Bioelectromagnetics 1 5 105-1 13(1994)

Superimposing Spatially Coherent

Electromagnetic Noise Inhibits Field-
Induced Abnormalities in Developing
Chick Embryos
T.A. Litovitz, C.J. Montrose, P. Doinov, K.M. Brown, and M. Barber
Department of Physics, The Catholic University of America, Washington, D.C.
(T.A.L., C.J.M., P. D., M.B.) and Biological Sciences Department, George Washing-
ton University, Washington, D. C. (K.M. B.)

Living cells exist in an electrically noisy environment. This has led t o the so-called “signal-
to-noise” problem whereby cells are observed to respond to extremely-low-frequency
(ELF) exogenous fields that are several orders of magnitude weaker than local endog-
enous fields associated with thermal fluctuations. T o resolve this dilemma, we propose
that living cells are affected only by electromagnetic fields that are spatially coherent
over their surface. The basic idea is that a significant number of receptors must he si-
multaneously and coherently activated (biological cooperativity) to produce effects on
the biochemical functioning of the cell. However. like all physical detection systems.
cells are subject t o the laws of conventional physics and can he confused by noise. This
suggests that a spatially coherent hut temporally random noise field superimposed on a
coherent ELF Fignal will defeat the mechanism of discrimination against noise. and any
observed field-induced hioeffects would be suppressed. An experimental test o f t h i s idea
was conducted using morphological abnormalities in developing chick embryos caused
by electromagnetic field exposure as the endpoint. At an impressed noise amplitude
comparable to the ELF field svengrh (but roughly one-thousandth of the thermal noise
field), the increased abnormality rate observed with only the ELF field present was re-
duced to a level essentially the same as for the control embryos. OIY& Wiley-Li\\. Inc.

Key words: noise, cooperativity, signal-to-noise, embryo development, abnormalities

INTRODUCTION
The association of biological effects at the cellular level with exposure to weak
ELF electromagnetic fields (EMFs) has remained a controversial subject despite
nearly a decade of such reports. Theoretical arguments based on signal-to-noise
considerations are in large measure responsible for the skepticism [Weaver and
Astumian, 1990; Adair, 19911. The dilemma arises because cells, existing in an

Received for review October 12, 1992: revision received J u l y 6 . 1993.

Address reprint requests to C.J. Montrose, Department of Physics, Catholic University of America.
Washington, DC 20064

0 1994 Wiley-Liss, Inc.

106 Litovitz et al.

electrically noisy environment, respond to external EMFs some 100 to 1000 times
weaker than local noise fields resulting from the thermally driven movements of
ions in the vicinity of cells. In this context Adair [ 19911 has concluded that “...it
does not appear to be possible for weak external ELF electromagnetic fields to affect
biological processes significantly at the cell level.” Yet the data unambiguously
demonstrate that they do [see Byus et al., 1987; Rubin et al., 1989; and Litovitz et
al., 199 I , for example]. How can the externally impressed fields possibly influence
cell behavior when cells have evolved in such a way as to function normally in the
presence of the much larger local noise fields?
To address this question, Weaver and Asturnian I19901 have proposed that cells
integrate the electromagnetic signals, effectively narrowing their acceptance band-
width, and thus averaging out the thermal noise. However, their calculations yield
averaging times much longer than exposure intervals observed to produce bioeffects.
For cells of 20 pm diameter, they estimate that to achieve the required signal-to-
noise improvement at 100 Hz requires averaging over 4.3 x 10‘ s (about 12 h ) yet
there are many cases of bioeffects being observed with exposure intervals on the
order of (or even less than) 1 h. It is clear that a simple time averaging mechanism
cannot explain the data.
The essence of this paper is the formulation of a hypothesis that confronts the
problem of how cells discriminate against the large endogenous thermal noise fields,
enabling them to respond to the weak exogenous fields. We propose that living cells
discriminate against thermal noise fields by recognizing them as .spatially itzcohewnt,
i.e., uncorrelated at d erent locations on the cell membrane. Adey [ 197Sa,b, 198 I ,
19881 and Blackrnan [ 1988, 19891 have conjectured that coherence plays some role
in the interaction of electromagnetic fields with cells. The specific character of the
hypothesis developed here is shaped by an experimental study designed to eluci-
date the effect of electromagnetic noise on biological response.

