Dar Rigol

You might also like

You are on page 1of 59

Stud. Hist. Phil. Sci.

34 (2003) 515–573
www.elsevier.com/locate/shpsa

Number and measure: Hermann von Helmholtz


at the crossroads of mathematics, physics, and
psychology
Olivier Darrigol
CNRS: Rehseis, 83 rue Broca, 75013 Paris, France

Received 3 April 2002; received in revised form 25 September 2002

Abstract

In 1887 Helmholtz discussed the foundations of measurement in science as a last contri-


bution to his philosophy of knowledge. This essay borrowed from earlier debates on the foun-
dations of mathematics (Grassmann / Du Bois), on the possibility of quantitative psychology
(Fechner / Kries, Wundt / Zeller), and on the meaning of temperature measurement (Maxwell,
Mach). Late nineteenth-century scrutinisers of the foundations of mathematics (Dedekind, Can-
tor, Frege, Russell) made little of Helmholtz’s essay. Yet it inspired two mathematicians with
an eye on physics (Poincaré and Hölder), and a few philosopher-physicists (Mach, Duhem,
Campbell). The aim of the present paper is to situate Helmholtz’s contribution in this complex
array of nineteenth-century philosophies of number, quantity, and measurement.
 2003 Published by Elsevier Ltd.

Keywords: Helmholtz; Measurement; Arithmetic; P. Du Bois-Reymond; H. and R. Grassmann; J. von


Kries

That [universal units] become more widely known through human transmission
does not change the business and concept of measurement, and appears instead
as merely accidental. (Helmholtz, 1887, p. 375)

The nineteenth century saw the systematic application of quantitative methods to


the entirety of physics. On the experimental side, this evolution implied the develop-

E-mail address: darrigol@paris7.jussieu.fr (O. Darrigol).

0039-3681/$ - see front matter  2003 Published by Elsevier Ltd.


doi:10.1016/S0039-3681(03)00043-8
516 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

ment of increasingly refined concepts and techniques of measurement. But it usually


went without reflection on the conditions that make a given property accessible to
measurement. The question was how to measure, not whether to measure. This state
of affairs explains why measurability does not come out in the rich literature recently
published on the history of measurement. Nor does the opinion, expressed in the
epigraph, that the social transmissibility of standards is irrelevant to the deeper mean-
ing of measurement.
The first physicist to provide a general discussion of the measurability of physical
properties was Hermann von Helmholtz, in his essay published in 1887 under the title
‘Zählen und Messen’ (counting and measuring). Helmholtz first introduced ordinal
numbers, in a manner independent of external experience and related to our ability
to order acts of consciousness in time. He then derived the usual arithmetic axioms
by mathematical induction from natural definitions of addition and multiplication.
Next, he characterized quantity through the equality of objects with respect to a
given method of comparison. To make quantity measurable, he further required a
concrete operation of addition (and difference). Measurement meant the division of
a given quantity into equal units, with a rest to be divided into sub-units. Lastly,
he sketched the indirect measurement of quantities of which no concrete addition
was known.1
The consistency of Helmholtz’s definition of measurable quantity required con-
crete equality and concrete addition to obey the corresponding arithmetic laws.
Helmholtz insisted that only experience could tell whether these properties were met.
If they passed the test, the method of comparison and the procedure of addition
defined a measurable quantity. Helmholtz’s concept of measurement thus combined
conventional and objective elements: it depended on a choice of concrete physical
operations that was partly free, partly determined by experiment.
The sort of consideration found in ‘Zählen und Messen’ was, and still is, extremely
unusual under a physicist’s pen. Yet it is now commonly regarded as a turning point
between an older concept of measurement in which quantity precedes number and the
modern concept in which quantity and number are defined separately. The purpose of
the present paper is to shed light on this transition by analyzing Helmholtz’s sources
and motivations, and by describing the early reception of his reflections.
Measurement was at the heart of Helmholtz’s science. All along his career as a
physiologist and as a physicist, he invented or improved scientific instruments. He
belonged to several commissions for setting metrological standards. He discussed
systems of electric units. The very year he wrote ‘Zählen und Messen,’ he became
the founding president of the Physikalisch-Technische Reichsanstalt and used this
position to promote precision instrumentation.2 As one who made measurement the
foundation of pure and applied physics, Helmholtz was predisposed to reflect on the

1
Insightful analyses of Helmholtz’s paper are found in Hertz (1921); Michell (1993, 1999); DiSalle
(1993).
2
Cf. Cahan (1993), pp. 574–575; Cahan (1989).
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 517

deeper meaning of measurement. Perhaps even greater determining factors were his
usual philosophical attitude and his familiarity with many fields of knowledge.
Helmholtz presented his analysis of counting and measuring as a natural counter-
part to his earlier empiricist analysis of the foundations of geometry, which had roots
in his physiology of perception. He also referred to a few external sources, including
Hermann and Robert Grassmann’s formalist foundation of arithmetic, Paul Du Bois-
Reymond’s phenomenological definitions of number and quantity, and Adolf Elsas’s
Kantian criticism of measurement in psychology. On the latter issue, Helmholtz was
probably aware of the debates in which participants included his former assistant
Wilhelm Wundt and his former student Johannes von Kries, and the Berlin philo-
sophy professor and friend to whom Helmholtz’s essay was dedicated, Eduard Zeller.
Lastly, Helmholtz must have known James Clerk Maxwell’s influential discussion
of temperature measurement.
A first striking feature of these sources is the deep connection that Helmholtz
perceived between the theory of measurement and the foundations of mathematics.
Also notable is the conflict between his various sources: Du Bois and the Grassmanns
held nearly opposite views of the foundations of mathematics; Elsas rejected Du
Bois’s concept of quantity. Most intriguing is the difference between Helmholtz’s
motivations and those expressed in his sources. Whereas Du Bois aimed at a rigorous
definition of limits in mathematics, Helmholtz judged this sort of consideration
superfluous in the present state of physics. Whereas the Grassmanns defended the
ideal of a pure mathematics, Helmholtz main concern was to clarify the empirical
usefulness of mathematics. Whereas Kries, Zeller, and Du Bois gave a prominent
role to psychology in their discussion of measurability, Helmholtz refrained from
any reference to psychology.
This last contrast calls for comments. The first global discussions of measurability
in science were motivated by the rise of empirical psychology. Although physicists
and astronomers had discussed methods of measurement, they did not address the
more fundamental question of measurability. The most thorough analysis of the
measurability of physical properties before Helmholtz, Kries’s memoir of 1882, was
meant to prepare the reader to the rejection of the measurability of sensations on
Kantian grounds. As Michael Heidelberger (1993) has suggested, Helmholtz prob-
ably had psychological measurement in the back of his mind when he wrote ‘Zählen
und Messen’. That he still remained silent on this issue may be regarded as a conse-
quence of a trait documented by Edward Jurkowitz (2002): his general reluctance
to enter controversies without decisive empirical arguments in his hands.
Helmholtz’s external sources are discussed in the first section of this paper. The
second section is devoted to the contents of Helmholtz’s essay, to their relation to
the sources, and to their connection with Helmholtz’s earlier reflections on the foun-
dations of geometry. A critical consideration of the parallel he perceived between
geometry and arithmetic will help to identify the empiricist elements of ‘Zählen und
Messen’ and the resulting departure from Kantian orthodoxy.
The imperfection of this parallel, and various details of Helmholtz’s essay suggest
that he heavily relied on the above-mentioned external sources to shape his argu-
ments. The disparity of these sources implies that he could only have selectively
518 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

borrowed from them. From the Grassmanns, he took the inductive derivation of all
axioms of arithmetic from a single axiom. From Du Bois, he may have borrowed the
general idea of giving characteristic properties of quantity, and empiricist concepts of
cardinal and fractional numbers. From Maxwell he plausibly borrowed the empirical
interpretation of the transitivity of equality.
Helmholtz articulated these imported elements in an original manner. His overall
conception of counting and measuring differed widely from those of Grassmann, Du
Bois, Kries, and Elsas. His empiricism obviously contradicted Grassmann’s formal-
ism and Kries’s and Elsas’s narrow Kantianism. It also differed from Du Bois’s
phenomenology: whereas Du Bois derived number from quantity, Helmholtz defined
number before quantity and inferred his concept of quantity from arithmetic.
The reception of ‘Zählen und Messen’ is the subject of the third part of this paper.
Helmholtz’s essay was published at a time when several eminent mathematicians
sought a logically impeccable foundation for arithmetic, itself regarded as the basis
or at least a model for the rest of mathematics. These mathematicians, including
Richard Dedekind, Georg Cantor, and Gottlob Frege, had a poor opinion of
Helmholtz’s arithmetic, and no interest in his theory of quantity. They regarded
numbers as logical constructs in a set-theoretical framework. In their eyes, Helmholtz
had regressed to an unclear psychologico-empiricist view of arithmetic.
Out of interest for physics and geometry, some other mathematicians wanted a
theory of quantity that did not reduce itself to arithmetic. Those were much more
favorable to Helmholtz’s views. Henri Poincaré’s discussion of the mathematical
continuum, and his discussions of measurement in various domains of physics were
clearly reminiscent of Helmholtz’s paper. Also, Poincaré based all mathematical gen-
eralizations on mathematical induction regarded as a synthetic a priori judgment, in
harmony with Helmholtz’s connection of this principle with internal intuition.
Another admirer of Helmholtz’s views was Otto Hölder, who produced a refined
version of Helmholtz’s concept of measurement in which quantity was defined in a
rigorous mathematical manner and measured in real numbers.
Helmholtz included his considerations on counting and measuring in the introduc-
tion to his lectures on theoretical physics. Other physics teachers did not follow him.
To this day, introductory physics texts quickly define measure as the ratio between
quantity and unit, and move on to discuss systems of units or errors of measurement.
Yet three influential philosopher-physicists, Ernst Mach, Pierre Duhem, and Norman
Campbell, addressed the question of measurability. Mach applauded Helmholtz’s
empiricist view of arithmetic and his measurability criteria, which he had partly
anticipated in his own reflections on the foundations of mechanics. Duhem also held
an empiricist, “common sense” view of the foundations of mathematics, and he faith-
fully reproduced Helmholtz’s definition of quantity. Campbell followed Helmholtz
even more closely, despite his Russellian emphasis on the ordinal properties of quan-
tity.
A running theme in this paper is the opposition between a narrow concept of
measurement in which the additivity and divisibility of quantities is required in a
concrete sense, and a more liberal concept in which the ordering of quantities is the
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 519

only requirement.3 Helmholtz clearly defended the narrow concept, which goes back
to the Greek theory of ratios. In his view intensive quantities, for which no concrete
addition is yet known, could only be measured through a law-like connection with
extensive quantities. Most of Helmholtz’s predecessors agreed on this point, even
psychologists like Wundt. The dividing issue between experimental psychologists
and their Kantian critics was not whether quantity should be additive, but whether
a natural addition could be found for psychological quantities.
The first proponents of the liberal concept of measurement were physicists who
discussed the basis of thermometry, Maxwell, Mach, and Duhem. They all regarded
thermometers as purely conventional means to identify and order thermal states.4 In
their opinion, the equality of two degrees of a given thermometer, and therefore the
addition of two temperatures, had no intrinsic meaning. Temperature was a quality,
artificially but usefully represented by numbers. As we will see, Mach and Duhem
had epistemological reasons to defend the numerical scales of qualities. In particular,
they believed that the confusion between quality and quantity belonged to the mech-
anical reductionism they both condemned.
The mathematician-philosopher Bertrand Russell defended a similar view, but for
different reasons. He had a mathematical interest in ordinal structures for which no
addition is defined, and a psychological interest in quantities such as pleasure that
have degree without being measurable. He had read and admired the Austrian philos-
opher Alexius Meinong, who defended the measurement of sensations against Kries’s
attack. In Russell’s view, for sensations and other intensive quantities, all useful
quantitative information was already contained in the ordinal definition of quantity.
The only function of numerical scales was to ease the access to this information.
Russell could not be further from Helmholtz’s view that quantity implied additivity,
which he did not even mention.
The longer term reception of Helmholtz’s paper is not discussed in this paper.
The following briefs remarks are based on Joel Michell’s relevant studies (1993,
1999) and some personal observations. In psychology, S. S. Stevens’s liberal concep-
tion of measurement (1946) which defines measurement as ‘the assignment of
numerals to objects or events according to rule’ has become dominant, with a few
notable exceptions including Michell himself. In physics, measurability is usually
not discussed, save for the specific difficulties encountered in quantum mechanics.
The modern ‘theory of measurement’ truly belongs to philosophy of science. It was
created under the guidance of a few philosophers with mathematical competence,
including Ernst Nagel, Brian Ellis, and Patrick Suppes. This literature often cites
Helmholtz and Campbell as outstanding precursors.5
Yet the nature of the conceptual filiation between the modern theory of measure-
ment and Helmholtz’s paper is not obvious. In 1993, Joel Michell addressed this

3
This distinction is still a matter a discussion, see for example Savage & Ehrlich (1992), and Mich-
ell (1999).
4
On the earlier evolution of thermometry toward this view, see Chang (2001).
5
A standard formulation of the modern theory of measurement is found in Krantz et al. (1971). For
a concise presentation, see Luce & Suppes (1981). On the origins, see Diez (1997) and Michell (1999).
520 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

issue in terms of the dichotomy between ‘classical’ and ‘representational’ concepts


of measurement.6 In the classical concept, measurement is the comparison between
a quantity and a unit of the same quantity; numbers are defined as the results of
measurement. In the representational concept, numbers are defined without reference
to quantity; quantity is defined without reference to number; and measurement is a
correlation between quantity and number. Since for Helmholtz numbers are still
defined empirically and measurement is still defined in terms of ratios, Michell tenta-
tively situated him at the end of the classical tradition, and made Russell the true
founder of the representational concept.
Although this view has the merit of clarity, it may not adequately represent the
historical significance of Helmholtz’s contribution. As Michell himself noted,
Helmholtz did not think in terms of the classical/representational dichotomy. If,
somewhat artificially, one insists in applying these categories to late nineteenth-cen-
tury philosophies of quantity, then ‘Zählen und Messen’ should be regarded as a
transitional contribution. Besides the above-mentioned classical aspects, Helmholtz’s
essay also has representational aspects: a definition of (ordinal) numbers independent
of measurement, and explicit criteria of measurability that suggest axioms of quan-
tity. Plausibly, these features explain why the founders of the modern theory of
measurement consistently regarded Helmholtz as a pioneer of their approach.
Scientists normally ignore this philosophers’ theory, despite its Helmholtzian ori-
gins. The reason may be that criteria of measurability à la Helmholtz probably never
served to identify new quantities, at least in the case of physics. The natural pro-
cedure is to tentatively regard some property as a quantity for some theoretical,
metaphorical, experimental, or sensorial reason, and then to try to measure it by
whatever means this reason suggests. If the numbers yielded by this measurement
are found to be related to other measured quantities in a regular manner, the measur-
ability criteria for equality are implicitly met, because these numbers would otherwise
be ill-defined. If the resulting laws are simple and general, some other criteria such
as those of concrete addition are likely to be met. In other words, the global testing
of a theory by quantitative experiments entails an implicit test of the measurability
of relevant quantities.

1. Sources

1.1. Grassmannian formalism

Helmholtz defined the problem of measurement as the identification of physical


operations that share structural properties of (whole) numbers. He therefore needed
a preliminary discussion of the foundations of arithmetic. To this end he selectively

6
This distinction is not quite the same as the distinction between narrow and liberal concepts of
measurement. Of course, the classical category excludes the liberal concept. However, Mach and Duhem
defended the liberal concept without yet belonging to the representational category, since they defended
an empirical view of arithmetic. Also, a representational theory may belong to the narrow concept if its
axioms of quantity allow for ratios.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 521

borrowed from the Grassmann brothers. Some familiarity with the Grassmannian
conception of mathematics is necessary in order to understand the nature of this
connection, and also as a contrast to another of Helmholtz’s sources, Du Bois’s
theory of functions.
In 1844, Hermann Grassmann, a mathematics school teacher in Stettin, Pomerania,
published the Ausdehnungslehre (doctrine of extension) that won him belated fame.
This profoundly original work has a double origin. On the one hand, Grassmann
wanted to generalize the relation AC = AB + BC between the segments defined by
three consecutive points A, B, C on a line to three arbitrary points in space. This
led him to the concept of vectorial addition. On the other, he inherited from his
father Justus a strange analogy between arithmetic and geometry.7
In this analogy, both numbers and geometric objects are engendered (geworden)
in thought. Numbers result from the repeated connection (Verknüpfung) of a unit e
with itself, as in 3e = e + e + e, while lines result from the motion of a point. The
product of two numbers is obtained by repeated connection of the first number
regarded as unit, for instance 3 × 2 = 2 + 2 + 2. Similarly, Justus Grassmann regarded
a rectangle as the product of its sides, since it is engendered by the motion of one
side along the other.8
Hermann Grassmann generalized the latter product in harmony with his own idea
of vector addition, and thus obtained the Grassmannian product xy of two segments
x and y as the corresponding parallelogram. In order to preserve the generality of
the relation x(y + z) = xy + xz, he oriented the parallelogram xy and replaced ordinary
commutativity (xy = yx) with anti-commutativity (xy = –yx).9 His initial hesitation
to admit this funny property disappeared when he realized how fit the new product
was to express the general theorems of mechanics.
Grassmann then worked out the properties of the new product, as well as a related
product of points (AB being a vector, ABC an oriented parallelogram, and so on).
He thus obtained an abstract theory of extension, far more general than the geometry
of the natural world. ‘The essential advantage of this conception’ (Grassman, 1844,
p. VII), he commented,

. . . is, with respect to form, that now all axioms [Grundsätze] that expressed
spatial intuitions completely disappear and thus the starting point becomes as
immediate as in arithmetic; and, with respect to content, that the limitation to
three dimensions disappears.10

7
Grassmann (1844), pp. III–IV. Cf. Crowe (1987); Flament (1994); Châtelet (1993).
8
Cf. Crowe (1987), p. 59.
9
The square products (x + y)(x + y), xx, and yy vanish since the corresponding parallelograms do.
Distributivity implies (x + y)(x + y) = xx + xy + yx + yy. Hence xy + yx must vanish.
10
Grassmann (1844), p. XXIII, distinguished between geometry, based on intuition in a Kantian sense
(eine Grundanschauung, die mit dem Geöffnetsein unseres Sinnes für die sinnliche Welt uns mitgegeben
ist), and the theory of extension, constructed without reference to our ability to perceive geometric struc-
tures. From which we may infer that unlike Kant he refused to base (pure) mathematics on synthetic a
priori judgments. In particular, he regarded arithmetic as purely analytic, whereas Kant regarded it as a
form of inner intuition.
522 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

As Grassmann believed to have created an entirely new domain of mathematics, he


decided to explain the underlying philosophy of mathematics. He first distinguished
between real and formal sciences. The real ones deal with thinking applied to the
real world. The formal ones deal with the mutual relation and connection of thought
processes; they are based on definitions whereas the real sciences are based on axi-
oms (Grundsätze). Among the formal sciences, he distinguished logic, defined as the
study of the general laws of thought, and pure mathematics, defined as the ‘science of
particular being (besonderes Sein) as being transformed (geworden) through thought’
(Grassman, 1844, p. XXII). Accordingly, he renamed pure mathematics as Formen-
lehre, or theory of the forms of thought (Denkformen). The transformations implied
in thought processes were of two kinds: discrete connection (Verknüpfung), and con-
tinuous setting (setzen). The first kind led to arithmetic and the theory of combi-
nations, the second to the theory of functions and to the new theory of extension.11
Grassmann associated this four part division of pure mathematics with their inte-
gration in the common frame of a general theory of forms (allgemeine Formenlehre).
On his path to the theory of extension, he had assumed a deep analogy between
arithmetic and geometry, as well as stability of some of the basic properties of
addition and multiplication. Accordingly, he believed that general laws of connection
(Verknüpfungsgesetze) were common to the four branches of pure mathematics. In
order to avoid repetition, he expounded them in full generality before applying them
to the theory of extension. They were essentially the properties of equality (such as
symmetry and transitivity), those of two composition laws (such as associativity,
commutativity, and distributivity), and their consequences in transforming mathemat-
ical formulas. The objects of these relations and operations were purely abstract.
They were denoted by signs or letters the referent of which did not matter.12
Hermann Grassmann’s Formenlehre encompassed all branches of pure mathemat-
ics, but left logic out. His brother Robert wanted to cover all the formal sciences,
including logic, with a similar formal apparatus of composition laws. He called this
framework Grösenlehre [sic], in homage to Leibniz’s sketch of a Scientia de magni-
tudine and to his father’s critical allusion to a generic theory of magnitudes. In 1847,
the two brothers began a collaboration on this project, but soon retreated to the more
manageable branches, the theory of numbers (arithmetic) and the theory of extension.
Their special interest in the former no doubt stemmed from its role as a model for
the latter.13
In late nineteenth-century parlance, Hermann Grassmann had arithmetized
geometry to reach a general, abstract theory of extension. Yet the brothers did not
regard the received arithmetic as a perfect model of mathematical reasoning. In their
opinion, earlier arithmeticians unconsciously relied on intuition where the rules of
mental connection should suffice. In particular, they assumed the associativity of

