Experimental History and Philosophy of Science

Coulomb’s Electric Torsion

Balance Experiments

exphps.org © 2009
 1

Coulomb’s Electric Torsion Balance

Experiments of 1785

Elay Shech
Department of History and Philosophy of Science
University of Pittsburgh
1017 Cathedral of Learning
Pittsburgh, PA 15260
USA
Email: [email protected]

Abstract
There has been a debate, among scholars, over whether Charles Augustin Coulomb
attained the results he presented in his 1785 and 1787 memoirs to the Paris Academy of
Sciences experimentally or theoretically. In an effort to contribute to this debate, we
study Coulomb’s famous electric torsion balance experiment of 1785 through a
historically more accurate replication than hitherto attempted, as well as through analysis
of relevant texts. In contrast to recent claims, we will show that (1) it has so far proved
impossible to obtain the same results reported by Coulomb in his paper of 1785, (2)
Coulomb’s published results are most likely atypical, and (3) electric torsion balance
experiments degenerate quickly when parameters are gradually altered.

Acknowledgments
We wish to thank Jed Buchwald for generously making available to us the glass pieces of a torsion balance
identical to that used by Alberto Martínez. Thanks also to Eric Hatleback and Paolo Palmieri for their
important roles in replicating the electric torsion balance, discussing results and recording trials, as well as
for the invaluable comments on earlier drafts of this paper. Thank you also to Katie Moriarti for her help in
translating sections of Coulomb’s memoirs. A part of this paper was presented at the 2008 History of
Science Society Annual Meeting in Pittsburgh, PA. We thank the participants of that meeting, especially
Alberto Martínez, for helpful discussion and comments. This research was supported by the Wesley C.
Salmon fund, University of Pittsburgh and the ‘HPS Lab’ of the History and Philosophy of Science
department in University of Pittsburgh.
 2

CONTENTS

1. Introduction: The Controversy Over the Experimental Justification of the

Fundamental Law of Electrostatics
2. Coulomb’s 1785 Memoir and the Electric Torsion Balance
3. Previous Replications: Martínez and Heering
4. Our Replication of Coulomb’s Experiment
5. Coulomb and the Electric Polarization of the Needle
6. Conclusions
 3

A NOTE ON THE VIDEO MATERIAL SUPPORTING THIS PAPER

(by Paolo Palmieri)

The majority of our experiments were video taped with a high quality digital video camera. Some
can be accessed by logging on to the HPSLab website, at: http://www.exphps.org/. The videos
that we present here (titles in italics) are arranged in a chronological order, which reflects well
our experimental labors, our ups and downs, as we witnessed the phenomena and struggled to
come to terms with them.

1. 08/08/2008 Loss of charge humidity of needle. From these trials, it became clear that the
needle was leaking a lot of the balls’ charge, that this leaking is due to the needle leaking
charge within the atmosphere of the base cylinder and/or the mechanical connection with
the silver wire and ultimately ground. The Spanish-wax needle had been made in April
2008, our best up to then, and it was used here after about three months. We hypothesized
that Spanish-wax being a natural material, unlike synthesized plastic, the needle might
have absorbed humidity, thus becoming useless.
2. 08/11/2008, morning. Exp 1 light needle, Exp 2 light needle. Two regular Coulomb
experiments, i.e., which mimic the experimental sequence of operations recounted by
Coulomb, with a light Spanish-wax needle sagging a lot but performing decently well.
3. 08-11-2008, afternoon. 08-11-2008 exp 3. These trials showed a new, perplexing
phenomenon: might the ball’s charge be spreading along the needle? If so, this new
phenomenon, we speculated, creates a net loss of force. That this explanation is plausible
is suggested by the fact that the total charge remained stable between the measurements
yet the exponent turned out to be well below the theoretical expectation of 2. No
definitive conclusions were reached, though. The Spanish-wax needle was subsequently
discarded.
4. 08/12/2008 Upper range 1. One out of five trials in what we call “the upper range”, that
is, with large angular separation between the two pith balls due to a large amount of
charge being imparted at the beginning of the experiment (note that it is not possible to
control how much charge is being imparted when touching the pith ball). The same
sagging needle as the day before was used.
5. 09/11-2008 Induction. This video shows the polarization phenomenon in a Spanish-wax
needle. Titles in the video will help the viewer follow the experiment.
 4

6. 25/09/2008 Plastic-straw acts like needle. A red-plastic needle with no pith ball attached
behaves more or less like one with pith ball. Titles in the video will help the viewer
follow the experiment.
7. 10/01/2008 Plastic-straw Needle Experiment. Same red-plastic needle with pith ball. A
regular, in the sense explained above, Coulomb experiment is showed.
8. 10/01/2008 Shellac Needle Experiment. A regular, in the sense explained above,
Coulomb experiment is showed with one of our best shellac needles.
9. 10/29/2008 White straw needle regular coulomb. A regular Coulomb experiment, in the
sense explained above, is showed with a white-plastic needle which replicates well the
range of results obtained by Alberto Martínez (see paper).
 5

Introduction: The Controversy Over the Experimental Justification of the

Fundamental Law of Electrostatics
In 1785 Charles Augustin Coulomb (1736-1806) presented to the Paris Académie Royale
des Sciences his first memoir on electricity and magnetism 1 which, along with his
subsequent memoirs on the same subject 2 , would lead physicists to name the fundamental
law of electrostatics as ‘Coulomb’s law.’ Anyone who is familiar with basic high school
physics knows of Coulomb’s law. It is a law that describes the force between two
electrically charged point-bodies as directly proportional to the product of the charges on
the point-bodies, and inversely proportional to the square of the distance between the
point-bodies. In this sense, it is analogous to Newton’s law of gravitation. Indeed,
Coulomb is best remembered for his work on electricity and magnetism, which is
considered by many his most important research. 3 Kragelsky and Schedrov, authors of
the locus classicus book on the history of friction, state that ‘Coulomb’s contributions to
the science of friction were exceptionally great. Without exaggeration, one can say that
he created this science.’ 4 Coulomb’s 1773 memoir on statics 5 , for which Coulomb won
an Académie prize competition, is described as ‘perhaps the finest engineering memoir
delivered at the Academy during Coulomb’s lifetime’ 6 ‘which laid the foundations of the
modern science of soil mechanics…’ 7 .
Yet, in recent times, Coulomb’s status as an experimenter and scientist has been
dramatically challenged. More specifically, the results announced in Coulomb’s 1785
memoir, and consequently the experimental data Coulomb attained with his torsion
balance in the following years, have come into question.

1
Coulomb (1884 [1785]).
See also http://www.academie-sciences.fr/Membres/in_memoriam/Coulomb/Coulomb_oeuvre.htm , for a
digital edition of Coulomb’s original memoirs.
2
Coulomb (1785, 1786, 1787, 1788, 1789), for a total of seven memoirs, three of which in 1875.
3
Falconer (2004), p. 109, is an example of such an opinion.
4
Kragelsky and Schedrov (1956), p. 52 (as quoted in Gillmor (1971), p. 136).
5
Coulomb (1773), republished in Heyman (1972).
6
Gillmor (1971), p. 81, original emphasis.
7
Heyman (1972), p. vii.
 6

In 1992, Peter Heering, noticing that Coulomb’s law was strongly contested in
parts of Europe (especially in Germany) after its publication, raised the question of
whether Coulomb attained the results described in his 1785 memoir experimentally or
from theoretical considerations. 8 By a close reading of Coulomb’s 1785 memoir and by
replicating the electric torsion balance, Heering provided both experimental and textual
evidence against Coulomb’s attaining his 1785 results experimentally. 9 Accordingly,
Heering claimed that ‘Coulomb did not find the inverse square law by doubtful
measurements from his torsion balance experiments but by theoretical considerations’ 10 .
John L. Heilbronn admits that it ‘…appears from Heering’s resourceful labor that
Coulomb either faked his number completely or obtained them under experimental
conditions materially different from those he reported.’11 Heering’s evidence included his
inability to successfully replicate Coulomb’s experiment as is described in the 1785
memoir, the identification of ‘unavoidable’ sources of error, and more recently 12 , his
failure to attain part of the results described in Coulomb’s 1784 wire-torsion
experiments 13 .
Heering’s work gave rise to plethora of claims, including his own, about the
experimental methods of Coulomb and his contemporaries as well as the practices of
scientific reporting of Coulomb’s time. 14 For example, Isobel Falconer has suggested that
it is possible that in his 1785 memoir Coulomb was attempting to establish priority over
the torsion balance as well as promote it as a valid experimental instrument rather then

