You are on page 1of 11

1

SEEKING NEW FOUNDATIONS FOR THE THEORY OF MAGNITUDES:


THE THEORY OF THE CONTINUUM AND APPLICATIONS IN MODERN PHYSICS

By David Braine

I wish to give the background of the treatment of our uses of expressions for numbers, and our
treatment of the magnitude of sets of discrete entities or subjects, and of continua such as occurs in the
Introduction and Chapters I, II, IV, and V of my book Language and Human Understanding. I wish to do
this by amplifying a paper I wrote in the 1990’s – also noting questions raised by quantum cosmology.

Aristotle's fundamental insight was that a point was not a part of a line but a dividing point or cut
in a line or motion into sub-lines or sub-motions. (In a parallel way, a line can be considered as a division
between sub-surfaces within a surface, and a surface as a division between sub-spaces within a space.)
This accords with Newton’s conception of a line as marking the path of a moving point, and this
idea signalled the essential connection between the continuity of a line and the continuity of motion.
Any straight line is segmentable, and so should be considered only as consisting of segments, not
as consisting of points: a point is not a segment or part of a line. Rather, a point is a cut, dividing point or
mark of division in a line, L, enabling its consideration as consisting of segments or sub-lines, L1 and L2,
which we shall represent {L1, L2}. This cut or mark of division between such sub-lines or segments,
called a point, has no length, and does not constitute a hole or gap in the line or motion.
The intuitionists, having treated the intuition of continuum as quite separate from the intuition of
number, nonetheless give the appearance of fabricating spreads such as would feel about to make sure the
series of what would be points in the real line had no gaps; and, using infinite decimals as a mode of
defining numbers, thinking of these as potentially (as opposed to actually) infinite sequences of the
integers from 0 to 9, the next integer in the decimal being chosen by free choice 1 without reference to any
geometrical construction or topological conception, and presumably not marked by any physical event,
these two alone identifying places in the simple order of cuts or dividing points in the real line. 2
Thus, if a series of points is known to have an upper bound, e.g., as the series 3.14, 3.141, 3.1415,
3.14159, 3.141592, 3.1415926, 3.14159265, 3.141592653, 3.1415926535, 3.14159265359, and so on,
unendingly, this will be because there is some construction such as a circle of radius length 1 around a
point, and opening out a string fixed to a point on the circumference of the circle into a straight line, and
marking its length by its end point—or, granted a vertical cylinder whose horizontal section was such a
circle, measuring the height of the liquid it can contain by decanting it into a rectangular jar whose square
sections have sides of length 1, and considering the height of the fluid contained. The constructive law
(excluding laws unconnected with topology or geometry, but instead simply dictating the adjustment of
digits in a decimal (or binary, or other such) representation 3) identifying such a converging series will
always originate in some construction or equation which gives reason for expecting there to be a least
1 L.E.J. Brouwer spoke at Göttingen (L.E.J. Brouwer—Topologist, Intuitionist, Philosopher: How Mathematics Is
Rooted in Life by Dirk von Dalen, 2012) in 1924 of the absurdity of the notion of the real numbers being represented by
actually infinite series of decimals. Such representation is required by Cantor’s diagonal procedure used in proving the absence
of any one-to-one correspondence between the real numbers and any inductively infinite series, even one embracing all
algebraic numbers. Such actual infinity allows a number to be defined in terms of whether or not there were a series of ten 9’s
in the decimal expansion of π.
2 In Brouwer’s conception, the idea of spread originated with the “mental intuition” of number (in his view quite separate
from the “mental intuition” of the continuum) and with exploring types of sequence of integer. It was, as it were, a bonus of
providence that the arithmetic of these sequences of integers can throw light upon the mathematics of the continuum.
3 I say this in order to exclude such definitions of a number as are mentioned in Foundations of Set Theory, Fraenkel, Bar
Hillel and Levy, 2nd Edition, 1973, pp. 266-67, in the following words: “It is unknown whether in the decimal expansion of a
sequence of (at least) seven 7’s will occur; if so, let h denote the place after the decimal point by Fraenkel (i.e. the index of the
respective digit in n =3a la2 ...ah...) where for the first time such a sequence starts. Now we define a decimal ρ, which begins
with 0.777 ..., as follows: if h exists as defined, replace the digit 7 at the hth place after the point in ρ by 6 if h is odd and by 8
if h is even, hence all digits of ρ are 7 ‘if no h exists.’ The real number ρ is effectively defined since, simultaneously with the
expansion of n, the successive digits of ρ can be calculated.”
1
2