11
Grassmann, Einleitung (1844), pp. XXI–XXXIV.
12
Grassmann (1844), pp. 1–9.
13
Cf. R. Grassmann (1890c), pp. III–VII.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 523

addition as an axiom, whereas it could be derived on more basic premisses. As


Robert later remembered:

The Grassmann brothers saw that . . . for the sake of scientific strength one should
generate every magnitude [Gröse] only in one manner, most simply by successive
connection [Knüpfung] with the preceding ones, and also that every strong proof
should be performed in a progressive manner, in such a way that a theorem should
hold for a + 1 when it holds for a. This was the most significant result of their
collaboration at that time.14

As will be explained in a moment, the Grassmann brothers constructed numbers by


iterated connection of a single unit or element; they defined operations and derived
their properties by mathematical induction.
Later on the two brothers worked separately, but with frequent exchanges. In 1861
Hermann published his arithmetic, which embodied the approach summarized in the
above sentence. In 1862 he offered a new version of his Ausdehnungslehre through
which he hoped to put an end to the general neglect of his treatise of 1844. Mean-
while Robert was busy with local and national politics, with theological writings,
and with the press and newspapers he run in Stettin. He returned to the Grösenlehre
in 1870. While working out logic and the theory of combinations (which he regarded
as the two parts of the ‘logical sciences’), he became convinced that the basic prin-
ciples of definition and demonstration where the same as those of his brother’s arith-
metic. This insight explains the wide scope of his Formenlehre of 1872.15 Robert
Grassmann started with two basic definitions:

A quantity [Gröse] is everything that is or can be the object [Gegenstand] of


thinking, in so far as it has only one, and not several values. The connection
[Knüpfung] of two quantities is every placing together or binding of these quan-
tities that is accessible to human thought, in so far as it has only one, and not
several values.16

R. Grassmann insisted on the univoqueness of quantities and connections, which


derived from his earlier linguistic studies. With his brother, he had co-authored man-
uals of German and Latin grammar. He believed that the Sprachlehre (linguistics)
necessarily preceded the Formenlehre, because a ‘philosophical grammar’, revealing
universal structures of thought could emerge from the comparison of natural langu-
ages. At the same time, he attributed the pitfalls of non-rigorous thought to the
pervasive ambiguities of words in any natural language. Universality and rigor could
only be achieved by the artificially constructed Formenlehre.17
Among the quantities, R. Grassmann distinguished the ‘elements’, which cannot

14
Ibid., p. VII.
15
Cf. ibid., p. VIII.
16
Grassmann (1872), pp. 7–8.
17
Cf. ibid., pp. IX-XII; R. Grassmann (1890a), p. XXVI on the philosophische Grammatik.
524 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

be obtained by connection of other quantities. All other quantities are obtained by


concatenation, that is, by a finite number of connections of a finite number of
elements. For a given connection, denoted ⴰ, and for two elements e and e⬘, an
example of a quantity is ((eⴰe⬘)ⴰ(eⴰe))ⴰe⬘. The order of connection, indicated by the
parentheses, matters in general. It becomes irrelevant if the composition has the
property of associativity (Einigung): aⴰ(bⴰc) = (aⴰb)ⴰc. In conformity with his bro-
ther’s treatment of arithmetic, R. Grassmann refused to regard this property as basic.
Instead, he defined associativity by the fundamental formula (Grundformel) aⴰ(bⴰe)
= (aⴰb)ⴰe, where e is any element. Then he proved by induction that general associa-
tivity resulted from this formula. Similarly, he defined commutativity (Vertauschung)
by the fundamental formula eⴰe⬘ = e⬘ⴰe for any two elements e and e⬘, together with
the fundamental formula of associativity. Again, the property aⴰb = bⴰa for any two
quantities a and b results from this definition by induction.18
R. Grassmann further introduced the notion of ‘orders of connection’ (Ordnungen
der Knüpfung), that is, a succession of mutually related composition laws, Fügen,
Weben, and Höhen, meant to generalize addition, multiplication, and exponentiation.
For example, the generalized multiplication had to be related to the generalized
addition by the fundamental formula of distributivity (Beziehung), a(b + e) = ab +
ae and (a + e)b = ab + eb, which implied general distributivity by induction.19
After developing the consequences of these definitions through the general prin-
ciple of induction, R. Grassmann turned to the various branches of the logical and
mathematical sciences. The composition properties of the elements characterize these
branches. R. Grassmann’s logic corresponds to e + e = e, ee = e, and ee⬘ = 0 for
any two elements e and e⬘ (0 being the quantity such that a + 0 = 0 + a = 0 for any
a); his theory of combinations to e + e = e, and ee = 0 (combinations without
repetition) or ee non-zero but still different from e (combinations with repetition)
for any element e; arithmetic to a single element for which the addition is associative
and commutative; the theory of extension to ee = 0, ee⬘ + e⬘e = 0.20
The spirit of Robert Grassmann’s Formenlehre is best seen on the arithmetic he
shared with his brother Hermann. He first introduced the concrete numbers
(Zahlgrösse or benannte Zahl) e + e + . . . + e obtained by concatenation of the same
element e (the parentheses may be dropped since associativity is assumed). Then he
introduced the (pure) numbers 1 + 1 + . . . + 1 obtained by concatenation of the unity
(die Eins) 1 such that 1·a = a·1 for any quantity a.21 As an illustration of the
Grassmanns’ use of induction, we may consider their deduction of associativity of
addition. Suppose that the property (a + b) + c = a + (b + c) is known to be true
for a given c (and for any a and b). The Grundformel (a + b) + 1 = a + (b + 1)
implies ((a + b) + c) + 1 = (a + b) + (c + 1) and (a + (b + c)) + 1 = a + ((b + c)

18
R. Grassmann (1872), pp. 20–21, 33.
19
Ibid., pp. 22–25.
20
Ibid., p. 13. Just to give an idea of what he had in mind in the case of logic, an example for e +
e is ‘Peter and Peter’, for ee ‘a stone that is a stone’, for ee⬘ ‘a stone that is a mushroom’.
21
Ibid., p. 10. In the foreword to his Arithmetik, Hermann Grassmann described it as ‘in its essential
features, a work done in common with my brother Robert’ (Grassmann, 1861, p. V).
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 525

+ 1) = a + (b + (c + 1)). Therefore, the property also holds for c + 1. As the property


obviously holds for c = 1, by induction it holds for any c.22
To sum up, the Grassmann brothers developed a formal theory of thinking as a
rigorous foundation for arithmetic, for the theory of extension (Hermann), and even
for logic (Robert). They believed that to be rigorous, mathematics had to be pure,
devoid of any recourse to experience or intuition. As arithmetic was closest to this
ideal, Hermann designed his theory of extension by analogy with it. He and Robert
then reworked arithmetic to make it perfectly pure. This implied a constructive,
generative conception of numbers. Robert further provided a general theory of the
composition of abstract quantities that embraced all branches of logic and mathemat-
ics. In this Formenlehre as in Hermann’s arithmetic, the basic constructive tool was
mathematical induction.
Mathematicians long ignored Hermann Grassmann’s Ausdehnungslehre, for they
were deterred by its philosophical outlook. Yet by the 1880s the importance of this
text was widely recognized, as were Hermann’s contributions to physics and physi-
ology. Robert’s Formenlehre was far less successful. It was part of a multi-volume
‘monument of knowledge’ (Gebäude des Wissens) which may never have been pub-
lished, had not the author run his own press. Robert’s boastful attitude, with his
ambition to be a new Aristotle, were not likely to attract the sympathies of sober
intellectuals. The Neue deutsche Biographie ignored him, and he now is a completely
forgotten figure.23
Yet Robert Grassmann had the satisfaction of inspiring an influential expert on
arithmetic and logic, Ernst Schröder, who counted Bertrand Russell and Charles
Sanders Peirce among his admirers. In arithmetic, Schröder (1873) acknowledged
and extended the Grassmanns’ criticism of the axioms. In his logic (Schröder, 1877,
pp. III, V), he preferred Robert Grassmann’s approach to George Boole’s (of which
Robert was unaware), for in his opinion the latter had followed arithmetic too closely
by allowing operations that had no direct logical meaning.24 His comments on
Grassmann’s work were ambiguous but altogether sympathetic: ‘Although this book,
with its mania of germanizing, is hard to enjoy, it contains, besides some strange
things, much worth consideration’ (Schröder, 1890–1895, p. 456).

1.2. Du Bois’s phenomenology

In 1882 a then renowned expert on the theory of functions, Paul Du Bois-


Reymond, expressed views on the foundations of mathematics that were antagonistic
to those of the Grassmann brothers. In his work on trigonometric series and through
his study of the asymptotic behavior of increasing functions, Du Bois had long been

22
Hermann Grassmann (1861, p. 4) regarded the formula a + (b + 1) = (a + b) + 1 as the definition
of the sum of a and b. Indeed, if a + 1 is defined as the successor of a, a + b is defined inductively
through this formula.
23
On Robert’s ambitions, see his autobiography (1890), pp. XIX–XXVIII; Grassmann (1894–1911),
Vol. 3, pp. 132, 225.
24
Yet Schröder preferred a set-theoretical approach to Grassmann’s method of irreducible elements.
526 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

aware of difficulties in the foundations of analysis. As he well knew, his egregious


colleagues Gustav Lejeune-Dirichlet, Bernhardt Riemann, and Karl Weierstrass had
exhibited ‘mathematical monsters’ (e.g. functions everywhere continuous and
nowhere differentiable) that contradicted earlier convergence theorems. The concept
of limit thus appeared to be problematic.25
Around 1870, several mathematicians, including Weierstrass, Dedekind, Cantor,
and Eduard Heine reacted to this situation by new theories of the argument of func-
tions, that is, by rational constructions of irrational numbers. Du Bois agreed that
the status of irrational numbers was at stake, for they were in essence limits of
rational numbers. Yet, for reasons to be explained in a moment, he did not believe
that the new constructions of irrationals solved the basic foundational problem.26
Du Bois’s own approach was a ‘metaphysics and theory of the fundamental con-
cepts of mathematics: quantity, limit, argument, and function’ (Du Bois-Reymond,
1882, subtitle). In his opinion, the basic difficulties of analysis were not in the rigor
of demonstration—the judgment of which he left to the likes of Boole and
Schröder—but in the starting point of the chains of demonstrations: the initial rep-
resentations (Vorstellungen) and derived concepts (Begriffe). In his usual convoluted
style, Du Bois defined representation as ‘everything that becomes conscious, that
can be object of memory, and is such in the condition in which it is received by
our memory. This notion naturally includes all that which appears to our conscious-
ness through thinking based on memories’.27
At first glance, Du Bois’s Vorstellung seems to be the same as Robert Grassmann’s
Gröse, defined as ‘unambiguous object of thought’ (Grassmann, 1872, p. 8). There
are however, essential differences. In his developments, Du Bois makes clear that
representations originate in perception (Wahrnehmung), both external (physics) and
internal (psychology). In contrast, Grassmann’s quantities can be any object of
thought. Their structure or content is irrelevant. Only their connection (Knüpfung)
matters to the construction of mathematical concepts.
Du Bois’s idea of the genesis of concepts (Begriffe) is radically different, and
inspired by his training in physiology under Johannes Müller:

Concepts emerge when commonness (Gemeinsames) in a group of representations


stimulates and captivates our attention. The differentness (das Verschiedenartige)
of these representations fades away, or, as Johannes Müller says, it is blacked out
by our consciousness. When the concept, that is, the commonness, thus remains,
it adheres—so have I observed—essentially to the word or sign with which langu-
age or science denotes the concept.28

25
Du Bois (1882), p. 10. Cf. Fischer (1981).
26
On the new theories of real numbers, cf. Pringsheim (1898); Dugac (1976), p. 35 and further refer-
ences there.
27
Du Bois (1882), pp. 11, X.
28
Ibid., p. 17.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 527

Du Bois applied this idea of concept formation to the desired clarification of the
concept of mathematical quantity (mathematische Grösse), as follows:

The right way is to pursue this concept in the different domains of knowledge in
which it conceivably occurs and to determine its commonness as it shows itself.
We will thus soon enough be led to a fundamental form of the concept of math-
ematical quantity that not only reigns over the external world of perception but
also extends far into mental life.29

Du Bois rejected ‘diplomatic definitions’ of mathematical quantity such as ‘common


property of things of different sort with respect to which they are numerically compa-
rable, such as their length or weight’. Instead he announced that he would survey
all fields of physical and psychological knowledge and thereby derive a more precise
and more relevant definition of mathematical quantity. Analysis would thus retrieve
‘its natural soil: natural sciences’.30
This approach may be called ‘phenomenological’ for it is based on abstraction
from the perceptual world. ‘Empiricist’ should be avoided, for Du Bois used the
word in a more specific sense, to be described in a moment. There is, however, a
clear connection between Du Bois’s conception of the foundations of mathematics
and the empiricist and mechanistic character of the physiology of perception culti-
vated by Johannes Müller and his disciples, to whom Emil Du Bois-Reymond and
Hermann Helmholtz belonged.31
Mathematical quantity, Du Bois goes on, may be discrete or continuous. In the
discrete category, we have the concept of number (Anzahl), defined as ‘that which
remains in our mind when everything that distinguishes the [contemplated] objects
vanishes and only the awareness of their being separated is retained’. Du Bois adds
that ‘the communication of this concept is done by means of words or signs, through
which the numbers are represented [dargestellt], and which are called numerals [Zah-
len]’. For small numbers, up to about seven, numbers are directly perceptible. The
precise definition of higher numbers, and the idea that their sequence is unlimited
requires a numeration system, designed in such a manner that ‘the next numeral
represents a number with one more object than the preceding numeral’. Du Bois
thus distinguished between cardinal and ordinal numbers, the latter being a system
of signs built to ‘measure’ the former.32
As the main kind of continuous quantity, Du Bois introduced the ‘linear quantit-
ies’, which he provisionally defined as mathematical quantities such that

29
Ibid.
30
Ibid., pp. 14–15.
31
Cf. Holmes (1994).
32
Du Bois (1882), pp. 16, 19.
528 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

they can be reduced to lengths; their differences, parts, and multiples are quantities
of the same kind, as is the case for lengths; they extend from the smallest to the
largest, as lengths do; they are comparable, measurable, as lengths are.33

He argued the paradigmatic role of lengths by noting that most measurements ended
with the reading of a pointer on a linear scale; that length measurements probably
were the first performed by mankind (besides counting); and that they constituted a
first step toward the mechanical understanding of nature: ‘The geometric measure-
representations are the origin and the constant refuge of exact thought, an assertion
that is not likely to meet any serious contradiction’.34
Du Bois then surveyed various fields of knowledge to identify what could count
as a linear quantity. His basic criteria were that the quantity should have sharply
defined values (in order to be a mathematical quantity), and that it should ‘vary by
degrees’ (dem Grade nach abgestuft). By the latter expression he not only meant
that two quantities of the same kind could be compared, but also that their difference
should be a quantity of the same kind.35
For lengths, this criterion is met by a simple physical operation on the measured
objects: a longer rod is obtained from a smaller rod by adjoining a rod piece to it.
Du Bois, however, did not necessarily associate a concrete physical operation with
this criterion. He only required that the quantity should be expressible in terms of
a quantity that was already known to be linear. For instance, he regarded temperature
as a linear quantity, because it represented the intensity of molecular agitation and
because the intensity of motion (kinetic energy) was already known to be a linear
quantity. More generally, he believed that all quantities pertaining to the external
world would ultimately be reduced to mechanical quantities, which were known to
be linear.36
Du Bois’s short survey of physics was followed by a much longer discussion of
the world of internal perception that is the object of psychophysics. There he found
‘a plentiful of extremely interesting kinds of quantities’. His eloquence on this subject
suggests that he may have moved from considerations of measurability in psychology
to the foundations of analysis rather than the reverse. Du Bois did not doubt that
some sensations, such as warmness, brightness or loudness, were linear quantities,
in so far as they met the first criterion of definiteness.37
The difference between a higher and a smaller sensation of the same kind, he
argued, still was a sensation of the same kind, because an accumulation of stimuli
of the same kind could yield a single sensation, higher than the individual sensation

33
Ibid., p. 23.
34
Ibid., p. 23.
35
Ibid., pp. 24–26. Du Bois remained unfortunately vague on the first criterion of definiteness. He
seems to have meant the existence of a method of comparison that gives stable results. For example,
he regarded lengths as well-defined, but degrees of pain as perhaps too subjective to be regarded as
mathematical quantities.
36
Ibid., pp. 23–27.
37
Ibid., p. 29.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 529

that the stimuli would individually have induced.38 Du Bois then praised Gustav
Fechner for his ‘first scientific attempt to measure sensations’, even though he saw
some arbitrariness in Fechner’s assumptions leading from Ernst Weber’s law to the
logarithmic measure of sensations (to be described in a moment). Du Bois similarly
argued for the linear character of mental dispositions (Stimmungen) such as wishes,
fear, anger, and pleasure; for he believed the relation between these and sensations
was similar to that between sensations and physical stimuli.39
After completing his survey, Du Bois offered a more precise definition of the
concept of linear quantity, to serve as a basis for analysis. He distinguished between
two kinds of axioms, those of comparison without measurement, and those of com-
parison with measurement. In the first kind, he had definitions of equality and order,
homogeneity, additivity and divisibility. In the second, he had the existence of differ-
ence and the Archimedean property. He did not try to minimize the number of axi-
oms, being content that they obviously applied to anyone’s concept of length. He
did not either try to eliminate reference to perception. For instance, his definition of
the equality of two linear quantities reads: ‘Their sensorial appearances produce the
same impression under the same circumstances’.40
Du Bois’s axioms permit the insertion of any quantity of a given kind between
two consecutive multiples of a standard quantity (unit) of the same kind. They also
permit the introduction of sub-units obtained by uniform division and their use for
refined measurement. In this combination of the concepts of (whole) number and
linear quantity (or length), Du Bois saw the plausible origin of fractional numbers.
He also believed that arithmetic operations on these numbers arose from the necessity
of measurement. Plausibly, at some stage men needed to know the area of a rectangle
with fractional sides: ‘One almost sees them, the settling hunters, as they now cease
their quarrels, and henceforth divide their land property according to this rule’.41
As Greek geometers proved long ago, geometric constructions lead to lengths the
ratio of which cannot be expressed as a fractional number. Fittingly, Du Bois axioms
of linear quantity did not require any two quantities of the same kind to be commen-
surable.42 They only permitted an indefinitely refined approximation of their ratio
by a sequence of rational numbers, for instance a decimal expansion. The concept
of linear quantity thus led to the question of the existence of limits representing
irrational numbers, the very question that motivated his whole enquiry.
Of recent constructions of irrationals, Du Bois certainly knew one, that of Eduard
Heine, based on Cauchy sequences and inspired by private hints by Cantor. In mod-
ern terms, Heine’s method, similar to Charles Méray’s, was to define a (real) number
as an equivalence class of Cauchy sequences, two sequences being equivalent when
the difference of their terms goes to zero. Heine then discussed arithmetic operations
on such classes, and showed that a subset of the new entities was isomorphic to the