8
Heering (1992).
9
See Heering (1992), (1994), and (2006). Heering (2006) concentrates on Coulomb’s 1784 memoir on
torsion (see Coulomb (1784)).
10
Heering (1992), p. 993.
11
Heilbron (1994), p. 151.
12
See Heering (2006).
13
Coulomb (1784).
14
See, among others, Blondel (1994), Falconer (2004), Heering (1992), (1994), and (2006), Heilbron
(1994), Licoppe (1994), Pestre (1994). Martínez (2006), p. 534, mentions insights, with regards to
Coulomb’s experimental methods, by Jed Buchwald and Maria Trumpler. Cf. also Dickman (1993), p. 501.
See Dickman (1993), for an overview of the controversy prior to Martínez (2006).
 7

demonstrate the fundamental law of electrostatics. 15 Further, Heering, discussing his

research concerning Coulomb’s 1784 memoir on torsion, as well the 1785 and 1787
memoirs on electricity and magnetism, goes so far as to suggest that the memoirs
represent a ‘rejection of politically dangerous theories’ through an ‘affirmation of the
fruitfulness of the new style of scientific experimentation’ as portrayed in Coulomb’s
1784, 1785 and 1787 memoirs 16 .
However, more recently, Alberto A. Martínez succeeded in replicating Coulomb’s
1785 results, thus bringing forth strong evidence that Coulomb’s results had been attained
experimentally. In consequence, in direct disagreement with prior scholars such as
Heering, Martínez has argued that ‘Coulomb obtained his numbers from experiment. His
results were not unusual, they were almost certainly typical.’ 17

In this paper, we engage this fascinating controversy over Coulomb’s 1785

memoir. Our goal is to contribute to it primarily through a historically more accurate
replication of the electric torsion balance described in Coulomb’s 1785 memoir. Between
the Spring of 2008 and Summer of 2009, we built a Coulomb torsion balance and
replicated Coulomb’s experiments. Our results, in part, tend to confirm those of Martínez,
but diverge from his in particularly important respects that will be described later on in
the paper (cf. sections 3 & 4). In short, we found that although it is possible to replicate
the torsion balance experiment, as Coulomb describes it, in an approximately successful
manner, historically exact results, i.e., very close to those published originally in
Coulomb’s memoir, and more recently by Martínez, are impossible to attain because of a
source of error that is most likely unavoidable and not identified in previous historical
accounts of the experiment 18 . A host of questions follow:
ƒ Was Coulomb aware of such sources of error?

15
Falconer (2004), pp. 111-112.
16
Heering (2006), pp. 61-62.
17
Martínez (2006), p. 547.
18
We gather from personal conversation that Martínez (2006) noticed these sources. We will elaborate on
this issue in section 3.
 8

ƒ If so did he account for them and if so, by what means did he account for such
error sources?
ƒ And more specifically, what exactly is the data set described by Coulomb in his
1785 memoir supposed to represent?
ƒ Is it a typical data set that anyone replicating the experiment should come across?
ƒ Is it an atypical data set that beautifully demonstrates the inverse square aspect of
the fundamental law of electrostatics and because of this was chosen by
Coulomb?
ƒ Does the data set represent an averaging of a larger data set?
This paper will, we hope, make some progress towards answering these questions 19 .

Here is an outline of the paper.

Section 2 presents Coulomb’s 1785 memoir along with his description of the
apparatus and the experiment. Section 3 discusses prior replications, concentrating on
Martínez’s replication, in order to motivate our current project. We first discuss the
manners by which recent replications diverged from Coulomb’s 1785 description, and
then go on to question the historical validity of Martínez’s replication. Section 4 will
present our replication of the torsion balance along with results, emphasizing those
aspects that diverge from prior replications. In section 5, we make some preliminary steps
in answering questions with regards to Coulomb’s awareness of error sources, and the
experimental methods he used for dealing with error sources. Section 6 ends the paper
with a summary of our conclusions.

19
Whenever we translate Coulomb without identifying the translator, the translation is our own with
extensive help from Katie Moriarty. The work presented here is very much a continuation of the work done
by others, above all, by Martínez. In order to avoid repetition, we will direct the reader to prior work done
on the torsion balance whenever it is appropriate, and concentrate on either our original contribution or
material that, though already covered by scholars, is essential for understanding and motivating our project.
 9

2. Coulomb’s 1785 Memoir and the Electric Torsion Balance

As is well known, the so-called fundamental law of electrostatics (Coulomb’s law)
describes the force between two point-bodies electrically charged as inversely
proportional to the square of the distances between the point-bodies:

Fe = C/d2 , (1)

where Fe is the electric force, d is the distance between the point-bodies and C is a
constant that depends upon the amount of charge on the point-bodies, and other constants
of nature. In order to demonstrate that the electric force behaves as an inverse square law,
Coulomb begins his 1785 memoir by recalling a result that he attained from his 1784
memoir on wire torsion experiments. There Coulomb found that the ‘laws governing the
torsion in a metal thread’ are of the following kind 20 :

Fτ = μBD4/l , (2)

where Fτ is the force due to torsion, μ a constant characteristic of a particular metal, B

the angle of torsion, i.e. the total angle the wire is twisted through, D the diameter of the
wire and l the length of the wire. By coupling electrically charged bodies, such as pith
balls, to a torsion balance, as will be described in detail below, and considering an
equilibrium point where the torsion and electric forces are equal, one can attain the
following relation:

Fe = C/d2 = μBD4/l = Fτ (3)

20
The choices of μ, B, D and l as the letters to signify the variable in the wire-torsion equation are
Coulomb’s. See Coulomb (1784), p. 247.
 10

Since C, μ, D and l can be chosen as constants, i.e., they do not vary in time between
consecutive measurements, the torsion angle of a wire B will be proportional to the
inverse square of the distance:

1/d2 ∝ B . (4)

In this manner, if the distance d between two bodies is halved to d/2, we expect the angle
of torsion, and the corresponding force of torsion, to quadruple: 1/(d/2)2 = 1/(d2/4) = 4/d2
∝ 4B ∝ 4Fτ. It is this technique that Coulomb uses in his 1785 memoir to demonstrate
the inverse square law of electrostatics.
After introducing the torsion-wire relation in equation (2), Coulomb goes on to
describe the construction of his electric torsion balance. This is followed by a
presentation of his experiment, and an explanation of the results. The memoir ends with
four remarks with regards to the experiment. We will follow the same order of
presentation, but for the description of the torsion balance we quote Coulomb in length.
His description is clear, concise and of the utmost importance to the project presented
here. The reader is urged to follow the description by focusing on Figure 1, which is an
illustration of Coulomb’s torsion balance and its parts as it appears in the published
memoir. 21

21
Figure 1 includes within it 5 figures which are referred to as Fig. 1, Fig. 2 and so on, in Coulomb’s
description of the torsion balance.
 11

Figure 1. Coulomb’s torsion balance of 1785.

