History of Constructivism in The 20 Century
History of Constructivism in The 20 Century
History of Constructivism in The 20 Century
A.S.Troelstra
1 Introduction
In this survey of the history of constructivism, more space has been devoted to
early developments (up till ca 1965) than to the work of the last few decades. Not
only because most of the concepts and general insights have emerged before 1965,
but also for practical reasons: much of the recent work is of a too technical and
complicated nature to be described adequately within the limits of this article.
Constructivism is a point of view (or an attitude) concerning the methods and
objects of mathematics which is normative: not only does it interpret existing math-
ematics according to certain principles, but it also rejects methods and results not
conforming to such principles as unfounded or speculative (the rejection is not always
absolute, but sometimes only a matter of degree: a decided preference for construc-
tive concepts and methods). In this sense the various forms of constructivism are
all `ideological' in character.
Constructivism as a specic viewpoint emerges in the nal quarter of the 19th
century, and may be regarded as a reaction to the rapidly increasing use of highly
abstract concepts and methods of proof in mathematics, a trend exemplied by the
works of R. Dedekind1 and G. Cantor2 .
The mathematics before the last quarter of the 19th century is, from the view-
point of today, in the main constructive, with the notable exception of geometry,
where proof by contradiction was commonly accepted and widely employed.
Characteristic for the constructivist trend is the insistence that mathematical
objects are to be constructed (mental constructions) or computed
thus theorems
asserting the existence of certain objects should by their proofs give us the means
of constructing objects whose existence is being asserted.
L. Kronecker3 may be described as the rst conscious constructivist. For Kro-
necker, only the natural numbers were `God-given'
all other mathematical objects
ought to be explained in terms of natural numbers (at least in algebra). Assertions
of existence should be backed up by constructions, and the properties of numbers
1 (Julius Wilhelm) Richard Dedekind 1831{1916
2 Georg Ferdinand Ludwig Philipp Cantor 1845{1918
3 Leopold Kronecker 1823{1891
1
History of constructivism in the 20th century 2
2 Finitism
2.1. Finitist mathematics
Finitism may be characterized as based on the concept of natural number (or nite,
concretely representable structure), which is taken to entail the acceptance of proof
by induction and denition by recursion.
Abstract notions, such as `constructive proof', `arbitrary number-theoretic func-
tion' are rejected. Statements involving quantiers are nitistically interpreted in
terms of quantier-free statements. Thus an existential statement 9x Ax is regarded
as a partial communication, to be supplemented by providing an x which satises
A. Establishing :8 xAx nitistically means: providing a particular x such that Ax
is false.
In this century, T. Skolem4 was the rst to contribute substantially to nitist
4 Thoralf Skolem 1887{1963
History of constructivism in the 20th century 3
of the same kind, whether we talk about very small numbers such9 as 3 or 5, or
extremely large ones such as 9 . In reality, we cannot handle 99 without some
9 9
(2) According to Poincare, the set-theoretic paradoxes are due to a vicious circle,
namely the admission of impredicative denitions: dening an object N by referring
to a totality E (in particular by quantiying over E ), while N itself belongs to E .
Poincare's standard example is J.Richard's14 paradox (Richard 1905): let E
be the totality of all real numbers given by innite decimal fractions, denable in
nitely many words. E is clearly countable, so by a wellknown Cantor style diagonal
argument we can dene a real N not in E . But N has been dened in nitely
many words! However, Poincare, adopting Richard's conclusion, points out that the
denition of N as element of E refers to E itself and is therefore impredicative. For
a detailed discussion of Poincare's philosophy of mathematics, see (Mooij 1966).
3.2. The semi-intuitionists
The term `semi-intuitionists' or `empirists' refers to a group of French mathemati-
cians, in particular E. Borel15 , H. Lebesgue16 , R. Baire17, and the Russian math-
ematician N.N.Luzin18. Their discussions of foundational problems are always in
direct connection with specic mathematical developments, and thus have an 'ad
hoc', local, character
also the views within the group dier.
What the semi-intuitionists have in common is the idea that, even if mathemati-
cal objects exist independently of the human mind, mathematics can only deal with
such objects if we can also attain them by mentally constructing them, i.e. have
access to them by our intuition
in practice, this means that they should be explic-
itly denable. In addition, pragmatical considerations occur: one is not interested
in arbitrary objects of a certain kind, but only in the ones which play an important
role in mathematics (which `occur in nature' in a manner of speaking).