38
This fallacious reasoning presupposes the linearity of the relation between sensation and stimulus.
39
Du Bois (1882), pp. 28–33.
40
Ibid., pp. 44–45.
41
Ibid., p. 53.
42
They are in this respect similar with Euclid’s axioms of quantity in the fifth book of the Elements.
530 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

set of rational numbers. This constructive approach, which quickly came to be


regarded as a successful arithmetization of analysis, failed to impress Du Bois. He
believed that Heine and consorts had only swept the difficulties under the rug, for
they no more proved the existence of the new objects (here classes of equivalence
of unlimited sequences) than previous analysts proved the existence of limits.43
There was perhaps one more reason for Du Bois’s lack of interest in Heine’s
construction. He knew and emphasized that the convergence of Cauchy sequences,
the existence of a limit for an increasing bounded function, and the convergence of
decimal expansions were all equivalent statements. He preferred to reason on the
most elementary of these statements, that regarding decimal expansions.44 The rep-
resentation of irrational numbers by such sequences is unique, unlike that by Cauchy
sequences. Therefore, in this representation there is no need for equivalence classes.
The only question is whether an unlimited sequence of decimals may be regarded
as always defining a number.
To his own puzzlement, Du Bois arrived at two antagonistic point of views from
which the existence of limits was equally well founded. The first is ‘idealism’, which
admits ‘the real existence not only of the representations, but also of the intuitions
(Anschauungen) that follow from the representations in a non-arbitrary manner’. One
of these intuitions is that of an infinite sequence of decimals, conceived as a whole.
In other words, the idealist mind is not only able to form the representation of finite
sequences of arbitrary lengths, it also has the intuition of an infinite sequence.45
Having thus admitted actual infinity, the idealist conceives an infinite number of
points in the segment [0, 1]. Moreover, the indefinite divisibility admitted in Du
Bois’s axioms of linear quantity can be pursued to infinity, thus yielding a concept
of infinitely small segments. According to the same axioms, these segments must
have the same linear structure as finite segments. They can be again submitted to
infinite division, which yields infinitesimals of second order, and so forth. Of course,
Du Bois recognized that the combination of infinitesimals of different order obeyed
special laws, different from that of regular addition and multiplication.46
Far from regarding actual infinity as a daring supposition, Du Bois compared it
to Helmholtz’s admission that our sensations were signs of a reality (Wirklichkeit),47
or to our willingness to imagine an ‘impenetrable machinery behind the scenes of
the world of appearances’. Yet he defended with equal energy the empiricist system,
which he defined as follows:

43
Du Bois (1882), p. 55; Heine (1872). Cf. Pringsheim & Molk (1904).
44
Du Bois (1882), pp. 56–57.
45
Ibid., p. 87.
46
Ibid., pp. 58–86. Cf. Fischer (1981), pp. 113–114.
47
According to Helmholtz’s theory of perception, sensations are signs for properties of external objects,
as letters are signs for certain sounds. The resulting sign systems have no resemblance with the objects
they represent. Their perceptual meaning is acquired (consciously or unconsciously) in the course of our
experience of the world.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 531

Everything that is scientifically well founded, every reliable knowledge proceeds


from immediate perception, and must be reducible to the system of representation
corresponding to the perceptible. By perception, I mean all that reaches our con-
sciousness, not only sensorial perception, but also perception through our
thought process.48

In particular, Du Bois’s empiricist tolerates the notion of arbitrary exactness, which


is obtained by imaginary extrapolation of a sequence of representations; but he rejects
complete exactness. Accordingly, the segment [0, 1] is arbitrarily thin, but not infi-
nitely thin. A point on this segment, is defined as a portion of it that is as long as
the segment is thin.49
Then, for every choice of the precision (thinness of the segment), and for every
unlimited sequence of decimals, there is an integer for which any truncated sequence
having a number of decimals higher than this integer approximately describes the
same point. This is all the empiricist needs to develop analysis. For him limits and
irrational numbers exist only as an unlimited process of approximation. He reinter-
prets the usual theorems of analysis as statements involving only rational numbers
for a given, arbitrary precision.
Du Bois believed the empiricist and idealist conceptions to be equally permissible.
In his subsequent presentation of the theory of functions, he opted for a strange
compromise: ‘empiricist language with idealist proofs’. Presumably, he preferred the
empiricist language because it left the decision between empiricism and idealism
open (the idealist assuming more than the empiricist). And he preferred idealist
proofs because of their much greater conciseness.50
Du Bois assorted his phenomenological approach with a criticism of the competing
‘formalist’ or ‘combinatorial’ approach:

A purely formalist-literal skeleton of analysis, which would imply the separation


of numbers and analytical signs from quantity, would ultimately depreciate this
science—which in reality is a science of nature, even though it only admits the
most general properties of perceived things in its domain of enquiry—to a mere
play with signs where the writing signs are given arbitrary meanings, as is the
case for chess pieces and play cards.51

Du Bois did not deny the possibility of constructing arithmetic on the basis of
axioms and conventions, without reference to geometry and quantity. He even
acknowledged the superior exactness of the resulting form of demonstration.52 But
he knew that several different systems of arithmetic and analysis would then be
possible, whereas in his view there could only be one system, that which was determ-

48
Du Bois (1882), pp. 111, 114.
49
Ibid., pp. 111, 114.
50
Ibid., p. 156.
51
Ibid., pp. 53–54.
52
Ibid., p. 290.
532 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

ined through the intimate relation between number and quantity. ‘Who would doubt’,
he interjected, ‘that the Martians’ analysis, if they had any, would essentially be the
same as ours!’53 This conviction derived from his phenomenological credo:

Fundamentally, analytical quantities still belong to the world of perception. The


problems and methods of mathematics all serve . . . our drive to understand the
world in mechanical terms.54

In sum, Du Bois founded arithmetic and analysis on the quantitative aspects of


perception. His definition of numbers depended on the notion of measurable quantity,
which he derived from a survey of measurement in various domains of knowledge.
In such a view, the relevance of numbers and analysis to physics and psychology
is inscribed in their very construction. The issue of measurability pertains to the
definition of numbers as much as to their application.
Among Du Bois’s readers was Robert Grassmann, perhaps the most extreme mem-
ber of the category that Du Bois contemptuously called Combinatoriker. Grassmann
strongly reacted to this implicit attack of his and his brother’s ideal of a pure math-
ematics. He found only confusion in Du Bois’s definition of quantity, the more so
because he failed to distinguish between Du Bois’s provisional, heuristic definitions
and the firmer axioms found at the end of his survey of measurement. He strongly
rejected Du Bois’s concept of mathematics as a natural science based on sensorial
perception:

Herr Du Bois-Reymond completely ignores the purely formal character of math-


ematics. He completely ignores that mathematics themselves set their quantities
without any content and without any sensorial substratum and that mathematics
connect the quantities according to purely mental laws, without relying on the
experience of natural sciences.55

Grassmann’s disgust reached its peak when he came to Du Bois’s empiricist, with
the coarse segments and thick points: ‘In fact, it would be difficult to imagine a
weaker achievement in this field’.56
Grassmann’s judgment, inserted in one of his little-read volumes, had little chance
to hurt Du Bois’s reputation. More dangerous criticism came from major mathema-
ticians who promoted the arithmetization of analysis. Especially hostile was Richard
Dedekind, the intellectual heir of Gauss, Dirichlet, and Riemann.57
In his Zürich lectures of 1858, Dedekind introduced his celebrated definition of
irrational numbers as ‘cuts’ in the set of rational numbers. A cut is a partition of
this set into two subsets such that every element of the first subset is inferior to

53
Ibid., p. 128.
54
Ibid., p. 39.
55
R. Grassmann (1891), pp. 1–3.
56
Ibid., pp. 1–3.
57
Cf. Dugac (1976), pp. 54–57.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 533

every element of the second (for example, the rationals the square of which is
smaller, respectively larger than two). Dedekind introduced order, addition, and mul-
tiplication among the cuts; showed that a subset of them was isomorphic to the set
of rationals; identified the complementary set with irrational numbers; and derived
the standard convergence theorems of analysis (existence of a limit for every increas-
ing bounded function, convergence of Cauchy sequences).58
Dedekind published this construction only in 1872, after reading Heine’s. He
believed to have found ‘a purely arithmetic and perfectly rigorous foundation’ of
analysis. Mathematicians, according to Dedekind, create new numbers (negative,
fractional, irrational) in order to surmount the limitation in performing inverse oper-
ations (subtraction, division, extraction of roots). Both motivation and construction
are in essence arithmetic. When, ten years later, Dedekind read Du Bois’s
Functionentheorie, he saw in it ‘the most extreme antithesis to [his own] conception
of arithmetic and analysis’. His reactions are detailed in a manuscript that, fortunately
for Du Bois, remained unpublished.59
Like Grassmann, Dedekind strongly reacted to Du Bois’s derogatory remark on
‘a purely formalist-literal skeleton of analysis’:

The whole of analysis is a necessary consequence of thought per se; only after
its throughout pure development (without recourse to the quantity-representations)
are we in a position to conceive the concept of quantity with full exactness.60

Dedekind further denied Du Bois’s claim to have proved the convergence of decimal
expansions, be it in the idealist or in the empiricist mode. He denied the capacity
of Du Bois’s idealist to think logically, and refused to follow him in his ‘labyrinth
of unclear and muddled thought’. The only consistency he found in the empiricist’s
view was in his denial of everything exact. Du Bois’s motto ‘empiricist language,
idealistic demonstrations’, reminded him of ‘the statesman-like’ slyness of certain
politicians: ‘Hence liberal words, reactionary actions’. Lastly, Dedekind argued that
his cuts, which Du Bois did not even mention, were a more convenient basis than
decimal expansions for the derivation of convergence theorems.61
Cantor’s opinion of Du Bois was equally negative, despite Du Bois’s comments
praising Cantor’s concept of power of a set. In a letter to Felix Klein, he wrote of
‘a loose proof, à la P. Dubois’, and denied Du Bois’s paternity for the theorem of
intermediate values.62 In a letter to Giulio Vivanti, he charged Du Bois with having
stolen from Johannes Thomae the idea of the infinitary calculus ‘for the gratification
of his personal ambition and vanity’.63 To his friend Dedekind, he confided:

58
Dedekind (1872). Cf. Dugac (1976), Ch. 3.
59
Dedekind (1872), p. 2; ‘Characteristische Stellen aus [Du Bois (1882)]’, MS reproduced in Dugac
(1976), pp. 199–203.
60
Ibid., p. 199.
61
Ibid., pp. 200, 201.
62
Cantor to Klein, 25 Feb 1882, in Dugac (1976), pp. 164–165.
63
Cantor to Vivanti (1893), quoted in Fischer (1981), p. 116.
534 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

‘Dubois’s works are, for the most, worth nothing; and if he once found something
good, his methods are still bad’.64 Dedekind heartily approved. Reciprocally, Dede-
kind’s work on the foundations of arithmetic (to be discussed later), appeared ‘hor-
rible’ to Du Bois.65
The Göttingen authority Felix Klein later echoed Dedekind’s judgment:

In his Allgemeine Theorie der Functionen, Tübingen, 1882, Paul Du Bois-


Reymond incessantly plays the difference between the ‘idealistic’ and the ‘empiri-
cal’ conceptions (which he transfers to spatial intuition as such). However, he
does not take any positive mathematical direction, but loses himself in spelling
out the resulting antinomies.66

Unlike Dedekind, Klein admitted a heuristic role of geometric intuition in the con-
struction of real numbers. But he basically approved the arithmetizing trend:

Surely the motivation for the construction of irrational numbers is in the apparent
continuity of spatial intuition. Yet I cannot draw the existence of irrationals from
spatial intuition, as I do not regard it as exact. To me, the theory of irrationals
rather is something that must be founded or circumscribed in a purely arithmetic
manner, and that we then import into geometry thanks to the axioms, with the
precision in the distinctions that is a precondition of mathematical treatment.67

In his article of 1898 on irrationals and convergence for Klein’s encyclopedia,


Alfred Pringsheim judged that the only idea of Du Bois’s worthy of the mathema-
tician’s interest was that for someone who wanted to found the theory of numbers
on the concept of quantity the existence of limits could only be an axiom. Pringsheim
regarded Du Bois’s ‘psychological’ justifications of this axiom as void. He also noted
that most contemporary mathematicians had opted for one of the arithmetic defi-
nitions of real numbers, and thus ‘for a separation of the pure theory of numbers
from the theory of quantities proper’.68
Yet Du Bois enjoyed a strong reputation in his lifetime; he won the posthumous
admiration of a few major mathematicians; and he could now be regarded as a precur-
sor of non-standard analysis. In 1893, Henri Poincaré saluted his infinitary calculus—
a method for comparing the infinities of diverging functions—as a creative step
beyond the usual mathematical continuum. Emile Borel, Jacques Hadamard, Felix
Hausdorff, and Godfrey Harold Hardy admired and developed this calculus. Hardy’s
judgment is most instructive:

64
Cantor to Dedekind, 16 Feb 1882, in Dugac (1976), pp. 252–253.
65
Dedekind to Cantor, 17 Feb 1882, in Dugac (1976), pp. 253–254.
66
Klein (1902), p. 15.
67
Klein (1890), p. 572.
68
Pringsheim (1898), pp. 56–57, who credited Helmholtz for this separation.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 535

The ideas of Du Bois-Reymond’s Infinitärcalcül are of great and growing impor-


tance in all branches of the theory of functions . . . Du Bois-Reymond was a
mathematician of such power and originality that it would be a great pity if so
much of his best work were allowed to be forgotten . . . He is also at times exceed-
ingly obscure; his work would beyond doubt have attracted much more attention
had it not been for the somewhat repugnant garb in which he was unfortunately
want to clothe his most valuable ideas.69

1.3. Measurement in psychology

Du Bois included sensations and mental dispositions among linear quantities,


because he believed they could be compared and compounded. He was not sure,
however, that they were determinate enough to be treated mathematically, nor that
they could actually be measured. Around 1880, the question of the measurability of
psychological processes was frequently debated. It may even be said that Fechner’s
psychophysics, and the more recent empirical psychology of Wilhelm Wundt were
the main contexts for posing the question of measurability in general.
In the 1840s, Gustav Fechner was known as the Göttingen physics professor who
provided the most accurate verification of Ohm’s law for electric conduction. Out
of distrust for his ability to follow the growing mathematization of physics, he gave
up his chair and began the research that led to the foundation of psychophysics. But
he did not give up his concern with measurement. On the contrary he made it the
basis of a new ‘psychophysics’. Sensations, he argued, are quantitative because a
given sensation can be judged to be more, less than, or equal to another sensation
the same kind. But the measurement of a sensation seems to require the further
possibility of expressing a given sensation as a sum of smaller, equal sensations
(Summirung eines Soundsovielmal des Gleichen).70
In the case of a length, Fechner went on, measuring rods provide the desired
divisibility. There is no such thing for sensations, because their material substratum,
the psychophysical activity, is unknown. But we may try to measure sensations
indirectly, through the physical stimuli that produce them. At that point Fechner
noted a paradox: indirect measurement can only be performed if the functional
relation between sensation and stimulus is known; but this relation can only be
known if sensation and stimulus can be measured independently.71
Fechner escaped the paradox by assuming the differential form of this relation to
be known. If we know the increment of a sensation that corresponds to a small
increment of the stimulus, we may follow the following prescription for measuring
a sensation:

69
Poincaré (1893), p. 33; Hardy (1910), preface. Cf. Fischer (1981), pp. 138–158.
70
Fechner (1859), p. 55. Cf. Heidelberger (1993).
71
Fechner (1859), pp. 57–58. This is what Chang (1995), pp. 153–154 calls ‘the paradox of nomic
measurement’.
536 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

Divide every sensation in equal parts, i.e. equal increments from which it grows
from the zero-state; and regard the number of these increments as determined by
the number of the corresponding, variable stimulus increments that produce equal
sensation increments.72

The situation, Fechner commented, is similar to the measurement of the length of a


curve by the number of small equal straight rods that are necessary to cover the curve.
The remaining question is: how do we know the differential relation between
stimulus and sensation? Fechner regarded it as obvious that the smallest noticeable
increment of a sensation S was a constant k independent of the intensity of the
sensation. This assumption, together with the measurement of the corresponding
increment e of the stimulus E, provides the desired relation. Fechner knew from
Ernst Weber that in a wide range of intensity, e is proportional to E. For a small
but largely noticeable variation dS of the sensation, the above prescription yields
dS / k = dE / e ⬀ dE / E. Integration leads to Fechner’s law: S ⬀ lnE.73
This simple law, and more generally Fechner’s and Wundt’s measurements of
psychological processes proved highly controversial. A significant example is the
debate that occurred in 1881–1882 between Wundt and the Berlin professor Edu-
ard Zeller.
Zeller was a philosopher and theologian, widely admired for his Philosophie der
Griechen. For moral and theological reasons, he opposed the contemporary rise of
materialism, and, more generally, any attempt to treat mental processes as if they
were physical processes.74 In March 1881, during a joint meeting of the various
sections of the Academy of Berlin, he delivered a speech ‘On the measurement of
mental processes’ in which he tried to prove the impossibility of this kind of
measurement. He began with a characterization of measurement borrowed from
physics, as the comparison of two quantities of the same kind, one known and one
unknown, following which the unknown quantity can be told to be a certain multiple
of the known one.
For Zeller, the crucial element in this definition was not so much the structural
properties required for the comparison (as in Du Bois’s analysis), but the existence
of fixed standards. In physics, Zeller argued every quantity is measurable, because
it can be reduced to the mechanical quantities of mass, time, and length, which
themselves can be determined by purely spatial measurements (mass by the size of
a cube of water, time by the length of a pendulum) (Zeller, 1881, pp. 3–4). Zeller
went on:

Although all other changes in nature can be reduced to mechanical motions or


complexes of such motions and therefore allow a precise measurement on constant
quantities [standards], this is not the case for the phenomena of consciousness as

72
Fechner (1859), p. 60.
73
Ibid., p. 71.
74
Cf. Vischer (1887).
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 537

such . . . Psychic processes as such are not at all measurable in the same sense
as space or time quantities, mechanical motions, and forces, for they do not allow
invariable standards [Massstäbe] by which their speed, their strength, their equal-
ity or inequality could be determined.75

The main reason for this impossibility was the relativity of psychic states, or contrast
effects. As the intensity of a given feeling or sensation depends on the intensity of
simultaneous or slightly earlier psychic states, it can never be taken as a standard.
Curiously, Zeller supported this assertion with ‘a general psychological law’ (ibid.,
p. 15), the differential form of Fechner’s law (which he attributed to Weber): the
increase of a sensation (dS) is proportional to the relative increase of the stimulus
(dE/E). He thus seemed to base the non-measurability of psychic phenomena on the
measurability of sensations!
Zeller did not name any of the implicit targets of his criticism. The most evident
one, Wilhelm Wundt, nonetheless reacted (1883). The Leipzig professor of experi-
mental psychology promptly noted the contradiction inherent in Zeller’s use of Fech-
ner’s law. The source of Zeller’s confusion, Wundt argued, was the belief in a clear-
cut division between external and internal world. The external world, just as much
as the internal world, is made of representations. In psychology as in physics,
measurement is the comparison between representations of the same kind. Wundt
thus defended not only the measurement of peripheral psychological processes
(sensations), but also that performed in his laboratory on the most internal processes,
such as apperception and volition times.
To which Zeller replied that he only objected to direct measurement of psychologi-
cal processes (by comparison with other psychological processes), but not quantitat-
ive determinations that ‘were found only through deductions and calculations made
on the basis of [physical] measurements’ (Zeller, 1882, p. 296). Measurements per-
formed in Wundt’s laboratory were legitimate to the extent that they met this broader
definition of measurement. Zeller nevertheless opposed Wundt’s view that Fechner’s
law provided a measure of psychological processes. The only thing Fechner and
Weber has measured was the stimulus. Fechner’s definition of the equality or equid-
istance of sensations depended on introspection, which could not pass as a measure-
ment.
On the first point, Wundt replied (1883) that indirect measurement was as
important in physics as Zeller admitted it to be in psychology. He thus rejected
Zeller’s implicit hierarchy between physics and psychology measurement. Moreover,
he argued that direct measurement was also possible in psychology. One could not
call Fechner’s relation between sensation and stimulus a law, as Zeller himself did,
without admitting a direct measurement of the sensation. The ability of our con-
sciousness to appreciate the equality (Gleichheit) of two sensations (or of their
increments) sufficed for this task, because in physics too comparison was the essence