 12

Over a glass cylinder [ABCD] 12 inches in diameter and 12 inches high, a flat plate of glass 13
inches in diameter was placed, which covers the entire structure. This plate has two holes of about
20 lines [≈4.5cm] in diameter, one of them at its center, f, on which a glass tube 24 inches high is
placed. This tube is cemented in place with the cement currently used in electrical apparatus. On
top of the tube at h is placed a torsion micrometer, shows in detail in Fig 2. The top part of this
micrometer, No. 1, has a knob b, index io, and a suspension clamp q, which fits into the hole G of
part No. 2. Part No.2 is made up of a circle ab, which has 360° scale on its edge, a copper tube Φ
that fits into the hole H of part No. 3, which is attached to the top of the glass tube fh of Fig. 1.
The clamp q (Fig. 2, No. 1) has nearly the form of the tip of a crayon holder, which can
be narrowed by means of the ring q. It is in this clamp that on end of a very fine sliver wire is
placed. The other end of this wire is attached at P (Fig. 3) by means of a clamp on the rod Po. This
rod is made of copper or iron and its diameter is barely one line [≈2.3mm.] The upper end P is
split, making a clamp that is closed by means of the sliding ring φ. This cylinder is enlarged and
pierced at C by a sliding needle ag. The whole weight of the cylinder must be such that it can keep
the silver wire stretched without breaking it. The needle ag, as can be seen in Fig. 1, is suspended
horizontally at its center and at about half the height of the glass container. It is made either of a
silk thread coated in Spanish wax or of a straw similarly coated, [and terminated from q to a in 18
lines (≈4cm) of length by] 22 cylindrical thread of shellac 23 ; at one end of this needle is placed a
small ball which is made of pith 24 and is two to three lines in diameter [4.5-6.8mm]. At the end g
is a small piece of paper soaked in turpentine; this paper counterbalances the ball a and slows
down the oscillations.
We have said that the glass cover AC has a second hole m; it is through this hole that a
small rod mφb is introduced. The lower part of this rod (φb) is made of shellac, and at b terminates
in another small pith ball. Around the glass container , at the height of the needle is a scale ZQ

22
Shamos (1959), p. 63 translates ‘The needle… is made of either silk or thread coated with Spanish wax…
It is about 18 lines long and terminates in a cylindrical thread of shellac…’ implying that the needle is 18
lines long but we, as well as Martínez (2006) p. 520, understand the cylindrical thread of shellac, and not
the needle, to be of 18 lines long: ‘L’aiguille que l’on voit (fig I) en ag, suspendue horizontalement à la
moitié à peu-prés de la hauteur du grand vafe qui la renferme, eft formée, ou d’un fil de soie enduit de cire
d’Espagne, ou d’une paille également enduite de cire d’Espagne, & terminée depuis q jufqu’en a, sur 18
lignes de longueur, par un fil cylindrigque de gomme-laque…’. Coulomb (1884 [1785]), p. 110.
23
In footnote 9, p. 520, Martínez (2006) identifies ‘shellac’ (or gomme-laque) as a sticky resin secreted
from various lac insects. Shellac, along with turpentine and additional resins, is the main ingredient in
Spanish Wax (cire d’Espagne).
24
The ‘pith’ referred to here is a soft substance found, for instance, on the inside elderberry shrubs: ‘une
petite balle de sureau’. Cf. Coulomb (1884 [1785]), p. 110.
 13

divided into 360 degrees. This scale was made for simplicity out of paper fastened around the
25
container at about the height of the needle… .

After bringing to two pith balls together at the zero degree mark of the ZQ scale
and grounding them, Coulomb goes on to electrify the two pith balls when they are in
contact. By symmetry, the charge distributes approximately equally onto both pith balls
and the balls repulse each other. Following a few oscillations the pith ball on the needle
settles at a distance away from the stationary pith ball which Coulomb records. He then
twists the micrometer an amount which quadruples the total force of torsion and
examines whether the distance between the pith balls converges to half, as would be the
case if the electrostatic law is an inverse square law. In his memoir Coulomb reports three
trials ‘which are easily reproduced, and which will make evident the law of repulsion:’ 26
First Trial. Having charged the two balls with the head of a pin with the micrometer
index set a O, the ball a of the needle is separated from the ball t by 36 degrees.
Second Trial. Turning the suspension thread through 126 degrees by means of the knob O
of the micrometer, the two balls are found separated and at rest at 18 degrees from one another.
Third Trial. After turning the suspension thread through 567 degrees, the two balls are
27
separated by 8 degrees and a half.
We refer the reader to Martínez (2006) and Coulomb’s original 1785 memoir for an in-
depth analysis of the results. In short, these results show that as the force of torsion,
which is proportional to the angle of torsion, is quadrupled from the first trial (36° of total
torsion angle) to the second trial (144°=36°+126° of total torsion angle), the distance
separating the balls halves, and the same is true from the second to the third trial. Here
we will use these results to calculate the exponent n of the distance d between the two
balls in the electrostatic law (see eq. (1)) with the following equation, assuming, as
Coulomb did, that the pith balls are small spheres (an assumption that can be regarded as
equivalent to assuming point-bodies) 28 :
25
Coulomb (1884 [1785]), pp. 108-110. The translation presented here is that of Shamos (1959), pp. 62-63.
Emphasis is ours. For comparison, the reader is referred to Martínez (2006), pp. 519-520, for another
translation.
26
Shamos (1959), p. 64.
27
Ibid., p. 64.
28
See Martínez (2006), pp. 523-525, for the derivation of this equation.
 14

n = ln[θacos(β/2)(αm+α)/θbcos(α/2)(βm+β)]/ln[sin(β/2)/sin(α/2)] , (5)

where θa is the elastic modulo of the wire in a first trial, θb the elastic modulo of the wire
in the subsequent trial, and generally θ=μD4/l. Assuming that the wire is not over-twisted,
θa= θb (and thus they cancel each other out). Further, α is the separation angle between the
two pith balls on a first trial, β the separation angle on a subsequent trial, αm is the
micrometer angle on a first trial and βm the micrometer angle on a subsequent trial. As
can be seen from equation (5), we need two trials to calculate a possible exponent n. Thus
we can calculate that Coulomb’s results predict an exponent of 1.98 for trials 1-2, an
exponent of 1.84 for trials 2-3, and an average exponent of 1.91 (we will discuss later on
the likely illegitimacy of this averaging process). These results should be compared with
the theoretical exponent value of 2 appearing in equation (1).
In order to evaluate the results of replications, we need to decide how they
compare to the exponent predicted by Coulomb’s results. This invites consideration of
the sources of error. The 1785 memoir ends with four relevant remarks (which we will
return to in sections 3, 4 & 5) explaining some sources of error and how one can account
for such sources. They may include small angle approximation (though equation (5) does
not make such an approximation), loss of charge during the trial, and error due to over-
twisting of the silver wire and consequent changes of its elastic modulo between trials.
In the following section, we will analyze recent replications and how their results
compare to the exponent which Coulomb’s results predict, concentrating on Martínez’s
work (since Martínez has already thoroughly analyzed Heering’s papers).

3. Previous Replications: Martínez and Heering

We will focus on three unresolved issues that will help us understand better the problems
of replication of the experiment, and how the results compare to the exponent which
Coulomb’s own results predict.
Firstly, the two most recent replications, Heering’s and Martínez’s, end with
directly contradicting conclusions. Heering claims that he found it impossible to attain
the results Coulomb reported in his 1785 memoir, while Martínez claims that such results
 15

are most likely typical. 29 This disagreement naturally calls for a third voice to confirm
one claim over another since, although Martínez goes a long way to discrediting much of
Heering’s evidence. Although on the latter count we agree with Martínez, his analysis
does not appear to be conclusive, as we will suggest.
Secondly, both Heering’s and Martínez’s replications diverge from Coulomb’s
1785 prescriptions in ways that we believe significant. For example, Heering’s
replication included a copper wire instead of silver that was soldered to the micrometer
rather then clamped and his needle was made out of PVC. It might very well be that
Heering’s replication failed because of these divergences. In fact Matinez’s appendix, in
his 2006 paper 30 , seems to confirm this idea. There, Martínez reports experimenting with
various wires such as aluminum. In one case, the wire was tied instead of clamped to the
micrometer. In these experiments Martínez produced poor results, very much comparable
to those of Heering. For the 14 successful experiments that Martínez published in his
paper (not those in the appendix), his apparatus diverged from Coulomb’s in the sense
that his silver wire was almost double the thickness of the one Coulomb used for his 1785
reported results and the needle Martínez used was made out of plastic (from the handle of
a blue Gillette shaving razor). In an attempt to experiment with a needle made from a
combination of wax and synthesized shellac, Martínez found that one ‘needle sagged a bit
too much and another reacted to the charge of the pith balls.’ 31 In particular, he found
that ‘the waxed needle itself accelerated toward the stationary ball when the two were
close… and was interfering with the repulsion of the pith balls.’ 32 Martínez concluded
that the material he used for his wax needles was not electrically neutral so instead he
used the plastic needles that did not show the above mentioned phenomena. But,
crucially, could it be that Coulomb’s needle, made from a combination of shellac and
turpentine, did not behave like Martínez’s plastic needle? In the following section, we
will indeed answer in the positive.

