Zach Weber-Paraconsistency - in - Mathematics

Weber
Paraconsistent logic makes it possible to study inconsistent

theories in a coherent way. From its modern start in the
mid-20th century, paraconsistency was intended for use in
mathematics, providing a rigorous framework for describing
abstract objects and structures where some contradictions
are allowed, without collapse into incoherence. Over the The Philosophy of
past decades, this initiative has evolved into an area of non-
classical mathematics known as inconsistent or paraconsistent
Mathematics
mathematics. This Element provides a selective introductory
survey of this research program, distinguishing between
`moderate’ and `radical’ approaches. The emphasis is on
philosophical issues and future challenges.
Paraconsistency
Paraconsistency in Mathematics
in Mathematics
About the Series Series Editors
This Cambridge Elements series provides Penelope Rush
an extensive overview of the philosophy University of
Zach Weber
of mathematics in its many and varied Tasmania
forms. Distinguished authors will provide Stewart Shapiro
an up-to-date summary of the results of The Ohio State
current research in their fields and give University
their own take on what they believe are
the most significant debates influencing
research, drawing original conclusions.
Cover image: CTRd / Getty Images ISSN 2399-2883 (online)

ISSN 2514-3808 (print)
https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press
Elements in the Philosophy of Mathematics
edited by
Penelope Rush
University of Tasmania
Stewart Shapiro
The Ohio State University
PARACONSISTENCY IN
MA T H E M A T I C S
Zach Weber
University of Otago

University Printing House, Cambridge CB2 8BS, United Kingdom
One Liberty Plaza, 20th Floor, New York, NY 10006, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
314–321, 3rd Floor, Plot 3, Splendor Forum, Jasola District Centre,
New Delhi – 110025, India
103 Penang Road, #05–06/07, Visioncrest Commercial, Singapore 238467
Cambridge University Press is part of the University of Cambridge.

It furthers the University’s mission by disseminating knowledge in the pursuit of
education, learning, and research at the highest international levels of excellence.
www.cambridge.org
Information on this title: www.cambridge.org/9781108995412
DOI: 10.1017/9781108993968
© Zach Weber 2022
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2022
A catalogue record for this publication is available from the British Library.
ISBN 978-1-108-99541-2 Paperback
ISSN 2399-2883 (online)
ISSN 2514-3808 (print)
Cambridge University Press has no responsibility for the persistence or accuracy of
URLs for external or third-party internet websites referred to in this publication
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.

Elements in the Philosophy of Mathematics
DOI: 10.1017/9781108993968
First published online: July 2022
Zach Weber
University of Otago
Author for correspondence: Zach Weber, zach.weber@otago.ac.nz
Abstract: Paraconsistent logic makes it possible to study inconsistent

theories in a coherent way. From its modern start in the mid-twentieth
century, paraconsistency was intended for use in mathematics, providing a
rigorous framework for describing abstract objects and structures where
some contradictions are allowed, without collapse into incoherence. Over
the past decades, this initiative has evolved into an area of nonclassical
mathematics known as inconsistent or paraconsistent mathematics. This
Element provides a selective introductory survey of this research program,
distinguishing between “moderate” and “radical” approaches. The
emphasis is on philosophical issues and future challenges.
Keywords: nonclassical mathematics, inconsistent mathematics,

paraconsistent logic, paraconsistent mathematics, nonclassical logic
JEL classifications: A12, B34, C56, D78, E90
© Zach Weber 2022

ISBNs: 9781108995412 (PB), 9781108993968 (OC)
ISSNs: 2399-2883 (online), 2514-3808 (print)

Contents
Introduction 1
1 Invitation to Paraconsistency in Mathematics:

Why and How? 1
2 Set Theory 19
3 Arithmetic 40
4 Calculus, Topology, and Geometry 53
5 Whither Paraconsistency in Mathematics? 65
References 71

Paraconsistency in Mathematics 1
Introduction
This Element is an introduction to some uses of paraconsistent logic for math-
ematics. It is for beginners – or at least, readers who know enough about what
the words in the title mean to have picked this up, but not much more. In writing
it, I have mostly tried to take a neutral “tour guide” approach, both in the selec-
tion of material and in avoiding trying to “sell” the reader anything, though
my biases have inevitably shown through, especially by the end. The views
expressed in this Element are the author’s and do not necessarily reflect the
position of Paraconsistency Inc or its affiliates.
Each of the first four little sections exposits some key ideas (including, inev-
itably, using some formal symbolism); the last section of each discusses a more
general philosophical issue that arises. The final section is a brief philosophical
and critical appraisal, looking to the future of this little field.
1 Invitation to Paraconsistency in Mathematics:

Why and How?
Is mathematics consistent? Must it be? Or could new mathematical discoveries
be found where previously no one had thought possible to look – in the
inconsistent?
Paraconsistency in mathematics allows the development of mathematics
that either is or could be inconsistent but without absurdity. On a mainstream
approach in logic, any inconsistent theory is as good (or bad) as another,
because all contradictions are equivalent. Standard mathematics uses classi-
cal logic, according to which there is only one inconsistent theory – the one
that contains every single sentence, the absurd or trivial theory. With classical
logic, if there is even a single contradiction, everything collapses. In a slogan,
the conventional view is that “the inconsistent has no structure” (Mortensen,
2010, p. 3).
A nonclassical, paraconsistent conception of logic takes a more fine-grained
approach to contradictions. Paraconsistency distinguishes between a theory
being inconsistent (it includes at least one contradiction) from the notion of
being incoherent, absurd, or trivial (it includes every sentence); from a para-
consistent point of view, a theory does not have to be consistent to be coherent.
Paraconsistency in mathematics thus provides a rigorous framework for both a
cautious approach to contradiction – for all we know, a theory might be incon-
sistent, so we use a logic that can handle it – and a much less cautious project:
of investigating and describing contradictory abstract objects and structures.
Until recently, even the possibility of a contradictory, yet coherent, mathe-
matical theory would have been taken as facially absurd. (In some quarters, of

2 Philosophy of Mathematics
course, it still is.) Developments in formal logic since the 1950s, though, have
established that such theories can indeed exist. “Paraconsistent logic” now has
an official mathematics subject classification code.1
One aim of this emerging, diverse field of work is to widen the horizons
of mathematics by discovering and studying new objects – much in the way
that historically mathematics has advanced by admitting the existence of “bad”
entities like zero, negative numbers, irrational numbers, imaginary numbers,
transfinite sets, geometries where parallel lines can meet, and so forth. Newton
da Costa suggests that “it would be as interesting to study the inconsistent sys-
tems as, for instance, the non-euclidean geometries (da Costa, 1974, p. 498).”
For paraconsistency in mathematics, as Robert Meyer puts it, “what is to be
hoped for most of all are not new routes to old truths, but an expansion of the
pragmatic imagination (Meyer, 2021a, p. 158).”
The aim of this Element is to give the interested reader a critical sense of
some of the work in this area to date, its strengths and weaknesses, and to
indicate what might be next.
1.1 Motivations
Let us begin with an example of possible inconsistency in mathematics that
motivates using paraconsistent logic. The discussion proceeds informally. Then
we will get into a few details of how the logic itself works.
Mathematics is, by standard accounts developed in the twentieth century,
based on set theory – itself a mathematical theory of collections that provides a
foundation for all other areas of mathematics. Even very cautious philosophers
like Quine have grudgingly accepted that sets are indispensable for mathemat-
ical (and so scientific) practice. Mathematics takes place in the universe of
sets.2
Consider, then, the universe of sets – the collection of all sets, U. This is,
one imagines, a very big collection, the most inclusive collection of sets there
could be, one containing every set. This collection is, intuitively, the domain of
discourse for statements that set theorists are interested in, such as “every set
can be wellordered” or “there are no self-membered sets”; and whether or not
those statements are true, they do clearly seem to be – have been taken to be –
meaningful. The universal collection is the prima facie basis of the meaning
1 Under the Mathematics Subject Classification 2010, database of the American Mathe-
matical Society, 03B53: “Logics admitting inconsistency (paraconsistent logics, discussive
logics, etc).”
2 For a good account, see Potter (2004).

of these statements: when you study set theory, U is “where” you work. It has
been traditional at least since Whitehead and Russell’s Principia Mathematica
in 1910 to define the universal class with a “property posessed by everything,”
for example, the collection of all sets x such that x = x, or
U = {x : x = x},
which is universal because everything is, after all, self-identical.3 Every set x
is a member of this collection, x ∈ U.
Within U are subcollections of all sorts – the continuous functions, the com-
mutative groups, the set of all collections with exactly three members, and so
forth. One such collection is all the singletons. Every x has a singleton, {x}, and
the collection of all singletons comprises the sets with just one member. Now,
that means every x in the universe, every x ∈ U, can be paired up exactly, into
a one-to-one correspondence, with its singleton, in pairs
⟨x, {x}⟩.
This shows a natural sense in which the universe and its subcollection of all
singletons have exactly the same “size”: they can be paired off perfectly. But
of course, there are more objects in the universe than just the singletons, since
most things are not singletons. So there is also a natural sense in which U is
not the same size as the subcollection of all singletons.
This is a little puzzling, but we are talking about the entire universe after
all, so we should be prepared for some surprises – and indeed, the outstanding
mathematician Richard Dedekind used exactly this fact in 1888 to define what
it means for a set to be infinite; namely, having a proper part of the same size.
He then used a variant on the aforementioned argument to prove that infinite
sets exist.4
Now consider all the subcollections of the universe collected together, P(U)
(called the powerset of U). Since U is maximally inclusive, both P(U) and
all its members are inside of U. Writing subsethood as “⊆,” then
P(U) ⊆ U. (1)
3 See Whitehead and Russell (1910, p. 216). Here, we are focusing on the universe of sets, so
“everything is self-identical” is short for “every set is self-identical.”
4 In the infamous Theorem 66 of his Was sind und was sollen die Zahlen? of 1888 (reprinted
in (Dedekind, 1901, p. 64)), he argues that the set of his thoughts is infinite because for each
thought x there is also the thought of that thought {x}, the thought of the thought of that thought
{ {x} }, and so forth. See Priest (2006, p. 33, sec. 10.1). For alternative ways to think about the
sizes of infinite sets, see Mancosu (2009).

But similarly, each member of U is a set, which in turn has only sets as
members; so for any x ∈ U, if z ∈ x then z ∈ U, which is to say that x
is also a subset of U and hence a member of the powerset of the universe,
showing
U ⊆ P(U) (2)
But, if all the members of U are members of P(U) and vice versa, then these
collections are exactly the same collections; by (1) and (2),
U = P(U) (3)
because sets with all and only the same members are the same set, by the
principle of extensionality.
A powerset contains all possible recombinations of elements of a set, so this
equality is a little odd – but also natural enough, maybe, since the set in question
is the universe. Note that, since U is a subset of itself, that is, U ∈ P(U),
then this means U ∈ U. The universe is contained in itself; the universe is
everything, after all.
But now we have a real problem. By (3), U and P(U) must be the same size
(since they are the same set). If they are the same size, there is a way to pair
off their members in a one-to-one correspondence. Call such a pairing f, that
matches members x ∈ U with members y ∈ P(U), as in f(x) = y. For members
x of U paired up with subsets y of U, sometimes x will be in that subset, and
sometimes not.5 So consider
r = {x : x < f(x)}
This is a subset of the universe comprising all the things that are not in the set
they are paired with. But then, since f pairs off everything, and r ∈ P(U), there
must be some x ∈ U such that f(x) = r. Now we just have to ask: is x ∈ r,
or not? If it is, then x < f(x) by definition, so x < r after all; yet if x < r, then
x ∈ f(x) again by definition, so x ∈ r after all. Since x is either in r or it is not,
it is both: contradiction.
Classically, this general argument has been taken as a reductio – requiring
the rejection of some assumption, usually the existence of a pairing off between
a set and its powerset, or a universal set, or both (see §2.3). But the existence of
a set theoretic universe is extremely hard to shake; and since we independently
established that U = P(U), it looks like there must be at least one such pairing
5 For instance, if a = {1, 2, 3} and b = {2, 4, 6} then if f(1) = a, we would have 1 ∈ f(1), and if
f(1) = b then 1 < f(1).

off, namely, identity f(x) = x. Under this identity mapping, where f(r) = r, then
writing “iff” for “if and only if,”
r ∈ r iff r ∈ f (r)
iff r < f (r)
iff r < r.
So we seem to have shown that

r ∈ r and r < r
as an apparently natural consequence of some apparently true premises. There

are no assumptions to reject, or at least, none the rejection of which would be
less puzzling than the conclusion that some “diagonal” subset of the universe
has contradictory properties. In brief, “naive” set theory is inconsistent.
This fact was discovered in the late nineteenth century,6 but was labeled a
paradox: an apparently sound argument for an apparently impossible result. In
more than a century since then, no fully satisfactory solution to this paradox has
been found, as witnessed at least by a continuing stream of dissatisfied research
on the subject, as we will see. A paraconsistent approach offers a unique take on
this situation: there is no problem here to solve. The paradox is simply a proof,
and its conclusion is a theorem: some sets are inconsistent. Put cosmically, the
universe is larger than itself.
Now, none of this forces us into “inconsistent mathematics”: historically,
this and other examples all have consistent accounts. In set theory, the official
solution is that there is no universal set; the collection U, the domain of set
theory (and mathematics) itself is not a set. If that seems good enough to you, I
do not plan to try to talk you out of it.7 Crucially, to see a place for paraconsis-
tency in mathematics, it seems like I do not need to talk you out of it. Indeed,
as we will be canvassing, the majority of people working in paraconsistency
are happy to accept standard mathematics. To move ahead with paraconsistent
investigations, all we need is the suggestion that there are interesting things
that seem to be inconsistent when we think about them. If there could be more
to learn than is allowed by the assumption of consistency, there is a place for
6 The fact that there cannot be a one-to-one correspondence between a set and its powerset was
proven by Cantor and is called Cantor’s theorem. The fact that Cantor’s theorem becomes incon-
sistent at the universe was known to Cantor by 1895 or so but was made especially public
by Russell in 1902 and is called Cantor’s Paradox. See van Heijenoort (1967, p. 124) and
Section 2.
7 If it does not seem good enough, though, I think you might be right and suggest you see Priest
(2006, ch. 2) and Weber (2021, ch. 1).

paraconsistency. Whether or not we should think of paraconsistent theories as

true or their investigations as a challenge to the official consistent approaches
in mathematics is not obvious and is a topic we will return to.
And foundational disciplines like set theory are not the only place some have
found more to learn. We will look at further topics in arithmetic, calculus, and
topology. Chris Mortensen, a key founder of the field of inconsistent mathemat-
ics, has focused on impossible pictures. These have philosophical motivations,
but from a purely mathematical point of view, perhaps it is enough just to
glimpse something new there to study: the inconsistent has some structure after
all, and we can pursue structure for its own sake. We may benefit from drop-
ping what Meyer, Routley/Sylvan, and Priest call the universal consistency
hypothesis – the limiting supposition that there is nothing of rational interest
outside of the bounds consistency8 – and widen our horizon of investigation to
points beyond.
Let us look at some precise ways one might do this.
1.2 Methods
1.2.1 Paraconsistent Logic Tutorial
A paraconsistent logic allows inconsistency without absurdity. In a non-
paraconsistent logic, any inconsistent premises p, ¬p will have any arbitrary
q as a valid conclusion (ex contradictione quodlibet, or explosion); a logic is
paraconsistent if and only if explosion is invalid. Denying explosion is all that
is required for a logic to be paraconsistent.
More precisely, let us say that a logic is determined by a consequence relation
⊢ that relates some sentences (premises) to another (conclusion).9 When the
consequence relation
p0, ..., pn ⊢ q
holds then the argument from p0, ..., pn to q is valid; and if not, not. And then
let us say that a theory is a set of sentences closed under logical consequence:
the “starting” sentences, and all the ones that validly follow under ⊢. An incon-
sistent theory contains both some sentence p and its negation ¬p. And so an
inconsistent theory under a non-paraconsistent logic will include every sen-
tence, which makes the theory trivial. Thus, if an inconsistent theory is to be
8 “...that all that can be spoken of or described (non-trivially) is consistent” (Priest, Routley, &
Norman, 1989, p. 4; cf. Routley and Meyer, 1976).
9 There is a lot of good (paraconsistent) work on multiple conclusion consequence relations; see
Beall and Ripley (2018, p. 744). The focus on single conclusion here is to keep it simple.

nontrivial, it must, it seems, be embedded in a paraconsistent logic, where the

principle of explosion is invalid:
p, ¬p ⊬ q.
Paraconsistent logics are the basis for the study of paraconsistent theories.
There are many strategies for making a logic paraconsistent, and within
these strategies there are many – infinitely many – paraconsistent logics. For
concreteness, let us look at one simple approach due to Asenjo (1966) (cf.
Asenjo and Tamburino, 1975) and then to Priest (1979). This is to general-
ize the standard truth conditions on logical evaluations and the definition of
semantic validity, opening some extra space that leaves classical conditions as
a special case.
Just as with classical logic, we are given a formal language with connec-
tives ¬ (negation, “not”), ∧ (conjunction, “and”), and ∨ (disjunction, “or”),
with propositional atoms p, q, ... connected by these connectives into complex
expressions A, B, .... The material conditional A ⊃ B is defined ¬A ∨ B and a
biconditional A ≡ B is defined as (¬A ∨ B) ∧ (¬B ∨ A). The new twist is in
an assignment ν taking sentences of the language to (two) truth values, t and f.
Sentences may be true, or false, or – now diverging from classical logic – both.
On this arrangement, no sentence must be “both” but some can. This is pos-
sible because ν is a relation, rather than a function.10 Relations can be multiple,
as in “y is a place x has lived” can take one x to multiple values for y. While an
evaluation function would treat a “true” contradiction as having a value equal
to both true and false, t = ν(p) = f, and hence11 t = f (which not even para-
consistentists will approve of), a relation will let t and f be among the values
of ν(p), one not always ruling the other out.
Relational truth conditions for negation, conjunction, and disjunction may
be spelled out in a standard-looking homophonic way:
ν(¬A) is at least t iff ν(A) is at least f

ν(¬A) is at least f iff ν(A) is at least t
ν(A ∧ B) is at least t iff ν(A) is at least t and ν(B) is at least t
ν(A ∧ B) is at least f iff ν(A) is at least f or ν(B) is at least f
10 Following an idea from J. Michael Dunn in the 1960s (see Omori and Wansing, 2019b), pub-
lished as Dunn (1976); cf. Priest (2008, p. 161). Both Asenjo and Graham Priest present this as
a three-valued logic, where the “both” value is a third distinct status along with truth and falsity.
They treat the relation ν as a function, assigning each sentence exactly one of the three possi-
ble statuses. There are strong reasons why it is philosophically preferable to take the relational
approach, as presented here; see Weber (2021, ch. 3).
11 At least, assuming the transitivity of identity: a = b, b = c ⊢ a = c. This is disputed in Priest
(2014).

ν(A ∨ B) is at least t iff ν(A) is at least t or ν(B) is at least t

ν(A ∨ B) is at least f iff ν(A) is at least f and ν(B) is at least f
On this account of negation, a sentence is evaluated as true iff its negation is

evaluated as false; a sentence is evaluated as false iff its negation is evaluated
as true; so if a sentence is evaluated as both true and false, then its negation is
also evaluated as both true and false.
An argument is valid if, whenever all the premises are at least true, then so
is the conclusion; an argument is invalid if there is some way for the premises
to be at least true but the conclusion to be not at all true. Heuristically, validity
is un-truth preservation backward, from conclusion to at least some premises.
If we also assume the relational valuation is exhaustive,
ν(A) is at least t or ν(A) is at least f
then this is the idea behind the logic now known as LP (for “logic of paradox”).
It is paraconsistent because if we just consider the sentences p and q, then it is
possible for p to be assigned both true and false, while q is not assigned true.
Then p and ¬p are both assigned at least true, but not at all so for q, and so ex
contradictione quodlibet is invalid – it has a counterexample.
This is a way of generalizing classical logic. If the relation ν were tightened
up to be a function, so that, for example, ν(¬A) = t iff ν(A) = f, then these con-
ditions simply are those of classical logic. Without assuming functionality, the
logic allows for gluts. We do assume that the relation ν assigns every sentence
at least value t or f, making negation exhaustive:
A ∨ ¬A
is always assigned at least value t. If one wanted to, this condition could be
dropped, and then the logic would allow gaps too; that would deliver the
logic FDE, which can then serve as a base to extend to several other sorts of
paraconsistent logics; see Belnap (1977). In LP there are gluts but no gaps.12
Nothing about the general idea of paraconsistency would seem to commit
to actual gluts (or true contradictions), only the hypothetical that even if there
were some true contradiction, still not everything would be true.13 This opens
12 In this Element, we will not be considering “gappy” approaches, because intuitionistic and con-
structive mathematics have already told us a lot about the “incomplete” (but consistent) side of
things, and it seems worthwhile to try to understand the dual. Also, without the law of excluded
middle (LEM), many of the standard paradoxical proofs of contradictions – those that motivate
inconsistent mathematics to begin with – fall apart; one can derive theorems of the form “A
iff ¬A” but no further. The question of whether one can still derive contradictions without the
LEM has been a point of contention between Priest and Brady; see Brady and Rush (2008);
Priest (2019).
13 See Barrio and Da Re (2018).

a space, then, for different approaches to paraconsistency in mathematics,

which we will look at. Moderate approaches will be based on a paraconsis-
tent consequence relation, without any commitment to gluts, whereas so-called
strong paraconsistency endorses some true contradictions, sometimes called
dialetheism.
1.2.2 A Very Brief History of Paraconsistency in Mathematics

There are many approaches to paraconsistency one could take; all that is
required for paraconsistency is that the argument from p, ¬p to arbitrary q be
invalid. There are almost as many approaches to paraconsistency as there are
logicians who have taken them. For present purposes, here is a potted history
(restricting attention only to mathematical applications, not the development of
paraconsistent logic). This also serves as a forecast for the sections ahead.
Paraconsistent logics were developed in various forms in the first half of the
twentieth century, most notably by Jaśkowski in 1948. The first paraconsistent
approaches connected with mathematical practice were developed independ-
ently by Florencio Asenjo in 1954 and then very prominently by Newton C. A.
da Costa in 1963. Asenjo presented what he called antinomic number theory.
Da Costa introduced paraconsistent set theory; this was investigated with Ayda
Ignez Arruda in the 1960s and 1970s. In 1986, da Costa proved that his sys-
tem is nontrivial. This flourishing line of research continues in work by many,
including Walter Carnielli, Marcelo Coniglio, Itala D’Ottaviano, João Marcos,
and many others, and is sometimes called paraconsistent mathematics.14
Diderik Batens, founder of the adaptive approach to paraconsistency, worked
with Arruda on set theory when she visited Europe in the late 1970s. Batens
had proposed what he then called dynamic dialectical logic, which is intended
to model reasoning in science and mathematics. He has most recently applied
adaptive logics to set theory, in light of some significant developments from
Peter Verdée.
In the 1970s in Australia, Richard Routley/Sylvan and Robert Meyer started
investigating what they called “dialectical” theories, including set theory and
arithmetic. Routley visited the State University of Campinas in 1976, and
da Costa visited the Australian National University in 1977.15 Unpublished
manuscripts by Routley and Meyer were widely distributed, not only to col-
leagues in Australia but also da Costa and others, containing the outlines
14 Interested readers may consult Gomes and D’Ottaviano (forthcoming).