upper bound, and there is no way in which there can be a hole or gap between such a series, and the limit
thus identified. In general, in this paper I shall use the expression “constructively definable” to exclude
cases where representation based laws unconnected with topology or geometry have entered into the
supposed “construction.”
Likewise, in some cases, Gödel’s Incompleteness theorem, and any succession of applications of
Gödel’s theorem up to but not including ε 0, allows a constructive proof of a certain result (one must not
reach ε0 in such a way as to be able to treat it as opening out room for further enumerations without going
beyond any kind of constructivity). But there is no reason for thinking that this would yield a converging
series, such as would allow a definition of a cut in the real line in the stricter sense I have explained.
Contrary to this underlying insight that in no way does the continuous line consist of the discrete
markers of conceptual or actual possible division, the modern theory of the continuum has represented a
line as a certain set of such discrete numbers or points in certain relations. The result of this has been that
the theory of the continuum has been considered derivative from the theory of sets of discrete entities.
From this, many awkwardnesses have followed: One, awkwardnesses in the definition of the
continuity of lines: intuitively, to be continuous, a function defining a line has to be differentiable at all
but a finite number of points within any finite length (have a gradient or direction). This is not true of
functions as treated in the classical analysis of modern times. Two, the fundamental theorem of analysis
called the Least Upper Bound Principle has an intuitive naturalness in application to dividing-points or
cuts in the continuum of a line through space presupposed by motion through space, together with the
magnitudes of things measured by relation to these (time, relative mass, momentum, energy, and so
forth).
By contrast, the postulates of those general theories of sets from which it is suggested that one should
derive this theorem will be found in each case to be subject to much greater doubt—not only because
supposedly natural postulates have led to paradoxes (e.g., the existence of nowhere-differentiable-lines
and the Banach-Tarski paradox), but also because in some cases the postulates have no great obviousness
(e.g., the axiom of choice where a set exists consisting of just one out of each of infinite number of sets,
so that the existence of this choice-set depends upon an actually infinite number of choices being made at
the same time). The non-constructive character, even of this relatively weak form of the axiom, appears in
§. 2.
In considering the roots of the Least Upper Bound Principle, we should consider the possibility of
treating the theory of the continuum along the lines exemplified in Euclidean geometry, its extensions,
and variants. We want a theory such that if there is any construction, using this word in a very extended
sense so that, e.g., it includes the straightening of curved or bent lines so long as length is kept constant,
which allows the identification of a cut or division point in a line, then there will be a number
corresponding to this. A proof of the Least Upper Bound Principle of what I describe as a geometrical
kind will only require that such genuinely constructible points have a least upper bound if they have an
upper bound.
One’s approach should be closer in spirit, method, and style to Euclid than to that of classical
analysis.
The challenge will be to develop a new form of analysis on a basis of this Euclidean kind adequate to
prove the results necessary to the applications of mathematics in legitimate modern science.

I. Comment on the status of general set theory and the arithmetic of transfinite cardinals, which
will be cardinals which do not assign magnitude

Any constructively ordered subset of the natural numbers taken in their natural order is finite but it is
not reflexive (having a one-to-one correspondence with a proper subset of itself).
The ordered set of all the natural numbers taken in their natural order constitutes a set of a kind we
may describe as recursively or inductively infinite, leaving open the question whether it should be
considered a totality (recursively as arising by a repetition of the same operation again and again, and

2
3

inductively in being subject to the application of the principle of mathematical induction; see below for
the discussion of the notion of potential infinity and the different ways of being potentially infinite). The
even numbers, the prime numbers, the perfect squares, perfect cubes, are all subsets, each capable of one-
to-one correspondence with the natural numbers. No such sets define magnitudes, as Galileo realized. (I
consider the fashionable restriction on the notion of set to be arbitrary, see n. 14.)
In Cantor’s conception, two sets have the same cardinality or cardinal number if they can be put into
one-to-one correspondence with each other. Evidently, the natural numbers, the odd numbers, the even
numbers, the squares, cubes, etc., of natural numbers, and by Euclid’s proof the prime numbers, can each
be put into one-to-one correspondence with each other, and it is from this that it is clear that cardinality is
not a measure of magnitude. All these sets are also reflexively infinite in that they can be put into one-to-
one correspondence with a proper subset of themselves.
Sets would appear to be of three kinds: (1) finite, having a one-to-one correspondence with an initial
segment of the positive integers and shown to be non-reflexive by mathematical induction, (2)
denumerable (“countably infinite”), having a one-to-one correspondence with the set of all the integers,
sets which are always reflexive, having a cardinality of 0‫אּ‬, and (3) non-denumerable sets, that is, infinite
sets lacking such correspondence with the set of all the integers but with denumerable subsets and
therefore also reflexive, conceived by Cantor as infinite sets of greater cardinality than denumerable sets.
At this stage, we are on relatively concrete ground. However, as I shall remark in §. 3, the set of
definitions of the real numbers is a subset of the set of finite strings of symbols in a language with a finite
set of symbols, and therefore certainly has a one-to-one correspondence with the positive integers, i.e., is
denumerable. But there is no way of determining whether something defines a real number or not, and
this raises the question as to whether being denumerable is not a form of being “indefinite in number.”
This will raise the question as to whether we can form a classification of different ways of being
“indefinite in number.”
It is against this background that we should consider Gödel’s question as to whether there may not be
different ways of being “denumerable” and different ways of being “non-denumerable.” Thus, amongst
non-denumerable sets (3), there are commonly supposed to be many sets of cardinality 2 0‫ אּ‬which, if one
adopts a Platonic approach whereby sets exist independently of there being any linguistically formulable
predicate of which they are the extension (contrary to §. 2 below), will be the cardinal number of the set
of all the subsets of the set of all the integers, and, if real numbers can be actually identified with what
would be denoted by an actually infinite decimal, the number of the real numbers.
Considered in the traditional way, obscurities arise even on a Platonic way of thinking of numbers.
Thus, on this Platonic way of thinking, two initial obscurities arise. First, there is a question raised by
Lebesgue in 1904 as to whether there are any “Dedekind-finite” sets (what the first (1958) edition of
Abraham A. Fraenkel and Yehoshua Bar-Hillel’s The Foundations of Set Theory, pp.62-64, called “of
mediate cardinality”), neither finite in the sense of having a one-to-one correspondence with an initial
segment of the positive integers (called “inductive” by Fraenkel and Bar-Hillel in 1958), nor reflexive.
Such sets, Fraenkel remarks, could not be well ordered.
The existence of such sets is incompatible with the axiom of choice in the weak form which involves
choices from a set of ordinal number ω, that is, with the axiom of dependent choice, which says that one
can pick a countable sequence where each element is in a non-empty set determined by the elements
chosen so far.
From the non-existence of a relation which allows well-ordering of such sets, it follows a fortiori that
there is no linguistically expressible relation which achieves this. But this will be required for the
identification and therefore of their members according to the position taken in §. 2.
Secondly, there is a question as to whether there are denumerable sets which include non-standard
elements in addition to zero and the positive integers, but conform to Peano’s axioms. However, the
existence of such sets is said to be a consequence of Gödel’s Completeness Theorem and its corollary the
Compactness Theorem, whose proofs are non-constructive. The non-constructive character of Gödel’s
proof of his Completeness Theorem (aimed at proving that every logically valid formula in first-order
predicate logic is provable) in his dissertation of 1929 was something he recognized in the introduction,
which includes the remark that "essential use is made of the principle of the excluded middle for infinite
3
4