75
Zeller (1881), p. 5.
538 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

of measurement. The non-existence of a fixed standard did not make measurement


impossible, it only confined measurement to relative values.
Among the witnesses of this vivid but somewhat obscure debate was Johannes
von Kries, professor of physiology in Freiburg, and expert on the perception of colors
with a Kantian philosophical inclination. Kries, like Zeller, denied that sensations
were measurable. Yet he agreed with Wundt that Zeller’s main argument was invalid:
one could well imagine reasonably constant standards even for psychological pro-
cesses. The problem lay somewhere else, as Kries proceeded to show in a thorough
essay entitled ‘On the measurement of intensive quantities and on the so-called psy-
chophysical law’.76
Kries borrowed from Immanuel Kant his understanding of the distinction between
extensive and intensive quantities. Extensive quantities refer to spatial and temporal
extension, which pertain to our intuition of phenomena (space and time being the
external and internal forms of intuition). They are such that ‘the representation of
the parts precedes the representation of the whole’. In other words, they satisfy a
property of additivity. Intensive quantities pertain to actual perception. They refer
to the degree of sensations or of their physical stimuli. They have a zero, denoting
the absence of sensation, and their degree indicates their distance from this zero.
Unlike extensive quantities, they are conceived as a whole, their division into parts
is meaningless.77
With this distinction in hand, Kries first addressed measurement in physics. The
simplest physical quantities are the extensive quantities, length and time. In this case,
the equality (Gleichheit) of two intervals has a clear meaning. So does the assertion
that a first interval is a multiple of a second, by juxtaposition (Zusammensetzung)
of equal copies of the latter interval. The situation is quite different for intensive
quantities, for which the equality of parts is not a priori defined.78
Mass for instance only becomes a quantity after a specific convention
(Festsetzung), for instance that two masses are equal when equilibrium obtains in a
balance. This convention is in a sense arbitrary: we could as well decide that two
masses are equal when the same quantity of heat is required to raise their temperature
by one degree. It is chosen, however, so that ‘certain empirical laws, so far as they
have become known to us, hold strictly and without exception’.79 In the present case,
the equality of weights for two equal masses is assumed. Then mass measurement
is brought back to counting equal parts of the same substance.
Kries went on to show that most physical quantities were defined through relations
(definitions or laws) with the more primitive quantities of length, time, and mass.
He thus espoused the Gaussian ideal of absolute measurement, as well as the Kantian
idea that the only continuous quantities susceptible of measurement were length,
time, and derived quantities. Equality and proportionality of two exemplars of the
same quantity are well defined because they can be brought back to the comparison

76
Kries (1882). Cf. Heidelberger (1993), p. 147.
77
Kant (1787), Axioms of intuition, Anticipations of perception.
78
Kries (1882), pp. 258–260.
79
Ibid., p. 261.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 539

of pure (unbenannte) numbers. For instance a velocity s / t is said to be n times a


velocity s⬘ / t⬘ if and only if the number s / s⬘ is equal to n times the number t / t⬘.80
The relation of a given quantity to length, time, and mass is not necessarily unique,
even in the practice of physicists. In cases such as velocity, acceleration, force, and
work, this relation follows from the requirement of a simple formulation of the laws
of mechanics. In the case of electric and magnetic quantities, Kries knew the exist-
ence of two systems of units, electrostatic and electromagnetic, for which the dimen-
sions of basic electric and magnetic quantities are different.81
Kries found greater arbitrariness in the case of temperature, for which there are
as many temperatures as there are possible conventions of uniform expansibility for
thermometric fluids. Yet a theoretical requirement, that the temperature of a low-
pressure gas should be proportional to its internal kinetic energy, eventually selected
one of these conventions (Kries, 1882, p. 285). Kries summed up:

In the last analysis, only the values of lengths, time, and mass can be compared;
the reduction of all other quantities to those quantities requires an appropriate
convention to be reached by consideration of factual [thatsächlicher] relations.82

At the border of physics, or outside physics the reduction of intensive quantities


to mechanical quantities could become impossible. While examining symptoms for
this impossibility, Kries identified the possible lack of transitivity of equality, and
the lack of additivity for a given convention of equality. Take for example the con-
vention that two lights of different color are equally bright when they appear to be
so in normal vision. It is not obvious that the brightnesses of two lights of two
different colors should be equal if they are equal to the brightness of a third light
of a third color. Worse, the equal brightness of two lights of different color does
not imply the equal brightness of ‘multiples’ of these two lights obtained by incoher-
ent superposition.83
Kries next came to the main target of his criticism, Fechner’s law. Following
Kant, Kries treated sensations as intensive quantities that would never be susceptible
of measurement. The heart of his argument was that the equality between two
increments of a given sensation could not be given any meaning (Kries, 1882, p.
273). Equivalently, ‘the higher intensity . . . does not lend itself to a decomposition
[zerlegen] that would allow us to perceive smaller intensities as parts in it’ (ibid.,
p. 275). One may well adopt the convention, as Fechner implicitly does, that the
smallest noticeable increment of a sensation should be regarded as independent of
its intensity. This convention, together with Weber’s law, leads to Fechner’s logarith-
mic law. However, ‘this law does not mean anything without the convention, and
together with the convention it means no more than the observed facts [Weber’s
law]’ (ibid., p. 276).

80
Ibid., pp. 262–264.
81
Ibid., pp. 265–266.
82
Ibid., pp. 267–268.
83
Ibid., pp. 268–270.
540 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

Kries admitted that the convention could become more fruitful and more natural
if it was made part of a predictive psychological theory that would correlate the
intensities of various psychic states. This would be similar to the change of status of
the thermometric convention of uniform expansivity of dilute gases after justification
through the kinetic theory of gases. However, Kries believed that well-established
results of the physiology of sensations made such a theory impossible. For instance,
he saw a basic inconsistency of Fechner’s convention in the fact that the sensibility
to increments of brightness varies according to their position in the visual field.84
Kries ended his memoir with a summary of his conclusions:

Intensive quantities are in themselves not measurable, because the equality of


different increments in an intensity scale has no clear meaning without further
consideration. When intensive quantities are determined in the domain of the so-
called exact sciences, it turns out to be only a matter of measurements of space
and time quantities, and of counting [for mass]; every measurement of intensity
is brought back to that through specific conventions. Similarly, one can only speak
of a measurement of sensations after an arbitrary convention about what should
be regarded as equal. The controversy over whether sensations grow as the logar-
ithm of the stimuli or are proportional to the stimuli therefore has no factual
content; on the contrary, it is a quarrel over words based on misunderstanding.85

In 1886 Adolf Elsas, a Marburg physicist with Kantian inclinations, published


a pamphlet against psychophysics without apparent knowledge of Kries’s previous
criticism. Elsas was aware of the Grassmannian concept of quantity (through O.
Stolz’s Arithmetik), and he had read Du Bois’s Theorie der Functionen. He agreed
with Du Bois that a philosophical basis was required for a proper theory of quantity.
But he rejected Du Bois’s phenomenology, in which he only saw muddled, naive
empiricism. He had no more sympathy for the way the Grassmanns severed the
concept of quantity from perception. He could not admit to quantities only being
unknowns submitted to equality, inequality, and combination.86
Like Kries, but more explicitly, Elsas preached a return to Kant’s principles of
understanding, with the distinction between extensive and intensive quantity. In his
opinion, nineteenth-century mathematicians, including Grassmann and Du Bois but
excluding Gauss, had lost the Kantian ‘thread’, and consequently lacked the means
to tackle the difficult issue of measurability. With this thread in hand, Elsas rejected
Fechner’s law and the whole of psychophysics. His criticism, like Kries’s, rested on
the arbitrariness of any reduction of the inherently intensive quantities of sensations
to extensive quantities susceptible of measurement.87
To sum up, the measurability of psychological phenomena was widely discussed
by philosophers, physiologists, psychologists, physicists, and mathematicians around

84
Ibid., pp. 282–285.
85
Kries (1882), p. 294.
86
Elsas (1886), pp. 54–56.
87
Ibid., p. 58.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 541

1880. From these discussions emerged various criteria of measurability. Zeller


insisted on the existence of fixed standards; Wundt on comparability in a vaguer
sense; Kries on the unambiguous definition of equality [Gleichheit], in particular
transitivity; Du Bois and Kries on additivity, even though they disagreed on whether
sensations met this criterion. All these authors, except Wundt, ascribed a paradig-
matic role to length measurement in determining these criteria. They did so in refer-
ence to Kant’s analysis of quantity, or because they approved the Gaussian doctrine
of absolute measurement and the underlying ideal of mechanical reduction.

1.4. Physicists on measurability

Du Bois, Zeller, and Kries discussed the measurability of physical quantities, if


only as a preliminary step for answering other questions: the nature of mathematical
quantities (Du Bois), and the measurability of psychological phenomena (Zeller and
Kries). Physicists usually ignored this basic issue. Laplace, Bessel, Gauss, Weber,
and Neumann who immensely improved the practice of measurement in physics
were more concerned with precision, reproducibility, and error analysis than with
measurability itself. When new physical quantities were introduced, their measur-
ability seemed to result from their very definition or from their assumed nature, as
had been the case with the caloric, electric, and magnetic fluids. Again, the question
was not whether to measure, but how to measure.
A first exception to this general attitude was Ernst Mach, physicist but also physi-
ologist with a keen interest in the foundations of his fields. While reflecting on the
history of mechanics, Mach found previous definitions of mass to be vague and
circular. He proposed a new definition according to which two bodies are of equal
mass if, mutually acting on each other, they induce on each other equal and opposite
accelerations. With his usual acumen, Mach realized that this definition, to be consist-
ent with the mathematical properties of equality, required a certain physical property
to hold:

Will two bodies B, C, which in mutual action with A have acted as equal masses,
also act as equal masses in mutual action with each other? No logical necessity
exists whatsoever, that two masses that are equal to a third mass should also be
equal to each other. For we are concerned here, not with a mathematical, but with
a physical question.88

Mach then showed that a negative answer to this question would imply the possibility
of perpetual motion.89
It should be noted that for earlier definitions of mass, the transitivity of equality
would have had a different status. For instance, the concept of mass can be introduced
in a specific picture of matter, as made of entirely similar discrete material points,

88
Mach (1872), pp. 52–53.
89
Mach (1883), p. 268.
542 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

or as made of a strictly homogenous universal substance with vacuous pores. In the


first picture mass is defined as the number of material points, in the second as the
total volume of the universal substance. Then transitivity is a logical consequence
of the definition. This is also the case for the more empirical definition of mass
evoked by Kries, through the equality of weights. Indeed the equality of two masses
is thus brought back to the equality, in the mathematical sense, of effects produced
by these masses, such as the compressions of a spring. In contrast Mach’s new
definition of mass turns the transitivity of equality into a measurability criterion.
Mach’s considerations appeared in 1872 in his Erhaltung der Arbeit, and in a
more developed form in his widely read Mechanik of 1883. James Clerk Maxwell’s
Theory of heat of 1871 contains a similar remark in regard to temperature measure-
ment. In general, Maxwell payed much attention to the nature of the physico-math-
ematical quantities, to their definition, and to their classification. He promoted Fouri-
er’s dimensional analysis, Faraday’s distinction between quantity and intensity for
electricity and magnetism, as well as William Rowan Hamilton’s quaternions.
Maxwell also had a vivid interest in precision measurement, and participated in the
British Association project of a new system of units.90
Maxwell defined the equality of the temperatures of two bodies as follows:

If when two bodies are placed in thermal communication neither of them loses
or gains heat, the two bodies are said to have equal temperatures or the same
temperature. The two bodies are then said to be in thermal equilibrium.91

Maxwell promptly complemented this definition with a law ‘of equal temperatures’:
Bodies whose temperatures are equal to that of the same body have themselves equal
temperatures. He explained:

This law is not a truism, but expresses the fact that if a piece of iron when plunged
into a vessel of water is in thermal equilibrium with the water, and if the same
piece of iron, without altering its temperature, is transferred to a vessel of oil,
and is found to be also in thermal equilibrium with the oil, then if the oil and
water were put into the same vessel they would themselves be in thermal equilib-
rium, and the same would be true of any other three substances. This law, there-
fore, expresses much more than Euclid’s axiom that ‘Things which are equal to
the same thing are equal to one another’, and is the foundation of the whole
science of thermometry.92

Again we may note, as Mach later did, that the empirical status of transitivity

90
Maxwell (1871), pp. 32–33; Maxwell (1873), Arts. 1–6, 10–12. Cf. Harman (1998) and further refer-
ence there. On BA units, see Schaffer (1995).
91
Maxwell (1871), p. 32. This definition has the disadvantage of regarding the notion of heat transfer
as given. Mach later replaced the transfer of heat by an observable physical change of the two bodies in
contact, for instance a change of volume.
92
Ibid., p. 33.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 543

depends on the selected definition of the equality of temperatures. Had Maxwell


defined this equality, as most of his predecessors did, by the equality of the expan-
sions produced in a thermometric device, then transitivity would have been a logical
consequence of the definition. But the equality of the temperatures of two bodies
would no longer have implied their thermal equilibrium.93
Maxwell went on to describe the construction of thermometers, emphasizing that
different thermometric fluids gave different temperatures, and that thermometry yet
permitted a consistent ordering of bodies from the colder to the hotter. After complet-
ing this task, he introduced some epistemological remarks on the kind of measure-
ment that thermometers permitted:

For the present, we regard [temperature] rather as a quality, capable of greater or


less intensity, than as a quantity which may be added or subtracted from other
quantities of the same kind. A temperature, as far as we have yet gone in the
science of heat [in this chapter of the present book], is not considered as capable
of being added to another temperature so as to form a temperature which is the
sum of its components. When we are able to attach a distinct meaning to such
an operation, and determine its result, our conception of temperature will be raised
to the rank of a quantity. For the present, however, we must be content to regard
temperature as a quality of bodies, and be satisfied to know that the temperature
of all bodies can be referred to their proper places in the same scale.94

The originality and depth of Maxwell’s position is best understood in comparison


with Kries’s idea of temperature measurement. For convenience, Maxwell’s scholas-
tic distinction between quality and quantity, which is roughly equivalent to Kries’s
Kantian distinction between intensive quantity and extensive quantity, is used in
this comparison.
For Kries, each kind of thermometer defines temperature as a quantity meeting
the additivity criterion. There can only be measurement if there is additivity; but the
convention of uniform expansivity of the thermometric fluid provides the required
additivity. The arbitrariness of this convention, which depends on the nature of the
thermometric fluid, is not prohibitive; it only implies that there are many different
temperature-quantities.
For Maxwell, thermometers do not (a priori) define quantities. The temperature
they define is not a quantity, because it does not meet the additivity criterion. The
convention of uniform expansivity is too artificial to make the additivity of expan-
sions pass for an additivity of temperature increments. Yet Maxwell regards the
quality of temperature as measurable, against the Kantian creed that only quantities
are measurable. Of course, measure has a weaker meaning in the case of qualities.
It only requires the establishment of a scale for ordering bodies from the colder to the
hotter. This scale is conveniently expressed in numbers, corresponding for example to

93
Cf. Mach (1896), p. 41.
94
Maxwell (1871), pp. 44–45.
544 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

the different degrees of expansion of a thermometric fluid. As Mach later put it,
these numbers are mere ‘inventory numbers [Inventarnummer] thanks to which we
can recognize, or, when necessary, seek or reproduce the same thermal state’ (Mach,
1896, p. 56).
In a subsequent chapter (Maxwell, 1871, pp. 156–158), Maxwell gave Thomson’s
definition of absolute temperature, thanks to which temperature becomes a quantity.
This definition amounts to regarding two temperature differences T 1⫺T 2 and T 2⫺
T 3 as equal if the works yielded by a perfect Carnot engine functioning between the
sources T1 and T2 is equal to the work of another perfect Carnot engine functioning
between the sources T2 and T3 when the heat received by the source T2 from the
first engine is equal to the heat given by the source T2 to the second engine. This
definition gives to the laws of thermodynamics their simplest expression, and hap-
pens to be very nearly identical to the temperature of a dilute-gas thermometer. The
latter coincidence did not prevent Maxwell from making a clear logical distinction
between two concepts of temperature: the thermometric quality, and the thermodyn-
amic quantity.
A definition of concrete addition, or of concrete equality of increments (which
amounts to the same) was essential to Maxwell’s presentation of absolute tempera-
ture. In contrast, Mach did not mention additivity in his definition of mass. There
is an obvious reason for that: the physical operation corresponding to the addition
of mass is too straightforward to be worth mention. Additivity only caught the atten-
tion of the investigators of measurability in non-trivial cases: brightness (Kries),
sensations (Du Bois and Kries), and temperature (Du Bois, Kries, Maxwell). There
is still another case that Maxwell judged problematic: electric charge.
For a continental physicist, electric charge was defined by the repartition of electric
fluid(s). Hence it was obviously additive. Maxwell, however, did not subscribe to
the continental concept. Following Faraday, he wanted to define electric properties
empirically, without recourse to a preconceived picture. He did this by means of
Faraday’s experiments on hollow conductors. The basic device is a hollow metallic
vessel externally connected to a gold-leaf electroscope. The basic effect is the devi-
ation of the electroscope when an electrified body is introduced into the vessel, which
has an opening with a lid.95
The electrification of two bodies is said to be equal when they produce the same
deviation of the electroscope in this operation. The electrification of two bodies is
said to be opposite and equal when their simultaneous introduction into the vessel
produces no deviation of the electroscope. The total electrification of a system of
bodies is said to be indicated by the deviation of the electroscope when the whole
system is introduced into the vessel. Second-order experiments, involving hollow
vessels introduced into a larger hollow vessel, show that the electrification of a con-
ducting body can be integrally transferred to a vessel by having the body touch the
inner walls of the vessel. By repeating the latter operation with a second conducting

95
Maxwell (1873), Arts. 28–33.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 545

body, the vessel acquires an electrification which is the sum of the electrification of
the two bodies.96
Maxwell thus gave concrete operations that had the structural properties expected
from the addition of electrifications. He concluded:

The electrification of a body is therefore a physical quantity capable of measure-


ment, and two or more electrifications can be combined experimentally with a
result of the same kind as when two quantities are added algebraically. We there-
fore are entitled to use language fitted to deal with electrification as a quantity as
well as a quality, and to speak of any electrified body as ‘charged with a certain
quantity of positive or negative electricity’.97

Maxwell’s insights into measurability were deep, but only occasional. He


expressed the additivity criterion in the limited contexts of electricity and thermom-
etry, and the transitivity criterion only for thermometry. In the ‘Preliminary on the
measurement of quantities’ found at the head of his Treatise, his only concern was
the definition of units and dimensions in ‘dynamical sciences’, for which length,
time, and mass have a privileged status. He reproduced the now standard notation
that he had given in 1863 with Fleeming Jenkins, for example l[L] for a length with
the ‘numerical value’ l and the ‘concrete unit’ [L]. The choice of units, and conver-
sion from one system of unit to another were indeed more important to the practical
electrician than the explicit formulation of general criteria of measurability.98

2. Helmholtz’s foundations

2.1. Debts

We have now surveyed a few nineteenth-century reflections on quantity and


measurement in mathematical, psychological, and physical contexts. How did Herm-
ann Helmholtz encounter these reflections? How relevant were they to his own philo-
sophy of counting and measuring?
Helmholtz came to know and appreciate Hermann Grassmann while studying his
theory of colors. ‘Zählen und Messen’ (1887) contains several references to the
Grassmann brothers. The most important one concerns the derivation of the various
axioms of arithmetic from ‘Grassmann’s axiom’, that is, (a + b) + 1 = a + (b + 1):
‘In the execution of arithmetic derivations, I will completely follow their way’
(Helmholtz, 1887, p. 357). A footnote refers to Hermann Grassmann’s Ausdehnungs-
lehre (inadequately), and to Robert Grassmann’s Formenlehre (adequately). Appar-

96
I have somewhat simplified Maxwell’s ideal processes.
97
Maxwell (1873), Art. 34.
98
Maxwell (1873), Arts. 1–6. Cf. Harman (1998).
546 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

ently, Helmholtz had not had in hand Hermann Grassmann’s Arithmetik, which was
most directly relevant.
In other aspects of his arithmetic, Helmholtz considerably departed from the
Grassmannian approach. As we will see, his notion of number had psychological
and empirical components that were totally foreign to the Grassmanns’ formalism.
Helmholtz himself mentioned some elements of this antagonism. He noted that Her-
mann Grassmann’s definition of equality (Gleichheit), as complete indifference with
respect to mutual substitution in all assertions, only applied to the quantities them-
selves, not to the objects they measured, because two objects could be alike (gleich)
in one respect, and differ in another (Helmholtz, 1887, p. 377n). Helmholtz thus
distinguished mathematical equality from empirical likeness, whereas Grassmann had
nothing to do with the latter notion. Similarly, Helmholtz noted that his use of the
term Verknüpfung (connection) was meant to be objective (corresponding to a physi-
cal operation) whereas for the Grassmanns it was purely subjective (concerning acts
of thought) (ibid., p. 381n).
Helmholtz was also aware of Ernst Schröder’s Arithmetik und Algebra (1873).
Although Schröder drew from the Grassmann brothers on the formal side, he also
anticipated Helmholtz’s consideration of psychological and empirical aspects of
numbers. Helmholtz praised him for his remark that the independence of the number
of objects (Anzahl) in a group of objects with respect to the order in which they
were counted should not be regarded as an obvious fact. This independence, Schröder
argued, had a double source: psychological (in the definition of ordinal numbers)
and empirical (in the stability of the counted objects).99
The nature of Helmholtz’s use of Du Bois is less obvious. Paul Du Bois-Reymond
was an old friend from his Königsberg years, as well as the brother of Helmholtz’s
closest friend.100 In his essay, Helmholtz mentioned Du Bois’s Funktionentheorie
for ‘relevant considerations’ including an ‘empirical . . . derivation of the concept
of quantity from the concept of line’ (Helmholtz, 1887, p. 358). He also noted that
Du Bois ended up with ‘a paradox, following which two opposite standpoints [ideal-
ist and empiricist], which both become muddled in contradictions, are possible’
(ibid., p. 359).
That Du Bois in reality claimed the two standpoints to be perfectly consistent did
not bother Helmholtz. On the contrary, Helmholtz used Du Bois’s ‘paradox’ to justify
his own intervention:

As [Hr. P. du Bois-Reymond] is an extremely sharp mathematician who has


probed the deepest foundations of his science with special intentness, his con-
clusion encouraged me to expound my own thoughts on this topic.101

The remark is the more puzzling because Helmholtz hardly touched the question of

99
Helmholtz (1887), p. 358; Schröder (1873), pp. 14–16.
100
See Olesko (1994).
101
Helmholtz (1887), p. 359.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 547

limits and the construction of irrational numbers, which were the subject of Du Bois’s
conclusion. Helmholtz’s subsequent reference to Du Bois’s rejection of the chess-
game conception of mathematics is equally confusing (ibid., p. 359). Whereas Du
Bois’s target was the formalist approach to mathematics, Helmholtz’s used the chess-
game analogy to characterize a purely psychological conception of arithmetic in
which the question of the applicability to real objects would never be raised.
Helmholtz’s failure to explain his relation to Du Bois’s work is easily under-
standable. Despite some formal similarity, their views on the relation between coun-
ting and measuring differed on a basic point: whereas Du Bois conflated the definition
of numbers and their applicability to real objects, Helmholtz clearly separated the
two issues. Nevertheless, we cannot be sure that Du Bois’s theory of quantity was
nothing to Helmholtz but an incentive to do better. Helmholtz’s criteria of measur-
ability may in part have been reminiscent of Du Bois’s definition of linear quantity.
Even more opaque is the relevance of psychological measurement to Helmholtz’s
considerations. Helmholtz’s memoir does not contain any reference to this question.
It is strictly restricted to measurement in physics. Yet several circumstances suggest
that Helmholtz had the case of psychology in the back of his mind.
Helmholtz had a deep interest in the rising empirical psychology, both in Fechner’s
and in Wundt’s form. He certainly approved Fechner’s law, as documented by his
attempts to generalize it, both in the Handbuch der physiologischen Optik and in
later publications. Wundt had been his assistant in Heidelberg, and Helmholtz sup-
ported him early in his career, even though he disagreed with the subsequent turn
of Wundt’s researches. Then Helmholtz could not be indifferent to the raging debates
on the measurability of psychological phenomena.102
Helmholtz was obviously aware of Du Bois’s contribution to this debate. He also
referred to Elsas’s book, though only to condemn its strict Kantian outlook
(Helmholtz, 1887, p. 358). Kries, to whom he did not refer, had been his student in
Berlin. Most significantly, the dedication of ‘Zählen und Messen’ to Zeller suggests
a connection with the Zeller–Wundt controversy.
Granted that the measurability of psychological phenomena almost certainly
belonged to Helmholtz’s motivations, why did he refrain from any explicit involve-
ment? A plausible answer to this question is that Helmholtz generally avoided to
enter controversies in which he thought that no decisive, consensual argument could
be reached.103 He used and extended Fechner’s law without ever discussing its
status.104 He never answered Wundt’s attacks against his physiological approach to

102
Cf. Turner (1982); Wundt (1920). As Cahan (2002) explains, the new empirical psychology relied
on some of Helmholtz’s physiological results and principles (propagation of nervous influx, sensations
as signs). Helmholtz also supported Hermann Ebbinghaus and his work on measuring memory.
103
Cf. Jurkowitz (2002).
104
In his Handbuch der Physiologischen Optik, Helmholtz justifies Fechner’s law with the sentence:
‘We regard differences of the degree of intensity [of a sensation] as equal when they are perceived with
the same distinctness’ (Helmholtz, 1856–1867, par. 21). This formulation may suggest that he recognized
the conventional aspect of the law. His extensions of the law are found ibid., and in later memoirs
published around 1890.
548 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

perception. Presumably, he perceived a recurrent lack of rigor in the promoters of


quantitative psychology. But at the same time he strongly supported the extension
of the empirical approach to new fields of knowledge. Had he used his tremendous
authority against Fechner or Wundt, he could thus have hampered the development
of new, promising fields.
Another question regarding Helmholtz’s sources is whether or not he relied on
Maxwell’s and Mach’s discussions of temperature, charge, and mass measurement.
No reference to these authors is found in ‘Zählen und Messen’. Whether Helmholtz
had read Mach’s Erhaltung der Arbeit or his Mechanik is not known. He was, how-
ever, quite familiar with Maxwell’s Treatise on electricity and magnetism and with
his Theory of heat. The additivity criterion could be found in another of Helmholtz’s
sources, Du Bois’s Theorie der Functionen. Kries had the transitivity criterion, but
we do not know whether Helmholtz read his essay. Most likely, Helmholtz’s special
emphasis on this criterion was inspired by a reminiscence of Maxwell’s discussions
of temperature measurement.

2.2. The foundations of geometry

If we believe the introduction of Zählen und Messen (Helmholtz, 1887, p. 356),


the main motivation for Helmholtz’s investigation was his previous study of the
foundations of geometry. Before recalling the contents of this study, we examine
Helmholtz’s early, Kantian views on quantity, space, and time.
In the early 1840s, Helmholtz wrote an essay on the foundations of the natural
sciences, a long fragment of which can be found in Leo Koenigsberger’s biography
(Koenigsberger, 1902–1903, Vol. 2, pp. 126–138). In this essay, Helmholtz intro-
duced quantity and quality as concepts of understanding. Quantität, by definition,
refers to the relation of an object to another of the same kind; Qualität instead refers
to the comparison of objects of different kind. Two objects are said to be of the
same kind (gleichartig) when both can be decomposed into parts that are all alike
(gleich) in some respect. Hence follows the definition of Grösse (quantity), measure,
and number:

An object, considered with respect to Quantität, is called Grösse; hence we can


regard as Grösse every object that can be thought as decomposed into equal
[gleich] parts. To measure means to determine the amount [Menge] of such parts;
a determined amount is called number; a single part is called unit of measure-
ment [Maasseinheit].105

Although this definition of quantity seems reminiscent of Kant’s ‘extensive quan-


tity’, there are notable differences. Helmholtz does not relate his definition to the
intuition of space and time. He gives a definition of equality (Gleichheit) that can
be applied to any physical property. The definition of quantity implies divisibility

105
Koenigsberger (1902–1903), Vol. 2, p. 128.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 549

into equal parts, whereas for Kant mere divisibility is enough. The definition of
numbers derives from the notion of quantity, against Kant’s anchoring of arithmetic
on the intuition of time (and against Helmholtz’s later view on this subject).
No matter how much the young Helmholtz thus departs from Kantian orthodoxy,
his concepts of counting and measuring still are transcendental, they are not inferred
from our past experience of the world. The same can be said of the concepts of
space and time found in this essay: they are basically Kantian, with some liberties.
Helmholtz introduces them as ‘general and necessary forms of our intuition of nature’
(ibid., p. 127). Yet he departs from Kant by favoring analytical over synthetic
geometry. A point is determined by a set of coordinates. Metric properties are funda-
mental (although Helmholtz does not formalize them). Rigid bodies and their congru-
ence are defined in these terms (ibid., pp. 132–135).
In the 1860s, Helmholtz returned to the problem of space through the physiology
of sensations. In his physiological work, he associated different manifolds to different
sensations. For example, ordinary space, visual space, and color space corresponded
to the sensations of touch, vision, and color. He also reflected on how we infer the
existence of external objects and how we develop spatial notions about them. His
general approach was kinesthetic, that is, based on our awareness of relations
between motor impulses and ensuing sensations. He concluded that spatial notions
derived from our ability to obtain and observe the congruence of physical objects.106
Accordingly, Helmholtz (1866) based geometry on the notion of congruence,
regarded as a necessary condition for any spatial measurement. This notion implies
the existence of freely mobile rigid objects. In order to express this notion mathemat-
ically, Helmholtz started with the general notion of continuous manifold, the points
of which are determined by a set of coordinates. This starting point was natural to
one who had encountered various sensorial manifolds. It also had the advantage of
avoiding the pitfalls of intuition inherent in the synthetic approach, Helmholtz noted.
From continuity and various assumptions permitting freely mobile rigid bodies,
Helmholtz derived the positive, quadratic form of the squared element of length with
respect to the differentials of the coordinates, as well as the constancy of the curva-
ture of the corresponding Riemannian space. When, in 1866, he first announced these
results, he believed that Euclidian space was the only space of that kind that was
infinite and three-dimensional. Hence he could declare to have found the ‘factual
foundation’ (thatsächliche Grundlage) of geometry in the free mobility of rigid bod-
ies.
Helmholtz often used the word Thatsache (fact) as a leitmotiv for the empiricist
approach to perception and natural sciences in general. The word includes That
(action), a marker of the kinesthetic origin of perception and knowledge, as in
Helmholtz’s favorite verse from Goethe’s Faust: ‘Am Anfang war die That’. In its
first form, Helmholtz’s understanding of the foundations of geometry was empirical

106
See for example Helmholtz (1856–1867), Vol. 3; Helmholtz (1878a); DiSalle (1993), and further
references there. Hyder (2001) (which I saw too late to take into account) pursues the connection between
Helmholtz’s work on colours and his reflections on the foundations of geometry, and emphasizes
Helmholtz’s concern with the measurement of space.
550 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

in two senses: it was based on a general precondition for the empirical determination
of spatial relations, and it made the structure of our sensorio-motor apparatus respon-
sible for the choice of Euclidian geometry. It still had some similarity with Kant’s
conception of space as a synthetic a priori form of intuition, although Helmholtz
“naturalized” Kant’s intuition by giving it a physiological substratum.
As Helmholtz soon came to know, he had overlooked the possibility of Lobatch-
evskian geometry, which permits free mobility and yet describes an infinite space.
From then on, Helmholtz insisted that the axioms of geometry did not pertain to an
a priori given form of our intuition (Helmholtz, 1868, 1876). Only the general notion
of space, with continuity, congruence, and dimensionality did. In order to decide
between various systems of axioms, one had to experiment. Helmholtz thus had a
third reason to regard geometry as empirical. He even regarded it as ‘the first and
most accomplished of the natural sciences’ (Helmholtz, 1878b, p. 642). Yet he did
not regard geometry as unambiguously determined by congruence experiments. He
clearly recognized that the concrete realization of rigid bodies depended on the
accepted laws of mechanics. A different choice of the laws of mechanics would lead
to a different geometry. What could be tested only was the combination of geometri-
cal and mechanical axioms (Helmholtz, 1876, pp. 30–31).
Helmholtz’s essays on the foundations of geometry induced responses from con-
temporary Kantian philosophers. Helmholtz was thus led to insist on the relation of
his views to Kant’s. Since for Kant, geometry and arithmetic had parallel foundations
in external and internal intuition, Helmholtz naturally faced the question of the foun-
dations of arithmetic.

2.3. ‘Zählen und Messen’

Helmholtz introduced his essay on counting and measuring through an analogy


with the foundations of geometry. Kant, he recalled, regarded the axioms of arithme-
tic and geometry as ‘a priori given truths that further determine the transcendental
intuition of time and space’ (Helmholtz, 1887, p. 356). In the case of geometry,
Helmholtz had already shown, against Kant, that the axioms were of empirical nature,
even though he agreed with Kant than the general notion of space remained a tran-
scendental form of intuition. Helmholtz continued:

Now it is clear that my empiricist theory, if it no longer admits that the axioms
of geometry cannot and must not be proved, must also apply to the origin of
the arithmetic axioms, which have a comparable relation to the temporal form
of intuition.107

Among the arithmetic axioms, Helmholtz included the transitivity of equality, the
associativity and commutativity of addition, and its compatibility with equality. His
empiricist revision of arithmetic had two steps. In the second and most straightfor-

107
Helmholtz (1887), p. 357.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 551

ward step (ibid., pp. 364–371), he reproduced the Grassmannian proof that all axioms
of arithmetic resulted from ‘Grassmann’s axiom’ (a + b) + 1 = a + (b + 1). In the
first step (ibid., pp. 360–364), he derived that axiom from a concept of number based
on psychological facts and on an adequate definition of addition. This was done
as follows.
The basic psychological fact is our capacity to order a succession of mental states:
‘Counting is a procedure which rests on our ability to memorize the order in which
acts of consciousness have occurred’ (ibid., p. 360). Numbers (Zahlen) are an unlimi-
ted sequence of arbitrarily chosen signs in a given, conventionally fixed order. Their
first purpose is to fix in our memory the temporal order of other series of acts of
consciousness. The numbers thus defined are ordinal numbers (Ordnungszahlen).
The decimal notation is a convenient way to generate an unlimited system of signs
with a definite order and no repetition, and thus to define a communicable num-
ber system.
Call Sa (successor of a) the number immediately following the number a.
Helmholtz defines the sum a + b of two numbers as the number Sba obtained by
repeating b times the operation S. For example a + 3 = SSSa. This definition implies
S(a + b) = SSba = Sb+1a = a + Sb. Since Sb = b + 1, this is the same as Grassmann’s
axiom (a + b) + 1 = a + (b + 1). Other arithmetic axioms follow by mathematical
induction. Multiplication is defined inductively by 1·a = a and (Sb)·a = b·a + a. Its
properties of associativity, commutativity, and distributivity also follow by induc-
tion.108
Having thus reconstructed arithmetic, Helmholtz considered its application to con-
crete objects. The most immediate application is the determination of the number
(Anzahl) of objects in a ‘group’ of objects (ibid., pp. 371–373). Helmholtz proved
that this number was independent of the order in which the number was counted.
He also emphasized, as Schröder had done, that objects of a given class were not
necessarily countable: ‘No object should be permitted to disappear, or to fuse with
others, or to split, or to be brought into existence’ (ibid., p. 372). Whether this
condition was met could only be decided empirically. Granted that objects of a given
class could be counted, the number of objects of the reunion of two groups obviously
equalled the sum of the numbers of object in each group. Helmholtz concluded to a
perfect agreement between ordinal and cardinal numbers (Zahl and Anzahl). Ordinal
numbers had, however, ‘the advantage of being accessible without recourse to exter-
nal experience’ (ibid., p. 372).
Helmholtz next considered the counting of objects that are alike (gleich) in a given
respect (ibid., p. 374). Such objects are called units. The result of this counting
is a concrete number (benannte Zahl), defined by a pure number (Zahl) and the
corresponding unit.
Next comes his definition of a quantity (Grösse) and its measure:

108
Helmholtz (1887), pp. 363–371. The ‘Sa’ notation is mine.
552 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

We call quantities objects or attributes of objects which, when compared with


other similar objects allow for the distinction of larger, equal, or smaller. When
we can express a quantity by a concrete number, we call this number the value
of the quantity, and we call the process through which we find this number the
measurement of the quantity.109

Helmholtz’s basic question then was whether a given attribute of objects could
be regarded as a quantity, and whether it could be measured. A specific attribute is
defined by ‘a method of comparison’ which enable us to decide whether two objects
are alike with respect to this attribute. In order to be compatible with arithmetic
axioms of equality, this method must meet the condition: When two objects are found
to be alike to a third, they must be alike. This is transitivity, stated in a manner that
includes symmetry (a = b if b = a).110
Helmholtz then reviewed the methods of comparison for weight, distance, length,
time, brightness, pitch, and electric intensity; and argued in each case that only
experiment could confirm the validity of the condition. He emphasized that transitiv-
ity was ‘not an objective law [ein Gesetz von objectiver Bedeutung], but a way to
decide which physical relations could be recognized as equality’. In other words, for
Helmholtz transitivity does not refer to pre-existing quantities. On the contrary, it is
a test for deciding what can be held as a quantity.111
At that point we might have expected Helmholtz to turn to the ordering relation
of quantity. He did not, because, for reasons to be discussed in a moment, he believed
that one could not decide whether a quantity was larger than another without first
defining a physical connection (Verknüpfung). His next step was to assume and dis-
cuss this procedure.
In order to be compatible with the arithmetic axioms of addition, the connection
must be compatible with the physical equality (a + b = a⬘ + b⬘ if a = a⬘ and b =
b⬘), it must be commutative, and associative. Again, Helmholtz regarded these con-
ditions as an experimental test for identifying possible quantities. That the condition
held for all connections encountered in physics only meant that physicists had already
done the selection (ibid., pp. 381–383).
Helmholtz then defined order among quantities: a first quantity is said to be larger
than a second when it can be obtained by composition of the latter quantity with a
third quantity (ibid., p. 383). We may incidentally note a flaw in Helmholtz’s reason-
ing. There are procedures of composition for which the above definition does not

109
Helmholtz (1887), p. 375.
110
Helmholtz (1887), pp. 375–376. Helmholtz sometimes speaks of Gleichheit of two quantities when
he means of two objects. He in fact uses the same word, quantity, to denote both a property defined by
a method of comparison, and an object considered in regard to this property.
111
Ibid., pp. 377–380. Similarly, in Helmholtz’s theory of perception the experienced correlations
between sensations and voluntary impulses determine what can be held as external object. As Michell
remarks (1999), pp. 70–71, Helmholtz’s methods of comparison, being dependent on a specific class of
necessarily imperfect instruments, can only bear on a limited range of quantities of a given kind, and
with a limited precision. Therefore, they are not sufficient to define the quantities of physico-mathemat-
ical theories.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 553

generate order. For example, the usual rule of composition of forces would lead to
the conclusion that any force is larger than any given force. Helmholtz should have
further required the composition to be such that it can be used to generate order. Or
he should have given up the idea of making order depend on physical addition.
This idea of Helmholtz’s rested on his awareness that for a given method of com-
parison, there could be several manners of composition that led to different orders.
For example, two electric conductors can be compared with the same method both
for resistance and for conductance. Yet there are two different operations of compo-
sition, in series and in parallel, that lead to different quantities, resistance and conduc-
tance, and opposite ordering of the conductors (ibid., pp. 383–384). Logically, this
fact only proves that order and method of composition are correlated, when the latter
exists. It does not follow that composition must precede order. Helmholtz thought
differently, presumably because he unconsciously projected the natural definition of
order for cardinal numbers or for the points of a line.
Having defined equality, addition, and order of physical quantity, Helmholtz pro-
ceeded to (uniform) division: ‘Quantities which can be added are also to be divided
in general’ (ibid., p. 384). Whether Helmholtz understood divisibility as a further
empirical condition on quantities or as a consequence of additivity is not clear. It
all depends on what he meant by ‘in general’ in the above extract. Most likely, he
only meant that all known physical quantities were divisible. This would explain his
introductory statement: ‘I believe it would be an unnecessary limitation of the the-
orems [about quantities] if physical quantities were required from the start to be
composed of units’ (ibid., p. 358). Helmholtz probably had in mind the traditional
introduction of measurement, in physics textbooks, as the division of a quantity into
equal units. In his view, this definition assumed too much in a single stroke and left
the true criteria of measurability in the dark.
For him divisibility only was the last of a series of criteria that led to the represen-
tation of quantities by concrete numbers. By divisibility, he not only meant the possi-
bility to regard a given quantity as the sum of a number of equal quantities, but also
the possibility to approximately express a given quantity as a multiple of a fixed
unit, and to indefinitely improve the approximation by introducing a series of sub-
units (ibid., p. 385). He thus implicitly assumed the Archimedean property, which,
together with the existence of difference, allows the arbitrarily precise approximation
of ratios by rational numbers.
Helmholtz went on to admit the existence of irrational ratios, which cannot be
exactly given by rational numbers. The existence of arbitrarily close approximations
by rationals was however sufficient for the almost everywhere differentiable func-
tions used by physicists. Helmholtz was aware of mathematical monsters that could
not be determined by rational approximations. But ‘in geometry and physics’, he
noted, ‘we have not yet encountered such kinds of discontinuity’ (ibid., p. 385).
Physics according to Helmholtz, just as geometry according to Euclid, needed
irrational ratios, not irrational numbers.
Helmholtz next discussed physical quantities for which no additive composition
was known. These occur as ‘coefficients’ in empirical laws or definitions relating
different additive quantities. For instance, the optical index, specific weight, and
554 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

conductivity of a substance are coefficients. Such quantities can be measured only


indirectly, by measuring the additive quantities which they relate. Helmholtz noted
that the distinction between coefficients and additive quantities was inessential,
because a coefficient could become an additive quantity upon the discovery of a
relevant additive connection. He also pointed to the similarity with ‘the ancient meta-
physicians’ opposition between extensive and intensive quantities’ (ibid., p. 386).
Note, however, that ‘extensive’ there referred to extension, whereas the additive
connections assumed by Helmholtz do not need to be spatial.
The value of a coefficient generally depends on the choice of the units in which
the related quantities are measured (unless the coefficient is dimensionless). For this
reason, Helmholtz introduced the multiplication and division of concrete numbers
(ibid., pp. 388–389). In modern words, we would say that he (and Maxwell) regarded
physical quantities as one-dimensional vectors the tensor product, power, or quotient
of which may have a physical meaning. A unit is a base in a tensor space. In the
Gaussian view according to which all quantities can be generated from length, time,
and mass, all units have the form [L]a[T]b[M]g, wherein [L], [T], [M], are the units
of space, time, and mass (Maxwell’s notation) and a, b, g three positive or nega-
tive integers.
Even more succinctly, Helmholtz discussed the addition of concrete numbers of
different kinds, as occurs in Hermann Grassmann’s theory of extension and in Wil-
liam Rowan Hamilton’s theory of quaternions. As he recalled, his main purpose only
was ‘to show the meaning and justification of calculation with pure numbers and
the possibility of their application to physical quantities’ (ibid., p. 390). He concluded
on the intimate connection of quantification with concept formation
(Begriffsbildung):

