%(1508!
)'*+'$%,"-'."&'*$("$/(!"#$%&*,%0%1&'*$('$(2,#"$'3+34((5(6%7()'*+"#$%&'3+!(pp.!1509!
223-232).!New!York:!Plenum!Press.!1510!
Saper,!C.!B.!(2002).!The!central!autonomic!nervous!system:!conscious!visceral!perception!1511!
and!autonomic!pattern!generation.!5$$>(N%?(6%>,*30'9(:C,!433-469.!1512!
doi:10.1146/annurev.neuro.25.032502.111311!1513!
Sastre,!A.,!Graham,!C.,!Cook,!M.!R.,!Gerkovich,!M.!M.,!&!Gailey,!P.!(2002).!Human!EEG!1514!
responses!to!controlled!alterations!of!the!Earth's!magnetic!field.!D-'$(6%>,*1@F3'*-9(1515!
;;B(9),!1382-1390.!doi:10.1016/s1388-2457(02)00186-4!1516!
Schulten,!K.!(1982).!Magnetic-Field!Effects!in!Chemistry!and!Biology.!U%3&L*,1%,1,*a-%+%O1517!
5/?"$0%3('$(8*-'/(8&"&%(S@F'039(::,!61-83.!doi:DOI!10.1007/BFb0107935!1518!
Schultheiss-Grassi,!P.!P.,!Wessiken,!R.,!&!Dobson,!J.!(1999).!TEM!investigations!of!biogenic!1519!
magnetite!extracted!from!the!human!hippocampus.!)'*0@'+()'*1@F3(50&"9(;M:A(1),!1520!
212-216.!doi:10.1016/s0304-4165(98)00160-3!1521!
Schwarze,!S.,!Steenken,!F.,!Thiele,!N.,!Kobylkov,!D.,!Lefeldt,!N.,!Dreyer,!D.,!.!.!.!Mouritsen,!H.!1522!
(2016).!Migratory!blackcaps!can!use!their!magnetic!compass!at!5!degrees!1523!
inclination,!but!are!completely!random!at!0!degrees!inclination.!80'(N%19(A,!33805.!1524!
doi:10.1038/srep33805!1525!
Semm,!P.,!&!Beason,!R.!C.!(1990).!Responses!to!small!magnetic!variations!by!the!trigeminal!1526!
system!of!the!bobolink.!),"'$(N%3()>--9(:C(5),!735-740.!doi:10.1016/0361-1527!
9230(90)90051-z!1528!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
52!
Tomanova,!K.,!&!Vacha,!M.!(2016).!The!magnetic!orientation!of!the!Antarctic!amphipod!1529!
Gondogeneia!antarctica!is!cancelled!by!very!weak!radiofrequency!fields.!H(JK1()'*-9(1530!
:;G(Pt!11),!1717-1724.!doi:10.1242/jeb.132878!1531!
Tukey,!J.!W.!(1949).!Comparing!individual!means!in!the!analysis!of!variance.!)'*+%&,'039(1532!
C(2),!99-114.!doi:10.2307/3001913!1533!
Veniero,!D.,!Bortoletto,!M.,!&!Miniussi,!C.!(2009).!TMS-EEG!co-registration:!On!TMS-induced!1534!
artifact.!D-'$'0"-(6%>,*1@F3'*-*#F9(;:<(7),!1392-1399.!1535!
doi:10.1016/j.clinph.2009.04.023!1536!
Walker,!M.!M.,!Dennis,!T.!E.,!&!Kirschvink,!J.!L.!(2002).!The!magnetic!sense!and!its!use!in!1537!
long-distance!navigation!by!animals.!D>,,%$&(21'$'*$('$(6%>,*a'*-*#F9(;:(6),!735-1538!
744.!doi:Doi!10.1016/S0959-4388(02)00389-6!1539!
Walker,!M.!M.,!Diebel,!C.!E.,!Haugh,!C.!V.,!Pankhurst,!P.!M.,!Montgomery,!J.!C.,!&!Green,!C.!R.!1540!
(1997).!Structure!and!function!of!the!vertebrate!magnetic!sense.!6"&>,%9(BG<(6658),!1541!
371-376.!!1542!
Wegner,!R.!E.,!Begall,!S.,!&!Burda,!H.!(2006).!Magnetic!compass!in!the!cornea:!local!1543!
anaesthesia!impairs!orientation!in!a!mammal.!H(JK1()'*-9(:<G(Pt!23),!4747-4750.!1544!
doi:10.1242/jeb.02573!1545!
Welch,!P.!D.!(1967).!Use!of!Fast!Fourier!Transform!for!Estimation!of!Power!Spectra!-!a!1546!
Method!Based!on!Time!Averaging!over!Short!Modified!Periodograms.!R%%%(1547!