15 According to Arruda (1989, p. 107). Lengthy handwritten correspondence from the late 1970s
between Routley and da Costa (“Dear Brother Richard/Dear Brother Newton”) are preserved
in the archive at the University of Queensland.

of what is now sometimes called inconsistent mathematics. These included

Routley (1977) and Meyer (2021a, 2021b).
Circa 1976, Meyer produced a model of inconsistent arithmetic; such models
were further developed by Meyer and Mortensen through the 1980s, and then
Priest in the 1990s, using techniques due to Dunn. Important limiting results
for Meyer’s arithmetic were found by Harvey Friedman and Meyer in 1992.
Circa 1978, Ross Brady produced a model of inconsistent set theory, which
he developed further with Routley and reached culmination in a 2006 book;
that set theory was studied further by Weber (me!). In 1995, Mortensen and
collaborators published a book outlining what he termed inconsistent math-
ematics across several new areas, including calculus, topology, and category
theory, and in 2010 followed this with further investigations in inconsistent
geometry.16
Today, with the widespread acceptance of the legitimacy (if not correctness)
of nonclassical logics, paraconsistent logics are being developed, sometimes
with applications to mathematics, in many directions across the globe. The
alternating terms “paraconsistent mathematics” or “inconsistent mathematics”
have been used to some extent as shorthand to demarcate regional traditions
and variations. Marcos has urged that logic does not carry a passport, and I
similarly think there is little to be gained by insisting on geographic labels.
What matters is that the formal development of paraconsistent logic(s) creates
an opportunity in mathematics: the rigorous study of theories that are possibly
or actually inconsistent.
1.2.3 Goals: Recapture, Expansion, Revision

Paraconsistency in mathematics deploys a nonclassical logic for mathematical
work. How sweeping a change, if any, is this to ordinary practice?
We just saw that there is no one project or program called “paraconsistency
in mathematics” or even “inconsistent mathematics.” The story of paraconsis-
tent mathematics told here is of two countervailing forces. On the one hand,
there is a moderate or conservative line of research, seeking mainly to fit a
paraconsistent logic into the wider standard mathematical picture, perhaps as
a kind of buttress or insurance policy against any future inconsistency.17 On
the other hand, there is a radical, revisionary line that not only seeks to extend
knowledge in novel nonclassical directions, but perhaps also to overturn pre-
viously accepted wisdom, rewriting the rules of the game. The conservative
aim is to prove that 1 + 1 = 2, even if there is some chance of inconsistency;
16 Interested readers may consult Brady and Mortensen (2014).

17 There is guarantee that mathematics is consistent, due to Gödel’s theorems; see §3.

the radical aim is to study paradoxical objects directly, on their own terms –
expanding mathematical horizons, ideally, in the way that irrational numbers
or non-Euclidean geometries opened new directions in the past.
This moderate/radical distinction is not cut and dry, and we will be revisiting
it throughout. Nevertheless, we can see at least two competing goals. First, there
is classical recapture. This is to show that most or all of accepted classical
mathematical theorems are true in a paraconsistent framework.
In order to sustain the [paraconsistent] challenge to classical logic it will have

to be shown that even though leading features of classical logic and theories
have been rejected, one can still get by. In particular, it will have to be shown
that by going [paraconsistent] one does not lose great chunks of the modern
mathematical megalopolis. (Routley, 1977, p. 927)
The metaphor of “great chunks” is a little imprecise, but the intent seems
clear enough. The idea is to reassure a would-be paraconsistentist that noth-
ing important is lost. Second, there is expansion: adding new theorems and
insights to the stock of mathematical knowledge, particularly about objects and
structures that are beyond the classical pale.18 This direction is more indefinite
and open-ended; as such, it has also received less attention. But I will be sug-
gesting that it is the direction that, ultimately, may decide the future of this
enterprise.
Some early proponents of inconsistent mathematics suggested that recapture
and expansion would come in a natural package, because using paraconsis-
tent logic for mathematics would involve no change at all to ordinary practice.
This is because, they argued, ordinary practice is not classical. That is, classi-
cal logic is not a good formalization of what actually goes on in real proofs.
Following Anderson and Belnap (1975), people like Routley/Sylvan, Meyer,
and Priest argued that if one formalized the canons and norms of reasoning
found in everyday mathematical use, where people do not in fact routinely
infer arbitrary conclusions from arbitrary contradictions, the result would be
a “relevant” (or “relevance”) paraconsistent logic.19 Priest says “we will have
to relinquish ‘classical’ logic” (Priest, 1979, p. 226); but Routley says this will
not be a problem if
18 An analogy might be how the study of transfinite numbers may require different, more strin-
gent rules, but does not change any facts about the good old natural numbers; for example,
addition in N is commutative (n + m = m + n) but not necessarily so for infinite ordinals
(ω + 1 , 1 + ω). To encourage this analogy, Priest coins his work “a study of the transcon-
sistent.” Carnielli and Coniglio use the same analogy to explain the place of their (different)
approach to paraconsistency (Carnielli & Coniglio, 2016, p. x).
19 “Imagine someone offering [an explosive argument] as a proof . . . in a class on number theory.
It is clear that it would not be acceptable” (Priest, 2008, p. 74, §4.74).

Insofar as mathematics relies on valid argument, its proper formalization

is not in terms of classical logic. . . . The bulk of intuitive mathematics is
not classical, except insofar as recent classical logical reconstructions have
pushed it in that direction. (Routley, 1977, p. 903)
Viewed this way, some fairly radical ideas can be presented in an apparently
moderate light. Priest states that “the programme of paraconsistent logic has
never been revisionist” and that “by and large, it has accepted the reasoning of
classical mathematics is correct.” He makes this claim on the basis that logic
as used by actual mathematicians would turn out to be paraconsistent:
What [paraconsistency] has wished to do is to reject the excrescence of ex
contradictione quodlibet which does not appear to be an integral part of clas-
sical reasoning but merely leads to trouble when reasoning ventures into the
transconsistent. (Priest, 2006, p. 221)20
A more mainstream view is that standard mathematics is couched in classi-

cal logic, perhaps, but paraconsistency in mathematics can be fruitfully applied
to amplify past work without revising or overturning it. For researchers of a
more moderate or conservative variety, classical recapture may be very easy;
for example, perhaps any classical theories can simply be added to a paracon-
sistent base, by assuming them via additional axioms.21 One way to do this is
by assuming various instances of explosive reasoning as correct, because they
involve consistent premises, and recovering standard proofs that way. Or relat-
edly, perhaps this dichotomy – conservative versus radical – can be undercut,
by separating logic from mathematics. This might be because logical valid-
ity concerns very general reasoning, whereas mathematics is structure-specific
reasoning.22
In practice the way recapture and expansion is studied is to work with models
of mathematical theories. This is probably the approach most used throughout
paraconsistent research. On this approach, one takes a classical model of, say,
arithmetic, and alters it in a controlled way to produce an interpretation of an
inconsistent theory (see §3). Mortensen, especially in his 1995 book, uses this
technique, with a focus on checking which theories obey substitution of identi-
cals (if a = b then A(a) implies A(b)) as crucial for the needs of doing (useful)
calculations. Working with “inconsistentized” classical models can be under-
stood as a riff on the idea that paraconsistency is a generalization of classical
20 Though Priest admits this conception results in “some tension” [p. 248], and his views have
developed since 2006 (see Priest [2013] and his Element on pluralism).
21 This is the approach taken by Brady in later chapters of Brady (2006) on the development of
mathematical theories. See Beall (2009, p. 111); cf. Beall (2013b), and Omori and De (2022).
22 Ideas suggested in various ways by Jc Beall, Mortensen, Stewart Shapiro, and more recently
perhaps Priest; cf. Mortensen (2009); Shapiro (2014).

logic. One can work with models to show that, for example, models of para-
consistent set theory already include models for classical set theory (see §2).
One advantage of working model theoretically is that a lot of work has already
been done with models.
For this approach to be coherent, it appears to require assuming that classical
mathematics, and model theory in particular, is correct. This assumption makes
sense from a moderate, conservative sort of standpoint that does not dispute
accepted mathematics or its formalization in classical logic. For example, on
some approaches to paraconsistency,
there is no point in advocating the replacement of classical logic with para-

consistent logic. . . . [T]here is no rejection of what has come before, but a
refinement of it (Carnielli & Coniglio, 2016, p. x).
Such a take on paraconsistency is clearly not mounting any “challenge” to clas-

sical logic, but rather “extends the classical stance” to build on the existing
modern mathematical megalopolis.
In a similar spirit, we might expect harmony between approaches because
“mathematics has a certain primacy over logic” (Mortensen, 1995, p. 5). As
Mortensen puts it,
The properties of inconsistent theories tend to be invariant over a large class

of background logics. . . . [L]ogical properties tend to take second place
behind mathematical calculations which are performed at the sub-atomic
level (sub-atomic relative to the atoms of logic, that is). (Mortensen, 2009,
p. 634)
The tantalizing idea here is that mathematical truth is invariant under changes
of logic, or alternatively that invariance under logic change is the mark of
mathematical truth.
These views reflect that most work in logic and mathematics is conducted
in a cooperative spirit, and there is a lot to show for it. However, recalling the
idea that looking for structure in the inconsistent is in some sense a repudi-
ation of the “universal consistency hypothesis,” there is a clear way in which
a more “fundamentalist” paraconsistent approach may be committed to saying
that some classical mathematics is not correct. For example, standard set the-
ory cannot make sense of the “paradoxical” arguments mentioned earlier about
the universal set, and this could be taken as a critical failing of the standard
approach – a failing at the very foundation of mathematics.23
23 There are non-paraconsistent set theories with a universal set. These raise other issues; see
Forster (1995).

So a more radical way still is to say that paraconsistency in mathematics may

call for a sweeping change to ordinary practice, and this would begin by trying
to give newly validated proofs of various theorems in elementary mathematics.
This means venturing onto less stable ground, beginning from some first prin-
ciples and trying to reason through the cannon of mathematical results going
back to before Euclid. An even-tempered way to do this might be to start with
Peano’s axioms for arithmetic and try to deduce in a paraconsistently accept-
able way that 2 + 2 = 4 (see §3). This approach has the advantage of honesty
(or some might say zealotry) but also means working without the reassurance
of models, since paraconsistent models do not exist until some proofs about
them do. This way lies work outside the comfort of “recognized” mathematical
practice.24
We will come back to this briefly at the end, when we have more evidence to
examine. For now, let us round out our logical preliminaries and demarcation
of different approaches by looking at some specifics choices of paraconsistent
tools.
1.2.4 Into the Paraconsistent

The way LP makes room for gluts – sentences that are both true and false – is
by its treatment of negation. From the conditions in §1.2.1, one can observe
that some familiar properties, especially of negation, are forthcoming. Letting
⊢ be the consequence relation for LP, then double negation elimination and
introduction, and the de Morgan laws, all hold:
¬¬A ⊣⊢ A
¬(A ∧ B) ⊣⊢ ¬A ∨ ¬B
¬(A ∨ B) ⊣⊢ ¬A ∧ ¬B.
That seems good from the standpoint of trying to generalize or extend classical
logic without changing it too much.25
The law of excluded middle (LEM) A ∨ ¬A is always evaluated as (at least)
true. That means, using de Morgan’s laws, that any conjunction of contradictory
24 Rosenblatt worries that “the project of proving [standard] results for one’s preferred deviant
logic is going to be either impossible to carry out or, in the best case scenario, a massive under-
taking” Rosenblatt (2022b). He has a point, though I do not see this as an objection to trying. For
small steps into a serious nonclassical metatheory, see Rosenblatt (2021b); cf. Weber, Badia,
and Girard (2016), Badia, Weber, and Girard (2022). Meyer already attempted some work like
this in Meyer (1985).
25 Compare this with intuitionistic negation, which does not obey all these laws. See Priest (2008,
ch. 6) and Posey’s Element on intuitionism.

sentences A ∧ ¬A, even glutty ones, will be at least false. A distinctive feature
of negation in LP, then, is that the law of noncontradiction
¬(A ∧ ¬A)
is a theorem: all contradictions are at least false; some may also be true.
It is worth saying that again. A logic designed to handle truth gluts has as a
theorem that there are no truth gluts. So the so-called glutty approach is itself
glutty.26 This is a feature, not a bug. The logic LP does not itself have any
contradictions in it, but it is a logic for seriously working with contradictions,
in an appropriately contradictory way. The language of LP, and in particular its
negation connective, is so open-minded, in fact, that there is nothing to prevent
an evaluation assigning every sentence of the language both truth values, a
“trivial” model (see §2.1).
And so LP is, by design, much weaker than classical logic. Most notably, the
common argument form of disjunctive syllogism is invalid:
A, ¬A ∨ B ⊬ B
since if A were assigned both t and f, but B not assigned t, then there would be
a way for the premises to be true but not the conclusion. And that means, as the
reader may have noticed, that LP has no implication connective; the “material
conditional” p ⊃ q as defined ¬p ∨ q will fail to obey modus ponens,
A, A ⊃ B ⊬ B
or indeed most other behaviors expected from a conditional; see Beall, Forster,
and Seligman (2013).
Given the absence of a conditional in LP one can try to add a conditional.
One way to do so – the one favored by, for example, Routley, Brady, Meyer,
and Mortensen in their investigations of inconsistent mathematics – is to adopt
some version of implication from relevant paraconsistent logic, →, where if
A→B
holds then there must be some “meaningful” connection between the anteced-
ent and the consequent.27 This conditional obeys modus ponens,
A, A → B ⊢ B
among other properties; but it is still much weaker (or stronger, depending on
your viewpoint) than classical “material” implication, and is not entirely happy,
26 See “Concluding Self-Referential Postscript” to Priest (1979).

27 For relevant implication studied using modal semantics, see Priest (2008, ch. 10).

as we will see. There are other conditionals to try adding to LP as we will see,
but none of them yet known are the “universal key, which opens, if rightly
operated, all locks.”28
Other paraconsistent approaches favor a different account of negation.
Asenjo (1966) states that the law of noncontradiction should not be a theorem
of any logic suitable for formalizing inconsistent theories. The pioneering work
of da Costa takes a similar attitude, where he states as his first adequacy condi-
tion on any paraconsistent logic that “¬(A ∧ ¬A) must not be a valid schema”
(da Costa, 1974, p. 498). Da Costa’s main logics are known as the C-systems,
and have since been subsumed and brought more sophistication under a richer
family of logics known as logics of formal inconsistency, or the LFIs.
In LFIs the idea is to allow explosion “in a local or controlled way” (Carnielli
& Coniglio, 2016, p. 31).29 A characteristic feature of this approach is to
include in the language an operator ◦ that may be read as “is consistent,” and
so is designed to separate out the “normal” or well-behaved sentences from
the paradoxical ones, demarcating sentences that may be reasoned about clas-
sically. (Note that if ◦(p) holds then ◦(¬p) holds, too.) So while in an LFI the
principle of explosion is invalid,
p, ¬p ⊬LFI q,
one does have “gentle” explosion,
p, ¬p, ◦(p) ⊢LFI q.
An explosive negation is definable: it holds when ¬p and ◦(p) both hold.

Initially, this was expounded by da Costa in a hierarchy of logics, where one
ascends the hierarchy through iterations of the consistency operator, asserting
ever more strongly that not only is p consistent, but it is consistent that p is con-
sistent, on to infinity; see §2.1. Whereas da Costa defined ◦(p) as ¬(p ∧ ¬p), in
the LFIs it is taken as primitive and has wider application (Carnielli & Coniglio,
2016, pp. 9, 71). By making the notion of “consistency” expressible in the logic
itself, these are not so much logics for inconsistency as logics of inconsistency.
They do not self-contradict themselves the way LP does about truth gluts but
provide a consistent external observation of what may be coherent but incon-
sistent internal reasoning, and mark off where classical reasoning can be safely
applied.
28 As Routley once envisioned (Routley, 1977, p. 893). Alternatively, the idea of sticking with
LP alone as a logic (or even the weaker FDE), without a new conditional, has been favored in
various moods by Beall (2013a, 2018), Omori (2015), and others, following a suggestion of
Laura Goodship in Goodship (1996). See Priest (2017b).
29 See Carnielli, Coniglio, and Marcos (2007) and Marcos (2005).

More widely, the idea of assuming or “recapturing” various bits of classical-

ity is a hallmark of prominent approaches to paraconsistency. On such moderate
ways of approaching paraconsistency, then, the idea is that “consistent” objects
can still be reasoned about classically.30
A way of operationalizing the idea is the adaptive logic approach developed
in detail by Batens and collaborators. Adaptive logics take a non-monotonic
approach, where a valid argument may become invalid under the addition of
more premises, for example, “If x is well-founded then x is not self-membered”
might be taken as true, but the augmented “If x is well-founded and x is the
universal set, then x is not self-membered” may not be true any more. So
proofs are dynamic. This is done via a model-selection strategy, based on three
components:
• A “lower limit logic” that is the stable, static base logic

• A set of “abnormalities” that are certain formulas of the lower limit logic
taken to be false
• a strategy for selecting models, most saliently, “minimize the abnormal-
ities”
In such systems “the certainty their proofs deliver is only pragmatic, not
irrevocable” (Verdée, 2013b, p. 661).31 These are mitigation strategies for
inconsistency in mathematics.32
This push-and-pull tension between conservative and radical tendencies
comes out most clearly in deciding on what language and techniques a para-
consistent mathematician might use to present their own work: using accepted
classical methods for their theories, or making a revolutionary (but also
untested) ascent to an inconsistent “metalanguage” beyond the classical? As
several critics have pointed out, most of the attention in paraconsistency has
gone to neither classical recapture nor expansion, but rather the more basic one
of paraconsistent recapture: confirming that various paraconsistent systems are
themselves coherent, by (classically) proving that they have classical models
30 For recent attempts at adding a consistency operator, see Barrio, Pailos, and Szmuc (2017); for
problems with this and other strategies, see Rosenblatt (2021a) and cf. Omori and Weber (2019).
31 The logic CLuNs is often taken as the lower limit logic for such an approach; this logic, or
variants thereof, have been independently proposed across the paraconsistency literature. It is
equivalent to LFI1 as well as da Costa and D’Ottaviano’s J3, and has the same underlying
algebra as Łukasiewicz’s Ł3 (Carnielli & Coniglio, 2021, p. 192). The underlying algebras of
Ł3 and “relevant” RM3 are also term equivalent, assuming the presence of a ⊥ constant. (Thanks
to Omori here.)
32 A similar proposal is Priest’s “minimally inconsistent LP,” which works by selecting the least
inconsistent models and recovering more classical reasoning for cases that appear to be safe, but
where such reasoning may become unacceptable if contradictions were to arise (Priest, 2006,
ch.16).

or are otherwise consistent relative to some classical structure (see Appendix

to §2).33 These choices will be kept in view throughout.
1.3 Realisms and Anti-realisms About (Inconsistent)

Mathematics
What sort of philosophical attitude best fits paraconsistent/inconsistent math-
ematics? In particular, what of the hoary old choice between some kind of
realism (that this field studies mind-independent existing objects) or some
variety of anti-realism (that it does not)? The short answer is that, insofar
as this is simply another area of mathematics, whatever stance one takes on
these questions elsewhere can carry over to here, perhaps with some attenuated
consequences.
The standard version of realism in mathematics is platonism. One very strong
version of this, plenitudinous platonism, asserts that any consistent mathe-
matical theory correctly describes some part of (abstract) reality (Balaguer,
1998). In this respect, perhaps surprisingly, plenitudinous platonism is not far
off from Hilbertian formalism, which asserts that any consistent theory is a
legitimate part of the syntactic symbol shuffling game called mathematics.
Since paraconsistency will allow some inconsistent mathematical theories,
too, Beall proposes upgrading plenitudinous platonism to “ really plenitudi-
nous platonism,” which relaxes the constraint on theories so that they only need
be non-trivial (Beall, 1999). Really plenitudinous platonism also has a corre-
sponding formalist version, a kind of “eclectic Hilbertianism” as in Shapiro
(2014).
In a similar vicinity, a Meinongian will accept any coherent mathematical
talk as being true of objects that may not exist but that nevertheless can stand
as the subjects of true statements; see Routley (1980b) and Priest (2005). On
this line, our theories are constrained only by our imaginations, as mathematical
objects are assumed to have the properties they are characterized by: there is a
universal set and it is both bigger than itself and not; it just does not exist.
On the more anti-realist end, one might take an epistemological inter-
pretation of paraconsistency. This is the approach favored by Carnielli and
Coniglio (2016, sec. 1.2). Here, inconsistency is not anything metaphysical,
but rather a feature of our beliefs or reasoning. As Carnielli and Coniglio
put it, by analogy, Brouwer’s intuitionistic logic features failures of excluded
middle because it is about mental constructions or proofs rather than truth;
dually, paraconsistent logics feature failures of explosion also because of a
33 See Burgess (2005) and Weir (2004).