When we form the concept of a class, we resume in it everything that is alike in


the objects which belong to this class. When we conceive a physical relation
[physisches Verhältnis] as a concrete number, we have also removed from the
concept of the units of the class every difference that belongs to them in reality.
Units are objects which we consider only as elements of their class, and the
expediency of which only depends on there being such exemplars. In the quantities
which are built from them, there remains only the most accidental [zufälligste]
of differences, that of number [Anzahl].112

In the introduction to the lectures on theoretical physics he gave in the early 1890s,
Helmholtz reproduced his considerations on counting and measuring in a modified
form (Helmholtz, 1903, pp. 26–46). The first difference concerns the order of expo-
sition. Helmholtz now started with concrete equality, transitivity, and additive combi-
nation. Then he examined the definition of numbers and derived the associativity
and commutativity of addition. Another difference concerns the content of his defi-
nition of number. He now started with the number (Anzahl) of objects in a group

112
Helmholtz (1887), p. 391.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 555

of physically stable objects, to be determined by counting with arbitrarily chosen


numeral signs (Zahl). That the possibility of counting rested ‘on the faculty of our
memory to retain such a sequence of arbitrarily chosen signs and to reproduce it at
any time in the same order’ (ibid., p. 30) was only mentioned in passing, with no
Kantian reference to internal intuition. These differences clearly resulted from
Helmholtz’s desire to adapt his discourse to physics students. They do not imply
any significant change of his views.
Helmholtz returned a third time to the foundations of measurement in his lectures
on the theory of heat (Helmholtz, 1908, pp. 1–6). There he defined the equality of
temperatures in Maxwell’s manner, and emphasized that the transitivity of this equal-
ity was not a priori given, that it was a result of experience. In conformity with his
general concept of measurement, Helmholtz made temperature measurement depend
on an associative addition of temperature increments. Like Kries and unlike Maxwell,
he believed the dilation of a thermometric fluid to provide the required additivity.
Helmholtz was well aware that the pioneers of temperature measurement never
worried about transitivity or associativity. In his conclusion, he offered a parallel
between this fact and the way in which the names and definitions of common langu-
age conceal many observations of everyday life. He then drew the Machian moral:
‘Who fails to note in which manner such concepts entered language, comes to believe
that the concept, as it exists in human mind, rules the world’ (Helmholtz, 1908, p. 6).

2.4. Comparisons

Were Helmholtz’s views on counting and measuring as similar to his views on


the foundations on geometry as he claimed in his introduction? Helmholtz maintained
the Kantian connections between geometry and external intuition on the one hand,
and between arithmetic and internal intuition on the other (Helmholtz, 1887, p. 373).
The basic fact of geometry according to him was the free mobility of rigid bodies
required for congruence; that of arithmetic was the possibility of ordering conscious
acts in time.113
These two facts, from a strict Kantian viewpoint, would be regarded as transcen-
dental. Helmholtz, however, tended to blur the border between transcendental and
empirical. His general endeavor to give a physiological basis to Kantian intuition
has already been mentioned. In the case of arithmetic, he also offered a psychological
basis, psychology being regarded as an empirical science. Thus he made counting
‘a method built on purely psychological facts’ (ibid., p. 359) as well as an ‘indispens-
able form of our internal intuition’ (ibid., p. 361).
Both in geometry and in arithmetic, Helmholtz derived a whole system from the
basic fact (free mobility of rigid bodies, ordering in time) and some definitions. The
parallel ends here. In geometry, several constant-curvature geometries are compatible
with the basic fact, so that experience (together with mechanical axioms) is required
to decide between these multiple options. In arithmetic, the basic fact is sufficient

113
Cf. DiSalle (1993), pp. 517–519.
556 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

to induce a single system of arithmetic (as was nearly the case in Helmholtz’s
geometry before he became aware of Lobachevski’s geometry). External experience
is no longer needed to decide between different sets of axioms. Rather, external
experience is needed to determine which physical properties can be measured by
numbers. In one case, the application decides the axioms; in the other, the axioms
control the applications.
Another instructive comparison is that between Helmholtz’s and Du Bois’s views
on counting and measuring. In Du Bois’s phenomenology, numbers were obtained
by a process of abstraction in which the identity of the contemplated objects was
ignored. Numerals were a system of signs used to ‘measure’ numbers. Helmholtz
shared with Du Bois this distinction between cardinal, and ordinal number, as well as
the Müllerian idea of concept formation applied to cardinal numbers. For Helmholtz,
however, ordinals came first and were independent of external experience, they
belonged to an empiricized form of internal intuition.
Du Bois and Helmholtz also shared the idea of defining quantity by a list of
characteristic properties including additivity and divisibility, and they both dis-
tinguished between comparison without measurement and comparison with measure-
ment. However, their choice of relevant criteria widely differed, as a consequence
of their diverging views on the relation between number and quantity. For Du Bois,
the concept of (linear) quantity was inferred from a survey of measurement in various
domains of knowledge and analogy with the concept of geometric line; quantity was
a mathematical concept of phenomenological origin. For Helmholtz, the concept of
quantity derived from the concept of number; it was a norm to be imposed on quanti-
tative science, a means to decide which physical properties were measurable; quantity
was a physical concept of mathematical origin. Consequently, Du Bois’s axioms of
quantity were inspired from geometric axioms, Helmholtz’s from arithmetic axi-
oms.114
Despite this deep difference, Helmholtz and Du Bois agreed that fractional and
irrational numbers derived from the necessities of measurement. Fractional numbers
resulted from the introduction of sub-units of measurement, irrationals from the need
to measure certain geometric quantities. Beyond this agreement we find another
divergence: whereas Du Bois focused on the construction of irrationals and the con-
cept of limit, Helmholtz declared these issues of no interest for the contemporary
physicist.
Was Helmholtz closer to Kries’s and Elsas’s Kantian view of measurement? Kries
and Helmholtz both recognized that conventions were required in the definition of
quantities, and they both formulated measurability criteria including transitivity and
additivity. Yet they held different conceptions of measurement. For Kries (and Elsas),
to measure a quantity was to express it in terms of the basic quantities of length
and time by means of appropriate conventions; the measurability criteria conditioned
the reducibility of a given property to space and time. For Helmholtz, to measure a

114
Helmholtz (1887, p. 386) misleadingly identifies his distinction between quantities and coefficients
with Du Bois’s distinction between linear and non-linear quantities.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 557

quantity was to express it in (concrete) numbers; the measurability criteria determ-


ined the compatibility of this procedure with arithmetic operations.
So far, we have encountered three different conceptions of counting and measur-
ing. In Du Bois’s conception, abstraction from external experience yields the con-
cepts of discrete and linear quantity. In turn these concepts lead to number concepts
as measure of quantity: integral numbers in the discrete case, rational and irrational
numbers in the linear case. In Kries’s conception, extensive and intensive quantity
are concepts of understanding. Number is a representation (pure scheme of quantity)
belonging to the internal intuition of time. Extensive quantity, being related to our
intuition of space and time through Kantian schematism, is automatically measurable
in numbers. An intensive quantity only becomes measurable after a conventional
connection with extensive quantities. In Helmholtz’s conception, number pertains to
psychological experience. Quantity results from our drive to extract numbers from
external experience.
In short, for Du Bois quantity and number both derive from external experience;
for Kries, number and quantity are both transcendental; for Helmholtz, number
derives from internal experience, quantity is number applied to external experience.

3. Reception

3.1. Mathematicians

Helmholtz’s memoir on counting and measuring was published at a time when


several leading mathematicians sought more solid foundations for arithmetic. The
need was most strongly felt by those who regarded arithmetic as the basis, or at
least a model for the whole of mathematics.

3.1.1. Kronecker
Among these mathematicians was the Berlin mathematics professor Leopold
Kronecker, who coined the word ‘arithmetisieren’ and defended the extreme view
that every mathematical theorem should be reduced to a finite system of relations
between integers. In an essay for the Zeller jubilee (1887), Kronecker sketched his
conception of the foundations of arithmetic. He started with ordinal numbers, defined
as a ‘stock [Vorrath] of signs [Bezeichnungen] ordered in a fixed sequence, and
which we can attach to a group of different and distinguishable objects’ (Kronecker,
1887, pp. 265–266). Cardinal numbers were the result of the counting process
implicit in this definition. Kronecker then introduced arithmetic operations and their
properties in a manner similar to Helmholtz, though without the rigor of Grassmann’s
inductive procedure. Helmholtz noted the similarity on the proofs of ‘Zählen und
Messen’ (Helmholtz, 1887, p. 372n).

3.1.2. Dedekind
The publication of Helmholtz’s and Kronecker’s definitions of numbers prompted
Dedekind to publish his own (1888), which he had matured for many years. Dedekind
558 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

noted that his views were ‘in many respect similar [to those of Schröder, Helmholtz,
and Dedekind] but in foundation essentially different’ (Dedekind, 1888, p. 31n). He
shared with his competitors the primacy of ordinals over cardinals, as well as similar
derivations of arithmetic axioms by mathematical induction. But he could not agree
with their psychological, semi-intuitive definition of ordinals. His own definition was
based on the sole assumption of infinite sets. Through amazingly abstract and subtle
reasoning based on the concept of chain, Dedekind managed to prove that any infinite
set contained an ordered subset that had all desired properties of ordinal numbers.
For him numbers were neither a form of internal intuition nor a system of signs,
they were ‘free creations of the human mind, serving as a means of apprehending
more easily and more sharply the difference of things’ (ibid., p. 31).115

3.1.3. Cantor
Whereas Dedekind avoided derogatory comments on Helmholtz’s approach, his
friend Cantor had no such inhibition:

While from my standpoint the words for ordinal numbers appear as the last and
most unessential things in the scientific theory of numbers, they have been taken
in two recently published works [by Helmholtz and by Kronecker] as the starting
point for the development of the concept of number . . . Helmholtz and Kronecker
plead the extreme empirico-phychological standpoint with a vigor which one
would scarcely deem credible, if it did not here appear twice embodied in flesh
and blood . . . With these investigators, number shall be signs first and foremost,
but not indeed signs for concepts that relate to sets, but symbols for things counted
by the subjective number process.116

In the same tone, Cantor compared Helmholtz and Kronecker with the enlightenment
mathematician Louis Bertrand, who believed numbers to be words invented by
shepherds to count their ship returning from pasture (Cantor, 1887, p. 383).
The violence of Cantor’s attack was not only a matter of temperament. He regarded
Helmholtz and Kronecker as promoters of a general ‘academic-positivistic skeptic-
ism’ that threatened his own conception of mathematics, especially the admission of
actual infinities (ibid., p. 383). His understanding of arithmetic depended on this
conception. He defined the cardinal number or ‘power’ [Mächtigkeit] of a set as the
power of a given set being ‘the general concept under which fall all sets that are
equivalent to this set’ (two sets being equivalent when there exists a bijection
between them) (ibid., p. 380). Similarly, he defined the ordinal number of a well-
ordered set as ‘the general concept under which fall all well-ordered sets that are
similar to this set’ (two sets being similar when there exists an order-preserving
bijection between them) (ibid., p. 380). Cantor was mostly interested in the case of
infinite sets, for which the cardinal and ordinal definitions not longer coincide. He

115
See Schubert (1898); Belna (1996).
116
Cantor (1887), p. 382.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 559

thus introduced the transfinite numbers which he believed to be threatened by aca-


demic skepticism.

3.1.4. Frege
Three years before Helmholtz’s ‘Zählen und Messen’ was published, Gottlob
Frege had published his Grundlagen der Arithmetik (1884). This originally little-
noticed book contained the now famous definition of (cardinal) numbers as equival-
ence classes for the equal-power relation among finite sets. Having thus conceived
a purely logicist foundation of arithmetic, Frege naturally condemned any trace of
empiricism he found among his competitors. In 1891 he publicly rejected
Helmholtz’s and Kronecker’s approach, which he mistook for a naive confusion
between numerals and numbers:117

The now very well-spread inclination not to admit any object [of study] that cannot
be perceived with our senses, misleads one to regard the numerals [Zahlzeichen]
themselves as the numbers, as the true objects under consideration. But such a
conception cannot be held, because one cannot at all speak of any arithmetic
property of the numbers without returning to the meaning of the numerals.118

3.1.5. Russell
The British mathematician-philosopher Bertrand Russell shared Cantor’s and
Dedekind’s idea of a purely logical foundation of mathematics in which Helmholtz’s
intuitive reasoning had no place. In his Principles of mathematics (1903), Russell
offered detailed considerations on number and quantity. He only referred to
Helmholtz to refute his, Kronecker’s, and Dedekind’s opinion that ordinals preceded
cardinals (Russell, 1903, p. 241).
Russell’s analysis of quantity shows a clear incompatibility with Helmholtz’s pos-
ition. Russell saluted ‘the purists’ Weierstrass, Dedekind, and Cantor for having
defined irrational numbers without reference to quantity and thus inflicted the strong-
est blow to the ‘Kantian theory of a priori intuitions as the basis of mathematics’
(ibid., pp. 157–158). In contrast, Helmholtz only discussed irrationality in the context
of quantity. Russell further insisted that number and quantity should be defined separ-
ately if measurement was to mean ‘the correlation, with numbers, of entities which
are not numbers or aggregates’ (ibid., p. 158). Helmholtz may have agreed that whole
numbers had to be defined before quantity. But he based his definition of quantity
on a transposition of arithmetic axioms, whereas Russell wanted a completely inde-
pendent definition.
It may be wondered why Russell, who professed a foundation of mathematics on
pure logic, felt at all the need to discuss quantity and measurement. He even asserted
that ‘quantity is not definable in terms of logical constants, and is not properly a
notion belonging to pure mathematics at all’ (ibid., p. 158). The answer to this puzzle

117
Edmund Husserl formulated a similar criticism in Husserl (1891), p. 171.
118
Frege (1891), pp. 17–18.
560 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

is that he could provide a logicist construction of ‘series of a certain kind’ that shared
some basic properties of quantity and could be measured by numbers. He discussed
quantity as a heuristic introduction to the series properties he later constructed in a
purely mathematical manner.
The most characteristic feature of Russell’s notion of quantity was the priority of
order over divisibility. Traditionally, he judged, capacity for the relations greater
and less was believed to imply divisibility. Helmholtz certainly did so since he
derived order from divisibility (or additivity). Russell instead defined quantity in
terms of equality and order only. He had two reasons to do so: his logico-mathemat-
ical interest in the theory of ordinal structures called series, and his concern with
psychological measurement. On the latter topic, he had studied the Austrian philos-
opher Alexius Meinong.
In 1896 Meinong published a long, difficult essay on the meaning of Weber’s law
in reaction to Kries’s paper of 1882, of which he had recently become aware. Mei-
nong accepted the Kantian distinction between intensive and extensive quantity, as
well as the Kantian definition of the former in terms of distance from a zero. But
he denied the conventional aspects of equality that Kries saw in every measurement
other than that of space and time. He put much emphasis on the indirect measurement
of intensive quantities, especially sensations. In his view, sensation was measurable
because its functional relation with stimulus could be determined from Weber’s law
together with the requirement that for any intensive quantity equal dissimilarities
[Verschiedenheiten] should correspond to equal ratios of its measure. The resulting
relation between sensation and stimulus was linear instead of logarithmic.119
Although Russell did not follow Meinong in these psychophysical subtleties, he
approved the emphasis on intensive quantity, defined by equality and order only.
His favorite example of a quantity was pleasure, for which no one would be tempted
to imagine divisibility. Yet his definition of quantity was dryly axiomatic, based on
a set of axioms for the equality and order relations (Russell, 1903, p. 163).
Russell carefully distinguished between quantity and magnitude (ibid., p. 159). A
quantity is a particular object, such as a ruler, a magnitude is the class of ‘equal’
objects with respect to a given relation of equality (reflexive and transitive), for
example the length of a ruler. The introduction of the more abstract magnitude justi-
fies common parlance: for instance, two objects are said to have the same length.
Most important for Russell, magnitude has a simpler set of axioms than quantity,
and is more directly related to his theory of series.
Magnitudes in Russell’s sense are not in general measurable, unless measurability
only stipulates an order-homomorphic bijection with real numbers. After evoking
the latter (Machian) possibility (ibid., p. 176), Russell promptly turned to a more
‘important and intimate sense’ of measurement in which an intrinsic meaning is
given to the proposition ‘one magnitude is double of another’ (ibid., pp. 176, 177).
Like Helmholtz and Maxwell, he found this meaning in an operation of addition of

119
A convenient summary of Meinong’s views is in Meinong (1896), pp. 366–370. Meinong did not
refer to Helmholtz (1887).
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 561

quantities that did not necessarily reduce to a mere conjunction. But his discussion
(ibid., pp. 178–182) was much more abstract, based on concepts of ‘divisibilities’
and ‘distance’ partly inspired from Meinong and to be precisely defined in a theory
of the correlation of ordinal series.
Russell believed that intensive quantities, for which no addition is known, were
philosophically as important as measurable quantities (‘extensive quantities’). A
numerical scale could still be introduced for these purely ordinal structures. But the
only benefit of this broader kind of measurement was practical. It only served to
reduce the margin of doubt in appreciating the difference of things. This difference
was completely given in the ordinal structure, and did not in principle require the
ascription of numbers.120

3.1.6. Poincaré
Much closer to Helmholtz was the French mathematician Henri Poincaré. His book
of 1902 La science et l’hypothèse had a first chapter on ‘Number and quantity’ drawn
from earlier essays written in 1892–1894. In an attempt to elucidate the nature of
mathematical reasoning, Poincaré began with the following dilemma:

The very possibility of mathematical science seems an insoluble contradiction. If


this science is deductive only in appearance, wherefrom does it draw this perfect
rigor that no one thinks of questioning? If, on the contrary, all its propositions
can be derived from each other by the rules of formal logic, how is it that math-
ematics do not reduce themselves to a huge tautology?121

To answer this question, Poincaré examined ‘the purest part of mathematical


thought’, arithmetic. His analysis has enough similarity with the arithmetic section
of Helmholtz’s ‘Zählen und Messen’ to suggest a historical connection.
Poincaré assumed the operation a + 1 to be already given for every number a.
Then he defined the addition a + b inductively through a + (b + 1) = (a + b) + 1,
and derived its commutativity and associativity also by induction. Mathematical
induction, Poincaré commented, offered a way to condense in a single formula an
infinity of syllogisms. It was the very origin of the mathematician’s ability to induce
general truths from particular ones. It was neither reducible to pure logic nor to
experience, because of its infinite character. It was ‘the prototype of the synthetic a
priori judgment,’ or ‘the affirmation of the power of the mind which knows itself
capable of conceiving the indefinite repetition of the same act as soon as this act is
once possible’ (Poincaré, 1902, pp. 12–13).
With Helmholtz Poincaré shared the Grassmannian emphasis on inductive defi-
nitions and derivations, as well as the reference to internal intuition as the true foun-
dation of arithmetic. There are, however, some differences. For Helmholtz, internal

