Neuroscience Letters 469 (2010) 164–168
Contents lists available at ScienceDirect
j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / n e u l e t
Mobile-phone pulse triggers evoked potentials
Simona Carrubba a, Clifton Frilot II b, Andrew L. Chesson Jr. c, Andrew A. Marino a,∗
a Department of Orthopaedic Surgery, LSU Health Sciences Center, Shreveport, LA 71130-3932, United States
b School of Allied Health, LSU Health Sciences Center, Shreveport, LA 71130-3932, United States
c Department of Neurology, LSU Health Sciences Center, Shreveport, LA 71130-3932, United States
a r t i c l e i n f o
a b s t r a c t
If mobile-phone electromagnetic ﬁelds (EMFs) are hazardous, as suggested in the literature, processes or
Received 2 October 2009
mechanisms must exist that allow the body to detect the ﬁelds. We hypothesized that the low-frequency
Received in revised form 27 October 2009
pulses produced by mobile phones (217 Hz) were detected by sensory transduction, as evidenced by the
Accepted 24 November 2009
ability of the pulses to trigger evoked potentials (EPs). Electroencephalograms (EEGs) were recorded from
six standard locations in 20 volunteers and analyzed to detect brain potentials triggered by a pulse of the
type produced by mobile phones. Evoked potentials having the expected latency were found in 90% of the
volunteers, as assessed using a nonlinear method of EEG analysis. Evoked potentials were not detected
when the EEG was analyzed using time averaging. The possibility of systematic error was excluded by
sham-exposure analyses. The results implied that mobile-phones trigger EP at the rate of 217 Hz during
ordinary phone use. Chronic production of the changes in brain activity might be pertinent to the reports
of health hazards among mobile-phone users.
© 2009 Elsevier Ireland Ltd. All rights reserved.
Mobile phones are increasingly more common, and questions have
tery currents altered the contingent negative variation triggered
been raised concerning whether the electromagnetic ﬁelds (EMFs)
by acoustic stimuli .
they emit are partly responsible for brain cancer or other dis-
Evidence that at least one of the types of EMFs produced by
eases . Mobile phones transmit and receive high-frequency
mobile phones was capable of eliciting brain potentials would pro-
EMFs (∼1 GHz), and also emit low-frequency magnetic pulses
vide a possible basis for explaining how chronic phone use leads to
(217 Hz) from the phone’s circuitry and battery currents 
disease. Consequently, in a study of 20 clinically normal volunteer
(Fig. 1a). If exposure to mobile-phone EMFs is hazardous, pro-
subjects, 7 males (age range 22–62 years) and 13 females (18–53
cesses or mechanisms must exist that allow the body to detect
years), we addressed the question of whether a low-frequency
at least one ﬁeld. One possibility is that the EMF is detected
pulse of the type produced by mobile phones was capable of trig-
by sensory transduction, like other environmental stimuli .
EMFs of the type produced by the electrical power system trig-
The subjects gave written informed consent prior to participat-
gered evoked potentials (EPs) having latencies of about 250 ms ,
ing in the experiment. They were informed of the goals, methods,
but the ability of mobile-phone EMFs to trigger EPs, as assessed
and general design of the investigation, but were not told exactly
using a standard stimulus-response protocol , has not been
when during the experimental session that the EMF stimulus would
be applied. The institutional review board at the LSU Health Sci-
Reports that brain electrical activity was affected during expo-
ences Center approved all experimental procedures.
sure to simulated mobile-phone EMFs supported the transduction
Mobile phones emit a complicated temporal array of elec-
hypothesis. As examples, high-frequency mobile-phone EMFs
tromagnetic, acoustic, thermal, and tactile stimuli. To avoid
altered the amplitude of the P50 component of the auditory evoked
confounding effects and to facilitate use of a standard protocol
potential , spectral coherence during an auditory memory
for detecting EPs  we applied a simulated mobile-phone pulse.
task , and alpha power during sleep . Both high-frequency
The strength of the pulse was chosen based on measurements of a
EMFs and low-frequency magnetic pulses from the phone’s bat-
typical mobile phone (Model 6085, Nokia, Helsinki, Finland) made
during a phone call (3 km between the base tower and phone). The
peak strength of the magnetic pulse 10 cm from the phone was
3 T, and the duration of each pulse was 0.7 ms (MAG-03, GMW,
∗ Corresponding author at: Department of Orthopaedic Surgery, LSU Health Sci-
Redwood City, CA, USA) (Fig. 1b).
ences Center, P.O. Box 33932, 1501 Kings Hwy., Shreveport, LA 71130-3932, United
Several factors entered into our considerations regarding
States. Tel.: +1 318 675 6180; fax: +1 318 675 6186.