mental, rather than ontological, focus (Carnielli & Coniglio, 2016, pp. 4, 14).
Or again, paraconsistent or inconsistent theories can capture the content of
our cognition or experience of apparently contradictory phenomenon; this is
the position taken by Mortensen (1995, 2010). Paraconsistency in mathemat-
ics might then be an account of mental constructions or imaginings, leaving
the logic of mind-independent truth values to classical logic or something
else.
Similar comments will apply to other positions. If you are a structuralist
about mathematics, you can be a structuralist about inconsistent mathemat-
ics; mutatis mutandis for fictionalism; and so forth. Inconsistent mathematics
may add some pressure to these positions by adding some unexpected commit-
ments, but where to draw the line has always been a question for such positions.
For example, perhaps you are happy to say that natural numbers really truly
exist, but become uncomfortable with a similar platonism about large trans-
finite cardinals. So here for realism, or not, about objects with inconsistent
properties. We will return to this issue in §4.4 on the topic of indespensibility
arguments.
The relationship between language and reality (whatever that is) does let
us answer a question about the meaning of expressions like “inconsistent
mathematical structure.” Inconsistency is a linguistic property, concerning a
sentence and its negation; whereas a structure, or its parts, cannot be “negated”
in a meaningful way. So it is worth clarifying that any talk about inconsist-
ent objects is a kind of shorthand for talking about an inconsistent theory, or
description, of an object. Whether one thinks that the inconsistency comes from
our mind, language, or the object itself, is a further question. As Priest puts it,
concerning the question of the “consistency of the world,”
[i]f something is true, there must be something that makes it so. Call this
the world. If some contradictions are true, then the world must be such as to
make this the case. In this sense, the world is contradictory. What it is in the
world that makes something true is another matter. (Priest, 2006, p. 299)
2 Set Theory
One of the first and most enduring motivating projects in paraconsistent math-
ematics has been to address the infamous paradoxes of “naive” set theory,
discussed in §1.1. For, ever since Frege’s Grundgesetze der Arithmetik at the
turn of the twentieth century, it has been clear that a basic assumptions about
sets is extremely plausible and yet inconsistent. This is the assumption that a
set is an object exactly composed of all and only the things sharing a common
property; Frege expressed this as

Basic Law V The set of all As is identical to the set of all Bs if and only if all
As are Bs and all Bs are As
This sounds very close to being some kind of trivially true or analytic state-
ment, and indeed Frege intended it to be a purely logical truth. The law can be
expressed as two principles:
Comprehension There is a set {x : A(x)} of all and only the xs that are A
Extensionality Sets that have all and only the same members are identical
As Russell said to Frege, it is alarmingly easy to find a contradiction here. Let’s

go through the reasoning (a simplification from §1.1 discussed earlier) to see
how naturally a contradiction presents itself to mathematical inquiry.
Some sets are members of themselves (like the universe) and some are not
(like the set of all natural numbers). Consider then the subset of the universe
consisting of all and only the sets that are not members of themselves, R = {x :
x < x}, the Russell set, which exists by comprehension. If R is a member of
itself, then by definition, it is not – so by reductio, this proves that R is not a
member of itself. But that is exactly the defining property of the Russell set, so
this proves that R is a member of itself. Thus,
R ∈ R and R < R.
This is a contradiction that appears to follow from very simple reasoning from
true premises.
That is not all. By extensionality, sets with the same elements are identical,
and conversely, sets with different elements are not identical. Since we have
shown that the Russell set has different members from itself, that means
R , R.
But then, if we recall the definition of the universe U = {x : x = x} from

above, then R < U. (It also is a member of U, assuming that everything is
self-identical; then it is still (also) the case that R = R.) Since R is still a subset
of U, this shows
U , P(U)
since U differs from P(U) with respect to membership. contradicting our

result (3) from Section 1. It also shows that U , U and P(U) , P(U),
or at least so it would seem on some natural (?) informal reasoning.
If inconsistency is to be avoided, either comprehension or extensional-

ity must be weakened; and not only that, but some of the more natural

consequences of comprehension must be blocked: to prevent reemergence of

the Russell set, standard consistent set theory teaches that there is no univer-
sal set (Halmos, 1974, p. 7). Further, on this sort of solution, sets do not have
unrestricted complements, a set of all their nonmembers – since without a uni-
verse there is nothing for them to have complements “into.” Although most set
theorists circa 1900 did not find this possibility very attractive, the considered
orthodox view about the set theoretic antinomies is that comprehension is false.
Not all predicates determine a set.
But whatever one thinks of the Russell set, so-called naive intuitions balk at
restrictions to the comprehension axiom. No universe of all sets? Then what is
the domain of investigation here – what is set theory even about? No unre-
stricted complements? Then what does the law of bivalence, that for every
x either it is A or not, even mean? And even if one accepts that these ques-
tions have counterintuitive answers, there is still no guarantee that consistency
has been achieved (indeed, there cannot be any such guarantee, ever, accord-
ing to Gödel; see Chapter 3, introduction); there is always the ongoing risk of
contradiction even in orthodox set theory.
What if, though, objects like the Russell set or the universal set are not so
much problems to be solved as structures to be investigated? What if a mathe-
matical study of these structures leads to new intrinsic insights and extrinsic
connections to other mathematical questions? Tending to the antinomies in
the foundations of mathematics is a strong motivation for both moderate and
immoderate approaches to paraconsistency: shoring up “classical” set theory
to withstand any future contradictions, understanding its logic, and expanding
past it to make an unapologetic study of the absolutely unrestricted universe of
sets.
2.1 In Search of Paraconsistent Set Theory

The great appeal of a paraconsistent approach is that Frege’s highly intuitive
principle, which was also assumed by Cantor, Dedekind, and other mathemati-
cians, may be restored (or something very close). Insofar as set theory is the
theory of collections, the naive paraconsistent set theorist can say that every
collection whatsoever is a set.
A naive set theorist also wants to preserve extensionality. We have in effect
just seen how flare-ups from comprehension might intuitively find their way
down into extensionality, in striking non-self-identities like U , U. One might
even be tempted to say that a statement like a , a” is a nice way to characterize
the inconsistency of a. But any would-be inconsistent set theorist needs to think
this through.

For Frege, the extension of a set (the “list” of all its members) automatically
determines its anti-extension (the “list” of all its nonmembers). Classically, the
claim that the set of As comprises of all and only the As is equivalent to the
claim that the set of non-As is comprised of all and only the ¬As. On a para-
consistent account of negation, though, this might no longer be automatic. The
extension of a predicate may be independent of the anti-extension of that pred-
icate, if contradictory membership is allowed. Should membership in the As
immediately entail nonmembership in the non-As?
Questions like this start to show that any details of a paraconsistent set theory
will depend very much on how the axioms are formulated. In the question just
posed, it is a matter of whether the biconditional in the axiom of comprehen-
sion contraposes34 or not; and here, the approaches of various paraconsistent
mathematicians have differed. So, just as there is no one paraconsistent logic,
there is no one paraconsistent set theory to describe to you in general terms; we
will need here, as in subsequent sections, to look at specific systems.
Let us follow a quasi-historical narrative, where a key takeaway is that,
whether moderate or radical, anyone who thinks paraconsistency is the “easy
way out” is wrong. Anyone looking for a miracle whereby we get a set the-
ory with all of the virtues and none of the vices will need to keep looking. As
Arruda and Batens already put it in the early days, “There can be no doubt that
the expectations of paraconsistent logicians with respect to set theory did not
come true (Arruda & Batens, 1982, p. 132).”
Some of the first work in paraconsistent set theory was done by da Costa
and Arruda going back to the early 1960s.35 Recall that the idea in da Costa’s
logics is to have an operator A◦ (now writing the ball as a superscript, following
the original notation) meaning that A is consistent; iterating, A◦◦ means that it
is consistent that A is consistent, and An abbreviates A◦...◦ where ◦ appears n
times. Then A(n) abbreviates A1 ∧ . . . ∧ An . This can even be iterated infinitely
many times. The propositional language of da Costa’s Cn , for 1 ⩽ n, takes
∧, ∨, ¬, and ⊃ as primitive; postulates are as follows.
1. A ⊃ (B ⊃ A)
2. (A ⊃ B) ⊃ ((A ⊃ (B ⊃ C)) ⊃ (A ⊃ C))
3. A ⊃ (B ⊃ A ∧ B)
34 That is, satisfies: if A then B, then if ¬B then ¬A. For a good account of different approaches to
models of paraconsistent set theory, see Libert (2005).
35 For recent scholarship on Arruda and her contributions to logic, see Lisboa and Secco (2022).
Arruda suggests that paraconsistency will “only subsist as far as professional mathematicians
occupy themselves with its problems” (Lisboa and Secco, 2022, §3, note 22).

4. A ∧ B ⊃ A, A ∧ B ⊃ B
5. (A ⊃ C) ⊃ ((B ⊃ C) ⊃ (A ∨ B ⊃ C))
6. A ⊃ A ∨ B, B ⊃ A ∨ B
7. ¬¬A ⊃ A
8. A ∨ ¬A
9. B(n) ⊃ ((A ⊃ B) ⊃ ((A ⊃ ¬B) ⊃ ¬A))
10. A(n) ∧ B(n) ⊃ ((A ⊃ B)(n) ∧ (A ∨ B)(n) ∧ (A ∧ B)(n) )
along with the rule of modus ponens, that from A and A ⊃ B we can derive B.
Postulates for quantifiers include standard axioms and rules, plus consistency
shifts like
∀x(A(x))◦ → (∀x(A(x)))◦
(If for everything it is consistent that A then “everything is A” is consistent.)

The logics developed by da Costa are paraconsistent, and so can handle, for
example, Russell’s paradox without explosion; so one might hope that they
can accommodate full comprehension.36
It was found, though, that simply being paraconsistent is not enough. This
is because of further paradoxes proven by Curry in the 1940s (also sometimes
attributed to Moh Shaw-Kwei in 1954). In set theoretic terms, consider the set
C = {x : if x ∈ x then p},
where p can be any sentence whatsoever. Suppose C ∈ C. Then the set’s defin-
ing condition gives us that, if C ∈ C then p. On assumption, then, p. That is
unacceptable, since p could be anything, so C ∈ C cannot be – but it looks like
we just proved that, if C ∈ C, then p, which is exactly what it would take for
C ∈ C to be true. So we are in trouble, and it looks like negation had nothing
to do with it.
For this reason, many even paraconsistent systems cannot accommodate the
full comprehension principle, because of Curry’s paradox.37 Arruda showed
that set theories based on da Costa’s systems under various formulations lead
to triviality, or at least something very close – the “unpleasant result” that there
is only one object, ∀x∀y(x = y). She and da Costa considered set theories based
on different logics, including logics without modus ponens, but similarly found
36 Here, I am mainly following Arruda (1989) (written in 1978; cf. Arruda (1979)). Da Costa ini-
tially worked with versions of Quine’s set theory NF, the topic of Arruda’s 1964 PhD thesis; she
eventually showed NF1 to be trivial in an unpublished talk from 1981 (D’Ottaviano & Carvalho,
2005, §2; Lisboa & Secco, 2022, §5).
37 One diagnosis is that postulate (2) for the conditional connective(s) in Cn is a form of
contraction; see §2.2.

that while these are nontrivial, that in them “every two sets are identical,” which
is bad.
Nevertheless, these investigations might be seen as some of the first bread-
and-butter mathematical work in this area. Since the logics in question are
paraconsistent and hence could presumably handle the Russell contradiction,
but seem to run into “unpleasantness” in the presence of a universal set, Arruda
came to wonder about the relationship between the Russell set and a universal
set, and in particular, whether it is possible to have one without the other. Under
more classical approaches, no sets are self-membered, so the Russell class is
the universe of all sets; but perhaps in a nonclassical setup matters could be
different?
Arruda shows that under relatively minimal conditions they cannot be kept
apart. Here is her original result, that the union of the union of the Russell set
is universal; see Arruda and Batens (1982).
∪∪
Theorem R = U.
Proof. We suppose two facts as given. One is about singletons:
• if x, y ∈ {z} then x = y.
This follows from defining a singleton as {x} = {z : z = x} and identity facts.

We also presume that the union of sets X ∪ Y is all the members of X or Y,
∪
and that the union of a set X, written X, is comprised of the members of the
∪
members of X. That means that if {z} ∈ X then z ∈ X.
Now, since the logic has argument by cases (postulate 5) and the LEM (postu-
late 8), we may argue as such. As a lemma, we wish to show that {R ∪{x}} ∈ R.
So suppose on the contrary that {R ∪ {x}} < R, that is, {R ∪ {x}} ∈ {R ∪ {x}}.
Then
R ∪ {x} = {R ∪ {x}} (4)
by our fact about singletons. But from this identity and the fact that the Russell
set is a member of itself, we have R ∈ R ∪ {x}. So by 4 again, R ∈ {R ∪ {x}}
and hence by singleton facts,
R = R ∪ {x}. (5)
By that identity and the fact that the Russell set is still self-membered, we get
R ∪ {x} ∈ R, and so
{R ∪ {x}} ∈ R
using 5. That proves the lemma.

∪
But then, R ∪ {x} ∈ R. Then if z ∈ {x}, z ∈ R. And therefore
∪∪
∀x(x ∈ R)
so the union of the union of R is universal. □
There are several questions this exercise raises. We have gone through it
because this is an early and (relatively) isolated example of seriously think-
ing through some basic properties of an inconsistent object, and trying to draw
sensible conclusions about its structure. This is paraconsistency in mathemat-
ics.
∪
Arruda and Batens then improved the result to show that U = R. Thus,
it seemed, a set theory with a Russell set would also have a universal set, and
then, depending on the other set existence axioms and the underlying logic,
be subject to Curry’s paradox. Still, the paraconsistent opportunity to study,
for example, the universal set, was there, and so it made more sense for da
Costa and those who followed to study paraconsistent versions of axiomatic set
theories that assume principles weaker than comprehension, but that can control
the existence of the Russell set or the universal set as additional axioms.38
At this fork in the road, then, one path points away from “naive” set theory
and considers instead paraconsistent versions of standard axiomatic set theo-
ries, like Zermelo–Fraenkel set theory, the theory proposed starting in 1908
to replace naive set theory, usually denoted ZFC (to include the axiom of
choice).39 This is a much more moderate project, in which we are dealing with
(presumably) consistent assumptions and a consistent theory, but nevertheless
embed it in a contradiction-tolerant logic – perhaps to demonstrate that the full
force of classical logic is excessive for the purposes of even standard mathe-
matics. As Carnielli and Coniglio put it, “Paraconsistent logic tries to find the
38 Da Costa proved nontriviality for a paraconsistent set theory in da Costa (1986). For more
see da Costa, Krause, and Bueno (2007) in Jacquette (2007) and (Carnielli and Coniglio
(2016, ch.8).
39 For reference, informally, the axioms of ZFC are, along with extensionality:
Pairing For any sets x, y there exits their pair z = {x, y}.
Union For any set there exists the set of all members of its members.
Powerset For any set there exists the set of all its subsets.
Separation For any set, there is a subset of it consisting of all and only the As.
Infinity There is an infinite set.
Replacement Every function with a domain also has a range.
Foundation No sets have infinitely descending membership chains.
Choice For every collection of non-empty sets, there is a function that chooses a member from
each.
Note that separation is a restricted form of comprehension, and would lead to Russell’s paradox
if the universe is a set. The standard reference is Jech (2003); cf. Bell (2005, p. 17).

minimal natural principles that are capable of permitting us to reason in generic

circumstances, even in the undesired circumstances of contradictions (Carnielli
& Coniglio, 2021, p. 189).”
Under one such approach, using variants of LFIs, one can still study possible
characterizations of inconsistent and “non-consistent” sets, with a consistency
predicate C(x) and two different negations, so that, for example, a set x is not
consistent if it is non-self-identical or self-membered, and as is characteristic
of this approach, a kind of gentle explosion is valid. For example, while x ∈
x, x < x ⊬ B for arbitrary B, because of paraconsistency, it is true that
C(x), x ∈ x, x < x ⊢ B
providing a means of internalizing the notion of consistency (Carnielli &

Coniglio, 2016, p. 353). It therefore also faces a similar ceiling as Arruda
found for da Costa’s approach: that there cannot be a universal set, else classi-
cal ZF itself would prove the existence of a universal set. Here, the result can
be expressed as saying that if the set
cons = {x : C(x)}
existed (the set of all consistent sets), then C(cons) would be true and hence
cons ∈ cons, which is trivializing, given other ZF assumptions.
On a similarly ZFC-directed tack, a detailed study of models of paracon-
sistent ZFC is being undertaken using an algebra-valued approach, by Sourav
Tarafder, Giorgio Venturi, and Santiago Jockwich-Martinez, following Löwe
and Tarafder (2015); see Jockwich-Martinez and Venturi (2021). The idea here
is to take the Boolean-valued models of set theory developed for forcing and
independence proofs (see Bell, 2005) but replace the Boolean algebra with
one(s) more amenable to paraconsistency. These efforts, based in classical alge-
braic methods, are some of the most active in the field at time of writing, and are
especially promising ways of building connections between paraconsistency
and more mainstream work in set theory.
Using adaptive logic, Verdée has developed paraconsistent set theories in
Verdée (2013a, 2013b) and argued that in the spirit of the adaptive approach
they can provide a “pragmatic” foundation for mathematics. Returning to the
motive of restoring naive set theory, Batens (2020) outlines some desiderata
for a naive set theory to be “Fregean”: it matches, at least in terms of symbols,
the axioms of Frege’s system; it does justice to the notions of membership,
nonmembership, identity, and nonidentity, and most characteristically for the
adaptive approach, “if a set (or sentence) can be selected as consistent” for a
“systematic or formal reason, then the set (or sentence) is consistent.” So to

balance the desire to maintain Basic Law V and the inevitable contradictions
that result, “the natural reaction is to allow for contradictions but to minimize
them (Batens, 2020, p. 912).”
The hope is to recover an inconsistent but coherent naive set theory that
“serves the purposes a decent set theory should serve” which on this approach
means including a consistent set theory. This would be, as Batens calls it, the
“adaptive gain.”
As with other attempts at paraconsistent set theory, Verdée found that the
adaptive approach runs in to trouble from Curry’s paradox. This leads to a
choice between developing adaptive set theories that do not entirely meet
Batens’ Fregean desideratum, or else to give up any “adaptive gain.” Batens
suggests as a solution defining a cumulative sequence of abnormalities leading
up to a “sequential superposition” called AFS1 (adaptive Fregean set theory 1),
which would have many of the desired properties. As Batens admits, though,
a proof that “there is an adaptive gain in AFS1 may have the same complexity
as a proof that ZFC is consistent” and so this theory is presented circa 2020 at
the level of a hypothesis or program for further investigation.
To avoid Curry’s paradox, among other problems, one option touched on by
da Costa and Arruda as mentioned is working with systems that do not obey
modus ponens (Arruda, 1989, p. 112). Without a “detachable” conditional, as
it is sometimes said, then there is no worry from Curry’s paradox or the like.
There is indeed a very natural paraconsistent logic with this feature, namely,
LP that we met in Section 1.2.1. Asenjo considers using this “antinomic logic”
for recovering naive set theory, but states that “Cantor’s axiom remains for us a
paradise lost, inaccessible to our most earnest efforts” (Asenjo, 1989, p. 399).
Nevertheless, by 1991, Greg Restall had made some strides in this direction,
which we now consider.
With no workable notion of modus ponens, there is not much hope of
using LP set theory to prove theorems, at least not in the “natural” deductive
way that Arruda’s abovementioned theorem is derived. But models of LP are
easy to devise and work with, and there has been ongoing efforts to under-
stand paraconsistent set theory in this way. Here is how models of (naive)
set theory in LP work, following (Restall, 1992, p. 422; cf. Priest, 2006, p.
223). It is worth noting that this is an entirely classical setup for modeling a
theory.
Given a first-order langauge with ∧, ∨, ¬, ∀, ∃, ∈ and variables x, y, z, ..., an LP
model structure consists of a non-empty domain of objects and an interpreta-
tion such that, for each a, b in the domain, the value of a ∈ b is either true or
false (or both). Variables x, y of the language are assigned objects ax, by in the

domain, and the truth value of x ∈ y is exactly that of ax ∈ by . Truth values are
then assigned recursively as in §1.2.1 earlier for LP, now with clauses for the
quantifiers:
∀xA(x) is at least true in a model iff A(a) is at least true

of every a in the domain
∀xA(x) is at least false in a model iff A(a) is at least false
for some a in the domain
The existential quantifier will have dual properties, so that ∃xA(x) is ¬∀x¬A(x).
Naive set theory in LP is obtained by taking every instance of comprehen-
sion40
∃y∀x(x ∈ y ≡ A(x))
as an axiom (where y does not appear free in A) and extensionality
∀z(z ∈ x ≡ z ∈ y) ⊃ x = y
as an axiom, where identity x = y is defined as ∀z(x ∈ z ≡ y ∈ z), that is, that

x is coextensive with y. Then A is a theorem of LP naive set theory iff every
LP model structure and evaluation that makes every instance of comprehension
and extensionality (at least) true will also make A (at least) true.
There are models of this theory – quite simple ones, in fact. Consider a
domain with just two elements a and b and the following assignment:
M1
a ∈ a is both true and false a = a is both true and false
a ∈ b is false a = b is false
b ∈ a is both true and false b = a is false
b ∈ b is true b = b is both true and false
This is a model of naive set theory in LP (Restall, 1992, Lemma 3, p. 425).