Q,"$3"0&'*$3(*$(5>/'*("$/(J-%0&,*"0*>3&'039(5>;C(2),!70-+.!doi:Doi!1548!
10.1109/Tau.1967.1161901!1549!
Westby,!G.!W.!M.,!&!Partridge,!K.!J.!(1986).!Human!Homing!-!Still!No!Evidence!Despite!1550!
Geomagnetic!Controls.!H*>,$"-(*I(JK1%,'+%$&"-()'*-*#F9(;:<,!325-331.!!1551!
Wiltschko,!R.,!Thalau,!P.,!Gehring,!D.,!Niessner,!C.,!Ritz,!T.,!&!Wiltschko,!W.!(2015).!1552!
Magnetoreception!in!birds:!the!effect!of!radio-frequency!fields.!H(N(8*0(R$&%,I"0%9(1553!
;:(103).!doi:10.1098/rsif.2014.1103!1554!
Wiltschko,!R.,!&!Wiltschko,!W.!(1995).!!"#$%&'0(*,'%$&"&'*$('$("$'+"-3!(Vol.!33).!Berlin:!1555!
Springer.!1556!
Wiltschko,!W.!(1972).!The!influence!of!magnetic!total!intensity!and!inclination!on!1557!
directions!preferred!by!migrating!European!robins!(Erithacus!rubeccula).!In!S.!R.!1558!
Galler,!K.!Schmidt-Koenig,!G.!J.!Jacobs,!&!R.!E.!Belleville!(Eds.),!5$'+"-(2,'%$&"&'*$(1559!
"$/(6"?'#"&'*$!(Vol.!NASA!SP-262,!pp.!569-578).!Washington,!D.C.,!USA:!U.S.!1560!
Government!Printing!Office.!1561!
Wiltschko,!W.,!Ford,!H.,!Munro,!U.,!Winklhofer,!M.,!&!Wiltschko,!R.!(2007).!Magnetite-based!1562!
magnetoreception:!the!effect!of!repeated!pulsing!on!the!orientation!of!migratory!1563!
birds.!H(D*+1(S@F3'*-(5(6%>,*%&@*-(8%$3(6%>,"-()%@"?(S@F3'*-9(;GB(5),!515-522.!1564!
doi:10.1007/s00359-006-0207-5!1565!
Wiltschko,!W.,!Munro,!U.,!Beason,!R.!C.,!Ford,!H.,!&!Wiltschko,!R.!(1994).!A!Magnetic!Pulse!1566!
Leads!to!a!Temporary!Deflection!in!the!Orientation!of!Migratory!Birds.!JK1%,'%$&'"9(1567!
C<(7),!697-700.!doi:Doi!10.1007/Bf01952877!1568!
Wiltschko,!W.,!Munro,!U.,!Ford,!H.,!&!Wiltschko,!R.!(1998).!Effect!of!a!magnetic!pulse!on!the!1569!
orientation!of!silvereyes,!zosterops!l.!lateralis,!during!spring!migration.!H(JK1()'*-9(1570!
:<;(bS&(:Bc(23),!3257-3261.!!1571!
Wiltschko,!W.,!Munro,!U.,!Ford,!H.,!&!Wiltschko,!R.!(2009).!Avian!orientation:!the!pulse!1572!
effect!is!mediated!by!the!magnetite!receptors!in!the!upper!beak.!S,*0()'*-(80'9(1573!
:PA(1665),!2227-2232.!doi:10.1098/rspb.2009.0050!1574!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;

!
53!
Wiltschko,!W.,!Munro,!U.,!Wiltschko,!R.,!&!Kirschvink,!J.!L.!(2002).!Magnetite-based!1575!
magnetoreception!in!birds:!the!effect!of!a!biasing!field!and!a!pulse!on!migratory!1576!
behavior.!H(JK1()'*-9(:<C(Pt!19),!3031-3037.!!1577!
Wiltschko,!W.,!&!Wiltschko,!R.!(1995).!Migratory!Orientation!of!European!Robins!Is!1578!
Affected!by!the!Wavelength!of!Light!as!Well!as!by!a!Magnetic!Pulse.!H*>,$"-(*I(1579!
D*+1","&'?%(S@F3'*-*#F("O8%$3*,F(6%>,"-("$/()%@"?'*,"-(S@F3'*-*#F9(;PP(3),!363-1580!
369.!!1581!
Yeagley,!H.!L.!(1947).!A!Preliminary!Study!of!a!Physical!Basis!of!Bird!Navigation.!H*>,$"-(*I(1582!
511-'%/(S@F3'039(;[(12),!1035-1063.!doi:Doi!10.1063/1.1697587!1583!
1584!
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/448449doi: bioRxiv preprint first posted online Oct. 20, 2018;