We shall illustrate semi-intuitionism by a summary of the views of Borel, the
most explicit and outspoken of the semi-intuitionists. According to Borel, one can
assert the existence of an object only if one can dene it unambiguously in nitely
many words. In this concern with denability there is a link with predicativism (as
described below in 3.4)
on the other hand, there is in Borel's writings no explicit
concern with impredicativity and the vicious circle principle, as in the writings of
Poincare and H.Weyl19 . (The Borel sets, introduced by Borel in the development of
measure theory, are a `continuous' analogue of the hyperarithmetic sets, which may
be regarded as a rst approximation to the general notion of a predicative subset of
the natural numbers, but this was not known to Borel.)
With respect to the reality of the countably innite, Borel takes a somewhat
pragmatical attitude: while conceding the strength of the position of the strict
14 Jules Richard
15 (Felix E douard) E mile Borel 1871{1970
16 Henri (-L eon) Lebesgue 1875{1941
17 (Louis) Ren e Baire
18 Nikolai Nikolaevich Luzin (Nicolas Lusin) 1893-1950
19 Hermann Weyl 1885{1955
History of constructivism in the 20th century 6
nitist, he observes that the countably innite plays a very essential role in math-
ematics, and mathematicians have in practice always agreed on the correct use of
the notion:
` The notion of the countably innite appears therefore : : : ] as a limit
possibility conceived by our imagination, just like the ideal line or the
perfect circle.' (Borel 1914, p.179)
In other words, the natural numbers and the principle of complete induction are
intuitively clear.
Borel explains the Richard paradox, not by referring to the impredicative charac-
ter of the denition of N , but by the observation that N has not been unambiguously
dened
the collection E is countable, but not eectively enumerable and hence the
construction of N cannot be carried out (Borel 1914, p.165). This distinction would
later obtain a precise formulation in recursion theory.
Borel explicitly introduced the notion of a calculable real number
a function f is
calculable if f (x) is calculable for each calculable x, and he observes that a calculable
function must necessarily be continuous, at least for calculable arguments (Borel
1914, p.223). This foreshadows Brouwer's wellknown theorem on the continuity of
all real functions (cf. section 6.1).
3.3. Borel and the continuum
Borel rejected the general notion of an uncountable set, as an essentially nega-
tive notion. According to Borel, Cantor's diagonal argument only shows that the
continuum cannot be exhausted by a countable set. Obviously, on this view the
continuum presents a problem. On the one hand it is the basic concept in analysis,
on the other hand it cannot be described as `the set of its (denable) elements' since
it is uncountable. Thus Borel remarks in 1908:
` : : : ] the notion of the continuum, which is the only well-known exam-
ple of a non-denumerable set, that is to say a set of which the mathe-
maticians have a clear idea in common (or believe to have in common,
which in practice amounts to the same thing). I regard that notion as
acquired from the geometric intuition of the continuum
it is well-known
that the complete arithmetical notion of the continuum requires that
one admits the legitimacy of an innity of countably many successive
arbitrary choices'. (Borel 1914, p.162)
(The term `arithmetic theory of the continuum', or `arithmetization of the con-
tinuum', frequently appears in discussions of constructivism in the early part of
this century. By this is meant the characterization of the continuum as a set of
reals, where the real numbers are obtained as equivalence classes of fundamental
sequences of rationals, or as Dedekind cuts. Since the rationals can be enumerated,
History of constructivism in the 20th century 7
4 Brouwerian intuitionism
4.1. Early period
In his thesis `Over de Grondslagen der Wiskunde' (Brouwer 1907) the Dutch mathe-
matician L.E.J. Brouwer27 defended, more radically and more consistently than the
semi-intuitionists, an intuitionist conception of mathematics. Brouwer's philosophy
of mathematics is embedded in a general philosophy, the essentials of which are
found already in (Brouwer 1905). To these philosophical views Brouwer adhered
all his life
a late statement may be found in (Brouwer 1949). With respect to
mathematics, Brouwer's main ideas are
1. Mathematics is not formal
the objects of mathematics are mental construc-
tions in the mind of the (ideal) mathematician. Only the thought constructions
of the (idealized) mathematician are exact.
2. Mathematics is independent of experience in the outside world, and mathemat-
ics is in principle also independent of language. Communication by language
may serve to suggest similar thought constructions to others, but there is no
guarantee that these other constructions are the same. (This is a solipsistic
element in Brouwer's philosophy.)
21 Solomon Feferman
22 Motokiti Kond^o
23 Andrzej Grzegorczyk
24 Georg Kreisel, born 1926
25 Kurt Schutte
26 Paul Lorenzen
27 Luitzen Egbertus Jan Brouwer 1881{1966
History of constructivism in the 20th century 9
3. Mathematics does not depend on logic
on the contrary, logic is part of math-
ematics.