With this as a setup, some preliminary facts can be shown. In every model,
as an instance of comprehension, there is a Russell set,
∀x(x ∈ R ≡ x < x)
Then instantiating the universal quantifier, we have R ∈ R ≡ R < R, and

then expanding the definition of the material biconditional, this is just a way
of writing R ∈ R ∧ R < R as earlier. So the theory is inconsistent. And not
just that; in every model structure, it will be true that ∃z¬(z ∈ R ≡ z ∈ R) (the
40 Recall from §1.2.1 that ≡ is the material biconditional.

Russell set is itself a witness for this claim) and so by the definition of identity,
R , R, which proves as a corollary that
∃x∃y(x , y)
One can also show that
∃x∀y(y ∈ x) and ∃x∀y(y < x)
are satisfied in every LP model: that is, there is a universal set and an empty
set. The basic idea of the proof is that, if there were no empty set, then the set
of all empty sets would be empty; and if there are no universal sets, then the
set of all nonuniversal sets is universal.41
The question arises of what, if anything, is not a theorem of this set theory.
It turns out that the (classical) axiom of foundation,
∀x(∃y(x ∈ x) ⊃ ∃y(y ∈ x ∧ ¬∃z(z ∈ y ∧ z ∈ x))),
which implies that no sets are self-membered, is not a theorem. Indeed, the little
model M1 is enough to provide a counterexample, since element b is not empty
but has no “least” member.
So the theory is inconsistent but not trivial – a paraconsistent theory. One
can go on to show that the axioms of Zermelo–Fraenkel set theory (without
foundation) are all theorems. For example, the axiom of infinity says that there
is a set containing the empty set and closed under a notion of successors (y+
the successor of y),
∃x(∅ ∈ x ∧ ∀y(y < x ∨ y+ ∈ x)).
This holds simply because the universal set is a witness: the universal set is
closed under . . . everything. So the first conjunct and the second disjunct are
automatic. We already saw in §1.1 how, for perhaps more intuitive reasons, the
existence of a universal set is already the existence of an infinite set.
On this setup, while the classical axioms of ZF (minus foundation) are the-
orems, it is not the case that all of their classical consequences are too. Indeed
it is unclear if much at all of real mathematical use can be extracted, since LP
is so weak; see Thomas (2014b). Undaunted, Priest has pursued increasingly
sophisticated methods for manipulating these models and found that there is at
least one nontrivial model of LP set theory that satisfies every theorem of ZF,
too (Priest, 2006, p. 257; 2017b, sec.11). Thus, it is possible that set theory in LP
41 To work in LP, these ideas need to be expressed as statements about models; see Restall (1992,
p. 426).

subsumes all of classical mathematics, though this is more of an outer-marker

than a full recapture.
Bridging the work done on LP set theory with ideas from da Costa’s origi-
nal investigations, Hitoshi Omori has shown that a consistency operator ◦ can
be added to the language of LP, and with it a classical negation can be defined
(Omori, 2015). The resulting theory is still nontrivial due to the use of the mate-
rial (bi)conditional in the formulation of comprehension. This is significant
philosophically because elsewhere Priest has claimed that the very existence
of classical negation must be rejected by a glut theorist, and mathematically
because Omori’s strengthened language can be used to define a conditional
that obeys modus ponens and so can do more theorem proving work, perhaps
given further axioms (again as in da Costa’s approach). Cf. Priest (2017b, p.
86).
Before the would-be paraconsistent foundationalist breaks out the cham-
pagne, it should be noted that LP set theory has further serious drawbacks.
As evidence, Weir (2004, pp. 393–396) points out that it has a one element
model,
M2
a ∈ a is both true and false a = a is both true and false,
which also satisfies all of ZF – at least, ZF when phrased in the language of

LP. This has echoes of Arruda’s finding that certain natural assumptions lead
to every “two” elements being identical. Now, truth in a one element model
is not a proof; a single possible interpretation is only that. (Which also means
Priest’s single model showing ZF contained in LP set theory is only just that,
too.) But such a blunt, structureless interpretation is a confirmation that such an
“unpleasant” result is at least not ruled out by LP models, and raises doubt about
whether the language of LP alone can express much of mathematical interest,
since for all we know it is talking about the “structure” of a single point.
Even if one is not concerned with the existence of “pathological” models –
after all, classical ZFC has a countable model even though it talks about
uncountable sets – the fact that LP does not have modus ponens suggests that
its various instances of comprehension only give the appearance of their stand-
ard mathematical meanings, but are not useful in practice. It is not possible to
reason inside of LP; we can only reason “over” it using (classical) model the-
ory; so it is hard to see how a foundationalist will find it satisfactory; and even
overlooking that issue, it appears that the space of possibilities allowed by LP
set theory, like the logic itself, is just too wide for most mathematical interests.

2.2 Relevant Set Theory

Since the start of investigations into paraconsistent mathematics, it has seemed
to many that more robust theories will require a conditional that obeys modus
ponens. Indeed, much of the work in paraconsistency was driven primarily
by the search for a decent theory of the conditional, with the idea of using
it for inconsistent mathematics as an offshoot. In the late 1970s, Richard Rout-
ley/Sylvan proposed what he called “ultralogic” as a kind of “ultramodal”
approach to conditionals using worlds semantics – and most crucially, models
including “inconsistent” worlds, so that an undesirable law like
(A ∧ ¬A) → B
will fail because there are points in the model where the antecedent holds
without the consequent. As in §1.2.4, logics in the family that Routley was
considering are now called relevant logics. The application proposed by
Routley, and then pursued especially by Ross Brady, is relevant set theory.
It “meets the paradoxes head-on, taking the paradoxical arguments as proofs
and the contradictory conclusions as holding in the theory (Routley, 1977, p.
912).”
So far, the set theories we have seen cannot accommodate Curry’s paradox,
or can only do so at the cost of having no conditional. A well-investigated
diagnosis here is to say that any conditional operator in comprehension cannot
obey contraction, the general principle that two uses of a premise are the same
as one. Curry’s paradox needs contraction to go through. In some weak relevant
logics, though, the principle of contraction is invalid, in the form
p → (p → q) ⊬ p → q.
This makes it possible to face Curry’s paradox more directly. We can have full
comprehension, but – as has been the paraconsistent plan all along – it must be
reasoned about more carefully.42
One such relevant logic without contraction is, for not very good reasons,
called DKQ and is given by closure of the following axioms and rules.43
42 Matters do not end there. While the classical approach drops comprehension and keeps exten-
sionality, keeping comprehension puts pressure on extensionality, as further Curry-esque
paradoxes show. Other nonclassical set theories keep comprehension but drop extensionality. It
appears that relevant set theory is highly unusual in that it can keep both axioms, but see Øgaard
(2016) and Weber (2021, ch. 4).
43 It is a very close relative of the logic DLQ, or dialectical logic with quantifiers, introduced in
Routley and Meyer (1976). It is also a very close relative of the logic DJQ, studied by Brady
(2006). As it extends the “extensional” connectives of LP, we define A ∨ B := ¬(¬A ∧ ¬B) and
again ∃xA(x) := ¬∀x¬A(x).

A1 A→A
A2 (A → B) ∧ (B → C) → (A → C)
A3 A∧B→A
A4 A∧B→B
A5 (A → B) ∧ (A → C) → (A → (B ∧ C))
A6 A ∧ (B ∨ C) → (A ∧ B) ∨ C
A7 ¬¬A → A
A8 (A → ¬B) → (B → ¬A)
A9 A ∨ ¬A
A10 ∀xA(x) → A(a)
A11 ∀x(A(x) → B) → (∃xA(x) → B) (x not free in B)
Rules:
R1 If A is a theorem and A → B is a theorem then B is a theorem

R2 If A and B are theorems then so is A ∧ B
R3 If A → B and C → D are theorems then so is (B → C) → (A → D)
R4 If A is a theorem then so is ∀xA.
To obtain set theory, one then adds to this extensionality,
∀x∀y(∀z(z ∈ x ↔ z ∈ y) → ∀z(x ∈ z ↔ y ∈ z))
and every instance of the axiom of comprehension,
∃y∀x(x ∈ y ↔ A(x))
all now suitably phrased with a robust relevant conditional. The extensionality
axiom states that if sets have the same members, they may be intersubstituted.
The comprehension axiom has no restrictions: every property whatsoever
determines a set. That means one very striking example is worth observing.
There is a set that is comprised of all and only the members that are not members
of that very set,
∀x(x ∈ Z ↔ x < Z)
That is, there is a set identical to its own complement. This extension of this
set completely coincides with its anti-extension. This is the Routley set, since
it was observed in Routley (1977) (although it may have been spotted earlier).
It is, as Routley said, a “completely bizarre” set, but there it is – a highly novel
object for mathematical investigation.
Less flamboyantly, on this approach several of the results we saw earlier
from LP set theory can be re-obtained via axiomatic deductions rather than
model manipulations, and improved. There is an empty set, a universal set, and

a Russell set. All the axioms of ZF (but still not foundation) are derivable, as
are some of their most elementary consequences.44 How many consequences
is not known precisely, although it seems clear that the answer will not be “all
of classical ZF”; but it is also clear that this can prove more than ZF (later).
Routley went as far as to attempt to derive the axiom of choice as a theorem,
which was rather audacious considering choice is known to be independent in
classical set theory.45
Beyond the axioms, one would like to know what standard results can be
derived. In particular, the centerpiece of set theory is the theory of ordinal num-
bers, including the finite ones that ground arithmetic, and also the transfinite
ones that keep track of processes beyond:
ω
..
ω ωω ω.
0, 1, ... ω, ω + 1, ... ω , ... ω , ... ω
2
, ... ω , ...
To show that the ordinals do their job, one proves that the set of all ordinals, Ω,
is structured in the right sort of way, where the ordinals form a single straight
line and any non-empty subset of ordinals has a least member. This can be done
in relevant set theory, given some suitable definitions – where each ordinal α
is the first one after all those before it, α = {β : β precedes α}. Strikingly,
though, once these properties are established it also establishes that Ω itself is
an ordinal,
Ω∈Ω (6)
since it is the first one after all those before it, Ω = {α : α precedes Ω}. But,
by the structure of the ordinals, no ordinal is self-membered,
∀α(α ∈ Ω → α < α) (7)
and so by (6) and (7),
Ω<Ω (8)
a contradiction. This was discovered by Burali-Forti in 1897 and perhaps even

more than the existence of a universal set or Russell set gives a mathematical
reason to think set theory might truly be inconsistent.46
The existence of inconsistent Ω can then be leveraged, rather easily, to prove
many of the “large cardinal axioms” that set theorists have considered adding to
ZFC to prove more results. Such axioms assert the existence of large sets with
44 See Weber (2010).

45 See editor’s introduction to Routley (1977).
46 This version of the Burali-Forti paradox is based on von Neumann’s notion of ordinals. Other
versions can be put in terms of order types of well-ordered sets or give the contradiction as
Ω = Ω + 1. See Moore (1982). Cf. Shapiro and Wright (2006) in Rayo and Uzquiano (2006).

various properties that are thought to “reflect” properties of the classically inef-
fable universal set. Proving they exist in relevant set theory is straightforward,
since Ω is the largest number; so it has all the properties one could dream of
for a large number.47
So we have theorems, some of which are also false. Which sentences are the
non-theorems? The language of relevant set theory is now substantially more
expressive, so that while in LP there is no sentence that cannot be satisfied by
some model (that is the point of Weir’s one element counterexample) in DKQ
set theory an “absurd” formula may be given:
⊥ := ∀x∀y(x ∈ y).
This formula is absurd because ⊥ → A is true for any A. (Just let y = {z : A}.)
So there are some useful theorems here, as well as some absolutely unsatisfiable
sentences.
Perhaps the most important result to date in paraconsistent set theory is Ross
Brady’s proof that naive set theory in a weak relevant logic has nontrivial mod-
els; the axioms of comprehension and extensionality, and their consequences
in relevant logic, can be true without sliding into absurdity. This proof was
devised in the late 1970s and has been refined continuously since.48 He con-
structs a classical model, M, which satisfies all the theorems of naive set
theory, but in which at least one sentence (⊥) is not satisfied. Because of the
importance of this result I will go through its main ideas, with details in an
Appendix.
The idea of the proof is to give a persistence argument, considerably extend-
ing Kripke’s fixed point method. Kripke’s construction is one-dimensional, a
straight line from a “ground” model to a fixed point. Brady takes the fixed
point at the end of the construction and uses it as the ground model for a
two-dimensional construction. A model is determined, made up of a transfi-
nite sequence of models, leading to the construction of a transfinite sequence
of transfinite sequences. Truth values for propositions are then shown to be
preserved along the sequence. An inductive argument is carried out on this
structure to show that it is a model of naive set theory, but that some sentences
fail to be true in the model, in particular, ∀x∀y(x ∈ y) and a few others; see
Appendix.
Philosophically, what does this proof show? It is a relative consistency proof.
It is not, I think, an attempt to show what the universe of relevant set theory
47 See Weber (2012).

48 See Brady (1971, 1989, 2006). It should be noted that Brady’s preferred logic DJQ in his 2006
book allows a consistent naive set theory and he does not endorse inconsistent set theory.

is “really” like. The proof is instrumental, showing that, if one believes that
classical set theory is consistent, then relevant set theory can be represented
within it, and given a nontrivial classical model. For the vast majority of people,
who believe in classical set theory, this shows the relative reliability of Routley-
Brady set theory. But what about someone who casts (extreme and vitriolic)
doubt on aspects of classical set theory – someone like, say, Routley? The point
is not lost on Brady:
It may seem ironic that hierarchies are being used to prove a result which is
designed to eliminate certain hierarchies. However, once our result is estab-
lished, hopefully one can chose a weaker and non-hierarchical logic to work
in. (Brady, 1989, p. 433)
In hoping for a “non-hierarchical logic,” Brady seems to be nodding to Rout-

ley/Sylvan’s vision of an untyped universal logic (§1.2.4). Decades later, this
hope has not been fulfilled. We will return to the question.
Setting aside such philosophical worries, Brady’s proof gives us classical
reassurance for a reasonably strong paraconsistent basis for mathematics. Is it
time for the paraconsistent set theorists to break out the champagne yet? This
approach is still limited in what it can “recapture” from standard mathematics.
There are problems with induction (both transfinite and for denumerable sets),
counting, and function symbols, among other challenges. So while relevant set
theory is “proof of concept” that there is an expressive and nontrivial theory of
naive comprehension, there is still (mostly) a lot of work to do in improving.49
2.3 Cantor’s Theorem and Infinity

What is the infinite? Aside from foundationalist questions, set theory is the
mathematical study of infinity. Set theory tells us there are transfinite orders
of infinity. And we have seen how a paraconsistent set theory in one sense has
very easy access to infinity – all the way to the “absolute” infinity at Ω. But it
is so easy as to be a worry. Yes, there is a universal set and it is (automatically)
infinite; but if as in LP set theory the universal set could also be a single point,
what does that really mean?
Cantor famously proved the existence of transfinite numbers, beginning with
his proof of the uncountability of the real numbers in 1874. We met the idea
of his proof in §1.1. An important aspect of a paraconsistent reconstruction
49 For very reasonable reasons to doubt recapture, see Thomas (2014a). There is also the ever-
present worry about Curry’s paradox. Contraction for → is avoided, but the theory still features
structural contraction (if A, A ⊢ B then A ⊢ B) and then runs into trouble if the theory can
internalize its own notion of validity. For more on this, and for my best, and still incomplete,
attempt with this, see Weber (2021, chs. 4, 5).

of set theory is to secure some of the cornerstone theorems of the subject –

and Cantor’s theorem would be the cornerstone par excellence. Let us focus
on the more abstract version of Cantor’s theorem: that there is no one-to-one
correspondence between a set and its powerset. How does this theorem fare in
a paraconsistent framework?
An immediate apparent obstacle is that the standard proofs are all by reduc-
tio: one assumes that there is a bijective mapping f between some X and P(X),
and arrives at a contradiction. By reductio, there is no one-to-one mapping
between a set and its powerset, so the latter is “bigger.” But without say-
ing more, in the paraconsistent case the argument could just show that such
bijections are inconsistent, not that P(X) is necessarily “bigger” than X.
This is more than an idle concern. There is a striking similarity between, on
the one hand, Cantor’s proof, and on the other hand, Russell’s paradox. (Indeed,
it was through studying Cantor’s proof that Russell arrived at his paradox.) If
we are prepared to accept the Russell set exists, rather than deny that there is
a closed universe, then why wouldn’t we be prepared to accept that there is a
contradiction, not a reductio, here too?
Interestingly, if we do accept a contradiction here, we can still prove Can-
tor’s theorem, at least up to the letter, by following out a version of Cantor’s
reasoning to the bitter end. Let us replace Cantor’s diagonal set with a subset
of X that is a little more lively,
Z(X) = {x ∈ X : x < Z(X)},
which is derived from the Routley set Z in §2.2. Every member of X is in Z(X)
iff it is not; but it must be one or the other, so it is both, as is now familiar. By
extensionality, for any non-empty X, then
Z(X) , Z(X) (9)
But then, we might appeal to a kind of “non-self-identity lemma”:
Lemma For any x, if x , x then for all y, x , y.
Proof : Suppose x , x. If x = y then y has all the same properties as x, including

not being identical to x. □
Using this lemma and 9,
∀z(z , Z(X)). (10)
But that means for any would-be bijection f that
∀z(f(z) , Z(X)). (11)

That proves the following.
Theorem (Cantor) For any set X and any mapping f from X to P(X), there is
a y ∈ P(X) such that f(x) , y for every x ∈ X.
Since this is true for any mapping whatsoever, a fortiori no bijection covers
P(X) after all, even allowing for inconsistency. Cantor’s theorem is still true,
even in the presence of inconsistency: recaptured.
Of course, this argument does not rule out that there also is a bijection. It can’t
rule that out because there is a bijection at the level of the universe U. This argu-
ment is not really a reconstruction of Cantor’s proof, so much as a new proof
using paraconsistency-specific methods. So in “recapturing” Cantor’s theorem
in the previous paragraph, what has been shown?
If you are familiar with other episodes in philosophical logic, consider, by
analogy, how Tarski responded to the liar paradox. He said that there must be
an infinite hierarchy of metalanguages, each to look down on the previous,
because if there ever were some flat or final level there would be a contradic-
tion. Cantor’s transfinite is in many ways the same response, but to Russell’s
paradox instead of the liar. Running with the analogy, perhaps the entire appear-
ance of transfinite set theory – the mathematics beyond infinity – is a way of
making classical sense of an inconsistency. There is “too much” information
in one place, reasons the classicist, so there must be “many places” for the
information to be spread across.
The paraconsistent mathematician can agree with this classical appraisal,
with the caveat that the many places are also just one inconsistent place. One
inconsistent place is many places, too. Far from saying that the inconsistent has
no structure, it turns out that the inconsistent has transfinite structure. Along
these lines, perhaps all sets are no bigger than the first place the contradiction
occurs, perhaps at the first (and last) transfinite ordinal, with the thought that
ω = Ω.
Maybe there is only one inconsistent infinity after all, which is what almost
everyone since Aristotle thought.50
Matters do not end there. What if the inconsistency (if there is one) occurs
even sooner, so that the last ordinal is some finite natural number? This will be
the topic of the next section.
50 On this “axiom of countability” see also Priest (2017a). As Scott memorably puts it when sur-
veying the vast array of incompatible models of ZF delivered by forcing methods, “Perhaps we
could be pushed in the end to say that all sets are countable . . . when at last all cardinals are
absolutely destroyed” (from the introduction to (Bell, 2005, p. xv)).

Appendix: Nontriviality Proof

Here is a sketch of Brady’s construction, proving that naive set theory in a
relevant logic is coherent. Note that this is an entirely classical proof. Lower
case greek letters α, β, κ, λ, ρ, τ . . . range over ordinals.
Let J·K be a function from sentences to three truth values t, f and now b (for
‘both’). So JAK = t means that A is true. At the start of Brady’s construction,
every atomic sentence is set to b, both true and false. Values are extended in
the following way.
∧ t b f ¬
t t b f t f
b b b f b b
f f f f f t
J∀xAK = t iff JA(a/x)K = t for every constant term a,

J∀xAK = f iff JA(a/x)K = f for some constant term a
J∀xAK = b otherwise.
These are the extensional connectives. Their values can be calculated at a point,
that is, if A is true at a point then ¬A is false at that same point.
To model the DKQ conditional →, which is not truth functional, we use a truth
functional conditional _. The truth table for it looks like this:
_ t b f
t t f f
b t b f
f t t t
This is the conditional from a logic called RM3. The DK conditional → is a modal
operator (at least in Routley–Meyer semantics) taking its meaning from varia-
tions across possible and impossible worlds; but it may be interpreted soundly
by the RM3 truth tables, making RM3 “the laboratory of relevant logic” accord-
ing to Meyer (according to Dunn, 1979, p. 81; cf. Priest, 2008, p. 125). Even
though many of the sentences that are true according the RM3 operator are not
true for naive set theory – for example, _ contracts, triggering Curry’s para-
dox, and so cannot simply be the conditional for our logic, still all DK theorems
are RM3 theorems.
In the model, values for a conditional at a point are determined by what
happens previously. A conditional is true at a point iff it is true (according to
the truth table) at all earlier points; it is false if there is a counterexample at
some earlier point. What truths persist to the limit, so to speak, are the truths.

Define a partial ordering ⩽ on structures as follows. For atomic formulas p,

M1 ⩽ M2 iff: if JpK1 = t then JpK2 = t
if JpK1 = f then JpK2 = f
This ordering is transitive (if M1 ⩽ M2 ⩽ M3 then M1 ⩽ M3 ) and antisymmet-
ric (if M1 ⩽ M2 ⩽ M1 then M1 = M2 ). A model M is made up of a transfinite
sequence of models,
{M0,λ0 , M1,λ1 , ..., Mκ,λκ }
So the models Mτ,λτ are fixed points, leading to the construction of a transfinite
sequence of transfinite sequences,
M0,0, M0,1, ..., M0,λ0 , ...