collections... It might perhaps appear that this would invalidate the entire completeness proof.” 4 In any
case, Gödel’s proof of this Completeness has been shown to depend on a weak form of the axiom of
choice, but only in the form of a dependence on the Weak König lemma (the König tree principle, that
every tree with infinitely many nodes, and only finitely many nodes at each level, has an infinite branch,
which although a very weak choice principle, is still strong enough to allow the proof of the existence of
non-computable sets, that is sets for which there is no finite procedure for determining whether an object
is a member of it or not5).
Skolem showed the general form of such denumerable sets conforming to Peano’s axioms, but
including non-standard elements in addition to zero and the positive integers. 6 Moreover, such sets are
describable in formulae using the elements forming a subset of a language with denumerable elements,
and these formulae can be explicitly well-ordered, so as to be specifiable without resort to the axiom of
choice7. In effect, the language will include predicates φ(x) and ψ(x) in such a way that one of the two
will appear in the specification of the non-standard set, but such that which occurs is determined by which
of the two occurs first in the well-ordering of the formulae defining the specification.
However, this method of defining a non-standard model conforming to Peano’s axioms without
resort to the axiom of choice, leaves it outside our knowledge which of φ(x) and ψ(x) has been
appropriately added, so that the system remains unconstructive as well as without application. 8
Accordingly, according to the principle I explain in §. 2 below, such systems still lack determinate
linguistic specification, and therefore describe fictions.
This is in a way fortunate inasmuch as such results as Euclid’s proof of the infinity of the prime
number series, Goodstein’s theorem and with it Gentzen’s proof of the consistency of natural arithmetic,
all fail within such models. These models are none of them recursive (inductive), and although having the
cardinality 0‫אּ‬, nonetheless will have an ordinality of ε0, or beyond, exceeding all the ω’s.
We may note that Skolem did not himself believe that these or any other denumerable (countable)
sets actually existed.

A. The question of the validity of the continuum hypothesis


Cantor’s continuum hypothesis is the hypothesis that the transfinite cardinal numbers form a series,
the cardinality of the set of the integers , the cardinality of the set of the subsets of the integers or of the
real numbers, the cardinality of the set of subsets of the real numbers, and so forth. In a non-Platonic
interpretation, this would be a matter, not of a succession of different magnitudes but a succession of
families of sets which cannot be supposed to have one-to-one correspondence with the natural numbers,
and cannot be supposed to have one-to-one correspondence with each other.
The denial of the continuum hypothesis amounts to there being multiple ways of being non-
denumerable intermediate between having the structure of the natural number series, i.e., the structure of
an algebraically complete field, and assigned the cardinality 0‫ אּ‬and having the structure of the real
numbers and assigned the cardinality 20‫אּ‬. Such intermediate cardinalities would identify different
structures, but not different magnitudes. .
Gödel raised this question as to whether there are cardinal numbers intermediate between the
cardinality of the set of the integers , the cardinality of the set of the subsets of the integers or of the real
numbers, that is between 0‫ אּ‬and 20‫אּ‬. Gödel expected the supposition that there are such intermediate