29
Heering (1992), p. 990. Martínez (2006), p. 547.
30
Martínez (2006), pp. 547-561.
31
Ibid. p. 536.
32
Ibid. p. 560
 16

More generally, both Heering’s and Martínez’s replication seem to confirm that
the electric torsion balance’s behavior is very much sensitive to its material make-up.
Hence the experimental success of the torsion balance as described by Coulomb must be
judged by using a replica of the torsion balance that is very close to that of Coulomb. It is
this kind of torsion balance and experiments that we have tried to replicate and will
present in the next section.
Thirdly, there is an issue concerning the accuracy with which prior results were
reported. One type of ‘unavoidable’ source of error is the limited accuracy range of any
apparatus. Coulomb does not report the error on his separation angle between the two
balls or micrometer angle, at least not in any straight forward manner. Nevertheless,
given Coulomb’s reported trials (cf. section 2), one can estimate that the upper limit of
error on his separation angle was 0.5 degree and that on his micrometer it was 1 degree.
This is confirmed by recent replications. Heering reports 0.5 degree error on separation
angle and 1 degree on the micrometer (when the experiment is done with a Faraday
cage) 33 , Martínez reports 0.3 degree on separation angle and 0.5 degree on micrometer
angle 34 , and we can report similar observational inaccuracies as those of Martínez.
Coulomb’s 1785 results predict exponents of 1.98 and 1.84. We propose to label
any set of trials that predicts exponents between 1.84-2.16 as successful Coulomb
experiments; that is, an exponent of n = 2.00 +- 0.16 is to be considered a successful
Coulomb type experiment 35 . One can also be more modest in experimental expectations
and label any exponents that ranges between 1.50-2.50 as moderately successful
experiments, i.e., an exponent of 2.00+-0.50 36 . In contrast, if we study Heering’s
published results and their corresponding exponents, as presented in Table 1, we see that
only 2/11 experiments produce successful Coulomb experiments, 2/11 produce

33
Heering (1992), p. 990.
34
Martínez (2006), p. 541.
35
Our calculation attribute a maximal effect of 0.14 on the calculated Coulomb exponents given the kind of
observational inaccuracies that Martínez reports. This would push the lower bound of our ‘Coulomb type
successful exponents’ to 1.70 (i.e. 1.70-2.16) but we ignore this issue for now and work with those
standards of Martínez (2006).
36
In counting ‘moderately successful experiments’ with exponents of n=2.0+-0.5 I do not include the
‘Coulomb type successful experiments’ of exponents n=2.00+-0.16.
 17

moderately successful experiments and the remaining 7/11 experiments are certainly
unsuccessful. 37

Table 1. An example of data from Heering’s replication, including micrometer and

separation angles, with corresponding calculated exponents for every two consecutive
trials.

micrometer 0 20 40 60 80 100 130 160 200 250

separation 32.5 26 22.5 19 16 14.5 13 10.5 8 6.5

micrometer 300 360

separation 5.5 4

trials 1-2 2-3 3-4 4-5 5-6 6-7 7-8 8-9 9-10 10-11
exponent 1.52 2.10 1.36 1.11 1.78 2.03 0.81 0.72 1.01 1.04

trials 11-12
exponent 0.54

Now, our results (reported in the following section), do demonstrate that an able
experimenter that builds a torsion balance sufficiently close to Coulomb’s design can
attain moderately successful exponents of 2.00+-0.50. But, crucially, Martínez’s reported
results seem to go further and confirm that Coulomb-type results are typical, though, as
we shall see in a moment, this cannot be the case. Table 2 reproduces the exponents
Martínez published for his 14 experiments (each experiment consisted of 3-4
measurements). We can see that Martínez reports 11/18 (≈61%) successful Coulomb-type
exponents, 4/18 moderately successful exponents and only 1/18 unsuccessful exponent,
in experiment B2. The latter can be explained by taking into account the error introduced
by Martínez when he twisted the wire beyond its linearity range prior to experiment B2.

37
Heering (1992), p. 14. See also Heering (1994), p. 55.
 18

Table 2. Set of fourteen of Martínez’s published experiments (not including his appendix
section) with their corresponding average exponent.

experiment # A1 A2 A3 A4 B1 B2 B3 B4 B5
exponent reported 1.894 1.956 2.277 2.068 2.216 2.707 1.678 2.357 2.266

experiment # C1 C2 C3 D1 D2 D3 D4 D5 D6
exponent reported 1.911 2.025 2.072 1.679 1.956 1.878 1.919 2.069 2.072

Thus, Martínez’s results seemingly confirm the claim that Coulomb-type results
are typical. Yet there is a catch. First, by reporting 3 digits beyond zero for each
exponent, Martínez seems to be suggesting that his accuracy in calculating exponents is
that of three significant digits, but his reported observational error is 0.5 degree on the
micrometer and 0.3 degree on separation angle. So, with such observational inaccuracy,
how is he able to know what the third digit actually is? In this manner, we think that
Martínez’s reported exponents inadvertently confirm a higher level of accuracy than his
experiments actually achieve.
In a similar vein, the exponents which Martínez reported are average exponents
of an experimental run 38 . Table 3 presents a list of exponents calculated for each of
Martínez’s consecutive trials, i.e., for every two consecutive measurements, as he himself
did before averaging them. Note that now only 18/43 (≈42%) trials can be considered
successful Coulomb-type experiments, and if one takes into account that, before
performing the set of the 14 experiments (corresponding to 43 trials) published in his
paper, Martínez performed over 60 experimental trials (also published in his paper, and
most of which resulted in unsuccessful exponents), it becomes far from obvious that his
data truly does confirm that Coulomb-type results are typical. In fact, Martínez’s results
become worse if one fixes them for the error due to loss of charge.
Consider, as an example, Martínez’s experiment D4, which he identifies as
‘particularly noteworthy’ due to low standard deviation and consistency of exponent.
Martínez reports that the total charge lost in experiment D4 is of 4 degrees.

38
An experiment or experimental run such as A,1 for example, consists of 3-4 measurements or trials with
corresponding 2-3 predicted exponents.
 19

Table 3. Martínez’s complete published experiments (not including his appendix section)
along with forty-three exponents corresponding to any two consecutive measurements in
an experimental run.

experiment # A1 A2 A3 A4 B1
exponent 2.11 1.67 2.26 1.64 2.44 2.10 2.31 1.81 2.56 2.15

experiment # (con. B1) B2 B3 B4 B5

exponent 1.92 2.76 2.72 2.62 1.81 1.54 2.43 2.43 2.28 2.40

experiment # (con. B5) C1 C2 C3 D1

exponent 2.12 2.22 1.59 2.30 1.74 2.27 1.86 2.03 1.64 1.35

experiment # D2 D3 D4 D5
exponent 2.10 1.92 1.83 1.87 1.85 1.90 1.99 1.88 1.87 2.00

experiment # (con. D5) D6

exponent 2.13 2.40 2.12

Since, roughly speaking, charge loss can be considered to be proportional to the

quantity of charge on the pith ball, we can conclude that, for example, in trial 1-2 at least
1.5 degrees were lost. If we take this loss into account when calculating Martínez’s
exponent for trial 1-2 of experiment D4, the exponent changes from 1.99 (Coulomb type
exponent) to a 2.24 (moderately successful exponent).
To drive the point home, consider the following. Categorizing experimental data
as successful includes not only getting the “right numbers” but also being able to explain,
at least to an extent, those results that diverge form the expected values because of error
sources. Martínez himself does precisely this when he states that the ‘likely cause of such
unusually high exponents [, as reported in experiments B1, B2, B4 and B5 in Table 2,]
was that once the silver wire had been subjected to a large total torsion, [in experiments
 20