These principles led to a programme of reconstruction of mathematics on intuition-
istic principles (`Brouwer's programme' or `BP' for short). During the early period,
from say 1907 until 1913, Brouwer did the major part of his work in (classical) topol-
ogy and contributed little to BP. In these years his view of the continuum and of
countable sets is quite similar to Borel's position on these matters. Thus he writes
` The continuum as a whole was intuitively given to us
a construction
of the continuum, an act which would create \all" its parts as individu-
alized by the mathematical intuition is unthinkable and impossible. The
mathematical intuition is not capable of creating other than countable
quantities in an individualized way.' (Brouwer 1907, p.62, cf. also p.10)
On the other hand, already in this early period there are also clear dierences
thus Brouwer did not follow Borel in his pragmatic intersubjectivism, and tries to
explain the natural numbers and the continuum as two aspects of a single intuition
(`the primeval intuition').
Another important dierence with Borel c.s. is, that Brouwer soon after nishing
his thesis realized that classical logic did not apply to his mathematics (see the next
section). Nevertheless, until circa 1913 Brouwer some of the views of the semi-
intuitionists and did not publicly dissociate himself from them.
4.2. Weak counterexamples and the creative subject
Already in (Brouwer 1908) a typically intuitionistic kind of counterexample to cer-
tain statements A of classical mathematics was introduced, not counterexamples
in the strict sense of deriving a contradiction from the statement A, but examples
showing that, assuming that we can prove A intuitionistically, it would follow that
we had a solution to a problem known to be as yet unsolved. Undue attention
given to these examples often created for outsiders the erroneous impression that
intuitionism was mainly a negative activity of criticizing classical mathematics.
Example. Consider the set X fx : x = 1 _ (x = 2 ^ F )g where F is any as yet undecided
mathematical statement, such as the Riemann hypothesis. X is a subset of the nite set f1 2g,
but we cannot prove X to be nite, since this would require us to decide whether X has only one,
or two elements, and for this we would have to decide F . (Intuitionistically, a set is nite if it can
be brought into a constructively specied 1-1 correspondence with an initial part of the natural
numbers.) So we have found a weak counterexample to the statement `a subset of a nite set is
nite'.
By choosing our undecided problems suitably, it is also possible to give a weak
counterexample to: `for reals x, x < 0 or x = 0 or x > 0', or to `for all reals x,
x
0 or x 0'. Brouwer used these examples to show the need for an intuitionistic
History of constructivism in the 20th century 10
revision of certain parts of the classical theory, and to demonstrate how classically
equivalent denitions corresponded to distinct intuitionistic notions.
G.F.C. Griss28, in a number of publications between 1944 and 1955, advocated a
form of intuitionistic mathematics without negation, since one cannot have a clear
conception of what it means to give proof of a proposition which turns out to be
contradictory (cf. the interpretation of intuitionistic implication and negation in
subsection 5.2).
In reaction to Griss, Brouwer published in 1948 an example of a statement in the
form of a negation which could not be replaced by a positive statement. Brouwer's
example involved a real number dened by explicit reference to stages in the ac-
tivity of the ideal mathematician trying to prove an as yet undecided statement A.
An essential ingredient of Brouwer's argument is that if the ideal mathematician
(creative subject in Brouwer's terminology) is certain that at no future stage of his
mathematical activity he will nd A to be true, it means that he knows that :A
(Brouwer seems to have used such examples in lectures since 1927).
The argument illustrates the extreme solipsistic consequences of Brouwer's intu-
itionism. Because of their philosophical impact, these examples generated a good
deal of discussion and inspired some metamathematical research, but their impact
on BP was very limited, in fact almost nihil.
4.3. Brouwer's programme
After 1912, the year in which he obtained a professorship at the University of Am-
sterdam, Brouwer started in earnest on his programme, and soon discovered that
for a fruitful reconstruction of analysis his ideas on the continuum needed revision.
Around 1913 he must have realized that the notion of choice sequence (appearing in
a rather dierent setting in Borel's discussion of the axiom of choice) could be legit-
imized from his viewpoint and oered all the advantages of an `arithmetical' theory
of the continuum. The rst paper where the notion is actually used is (Brouwer
1919).
In the period up to 1928 he reconstructed parts of the theory of pointsets, some
function theory, developed a theory of countable well-orderings and, together with
his student B. de Loor29, gave an intuitionistic proof of the fundamental theorem of
algebra.
Brouwer's ideas became more widely known only after 1920, when he lectured
on them at many places, especially in Germany.
After 1928, Brouwer displayed very little mathematical activity, presumably as
a result of a conict in the board of editors of the `Mathematische Annalen' (cf.