4 Gödel 1929, vol I, Collected works, Oxford University Press, New York, 1986–2003, ed. Solomon Feferman et al., p.
63. This is discussed in “Gödel and the metamathematical tradition,” Jeremy Avigad, in Kurt Gödel: Essays for his Centennial,
ed. Solomon Feferman, Charles Parsons, and Steven G. Simpson (Cambridge: Cambridge University Press, 2010).
5 I owe these and other clarifications in this section chiefly to Wilfrid Hodges, but in part to “Reverse Mathematics” in
Wikipedia.
6 “Über die Nicht-charakterisierbarkeit der Zahlenreihe mittels endlich oder abzählbar unendlich vieler Aussagen mit
ausschliesslich Zahlenvariablen,” Fundam. Math. 23 (1934), pp. 150-161.
7 Cf. Stephen Kleene, Introduction to Metamathematics, North-Holland, Amsterdam 1952, Chapter XIV, §. 72, pp. 389-
98, especially noting pp. 394-98, cf. general discussion in pp. 425-32.
8 In the whole of my remarks on non-standard arithmetic I have drawn on the masterly explanations of Wilfrid Hodges,
but he is not to blame for any mistakes in my abridged summary of parts of these explanations.
4
5

cardinals might be fruitful, and it has been suggested that the suggestion that the cardinality of the
smallest number of nowhere dense subsets which are needed to cover a separable compact topological
space, p, which is said to be provably equal to t, which is the cardinality of something else topological, is
intermediate.9
This has the appearance of not being meaningless, but in an area having nothing to do with the
assessment of magnitude. It would be as Gödel expected that the denial of the continuum hypothesis
would have applications—although, not as he hoped in the sphere of solving Diophantine equations,
unless via topology. (Even with the arithmetic of the natural numbers, much of its interest lies in dealing
with the solution of Diophantine equations.)

II. A set has no existence independently of a predicate indicating its membership: the status of the
axiom of choice

I take the position that in general a set or class has no existence unless there is a law or predicate
constructively defining its membership. This was substantially the view of the “Paris school,” pre-
eminently Lebesgue and Borel.10
That is, I reject an extensional conception of set. If there are many things to which a particular
predicate applies, it does not follow that there is a set comprising all the things to which the predicate
applies. The existence of a predicate does not imply the existence of a set constituted of its extension as
its members.
It is standardly agreed that such predicates as “is a set,” “is an ordinal,” “is a class or set which is not
a member of itself” and such like do not define a set as their extension. These show the difficulty that
there can be in applying the law of the excluded middle or its corollary, the law of bivalence. The same
difficulty arises in such cases as that of the Berry paradox—the suppositions that there is or that there is
not a least number definable in less than a hundred words each lead to a contradiction.
However, there are other cases where a problem arises, e.g., in applying the predicate “is a definition
of a real number,” which has been supposed to lead to the Richard paradox (1905), but leads to paradox
only when a particular style of definition is proposed, viz. the style suggested by Cantor’s Diagonal
Procedure, in which it is suggested that we model our definitions of real numbers on Cantor’s method of
proving that the real numbers can not constitute a set in one-to-one correspondance with the natural
numbers. This method of definition is itself constructive in the sense that we can state what number n
occurs in the nth decimal place sequentially, compatible with treating the resulting decimal as a potential
infinite, the process of determining the occupants of each decimal place decided as far as we want.
However, Cantor’s further argument on the basis of this method of definition to the effect that the
whole set of such a series of infinite decimals allows the definition of a real number which is not a
member of the series, firstly requires that each decimal be regarded as an actually infinite sequence, and
secondly requires that the whole series be considered actually infinite.
The distinction is between considering a sequence as potentially infinite and considering it as actually
infinite, and treating first a sequence and second a set of such sequences as actually infinite totalities.
(If constructions were taken narrowly as algorithms (or more technically, computable functions),
then the reals as constructed here are essentially what classically would be called the computable
numbers.)
If we say that the natural numbers taken in their natural order are potentially infinite in number then
this would be a kind of potential infinity that is relatively clear, and shared by all sets which can be put
into a constructive one-to-one correspondence with the natural numbers. By describing such sets as

9 Proc. National Acad. Sci., USA. Aug 13, 2013; 110(33): 13238–13239.
10 Fraenkel, BarHillel and Levy, Foundations of Set Theory, 2nd ed., p. 216: “(i) According to H. Lebesgue, ‘Sur les
fonctions representables analytiquement’ (J. Math Pures Appliquées 60, 1905, pp. 139-216), 205, an object exists when it has
been defined in finitely many words. Borel in this context speaks of ‘effectively defined’ objects. (ii) Lebesgue rejects the
concept of an arbitrary number sequence, he only accepts those determined by a law.”
5
6