A3 and A4,] the wire’s elastic modulus decreased’ resulting in higher exponents 39 .
However, when we look at the individual exponents of the experimental trials of
experiments B1, B2, B4 and B5 (as shown in Table 3), rather than at the average
exponents, we notice that there are numerous experimental trials that give results such as
1.92 and 2.12 (Coulomb type results) and 2.15, 2.43, 2.28, 2.40 (moderately successful
results).
Hence, we are left with the conclusion that only when averaging them do
Martínez’s results really support the claim that Coulomb’s data were typical and that the
torsion balance experiment behaves in a non-erratic manner (i.e., such that high or low
exponents can be explained by taking the appropriate error source into account).
But would Coulomb, or another experimenter-scientist working in Paris at the end
of the eighteenth century, consider Martínez’s averaging technique as a legitimate
manner by which to demonstrate the electrostatic law or any other law of nature? And if
so, are we to understand Coulomb’s reported data not as actually observed trials but as an
averaging of many experimental trials?
At first glance Coulomb’s various memoirs suggests that he does not make use of
averaging techniques. Here we come face to face with the question of Coulomb’s views
on probability, averaging techniques, error theory, and how these techniques might be
applicable to observation and experiment. Lorrain Daston notes that the commission of
academicians, of which Coulomb was a member, that evaluated the Essai sur l’
application de l’ analyse à la probabilité des décisions rendues à la pluralité des voix
(1785) by the Marquis de Condorcet (1743-1794), stated that the essay was “full of
precious applications, luminous and profound reflections… [and] new views both for
changing and improving the methods upon which depend the security and happiness of
the people” 40 . Yet, Daston says, Condorcet’s Essai “seems to have been eclipsed by the
events of the Revolution in the minds both of its author and its readers 41 . Further, we note
that Coulomb landmark Mémoire of 1785 on electricity and magnetism was published a

39
Martínez (2006), p. 543.
40
Daston (1988), p. 352.
41
Ibid. Fascinating insights into the rise of quantitative statistics in France under Napoleon are given by
M.-N. Bourguet (1987).
 21

full ten years before P. S. Laplace (1749-1827) delivered his famous lecture on
probability at the Écoles Normales in 1795. The latter would be recast into the
Philosophical essay on probability prefaced to his celebrated Analytical theory of
probabilities. In the Philosophical essay, Laplace states that “Natural phenomena are
most often surrounded by so many foreign circumstances, and so many perturbing causes
confound their influence, that it is very difficult to recognize them. In such a case they
can be discovered only by increasing the number of observations or experiences {or
experiments}, so that as the foreign effects finally cancel each other out, these
phenomena and their various components are clearly revealed by the mean results” 42 . At
present, we can only speculate about what Coulomb might have made of such a crystal
clear statement.
Thus, all being considered, we tend to regard Martínez’s averaging techniques
applied to his replication of Coulomb’s experiments as lacking historiographical validity.
In conclusion, we reiterate that the success of electric torsion balance experiments
is highly dependent on reproducing a historically-accurate material makeup of the
balance, and that averaging calculated exponents from measured trials is a rather
anachronistic manner by which to judge the success or failure of Coulomb’s experiment.

4. Our Replication of Coulomb’s Experiment

In this section, we discuss those aspects of our replication of Coulomb’s torsion balance
that diverge from recent replications by Heering and Martínez. We do not present a large
set of data since this would be very much a repetition of Martínez’s work. Instead we
present short data sets, which exemplify our results in a spirit similar to Coulomb’s 1785
memoir.
Our presentation comes in two parts. Firstly, we will describe our experience in
performing the torsion balance experiment with various, historically-accurate replicas of
the needle and the unavoidable error source we believe to have identified. Secondly, we
will discuss our hypothesis that the success of the torsion balance experiment is highly

42
See Dale (1995), p. 43.
 22

dependent on the materials involved and that the experiment degenerates when
parameters are gradually altered.

Table 4. Comparison of dimensions of torsion balances.

Components Coulomb’s 1785 Martínez’s 2005 HPS 2008/09

glass cylinder, diameter 32.48cm 30.0cm 29.8cm
glass cylinder, height 32.48cm 30.0cm 29.8cm
apertures on lid, diameter 4.51cm 3.1cm 3.3cm
glass tube, length 64.97cm 61.5cm ???
silver wire, length 75.79cm 75.5cm 75.0cm
silver wire, diameter 0.035mm 0.051mm 0.051/0.031mm┼
needle, length 21.66cm 20.5cm 20.3-25.0cm*
pith balls, diameter 4.5 to 6.8mm 6mm ???
┼
The experiments with the Spanish wax and Plastic needles were conducted with a silver wire of 0.051mm and the experiments with
the Shellac needles were conducted with a silver wire of 0.031mm.
* The Spanish wax needle was 25cm, the plastic needle was 20.3cm and the shellac needle was 21cm.

4.1 Shellac versus Plastic

The chronology of our replication of Coulomb’s torsion balance experiments somewhat
mimicked Martínez’s (see Figure 2). In the beginning we attained data that predicted
exponents ranging between 0.5 to 2.5, in a relatively erratic manner, but with time, as our
experimental techniques matured, we were able to attain consistent results that were
moderately successful. Unsuccessful results were explained by various sources of error
such as charge loss (lower exponent) and over-twisting of the silver wire due to its elastic
modulo being modified (higher exponent). We refer the reader to Table 4 for a
comparison of various sizes of the torsion balance apparatus as reported in Coulomb’s
1785 memoir, Martínez’s 2006 replication, and our 2008/09 replications 43 .

43
A similar table appears in Martínez (2006), p. 537. The construction data here reported for Coulomb’s
1785 memoir are taken directly from Martínez (2006) and were confirmed by us through a reading of the
1785 first memoir on electricity and magnetism (we did not recalculate Martínez’ approximation of the
sliver wire’s or pith balls diameter).
 23

Figure 2. Replication of Coulomb’s torsion balance in the HPSLab. Left: the complete
balance. Right: a detail showing the micrometer and the clamp holding the silver wire.

In what follows, we will present data obtained with three different kinds of
needles, which we shall label, for convenience, Spanish wax needle, plastic needle and
shellac needle.
In contrast to Martínez, we were able to acquire “wax” that was most likely
similar to Coulomb’s Spanish wax, in the sense that our wax’s main compound was
shellac (not synthesized plastic) along with turpentine and other colorants. We managed
to make very thin needles, weighing only about 0.3-0.2 gram, by melting the wax on an
electric burner covered in aluminum foil and dipping in it a thin thread of silk44 . Yet, in
the replications in which we used a silver wire of thickness 0.051mm, as Martínez did,
and which were most successful― i.e., behaved consistently and portrayed minimal
charge loss―, we were only able to attain exponents ranging between 1.6-1.8. We were
unable to push the exponent closer to Coulomb’s exponent range. Table 5 presents an

44
This process took weeks to master since it is very difficult to create a thin, light needle that is
approximately homogenous and decently straight.
 24

example of such data. In addition to calculating the exponents predicted by our

measurements, we present exponents that were fixed for maximal charge loss effects―
that is, we assumed that the entire charge loss in the experiment, 0.5 degrees, was lost
between consecutive trials, and recalculated the exponent appropriately. Notice that all
four exponents reported for trials 1-2 and 2-3, for both experiments S1 and S2, are
moderately successful results that fall in the 1.6-1.8 range.
Furthermore, even from those exponents fixed for maximal charge loss, only one
exponent borders the Coulomb range of 1.84, but of course maximal charge loss is
unrealistic since charge is lost over the entire set of measures and not just, say, between
trials 1-2. So, charge loss could not explain our low exponents.
Since the experimental results were consistent, our apparatus very similar to
Coulomb’s and Martínez’s (diverging from Martínez’s mainly in the material makeup of
the needle, clamps and micrometer), we decided to study the needle further. Our shellac
needle seemed to be sucked in by the charged stationary pith ball (although our best
needle did not sag and did not behave erratically).