M1,0, M1,1, ..., M1,λ1 , ...,
..
.
Mκ,0, Mκ,1, ..., Mκ,λκ , ...
..
.
Truth values for propositions are shown to be preserved along the sequence,
thus establishing properties of the model M by induction. The proof proceeds
by a series of lemmas, starting at Brady and Routley (1989, p. 422) and Brady
(2006, p. 205). The first shows persistence.
Lemma 1 (Persistence) For all formulae A, if M1 ⩽ M2 for structures

M1, M2 , then if JAK1 = t then JAK2 = t, and if JAK1 = f, then JAK2 = f.
We define the standard wellorder on pairs of ordinals:
⟨α, β⟩ ⩽ ⟨γ, δ⟩ iff (α < γ) ∨ (α = γ ∧ β < δ)
Then the next lemma states that, if ⟨α, β⟩ ⩽ ⟨γ, δ⟩, then Mα,β ⩽ Mγ,δ .
Lemma 2 (Wellorder) For all α ⩽ β, Mτ,α ⩽ Mτ,β for all τ.
Lemma 3 (Fixed Points) For all sequences Mτ,0, Mτ,1, ..., Mτ,λµ , ... there is
a countable ordinal λτ that is a fixed point; that is,
Mτ,λτ = Mτ,β
whenever λτ < β. Further, some λτ is the first such fixed point. For sequences,
there is a first ordinal κ such that
Mκ,λκ = Mκ+1,λκ+1 .

A double induction, which is divided into five cases (each case itself having
multiple subcases) to cover successor and limit ordinals (Brady, 1989, p. 445),
then determines the model structure
M = {M0,λ0 , M1,λ1 , ..., Mκ,λκ },
with Mκ,λκ called the limit structure. Finally, we can define validity in M: a
formula A is valid in M iff JAK = t or b at the limit structure, and invalid in M
otherwise.
Theorem 4 The axioms of naive set theory are valid in M and all the rules
preserve truth.
Finally, we need that some formula outright fails in the model.
Theorem 5 The formula ∀x∀y(x ∈ y) is false in M.
Theorem 6 Naive set theory is nontrivial in the logic DKQ.
3 Arithmetic
It is not so hard to imagine that, somewhere in the vast transfinite towers of
Cantorian set theory, in some infinitely large sets that exceed even very immod-
est conceptions of infinity, there may be some inconsistency. But arithmetic is
the theory of the finite counting numbers 0, 1, 2, 3, . . .. It is harder to imagine
that there is any inconsistency in arithmetic. And yet it is here that some of
the deepest philosophical and technical investigations using paraconsistency
in mathematics have been made.
Hilbert’s program was a call in the first decades of the twentieth century to
show that mathematics is complete, consistent, and decidable, and to prove
it using finitary methods. Gödel proved in 1931 in his first incompleteness
theorem that for any consistent axiomatization of arithmetic, there is some sen-
tence (roughly, “this sentence is not provable”) that can be neither proved nor
its negation proved. In his second incompleteness theorem, he further showed
that no consistent axiomatization of arithmetic can prove itself to be consistent.
Gödel’s theorems showed that Hilbert’s program – as standardly understood –
is impossible.51 It turns out that both of Gödel’s theorems are fertile grounds
for paraconsistency.
While the standing wisdom on the first theorem is to accept that, as the name
suggests, arithmetic is consistent but incomplete, an alternative interpretation
51 See Franks (2009).

is that arithmetic may well be either potentially or actually inconsistent – it

looks, after all, very much like Gödel has himself proven an “unprovable” sen-
tence in an apparently arithmetically formalizable way; see §3.3. And while
standing wisdom on the second theorem says that we must ultimately accept
consistency on faith, not proof, an alternative result developed by Meyer sug-
gests that a coherent (inconsistent, but not trivially so) arithmetic can confirm
its own coherence after all. And while standing wisdom about Gödel’s theorems
in general is that there is no hope of a complete, effective, and reliable mathe-
matical foundation – seen today in the logical limits on what any computer can
do – paraconsistency calls us, as ever, to reconsider.
And if all of that sounds too heretical, there are some very even-tempered rea-
sons to consider this area. Dunn suggests that paraconsistent arithmetic might
be adopted on the basis of a kind of Pascal’s wager (Dunn, 1980, p. 408). If
you believe arithmetic is consistent, and you turn out to be wrong, there is hell
to pay; if you hedge your bets and use a paraconsistent arithmetic, then that
disaster is insured against, and if nothing too valuable is lost, then why not?
3.1 Inconsistent Arithmetic?

Asenjo was one of the very first paraconsistent mathematicians. He developed
an antinomic number theory in 1966 based on his calculus of antinomies from
1953.52 He introduces the rationale for his project in very strong terms.
Current opinion considers antinomies as unintuitive. This is a prejudice
which originates in attitudes that must be changed if we wish to honor the
facts. An antinomy can be as natural as the most ordinary physical state-
ment. . . . [R]eality itself is intrinsically antinomic. . . . An antinomic reality
can be better served by an antinomic logic, an antinomic mathematics, and
an antinomic science. (Asenjo, 1989, p. 394)
Asenjo jumps in with both feet, so to speak, and we will too.

In devising an antinomic account of numbers, Asenjo works with a strict
order relation < as the basic notion, with the envisioned contradictions being
something like
(n < m) ∧ (n ≮ m).
While inconsistency is at the level of propositions, then, he considered some

numbers to be antinomic: any j is antinomic if j < k and k < j for all k. This
means that, if j is antinomic,
j<j
52 Terminology in this area is, yes, inconsistent. Elsewhere, da Costa uses the term “antinomy” to
denote “bad” contradictions: “An antinomy implies triviality. A paradox is not in general an
antinomy” (da Costa, 1974, p. 498). This is not Asenjo’s (prior) usage.

He introduces two predicates into the language, “x is an ordinary number” and

“x is antinomic,” with the idea that every number is either antinomic or not,
none is both, zero is non-antinomic, and there is at least one antinomic number.
The intended model for this theory would be pictured as follows. We have the
standard (non-antinomic) natural numbers in a line, with the antinomic numbers
each having two locations on the line, to the left and to the right of the standard
numbers, each set of antinomic numbers the mirror image of the other.
. . . < x3 < x2 < x1 ... 0<1<2<3< ... x1 < x2 < x3 < . . .
The antinomicity of order manifests itself in the bilocation of each antinomic
number (Asenjo, 1989, p. 405).
This construction bears some (distant) resemblance to some nonstandard mod-

els of Peano arithmetic.
Asenjo goes on to consider identity in antinomic number theory and divides
numbers into three classes: those which are self-identical only; those which are
both self-identical and not; and those which are only not self-identical. With
suitable postulates, it follows that some inconsistency spreads to all numbers,
∀x∃y(x = y ∧ x , y)
“No object, then, is totally non-antinomic” [p. 408]. As a consequence, since

“every individual has its touch of antinomicity,” Asenjo determines that Leib-
niz’s principle of the identity of indiscernibles only holds over the “consistent”
numbers; the failure of substitution and the spread of inconsistent identity
would become major parts of other LP-based investigations decades later
(Priest, 2014).
Asenjo ends his 1989 study by reflecting on how attempts were received
in 1953 (at a mathematics seminar at the University of La Plata) with skep-
ticism, but how since then paraconsistency has become accepted as at least
legitimate, and that the future of paraconsistent mathematics will be decided
by its usefulness – especially
if by allowing no more than a touch of antinomicity basic mathematical sys-
tems can become complete, thus circumventing the devistating consequences
of Gödel’s incompleteness theorem. (Asenjo, 1989, p. 413)
Asenjo’s work is bold, creative, and ahead of its time. The development of
arithmetic in LP has been more recently and prominently studied by Priest.
Asenjo is forceful about the need for antinomic mathematics, but less clear on
specific justifications for it or how it would work. Priest tells the following
story.53
53 In papers Priest (1994a, 1994b, 1997, 2000) and (Priest, 2006, ch. 17). It appeals to finite models,
something we already saw for LP set theory in §2.1.

Consider a natural number n so large that, as Priest puts it, it has “no phys-
ical meaning or psychological reality.” Suppose it is a number that is simply
larger than anything that could be specified before the end of the universe. If we
suppose n is the least such number, this comes perilously close to something
like Berry’s paradox – giving a precise description of the first number too big
to give a precise description – and so we might entertain the idea that it is the
least inconsistent number. In particular, this number has the property that
n = n + 1.
If there is such a number, then we can pursue the intuition further. With N the
(infinite) set of natural numbers, consider the set
Nn = {m ∈ N : m ⩽ n}.
These are the (presumably) consistent numbers up to n. But since n is its own
successor, for any number k > n we might think of k as also being identical to
n. That is, in being the first inconsistent number, this is also (in some sense)
the last number.54 Anything true (or false) of any k > n is also true (or false)
of n. The picture might be illustrated like this:
In addition to looping at n, one might wonder whether the line also continues,
since all k > n are after all distinct, k , n. They are also not, k = n (and so n ,
n) but what precisely to make of this structure past the point of inconsistency
is up for further discussion.
Getting more precise: let N be the set of sentences in the language of first-
order arithmetic that are true about N, and Nn be the set of true sentences about
Nn . Some rather astonishing results follow (Priest, 1994a, p. 337).
The theory Nn , as governed by the logic LP, is inconsistent but not trivial.
It is inconsistent because it contains sentences that say some number both is
and is not its own successor. It is nontrivial because all the numbers m < n
are consistent, in the sense that if, for example, m is an even number, a sen-
tence saying m is not even is not in Nn . That is, N and Nn agree up to the first
inconsistency.
The theory Nn is complete and decidable. It is complete because every sen-
tence of N is a sentence of Nn : if A is a sentence about numbers before n then
both theories say exactly the same thing; and if A is a sentence about numbers
54 For connections to strict finitism, then, see van Bendegem (2003).

after n (which are all, really, just n itself) then Nn will at least agree with N
(although it might say more). The theory is decidable because the structure it
describes is finite. It is therefore also axiomatizable.
And the theory Nn can represent itself. It contains a truth predicate and its
own proof predicate. It proves its own soundness, its own nontriviality (by finite
means), and its Gödel sentence (“this sentence is not provable”) as well as its
negation.
These properties can be rigorously proved to hold in an LP-model of arithme-
tic, using a technique called collapsing.55 The rough idea is to take the standard
natural numbers up to n, and “collapse” the model after this, so that all subse-
quent numbers are interpreted as identical and not identical to n. To make this
precise and a little more general, consider some equivalence relation ∼ on the
numbers, and break the numbers into equivalence classes [n] = {m : m ∼ n},
giving a quotient algebra. Then a collapsed interpretation is obtained by, effec-
tively, treating any numbers in the same equivalence class as identical, an object
that has all the properties of all its members. Notably, nonidentities of members
of the equivalence classes are inherited by the classes themselves: if x , y and
x, y ∈ [z] then [z] , [z].
The key result from this construction is as follows.
Collapsing Lemma If M is a model and M ∼ is its collapse, everything true
(or false) in M is still true (or false) in M ∼ .
The proof is an induction on formulas, which turns on the monotonicity of
the connectives: once truth values go in, they never go out. It is important to see
that this would not work with classical negation, since adding more negation
facts to a theory would rule out previous ones – or destroy the theory.
A collapsed model can be an interpretation of a nontrivial but inconsistent
theory. This gives a neat way of generating models for paraconsistent theories,
and has antecedents in classical model theory. The (downward) Löwenheim-
Skolem theorem states that any theory with an uncountably infinite model has
a countably infinite model. The collapsing lemma is a kind of elaboration on
this – the “ultimate downward Löwenheim-Skolem theorem,” as Priest puts
it – in that it takes a consistent theory with an infinite model and reinterprets
it as an inconsistent theory with a finite model. The general structure of these
models is now well understood and involves sequences of successor-cycles off
to infinity.56
55 From Dunn (1979).

56 See Priest (2000, p. 1523). For some (negative) answers to open problems set by Priest, and
in particular some LP-models that are not obtainable by collapse, see Paris and Sirokofskich
(2008). For early discussion of finite, inconsistent models of arithmetic see Nelson (1959).

A downside of this theory is that, because it is embedded in the logic LP, it

has only the material conditional ⊃ as its implication operator, which does not
obey modus ponens. Thus, it may be the case that
n+1=m+1 and n+1=m+1 ⊃ n=m
and yet it may not be the case that n = m. This is what allows there to be
inconsistencies, but apparently limits the use of the theory.57 An advocate of
the theory might reply that failures of modus ponens will only occur “above”
n and that for lower-down consistent numbers “subtraction” works just fine.
A critic might reply that arithmetic subtraction is not supposed to be the sort
of thing that only works sometimes. In any case, as with set theory, it seems
reasonable to consider working with a logic that has a conditional.
A candidate conditional that can easily be added to LP (indeed, conserv-
atively extending it) for the purposes of arithmetic has been investigated in
Tedder (2015). Suppose A ⇒ B is true when A is false, and otherwise takes the
value of B. In three-valued terms this is
⇒ t b f
t t b f
b t b f
f t t t
This gives an appealing conditional that obeys modus ponens, which has
been investigated by Arnon Avron, Batens, and others. Tedder applies collaps-
ing techniques. Then a sound and complete axiomatization of paraconsistent
arithmetic can be given in this logic; cf. Tedder (2021).
Another, more radical, direction is to move away from the idea of “gen-
eralizing” classical logic and considering instead contra-classical logics – in
particular, connexive logics. These are systems that include the principles
Aristotle ¬(A → ¬A)

Boethius (A → B) → ¬(A → ¬B)
Paraconsistent connexive logic has been studied by Routley/Sylvan, and then

prominently by Heinrich Wansing, and Omori.58 Ferguson has considered the
idea of arithmetic in connexive logic and again makes extensive use of col-
lapse models in Ferguson (2016, 2019). There are some intriguing features of
this approach, notably that some connexive logics are inconsistent even in their
propositional fragment. Investigating connexive mathematics though would be
57 As pointed out by Denyer (1995).

58 See Omori and Wansing (2019a).

in some ways an even more radical departure than anything considered here,
since it would seem to rule out any reductio proofs, or at least render key
premises in a reductio at least false.
As with set theory, then, we find ourselves looking for a developed theory
of paraconsistent arithmetic with a robust conditional; as with set theory (but
in fact chronologically prior), we turn to relevant logic.
3.2 Relevant Arithmetic

A relevant arithmetic begins with an axiomatization of first-order relevant
logic, like DKQ mentioned earlier, or a stronger logic like RQ, which has the
additional axiom
A → ((A → B) → B).
Add to the language the symbols 0, ×, +, and ′, and add the following version
of Peano’s postulates:
∀x∀y(x = y ↔ x′ = y′)
∀x∀y(x = y → (x = z → y = z))
∀x(x′ , 0)
∀x(x + 0 = x)
∀y∀y(x + y′ = (x + y)′)
∀x(x × 0 = 0)
∀x∀y(x × y′ = (x × y) + x)
This looks extremely conservative. The third axiom, for example, says that 0 is
not the successor of any number. But recall that the “not” here is paraconsistent
(LP) negation.
Atop these one adds some form of induction axiom. If we add
∀x(x , 0 → ∃y(x = y′))
(which says that every number other than zero is a successor) then we get rel-
evant Robinson arithmetic, which has been investigated by Dunn. (Notice that
this phrasing of the axiom has a built-in disjunctive syllogism.)59 Instead of
that we can add the rule
If A(0) is a theorem, and ∀x(A(x) → A(x′)) is a theorem, then ∀xA(x) is a

theorem,
59 One would also need to add a few further postulates that are derivable from stronger forms of
induction. See Dunn (1980, p. 408).

which is mathematical induction. The quantified logic R plus the Peano postu-
lates and mathematical induction rule is called R# , or “arr sharp.”60 If instead
of the induction rule we add the infinitary omega rule
If A(0), A(1), A(2), . . . are theorems, then so is ∀xA(x)

then for logic R we get R## .
Let us focus on R# as it has received the most attention, being a fairly strong
and natural logic and one that held out some early promise to investigators in
the 1970s and beyond.61
As a candidate for working arithmetic, R# has some strengths. Basic arith-
metic properties can be developed, for example, that every positive number can
be divided (Meyer, 2021a, p. 272), and indeed that through some translation
arguments, many classical theorems appear to be recovered. R# can represent
all recursive functions. It is also, like its better known classical counterpart,
essentially undecidable, and incomplete. So far, so arithmetic.
We have already seen that theories based in LP admit some very simple mod-
els. But the language of relevant logic is substantially more expressive; we can
say things like A → 0 = 1, which might intuitively be thought of as a way
to say that A must be ruled out. And yet, perhaps the seminal discovery in the
history to date of inconsistent mathematics is that R# has some simple models,
too.
The idea is to borrow some ideas from modular arithmetic, inaugurated by
Gauss. Modular arithmetic wraps numbers around in a circle, like on the face
of an analog clock; so there, if it is 9AM and we ask what time it will be in five
hours, the answer is 2PM, or 9 + 5 = 2 modulo 12. Applying that thought to
a model for inconsistent arithmetic, take the domain of the model to be {0, 1},
and interpret everything as arithmetic modulo 2, so that the successor of 0 is
1, the successor of 1 is 0, and any other numeral is identified sequentially with
either 0 or 1. This will have the effect that, in the model, all self-identities are
both true and false. But then, because in the model it is simply false that 0 = 1,
there is at least one sentence that is not satisfied, and the whole theory is proved
nontrivial.
That is notable because the proof is “finitary” in Hilbert’s sense. Any quan-
tified claim like ∀x(A(x)) can be replaced with a finite conjunction, A(0) ∧ A(1),
and similarly a disjunction for existentials. That means the nontriviality of the
theory can be established “by anyone who can count to three.” Meyer declares
this grounds to “overturn” Gödel’s second incompleteness theorem.
60 Using quantified logic DK instead would give DK# as in Routley (1977).

61 Along with Meyer’s recently reprinted opuses Meyer (2021a, 2021b), see Meyer and Mortensen
(1984, 1987); Meyer and Restall (1999); Mortensen (1988) and elsewhere.

This is evidently a finitary proof refuting 0 = 1 in first-order arithmetic,

as Hilbert desired. (Moral: Hilbert should have used R# ) (Meyer, 2021c, p.
151).
By accepting some simple inconsistency in some models (like 2 + 2 , 4 and

2 + 2 = 4, conceding that the simple consistency of the theory is simply not
provable) then the absolute consistency of the theory is forthcoming, and that
by methods that arguably would have been acceptable to Hilbert.
“Overturn Gödel”? How radical a claim was this? Meyer was not, I think,
disputing the validity of Gödel’s proof for classical Peano Arithmetic or related
systems. He was rather disputing that these are good formulations of arithmetic,
and was arguing this on the basis of other relevant paraconsistent formulations
not suffering from similar deficiencies. So Meyer was staking a revisionary
claim about the correct way to formalize mathematics, but he was doing so with
some fairly conservative targets in mind. In particular, he was hoping that this
better paraconsistent formulation would not require any revision of accepted
“properly arithmetic” theorems.
What we wish to be sure of, as Hilbert might have put it, is that excur-
sions through general logical laws . . . do not render dubious what we rightly
regard as indubitable . . .. From the present viewpoint, the task of furnishing
a non-mythological and demonstrably secure reconstruction of all mathe-
matics was interrupted over trivia, and it is time that these trivia were placed
once more in proper perspective. Again, I do not propose to change the log-
ical superstructure – only to understand it more clearly, by making explicit
in a formal way features that have belonged to our intuitive logic all along
(Meyer, 2021b, pp. 298–299).
By Meyer’s own lights, this hope was not realized, as we now see.
Buoyed by the prospects of an improved formalization of arithmetic that
“removes the sting” of Gödel’s theorem (something that LP arithmetic does) but
still apparently has some proving power (something that LP arithmetic does not)
Meyer and collaborators wondered whether R# could recapture all of standard
Peano Arithmetic. This turned out to be an instance of a more general question
in relevant logic, called (for no very good reason62 ) the gamma problem.
The gamma problem asks about classical recapture. We know that, in order
to be paraconsistent, these logics must not validate disjunctive syllogism. But
in situations that are consistent, perhaps it would turn out that all the effects
of disjunctive syllogism still obtain “for free” as it were? That is, anything
provable with disjunctive syllogism would turn out already to be a theorem
62 The name comes from work by Ackermann in 1956, via Anderson and Belnap (1975), where
disjunctive syllogism is labeled γ (gamma).

without it. In such a situation, disjunctive syllogism would be an invalid but

admissible rule. And “by happy accident (or the beneficent management of
the Universe) all relevant theories for which the question had hitherto been
serious have admitted gamma” (Friedman & Meyer, 1992, p. 827) so Meyer
was optimistic.
But in 1992, Meyer and Friedman showed that, perhaps surprisingly, the ring
√
of complex numbers C (which includes imaginary numbers such as i = −1)
can soundly interpret a fragment of arithmetic that includes all the positive
theorems of R# . But then there is a formula,
∀x∃y∀z∃a∃b(a(2x + 1) + b(y − z2 ) = 1),
which is a theorem of Peano Arithmetic but is false in the complex numbers

and hence not a theorem of R# . So it is not even the case that every strictly
positive theorem (let alone any negations) of PA is a theorem of R# , and as
a corollary, disjunctive syllogism is not admissible. And that, in turn, means
that there are apparently true theorems about arithmetic that relevant arithmetic
(especially logics weaker than R) cannot prove. Perhaps in retrospect this is not
so surprising; given that R# has a two-element model, it might be suspected that
it has many other unintended models. But the upshot is unmistakable. Classical
recapture: lost.
Whither relevant arithmetic? Meyer went on to speculate about R## , which
has the omega rule, and which he argued contains all truths about the natural
numbers (Meyer, 1998). This “true arithmetic” would have disjunctive syllo-
gism as an admissible rule. It would still be incomplete, since 0 = 2 → 0 = 1
is not provable, nor is its negation. But the omega rule is infinitary, and so
adding it removes any of the “finitary” gains on Gödel’s second theorem and
Hilbert’s Program – not to mention removing the effectiveness of the theory.
So Meyer conjectured the existence of some theory between R# and R## in
the Goldilocks zone. Such a theory has not been found.
For all its trappings as radical and revisionary mathematics, the relevant
arithmetic project was a conservative effort. Meyer was happy with some funny
models of arithmetic making some of its sentences inconsistent in a model, but
did not seem to think that true arithmetic is, in the end, inconsistent.
We do not think of such negated formulas as 2 + 2 , 4 as true. But the point
is, even if we did, it still wouldn’t follow that 2 + 2 = 5. That’s the idea,
anyway. (Meyer & Urbas, 1986, p. 49)
Meyer’s proofs involved the entirely classical construction of classical models

for theories. Priest’s excursions with LP are more serious about contradictions,
but they are still based squarely in classical models and methods.