EXPERIMENTAL METHODS AND RESULTS

Fertilized White Leghorn eggs (Truslow Farms, Chestertown, MD) were used
within 24 h of their being received. The apparatus and techniques followed the “Project
Henhouse” protocols [Berman et al., 19901 with several exceptions. Six VWR water-
jacketed incubators were employed in the study. Their temperature control systems
were modified to avoid the possible introduction of unwanted stray magnetic fields:
the coiled heater elements located below the water jackets at the bottom of the
incubators were disconnected; the water was heated externally using RTE Model
1 10 FRC Bath/Circulators. Incubators were used interchangeably and randomly for
the various exposure configurations.
Electromagnetic fields were produced by passing current through Helmholtz
coils wound and connected as described by Berman et al. [ 19901. Three exposure
conditions were examined: 1 ) control or “sham” exposed, 2) “pulse” exposed, and
3) “pulse-plus-noise” exposed. In the first, no current was allowed to flow in the
Helmholtz coils. I n the “pulse” exposure, the pulsed EMF signal prescribed by
“Henhouse” (100 Hz, 500 p s pulse duration, 2 ps rise and decay times and, 1 pT
nominal peak strength) was imposed on the incubating embryos. The signal in the
“pulse-plus-noise” condition consisted of the same pulsed EMF as i n (2) on which
was superimposed a noise EMF-a coizfusion field.
 Spatial Coherence and EM Fields 107

The pulsed signal was produced using a Wavetek Model 801 SO Mhz pulse
generator. The noise was generated by a General Radio Model 13090-B random
noise generator. The output of the noise generator was filtered with a Krohn-Hite
Model 3323 bandpass filter and used to drive a Realistic Model MPA-90 audio
amplifier. The filter was set for a nominal 30 to 100 H z bandpass. The amplitude
of the noise was adjusted to produce an rms magnetic field strength of I pT. I n Figure
1 the Fourier spectra of the pulse and the noise are presented. The spectrum of t h e
pulse contains significant contributions out to nearly 10 kHz, as is to be expected
of a nearly “square” pulse; only the first few harmonics are shown here to provide
a comparison with the spectrum of the noise. These frequency spectra were mea-
sured with an Ono Sokki Model CF-910 dual channel FFT spectrum analyzer.
The eggs were exposed during the first 48 h of incubation, with the control
sample being incubated simultaneously. After the 48 h incubation, embryos were
removed from their shells and examined by the procedures used in “Henhouse.”
This evaluation was performed under blind conditions. Eggs were first examined
for fertility and the embryos were determined to be live or dead. Live embryos were
examined for abnormal gross morphologies. Embryos were considered t o be ab-
normal if they differed markedly from the Hamburger and Hamilton [ 195 I ] 48 h
developmental stages. Malformations were classified as cephalic nervous system,
truncal neural tube, heart, blood vessels, and somites. Of these only the first two
occurred with any significant frequency. Indeed, over 90% of those embryos (both
control and exposed samples) that were characterized as abnormal exhibited truncal
neural tube malformations (as well as possibly other abnormalities).
The experiments were carried out in two groups of replicates (two runs), the
second run being done roughly one year after the first, Each replicate consisted of
10,20, or 30 eggs i n each sample group. Summaries of the results are presented in
Table 1 and displayed i n Figure 2. It seems apparent that there is an increase i n the
fraction of abnormally developed embryos when the eggs are subjected to a pulsed
external EMF. It also seems evident that this increase is greatly reduced when a

PULSE

1 00 200 300 400 fi

c

FREQUENCY/ Hz
Fig. 1 . Frequency spectra of the pulsed signal and of the superimposed noise field.
108 Litovitz et al.

TABLE 1. Results for Embryos Subjected to Pulse EMFs and Pulse

EMFs With Noise EMFs.