potentially infinite, we mean to exclude any implication that they are totalities containing an actually
infinite number of members.
By contrast, the real numbers and their definitions will be potentially infinite in a different way, a
different way which excludes such constructive one-to-one correspondence with the natural numbers. It
cannot be supposed that the definitions of such numbers constitute a larger set than the set of natural
numbers, because every definition is a finite string of symbols taken from a finite list of symbols, and the
set of all such strings, given an enumerable vocabulary of symbols, can be put into one-to-one
correspondence with the natural numbers. This makes Skolem’s paradox (1922) unsurprising. This is
often stated in the form, “Skolem's paradox is that every countable axiomatization of set theory in first-
order logic, if it is consistent, has a model that is countable.” Skolem himself, as I remarked earlier,
appears to have believed that sets which were incapable of being put into one-to-one correspondence with
the natural numbers for the reason that they were “too many” were a fiction. It is rather that the real
numbers, like their definitions, cannot be well-ordered so as to be taken one by one. Perhaps, when it is
said that “the continua of Weyl, Lebesgue, Lusin, etc. are denumerable, though they cannot be
enumerated within the system,”11 this should be interpreted along these lines, that the constructively
identifiable points within the real line cannot be put into one-to-one correspondence with the natural
numbers, and therefore equivalently cannot be put into one-to-one correspondence with the rational
numbers, but not because they are “too many” but because they cannot be well-ordered and so are not
enumerable.
Since an infinite set such as can be put in one-to-one correspondence with the natural numbers only
exists if there is a constructive (in a way which is not based on the adjustment of some decimal
representation, cf. nn. 1 and 2) law determining its membership, no set can exist whose members would
be decided by infinite number of choices.
In particular, the plausibility of key axioms, of which the simplest is the axiom of choice for
denumerable sets, has depended on adopting a thoroughly extensional view of all classes and relations
according to which the existence of a class or relation (considered as set of ordered pairs of things related
by the relation, albeit that a language-independent conception of ordered pair has not been provided) does
not depend on the existence of linguistically expressible predicates defining the class or relation. I reject
this (cf. nn. 1 and 2).
This suggests that being “non-denumerable in number” should be considered a particular way of
being “indefinite in number.”

III. Note: the significance of the Richard paradox

We have proofs of the existence of a one-to-one correlation between, e.g., the natural numbers and
the even numbers. And, if such sets exist, we have proofs of the existence of a one-to-one correlation
between the real numbers and subsets of the integers.
We also have proofs of the absence of any one-to-one correlation between definitions of the real
numbers and the natural numbers, and also of the absence of any one-to-one correlation between the real
numbers and the natural numbers.
But when we have classes between which there is a one-to-one correlation, this is because they are of
the same magnitude, as we saw in the case of the natural numbers and various subsets of the natural
numbers (see §2 above). Accordingly, when there is no such one-to-one correlation, this will not be
because they are of different magnitude. There is no reason to suppose that there are more real numbers
than there are natural numbers, only that there is no one-to-one correlation between them.12

11 Fraenkel, BarHillel and Levy, Foundations of Set Theory, 2nd ed., pp. 254-55, n. 1.
12 Cf. the Löwenheim–Skolem theorem, stating that if a first-order theory has an infinite model, then it has a countable
model, whose 1920 proof employed the axiom of choice, but which Skolem later, 1922 and 1928, proved using König's
lemma (said to restrict the axiom of choice in a way conducive to constructive approaches) instead, and Skolem's so-called
paradox, whereby, if Zermelo's axioms are consistent, then they must be satisfiable within a countable domain, even though
6
7

Moreover, it is silly to say that the reason why there can be no one-to-one correlation between
definitions of the real numbers and the natural numbers is because there is something queasy about the
notion of definition, whereas the reason why there can be no one-to-one correlation between the real
numbers themselves and the natural numbers is because there are too many real numbers—as if infinity of
any kind is a magnitude, and as if the real numbers existed independently of the possibility of their being
constructively definable.
What we have in the case of inductively ordered infinite sets of an order less than Epsilon 0 is a
constructive finitary approach to identifying their membership, and in other cases the lack of this.
There should be no difficulty in supposing that the existence of a cut (or division between segments)
of a line depends on the possibility of an already existing constructive definition of the location of the cut.
What is essential to continuity is that such a cut should not fall into a hole or gap in a line, but if a
line is considered as a continuum and not as consisting of points or cuts, there will be no such holes or
gaps. Accordingly, there is no reason to suppose that the “class” of cuts constitutes a complete totality—
or in the peculiar terminology of Zermelo-Fraenkel, is a “proper class” as opposed to a “small class.” 13
Unless a class or set14 constitutes a totality, one cannot reason from the falsehood of a generalization
about the members of a proper class ¬(∀x) (¬fx) to (∃x) (¬fx). To do this would be to treat the range of
the quantifiers here, including where actually infinite sets are concerned, as still allowing the application
of the laws of classical logic.