Table 5. Two experiments from our replication with 0.051mm silver wire and Spanish
wax needle. ┼

experiment # S1 S2
micrometer 5 90 180 5 90 180
separation 45 28 20.5 42.5 25 18
total deg. lost 0.5 0.5
trial 1-2 2-3 Ave. 1-2 2-3 Ave.
exponent 1.76 1.68 1.72 1.62 1.63 1.63
fix* 1.84 1.83 1.83 1.69 1.79 1.74
┼
wire diameter: 0.051mm needle used: Spanish wax needle weight: ~0.3g
* Accommodating charge loss, maximal effect assumed

Here are some noteworthy aspects of the phenomenon that we witnessed. (i) We
manage to suck the needle inwards, by placing charge on the stationary pith ball, from
separation angles as large as 45 degrees, i.e., the general range of torsion balance
experiments. (ii) The strength by which the needle was pulled seemed to be roughly
proportional to the charge placed on the stationary pith ball and approximately inversely
 25

proportional to the separation angle between the two pith balls. (iii) The needle would
stick to the stationary charge pith balls for up to around 30 min.
Thus, it is cannot be the case that the needle was conductive, for, if the needle had
been conductive 45 the charge would have “evaporated” quickly through the silver wire
(more quickly than it would have without contact with the needle), but this was not
witnessed. Our conclusion was that the Spanish wax needle behaves like a dielectric
material. All dielectric materials, as we known, electrically polarize in the presence of an
electric field. The strength of polarization depends on the electric field (the charge placed
on the stationary pith ball) and a constant characteristic of the material (the electric
susceptibility). Our Spanish wax needle, unlike Martínez’s plastic needle but like his wax
needle, portrayed a strong polarization phenomenon that accounts for our lower
exponents. In addition, notice that, once we take the polarization phenomena into
account, we can better explain the lower exponent of 1.84 which Coulomb attained for
his trials 2-3 since according to Coulomb charge loss in his experiment did not contribute
significantly.
In order to test our reasoning about the wax needle, we decided to run some
experiments with a plastic needle. We originally used a heavy and short plastic needle
made out of Starbuck’s coffee stirrer (red plastic). However, this needle behaved
erratically, so we attempted to make a needle out of a Gillette shaving razor as Martínez
did. We were unsuccessful. Eventually we managed to make a light and thin needle from
some stirrers made of white plastic bought at a supermarket and we ran some
experiments, an example of which is given in Table 6.

45
In effort to confirm the needle was not conductive, we attempted to run large quantities of electricity
generated with a whimshurst machine through the needle to satisfy ourselves that it did not conduct.
 26

Table 6. Two typical experiments from our replication with 0.051mm silver wire and
white-type plastic needle.┼

experiment # P1 P2
micrometer 5 90 180 5 90 180
separation 44 30 23 38 25 19
total deg. lost 0.5 0.5
trial 1-2 2-3 Ave. 1-2 2-3 Ave.
exponent 2.31 1.95 2.13 2.33 1.98 2.15
fix* 2.43 2.14 2.28 2.46 2.20 2.33
* Accommodating charge loss, maximal effect assumed
┼
wire diameter: 0.051mm needle used: White Plastic needle weight: 0.3g

We found that the white plastic needle barely showed the polarization
phenomenon— it only did so at small angles less than 6-8 degrees— and, as can be seen
from Table 6, we attained Coulomb type exponents of the range that Martínez did. In
fact, we found that we were able to perform the torsion balance experiment with just the
plastic needle with no pith ball, i.e., we observed that the needle’s tip picked up charge,
though the needle itself did not conduct. The extra bit of charge picked up by the needle,
which is not distributed spherically, might explain the higher than expected exponent we
found in trails 1-2 in experiments P1 and P2. This, though needs further scrutiny.
Thus, the main point to emphasize from these results is that the experiment’s
success is highly dependent on the materials used. Ultimately, we conclude that most
likely Coulomb’s own needles were of material closer to our Spanish-wax and shellac
than to Martínez’s plastic.
To confirm our hypothesis that shellac is subject to strong polarization effects, we
ordered several different kinds of shellac and made various different shellac needles 46 .
All such needles showed similar polarization effects.
In addition, to abide by the historical standards of replication that we set ourselves
(of course within the limit of our resources), we ran sets of experiments with our best
shellac needle and with a new silver wire of diameter 0.035mm such as reported in

46
Our shellac, ordered from www.shellac.net, is not synthesized or heavily processed. The gamut includes
shellac of the following types: Kusmi See, Dark Seed, Kusmi #1 Button, Kusmi #2 Button, Bysakhi Button,
Gossamer Flake, Hand Made Yellow and Lemon Yellow Orange.
 27

Coulomb’s 1785 memoir. The results are similar to those attained with the Spanish wax
needle and are reported in Table 7.

Table 7. Four typical experiments from our replication with 0.035mm silver wire and
shellac needle.┼

experiment # E1 E2
micrometer 0 90 180 0 90 180
separation 35.5 18.5 13 27 13 9
total deg. lost 0.5 0.5
trial 1-2 2-3 Ave. 1-2 2-3 Ave.
exponent 1.75 1.62 1.68 1.62 1.50 1.73
fix* 1.83 1.82 1.82 1.71 1.77 1.74
┼
wire diameter: 0.035mm needle used: Shellac needle weight: 0.2 grams
humidity: 47-54% temp.: 74-78° F
* Accommodating charge loss, maximal effect assumed

experiment # E3 E4
micrometer 0 90 400 0 90 400
separation 40 22 9.5 41.5 23 10
total deg. lost 0.5 0.5
trial 1-2 2-3 Ave. 1-2 2-3 Ave.
exponent 1.68 1.53 1.59 1.66 1.53 1.59
fix* 1.76 1.63 1.69 1.73 1.63 1.68
┼
wire diameter: 0.035mm needle used: Shellac needle weight: 0.2 grams
humidity: 47-54% temp.: 74-78° F
* Accommodating charge loss, maximal effect assumed

To summarize, 5/8 exponents are moderately successful and fall between 1.6-1.8,
while the remaining 3/8 are moderately successful and fall under the 1.6 range and once
we fix for maximal charge loss all results fall within 1.6-1.8 range. There is one last
noteworthy point to mention with regards to the 0.035mm wire. It is not possible that
Coulomb’s set micrometer angles and measured separation angles are typical since once
the 0.051mm wire is twisted to 567° (a total torsion of 575.5°), its elastic modulo is
modified and, as is clearly described in Coulomb’s 1785 paper, in the first remark on the
experiment, the zero point of the needle is offset by
 28

2 or 3 degrees. As such, to have a first attempt to compare with the [reported experiment] we
must, after having electrified the two balls, twist the wire by 30 or 40 degrees … so that the 2 or 3
47
degrees of incertitude in the first position … does not produce in the results sensible error.
This means that any consecutive experiments, after the first Coulomb experiments cannot
start at the zero micrometer mark. But is the same true of the actual wire Coulomb used?
We found that it is possible to run 3-4 Coulomb type experiments before the
0.035mm wire’s elastic limit is exceeded, at which point one can either change a wire or
compensate the results for error as Coulomb describes. Since changing a wire can take a
long time because of the delicateness of the 0.035mm wire 48 it would seems unlikely that
Coulomb conducted more than 4 experiments with the actual numbers he published in a
typical run of many experiments. Also, the fact that all needles composed out of shellac
portrayed a strong polarization effect implies that it is likely that Coulomb’s needle
behaved similarly to ours and that his 1785 results were not typical. Rather, we suggest,
after running sets of experiments that ranged in exponents between 1.6 and 2.0 (due to
various sources of error), Coulomb chose to publish that point which best demonstrates
the fundamental electrostatic law and exemplifies the application of his torsion balance.