120
Russell (1903), pp. 182–183. In their Principia mathematica (1913) Russell and Whitehead adopted
Whitehead’s definition of magnitude, which, unlike Russell’s, automatically implies measurability. Cf.
Michell (1999), p. 64.
121
Poincaré (1902), p. 1.
562 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

intuition was already implied in the introduction of ordinal numbers, which contained
mathematical induction in germ. Moreover, Helmholtz conflated internal intuition
with a form of psychological experience whereas Poincaré emphasized the mathema-
tician’s need to transcend experience.122
Poincaré next discussed ‘mathematical quantity and experience’ in order to show
the origins and consistency of the mathematical continuum (Poincaré, 1902, pp. 17–
34). To this end he introduced Helmholtz’s distinction between arithmetic equality
and concrete equality:

The properties of equality are not an experimental truth exposed to refutation


through future, refined experiments. I prefer to believe with Helmholtz that we
give the name of equality to everything in the external world that meets the pre-
conceived idea we have of mathematical equality.123

He then gave a new twist to the transitivity criterion: whereas Helmholtz took
physical quantity to comply to this criterion, Poincaré argued that experiment sug-
gested the violation of transitivity. Indeed Weber had shown that the sensations pro-
duced by two weights A and B could be mutually discernable without being dis-
cernable from the sensation produced by a third, intermediate weight C. Even if a
more precise method was used to compare the weights, the precision of this compari-
son would always be finite, so that there would always be choices of the weights
for which
A ⫽ C, B ⫽ C, A ⬍ B.
This paradoxical triplet of relations, Poincaré went on, could be taken to define the
physical continuum. The mathematical continuum then emerged from our attempts
to solve the contradictions inherent in this empirical notion (Poincaré, 1893; Poin-
caré, 1902, pp. 21–25).
After these preliminaries, Poincaré espoused the arithmetic construction of the
continuum in Dedekind’s form. He agreed with the arithmetizers of mathematics that
a rigorous concept of the continuum had to rest on integral numbers and adequate
definitions. But he insisted that experience had motivated and inspired this construc-
tion: ‘The mathematical continuum . . . has been created from bits and pieces [de
toutes pièces] by our mind, but it is experience that has provided the occasion’
(Poincaré, 1902, p. 22). For example, the source of Dedekind’s cuts presumably was
the observation that the partition of a concrete line of finite thickness in two halves
always involved a separating point (of finite extension), no matter how thin the line
could be imagined.124
By continuum, Poincaré meant any linear quantity, not only real numbers. As the
indefinite divisibility and continuity of a quantity did not necessarily imply measur-

122
On Poincaré’s rejection of an empirical foundation of mathematics, see Karagozowski (forthcoming).
123
Poincaré (1892b), p. 75.
124
Poincaré may here have been inspired by Du Bois’s ‘empiricist’.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 563

ability, Poincaré introduced the equality of intervals, to be defined by a special con-


vention of measurement. He also noted that this convention, once given, provided a
definition of addition (of intervals starting from the same point): the interval AD is
said to be the sum of the intervals AB and AC when the interval BD is equal to the
interval AC. Poincaré of course required this addition to be commutative and associ-
ative. In the earlier version of this text, he referred to ‘von Helmholtz’s magisterial
work’ (1887) for a fuller discussion of this issue (Poincaré, 1893, p. 33).
Poincaré’s conception of measurement had indeed much similarity with
Helmholtz’s. He believed that any measurement required conventions of equality
and addition, and that the arbitrariness of these conventions was only limited by
arithmetic properties such as transitivity, commutativity, and associativity. For
example, in his lectures on heat theory (Poincaré, 1892a, pp. 16–17) Poincaré defined
the equality of temperatures of two bodies through the absence of expansion or
contraction during thermal contact;125 noted that only experiment could tell whether
this convention led to a transitive equality; and (momentarily) adopted the further
convention that equal temperature intervals corresponded to equal scale intervals on
the mercury thermometer. In his article of 1898 on the measurement of time, he
insisted that simultaneity as well as the equality of time intervals were conventional,
the best conventions being those for which the laws of mechanics and optics had
their simplest expression. In later discussions (1908) of the same topic he verified
that the conventional definition of simultaneity through the exchange of light signals
met the transitivity criterion for moving observers.126

3.1.7. Hölder
In 1901 the Leipzig professor of mathematics Otto Hölder proposed a new theory
of quantity that had some resemblance with Poincaré’s views. In a preliminary dis-
cussion of the concept of number, Hölder questioned Helmholtz’s qualification of
the formula a + (b + 1) = (a + b) + 1 as ‘axiom of arithmetic’ and supported the
view that ‘the arithmetic of whole numbers could be built in a purely logical manner
and required no axiom’ (Hölder, 1901, pp. 1–2). A more detailed consideration of
his position shows that it was not so far from Helmholtz’s. By pure logic, he did
not mean formal logic, but the unlimited application of certain thought processes
involved in inductive reasoning. This practice, he explained, is ‘a sort of experience,
which however cannot be counted as proper sensorial experience, since we can also
have it in thought’ (ibid., p. 2). The formula a + (b + 1) = (a + b) + 1 was to be
regarded as an inductive definition of addition, as Poincaré did and as Helmholtz
also meant despite his naming it ‘axiom’.
Hölder’s interest in the theory of quantity presumably derived from his study
(1900) of David Hilbert’s (1899) work on the foundations of geometry. Hilbert’s
central achievement was to base synthetic geometry on a small number of demon-
strably independent and non-contradictory axioms. In particular, he showed that met-

125
In conformity with Duhem’s and Mach’s definitions.
126
Cf. Darrigol (1995), p. 37.
564 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

ric properties could be derived from purely geometric axioms for the equality of
segments (Strecken) and the disposition of points on a line. To Hölder this meant
that an axiomatic theory of quantity could be given in the special case of segments.
His personal goal was to extend the axiomatic approach to more general quantities.
This extension could serve as ‘a preparation for the case when a physical state can
be measured’ (Hölder, 1900, p. 54). Hölder knew Maxwell’s additivity argument for
the measurability of charge, as well as the existence of properties, such as hardness,
that had degree but no quantity.
For the sake of historical perspective, Hölder recalled the theory of proportions
found in the fifth book of Euclid’s Elements. This theory concerns quantities of any
kind for which equality, order, and addition have been defined. It is built on the
compatibility of equality and order with addition, on the Archimedean property, and
on a few implicit properties such as the associativity of addition. Two quantities a
and b are said to be in the same ratio as two other quantities a⬘ and b⬘ when for
any two integers n and m the order of ma and nb is the same as the order of ma⬘
and nb⬘. Thanks to this ingenious definition, Euclid could treat commensurable and
incommensurable ratios on the same footing.127
A few centuries later, Isaac Newton simplified Euclid’s theory by defining number
as ‘ratio between a given abstract quantity and a quantity of the same kind regarded
as unit’ (Newton, 1707, p. 2). Then the ratio a:b always is a number, even when a
and b are incommensurable; and the proportionality a:b = a⬘:b⬘ simply means the
equality of the corresponding numbers. Hölder welcomed the simplification. But he
judged it so far unfounded (Hölder, 1901, p. 18), because in his view numbers had
to be defined by arithmetic means. The purpose of his own theory of quantity was
to provide the missing foundation.
Hölder had seven axioms of quantity, the independence of which he could prove.
Call Q the set of quantities of a given kind. In modern notation, the axioms read
(ibid., pp. 5–6):
I. ∀(a,b)苸Q2, a ⫽ b or a ⬍ b or b ⬍ a
II. ∀a苸Q, ∃b苸Q s.t. b ⬍ a
III. ∀(a,b)苸Q2, ∃c苸Q s.t. c ⫽ a ⫹ b
IV. ∀(a,b)苸Q2, a ⬍ a ⫹ b, b ⬍ a ⫹ b
V. ∀(a,b)苸Q2 s.t. a ⬍ b, ∃x苸Q s.t. a ⫹ x ⫽ b, ∃y苸Q s.t. y ⫹ a ⫽ b
VI. ∀(a,b,c)苸Q3, (a ⫹ b) ⫹ c ⫽ a ⫹ (b ⫹ c)
VII. ∀(A,B)苸(P (Q)⫺⭋)2 s.t.
[Q ⫽ A傼B, A傽B ⫽ ⭋, (∀(a,b)苸A ⫻ B, a ⬍ b)],
∃x s.t. [{a苸Q兩a ⬍ x}傺A, {b苸Q兩x ⬍ b}傺B]

127
Hölder (1901), pp. 3–4, 17–18. Euclid’s theory of ratios does not presuppose arithmetic. The only
notion that his definition of equal ratio requires is ‘equimultiplicity’. For example, instead of ma and
ma⬘, Euclid speaks of two quantities that are equimultiples of a and a⬘. Cf. Vitrac (1994), pp. 15–19.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 565

Hölder did not have axioms of equality such as symmetry and transitivity, because
he regarded equal quantities as identical. He was aware, however, that in concrete
cases this simplification required empirical testing.128 Also missing are the commuta-
tivity axiom and the Archimedean axiom. This is because Hölder could derive both
from his seven axioms. The last axiom is an axiom of continuity obtained by transpo-
sition of Dedekind’s cuts.
Hölder could easily prove that every couple (a, b) of quantities defined a Dedekind
cut among rational numbers, that is, a real number. In place of axiom VII, the Archi-
medean axiom would have led to the same result; but the quantities would not neces-
sarily have been continuous and their ratios could have all been rational.129
Hölder perceived a perfect agreement between his and Helmholtz’s conceptions
of the relation between arithmetic and the theory of quantity.130 To some extent,
Hölder’s idea of measurement may indeed be regarded as a refinement of
Helmholtz’s. They both defined quantity so as to fit structural properties of numbers.
Hölder’s refinement consisted in enlarging the arithmetic basis. Whereas Helmholtz
continued the Euclidian tradition of whole-number or rational measure, Hölder admit-
ted the newly arithmetized irrationals.
There were more significant differences in their intentions. Helmholtz regarded
his axioms of quantity as criteria for selecting measurable properties of concrete
objects. In contrast, Hölder wanted to build an abstract, mathematical theory of quan-
tity on the same footing as Hilbert’s abstract geometry. To him the non-contradiction
and independence of axioms were more important than applications to physics.
Consequently, his axioms were not well adapted to empirical tests of measurability.
He had neither the easily testable transitivity of equality nor the commutativity of
addition; and the continuity expressed in his last axiom was rigorously untestable.131

3.2. Physicists

Helmholtz’s ‘Zählen und Messen’ had little impact among physicists. Introductory
physics texts kept defining measurement as the comparison of a quantity with a unit
of the same kind, and kept avoiding any discussion of measurability. So did metro-
logical texts, save for one written by Carl Runge, the editor of Helmholtz’s introduc-
tory lectures on theoretical physics. The most interesting exceptions to this general

128
‘These facts [transitivity and compatibility of equality with addition] must of course acquire their
legitimacy in the applications’ (Hölder, 1901, p. 4n). In Russell’s terminology, Hölder’s quantities truly
are magnitudes.
129
In a last section, Hölder showed that Hilbert’s axioms for segments implied that distances obeyed
Hölder’s axioms of quantity. Cf. Michell (1999), pp. 74–75, who sees here an anticipation of the theory
of conjoint measurement.
130
Helmholtz ‘conceives the relation between arithmetic and theory of quantities exactly in the manner
that is executed in the present work’ (Hölder, 1901, p. 1n).
131
Ernst Nagel (1931) modified Hölder’s axioms so as to make them verifiable, in Helmholtz’s spirit.
For Nagel quantities are particular objects like for Helmholtz; for Hölder they are universals applying to
a class of objects. Patrick Suppes (1951) imitated Nagel, but with weaker axioms corresponding to a
broader concept of measurement. Cf. Michell (1993), pp. 198–199.
566 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

lack of interest are found among philosophically inclined physicists such as Ernst
Mach, Pierre Duhem, and Norman Campbell. We first consider the less philosophical
Carl Runge.132

3.2.1. Runge
A former student of Helmholtz’s and skilled spectroscopist, Carl Runge wrote the
article on measurement in Felix Klein’s encyclopedia (Runge, 1902). His familiarity
with Helmholtz’s discussion may explain why he began with general conditions of
measurement. His presentation was nonetheless closer to Maxwell’s and Mach’s. For
the introduction of a ‘scale of measurement’ adapted to a given property, he first
required the comparability of objects with respect to this property. This possibility
permitted the assignment of ‘order numbers’, such as the intensities of spectral lines
or the degrees of hardness of various solids in Mohr’s scale. In a second step, as a
means to reduce the arbitrariness of the scale, Runge introduced the comparability
of increments on this scale.
Worth noting is Runge’s emphasis on the conventional aspects of measurement.
Different laws could be used for the indirect measurement of a given quantity; and
the choice of the base units in a system of units was largely arbitrary. Against the
Gaussian creed, Runge declared that the choices of a mass, a length, and a time for
the fundamental units merely resulted from the practical requirement of reproduci-
bility (Runge, 1902, p. 10). Charles Edouard Guillaume, who augmented Runge’s
article for the French edition of Klein’s encyclopedia, warned the reader against an
anti-realist interpretation of this metrological relativity: ‘Nonetheless, independently
of the measure that can be made of it, every quantity has an existence of its own;
it constitutes a distinct entity of nature’ (Runge & Guillaume, 1916, p. 21).

3.2.2. Mach
The Viennese physicist-philosopher Ernst Mach essentially agreed with Helmholtz
on the nature of numbers. In his Mechanik, he introduced numbers as ‘order signs
[Ordnungszeichen] which, for the sake of perspicuity and economy, are themselves
arranged in a simple system’ and operations as a means to avoid direct numeration
(Mach, 1883, p. 583). After reading Helmholtz, he defined numbers as ‘a system of
order signs that can be indefinitely extended’ (Mach, 1896, p. 67).
He introduced addition in Grassmann’s and Schröder’s manner, and he embraced
Helmholtz’s empiricist view of arithmetic: ‘I regard the laws of arithmetic as experi-
mental laws, if only as laws that are drawn from internal experience’ (ibid., p. 67).
One difference is worth noting. Mach, unlike Helmholtz and more like Du Bois,
founded his empiricist understanding of numbers on a cultural and psychological
history of their introduction, from the comparison of groups of objects to the success-
ive extensions required for measurement or for inversion of operations. In Erkenntnis

132
Ludwig Boltzmann alluded to Helmholtz’s ‘Zählen und Messen’ in his fourth lecture on natural
philosophy (9 Nov, 1903), but only to criticize its arithmetic section. According to Boltzmann, (whole)
numbers were too primitive a notion to be defined at all. They could only be described through the
practice of calculation inherited from our ancestors. Cf. Courtenay (1999), pp. 393–400.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 567

und Irrtum (Mach, 1905, p. 328), he acknowledged the recent logicist approach to
arithmetic, but maintained that his historico-critical approach was a ‘necessary comp-
lement’ to the latter.133
In his definitions of mass and temperature, Mach anticipated Helmholtz’s emphasis
on the transitivity of equality. Yet he departed from Helmholtz by admitting, like
Maxwell, the possibility and the expediency of quantification by ‘inventory num-
bers,’ for which no natural operation of addition is known. The main purpose of his
detailed discussion of thermometry (Mach, 1896, Ch. 1–3) was to confine tempera-
ture to this limited role. According to Michael Heidelberger, Mach’s interest in a
weaker concept of quantity derived from his sympathy for Fechner’s psycho-physical
parallelism and from his general endeavor to make sensations the basis of all empiri-
cal knowledge (Heidelberger, 1986, pp. 163–167). Note, however, that Mach care-
fully avoided the word ‘measure’ in this context.
Mach defined measurement conservatively, as the comparison of a segment of a
linear continuum with a concrete unit. Measure, like numerical index, served the
purpose of refining the sensorial judgments of equality and difference (Mach, 1896,
Ch. 5). But besides a convention of concrete equality it also required an operation of
composition that shared the properties of arithmetic addition: ‘Whenever we regard a
physical quantity as composed of similar parts, we must consider whether this com-
position corresponds to a real addition’ (Mach, 1905, p. 330). For these conditions
of measurability Mach referred to Helmholtz’s ‘Zählen und Messen’. The very title
of the relevant chapter of Erkenntnis und Irrtum (Mach, 1905, pp. 315–330), ‘Zahl
und Mass’, was probably reminiscent of that essay.

3.2.3. Duhem
Pierre Duhem’s positions with respect to number and measure were similar to
Mach’s. With Helmholtz and Mach, Duhem pleaded for an empirical origin of math-
ematics:

Most of the abstract and general ideas which spontaneously come to us in the
occasion of our perceptions are complex and unanalyzed conceptions; yet some
of them appear to us clear and simple, almost without effort; those are the diverse
ideas clustering around the notions of number and figure; vulgar experience
induces us to connect these ideas through laws which, on the one hand, have the
immediate certitude of common-sense judgments, and which, on the other hand,
have extreme clarity and precision. It has thus been possible to take a few of
these judgments as premisses of deductions in which the incontestable truth of
common knowledge was inseparably bound to the perfect clarity of strings of
syllogisms. So were constituted arithmetic and geometry.134

133
Another singularity of Mach was his insistence on the discrete character (finite resolution) of percep-
tion. He nevertheless admitted the necessity of irrationals for representing geometric quantities (Mach,
1896, pp. 71–77).
134
Duhem (1914), pp. 404–405.
568 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

This idea of a spontaneous genesis of arithmetic (and geometric) axioms by abstrac-


tion differs, however, from Mach’s emphasis on a conscious striving for the economy
of thought. It also excludes Helmholtz’s and Poincaré’s admission of a synthetic a
priori component in the foundations of arithmetic. In 1912, Duhem vehemently
rejected Poincaré’s understanding of the principle of mathematical induction, and
argued that arithmetic reasoning was purely syllogistic and based on axioms of
empirical origin.135
Duhem was closest to Helmholtz (but didn’t refer to him) when he defined quantity
through relations of equality, inequality, and a concrete operation of addition, with
the usual properties of transitivity, associativity, commutativity, and compatibility
(Duhem, 1914, pp.159–163). He also indicated how this definition permitted the
expression of every quantity in terms of a concrete number made of a real number
and a unit. However, Duhem departed from Helmholtz by rigidifying his distinction
between direct and indirect measurement, or Maxwell’s similar distinction between
quality and quantity. For Helmholtz and Maxwell, a property was a quality in the
temporary lack of a concrete addition.136 For Duhem, a property identified as a qual-
ity had to remain a quality for ever (Duhem, 1914, pp. 163–166).
The reason for this difference is that the only concrete additions Duhem admitted
were those inherent in ‘physical notions’ belonging to the everyday experience of
common men. Whereas measured temperature always was a quantity for Helmholtz
and could become so for Maxwell, Duhem believed it could only be a quality,
because there was no additivity for the common-sense notion of heat:

These words to be warm correspond to a property of all the infinitesimal parts in


which bodies can be divided. What is this property in itself? Is it reducible, in
its very nature, to quantitative elements? Those are questions that physics does
not have to resolve. As we conceive it, this property is not quantitative. It appears
to us as susceptible of identical reproduction, of increase and decrease, but not
susceptible of addition.137

Of course, Duhem admitted a mathematical concept of temperature, represented


by numbers. But this temperature ‘served to mark out [repérer], not to measure’
(Duhem, 1892–1894, p. 286). In general, the ill-named thermometers supposed a
conventional correspondence between a quality and a measurable quantity. But addi-
tivity for this quantity was not to be confused with additivity for the quality. In a
partial rehabilitation of Aristotelian physics, Duhem pleaded for the irreducibility of
quality to quantity. He held the Cartesian drive to reduce every quality to quantity
responsible for the proliferation of mechanistic theories of everything (Duhem, 1914,
pp. 166–170). In his conception of physical theory, qualities and quantities had equal

135
Duhem (1912). Cf. Karagozowski (forthcoming).
136
Maxwell and Helmholtz nevertheless disagreed on whether thermometric temperature was a quality
in this sense.
137
Duhem (1892–1894), p. 286.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 569

legitimacy, they were both represented by numbers, they both entered the fundamen-
tal equations of physics (ibid., pp. 170–178).