IV. The intuitive basis of the theory of continuity

The intuitive basis of the theory of continuity lies in our perception and grasp of what is involved in
continuous motion, including as its first essential the movement of one and the same physical body from
one place to another, passing through intervening places and continuously existing, not ceasing to exist at
one place and coming to exist at another.
Apart from the phenomena associated with continuous space and continuous motion, we would not
have any use for the mathematics of the continuum. Every application of this mathematics is either to
these or to something measured by means of measuring them, time, velocity, acceleration, mass,
momentum, energy, etc. Newton's remark that “a line is the path of a moving point” strikingly exhibits
the incipient presence of this conception in Newton.

V. Contrasting continua with discrete magnitudes

In approaching the theory of magnitudes, we should treat the fundamental relation as one analogous
to class-inclusion, not to class-membership. Let us call this relation “being a part of.” A point is not part
of a line, although we should say it “marks a cut or division between segments” of the line.

they prove the existence of uncountable sets, not in the sense that there are too many of them to be put in one-to-one
correspondence with the natural numbers, but perhaps because such correspondence is impossible for some other reason, e.g.,
for the sort of reason that the definitions of real numbers cannot be put in such correspondence.
13 The argument in Fraenkel, BarHillel, and Levy (2nd Edition, pp. 266-67) is inadequate. The trichotomy still obtains
between the numbers themselves considered as cuts in the continuum, even though one cannot tell the order of the cuts within
the line because of the discrepancies in the ways the cuts have each been constructively defined.
14 The current division between “sets” and “classes” stems from von Neumann’s development (1928) of “Zermelo-
Fraenkel set theory with the axiom of choice” [ZFC], the system proposed by Fraenkel in 1922, as successively modified in
1925 and, by Skolem, in 1929 (cf. From Frege to Gödel, A Source Book in Mathematical Logic, 1879-1931, ed. J. van
Heijenoort, pp. 284-85, shows how wrong it is to treat ZFC as the fixed form of set theory, as if it already existed in 1908).
For von Neumann, classes cannot be elements of sets, and sets alone constitute totalities, intuitionistic restrictions being
placed only on classes. However this terminological approach provides only a superficial explanation of the paradoxes.
Rather, it is always the predicate used in the attempt to define a set or class which is the source of illegitimacy, and in
special cases of paradox. Some kind of constructivity is required in applications of the predicate, and, in sections 9, 10, and 11
below, I indicate the lengths to which we may go in seeking various forms of constructivity.
7
8

A continuum is something of which every part is a continuum and of which no part has a least
partition. When we refer to one line as a part of another, we are utilizing “...is a part of...” as one of the
primitives of our theory. I note that the positive whole numbers (integers) are used amongst the rational
and other real numbers in measuring lengths of line and marking division points in a line.
By contrast, a discrete magnitude has a least partition. This means that because, if we take, for
instance, the finite collection, {a, b, c, d, e}, we find it has a least partition, viz. {{a}, {b}, {c}, {d}, {e}},
it counts as a discrete magnitude. Whereas sets defined as extensions of predicates can be regarded as
abstract objects, finite collections can be treated in two ways. The first is as sets and so as such abstract
objects. The second is as being as concrete as their members, a pair of shoes being as concrete as a single
shoe, but in this way a collection with no members is a fiction.

VI. Some observations on Euclidean Geometry

A.
In the Euclidean case, we postulate that any two congruent lines which are different in place, as well
as those that overlap or have one including the other, are comparable as if one could be superposed on
part of the other.
Congruency implies either that both lines are straight, or that they are of the same shape, size, and
handedness.
Preliminary (definitional): If two points are joined by lines, then if one of these lines is partitioned
into disjoint adjacent equal sub-lines, in such a way that none of the others can be partitioning it into a
smaller number of sub-lines equal or shorter than the first set of sub-lines, then this is a maximally short
line between these points. (We need to distinguish between two forms of the familiar Euclidean axiom:
we can say between any two points there exists at least one maximally short line, or we can say that
between any two points there exists one and only one maximally short line.)
Things with a kind of magnitude which have only one dimension of variation, e.g., periods of time,
i.e., have only scalar magnitude, can make do with this concept, without the concept of straightness.
Things with a kind of magnitude which can vary in more than one dimension require us to introduce
other concepts, straightness, change of angle at a point, curvature.
We have to distinguish between the measure of a magnitude and the thing with magnitude
(sometimes referred to as the magnitude). The measure of a magnitude will be a number of the kind called
a ratio, i.e., a numerical measure of the proportion or ratio between the magnitude concerned and some
standard magnitude. I would allow for such ratios being in some cases called integers, in others rational
numbers (proportions between integers, what is common to equal fractions), in others real numbers
(including all other proportions between the measures of geometrically definable magnitudes).
Some things with magnitude are uni-dimensional, corresponding to lines over which there is motion,
or to the number of times a motion is repeated, sometimes counted by observation of such motions, and
sometimes by the continuous time over which such repetition would have occurred, e.g., time.