E-mail address: firstname.lastname@example.org (A.A. Marino).
the design of the apparatus necessary to repetitively apply a
0304-3940/$ – see front matter © 2009 Elsevier Ireland Ltd. All rights reserved.
S. Carrubba et al. / Neuroscience Letters 469 (2010) 164–168
Fig. 1. Mobile-phone EMFs in the brain. (a) Mobile phones produce high- and low-
frequency EMFs. A battery supplies current to a high-frequency circuit (HFC) that
encodes speech (transducer T1) on the transmitted signal and decodes speech (trans-
ducer T2) in the received signal. The battery current produces magnetic pulses whose
strength depends on D and d, the distances between the base tower and the phone
and between the phone and the ﬁeld sensor, respectively. (b) Portion of the magnetic
pulse train from a Nokia 6085 mobile phone (D = 3 km, d = 10 cm).
3- T, 0.7-ms pulse and an appropriate inter-stimulus period. To
minimize variability in the responses of the subjects, it was desir-
able that the pulse strength be relatively uniform throughout the
brain, irrespective of the motion of the subject’s head relative to
the source of the ﬁeld (actual mobile-phone magnetic ﬁelds do not
Fig. 2. Procedures for determination of effect of mobile-phone pulse on brain activ-
exhibit this property). The standard method for applying uniform
ity. (a) Schematic diagram of the exposure and EEG-detection systems; E (C), latency
magnetic ﬁelds involves the use of multiple current-carrying coils
(control) epoch for assessment of the presence of an evoked potential. A/D, analog-
of magnet wire , but the impedance of typical coil systems 
to-digital conversion. (b) Sequence of experimental conditions; number of 3-s trials
shown in parentheses.
prevents generation of the narrow pulse (Fig. 1b) required for the
present study . Fortunately, however, we recently discovered
that subjects exposed to an EMF stimulus responded not to its mag-
in 90% of the derivations). The signals, V(t), were ampliﬁed (Nihon
netic component but rather to the electric component induced in
Kohden, Irvine, CA, USA), analog ﬁltered to pass 0.3–35 Hz, sampled
the brain as a consequence of the rate of change of the magnetic ﬂux
at 300 Hz, and analyzed ofﬂine.
. External electric ﬁelds that produce the same induced electric
Application of the pulse produced a spike in the EEG of less
ﬁeld as that from a magnetic pulse can conveniently be produced
than 1 ms which, in studies on electrical phantoms of the head, we
by applying a voltage to a pair of parallel metal plates.
established had been generated by Faradaic induction. The spike
An external electric-ﬁeld pulse was generated by applying a 65-
was removed from the signal by deleting the ﬁrst 9 points (30 ms)
V, 0.7-ms pulse (Hewlett Packard 8015A, Palo Alto, CA, USA; Krohn-
from each epoch (see below) prior to analyses of the signal. Tri-
Hite 7500, Avon, MA, USA) to two metal plates (65 cm apart) located
als containing movement or other artifacts (as assessed by visual
on each side of the head (Fig. 2a). It can be shown that this pulse
inspection) were discarded (<5% of the trials). The remaining tri-
produced an unperturbed electric ﬁeld (ﬁeld in the absence of a
als were digitally ﬁltered between 0.5 and 35 Hz, and the latency
subject) of about 100 V/m that, in turn, induced a brain electric ﬁeld
interval 76–471 ms (E epoch) was analyzed for the presence of an
comparable to that induced by the magnetic pulse (Fig. 1b).