Van Dalen 1990). From 1923 onwards, M.J. Belinfante30 and A. Heyting31, and
28 George Francois Cornelis Griss
29 Barend de Loor
30 Maurits Joost Belinfante
31 Arend Heyting 1898{1980
History of constructivism in the 20th century 11
G!odel's embedding made it clear that intuitionistic methods went beyond nitism,
precisely because abstract notions were allowed. This is clear e.g. from the clause
explaining intuitionistic implication in the BHK-interpretation, since there the ab-
stract notion of constructive proof and construction are used as primitives. This
fact had not been realized by the Hilbert school until then
Bernays was the rst
one to grasp the implications of G!odel's result.
Quite important for the proof theory of intuitionistic logic was the formulation in
(Gentzen 1935) of the sequent calculi LK and LJ. Using his cut elimination theorem,
Gentzen showed that for IQC the disjunction property DP holds: if ` A _ B , then
` A or ` B . Exactly the same method yields the explicit denability or existence
property ED: if ` 9x A(x) then ` A(t) for some term t. These properties present
a striking contrast with classical logic, and have been extensively investigated and
established for all the usual intuitionistic formal systems.
The earliest semantics for IPC, due to the work of S.Jaskowski41, M.H. Stone42,
A.Tarski43, G.Birkho44, T. Ogasawara45in the years 1936-1940 was algebraic se-
mantics, with topological semantics as an important special case. In algebraic se-
mantics, the truth values for propositions are elements of a Heyting algebra (also
known as Brouwerian lattice, pseudo-complemented lattice, pseudo-Boolean alge-
bra, or residuated lattice with bottom). A Heyting algebra is a lattice with top
and bottom, and an extra operation ! such that a ^ b
c i a
b ! c, for all
elements a b c of the lattice. An important special case of a Heyting algebra is
the collection of open sets of a topological space T ordered under inclusion, where
U ! V := Interior(V (T n U )). The logical operations ^ _ ! : correspond to
the lattice operations ^ _ ! and the dened operation :a := a ! 0, where 0 is the
bottom of the lattice. (A boolean algebra is a special cases of a Heyting algebra.)
5.4. Metamathematics of intuitionistic logic and arithmetic after 1940
In the early forties Kleene devised an interpretation which established a connection
between the notion of computable (= recursive) function and intuitionistic logic, the
realizability interpretation (Kleene 1945).
The essence of the interpretation is that it so to speak hereditarily codes informa-
tion on the explicit realization of disjunctions and existential quantiers, recursively
in numerical parameters. The denition is by induction on the number of logical
symbols of sentences (= formulas without free variables): with every formula A one
associates a predicate `x realizes A', where x is a natural number. Typical clauses
are
1. n realizes t = s i t = s is true.
41 Stanislaw Jaskowski
42 M.H.Stone?
43 Alfred Tarski
44 Garrett Birkho
45 T^
ozir^o Ogasawara
History of constructivism in the 20th century 15
is intuitionistically valid if for all domains D and all relations Ri of the appropri-
ate arity (i.e. the appropriate number of argument places), AD (R1 : : : Rn) holds
to D, and replacing Ri by Ri .
The analogy goes deeper: one can use terms of a typed lambda calculus to denote
natural deduction proofs, and then normalization of proofs corresponds to normal-
ization in the lambda calculus. So pure typed lambda calculus is in a sense the same
as IPC in natural deduction formulation
similarly, second-order lambda calculus
(polymorphic lambda calculus) is intuitionistic logic with propositional quantiers.
The formulas-as-types idea is a guiding principle in much recent research in type
theory on the borderline of logic and theoretical computer science. It is used in the
type theories developed by P.Martin-L!of55 (Martin-L!of 1975, 1984) to reduce logic
to type theory
thus proof by induction and denition by recursion are subsumed
under a single rule in these theories. Formulas-as-types plays a key role in the proof-
checking language AUTOMATH devised by de Bruijn and collaborators since the
late sixties.
The formulas-as-types idea has a parallel in category theory, where propositions
correspond to objects, and arrows to (equivalence classes of) proofs. J. Lambek56
investigated this parallel for IPC in (Lambek 1972) and later work, culminating in
the monograph (Lambek & Scott 1986).
nite branches, corresponding to the trivial L satisfying L((n + 1)) = n. Since it
is not the denition of a spread, but the way the choice sequences are given to us,
which determines the properties of the continuum, we shall henceforth concentrate
on the choice sequences.
The notion of spread is supplemented by the notion of species, much closer to
the classical concept of set
one may think of a species as a set of elements singled
out from a previously constructed totality by a property (as in the separation axiom
of classical set theory).