Exposure conditions Live Abnormal Percent

Run I
Sham exposed (control) 255 I6 6.3
Pulse EMF exposed I52 29 19.1
Pulse E M F + n o i w exposed II0 8 7.3
Run 2
Sham exposed (control) 206 6 2.9
I’tilse EMF: exposed 203 22 10.8
Pulsc EMF + noise exposed 181 6 3.3

confusion field is present. The data were not pooled because during the nearly 1
year interval between the two runs, the supplier changed the laying flock. From the
difference i n the abnormal development frequencies in the sham-exposed control
samples for the two runs, about a factor of two, i t is clear that the two flocks rep-
resent different experimental models.
Analysis of the Data
A detailed chronicle of the experiments is given in Table 2. The record is a
rather consistent one: no single replicate gave results that stand out as so unusual
that they should be regarded as “tainted.” Indeed there is quite a remarkable consis-

G
E
251 PEMFEXPOSED
PEMF + NOISE EXP

J1
15

I-
z
W
0
a
W
e
5

0
1LL
RUN #1 RUN #2
Fig. 2. The abnormal development ratc for chick embryos following a 48 h incubation period during
which the embryos were exposcd to pulsed electromagnetic fields. T h e abnormality rate\ for sham-
exposed (control) embryos and for embryos exposed to ;I field i n which a noise signal was siiperim-
posed on the ternporally cohcrcnt signal are also given.
 Spatial Coherence a n d EM Fields 109

TABLE 2. Chronological Summarv of the Exaerimental Realicates

Replicate number Control sample Pulse exposed Pulse + noise exposed

and date* Live Abn Remarks Live Abn Remarks Live Abn Remarks
1 24 Nov 91 19h 2 10 2 10 2
2 27 Nov 91 26 2 9 2 1De 16 3 2 ~ e I ,'rw
3 29 Nov 91 27 2 2De 10 2 19 I I De
4 I Dee91 29 1 I9 2 I De 10 I
5 4 Dee91 29 2 I De I8 3 1 De 10 0
6 6 Dee 91 28 3 2O 5 8 0 2De
7 8 Dec91 28 1 I De, 1 Tw 18 3 I De, I Tw 10 I
8 lIDec91 30 2 10 2 17 0 2De
9 I3 Dec 91 30 I 29 7 1 De None'
10 15Dec91 9 0 9 1 I0 0
RUN 1 totals 255 16 (6.3%) 152 29 (19.1%) 110 8 (7.3%)

11 9 Dec 92 19 0 20 2 17 I 1 De. I Tw
12 I 1 Dec 92 18 I 2De 17 2 3 De 18 0
13 I3 D e e 9 2 18 0 1 De 19 2 20 I
14 16Dec92 20 I 19 2 I De 18 0
15 18DecY2 19 0 18 2 1 De 20 I
16 20 Dec Y2 20 1 19 2 19 0 I De
17 23 D e e 9 2 20 2 19 0 I De 16 2
18 6 Jan 93 15 0 3De 19 I 19 0
19 9 Jan 93 20 0 17 3 2 D e , I Tw None''
20 12 Jan 93 20 0 19 2 I De 18 0 I De
21 18Jan93 17 1 3De 17 4 2De 16 I 3 D e , I Tw
RUN 2 totals 206 6 (2.9%) 203 22 (10.8%) 181 6(3.3%)
"his is the date on which the 48 h incubation interval was completed.
hUnless r;oted (De, dead; Tw, twin embryo). missing embryos were broken during analysis.
'The noise Field was inadvertently not hwitched o n : thus there is no "pulse + noise exposed" sample
for this replicate.
T h e noise field was switched off alter 30 h incubation; there is no "pulw + noise exposed" sample
for this replicate.

tency to the results with the pulsed-EMF (PEMF) exposed abnormality rate ex-
ceeding the sham-exposed rate in 20 of the 2 1 replicates and exceeding the pulsed-
EMF-plus-noise (PEMF + N) exposed abnormality rate in 17 of 19 replicates (with
one being tied). The data lend themselves to various statistical analyses so that
the tentative conclusions of the preceding paragraph can be subjected to quanti-
tative scrutiny.
The overall null hypothesis that all three treatments, sham-exposure, PEMF-
exposure, and PEMF + N-exposure, yield the same rate of abnormal development
was tested using the non-parametric Friedman rank test [Lehmann, 19751. This
procedure ranks the abnormality rate for each of the three treatments (from low
to high) within each replicate. For run 1 , the hypothesis can be rejected with a P
value < .001 and for run 2 with P < .004. The Friedman test can also be used to
study whether or not differential treatment effects exist [Neder et a]., 198.51. The
results are that for both runs the differences between the PEMF-exposed and the
sham-exposed samples are significant at P < .OS. Similarly, the differences between
the PEMF-exposed and the PEMF + N-exposed samples are significant at P < .05.
110 Litovitz et al.