B.
1. Distinguish (4b1) “any maximally short line can be extended some part of its length, the result
being also a maximally short line” (Gauss’s idea) from (4b2) “any maximally short line can be extended
the same length again, the result being also a maximally short line.” Euclid requires (4b2).
2. Why it appeared so obvious to Locke that the proposition that the straight line is the shortest was
not “trifling,” and to Hume that the notions of the “shortest” and the “straight” line, as descriptions of a
line between two points, were not the same.
Intuitive concept: A straight line is undeviating in any direction.
This implies that the notion of straight line makes no sense except in a setting of at least two
dimensions. To say that the line is undeviating makes no sense unless, as it were, we can look along its

8
9

length, comparing it with lines parallel to it without running into them whichever side we look along it.
Unsurprisingly, Euclid and other ancient geometers sometimes used optical ideas and the notion of being
level in order to capture this idea (Εὐθεῖα γραμμή ἐστιν, ἥτις ἐξ ἴσου τοῖς ἐφ᾿ ἑαυτῆς σημείοις κεῖται—a
straight line is (any) one which lies evenly with points on itself). One can also grasp why Posidonius tried
to characterize parallelism in terms of equidistance.
One can also see why the intuitive conceptions of straightness and parallelism naturally appear
inseparable if lines can be produced the same length any number of times.
3. What reasons are there for supposing more than one dimension of space?
a. The organization of the perceptual field, especially (i) touch and all the senses considered
in relation to exploratory movement, experienced as in three dimensions, and (ii) sight considered
as the things as spread before it through at least two dimensions either at a uniform or variable
distance.
b. There is no reason to suppose either space or time to be continuous unless there is such a
thing as continuous movement through space. In Aristotle's Physics, Bk VI, c. 10, it is shown that
the reality of continuous movement through space requires that bodies, and not just imaginarily
idealized mathematical points, pass across dividing points in space and time, one part of the body
passing the division point in space before another one, i.e., a process taking a period of time. A
body has at least three dimensions.
c. I introduced the notion of comparing lines which are set at different places as if they were
superposed—this is implicit in the method of taking the measure of a line by the points of
compasses.
Euclid implicitly compares right angles, as if they were imagined superposed.
We require a theory of constructions and axioms about them in order to allow the introduction of
methods of mensuration, comparing lines as to length.
A theory of construction is already implicit, involving at least imaginary transposition, to deal with
the earlier problem of superposing lines and angles?

VII. Question: by what geometrical constructions can lines of lengths π and e be laid out on a
straight line (I speak of straight lines rather than "uni-dimensional magnitudes," because it is
the way we imagine them in doing mathematics)? How much are the results of these
constructions dependent on supposing a Euclidean space?

I have explained a measure as a ratio between geometrically defined magnitudes (by “ratio,” I do not
mean a rational measure, such as might be defined by a rational number).
It is going to be important to establish an order of greater and less great (a simple ordering) between
these geometrically defined ratios—but there will not be a doubt about the placing of any such
geometrically defined ratios amongst the rational numbers, and there will not be a doubt as to its being
well defined as a ratio as this notion has been explained.
However, there are certain non-geometrical constructions which give a law for determining the
membership of an unending sequence of digits, each between 0 and 9, for any place in the sequence. Such
laws have been thought to identify real numbers thought of as infinite decimals. There is no proof that
every such non-geometrical construction has a geometrical construction corresponding to it, and no
general reason why we should be able to determine its placing as greater or less in an ordering of infinite
decimals corresponding to geometric ratios.
Therefore, if we consider real numbers as being these ratios, i.e., as part of a general theory of the
nature of numbers as being measures within systems of mensuration, all mensuration being either of
continua or of discrete magnitudes, then there is no reason for supposing these non-geometrically defined
infinite sequences of digits to define real numbers at all.
In effect, we are dividing what the intuitionistic mathematicians call species into two kinds, in effect
those definable by geometric law, and as opposed to what they call spreads, which would include any
non-geometrically defined sequences of integer.
9
10

Now, the application of the differential and integral calculus to measures of continua requires the
applicability of the Least Upper Bound Principle only to species definable by geometric law.
Nothing is required in respect of the non-geometric spreads as such. The latter are of interest to set
theory, but not particularly to the algebraic theory of fields considered as ranging over cuts in the
continuum of continuous motion and change as has been explained above.

VIII. What we need of an indefinitely large number of kinds of construction or procedure properly
described as geometric, but which are not amongst Euclid's methods of construction

For instance, consider the problems of integration and the means of constructing lengths on a line
precisely corresponding to the logarithms of particular numbers: now, given an ideal liquid whose amount
in no way changes in the course of putting it into vessels of different shape, then one would have an
automatic way of identifying a line of length corresponding to the logarithm of a certain number set to a
certain base. Again, considering the problems of differentiation, in effect one wishes to establish a
construction which will assign a length corresponding to the gradient of a curve, taken at any assigned
point, what we need is an ideal plane piece of wood, and then measure the angle of the slope more simply,
by assessment of the value of its tangent (by preference to its sine). This will enable us to assign values to
the exponential function. However, even if we get over the problems of identifying ideal plane pieces of
wood, things which have no existence except within Euclidean space, we would still have the problem of
the theory of ideal liquids. We would seem to have to construct laws in terms of the theory of
differentiation and integration in order to define what it was for a liquid to be in the relevant sense ideal. 15
According to my suggestion here, the validity of the differential and integral calculus, inseparable
from the L.U.B., is correlative with the ideas of ideal solids and liquids. Ideal solids and liquids allow
even π and e, together with logarithms and any constructively definable ordinals, to define areas and
volumes.