Research on inconsistent arithmetic pulls in two directions at the same time:

“classical logic is wrong! Gödel is a dirty trick!” but “nothing is lost by going
this way!” The motives point toward very radical challenges to mathematics,
while the methods speak of calm and conservative tendencies. This can be, I
think, a little confusing. There is nothing controversial about the integers mod-
ulo 2, or dividing the numbers into equivalence classes, and nothing to stop
anyone from naming nonstandard objects with the standard names for num-
bers. In the two-element model of Meyer’s arithmetic, indeed, “2 + 2 , 4” is
actually a funny way of writing that “0 + 0 = 0” is assigned the value “both”
in RM3. The proof that C is a a model for some subtheory of arithmetic is, sim-
ilarly, a completely classical piece of mathematical reasoning. Priest assumes
an initial segment of the numbers in N are standard. And the aim of recovering
accepted mathematical truths within this epistemologically more secure theory
was an attempt to repair the conventional mathematical edifice, not break free
of it.
For an enterprise self-styled as “inconsistent mathematics” this is all sur-
prisingly moderate. There is a tension inherent here, for example, in saying
there might be an inconsistent natural number, but still relying on methods and
proofs assuming the consistency of the natural numbers. It seems to me that
this tension becomes difficult sustain – at least, under an ultimately classical
methodology.
There is, in all of this, a generous helping of the Standard Mythology. It may
be that we are confused about truth in the standard model N of the natural
numbers. One could demonstrate this simply by improving Gödel to the point
where an actual proof of A&¬A for some A is classically forthcoming (in,
say, first-order Peano arithmetic). (Meyer, 1998, p. 349)
Let us see about that.
3.3 Incompleteness and Computability

The clearest reason to engage with paraconsistent logic for mathematics is if we
think that there are contradictions in mathematics. We might find Dunn’s pre-
cautionary Pascal’s wager motive enough, or we might just like imagining, but
it would make paraconsistency undeniable if some (arithmetic) contradictions
were true.
An argument to this effect was given by Priest in his seminal 1979 paper
on the “logic of paradox,” building off of work with Routley and modulating
the idea floated by Anderson, Belnap, Meyer, and others that there is room
to dispute and improve the classical formalization or translation of mathe-
matics. If Priest’s argument succeeds, it would amount to a proof that there

are true inconsistent sentence pairs in arithmetic, the most elemental part
of mathematics. If so, that would make paraconsistency in mathematics not
merely a curiosity or side project, but urgently needed in order to retain (or
explain) the coherence of arithmetic, if not mathematics overall. Here is the
argument.63
Mathematics is a practice, conducted by humans using natural language.
This language (let us suppose English) is augmented with technical jargon
and many special symbols, all of which are assigned unambiguous meanings.
The language is used to communicate ideas and results, especially through
proofs.
A proof begins from some accepted statements and proceeds in a replicable
step-by-step way to a conclusion. All of this needs to be comprehensible and
checkable by other mathematicians. Depending on the level of doubt about a
given purported proof, the language may be subjected to more and more scru-
tiny, each statement and steps between the statements analyzed further and
further down into clear and unambiguous language, until ultimately it is decided
that the proof is valid or invalid (perhaps invalid because insufficiently clear).
In principle, any piece of mathematical reasoning should be analyzable in this
way. The practical details of all this might be very daunting (e.g., Perelman’s
proof of the Poincaré conjecture was posted in 2002 but not accepted as verified
until 2006 – and even that was not done by reducing everything to predicate
logic or the like) but such concerns are merely medical; if a proof is not for-
mally checkable in principle, it is not a proof. We can effectively assess the
validity of any proof. And not just that, but because proofs begin from state-
ments accepted as true and proceed by valid steps, we can see that mathematical
proofs are sound, that theorems are true. Informal proofs effectively prove true
theorems.64
But then – argues Priest – our “informal” mathematical discourse meets the
conditions for being a formal axiomatic system, where the provability relation is
effective. By the Church-Turing thesis, all effective procedures are recursive,
and in our formalized “naive” mathematics, all recursive relations are repre-
sentable. And if all that is correct, then informal or “naive” mathematics meets
all the conditions for Gödel’s incompleteness theorem. This has some dramatic
consequences.
63 Compare with Priest (2006, chs. 3, 17), Berto (2007, chs. 4, 12), and Routley (1979).
64 In classical arithmetic Löb’s theorem states that the soundness of arithmetic cannot be estab-
lished within arithmetic itself: that if “A is provable → A” were a theorem, then A would be a
theorem – for any A, which is absurd. But the proof of Löb’s theorem uses contraction. In logics
without contraction this version of incompleteness could be evaded. See van Benthem (1978,
p. 54). For more on the idea of informal provability, see Leitgeb (2009).

Gödel’s theorem states that any formal system that is able to represent all its
own recursive relations, and which has an effective proof relation, will contain
what he called “undecidable” propositions. One such undecidable proposi-
tion may be informally glossed as a statement in the language of the system
that says “the system cannot prove this very statement.” This is a gödel sen-
tence g for the system. If the system did prove either g or its negation then
the system would be inconsistent. Crucially, on the standard story, we estab-
lish all this from outside the system, showing that it is incomplete – and in
doing so, we can see that g is true. There is an unprovable truth, on pain of
contradiction.
Gödel gets to incompleteness by stepping outside the formal system. The
trick when it comes to the system of naive mathematical proof is that there is no
“outside” the system to retreat to. Rather than showing there is some sentence
our consistent system cannot decide, confronted with the gödel sentence g we
end up proving both g and its negation. The system itself, our system, is incon-
sistent. The system in question is our colloquial, informal, sound mathematical
discourse; so mathematics itself is inconsistent.
There is a lot to say about this argument and the issues it raises.65 We do not
have time to say any of it. I will only point in one of its most audacious, and
untapped, directions.
As Shapiro points out,66 and Priest agrees,67 if the gödel sentence is both
true and false (has a false negation), then this not only shows an inconsistency
in arithmetic, but that proof itself is inconsistent: for we have shown that there
is a proof of g, but then because g is true there is also no proof of g (since
that’s what g says). So the proof of g is also not a proof of g. Since a proof
is a precise object, a sequence of sentences beginning with axioms and ending
with a theorem, it is fair to ask where exactly in the (valid) proof things go
wrong.
If our informal proof relation is both effective and inconsistent, then the
this question approaches a very intriguing, and mostly uninvestigated, topic:
paraconsistent computability theory. What would an inconsistent computa-
tion be? Gödel (and Turing and Church) showed that there are hard limits on
what any consistent computer can do; but now we are beyond consistency.
What, in the end, about the prospect of reviving Hilbert’s failed dream of a
universal truth machine that effectively decides all mathematical questions?
This would have to be an inconsistent machine, of course, but so be it. The
65 See Carrara and Martino (2021) and Tanswell (2016).

66 In Shapiro (2002); cf. Shapiro (1998).
67 In Priest (2006, pp. 239–243).

familiar limits on computability from Turing et al. are really just variations
on the sorts of paradoxes that seem amenable to paraconsistent treatment –
the existence of uncomputable procedures is a diagonalization argument like
Cantor’s theorem – and so perhaps these apparently unbreakable limits can be
broken?
But to answer any of these questions is again to face Shapiro’s question,
of what, for example, a computation that is also not a computation, might look
like. Priest for the most part deflects these questions, and instead answers ques-
tions about what it would mean for the description of a computation to be
inconsistent; the answer to that is relatively easy: there would be some incon-
sistent natural number n , n, which he has already made peace with. This does
not tell us, though, what the description is describing. And here Priest deems
this “metaphysical speculations” (Priest, 2006, p. 243).
Saying something more meaningful about the hardware of paraconsistent
proofs either requires some advances in physics, as Copeland and Sylvan seem
to suggest (Copeland & Sylvan, 1999; Sylvan & Copeland, 2000; cf. Weber,
2016), or some other insight, perhaps something mundane along the lines of
Weber, Badia, and Girard (2016). Or, this is evidence that paraconsistent math-
ematics reaches its own limits at the horizon of mechanical realizability, which
is more or less what Gödel said about classical mathematics anyway.
Priest’s argument, I would say, is based on a fair gloss on the received
mythology about mathematical proof. It would also be fair to say that, espe-
cially since Gödel’s theorems, that mythology has been called in to question
as too simplistic and maybe the lesson to take is that our naive proof prac-
tices are more open ended. As is often the case, there is agreement about a
conditional statement, “if naive proof is recursive then it is inconsistent” and
then disagreement as to whether modus ponens or modus tollens is called
for.
4 Calculus, Topology, and Geometry

The foundations of mathematics – set theory, arithmetic – are some natural
places to think about applications of paraconsistency, since they are philosoph-
ically rich and also sites where some infamous paradoxes (or close relatives)
can be considered: the antinomies of set theory, the Gödel-cum-liar sentences
of arithmetic. But there are also applications of paraconsistency in other more
working-class areas of mathematics, which we will now consider. And these
touch on a different, even older sort of paradox that does not get as much play
in the philosophical logic literature anymore: Zeno’s paradoxes of motion and
change.

4.1 Infinitesimals
“Questions about the immeasurably large are idle questions for the explanation
of Nature,” says Riemann. “But the situation is quite different with questions
about the immeasurably small.”68 Let us see.
It is easy to measure a straight line. It is not so easy to measure a curved
line. An idea for tackling the latter problem goes back to Archimedes, and is
emblematic of a lot of mathematical ingenuity: to break one big difficult prob-
lem up into many small but easy problems, or in this case, to imagine a curved
line as composed of infinitely many infinitely small straight lines. Archime-
des thought of this as a way to approximate the correct answer, but eventually
methods that do not merely approach but arrive at the precise answer were
accepted – the Calculus.69
As conceived by Newton and Leibniz (independently), the differential cal-
culus is concerned with “infinitesimal” changes. In its original versions, there
were no scare quotes: calculations involved terms (Newton called them “flux-
ions”) that were infinitely small, in the sense of being big enough to divide by
(since division by zero is undefined), but small enough to disregard completely.
These “ghosts of departed quantities,” as they were unflatteringly called, were
remarkably effective for getting results, but also apparently inconsistent. While
methods developed in the 1800s by Weierstrauss et al. make it possible now
to avoid use of infinitesimals, and moreover innovations by Robinson in non-
standard analysis make it possible to simulate the appearance of infinitesimals
using some loopholes in first-order logic,70 a return to the original intuitive
calculus has remained an attractive prospect for paraconsistent research.
For a real number x, let δ(x) be, intuitively, an infinitesimal part of x. For a
continuous function f, the original definition of its first derivative was
f(x + δ(x)) − f(x)
f ′(x) = .
δ(x)
This can only be allowed if δ(x) , 0, since we are dividing by it.71 For example,
for the function g(x) = x2 , then we have
(x + δ(x))2 − x2
g′(x) =
δ(x)
2xδ(x) − (δ(x))2
=
δ(x)
68 From “On the Hypotheses That Lie at the Foundations of Geometry,” §3, translated in Spivak
(1999, vol. 2, pp. 151–162).
69 See Bell (2008, introduction). Cf. McKubre-Jordens and Weber (2016).
70 See Goldblatt (1998).
The standard modern definition is f ′ (x) = limh→0
71 f(x+h)−f(x)
h , which allows division by h as it
approaches zero.

= 2x + δ(x)
= 2x.
But then at the last step, the δ(x) term is thought of as being so small as
to not matter at all, whence we have the standard rule for powers, that the
derivative of x2 is 2x. It looks like this reasoning involves two assumptions,
namely,
∀x(δ(x) , 0) and ∀x(δ(x) = 0).
Assuming the quantifiers quantify over anything, this is plainly inconsistent,

and hence looks like a fertile site for paraconsistency in mathematics.72
It is not so easy, though, just to “accept” that infinitesimals are inconsistent,
for similar reasons as to why division by zero is undefined. Suppose we do
so try to accept that, for some number a, δ(a) is both zero and not zero – not
approximately, but exactly. Then since 0 = 2 × 0, by the assumption that δ(a) =
0 and substitution, δ(a) = 2δ(a). Then by the assumption that δ(a) , 0, we can
divide both sides by δ(a), and end up with 1 = 2. (And then, subtracting from
both sides, 0 = 1.) The trivial conclusion that a = b for any “two” numbers
a , b does not look far away.
Even for someone prepared to accept that some unimaginably tiny numbers
are inconsistent, this looks like much too much. Avoiding it would seem to
require tampering with some of the basic field properties. Mortensen notes that
there are finite, inconsistent, but nontrivial models of classical field theory, but
this is done by phrasing the key axiom about division as
x
∀x(x = 0 ∨ = 1)
x
(Mortensen, 1995, p. 29), which avoids the issue because without disjunctive
syllogism this says very little. To capture the intuitive idea of infinitesimals
without collapse into absurdity, Brown and Priest propose what they call a
“paraconsistent reasoning strategy” called chunk and permeate. It is based on
72 A related example: In 1930, Dirac introduced as “convenient notation” a function δ to represent

a mass that is everywhere 0, except at 0, but with an integral over the reals equal to 1. That is,
the delta function is such that
(i) ∫δ(t) = 0 for all t , 0

+∞
(ii) −∞ δ(t)dt = 1
There is an instantaneous-but-not jump at zero. This turns out to be extremely useful, for exam-
ple, modeling the dynamics of force of impact. This also turns out to be classically impossible –
there is no function like this, as Dirac well knew. Proponents of inconsistent mathematics have
suggested that “bad” objects like the delta function could be prime candidates for paraconsistent
rehabilitation. See Priest, Routley, and Norman (1989, p. 376) or Mortensen (1995, ch. 7).

an approach to paraconsistency called preservationism, developed by Schotch

and Jennings; see Schotch, Brown, and Jennings (2009).
The basic idea is to take an inconsistent set of information and break it
up into “chunks” that are individually consistent, following one of the earli-
est strategies in paraconsistency due to Jaśkowski. Information is allowed to
“permeate” from one chunk to another in a controlled way. If we imagine this
as a sequence of sets of information, starting from a set of sentences closed
under classical logical consequence, then each step in the sequence comprises
what can be inferred from the previous step, plus whatever flows in from other
chunks.
Then a paraconsistent consequence relation may be defined over such a
structure: the “preserved” consequences of a set of sentences are those that
remain at the limit of this sequence, the sentences that can be inferred when
“all information of the appropriate form has been allowed to flow along the
permeability relations” (Brown & Priest, 2004, p. 381).
To model the infinitesimal calculus with chunk and permeate,we consider
just two chunks, a source and a target. The source ΣS containing the axioms
for a real closed field and its (classical) consequences, along with additional
axioms for taking derivatives and stating that ∀x(δ(x) , 0). The target ΣT does
not contain these axioms but instead has an axiom asserting that ∀x(δ(x) =
0). The set ΣS together with ΣT is inconsistent, but each chunk is consis-
tent. Information about derivatives is allowed to flow from ΣS to ΣT but not
back. A similar setup could be defined for integration, and one can prove
that in this binary setup, the flow of information into the target preserves
consistency.
Chunk and permeate is a way of modeling inconsistent reasoning. It is not a
working system for using infinitesimal calculus. There have been other more
direct attempts to embed calculus (or real analysis more broadly) into a para-
consistent logic. These adopt Robinson’s theory of “hyperreals” but with a
background of various paraconsistent logics, as in D’Ottaviano and Carvalho
(2005). Mortensen presents a kind of “microscope” that cannot distinguish
between quantities below a certain order of “infinitesimality” (Mortensen,
1995, p. 58). A paraconsistent account of real numbers is given in McKubre-
Jordens and Weber (2012) but without broaching the infinitesimal problem.
More radically, the abovementioned triviality problem is tackled by tampering
with the field properties of the reals in (Weber, 2021, ch. 8); and perhaps most
radically, the idea of accepting that the field collapses to a point is considered
by Estrada-González (2016). Formulating a satisfactory theory of inconsistent
infinitesimals remains an alluring but difficult and unsolved problem.

4.2 Topology
Calculus works by thinking about space with a metric, so that distances and
areas are assigned a quantifiable number. Topology drops the metric. The intu-
ition that there is something interesting but inconsistent about infinitesimals
has in some ways received better paraconsistent treatment in more qualitative
theories of space, where the metric is dropped, and the formalism is closer to
an underlying set theory than the equations and rules of calculus. We can focus
simply on regions as sets of points, where we can look to model impossibly
small-but-not-zero “instants.”
Think again of the idea of a curve as composed of infinitely many infinitely
small lines. Suppose this curve is describing the trajectory of an object moving
through space, for example, you walking from your bed to the door, stepping
around a pile of socks. You are moving, but if we think of the path as composed
solely of unextended points, none of these are moving; there is no motion at an
instant. No matter how many instants we include that depict the transition from
one to another, nothing moves. As Russell puts it (approvingly), “Change does
not involve a state of change” (Russell, 1903, p. xxxiii).
Put this way, the problem seems to be about capturing the “dynamics” in a
static formalism, but we can remove the movement and still see the problem.73
Suppose you step around your socks and reach the door, and step through it.
At some location along your path you were in your room (a ∈ A), and then at
a location further along the path, you were not in your room any more (a < A).
What happened? In particular, what happened at the plane of the doorway?
There are four options:
(i) a ∈ A only
(ii) a < A only
(iii) neither a ∈ A nor a < A
(iv) both a ∈ A and a < A
The fourth option would lead us to consider a paraconsistent account. By way

of motivation, let us consider the alternatives.
Option (iii) is that there is some kind of gap, or indeterminacy. This would
be a violation of the LEM. There may or may not be good reasons to doubt the
LEM. Perhaps, for example, someone who ate lunch five hours ago is neither
hungry nor not hungry. Without taking a stand on the LEM, we can just say
that, when you are standing in the doorway, on the precipice between two sets,
73 Here, following Priest (2006, ch. 11). Priest considers what he calls the (classically false) Leibniz
continuity condition, which is taken up by Mortensen (1995, ch. 6).

it is hard to see why there is any indeterminacy. If we identify you with your
center of gravity and the doorway with the plane of the door, then there is an
instant when the point is on the plane, and either that is in the room, or not.
Any theory of your journey that treats this as indeterminate is, in a basic sense,
an incomplete theory of what happened. Option (iii) is just not answering the
question.
The broadly accepted classical theory, then, is to say that when you are
exactly in the doorway exactly one of (i) or (ii) is the case, and it is more or
less arbitrary which one. Either you are at the last in-the-room point, or the first
not-in-the-room point, and only one of these is true, and one of these must be
true, but there is no way to settle which beyond just flipping a coin. This has
an advantage over the gappy response insofar as it answers the question, but
it leaves a nagging sense of asymmetry. The very fact that we are free, more
or less, to stipulate whether it is (i) or (ii) is itself an indication that there is no
meaningful difference between these options. The classical theory is not merely
arbitrary, but also missing something about the scenario, namely that options
(i) and (ii) are symmetrical.
Maybe, then, inconsistency (option (iv)) can capture instantaneousness. To
model this, we can go from the metric idea of being “infinitesimally close” to
the topological idea of being “very nearby.” For a set of points A, the set of
points that are very nearby A is called the closure of A, usually written A− .
Every point in A is in A− , and A is called closed iff A = A− . Now, if we think
of a space X divided into two sets A, B such that every point in X is either in A
or B but not both, then B is the complement of A in X and we can define the
boundary (or sometimes ‘frontier’) of A as being all the points both very close
to A and to its complement:
bdy(A) := A− ∩ B−
So now we can rephrase the question from before: Is the boundary of A a part
of A, or a part of the complement of A? It must be one of the two, since A and B
exhaust the space by construction. But again it would seem asymmetrical and
arbitrary just to pick one and not the other. It seems like the boundary belongs
to both, that the spaces A and B touch each other at their shared boundary. But
a closed set has its boundary as a part: A is closed iff A includes its boundary.
If A and B are both closed, then points on the boundary will be in both A and
B; and since points in B are not in A, that is a contradiction.
Classically, that means that, if A and B are both closed, their boundary must
be empty, by reductio. But you are standing on the boundary; it is palpably
not empty. If your room and the abutting hallway are touching, and the overall
space has no rips or tears in it, then the reductio is not a reductio but a proof:

when you are on the boundary of A (as you are) then you just are both in A and
not. Or so goes this line of thought.
A formal theory of paraconsistent topology, or mereotopology (the theory
of connected parts), can be developed to make all this precise. Accounting for
inconsistency, the following can be shown:74
An object and its complement share a boundary, up to inconsistency: if B is
the complement of A then
bdy(B) = bdy(A ∪ (A ∩ B))
The boundary of the complement of A is the boundary of A plus anything

both in and not in A.
A nontrivial model of this scenario using two elements and the logic RM3 is pro-
vided in Weber and Cotnoir (2015) and further point set topology is in (Weber,
2021, ch.9).
Some of these topological notions can be fed back into logic. The closure
operator has some inherently attractive properties:
• A is included in its closure

• The closure of the closure of A is included in the closure of A
• If A is included in B, then the closure of A is included in the closure of B
These properties are attractive not least because, if instead of thinking of sets
of points we think of sets of sentences, and we read the closure operator as the
set of logical consequences of those sentences, then this provides a nice model
of logical consequence, following an observation by Tarski.
If we think of sentences as indexed by a set of points (such as the “worlds” at
which the sentence is true), then we can think about the topological structure of
those sets. If we think of sentences as holding only on closed sets, this gives rise
to a paraconsistent logic, which as a result Mortensen declares to be the “most
natural.” Conjunction and disjunction are the union and intersection of closed
sets, respectively, and negation is the smallest closed set containing the com-
plement. (As a three-valued logic, it is the same as RM3 except that in the truth
table for negation, “not both” is identified with “true.”) The distinctive feature
is that at the boundary of a proposition, both it and its negation hold. This is
exactly the dual of open set semantics for intuitionistic logic. See Mortensen
(2010, ch. 2).
Generalizing, in the area of category theory it is known that a certain
class of structures, “toposes,” correspond to intuitionistic logic; they are
74 Cf. Weber and Cotnoir (2015, p. 1284); cf. Weber (2021, p. 280).