4.2 Sensitivity to Materials and Parameters

Our hypothesis that the success of the torsion balance experiment is highly dependent on
its material make up and parameters involved is partially confirmed by prior replications.
As mentioned earlier, Heering’s replication, as well as Martínez’s original replications,
described in the appendix of his paper, diverge from Coulomb’s description and do not
achieve Coulomb type results. The success of the experiment, we submit, depends on the
wire material that is used, on the manner in which the wire is attached to the micrometer
and more generally on the dimensions of the apparatus. We further strengthen this
hypothesis with the experimental data presented above as regards to needle. Whether or
47
Coulomb (1884 [1785]), p. 113.
48
Martínez (2006), p. 545, notes that it took him several hours of work to change a wire. We experienced
something similar. It takes some patience to manipulate wires so thin that the sense of touch of the finger
tips becomes useless, and sight does not help much since the wire is almost invisible, though with time we
learned how to clamp the wire with our pencil-holder device pretty efficiently (especially when this
operation was done by two experimenters).
 29

not one can attain Coulomb type results depends on the type of needle one uses, its
material, thickness, weight and so on. In addition, the sensitivity of the apparatus
generally depend on the parameters involved. We found that separation angles smaller
than 10 degrees, resulting from a very small amount of charge, or from over-twisting the
micrometer so as to force the pith balls close together, afford lower exponents. In such a
situation, the force between the pith balls is weakened. Thus, as the separation angle
decreases into the region of very low angles the experiment degenerates.
Similarly, as mentioned above, large separation angles and large micrometer
angles modify the wire’s elastic modulo thereby preventing experimental success. Also, it
becomes impossible to run the experiment with large separation angles, since the amount
of charge necessary to produce angles larger than 60-70 degrees repulses the pith balls so
violently that the needle flies outward, either bumping into the stationary ball or, if the
latter is quickly removed, over twisting the wire. One can counter this effect by using a
thicker wire or beginning the experiment with large micrometer torsion, but both of these
techniques render the experiment unsuccessful since the large torsion force overpowers
the electrostatic force quickly as charge is being lost.
It becomes an art of sorts to figure out what kind of torsion balance with what
kind of materials, dimensions and variables is appropriate for what kind of phenomenon
is being studied, whether it be fluid resistance, repulsive electric force, attractive electric
force, magnetic force, and so on. Hence, we are left with the following question. If, in
order to demonstrate the electrostatic law, Coulomb must choose the “correct” torsion
balance, by which criteria can he make such a crucial choice? This question is left open
but we speculate that a large part of the answer involves the experimenter’s ability to
identifying sources of error and have a theoretical story to tell about why experiments
degenerate in the manner they do.

5. Coulomb and the Electric Polarization of the Needle

In section 1, we explained how our study of the torsion balance leads us to the historical
question regarding Coulomb’s awareness of sources of error. Specifically, we would like
to know if Coulomb observed the needle’s polarization phenomenon (described above)
and, if so, how he dealt with such error source. However, answering this question entails
 30

an in depth look at the entirety of Coulomb’s memoirs. This is certainly beyond the scope
of this paper. So, in this section, we will only make some preliminary steps in this
direction by focusing on Coulomb’s first and third memoirs on electricity and magnetism.
Did Coulomb observe the needle’s electric polarization effect without a pith ball
and realize that this was a significant source of error (significant here meaning that
Coulomb type results cannot be obtained typically)? In short, we think not. For, even
though Coulomb’s 1785 memoir included remarks detailing those sources of error that
Coulomb thought were significant, there is no hint at polarization phenomena such as
those we encountered. Moreover, not all of the error sources he mentioned were
applicable to his experiment, though he thought it appropriate to discuss them as they
might be potentially significant for a replication of the experiment. He explicitly
mentioned that when such error sources become significant they must be corrected for.
Such error sources include charge loss due to contact with air over time, small angle
approximations, and exceeding the wire’s torsion limit. 49 Now, Coulomb certainly knew
of the common phenomena of static electricity which had become so popular in the
eighteenth century, and he was aware that conductors, after being grounded, react to
electrically charged bodies. For example, in his second memoir on electricity and
magnetism, when discussing possible sources of error, he explicitly states that when
performing torsion balance type experiments one must ‘distance all conductive bodies at
least three feet from the electrified globe and the needle.’ 50 Yet he makes no mention of
the effects dielectric bodies might have on the experiment.
In order to investigate further whether Coulomb knew of the needle’s polarization
phenomenon, we can take a look to his third memoir on electricity and magnetism 51 . The
third memoir is especially relevant to our torsion balance experiment for two reasons.
First, Coulomb announces in the beginning of his memoir that the entirety of
experiments in the third memoir are conducted with the same type of electric torsion
balance described in his first memoir and he even repeats a description of the apparatus

49
Coulomb, Premier mémoire (1785), pp. 574-577.
50
Coulomb, Second mémoire (1785), p. 586.
51
Coulomb, Troisième mémoire (1785).
 31

and experiment there 52 . In this sense, the third memoir, as apposed to the second
experiment which uses a different apparatus, allows scholars to investigate additional
experiments conducted with the electric torsion balance with corresponding extra data.
For instance, one can extract from Coulomb’s description of his experimental procedure
an additional exponent of 1.97 from two reported experimental trials with the torsion
balance of 0 micrometer angle and 40 separation angles for the first reading and 140
micrometer angle with 20 separation angles for his second reading. 53
The second reason for the third memoir’s significance is that it is wholly devoted
to the study of charge loss in conductors in contact with air and dielectric material. So,
for example, Coulomb describes how, in search for the best insulator to study charge loss
of conductors due to surrounding air, he experimented with various needles and chose the
shellac needle as the best insulator. 54 If Coulomb had witnessed the polarization
phenomenon and attributed some significance to it, it would surely have been mentioned
in this third memoir.
In fact, Coulomb knew that the presence of dielectric material between conductive
bodies diminishes the electric force between the two. But he believed that the cause for
such behavior was either that no perfect dielectric existed or that the surface of the
dielectric might conduct electricity when covered with conducting water or gas
molecules. This second cause is one of the reasons why Coulomb made thin needles— to
minimize surface area and, thus, maximize insulation.55 Yet at no point in the memoir,
which describes in detail charge loss in conductors in contact with dielectric material,
does Coulomb mention the idea that, even with a near perfect insulator, where no charge
leaks across its surface, the insulator will react to an electric field as in the polarization
phenomenon of the needle that we witnessed in our replication. He seemed to have no
conception of electric molecular polarization although, in stark contrast, Coulomb had a
conception of magnetic molecular polarization. His biographer notes that as early as
1777, and certainly by 1787, Coulomb thought that magnets were composed of small

52
Ibid., pp. 616-617.
53
Ibid., p. 617
54
Ibid., p. 615.
55
Ibid., pp. 612-614.
 32

particles which themselves acted as small magnets: ‘Each point of a magnet or of a

magnetized bar can be regarded as the pole of a tiny magnet…’ 56 . Furthermore, Gillmore
says that, even in his later memoirs, Coulomb pointed out that he planned to study the
distribution of electricity within the body of dielectrics― study that might have informed
him of electric polarization― but such work was never published. 57
In summary, it might very well be the case that Coulomb noticed the polarization
phenomenon, interpreted as above— that is, as if due to imperfect insulation and, more
important, due to conducting water/gas molecules on surface of the dielectric—, but
ultimately disregarded it as an insignificant source of error. A more comprehensive study
of Coulomb’s works would most likely shed light on this issue.
We conclude with one last remark. Nowhere, either in the third memoir or in the
first or second, does Coulomb make use of the type of averaging data technique Martínez
made use of. There seems to be no cognizance, on Coulomb’s part, of the possibility of
smoothing out observational errors by averaging data.