evoked potential by comparing each point in the interval with the
The stimulus duration and inter-stimulus period were 0.7 ms
corresponding point in 2.076–2.471 s (the control (C) epoch). All
and 2.9993 s, respectively, resulting in 3-s trials (Fig. 2a). The sub-
results were based on data from at least 50 trials.
jects were exposed in an isolation chamber to reduce the effect of
Following an acclimation period, there were two experimental
random ambient stimuli; all electrical equipment was located out-
periods during which either a ﬁeld or a sham ﬁeld (no-ﬁeld condi-
side the chamber. The absence of both uncontrolled sensory cues
tion) was presented (Fig. 2b); the order of presentation of the ﬁeld
and direct perception of the ﬁeld was veriﬁed by interviewing each
and sham varied randomly from subject to subject.
subject at the end of the experimental session.
The method used to analyze the data was the same as previously
Electroencephalograms (EEGs) were recorded from O1, O2, C3,
described [3,5,6] and will be mentioned here only in summary.
C4, P3, and P4 (International 10–20 system) referenced to linked
Epochs of interest in the EEG were embedded in ﬁve-dimensional
ears, using gold-plated electrodes attached to the scalp with con-
phase space, and the resulting trajectory was mapped to a two-
ductive paste. Electrode impedances (measured before and after
dimensional recurrence plot. The plots were quantiﬁed using two
each experiment) were below 10 k
in all subjects (below 5 k
recurrence variables : (1) percent recurrence (%R), deﬁned as
S. Carrubba et al. / Neuroscience Letters 469 (2010) 164–168
the ratio of the number of points in the plot to the total number
We therefore computed the contributions to the family-wise error
of points in the recurrence matrix; (2) percent determinism (%D),
rate separately for the central, occipital, and parietal electrodes,
deﬁned as the fraction of points in the plot that formed diagonal
using the binomial formula; PFW, the error rate for the occurrence
lines. The process was iterated, yielding the time series, %R(t) and
of EPs in each subject, was determined by the law of compound
%D(t) , which contained the determinism in V(t) but in a more
compact time interval. For example, the analyzed interval in %R(t)
To examine for the presence of linear evoked potentials, the
(176–371 ms) corresponded to 76–471 ms in V(t).
EEG was also evaluated directly (no unfolding in phase space) by
Each of the 60 points at 176–371 ms in %R(t) and %D(t) was com-
time averaging . The estimation of the a posteriori false-positive
pared with the corresponding point in the control epochs (Fig. 2a)
rate and the family-wise error was identical to the analysis used to
using the paired t-test at a pair-wise signiﬁcance level of p < 0.05. As
evaluate the recurrence time series.
previously , when ≥10 tests were pair-wise signiﬁcant at p < 0.05,
We regarded a potential as nonlinear if it was detected by recur-
we regarded the result as demonstrating the presence of an EP.
rence analysis but not by time averaging.
Filtering the EEG to remove alpha frequencies facilitates detec-
Using the nonlinear variable %R(t), brain potentials evoked by
tion of EMF-induced evoked potentials [4–6]; sometimes removal
the simulated mobile-phone pulse were detected in 14 of 20 sub-
of 9–12 Hz but not 8–10 Hz was effective, and sometimes con-
jects (Table 1 , ﬁrst data column). In subject S3, for example, when
versely. Use of %R and %D often gave the same result, but sometimes
the E and C epochs in %R(t) were compared point by point, an EP
only one of them revealed a ﬁeld-induced change in the EEG .
(>10 pair-wise signiﬁcant tests) having the expected latency was
Based on these prior observations, we systematically considered
observed at C3 and P3 (Fig. 3, left panels); sham-ﬁeld exposure (the
all conditions of analysis previously shown capable of revealing an
negative control procedure) yielded no false-positive results (<10
EMF-induced EP . First, we analyzed %R(t) in all 6 electrodes. If
signiﬁcant tests in each derivation (Fig. 3, right panels). A total of
we found an EP (≥10 pair-wise signiﬁcant tests within the expected
120 statistical tests involving the %R(t) time series were performed
latency interval) in at least 3 electrodes, no further analyses were
to evaluate the effect of the mobile-phone pulse (6 derivations × 20
conducted. If fewer than 3 EPs were found, we analyzed %D(t). If a
subjects), resulting in 19 EPs (Table 1, ﬁrst data column).