(vastly) abstracted from open sets. Then it is straightforward to define a

dual-topos, which corresponds to a paraconsistent logic as its internal logic.
See Mortensen (1995, chs. 11, 12) and bold strides from Estrada-González
(2010, 2015).
There is a great deal more to be done in developing topological semantics
for paraconsistent logics, especially those based on a paraconsistent topology.
4.3 Impossible Pictures

According to Brown, mathematical pictures can be “windows into Plato’s
heaven” (Brown, 2008). According to Mortensen and his group (especially
Steve Leishman and Peter Quigley) working in inconsistent geometry, so-
called impossible pictures are sites for fruitful paraconsistent study: “making
mathematical sense of . . . real pictures whose content is of logically impossible
or contradictory objects” (Mortensen, 2010, p. 69).
The Necker Cube is, famously, a two-dimensional image that calls out
for a three-dimensional interpretation while at the same time making such
interpretation very difficult. Here is the cube:
Here are two consistent resolutions of it:
But here are two inconsistent resolutions.
They are inconsistent because some of the faces are both in front of, and not in
front of, the other.

These are examples of what Mortensen and his collaborators call impossible
pictures – images that offer some visually immediate material for inconsistent
mathematics. These are the sorts of pictures that are found canonically in the
artworks of M. C. Escher. Impossibilia are pictures that upon analysis have
some propositions A and ¬A as part of their description. Mortensen is con-
cerned with capturing how these pictures “seem” to us, how we cognize them
in our minds, and argues that any consistent interpretation of them will inevi-
tably leave something out: namely, that they present as contradictory in some
essential way (Mortensen, 2010, p. 74).
For example, at the end of the movie Labyrinth (1986)75 there is a scene that
takes place in an impossible room filled with staircases going in all directions;
the content of the scene clearly indicates that at some points the characters are
both “above” and “not above” in relation to each other. This scene would make
no sense if it were resolved into noncontradictory terms; the logical vertigo is
a key component of the story. This is a fictional fantasy story, of course, but
the images on screen present clear geometric properties and describing these
faithfully would seem to call for inconsistent statements.
The edges of a necker cube in the above diagrams are either dotted or solid;
to be more colorful, they may be thought of as painted either red or blue. So
when two lines cross, the image may indicate that either (1) the red line occludes
(blocks) the blue line, (2) the blue line occludes the red line, (3) neither occludes
the other, or (4) they overlap and occlude each other. To get at the underlying
impossible content, we specify that a local contradiction about crossings would
violate a “Local Consistency Axiom,”
If x occludes y then y does not occlude x.
A Necker is locally inconsistent iff it has both red and blue at some crossing.
Then a face of a necker is a figure with four edges intersecting at the corners
in four vertices. One face is in front of another face if some edge of one occludes
some edge of the other, violating a “Global Consistency Axiom”
If x is in front of y then y is not in front of x.
And we have a definition of global inconsistency: a (theory of) a Necker is

globally inconsistent iff (it says that) there are faces that are both in front of
and not in front of each other, that is, it is inconsistent with respect to the “in
75 Directed by Jim Henson. Written by Terry Jones. There is also a part where the protagonist
faces a classic liar/truth-teller riddle, of the sort made famous by Smullyan in his “knights and
knaves” scenarios.

front of” relation. It would be interesting to have conditions that connect local
and global inconsistency.
To carry on more precise study requires some formal representation of the
targets. For this, Mortensen employs linear algebra , which works with systems
of equations; cf. Mortensen and Leishman (2009). In any drawing of a Necker
Cube there are only two places where lines cross. If we denote these C1 and C2
then each cube can be described by a pair of simultaneous linear equations
aR + bB = C1
cR + dB = C2
with a, b, c, d ∈ {0, 1} so that e.g. 0R + 1B = C2 is read “crossing C2 is com-

posed of zero red lines and one blue line.” This can be presented as matrix
multiplication,
[ ] [ ] [ ]
a b R C1
× =
c d B C2
Call this the primary equation and of the cube and the two-by-two matrix in it
its primary matrix. A cube is thus locally inconsistent iff it has all ones in one of
its rows. With cubes represented as matrices, we can[ then ]appeal to the standard
a b
notion of the determinant of a two by two matrix as (a × d) − (b × c),
c d
and provide a sufficient condition for contradiction:
If the determinant of a primary matrix of a Necker is not zero, then the Necker
is locally inconsistent.
This is not very illuminating – it amounts to checking a small number of cases –

and the converse fails; but it can be improved somewhat:
A Necker is inconsistent iff its primary equation is such that R = 0 = B or

R = 1 = B.
Again this is a matter of checking cases, but it gives a quick necessary-and-

sufficient means of capturing inconsistency (Mortensen, 2010, pp. 93–94).
Faces of a Necker can also be represented. A face with only red edges is a
red face, RF, and a face with only blue edges is a blue face, BF; if we focus
on the two inconsistent neckers then there are unique red and blue faces, so we
can write
jC1 + kC2 = RF
nC1 + mC2 = BF

with j, k, n, m ∈ {0, 1} so to say, for example, “the red face is composed of one
red at C1 and no red at C2 ” as 1C1 + 0C2 = RF. Then this information is written
as
[ ] [ ] [ ]
j k C1 RF
× =
n m C2 BF
with the two-by-two array as the secondary matrix of the cube. Then there is
an account of how both local and global inconsistency is generated:
A Necker is inconsistent iff it has a column of all ones in its secondary matrix.
This “supplies the desired assimilation of local and global inconsistency”

(Mortensen, 2010, p. 96). If we think of the “unit space” as the set of all the
columns in the secondary matrix that have all ones, then the unit space contains
all the information about how inconsistent Neckers are generated. The degree
of inconsistency of a Necker can be defined as the number of independent solu-
tions to the secondary equation with a column all ones. These methods can be
dualized to catalogue complete Neckers, now focusing on the “null space” as
the set of all the columns in the secondary matrix that have all zeros; and the
methods can be extended to represents chains of Necker cubes.
Other impossible images are the impossible window and the Schuster fork:
There are images that seem to us to depict impossible situations – in particular,

contradictory properties: parts of the fork, such as the region between the top
and middle tines, seem to be both solid and not solid (Mortensen, 2010, ch. 15).
Interestingly, Thro points out that the window is built from the Necker cube and
is two Schuster forks put together (Thro, 1983):
Mortensen’s group has gone on to undertake a classification of all impossible

pictures. Using computer modeling, they disproved a conjecture of Ernst, that
an impossible picture cannot be rotated, by rotating some.76
76 For further history and analysis, and an extensive gallery of impossible pictures, see Mortensen
(2022).

4.4 Indispensability Arguments and Ontological Commitment

In §1.3 we saw that realism is not required by inconsistent mathematics. But,
how much from our theories of mathematics – orthodox or otherwise – are we
obliged to believe? In particular, what should we believe about the existence
of mathematical objects? Assuming that we would prefer to limit our beliefs
about the existence of objects so that we do not take on more than is required,
Quine famously posited criteria for ontological commitment: we should believe
in the existence of those objects that are needed for our best scientific and
mathematical theories. That is, Quine recommends we commit only to what
is indispensable. This is a much more stringent criteria than, say, a mathemat-
ical plenitudinous platonist has, but it is not so severe as fictionalism or the
outright denial of all abstracta; the Quine/Putnam indispensibility argument
has proved appealing to many, especially philosophers of a naturalistic bent,
who think that philosophy should take its cues from science.
This raises the interesting possibility, from a paraconsistent perspective, that
if some inconsistent mathematical theory turned out to be applicable and useful
in science, then Quine’s seemingly conservative criteria would commit us to the
existence of “inconsistent” objects – objects that have a canonical inconsistent
description. This idea has been developed by Colyvan, as a possible unexpected
consequence of paraconsistent mathematics.77
For, especially if one were taking a more measured approach to
inconsistency – for example, where any inconsistencies are thought of as con-
ceptual, as in Mortensen, or epistemic, as in Carnielli and Coniglio – then it
might come as an unwelcome consequence that one is suddenly obliged to
accept ontological contradictions, contradictions in the world. And yet, if say
the best current theory of galactic spiral rotation turns out to be couched in a
paraconsistent geometry, then that is just what one would find.
Anyone persuaded by the indispensability argument for scientific and math-

ematical realism, should also (perhaps reluctantly) sign up for belief in
inconsistent objects. Note that this is not an argument that the world is incon-
sistent or that the world contains inconsistent objects, just that there were
times when we had warrant to believe in such inconsistent objects. So the
conclusion is not as radical as it might first seem. But even this conclusion
many will find hard to swallow. (Colyvan, 2008, p. 119)
For those sympathetic to Colyvan’s idea (and if you are this far into an Ele-
ment on paraconsistency, maybe you are), there is now a way to make better
sense of what “realism” about inconsistent mathematics might mean. For we
already clarified that consistency (or not) is a feature of language or theories,
77 See Colyvan (2008, 2009, 2012, ch. 7).

and there is something of a category mistake in talking about an inconsistent

“object.” The onotological commitments of an inconsistent theory suggest that
an inconsistent object is the value of a bound variable that occurs in both con-
juncts of a contradiction, ∃x(A(x) ∧ ¬A(x)). The entity in question might be
something very boring, like your socks from §4.2.
For those who shrink from such a conclusion, it is worth pointing out that
even serious Quine-inspired naturalists do not uncritically commit to every
value of a bound variable from science. Maddy, for example, suggests that
we can use science as a guide to ontology, but that not even scientists them-
selves always commit to every bit of their own theories; science often traffics
in idealizations, which are not meant to be taken literally but are pragmati-
cally indispensable nonetheless. “We must allow a distinction,” Maddy concils,
“between the parts of a theory that are true and the parts that are merely use-
ful” (Maddy, 1992, p. 280). Perhaps paraconsistency, if it ever were to be used
in an indispensable scientific theory, serves to further emphasize the impor-
tance of not taking naturalism too far. The fact that some current best theory
includes some “unresolved” inconsistencies might be taken as a reductio of the
Quine/Putnam indispensibility argument for realism to begin with.
Perhaps the biggest objection to Colyvan’s idea would be that an inconsistent
theory can never qualify as our “best” theory, so the indispensibility argument
never gets off the ground. Colyvan counters that this objection is at odds with
actual scientific practice and hence a odds with naturalism – the methodological
principle that we should follow our best scientific practices and the basis for
indispensibility to begin with. As we saw in §4.1, the original calculus was
inconsistent. Or, it seems for instance that our current best physics includes both
relativity and quantum mechanics, and that these are incompatible, pending
some as-yet-undiscovered uber-theory.
Colyvan ultimately thinks that an inconsistent theory is only ever going to
be instrumentally and temporarily best, because he (like many, including work-
ing paraconsistent logicians) recoils from the idea of the world “really” being
inconsistent. Contradiction here is confined to our theories, which on some
accounts of science are never “ultimately” true (whatever that means). If glut
theory is correct and some contradictions are true, though, then commitments
could be binding.
5 Whither Paraconsistency in Mathematics?

Paraconsistent logic and inconsistent mathematics are relatively new human
endeavors, and it would be misleading to give the impression (in case I have
done so) that much of this is settled, uncontroversial, or fully worked out. This

Element does not arrive at the endpoint of a great discovery or the culmination
of a unified philosophical program, but is rather a start, for those not afraid of
risk, into an unknown future.78
For what it is worth, I think Asenjo, Arruda, and others are correct that
the future will hinge on the emergence of more mathematical results from
this diverse field. I think that, just as mountains are eventually climbed sim-
ply because they are there – even mountains that once appeared impossible
to climb – there is more to come from inconsistent mathematics; and if what
comes is useful or at least mathematically interesting, this will be more impor-
tant than the outcome of further philosophical debate. How might such future
efforts go?
Throughout this selective survey we have been taking note of when various
approaches are moderate, and when they are more revisionary. This can be
along two dimensions: in the content of the claims and results (i.e. whether
a paraconsistent mathematician is saying something that sounds like it is at
odds, or not, with standard mathematics); and in the methods used to support
those claims (e.g. whether a paraconsistent mathematician is using standard
mathematics as a framework).
In terms of content, we have seen some very conservative projects, includ-
ing embedding classical ZFC in a paraconsistent logic and Meyer’s dismay
that classical PA is not recovered in relevant arithmetic. We have seen also
some more risqué efforts with inconsistent sets, inconsistent natural numbers,
and inconsistent pictures – and a lot of finite inconsistency approximating the
infinite; but in terms of methods even these are more conservative in the sense
that they appeal to classical devices (Brady’s nontriviality proof, the collapsing
lemma).
This might all seem a bit off kilter, if Routley is right that
It is not satisfactory to reject classical logic systematically, e.g. as involving

mistakes or illegitimate assumptions, and to use it metasystematically with-
out further ado or qualification; for to do so would be to proceed by what
are confessedly mistaken paths. Such choices are not generally satisfactory:
they fail to cohere. (Routley, 1980a, p. 94)
Various people, both friends and critics, have wondered aloud about this issue.
It turns, once again, on whether paraconsistency is ultimately at odds with
standard mathematics.
Many working in this area have suggested this worry is somehow misguided,
that we do not need to think of paraconsistency as a dispute with classicality,
78 More than in previous sections, I will be articulating a position that is, I suspect, not represen-
tative of the paraconsistent community (such that there is one).

so the use of classical methods is fine. If there is a dispute here, what sort of
disagreement would it be? Warren distinguishes between descriptive disputes
about which logic is used in some particular natural language; normative dis-
putes about which logic should be used in some particular natural language; and
metaphysical disputes about which logic is objectively correct in a language
independent sense (Warren, 2018, p. 437). Let us consider these.
The original Meyer/Routley/Priest approach, at least, was a serious descrip-
tive dispute about the correct formalization of mathematical practice, as we saw
in §1.2.3. Routley bragged “we can do everything you can do, only better, and
we can do more (Routley, 1977, p. 929).”
Circa the first quarter of the twenty-first century, this has not come to pass.
The idea that the classical and paraconsistent truths would somehow coincide,
or that the paraconsistent truths might completely subsume the classical, has
been painstakingly tested and appears to be out of reach. There is currently
no consensus on what logic people really are using (if there even is an answer
to this question); Gillian Russell’s logical nihilism suggests there simply is no
descriptive logic of natural language (Russell, 2018). But I think it is reasonable
to say that the early Meyer et al. hypothesis has not been vindicated; various
attempts to reconstruct standard proofs in paraconsistent logics have repeatedly
failed.79 Ordinary mathematical practice may not be classical, but hard-won
evidence would suggest it is not straightforwardly paraconsistent, either. There
was a descriptive dispute here about the logic of mathematical proofs, and I do
not think the classical side unequivocally won, but the relevant paraconsistent
side did not either.
What about sidestepping this with the idea of mathematical truth as invari-
ant to logic, by making logic secondary to mathematics? My sense, after long
exposure to useage at odds with classicality, is that the idea of mathematical
truth as invariant or prior to logic (suggested by Mortensen, but also Brouwer)
looks harder and harder to maintain. If one is looking at classical manipulations
of classical models, or translations into “classicalese” (as Meyer once called it)
then yes, things may all seem to exhibit some similarities. But taken more on
their own terms, without classical representations, I see very little by way of
invariance; even small decisions about the phrasing of axioms in set theory
or real analysis, as we have observed, will lead to dramatic differences down-
stream. Similarly, for the idea of some kind of “commonsense” cutting across
79 Much of this evidence is unpublished, since it involves unsuccessful work. See Priest (2006, p.
221), Istre and McKubre-Jordens (2019). Note that the proof of Cantor’s theorem in §2.3 is not
the classical proof but a new proof.

different logics (as Shapiro suggests in (Shapiro, 2014, ch. 7)), friendly though
the sentiment may be, the subtle differences between weak logics tend to render
what was once classically familiar as multitudinous, strange, and new, a chal-
lenge to our old ways. Saying that maybe some natural number n = n + 1 is no
small thing, more than some kind of idiosyncrasy or regional dialect.
If there is still a dispute here, then, what is left open are the normative and
metaphysical lines, the idea that paraconsistency in mathematics can be a guide
to new truths – truths that are currently on the far side of what is currently
acceptable. Without being polemical or trying to pick a fight, I want to suggest
that something closer to this radical, rivalrous understanding of the situation
makes more sense of some of the efforts that have gone into this area. Meyer
says Hilbert “should” have used relevant arithmetic, because Meyer thinks rel-
evant arithmetic is true arithmetic. Meyer might have been wrong, but it is
more charitable to say so than to tell a story about why he really meant some-
thing more agreeable or pluralistic.80 Seeing things this way may be ultimately
healthier for the growth of all fields involved. Classicists and nonclassicists
alike stand to learn more if we think that there is, indeed, some truth-seeking
activity going on, and by comparing notes, especially notes that do not always
agree, we might get closer to truth.
Whither paraconsistency in mathematics? A very standard story goes that
a mathematical theorem, once proven, cannot be undone. Mathematics only
accumulates, generations building on previous generations, like a tower. And
it is a very good tower, perhaps one of the finest things humans have to show
for themselves. If you like that story, there is a clear place for paraconsis-
tency in mathematics. The idea that paraconsistent logics are a generalization
of classical logic is right there in the semantics. For observe:
Every classical tautology is an LP tautology.
Proof: suppose p is not an LP tautology; there is some three-valued assignment

making it untrue; modify this assignment to a two valued one by changing any
b to f; by checking the truth tables, p it is not a classical tautology, either, QED
(Priest, 1979, p. 228). And any argument that is LP invalid, but classically valid,
can be made LP valid by generalizing, by making explicit the case ignored by
classicality – namely, the possibility of inconsistency. The argument
p ⊃ q, q ⊃ r therefore p ⊃ r
80 This hews very close to detailed debates in logical pluralism about the role of normativity, which
I cannot do justice here. See Steinberger (2019). Priest separates mathematical pluralism from
logical pluralism, leaving a place for truth, in Priest (2021). Cf. Meadows and Weber (2016).

is classically but not LP valid; but the argument
p ⊃ q, q ⊃ r therefore (p ⊃ r) ∨ (q ∧ ¬q)
is LP valid. Then an LP theory can similarly be transformed back to a

classical one by discarding the inconsistent case (cf. Beall, 2011). Paracon-
sistency extends into the transconsistent leaving the consistent undisturbed.
Simpatico.
The story that mathematics is only cumulative, never revisable, though, has
been credibly challenged. Lakatos argues that more careful attention to the
actual history and details of mathematical development show a much mes-
sier process, prone to missteps and corrections, that are only later cleaned up
for posterity (Lakatos, 1976). There is room for sober, careful dispute of what
appears to be established theory. That is how the tower of mathematics came
to be; that is how it will keep being good.
What we come to is that, methodologically, the stance one takes here about
revision is in part dependent on what “meta” logic one adopts. This can effect
even what the facts are, facts like “paraconsistency generalizes classical logic.”
Apropos: the argument that all classical tautologies are LP tautologies is proven
by contraposition. Any LP counterexample can be converted into a classical
counterexample, and therefore by contraposition all classical theorems (with
no counterexample) are LP theorems (with no counterexample). But the con-
sequence relation of LP does not contrapose, and the language of LP does not
express “no counterexample” in the way needed for the proof.81 For a para-
consistent logician who “takes such reasoning not to be correct, at least in part,
things cannot be left like this. The classical ladder must, so to speak, be kicked
away” (Priest, 2008, p. 585).
So, absent some other way of proving the result, a committed LP-
paraconsistentist cannot after all take it as writ that all classical tautologies are
valid. What should they say? That way lies uncharted waters. But many of us
would not be where we are today if people had not, at various times, set sail
into uncharted waters.
If one is not persuaded that inconsistent mathematics might, at the end of
the day, be true, and so in a basic logical sense at odds with classical math-
ematics, then it is not clear to me at least what the project here is. Why do
something so difficult, unappreciated, and weird when compared to its classical
81 Even if there is no counterexample, maybe there is still a counterexample. See Omori and Weber
(2019) and Weber (2021, ch. 10).

counterpart?82 If one is persuaded to pursue some inconsistent mathematics

in the expectation that it might be true, then it is not clear that classicality-
preserving strategies are available, or how they would work (cf. Weber, 2021,
ch. 3). That might be a more uncomfortable position than a would-be paracon-
sistent mathematician meant to sign up for, but pursing theories for their truth
can be uncomfortable.
A future for paraconsistency in mathematics, I submit, is to follow out a
more radical project. A future radical paraconsistentist is committed to finding
new rules, overturning some old ones, and risks not being understood (literally)
by others. More conservative projects in paraconsistency are subsumed under
standard mathematics, and so will remain on safe, previously trodden ground.
The more radical approach is to embark on a project that is not safe and may
ultimately fail; but that is what makes it research, and that is what holds out the
chance of achieving something genuinely new.
82 An Ivy League professor took me aside and asked me this very question when I was still a
graduate student. In this footnote, I am taking you aside and asking you, even though I am not
an Ivy League professor, and I did not listen to him anyway.