IX. Strange dependence upon Euclidean and Cartesian ideas in utilizing the imagination in
developing extensions and variants to these ideas in forming models for modern physics

I would note that Euclid's ruler and compass constructions depended on his freedom from the ideas
of Special Relativity, viz. change of appropriate judgement of length between compass points according
to direction, depending on relative velocity of the observer.
Likewise the background idea of undeviatingness, involving the inseparability of the notions of
straightness and the parallels' postulate is indispensable for Descartes' arithmetical representation of
Euclid's ideas in his Analytic Geometry. In this way, the correlation of certain parts of Cartesian
mathematics with certain physical idealizations seems to reach to an even deeper level, viz. to certain
presumptions about how light proceeds when not interfered with, e.g., as in reflection and refraction.
Strangely, we continue to draw upon this Cartesian mathematics, its extensions and variants, even
when the physical suppositions on which it was founded have turned out to be inapplicable to the real
world. However, as the intellectually structured imagination was the sole basis upon which Cartesian
mathematics relied, so the intellectually structured imagination will be and is the sole available basis upon
which extensions or variants of this mathematics can be developed, normally within reach of some
imagined field of possible application. Constructive proof of existence in cases where its denial is absurd
will commonly rest on a combination of considerations of elegance and of fruitfulness within some
empirical field of application. The principle ab esse ad posse has priority. If a theory or model is fruitful
in application, as, for instance, the introduction of the idea of quantum levels and the quantization of

15 Poincare, Science and Hypothesis, chapter XI, pp. 193-206, has relevant thoughts. Measure theory provides the
foundation for the modern notion of integral, the Lebesgue integral. It is said that it is possible to rework measure theory on
the basis of the computable real line, where the set-theoretic basis for measurability is replaced by notions from order theory.
This constructive measure theory provides the basis for computable analogues for Lebesgue integration.
10
11

energy together with the combining of particle models with wave models has proved so fruitful despite its
involving indeterminacy, then it promises to represent a further move towards the truth—even though
there are more difficulties and disputes as to its interpretation than arise with relativity theory.
I was led to these remarks by the discussion in Paul Davies, The Goldilocks Enigma, pp. 267-273, of
the limits to the applicability of mathematics shown up by discussions of firstly quantum theory and
quantum mechanics in general and secondly as applied in quantum cosmology. Limits to the applicability
of classical mathematics are limits to our knowledge of the early universe.
These limits render theories of a cyclical universe and of multiverses scientifically baseless on the
Popperian ground of unfalsifiability. Penrose himself has given reason for supposing that gravity as well
as complexity limit to the diversification or scatter introduced by quantum theory; 16 these will be limits to
be set theoretically according to considerations partly by theoretical elegance, but confirmed by lack of
evidence of their absence.
Platonism is okay with the natural numbers, but not with the continuum. 17 Kronecker said, “God
made the integers, and all the rest is the work of man” (Die ganzen Zahlen hat der liebe Gott gemacht,
alles andere ist Menschenwerk). And the logician can see reason for this in the capacity of natural number
theory to be applied in just any field, abstract or concrete, counting people, angels, colors, musical notes,
numbers, and sets.
By contrast, the mathematics of the continuum has application only to the concrete related fields
within which the rational and real numbers come into play. Accordingly, we should rephrase Kronecker
and say, “God gave the integers in giving any capacity of thought in respect of anything concrete or
abstract, but all the rest is the work of man in finding ways to measure and understand the concrete
world.” In respect of the latter, the best reasons for thinking propositions meaningful are whatever reasons
we have for thinking them true, or whatever puts us in the position of certain knowledge that they are
true, and then we can reason that what is true must have meaning, and in this way more generally reason
ab esse ad posse.

16 e.g., The Road to Reality, 859-60, and the chapter following 816Cf. Quantum Concepts in Space and Time, eds. R.
Penrose and C.J. Isham, OUP, 1986, especially Chapters 8 and 9, and Shadows of the Mind, R. Penrose, OUP, 1994, pp. 335-
347.
17 The different roots of the mathematics of natural arithmetic and finite sets and of the mathematics of the continuum
was part of the reason for Imre Lakatos’ guardedness in approaching mathematics. Lakatos’ position and my own would not go
back to the view expressed by Lee Smolin who seems to follow John Stuart Mill in saying “Mathematics is derived from
experience as a generalisation of observed regularities, when time and particularity are removed” in an article “The Unique
Universe” in Physics World, June 2009, published by the Institute of Physics. One needs to distance oneself from mathematical
platonism, and reject any Platonic conclusions, but observe the "The Unreasonable Effectiveness of Mathematics in the Natural
Sciences” discussed by Eugene Wigner, and avoid the extremes to be found in John Stuart Mill.
11

You might also like