3.2.4. Campbell
Despite the richness of previous considerations of measurability, the British exper-
imenter Norman Campbell is often regarded as the founder of the theory of measure-
ment. The reason may be that he devoted several chapters in 1920, and a whole
book in 1928 on measurement. Also, instead of references to the work of others he
presented the excuse that he had lost his ‘large index of references’ during the war
(Campbell, 1920, p. vi). This attitude and his prolix style could suggest that his
arguments were new, even though the preface of Physics: the elements contains the
unusual admission: ‘There is very little original in the substance of the book; there
is hardly a paragraph which is not a paraphrase of something that can be found in
well-known treatises or papers’ (ibid., p. vi).
Campbell’s acknowledged masters were Poincaré and Russell (Campbell, 1920,
p. vii). Some aspects of his approach to number and measurement clearly depended
on these authors. For instance, Campbell adopted Poincaré’s characterization of the
physical continuum (ibid., p. 543); and he followed Russell in placing order before
addition in his definition of measurable properties. However, Campbell most system-
atically relied on a text he never mentioned: Helmholtz’s ‘Zählen und Messen’. The
similarities are numerous and striking.
Like Helmholtz, Campbell gave priority to ordinal numbers and characterized
them through the corresponding numerals (Campbell, 1920, pp. 269–270).138 He
introduced ‘physical numbers’ as resulting from a peculiar kind of measurement, the
counting of a group of similar objects (ibid., Ch. 11). His definition of fundamentally
measurable properties required, besides Russellian order, a concrete addition with
the proper arithmetic properties (ibid., Ch. 10). He insisted that ‘it is experiment and
experiment only that can determine whether a property is fundamentally measurable’
(ibid., p. 267). He introduced fractional numbers as a consequence of the division
of units (ibid., Ch. 12). He denied the relevance of irrational numbers to measurement
(ibid., p. 543). He admitted indirectly measurable properties, for which no addition
is known but relations with additive properties are assumed (ibid., Ch. 13). He noted
the similarity between derived and fundamental magnitudes: ‘Both contain an arbi-
trary element connected with the choice of unit, but both are definite properties of
a system’ (ibid., p. 328). He recognized that the same property could sustain both
a fundamental and a derived magnitude.
In all of that Campbell followed Helmholtz. There was, however, a difference in
the intended audience. Whereas Helmholtz wrote his ‘Zählen und Messen’ for an
elite of mathematicians, philosophers, and theoretical physicists, Campbell believed
in the pedagogical virtues of a general discussion of measurement in introductory

138
Campbell did not formalize the ordinal structure, and rather insisted on the ‘quasi-material’ nature
of numerals. He introduced ‘Numbers’, with capital N, to refer to the Frege–Russell concept of cardinal
numbers as classes of similar classes. But he made relatively little use of them.
570 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

physics courses (Campbell, 1928, p. V). Physics teachers did not hear his call. They
stuck to the succinct definition of measurement as comparison with a unit of the
same kind. They ignored the Helmholtzian criteria of measurability, and they left it
to philosophers and psychologists to decide what could in principle be measured.

Acknowledgements

I am thankful to Anouk Barberousse, Hasok Chang, and Nadine de Courtenay for


their comments on an early draft of this paper; to an anonymous reviewer for useful
suggestions; to Cathy Carson for her hospitality at Berkeley’s OHST, in which most
of this research was done.

References

Belna, J.-P. (1996). La notion de nombre chez Dedekind, Cantor, Frege: Théories, conceptions, philoso-
phie. Paris: Vrin.
Cahan, D. (1989). An institute for an empire: The Physikalisch-Technische Reichsanstalt 1871–1918.
Cambridge: Cambridge University Press.
Cahan, D. (1993). Helmholtz and the civilizing power of science. In D. Cahan (Ed.), Hermann von
Helmholtz and the foundations of nineteenth-century science (pp. 559–601). Berkeley: University of
California Press.
Cahan, D. (2002). Helmholtz, Hermann Ludwig Ferdinand von (1821–1894). In International encyclo-
pedia of the social and behavioral sciences. Amsterdam: Elsevier.
Campbell, N. (1920). Physics: the elements. Cambridge: Cambridge University Press.
Campbell, N. (1928). An account of the principles of measurement and calculation. London: Longman,
Green & Co.
Cantor, G. (1887). Mitteilungen zur Lehre vom Transfiniten. Zeitschrift für Philosophie und philosophis-
che Kritik. In G. Cantor (Ed.), Gesammelte Abhandlungen (pp. 378–439). Berlin: Springer.
Chang, H. (1995). Circularity and reliability in measurement. Perspectives on Science, 3, 153–172.
Chang, H. (2001). Spirit, air, and quicksilver: The search for the ‘real’ scale of temperature. Historical
Studies in the History of Physical and Biological Sciences, 31, 249–284.
Châtelet, G. (1993). Les enjeux du mobile. Mathématique, physique, philosophie. Paris: Le Seuil.
Courtenay, N. de (1999). Science et philosophie chez Ludwig Boltzmann: La liberté des images par les
signes. PhD thesis, Université de Paris IV.
Crowe, M. (1987). A history of vector analysis. Notre Dame: University of Notre Dame Press.
Darrigol, O. (1995). Henri Poincaré’s criticism of fin de siècle electrodynamics. Studies in History and
Philosophy of Modern Physics, 26, 1–44.
Dedekind, R. (1872). Stetigkeit und irrationale Zahlen. Braunschweig: Vieweg. Translated in R. Dedekind,
Essays on the theory of numbers (pp. 1–27). Chicago: Open Court.
Dedekind, R. (1888). Was sind und was sollen die Zahlen? Braunschweig: Vieweg. Translated in Essays
on the theory of numbers (pp. 31–115). Chicago: Open Court.
Diez, J. A. (1997). A hundred years of numbers, an historical introduction to measurement theory 1887–
1990. Studies in History and Philosophy of Science, 28, 167–185, 237–265.
DiSalle, R. (1993). Helmholtz’s empiricist philosophy of mathematics. Between laws of perception and
laws of nature. In D. Cahan (Ed.), Hermann von Helmholtz and the foundations of nineteenth-century
science (pp. 498–521). Berkeley: The University of California Press.
Du Bois-Reymond, P. (1882). Die allgemeine Functionentheorie. Erster Theil. Metaphysik und Theorie
der mathematischen Grundbegriffe: Grösse, Grenze, Argument und Function. Tübingen: Laupp.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 571

Dugac, P. (1976). Richard Dedekind et les fondements des mathématiques. Paris: Vrin.
Duhem, P. (1892–1894). Commentaire aux principes de la thermodynamique. Journal de mathématique
pure et appliquée, 8, 269–330, 9, 293–359, 10, 207–285.
Duhem, P. (1914). La théorie physique. Son objet, sa structure (2nd ed.). Paris: Rivière.
Duhem, P. (1912). La nature du raisonnement mathématique. Revue de philosophie, 21, 531–543.
Elsas, A. (1886). Über die Psychophysik. Marburg: Elwert.
Fechner, G. (1859). Elemente der Psychophysik. Leipzig: Breitkopf und Hartel.
Fischer, G. (1981). The infinite and infinitesimal quantities of Du Bois-Reymond and their reception.
Archive for the History of Exact Sciences, 24, 101–163.
Flament, D. (1994). Hermann Günther Grassmann: L’homme et l’oeuvre. In H. Grassman (Ed.), La science
de la grandeur extensive (pp. 7–47). Paris: Blanchard. Translation of Grassmann (1844).
Frege, G. (1884). Die Grundagen der Arithmetik. Breslau: Köbner.
Frege, G. (1891). Funktion und Begriff. In Funktion, Begriff, Bedeutung. Fünf logische Studien. Göttingen:
Vandenhoeck & Ruprecht (1975).
Grassmann, H. (1844). Die lineale Ausdehnungslehre. Leipzig: Wigand.
Grassmann, H. (1861). Lehrbuch der Arithmetik für höhere Lehranstalten. Berlin.
Grassmann, H. (1894–1911). Gesammelte mathematische und physikalische Werke (3 vols.) (F. Engel,
Ed.). Leipzig: Teubner.
Grassmann, R. (1872). Die Formenlehre oder die Mathematik. Stettin: Grassmann. Five books bound as
one. Page numbers in citations refer to the first book, Die Grösenlehre [sic]. The other books are Die
Begriffslehre oder Logik, Die Bindenlehre oder Combinationslehre, Die Zahlenlehre oder Arithmetik,
and Die Ausenlehre [sic] oder Ausdehnungslehre.
Grassmann, R. (1890a). Das Gebäude des Wissens. Stettin: Grassmann.
Grassmann, R. (1890b). Die Sprachlehre. Stettin: Grassmann.
Grassmann, R. (1890c). Die Denklehre. Stettin: Grassmann.
Grassmann, R. (1891). Die Zahlenlehre oder Arithmetik. Stettin: Grassmann.
Hardy, G. H. (1910). Orders of infinity: The Infinitärcalcül of Paul Du-Bois Reymond. Cambridge: Cam-
bridge University Press.
Harman, P. (1998). The natural philosophy of James Clerk Maxwell. Cambridge: Cambridge Univer-
sity Press.
Heidelberger, M. (1986). Zur Philosophie der Messung im 19. Jahrhundert. In H. Witthöft (Ed.), Die
historische Metrologie in den Wissenschaften (pp. 159–168). St. Katharinen: Scripta Mercaturae.
Heidelberger, M. (1993). Fechner’s impact for measurement theory. Behavioral and Brain Sciences, 16,
146–148.
Heine, E. H. (1872). Die Elemente der Funktionenlehre. Journal für die reine und angewandte Mathema-
tik, 74, 172–188.
Helmholtz, H. (1856–1867). Handbuch der physiologischen Optik (3 vols.). Leipzig: Voss.
Helmholtz, H. (1866). Über die Thatsächlichen Grundlagen der Geometrie. In Helmholtz, Wissen-
schaftliche Abhandlungen (3 vols.) (Vol. 2, pp. 610–617). Leipzig: Barth.
Helmholtz, H. (1868). Über die Thatsachen, die der Geometrie zum Grunde legen. In Helmholtz, Wissen-
schaftliche Abhandlungen (3 vols.) (Vol. 2, pp. 618–639). Leipzig: Barth.
Helmholtz, H. (1876). Über den Ursprung und die Bedeutung der geometrischen Axiome. In Helmholtz,
Vorträge und Reden (2 vols.) (Vol. 2, pp. 1–31). Braunschweig: Vieweg.
Helmholtz, H. (1878a). Die Thatsachen in der Wahrnehmung. In Helmholtz, Vorträge und Reden (2 vols.)
(Vol. 2, pp. 215–247). Braunschweig: Vieweg.
Helmholtz, H. (1878b). Über den Ursprung und Sinn der geometrischen Axiome: Antwort gegen Herrn
Professoren Land. In Helmholtz, Wissenschaftliche Abhandlungen (3 vols.) (Vol. 2, pp. 640–660).
Leipzig: Barth.
Helmholtz, H. (1882–1895). Wissenschaftliche Abhandlungen (3 vols.). Leipzig: Barth.
Helmholtz, H. (1887). Zählen und Messen, erkenntnisstheoretisch betrachtet. In Philosophische Aufsätze,
Eduard Zeller zu seinem fünfzigjährigen Doctorjubiläum gewidmet. Leipzig: Fues. Also in Helmholtz,
Wissenschaftliche Abhandlungen (3 vols.) (Vol. 3, pp. 356–391). Leipzig: Barth.
Helmholtz, H. (1903). Einleitung zu den Vorlesungen über theoretische Physik. Leipzig: Barth. Edited
by A. König & C. Runge after steno of 1893.
572 O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573

Helmholtz, H. (1908). Vorlesungen über die Theorie der Wärme. Leipzig: Barth. Edited by F. Richarz
after notes of 1890 and steno of 1893.
Hertz, P. (1921). Notes to H. Helmholtz, Schriften zur Erkenntnistheorie. Berlin: Springer.
Hilbert, D. (1899). Grundlagen der Geometrie. Leipzig: Teubner.
Hölder, O. (1900). Anschauung und Denken in der Geometrie. Leipzig: Teubner. Inaugural lecture for
the Leipzig Academy, 22 July 1899.
Hölder, O. (1901). Die Axiome der Quantität und die Lehre vom Mass. Königlich Sächsische Gesellschaft
der Wissenschaften zu Leipzig, Mathematisch-physikalische Classe. Berichte über die Verhandlungen,
53, 1–46.
Holmes, F. L. (1994). The role of Johannes Müller in the formation of Helmholtz’s physiological career.
In L. Krüger (Ed.), Universalgenie Helmholtz. Rückblick nach 100 Jahren (pp. 3–21). Berlin: Akade-
mie Verlag.
Husserl, E. (1891). Philosophie der Arithmetik. Niemeyer: Halle.
Hyder, D. J. (2001). Physiological optics and physical geometry. Science in Context, 14, 419–456.
Jurkowitz, E. (2002). Helmholtz and the liberal unification of science. Historical Studies in the Physical
and Biological Sciences, 32, 291–317.
Kant, I. (1787). Kritik der reinen Vernunft (2nd ed.). Riga: Hartnoch.
Karagozowski, S. (forthcoming). Etude comparative des philosophies de Pierre Duhen et de Henri Poin-
caré à travers leurs pratiques scientifiques. Ph.D. thesis, Université de Paris VII.
Klein, F. (1890). Zur nicht-Euklidischen Geometrie. Mathematische Annalen, 37, 544–572.
Klein, F. (1902). Elementarmathematik vom höheren Standpunkte aus. Berlin: Springer.
Koenigsberger, L. (1902–1903). Hermann von Helmholtz (3 vols.). Braunschweig: Vieweg.
Krantz, D. H., Suppes, P., Luce, R. D., & Tversky, A. (1971). Foundations of measurement (Vol. 1).
New York: Academic Press.
von Kries, J. (1882). Über die Messung intensiver Grösse und über das sogenannte psychophysiches
Gesetz. Vierteljahrschrift für wissenschaftliche Philosophie (Leipzig), 6, 257–294.
Kronecker, L. (1887). Über den Zahlbegriff. In Philosophische Aufsätze, Eduard Zeller zu seinem
fünfzigjährigen Doctorjubiläum gewidmet (pp. 263–274). Leipzig: Fues.
Luce, R. D., & Suppes, P. (1981). Measurement, theory of. In Encyclopedia Britannica, 15th ed., Macro-
paedia, 11, 739–745.
Mach, E. (1872). Die Geschichte und die Wurzel des Satzes von der Erhaltung der Arbeit. Prague: Calve.
Mach, E. (1883). Die Mechanik in ihrer Entwicklung. Historitisch-kritisch dargestellt. Leipzig: Brockhaus.
Mach, E. (1896). Die Principien der Wärmelehre. Historitisch-kritisch entwickelt. Leipzig: Barth.
Mach, E. (1905). Erkenntnis und Irrtum. Leipzig: Barth.
Maxwell, J. C. (1871). Theory of heat. London: Longman, Greens & Co.
Maxwell, J. C. (1873). A treatise on electricity and magnetism (2 vols.). Oxford: Clarendon Press.
Meinong, A. (1896). Über die Bedeutung des Weber’schen Gesetzes. In A. Meinong, Gesammelte Abhand-
lungen (2 vols.) (Vol. 2, pp. 215–372). Leipzig: Barth.
Michell, J. (1993). The origins of the representational theory of measurement: Helmholtz, Hölder and
Russell. Studies in History and Philosophy of Science, 24, 185–206.
Michell, J. (1999). Measurement in psychology. Critical history of a methodological concept. Cambridge:
Cambridge University Press.
Nagel, E. (1931). Measurement. Erkenntnis, 2, 313–333.
Newton, I. (1707). Arithmetica universalis. Cambridge: Typis Academicis.
Olesko, K. (1994). Civic culture and calling in the Königsberg period. In L. Krüger (Ed.), Universalgenie
Helmholtz. Rückblick nach 100 Jahren (pp. 22–42). Berlin: Akademie Verlag.
Poincaré, H. (1892a). Thermodynamique. Paris: Gauthier-Villars. Sorbonne lectures of 1888–1889.
Poincaré, H. (1892b). Sur les géométries non euclidiennes. Revue générale des sciences pures et
appliquées, 3, 74–75.
Poincaré, H. (1893). Le continu mathématique. Revue de métaphysique et de morale, 1, 26–34.
Poincaré, H. (1898). La mesure du temps. Revue de métaphysique et de morale, 6, 371–384.
Poincaré, H. (1902). La science et l’hypothèse. Paris: Flammarion.
Poincaré, H. (1908). La dynamique de l’électron. Revue générale des sciences pures et appliquées, 19,
386–402.
O. Darrigol / Stud. Hist. Phil. Sci. 34 (2003) 515–573 573

Pringsheim, A. (1898). Irrationalzahlen und Konvergenz unendlicher Prozesse. In Encyklopädie der


mathematischen Wissenschaften mit Einschluss ihrer Anwendungen (Vol. 1:1, pp. 47–147). Leipzig:
Teubner.
Pringsheim, A., & Molk, J. (1904). Nombres irrationnels et notion de limite. In Encyclopédie des sciences
mathématiques pures et appliquées (Vol. 1:1, pp. 133–238). Paris.
Runge, C. (1902). Mass und Messen. In Encyklopädie der mathematischen Wissenschaften mit Einschluss
ihrer Anwendungen (Vol. 5:1, pp. 3–24). Leipzig: Teubner.
Runge, C. & Guillaume, E. (1916). La mesure. In Encyclopédie des sciences mathématiques pures et
appliquées (Vol. 5:1, pp. 1–64). Paris.
Russell, B. (1903). The principles of mathematics. London: Allen & Unwin.
Russell, B., & Whitehead, A. N. (1913). Principia mathematica (Vol. 3). Cambridge: Cambridge Univer-
sity Press.
Savage, C. W., & Ehrlich, P. (1992). A brief introduction to measurement theory and to the essays. In
C. W. Savage, & P. Ehrlich (Eds.), Philosophical and foundational issues in measurement theory (pp.
1–14). Hillsdale: Erlbaum.
Schaffer, S. (1995). Accurate measurement is an English science. In N. Wise (Ed.), The values of precision
(pp. 135–172). Princeton: Princeton University Press.
Schröder, E. (1873). Lehrbuch der Arithmetik und Algebra. Leipzig: Teubner.
Schröder, E. (1877). Operationskreis des Logikkalkulus. Leipzig: Teubner.
Schröder, E. (1890–1895). Vorlesungen über die Algebra der Logik (3 vols.). Leipzig: Teubner.
Schubert, H. (1898). Grundlagen der Arithmetik. In Encyklopädie der mathematischen Wissenschaften
mit Einschluss ihrer Anwendungen (Vol. 1:1, pp. 1–27). Leipzig: Teubner.
Stevens, S. S. (1946). On the theory of scales of measurement. Science, 103, 667–680.
Suppes, P. (1951). A set of independent axioms for extensive quantities. Portugaliae mathematica, 10,
163–172.
Turner, S. (1982). Helmholtz, sensory physiology, and the disciplinary development of German psy-
chology. In W. R. Woodward, & G. A. Mitchell (Eds.), The problematic science: Psychology in
nineteenth-century thought (pp. 147–166). New York: Praeger.
Vischer, F. (1887). Widmung zum Zeller Jubiläum. In Philosophische Aufsätze, Eduard Zeller zu seinem
fünfzigjährigen Doctorjubiläum gewidmet. Leipzig: Fues.
Vitrac, B. (1994). Introduction to Euclide, Les éléments, Vol. 2, Livres V–VI. Paris: Presses Universitaires
de France.
Wundt, W. (1883a). Über die Messung psychischer Vorgänge. Philosophische Studien, 1, 251–260.
Wundt, W. (1883b). Weitere Bemerkungen über psychische Messung. Philosophische Studien, 1, 461–
471.
Wundt, W. (1920). Erlebtes und Erkanntes. Stuttgart: Kröner.
Zeller, E. (1881). Über die Messung psychischer Vorgänge. Königliche Akademie der Wissenschaften zu
Berlin, Philosophische und historische Abtheilung. Abhandlungen, 3, 1–16.
Zeller, E. (1882). Einige weitere Bemerkungen über die Messung psychischer Vorgänge. Königliche
preussische Akademie der Wissenschaften, Sitzungsberichte, 295–305.

You might also like