The admissibility of impredicative denitions is not explicitly discussed in Brouwer's
writings, though it is unlikely that he would have accepted impredicative denitions
without restrictions. On the other hand, his methods allow more than just predica-
tive sets over N: Brouwer's introduction of ordinals in intuitionistic mathematics is
an example of a set introduced by a so-called generalized inductive denition, which
cannot be obtained as a set dened predicatively relative to N.
A choice sequence of natural numbers may be viewed as an unnished, ongoing
process of choosing values 0, 1, 2, : : : by the ideal mathematician (IM)
at any
stage of his activity the IM has determined only nitely many values plus, possibly,
some restrictions on future choices (the restrictions may vary from `no restrictions'
to `choices henceforth completely determined by a law'). For sequences completely
determined by a law or recipe we shall use lawlike
other mathematical objects not
depending on choice parameters are also called lawlike. An important principle
concerning choice sequences is the
Continuity principle or continuity axiom. If to every choice sequence of
a spread a number n() is assigned, n() depends on an initial segment
m = 0 1 : : : (m ; 1) only, that is to say for all choice sequences
starting with the same initial segment m, n( ) = n().
This principle is not specially singled out by Brouwer, but used in proofs (for
the rst time in course notes from 1916-17), more particularly in proofs of what
later became known as the bar theorem. From the bar theorem (explained below)
Brouwer obtained an important corollary for the nitary spreads, the
Fan theorem. If to every choice sequence of a nitely branching spread
(fan) a number n() is assigned, there is a number m, such that for all
, n() may be determined from the rst m values of (i.e. an initial
segment of length m).
The fan theorem may be seen as a combination of the compactness of nite trees
with the continuity axiom
Brouwer uses the fan theorem in particular to derive
The uniform continuity theorem. Every function from a bounded closed
interval into R is uniformly continuous.
The essence of the bar theorem is best expressed in a formulation which appears
in a footnote in (Brouwer 1927), and was afterwards used by Kleene as an axiom in
his formalization of intuitionistic analysis.
History of constructivism in the 20th century 19
Bar theorem.If the `universal spread' (i.e. the tree of all sequences of
natural numbers) contains a decidable set A of nodes such that each
choice sequence has an initial segment in A, then for the set of nodes
generated by (i) X A, (ii) if all successors of a node n are in X , then
n 2 X , it follows that the empty sequence is in X .
Originally, in (Brouwer 1919), the fan theorem is assumed without proof. In 1923
Brouwer presents an unsatisfactory proof of the uniform continuity theorem, in 1924
he proves this theorem via the fan theorem which in its turn is obtained from the bar
theorem
his 1924 proof of the bar theorem is repeated in many later publications
with slight variations.
Brouwer's proof of the bar theorem has often been regarded as obscure, but has
also been acclaimed as containing an idea of considerable interest. For in this proof
Brouwer analyzes the possible forms of a constructive proof of a statement of the
form 89nA(n) for decidable A.
More precisely, the claim made by Brouwer, in his proof of the bar theorem, amounts to the
following. Let Sec(n) (`n is secured') mean `All through the node n pass through A', Now
Brouwer assumes that a `fully analyzed proof' of the statement h i 2 Sec or Sec(h i) (i.e. all pass
through A), can consist of three kinds of steps only:
(i) n 2 A, hence Sec(n)
(ii) if for all i Sec(n hii), then Sec(n)
(iii) if Sec(n), then Sec(n hii).
Brouwer then shows that steps of the form (iii) may be eliminated, from which one readily obtains
the bar theorem in the form given above.
Nowadays the proof is regarded not so much obscure as well as unsatisfactory.
As Kleene (Kleene & Vesley 1965, p.51 ) rightly observes, Brouwer's assumption
concerning the form of `fully analyzed proofs' is not more evident than the bar
theorem itself. On the other hand, the notion of `fully analyzed' proof or `canonical'
proof was used later by M.A.E.Dummett57 (Dummett 1977) in his attempts to give
a more satisfactory version of the BHK-interpretation of intuitionistic logic.
6.2. Axiomatization of intuitionistic analysis
In Heyting's third formalization paper from 1930 we nd for the rst time a formal
statement of the continuity principle. For a long time nothing happened till Kleene
in 1950 started working on the axiomatization of intuitionistic analysis
his work
culminated in the monograph (Kleene & Vesley 1965). Kleene based his system on
a language with variables for numbers and choice sequences
to arithmetical axioms
he added an axiom of countable choice, the axiom of bar induction (equivalent to
the bar theorem as formulated above) and a continuity axiom, a strengthening of
the continuity principle as stated above (`Brouwer's principle for functions').