total of 3 EPs were still not detected, we ﬁltered V(t) prior to calcu-
For subjects who exhibited EPs from fewer than 3 derivations,
lating %R(t) and %D(t) and continued the analysis until either 3 EPs
%D(t) was computed and analyzed; EPs were found in S1, S7, S12,
were detected or all the 6 predetermined conditions (combinations
S13, and S15 that had not been detected with %R(t) (Table 1, sec-
of recurrence variable and ﬁltering conditions) were considered.
ond data column). Filtering the EEG to remove 8–10 Hz or 9–12 Hz
The overall results did not depend on the order; for presentation,
prior to computing %R(t) or %D(t) revealed additional potentials.
we chose the sequence %R(t), %D(t), %R(t) after ﬁltering out 8–10 Hz,
For example, when the 8–10-Hz energy was removed from the EEG
%D(t) after ﬁltering out 8–10 Hz, %R(t) after ﬁltering out 9–12 Hz,
signals prior to computing %R(t), previously undetected potentials
%D(t) after ﬁltering out 9–12 Hz.
were found in 5 subjects (S2, S7, S10, S14, S16). Overall, 90% of
Whenever tests were done to compare evoked potential and
the subjects (18/20) satisﬁed the criterion for the presence of an
control epochs, the conditions being evaluated were also applied
effect (at least 3 pair-wise signiﬁcant tests from any combination
to the sham data (sham evoked potential versus sham control).
We calculated the a posteriori false-positive rate (number of false-
The a posteriori comparison-wise error rate was 21 false-positive
positive effects in the sham data divided by the total number of tests
tests in the sham data/470 total tests = 0.0413. We used this error
performed), and used that error rate to estimate the family-wise
rate to compute PFW, the family-wise error for a decision that a
error (PFW) for the decision that a subject had exhibited ﬁeld-
subject detected the ﬁeld; PFW < 0.05 in 78% of the subjects (14/18),
induced evoked potentials.
and PFW < 0.085 in the remaining 22% of the subjects. There were
Evoked potentials were more likely to be observed at some
no cases of false-positive results (no instances where >3 pair-wise
derivations compared with others, depending on the stimulus
signiﬁcant tests were found in the sham data).
[4,7]. Prior to the study we were unaware of how the probabil-
Neither the latency nor duration of the potentials depended on
ity for detection of pulse-induced EPs depended on derivation.
gender or electrode derivation (Table 2). When the value of the
Evoked potentials in subjects exposed to mobile-phone pulse. Column heads indicate conditions of analysis. Effects in %D(t) are shown in bold. X, evoked potentials not
detected. Dashes indicate conditions not analyzed. PFW, family-wise error for the decision that the subject exhibited evoked potentials. NE, no effect.
%R (8–10 Hz)
%D (8–10 Hz)
%R (9–12 Hz)
%D (9–12 Hz)
O1 O1 O2 P4
C3 C4 C4 P3
C3 C3 P3 P3
C3 C4 P3
C3 C4 P3
O1 O1 O2
O1 C4 C4
O2 O2 P4
O2 P3 P4
O2 C3 C3
O2 O2 C3
C3 P3 P4
O1 O1 P3
O1 O2 O2
C3 P3 P3
C3 C3 C4 C4
C4 C4 P3
C3 C3 P3 P3
P3 P4 P4
S. Carrubba et al. / Neuroscience Letters 469 (2010) 164–168
Latency and duration of evoked potentials stratiﬁed by gender and electrode derivation. Mean ± SD. N, number of evoked potentials (from Table 1).