6. Conclusions
The purpose of this paper was to contribute to the controversy surrounding the
experiments reported in Coulomb’s 1785, especially with a more accurate replication of
the experiments. We are led to the following conclusions.
In contrast to Heering’s research and along the lines of Martínez’s work, our
replication confirms that it is possible to replicate Coulomb’s experiment, as he described
it, with moderately successful results predicting exponents of 1.5-2.0, wherein the best
data, where error sources are minimized, predict exponents of 1.6-1.8. In contrast to
Martínez, we suggest that Coulomb’s reported results are not typical. We found it

56
Coulomb (1777), translation by Gillmor (1971), p. 181.
57
Gillmore (1971), p. 206.
 33

exceedingly difficult to attain data that were as close to the theoretical predicted value as
Coulomb reported. We suggest that the reason for such difficulty is that shellac-based
needles portray a strong electric polarization phenomenon which plastic needles, such as
Martínez’s, do not mimic. In fact, even Martínez’s published data are undermined once
we reject the averaging technique he used to analyze them. Instead, we found that the
experimental success of the torsion balance is highly dependent on the materials involved
and that the experiment degenerates quickly when either parameters or materials are
gradually altered.
Further, when looking at some of Coulomb’s mémoires, concentrating on his first
and third mémoires on electricity and magnetism, we find no evidence supporting the
idea that Coulomb was aware that a perfectly insulating dielectric, with almost no
conductive molecules surrounding its surface area, causes the electric force between two
electrified bodies to decrease significantly, in the manner shellac needles do in the torsion
balance experiment. We submit that this issue calls for further scrutiny.
In sum, revisiting a landmark experiment in the history of science is much more
than replicating its purported original results, it is like an archaeology of knowledge.
There is no such thing as a scientific experiment in the sense of an event which has been
accomplished once and for all. Two centuries after Coulomb, we must conclude that
Coulomb’s experiment with the electric torsion balance has not yet been successfully
replicated, and that we must keep learning from twisting wires and melting shellac.
 34

References

Blondel, Christine. (1994). “La ‘Mécanisation’ de l’électricité: idéal de mesures exactes

et savoirfaire qualitatifs,” in Blondel and Dörries, pp. 99–119.

Blondel, Christine, and Matthias Dörries, [Eds]. (1994). Restaging Coulomb, usages
controverses et réplications autour de la balance de torsion. Florence: Leo S. Olschki.

Bourguet, M.-N. (1987). Décrire, compter, calculer: The debate over statistics during the
Napoleonic period. In: Krüger, L., Daston, L., Heidelberger, M., pp. 305-316.

Coulomb, Charles Augustin. (1773). “Essai sur une application des règles de maximis et
minimis à quelques problèmes de statique, relatifs à l'architecture”, Mémoires de
mathématiques et de physique présentés à l'Académie royale des sciences par divers
savants, et lus sans ses assemblées (Vol. 7, 1776), pp. 343-382.

(1777). “Recherches sur la meilleure manière de fabriquer les aiguilles aimantées, de les
suspendre, de s'assurer qu'elles sont dans le véritable méridien magnétique, enfin de
rendre raison de leurs variations diurnes régulières”, Recueil des savants étrangers de
l'Académie royale des sciences T. 9, pp. 165-264.

(1784) “Recherches Théoriques et Expérimentales sur la force de torsion, & sur

l’élasticité de fils de métal: Application de cette théorie à l’emploi des métaux dans les
Arts & dans différentes balances de torsion, pour mesurer les plus petits degrés de force.
Observations sur les loix de l’élasticité & de cohérence,” Mémoires de l’Académie
Royale des Sciences (1787), pp. 229–269.

(1785) “Premier mémoire sur l'électricité et le magnétisme. Construction et usage d'une

balance électrique fondée sur la propriété qu'ont les fils de métal d'avoir une force de
réaction de torsion proportionnelle à l'angle de torsion. Détermination expérimentale de
la loi suivant laquelle les éléments du même genre d'électricité se repoussent
mutuellement”, Mémoires de l'Académie royale des sciences (vol. 88, 1788), pp. 569-
577).

(1785). “Second mémoire sur l'électricité et le magnétisme, où l'on détermine, suivant

quelles lois le fluide magnétique, ainsi que le fluide électrique, agissent, soit par
répulsion, soit par attraction”, Mémoires de l'Académie royale des sciences (vol. 88,
1788), pp. 578-611.

(1785). “Troisième mémoire sur l'électricité et le magnétisme. De la quantité d'électricité

qu'un corps isolé perd dans le temps donné, soit par le contact de l'air plus ou moins
humide, soit le lond des soutiens plus ou moins idio-électriques”, Mémoires de
l'Académie royale des sciences (vol. 88, 1788), pp. 612-638.
 35

(1786). “Quatrième mémoire sur l'électricité ; où l'on démontre deux principales

propriétés du fluide électrique : la première, que ce fluide ne se répand dans aucun corps
par une affinité chimique ou par une attraction élective, mais qu'il se partage entre
différents corps mis en contact uniquement par son action répulsive ; la seconde, que
dans les corps conducteurs le fluide parvenu à l'état de stabilité, est répandu sur la surface
du corps, et ne pénètre pas dans l'intérieur”, Mémoires de l'Académie royale des sciences
(vol. 89, 1788), pp. 67-77.

(1787). Cinquième mémoire sur l'électricité. De la manière dont le fluide électrique se

partage entre deux corps conducteurs mis en contact, et de la distribution de ce fluide sur
les différentes parties de la surface de ces corps”, Mémoires de l'Académie royale des
sciences (vol. 90, 1789), pp. 421-467.

(1788). “Sixième mémoire sur l'électricité. Suite des recherches sur la distribution du
fluide électrique entre plusieurs corps conducteurs: détermination de la densité électrique
dans les différents points de la surface de ces corps”, Mémoires de l'Académie royale des
sciences (vol. 91, 1791), pp. 617-705.

(1789). “Septième mémoire sur l'électricité et le magnétisme”, Mémoires de l'Académie

royale des sciences (vol. 92, 1793), pp. 455-505.

(1884). Collection des mémoires relatifs à la physique. Tome I. Mémoires de Coulomb.

Paris: Guathiers-Villars.

Dale, A. [Ed.]. (1995). Pierre-Simon Laplace. Philosophical essay on probabilities. New

York, Berlin: Springer-Verlag.

Daston, Lorraine. (1988). Classical probability in the Enlightenment. Princeton:

Princeton University Press.

Dickman, Steven. (1993). “Could Coulomb’s experiment result in Coulomb’s law?”

Science 262, pp. 500–501.

Falconer, Isobel (2004). “Charles Augustin Coulomb and the fundamental law of
electrostatics,” Metrologia 41, S107–S114.

Gillmor, C. Stewart. (1971). Coulomb and the evolution of physics and engineering in
eighteenth-century France. Princeton: Princeton University Press.

Heering, Peter. (1992). “On Coulomb’s Inverse Square Law,” American Journal of
Physics 60, 988–994.

(1994) “The Replication of the Torsion Balance Experiment: The Inverse Square Law
and its Refutation by early 19th-Century German Physicists,” in Blondel and Dörries, pp.
47–66.
 36

(2006). “Regular twists: Replicating Coulomb’s wire-torsion experiments”, Physics in

Perspective 8, pp. 52-63.

Heilbron, John L. (1994). “On Coulomb’s Electrostatic Balance,” in Blondel and Dörries,
pp. 151–161.

Heyman, Jacques. (1972). Coulomb’s memoir on statics; an essay in the history of civil
engineering. Cambridge, England: Cambridge University Press.

Krüger, L., Daston, L., Heidelberger, M., [Eds]. (1987). The probabilistic revolution. 2
vols. Cambridge, Massachusetts, and London, England: The MIT Press.

Licoppe, Christian. (1994). “Coulomb et la ‘Physique experimentale’: Pratique

instrumentale et organisation narrative de la preuve,” in Blondel and Dörries, pp. 67–83.

Martínez, A. “Replication of Coulomb’s Torsion Balance Experiment”, Archive for

History of exact Sciences 60, pp. 517-565.

Pestre, Dominique. (1994). “La pratique de reconstitution des expériences historiques,

une toute première réflexion,” in Blondel and Dörries, pp. 17–30.

Shamos, Morris. (1959). Great experiments in physics. New York: Holt.

37