The differences between the sham-exposed and PEMF + N-exposed samples are
not statistically significant. Paired t tests were also used to evaluate the differences.
For run I the abnormal development rates in the PEMF-exposed samples exceeds
those for the sham-exposed and PEMF + N-exposed samples at the P < ,001 level;
for run 2 the same comparisons yield P < .O 1. While these tests are not indepen-
dent, they do serve to bear out the results obtained using the Friedman test, cor-
rected for multiple comparisons.
The data demonstrate that imposing the pulsed EMF on the developing em-
bryos brings about an increase in the frequency of abnormal development (at least
at the 48 h point in the incubation process). Moreover, they show that superimpos-
ing a temporally incoherent noise field on an electromagnetic field that by itself
induces embryo abnormalities, greatly curtails, or essentially eliminates its capacity
to induce adverse developmental effects.
The conclusions that can be drawn are the following: 1 ) Impressing pulsed
electromagnetic fields with amplitudes on the order of a microTesla significantly
increases the fraction of embryos exhibiting developmental abnormalities after a
48 h incubation period; 2) Superimposing a confusion field (a spatially coherent,
temporally incoherent field) of rms amplitude comparable to the pulsed field sup-
presses the effect of the latter on the abnormal development frequency: 3) For the
noise fields considered here, the abnormality rates with no field imposed and with
both the temporally coherent and noise fields are nearly the same, that is, the con-
fusion field masks the effect of the imposed EMF even though i t contains signifi-
cant energy only near the fundamental frequency of the pulsed field.

DISCUSSION
The results presented above establish that the presence of noise, comparable
in magnitude to the EMF signal, can nullify the bioeffects of that signal. In so doing,
they confer on those effects the kind of respectability that derives from their being
in harmony with fundamental scientific principles. I n addition, they focus atten-
tion on the issue of how the externally imposed noise, the confusion field, differs
from the thermal noise that is always present and several orders of magnitude larger.
A n obvious, and we believe the crucial, difference has to do with the spatial varia-
tion of thermally generated fields compared with that of external fields. Because
the Debye screening length (roughly the range over which a given ion is not shielded
from other ions) i n the extracellular fluid is about 1 nm, localized charge density
fluctuations produce thermal noise fields that are spatially uncorrelated. Conversely,
externally impressed fields (including externally imposed noise) are spatially co-
herent over distances substantially larger than cellular dimensions.
We assume that the vital electromagnetic field-cell interaction occurs i n the
immediate vicinity of the cell surface; this interaction triggers a transmission to
the cell interior where some modification of the biochemical reaction pathway is
effected. A plausible supposition is that the field affects the binding of ligands to
the roughly 10000 receptor proteins (sensors) that are integral to the cell membrane
(we envision :he effect as arising from charge redistribution caused by the local
induced electric field, although this specific mechanism is not crucial to the pic-
ture). Ligand binding causes the production of effector molecules (second messengers)
within the cell; the net effect is the transducing of the extracellular signal into an
 Spatial Coherence and EM Fields 111