281 ± 51
267 ± 54
293 ± 55
275 ± 44
251 ± 55
264 ± 29
262 ± 27
254 ± 24
270 ± 28
263 ± 27
recurrence variable for each potential (Table 1) was compared with
iﬁed the absence of electrode signals within the expected latency
its control (expressed as a percent of the average of the sum), the
change was sometimes greater than the control, and sometimes
Filtering within the alpha band was sometimes necessary for
less, which is characteristic behavior for a nonlinear system. The
detection of the EPs (Table 1), as observed previously [4–6]. The
average absolute value of the ﬁeld-induced changes in the recur-
rationale for removing alpha energy was that it did not contribute
rence variables was 32% of the control, indicating a more robust
to the response, and therefore that removal of alpha increased sen-
effect than that typically observed in auditory or visual EPs.
sitivity for detection of the EPs by increasing the signal-to-noise
Using time averaging, EPs were not detected in the EEG from
ratio in the system. The increased sensitivity afforded by alpha ﬁl-
tering might mean that the brain region where the alpha activities
A stimulus equivalent to a pulse produced by mobile phones
originate, usually assumed to be the cerebral cortex , was not
resulted in statistically signiﬁcant effects on brain electrical activ-
crucial in the brain processing that gave rise to the EPs. This sug-
ity in 18 of 20 subjects (Table 1). Several considerations indicated
gestion is consistent with the ﬁnding that the subject did not know
that the effects were true EPs. First, the analysis incorporated pro-
the mobile-phone ﬁeld was present even though the subject’s brain
tection against family-wise error, which obviated an explanation
did. Alternatively, the increased sensitivity afforded by alpha ﬁlter-
based on chance. Second, comparable changes were not observed
ing might be related to differences among the subjects in their level
in the sham data. Third, the changes occurred several hundred mil-
of alertness during the experimental session.
liseconds after the pulse, which was consistent with the inference
The EPs were not detected when the EEGs were analyzed by
that the changes arose from brain processing of afferent signals that
time averaging, indicating that they were nonlinear in origin, as
resulted from transduction of the pulse. The observed latency was
observed previously [4–6]. The ﬁnding that the changes in recur-
inconsistent with the possibility that the changes could have been
rence variables could be either an increase or a decrease further
generated by a ﬁeld-electrode interaction because that process has
conﬁrmed the nonlinearity of the response, because only nonlinear
no latency. Fourth, studies using phantoms of the human head ver-
systems can exhibit such behavior.
We did not address the question of the anatomical location of the
electroreceptor cell. The observed latencies (Table 2) were consis-
tent with a location anywhere in the body. Animal studies, however,
suggested the electroreceptor cell was located in the head, possibly
the cerebellum [9,13]. The actual transduction process may involve
ion channels having ﬁeld-sensitive gating characteristics .
We used a simulated rather than actual mobile-phone pulse
as the stimulus. We accepted this limitation to avoid confounding
effects, and to increase the reproducibility of the effective stimulus.
Nevertheless the pulse was represented verisimilarly enough that
the brain potential it triggered could reasonably be imputed to that
produced by an actual mobile-phone pulse.
In summary, a pulse of the type produced by mobile phones was
transduced by 90% of the subjects studied, as indicated by the occur-
rence of EPs. The implication of our results is that mobile phones
trigger EP at the rate of 217 Hz during ordinary phone use. One
possibility is that chronic administration of the periodic changes in
brain electrical activity causes or promotes disease.
 R.O. Becker, A.A. Marino, Electromagnetism & Life, State University of New York
Press, Albany, 1982.
 A.A. Borbély, R. Huber, T. Graf, B. Fuchs, E. Gallmann, P. Achermann, Pulsed
high-frequency electromagnetic ﬁeld affects human sleep and sleep electroen-
cephalogram, Neurosci. Lett. 275 (1999) 207–210.
 S. Carrubba, C. Frilot, A. Chesson, A. Marino, Detection of nonlinear event-
related potentials, J. Neurosci. Methods 157 (2006) 39–47.
 S. Carrubba, C. Frilot, A.L. Chesson Jr., A.A. Marino, Evidence of a nonlinear
human magnetic sense, Neuroscience 144 (2007) 356–367.