The continuity axiom was the only non-classical principle, and Kleene estab-
lished its consistency relative to the other axioms using a realizability interpretation
57 Michael A.E. Dummett
History of constructivism in the 20th century 20
The publication of (Bishop 1967) led to several proposals for an axiomatic frame-
work for Bishop's constructive mathematics, in particular a type-free theory of op-
erators and classes (Feferman 1975, 1979), and versions of intuitionistic set theory
(Friedman 1977). Martin-L!of's type theories have also been considered in this con-
nection (Martin-L!of 1975, 1984). As shown by Aczel in 1978, one can interpret
a constructive set theory CZF in a suitable version of Martin-L!of's type theory.
(CZF does not have a powerset axiom
instead there are suitable collection axioms
which permit to derive the existence of the set of all functions from x to y for any
two sets x and y
moreover, the foundation axiom has been replaced by an axiom
of 2-induction). Much of this work is reported in the monograph (Beeson 1985)
Beeson60 himself made substantial contributions in this area.
6.3. The model theory of intuitionistic analysis
Topological models and Beth models turned out to be very fruitful for the metamath-
ematical study of intuitionistic analysis, type theory and set theory. D.S. Scott61
was the rst to give a topological model for intuitionistic analysis, in two papers
from 1968 and 1970. In this model the real numbers are represented by the contin-
uous functions over the topological space T underlying the model. For suitable T ,
all real-valued functions are continuous in Scott's model.
This later developed into the so-called sheaf models for intuitionistic analysis,
type theory and set theory, with a peak of activity in the period 1977-1984. The
inspiration for this development not only came from Scott's models just mentioned,
but also from category theory, where W.Lawvere62(Lawvere 1971) developed the
notion of elementary topos | a category with extra structure, in which set theory
and type theory based on intuitionistic logic can be interpreted. The notion of
elementary topos generalized the notion of Grothendieck topos known from algebraic
geometry, and is in a sense `equivalent' to the notion of an intuitionistic type theory.
Even Kleene's realizability interpretation can be extended to type theory and be
recast as an interpretation of type theory in a special topos (Hyland 1982).
Some of the models studied are mathematically interesting in their own right,
and draw attention to possibilities not envisaged in the constructivist tradition (e.g.
analysis without an axiom of countable choice, where the reals dened by Dedekind
cuts are not isomorphic to the reals dened via fundamental sequences of rationals).
60 Michael J. Beeson
61 Dana Stewart Scott, born 1932
62 William Lawvere
History of constructivism in the 20th century 22
f0 ;
;; (3 1)
;
(1 0) ;;
;; (2 0)
;
;
;;
(0 ;1)
The function fa(x) given by fa (x) = f0(x) + a, where f0 is as in the picture, cannot
be constructively proved to have a zero as long as we do not know whether a
0
or a 0. But even classically we can show that fa does not have a zero recursively
in the parameter a - and this is typically a result of RM.
Where in constructive recursive mathematics the recursivity in parameters is
built into the constructive reading of the logical operators, in RM the recursiveness
has to be made explicit.
RM is almost as old as recursion theory itself, since already Turing introduced
\computable numbers" (Turing 1937), and it is still expanding today. We mention
a few examples.
E.Specker63 was the rst to construct an example of a recursive bounded mono-
tone sequence of rationals without a recursive limit (Specker 1949). Such sequences
are now known as Specker sequences.
Kreisel and D.Lacombe64 constructed singular coverings of the interval in 1957
Kreisel, Lacombe and J.Shoeneld65 showed, also in 1957, that every eective oper-
ation of type 2 is continuous (an eective operation of type 2 is a partial recursive
operation with code u say, acting on codes x y of total recursive functions such that
8z (x z = y z ) ! u x = u y
continuity of u means that u x depends on nitely
many values only of the function coded by x).
63 Ernst Specker
64 Daniel Lacombe
65 Joseph Shoeneld
History of constructivism in the 20th century 23
formula 9x1 : : : xi A where A is normal, i.e. does not contain _ 9 and the string
of existential quantiers may be interpreted in the usual way. Shanin's method is
essentially equivalent to Kleene's realizability, but has been formulated in such a
way that normal formulas are unchanged by the interpretation (Kleene's realizabil-
ity produces a dierent although intuitionistically equivalent formula when applied
to a normal formula).