intracellular one. Cooperativity is required in such processes in that “more than one
intracellular effector molecule must bind to some target niacromolecule in order
to induce a response” [Alberts et al., 19831.
The average spacing between receptors for a particular hormone at the
cell membrane is on the order of 100 nm. This is estimated by considering
that there are roughly 10000 such receptors per cell [Darnell et al., 19901 and
by using 10 pm as t h e approximate cell diameter. Hence, t h e idea that a
multiplicity of cellular receptors must be simultaneously activated i f modi-
fied cell functioning is to result requires that an essentially uniform stimu-
lating field be presented to the receptors for some required “coherence interval.”
Litovitz et al. 1199 I ] have already demonstrated that temporal coherence for
times on the order of 5-10 s is necessary if EMF-produced bioeffects are to
be observed. The operative mechanism i n discriminating against fields from
thermal fluctuations is analogous to a coincidence detection scheme, a sig-
nificant number of receptors at the cell membrane must be simultaneously and
coherently activated to produce an effect on the biochemical functioning of
the cell. Externally imposed electromagnetic fields are thus able to affect cell
functioning because they are spatially coherent at various receptor sites i n the
membrane; consequently, they produce the required number of effector rnole-
cules to initiate a cytoplasmic response. I n contrast, t h e uncorrelated thermal
noise fields can produce no such synchronous effects. This biological coin-
cidence detection scheme allows the cell to be exquisitely sensitive to very
weak sputially correlared electromagnetic fields while discriminating against
the much stronger but spatially random (on the relevant distance scale) ther-
mal noise fields.
This discrimination mechanism cannot operate if the noise signal is itself
spatially correlated. The spatially coherent but temporally random confusion fields
are thus able to frustrate the cellular detection scheme and eliminate any electro-
magnetic field-induced bioeffects. The experimental result given above. nullifying
the EMF-caused developmental abnormalities in chick embryos by the superposi-
tion of a confusion field, is completely consistent with this interpretation. Preliminary
work by Farrell et al. [ 19931on field-modified ornithine decarboxylase (ODC) activity
in developing embryos, and the abolition of this effect with a confusion field fur-
ther supports the conclusion. Mullins et al., [ 19931 have reported a similar elimi-
nation of sinusoidal EMF-enhanced ODC activity in murine fibroblast cell cultures.
Comparison of the results reported here with those of Mullins et al. [ 19931 is
informative. They find that imposition of a weak sinusoidal 60 Hz EMF induces
roughly a two-fold increase in ODC activity in L929 cells. Similar to the result
reported here they find that a 30-100 H z bandwidth confusion field essentially
eliminates this enhancement. Note that in their case the Fourier components of the
coherent signal (the 60 Hz sinusoid) lie completely within the noise spectrum. In
contrast for the pulsed EMF signal, most of the energy lies outside of the spectral
range of the noise. A straightforward interpretation of these results is that living
cells respond to coherent signals as a broadband detector, not distinguishing be-
tween different frequencies over a rather wide range. This is supported by the work
of Juutilainen and Saali [ 19861 and Juutilainen et al. 119871 who have shown that
developmental abnormalities in chick embryos result from exposure to sinusoidal
fields stronger than about I pT at frequencies ranging from 50-100 kHz.
112 Litovitz et al.

SUMMARY
We have hypothesized that living cells discriminate against random thermally
generated electromagnetic fields by a cellular coincidence detection mechanism that
responds only to fields that are spatially coherent over the cell membrane. We have
rationalized this hypothesis by ascribing i t to biological cooperativity i n which a
significant number of membrane-integral receptor proteins must be simultaneously
activated in order to cause signal transduction to the cytoplasm. Because all physical
detection systems are ultimately limited by noise constraints, one test of this hy-
pothesis involves the superposition of spatially coherent temporally random noise
on a pulsed electromagnetic field that has been demonstrated to cause a bioeffect.
We have carried out such an experiment using developmental abnormalities
in incubating chick embryos as the endpoint. Initially, i t was established that weak
pulsed EMFs (100 HI. rep rate) caused an increase in the embryo abnormality rate
following exposure during the first 48 h of incubation. It was found that superimposing
externally generated noise fields of roughly the same strength (although several orders
of magnitude smaller than the endogenous thermal noise field) on the pulsed fields
essentially eliminated the field-induced abnormalities. The results with the exog-
enous noise fields were statistically indistinguishable from those for the sham-exposed
embryos.
This finding supports our proposal that it is their sensitivity o n l y to s p r f i u I f y
coherent EMFs that enables living cells to respond to weak exogenous fields while
remaining unaffected by the relatively large (but spatially incoherent) endogenous
noise fields that are always present. The idea that cells distinguish between exog-
enous and endogenous fields by recognizing the latter’s spatial incoherence offers
a viable explanation of the signal-to-noise dilemma. Far from being “magically”
exempt from signal-to-noise considerations, cellular detection mechanisms are subject
to the same laws of nature as all physical systems. The masking that is provided
by a spatially coherent confusion field could prove to be a basis for protecting humans
against possible adverse health risks associated with environmental electromagnetic
fields.

ACKNOWLEDGMENTS
The authors thank Dr. Sam Greenhouse of the George Washington University
Biostatistics Center for his help with the statistical analysis of the data. They also
acknowledge the Walter Reed Army Institute of Research for support of this work
under contract DAMD 17-86-C-6260.