Evoked potentials in subject S3, observed using the recurrence analysis vari-
 S. Carrubba, C. Frilot, A.L. Chesson Jr., A.A. Marino, Nonlinear EEG activation
able %R(t). Potentials from C3 and P3 are respectively shown in (a) and (b). Left panels,
by low-strength low-frequency magnetic ﬁelds, Neurosci. Lett. 417 (2007)
mobile-phone pulse stimulus; right panels, sham stimulus. The curves at the tops
of the panels show the average values of %R in the E and C epochs (Fig. 2a) (N ≥ 50
 S. Carrubba, C. Frilot, A.L. Chesson Jr., C.L. Webber Jr., J.P. Zbilut, A.A. Marino,
trials). The P(t) curves are the probability that the difference between the means of
Magnetosensory evoked potentials: consistent nonlinear phenomena, Neu-
%R in E and C at time t was due to chance. Bar graphs indicate the average value
rosci. Res. 60 (2008) 95–105.
of %R over the latency interval for which P(t) < 0.05 (horizontal line); the standard
 S. Carrubba, C. Frilot II, F.X. Hart, A.L. Chesson Jr., A.A. Marino, The electric ﬁeld
deviations are not resolved at scale shown. The stippled regions show the expected
is a sufﬁcient physical determinant of the human magnetic sense, Int. J. Radiat.
latency intervals (0.176–0.371 s).
Biol. 85 (2009) 622–632.
S. Carrubba et al. / Neuroscience Letters 469 (2010) 164–168
 M. de Tommaso, P. Rossi, R. Falsaperla, V. Francesco, R. Santoro, A. Federici,
 R. Merritt, C. Purcell, G. Stroink, Uniform magnetic ﬁeld produced by 3, 4, and
Mobile phones exposure induces changes of contingent negative variation in
5 square coils, Rev. Sci. Instrum. 54 (1983) 879–882.
humans, Neurosci. Lett. 464 (2009) 79–83.
 C.C. Papageorgiou, E.D. Nanou, V.G. Tsiafakis, E. Kapareliotis, K.A. Kontoangelos,
 C. Frilot II, S. Carrubba, A.A. Marino, Magnetosensory function in rats: localiza-
C.N. Capsalis, A.D. Rabavilas, C.R. Soldatos, Acute mobile phone effects on pre-
tion using positron emission tomography, Synapse 63 (2009) 421–428.
attentive operation, Neurosci. Lett. 397 (2006) 99–103.
 C.D. Hountala, A.E. Maganioti, C.C. Papageorgiou, E.D. Nanou, M.A. Kyprianou,
 G.F. Pedersen, Amplitude modulated RF ﬁelds stemming from a GSM/DCS-1800
V.G. Tsiafakis, A.D. Rabavilas, C.N. Capsalis, The spectral power coherence
phone, Wireless Netw. 3 (1997) 489–498.
of the EEG under different EMF conditions, Neurosci. Lett. 441 (2008)
 D. Regan, Evoked Potentials in Science and Medicine, John Wiley & Sons, Hobo-
ken, NJ, 1985.
 M. Kundi, The controversy about a possible relationship between mobile
 D.S. Ruchkin, Measurement of event-related potentials: signal extraction, in:
phone use and cancer, Environ. Health Perspect. 117 (2009) 315–
T.W. Picton (Ed.), Human Event-related Potentials, Elsevier, New York, 1988.
 J.C. Shaw, The Brain’s Alpha Rhythms and the Mind, Elsevier, Amsterdam, 2003.
 A.A. Marino, S. Carrubba, C.A.L.C. Frilot Jr., Evidence that transduction of elec-
 C.L. Webber Jr., Recurrence Quantiﬁcation Analysis (2007) http://homepages.
tromagnetic ﬁeld is mediated by a force receptor, Neurosci. Lett. 452 (2009)
luc.edu/∼cwebber. Accessed June 30, 2009.
 J.P. Zbilut, C.L. Webber Jr., Recurrence quantiﬁcation analysis, in: M. Akay (Ed.),
 A.A. Marino, E. Nilsen, C. Frilot, Localization of electroreceptive function in
Wiley Encyclopedia of Biomedical Engineering, John Wiley & Sons, Hoboken,
rabbits, Physiol. Behav. 79 (2003) 803–810.
2006, pp. 2979–2986.
- Mobile-phone pulse triggers evoked potentials