The deciphering method systematically produces a constructive reading for no-
tions dened in the language of arithmetic
classically equivalent denitions may
obtain distinct interpretations by this method. But not all notions considered in
CRM are obtained by applying the method to a denition in arithmetic (example:
the `FR-numbers' are the reals corresponding to the intuitionistic reals and are given
by a code for a fundamental sequence of rationals, together with a code for a
modulus of convergence, i.e. relative to a standard enumeration of the rationals
hrnin the sequence is hr n in, and 8mm (jrn+m ; rn+m j < 2 k ). An F-sequence is
0
0
;
an FR-sequence with the omitted. The notion of an F-sequence does not arise as
an application of the deciphering algorithm.)
Markov accepted one principle not accepted in either intuitionism or Bishop's
constructivism: if it is impossible that computation by an algorithm does not ter-
minate, then it does terminate. Logically this amounts to what is usually called
Markov's principle:
MP ::9x(fx = 0) ! 9x(fx = 0) (f : N ! N recursive:)
The theorem by Tsejtin, mentioned above, needs MP for its proof.
The measure-theoretic paradox (cf. 3.3) is resolved in CRM in a satisfactory
way: singular coverings of 0 1] do exist, but the sequence of partial sums Pkn=0 jInj
does not converge, but is a Specker sequence
and if the sequence does converge, the
limit is 1, as shown by Tsejtin and I.D.Zaslavski&i75 in 1962.
In recent years (after ca. 1985) the number of contributions to CRM has consid-
erably decreased. Many researchers in CRM have turned to more computer-science
oriented topics.
8 Bishop's constructivism
8.1. Bishop's constructive mathematics
In his book `Foundations of constructive mathematics' the American mathemati-
cian E. Bishop76 launched his programme for constructive mathematics. Bishop's
attitude is both ideological and pragmatical: ideological, inasmuch he insists that
we should strive for a type of mathematics in which every statement has empirical
content, and pragmatical in the actual road he takes towards his goal.
75 Igor' Dmitrievich Zaslavskii, born 1932
76 Errett Bishop 1928-1983
History of constructivism in the 20th century 25
'' $$
denition of algorithm.
Thus BCM appears as a part of classical mathematics, and the situation may be
illustrated graphically in the diagram below, where `INT' stands for intuitionistic
' $ %
mathematics, and `CLASS' for classical mathematics.
&&
CLASS
% %
BCM
However, it should not be forgotten that this picture is not to be taken at face
value, since the mathematical statements have dierent interpretations in the various
forms of constructivism.
In the actual implementation of his programme, Bishop not only applied the
three principles above, but also avoided negative results (only rarely did he present
a weak counterexample), and concentrated almost exclusively on positive results.
History of constructivism in the 20th century 26
9 Concluding remarks
In the foundational debate in the early part of this century, constructivism played
an important role. Nevertheless, at any time only a handful of mathematicians have
been actively contributing to constructive mathematics (in the sense discussed here).
In the course of time, the focus of activity in constructive mathematics has shifted
from intuitionism to Markov's constructivism, then to Bishop's constructivism. In
addition there has been a steady ow of contributions to classical recursive mathe-
matics, a subject which is still ourishing.
Much more research has been devoted to intuitionistic logic and the metamathe-
matics of constructive systems. In this area the work has recently somewhat ebbed,
but its concepts and techniques play a signicant role elsewhere, e.g. in theoret-
ical computer science and articial intelligence, and its potential is by no means
exhausted (example: the notion of `formulas-as-types' and Martin-L!of-style type
theories).
New areas of application lead to renement and modication of concepts devel-
oped in another context. Thus a recent development such as `linear logic' (Girard
1987) may be seen as a renement of intuitionistic logic, obtained by pursuing the
idea of a `bookkeeping of resources' seriously.
References
A very extensive bibliography of constructivism is (M!uller 1987), especially under
F50{65.
Barzin,M., Errera, A., Sur la logique de M. Brouwer, Academie Royale des Sciences,
des Lettres et des Beaux Arts de Belgique. Bulletin de la Classe des Sciences. Cin-
quieme Serie 13, 1927, pp. 56{71.
Beeson,M.J., Foundations of Constructive Mathematics, Springer Verlag, Berlin 1985.
Beth, E.W., Semantic construction of intuitionistic logic, Koninklijke Nederlandse
Akademie van Wetenschappen. Afdeling Letterkunde: Mededelingen. Nieuwe Reeks
19, 1956, pp. 357{388.
Beth, E.W., The Foundations of Mathematics, North-Holland Publ. Co., Amsterdam
1959, 19652 .
Bishop, E., Foundations of Constructive Analysis, McGraw-Hill, New York, 1967.
Borel,E., Lecons sur la Theorie des Fonctions, Gauthier-Villars, Paris, 19142 , 19283 ,
19594 .
Brouwer, L.E.J., Leven, Kunst, Mystiek (Dutch) (Life, Art and Mysticism). Waltman,
Delft, 1905.