REFERENCES
Adair RK ( I Y Y I ] : Constraints on hiological effects of weak extremely-low-frequency electromagnetic
fields. Phys Rev A 41:1039-1048.
Adey WR (1975a): Introduction: Effects of electromagnetic radiation on the nervous system. Ann NY
Acad Sci 247: 15-20.
Atley WR ( I97Sb): Evidence tor cooperative mechanisms in the ausccptibility of cerebral tissue t o
environmental and intrinsic electric fields. In Schmitt FO. Schneider DM, Crothers DM (eds):
“Functional Linkage in Biomolecular Systems.” New York: Raven Press, pp 325-342.
Adey WR ( I98 I ) : Tissue interaction with non-ionizing electromagnetic fields. Physiol Rev 61 :405-5 14.
 Spatial Coherence and EM Fields 113

Adey W R (1988): Physiological signalling across cell membranes and cooperative influences of ex-
tremely low frequency electromagnetic fields. In Frohlich H (ed): “Biological Coherence and
Response to External Stimuli.” Heidelberg, Germany: Springer, pp 148- 170.
Alberts B, Bray D, Lewis J. Raff M. Roberts K, Watson JI) (1983): “Molecular Biology o f t h e Cell.”
New York: Garland Publishing, Inc., p 752.
Berman E, Chacon L, House D, Koch BA, Koch WE. Leal I , Lfivtrup S , Mantiply E, Martin AH, Martucci
GI, Mild KH, Monahan JC, SandstrBm M, Shamsaifar K. Tell R, Trillo MA, Ubeda A. Wagner
P (1990): T h e effects of ii piilsed magnetic field on chick embryos. Bioelectromagnetics
1 1:169-187.
Blackman C F (1988): In!luence of electromagnetic fields on the efflux of calcium ions from brain tisstie
in vitro: Three models consistent with the frequency response u p to 510 Hz. Bioelecti-ornag-
netics 9:21S-227.
Blackman C F (1989): Multiple power-density windows and their possible origin. Bioelectromagnetics
10:l 15-128.
Byus CV. Pieper SE, Adey WR (1987): The effects of low-energy 60-Hr environmental t’ields upon
the growth-related enzyme ornithine decarboxylase. Carcinogenesis (Lond.) 8: 1385-138’).
Darnell J , Lodish H, Baltimore D (1990): “Molecular Cell Biology.” 2nd Ed. New York: W.H. Frec-
man and Co., p 7 19.
Farrell JM, Barber M, Doinov P, Krause D, Litovitz TA (1993): Effect of 60 Hz sinusoidal magnetic
fields on ornithine decarboxylase activity in developing chick embryos. In Blank M (ed): “Elec-
tricity and Magnetism in Biology and Medicine.” San Francisco, CA: San Francisco Press, Inc.
Hamburger V, Hamilton HL (1951 ): A series of normal stages in the dcvclopment o f the chick e m -
bryo. J Morphol 88:49-92.
Juutilainen J, Liiiira E, Saali K ( 1987): Relationship between field strength and abnormal developments
in chick embryos exposed to SO Hz magnetic fields. J Radiat Biol 52:787-793.
Juutilainen J, Saali K (1986): Development of chick embryos in I Hz to I Khz magnetic fields. Radiat
Environ Biophys 25: 135-140.
Lehmann E (1975): “Non-Parametric Statistics.” San Francisco. CA: Holden-Day Puhlishcrs,
pp 262-264.
Litovitr TA, Mullins JM, Krause D ( I99 I ): Effect of coherence time of the applied magnetic field on
ornithine decarboxylase activity. Biochem Biophys Res Comni I78:862-865.
Mullins JM. Krause D, Litovitz TA ( 1993): Simultaneous application o f a spatially coherent noise field
blocks the response of cell cultures to a 60 Hz electromagnetic field. In Blank M ( e d ) : “Elcc-
tricity and Magnetism in Biology and Medicine” San Francisco. CA: San Francisco Press, Inc.
Neder J. Wasserman W , Kutner MH (1985): “Applied Linear Statistical Models.” 2nd Ed. Homewood.
IL: Irwin, pp 950-Y52.
Rubin CT, McLeod KJ, Lanyon LE ( 1989) Prevention of osteoporosis by pulsed electromagnetic fields.
J Bone Joint Surg 71A:41 1-417.
Weaver JC, Astumian RD (1990): The response of living cells to very weak electric fields: The ther-
mal noise limit. Science 247:459-462.