History of constructivism in the 20th century 28
Feferman, S., A language and axioms for explicit mathematics, in Algebra and Logic,
edited by J. Crossley, Springer Verlag, Berlin 1975, pp. 87{139.
Feferman, S., Constructive theory of functions and classes, in Logic Colloquium '78,
edited by M. Boa, D. van Dalen, K. McAloon, North-Holland Publ. Co., Amster-
dam 1979, pp. 159{224
Feferman, S., Weyl vindicated: `Das Kontinuum' 70 years later, in: Temi e Prospettive
della Logica e della Filosoa Contemporanee I, Cooperativa Libraria Universitaria
Editrice Bologna, Bologna 1988, pp.59{93.
Feferman, S., Hilbert's program relativized: proof-theoretical and foundational reduc-
tions, The Journal of Symbolic Logic 53, 1988, pp. 364{384.
Friedman, H., Set-theoretic foundations for constructive analysis, Annals of Mathemat-
ics 105, 1977, pp. 1-28.
Gentzen, G., Untersuchungen uber das logische Schliessen, Mathematische Zeitschrift
39, 1935, pp.176{210, 405{431.
Girard, J.-Y., Linear logic. Theoretical Computer Science 50, 1987, pp. 1{101.
Godel, K., Zur intuitionistischen Arithmetik und Zahlentheorie, Ergebnisse eines math-
ematischen Kolloquiums 4, 1933, pp.34{38.
Goodstein, R.L., Recursive Number Theory, North-Holland Publ. Co., Amsterdam,
1957.
Goodstein, R.L., Recursive Analysis, North-Holland Publ. Co., Amsterdam, 1959.
Heijenoort, J. van, From Frege to Godel, Harvard University Press, Cambridge
(Mass.) 1967.
Heyting, A., Die formalen Regeln der intuitionistischen Logik, Sitzungsberichte der
Preussischen Akademie von Wissenschaften, Physikalisch Mathematische Klasse,
1930, pp.158{169 Die formalen Regeln der intuitionistischen Mathematik, Ibidem
pp.57{71, 158{169.
Heyting, A., Mathematische Grundlagenforschung. Intuitionismus. Beweistheorie,
Springer Verlag, Berlin 1934.
Heyting, A., Intuitionism, an introduction, North-Holland Publ. Co., Amsterdam 1956,
19662 , 19713 .
Hilbert, D., Bernays, P., Grundlagen der Mathematik Vol. I, Springer Verlag, Berlin,
1934, 19682 .
Hyland, J.M.E., The eective topos, in The L.E.J. Brouwer Centenary Symposium,
edited by A.S. Troelstra, D. van Dalen, North-Holland Publ. Co., Amsterdam 1982,
pp. 165{216.
History of constructivism in the 20th century 30
Stigt, W.P. van, Brouwer's intuitionism, North-Holland Publ. Co., Amsterdam 1990.
Tait, W.W., Finitism, The Journal of Philosophy 78, 1981, pp.524{546.
Troelstra, A.S., Choice Sequences, a Chapter of Intuitionistic Mathematics. Claren-
don Press, Oxford 1977.
Troelstra, A.S., Analyzing choice sequences, The Journal of Philosophical Logic 12,
1983, pp.197{260.
Troelstra, A.S., Logic in the writings of Brouwer and Heyting, in: Atti del Convegno
Internazionale di Storia della Logica, San Gimignano, 4-8 dicembre 1982, edited by
V.M Abrusci, E. Casari, M. Mugnai, Cooperativa Libraria Universitaria Editrice
Bologna, Bologna 1983, pp. 193{210.
Troelstra, A.S., On the origin and development of Brouwer's concept of choice se-
quence, in: The L.E.J. Brouwer Centenary Symposium edited by A.S. Troelstra
and D. van Dalen, North-Holland Publ. Co., Amsterdam 1990, pp. 465{486.
Troelstra, A.S., On the early history of intuitionistic logic, in: Mathematical Logic
edited by P.P.Petkov, Plenum Press, New York, 1990, pp. 3{17.
Troelstra, A.S., Dalen, D. van, Constructivism. An Introduction, 2 vols., North-
Holland Publ. Co., Amsterdam 1988.
Turing, A.M., On computable numbers, with an application to the Entscheidungsprob-
lem, Proceedings of the London Mathematical Society 42, 1937, pp. 230{265 Cor-
rections, Ibidem 43, pp. 544{546.
Weyl, H., Das Kontinuum. Kritische Untersuchungen uber die Grundlagen der Analy-
sis, Veit, Leipzig 1918.