References
Anderson, A. R., & Belnap, N. D. (1975). Entailment: The logic of relevance
and necessity (Vol. 1). Princeton: Princeton University Press.
Arruda, A. I. (1979). The Russell Paradox in the systems NFn. In A. Arruda, N.
da Costa, & A. Sette (Eds.), Proceedings of the third Brazilian conference on
mathematical logic (pp. 1–12). Campinas: Sociedade Brasileira de Logica.
Arruda, A. I. (1989). Aspects of the historical development of paraconsistent
logic. In G. Priest, R. Routley, & J. Norman (Eds.), Paraconsistent logic:
Essays on the Inconsistent (pp. 99–130). Munich: Philosophia Verlag.
Arruda, A. I., & Batens, D. (1982). Russell’s set versus the universal set in
paraconsistent set theory. Logique et Analyse, 25(8), 121–133.
Asenjo, F. G. (1966). A calculus of antinomies. Notre Dame Journal of Formal
Logic, 7, 103–105.
Asenjo, F. G. (1989). Toward an antinomic mathematics. In G. Priest, R. Rout-
ley, & J. Norman (Eds.), Paraconsistent logic: Essays on the Inconsistent
(pp. 394–414). Munich: Philosophia Verlag.
Asenjo, F. G., & Tamburino, J. (1975). Logic of antinomies. Notre Dame
Journal of Formal Logic, 16(1), 17–44.
Badia, G., Weber, Z., & Girard, P. (2022). Paraconsistent metatheory:
New proofs with old tools. Journal of Philosophical Logic https://link
.springer.com/article/10.1007/s10992-022-09651-x
Balaguer, M. (1998). Platonism and anti-platonism in mathematics. Oxford:
Oxford University Press.
Barrio, E., & Da Re, B. (2018). Paraconsistency and its philosophical interpre-
tations. Australasian Journal of Logic, 15(2), 151–170.
Barrio, E., Pailos, F., & Szmuc, D. (2017). A paraconsisitent route to semantic
closure. Logic Journal of the IGPL, 25(4), 387–401.
Batens, D. (2020). Adaptive Fregean set theory. Studia Logica, 108, 903–939.
Beall, J. (1999). From full blooded platonism to really full blooded platonism.
Philosophia Mathematica, 7(3), 322–325.
Beall, J. (2009). Spandrels of truth. Oxford: Oxford University Press.
Beall, J. (2011). Multiple-conclusion LP and default classicality. Review of
Symbolic Logic, 4(2), 326–336.
Beall, J. (2013a). Free of detachment: Logic, rationality, and gluts. Noûs, 49(2),
410–423.
Beall, J. (2013b). Shrieking against gluts: the solution to the ‘just true’ problem.
Analysis, 73(3), 438–445.

72 References
Beall, J. (2018). The simple argument for subclassical logic. Philosophical

Issues, 28(1), 30–54.
Beall, J., Forster, T., & Seligman, J. (2013). A note on freedom from detachment
in the logic of paradox. Notre Dame Journal of Formal Logic, 54(1), 15–20.
Beall, J., & Ripley, D. (2018). Nonclassical theories of truth. In M. Glanzberg
(Ed.), Oxford handbook of truth (pp. 739–754). Oxford: Oxford University
Press.
Bell, J. L. (2005). Set theory: Boolean-valued models and independence proofs.
Oxford: Oxford University Press.
Bell, J. L. (2008). A primer of infinitesimal analysis (2nd ed.). Cambridge:
Cambridge University Press.
Belnap, N. D. (1977). A useful four-valued logic. In J. Dunn & G. Epstein
(Eds.), Modern uses of multiple-valued logics (pp. 8–37). Dordrecht: Reidel.
Berto, F. (2007). How to sell a contradiction: The logic and metaphysics of
inconsistency. London: College.
Brady, R. (1971). The consistency of the axioms of the axioms of abstraction
and extensionality in a three-valued logic. Notre Dame Journal of Formal
Logic, 12, 447–453.
Brady, R. (1989). The non-triviality of dialectical set theory. In G. Priest,
R. Routley, & J. Norman (Eds.), Paraconsistent Logic: Essays on the
inconsistent (pp. 437–470). Munich: Philosophia Verlag.
Brady, R. (2006). Universal logic. Stanford: CSLI.
Brady, R., & Mortensen, C. (2014). Logic. In G. Oppy & N. Trakakis (Eds.),
History of philosophy in Australia and New Zealand (pp. 679–705). Dor-
drecht: Springer.
Brady, R., & Routley, R. (1989). The non-triviality of extensional dialectical set
theory. In G. Priest, R. Routley, & J. Norman (Eds.), Paraconsistent logic:
Essays on the inconsistent (pp. 415–436). Munich: Philosophia Verlag.
Brady, R., & Rush, P. (2008). What is wrong with Cantor’s diagonal argument?
Logique et Analyse, 51(202), 185–219.
Brown, B., & Priest, G. (2004). Chunk and permeate, a paraconsistent inference
strategy. Part I: The infinitesimal calculus. Journal of Philosophical Logic,
33, 379–388.
Brown, J. (2008). Philosophy of mathematics (2nd ed.). New York: Routledge.
Burgess, J. P. (2005). No requirement of relevance. In S. Shapiro (Ed.), The
Oxford handbook of philosophy of mathematics and logic (pp. 727–750).
Carnielli, W., Coniglio, M., & Marcos, J. (2007). Logics of formal inconsist-
ency. In D. Gabbay & F. Guenthner (Eds.), Handbook of philosphical logic
(Vol. 14, pp. 1–93). Dordrecht: Springer Verlag.

References 73
Carnielli, W., & Coniglio, M. E. (2016). Paraconsistent logic: Consistency,

contradiction and negation. Cham, Switzerland: Springer.
Carnielli, W., & Coniglio, M. E. (2021). Twist-valued models for three-
valued paraconsistent set theory. Logic and Logical Philosophy, 30(2),
187–226.
Carrara, M., & Martino, E. (2021). A note on Gödel, Priest and naïve proof.
Logic and Logical Philosophy, 30, 79–96.
Colyvan, M. (2008). The ontological commitments of inconsistent theories.
Philosophical Studies, 141(1), 115–23.
Colyvan, M. (2009). Applying inconsistent mathematics. In O. Bueno and
Ø. Linnebo (Eds.) New waves in philosophy of mathematics pp. 160–172.
London: Palgrave MacMillan.
Colyvan, M. (2012). An introduction to the philosophy of mathematics. Cam-
bridge: Cambridge University Press.
Copeland, B., & Sylvan, R. (1999). Beyond the universal turing machine.
Australasian Journal of Philosophy, 77, 46–66.
da Costa, N. C. (1986). On paraconsistent set theory. Logique et Analyse,
29(15), 361.
da Costa, N. C., Krause, D., & Bueno, O. (2007). Paraconsistent logics and
paraconsistency. In D. Jacquette (Ed.), Philosophy of logic (pp. 791–912).
Amsterdam: North Holland.
da Costa, N. C. A. (1974). On the theory of inconsistent formal systems. Notre
Dame Journal of Formal Logic, 15, 497–510.
Dedekind, R. (1901). Essays on the theory of numbers. New York: Dover
(Edited and translated by W. W. Beman, 1901. Includes Stetigkeit und irra-
tionale Zahlen [1872] and Was sind und was sollen die Zahlen? [1888].
Dover ed. 1963).
Denyer, N. (1995). Priest’s paraconsistent arithmetic. Mind, 104, 567–575.
D’Ottaviano, I., & Carvalho, T. (2005). Da Costa’s paraconsistent differential
calculus and the transference theorem. 2nd Indian International Conference
on Artificial Intelligence, 1659–1678.
Dunn, J. M. (1976). Intuitive semantics for first-degree entailments and “cou-
pled trees.” Philosophical Studies, 29, 149–168.
Dunn, J. M. (1979). A theorem in 3-valued model theory with connec-
tions to number theory, type theory, and relevant logic. Studia Logica, 38,
149–169.
Dunn, J. M. (1980). Relevant Robinson’s arithmetic. Studia Logica, 38, 407–
418.
Estrada-González, L. (2010). Complement-topoi and dual intuitionistic logic.
Australasian Journal of Logic, 9, 26–44.

74 References
Estrada-González, L. (2015). From (paraconsistent) topos logic to universal

(topos) logic. In A. Koslow & A. Buchsbaum (Eds.), The road to universal
logic (pp. 263–295). Basel, Switzerland: Birkhauser.
Estrada-González, L. (2016). Prospects for triviality. In H. Andreas & P.
Verdée (Eds.), Logical studies of paraconsistent reasoning in science and
mathematics (pp. 81–90). Cham, Switzerland: Springer.
Ferguson, T. M. (2016). On arithmetic formulated connexively. IfCoLog Jour-
nal of Logics and their Applications, 3, 357–376.
Ferguson, T. M. (2019). Inconsistent models (and infinite models) for arith-
metics with constructible falsity. Logic and Logical Philosophy, 28, 389–
407.
Forster, T. (1995). Set theory with a universal set. Oxford: Clarendon Press.
Franks, C. (2009). The autonomy of mathematical knowledge. Cambridge:
Friedman, H., & Meyer, R. (1992). Whither relevant arithmetic? Journal of
Symbolic Logic, 57, 824–831.
Goldblatt, R. (1998). Lectures on the hyperreals. Berlin: Springer.
Gomes, E., & D’Ottaviano, I. (forthcoming). Illuminating contradictions: A
history of paraconsistent logic. Synthese Library, Springer.
Goodship, L. (1996). On dialethism. Australasian Journal of Philosophy, 74(1),
153–161.
Halmos, P. (1974). Naive set theory. New York: Springer.
Istre, E., & McKubre-Jordens, M. (2019). The difficulties in using weak rel-
evant logics for naive set theory. In C. Baskent & T. Ferguson (Eds.),
Graham Priest on dialetheism and paraconsistency (pp. 365–381). Cham,
Switzerland: Springer.
Jacquette, D. (Ed.). (2007). Philosophy of logic. Amsterdam: North Holland.
Jech, T. (2003). Set theory (Third millennium edition, revised and expanded).
Berlin: Springer.
Jockwich-Martinez, S., & Venturi, G. (2021). Non-classical models of ZF.
Studia Logica, 109, 509–537.
Lakatos, I. (1976). Proofs and refutations: The logic of mathematical discovery.
Cambridge: Cambridge University Press.
Leitgeb, H. (2009). On formal and informal provability. In New waves in
philosophy of mathematics (pp. 263–299). New York: Palgrave Macmillan.
Libert, T. (2005). Models for paraconsistent set theory. Journal of Applied
Logic, 3, 15–41.
Lisboa, M. A., & Secco, G. D. (2022). History of logic in Latin America: The
case of Ayda Ignez Arruda. British Journal of History of Philosophy, 30(2),
384–408.

References 75
Löwe, B., & Tarafder, S. (2015). Generalized algebra-valued models of set

theory. Review of Symbolic Logic, 8(1), 209–225.
Maddy, P. (1992). Indispensability and practice. Journal of Philosophy, 89(6),
275–289.
Mancosu, P. (2009). Measuring the size of infinite collections of natural
numbers. Review of Symbolic Logic, 2(4), 612–646.
Marcos, J. (2005). Logics of formal inconsistency. Brazil: Fundaçäc Biblioteca
Nacional.
McKubre-Jordens, M., & Weber, Z. (2012). Real analysis in paraconsistent
logic. Journal of Philosophical Logic, 41(5), 901–922.
McKubre-Jordens, M., & Weber, Z. (2016). Paraconsistent measurement of the
circle. Australasian Journal of Logic, 14(1), 268–280.
Meadows, T., & Weber, Z. (2016). Computation in non-classical foundations?
Philosophers’ Imprint, 16, 1–17.
Meyer, R. K. (1985). Proving semantical completeness “relevantly” for R.
Logic Research Paper(23), RSSS, Australian National University.
Meyer, R. K. (1998). ⊃-E is admissible in “true” relevant arithmetic. Journal
of Philosophical Logic, 27, 327–351.
Meyer, R. K. (2021a). Arithmetic formulated relevantly. Australasian Journal
of Logic, 18(5), 154–288 (typescript circa 1976).
Meyer, R. K. (2021b). The consistency of arithmetic. Australasian Journal of
Logic, 18(5), 289–379 (typescript circa 1976).
Meyer, R. K. (2021c). Relevant arithmetic. Australasian Journal of Logic,
18(5), 150–153. (Abstract originally published in the Bulletin of the Section
of Logic 1976).
Meyer, R. K., & Mortensen, C. (1984). Inconsistent models for relevant
arithmetics. Journal of Symbolic Logic, 49, 917–929.
Meyer, R. K., & Mortensen, C. (1987). Alien intruders in relevant arithmetic.
Technical Report, TR-ARP(9/87).
Meyer, R. K., & Restall, G. (1999). “Strenge” arithmetics. Logique et Analyse,
42, 205–220.
Meyer, R. K., & Urbas, I. (1986). Conservative extension in relevant arithmetic.
Mathematical Logic Quarterly, 32(1–5), 45–50.
Moore, G. H. (1982). Zermelo’s axiom of choice. New York: Springer Verlag.
Mortensen, C. (1988). Inconsistent number systems. Notre Dame Journal of
Formal Logic, 29, 45–60.
Mortensen, C. (1995). Inconsistent mathematics. Dordrecht: Kluwer Aca-
demic.
Mortensen, C. (2009). Inconsistent mathematics: Some philosophical
implications. In A. Irvine (Ed.), Handbook of the philosophy of science,

76 References
volume 9: Philosophy of mathematics (pp. 631–649). Amsterdam: North

Holland.
Mortensen, C. (2010). Inconsistent geometry. London: College.
Mortensen, C. (2022). The impossible arises. Bloomington: Indiana University
Press.
Mortensen, C., & Leishman, S. (2009). Linear algebra representation of Necker
cubes I: The crazy crate. Australasian Journal of Logic, 7, 1–9.
Nelson, D. (1959). Negation and separation of concepts in constructive sys-
tems. In A. Heyting (Ed.), Constructivity in mathematics (pp. 208–225).
Amsterdam: North Holland.
Øgaard, T. F. (2016). Paths to triviality. Journal of Philosophical Logic, 45(3),
237–276.
Omori, H. (2015). Remarks on naive set theory based on LP. Review of
Omori, H., & De, M. (2022). Shrieking, shrugging, and the Australian plan.
Notre Dame Journal of Formal Logic, (forthcoming).
Omori, H., & Wansing, H. (2019a). Connexive logics. an overview and current
trends. Logic and Logical Philosophy, 28, 371–387.
Omori, H., & Wansing, H. (Eds.). (2019b). New essays on Belnap-Dunn logic.
Cham, Switzerland: Synthese Library, Springer.
Omori, H., & Weber, Z. (2019). Just true? On the metatheory for paraconsistent
truth. Logique et Analyse, 248, 415–433.
Paris, J., & Sirokofskich, A. (2008). On LP-models of arithmetic. Journal of
Potter, M. (2004). Set theory and its philosophy. Oxford: Clarendon.
Priest, G. (1979). The logic of paradox. Journal of Philosophical Logic, 8, 219–
241.
Priest, G. (1994a). Is arithmetic consistent? Mind, 103 (411), 337–349.
Priest, G. (1994b). What could the least inconsistent number be? Logique et
Analyse, 145, 3–12.
Priest, G. (1997). Inconsistent models of arithmetic part I: Finite models.
Journal of Philosophical Logic, 26, 223–235.
Priest, G. (2000). Inconsistent models of arithmetic part II: The general case.
Journal of Symbolic Logic, 65, 1519–1529.
Priest, G. (2005). Towards non-being. Oxford: Oxford University Press.
Priest, G. (2006). In contradiction: A study of the transconsistent (2nd ed.).
Priest, G. (2008). An introduction to non-classical logic (2nd ed.). Cambridge:
Priest, G. (2013). Mathematical pluralism. Journal of the IGPL, 21, 4–13.

References 77
Priest, G. (2014). One. Oxford: Oxford University Press.

Priest, G. (2017a). A note on the axiom of countability. If CoLog Jour-
nal of Logics and their Applications, 4, 1351–1356. (First published in
Al-Mukhatabat 2012).
Priest, G. (2017b). What if? The exploration of an idea. Australasian Journal
of Logic, 14(1), 55–127.
Priest, G. (2019). Some comments and replies. In C. Baskent & T. Ferguson
(Eds.), Graham Priest on dialetheism andparaconsistency (pp. 575–675).
Cham, Switzerland: Springer.
Priest, G. (2021). A note on mathematical pluralism and logical pluralism.
Synthese, 198, 4937–4946.
Priest, G., Routley, R., & Norman, J. (Eds.). (1989). Paraconsistent logic:
Essays on the inconsistent. Munich: Philosophia Verlag.
Rayo, A., & Uzquiano, G. (Eds.). (2006). Absolute generality. Oxford: Oxford
University Press.
Restall, G. (1992). A note on naïve set theory in LP. Notre Dame Journal of
Formal Logic, 33, 422–432.
Rosenblatt, L. (2021a). Expressing consistency consistently. Thought, 10(1),
33–41.
Rosenblatt, L. (2021b). Towards a non-classical meta-theory for substructural
approaches to paradox. Journal of Philosophical Logic, 50, 1007–1055.
Rosenblatt, L. (2022). Should the non-classical logician be embarassed.
Philosopy and Phenomenological Research, 104(2), 388–407.
Routley, R. (1977). Ultralogic as universal? (First appeared in two parts in
The Relevance Logic Newsletter 2(1), 51–90, January 1977 and 2(2), 138–
75, May 1977; reprinted as appendix to Exploring Meinong’s Jungle and
Beyond 1980, pp. 892–962; new edition as The Sylvan Jungle, vol. 4, edited
by Z. Weber, Synthese Library, 2019).
Routley, R. (1979). Dialectical logic, semantics and metamathematics. Erken-
ntnis, 14, 301–331.
Routley, R. (1980a). The choice of logical foundations: Non-classical choices
and the ultralogical choice. Studia Logica, 39(1), 77–98.
Routley, R. (1980b). Exploring Meinong’s jungle and beyond. Canberra: Phi-
losophy Department, RSSS, Australian National University (Departmental
Monograph number 3. Reprinted in four volumes as The Sylvan Jungle,
edited by Dominic Hyde et al., Synthese Library, 2019–2020).
Routley, R., & Meyer, R. K. (1976). Dialectical logic, classical logic and the
consistency of the world. Studies in Soviet Thought, 16, 1–25.
Russell, B. (1903). The principles of mathematics. London: George Allen &
Unwin.

78 References
Russell, G. (2018). Logical nihilism: could there be no logic? Philosophical

Issues, 28(1), 308–24.
Schotch, P., Brown, B., & Jennings, R. (Eds.). (2009). On preserving: Essays
on preservationism and paraconsistent logic. Toronto: University of Toronto
Press.
Shapiro, S. (1998). Incompleteness, mechanism, and optimism. Bulletin of
Shapiro, S. (2002). Incompleteness and inconsistency. Mind, 111, 817–832.
Shapiro, S. (2014). Varieties of logic. Oxford: Oxford University Press.
Shapiro, S., & Wright, C. (2006). All things indefinitely extensible. In A. Rayo
& G. Uzquiano (Eds.) Absolute generality, (pp. 255–304). Oxford: Oxford
University Press.
Spivak, M. (1999). A comprehensive introduction to differential geometry (3rd
ed., 4 vols.). Berkley: Publish or Perish.
Steinberger, F. (2019). Logical pluralism and logical normativity. Philoso-
pher’s Imprint, 19(12), 1–19.
Sylvan, R., & Copeland, J. (2000). Computability is logic relative. In D. Hyde
& G. Priest (Eds.), Sociative logics and their applications (pp. 189–199).
Surrey: Ashgate.
Tanswell, F. (2016). Saving proof from paradox: Gödel’s paradox and the
inconsistency of informal mathematics. In H. Andreas & P Verdée (Eds.),
Logical studies of paraconsistent reasoning in science and mathematics (pp.
159–174). Cham, Switzerland: Springer.
Tedder, A. (2015). Axioms for finite collapse models of arithmetic. Review of
Tedder, A. (2021). On consistency and decidability in some paraconsistent
arithmetics. Australasian Journal of Logic, 18(5), 473–502.
Thomas, M. (2014a). A conjecture about the interpretation of classical math-
ematics in naive set theory. Presented at the Paraconsistent Reasoning in
Science and Mathematics conference, LMU 2014.
Thomas, M. (2014b). Expressive limitations of naive set theory in LP and
minimally inconsistent LP. Review of Symbolic Logic, 7(2), 341–350.
Thro, E. B. (1983). Distinguishing two classes of impossible objects. Percep-
tion, 12, 733–751.
van Bendegem, J. P. (2003). Classical arithmetic is quite unnatural. Logic and
Logical Philosophy, 11, 231–249.
van Benthem, J. (1978). Four paradoxes. Journal of Philosophical Logic, 7(1),
49–72.
van Heijenoort, J. (Ed.). (1967). From Frege to Gödel: A source book in
mathematical logic, 1879–1931. Cambridge, MA: Harvard University Press.

References 79
Verdée, P. (2013a). Non-monotinic set theory as a pragmatic foundation of

mathematics. Foundations of Science, 18, 655–680.
Verdée, P. (2013b). Strong, universal and provably non-trivial set theory by
means of adaptive logic. Logic Journal of the IGPL, 21, 108–125.
Warren, J. (2018). Change of logic, change of meaning. Philosophy and
Phenomenological Research, 96(2), 421–442.
Weber, Z. (2010). Transfinite numbers in paraconsistent set theory. Review of
Weber, Z. (2012). Transfinite cardinals in paraconsistent set theory. Review of
Weber, Z. (2016). Paraconsistent computation and dialetheic machines. In H.
Anreas & P. Verdée (Eds.), Logical studies of paraconsistent reasoning in
science and mathematics (pp. 205–223). Cham, Switzerland: Springer.
Weber, Z. (2021). Paradoxes and inconsistent mathematics. Cambridge: Cam-
bridge University Press.
Weber, Z., Badia, G., & Girard, P. (2016). What is an inconsistent truth table?
Australasian Journal of Philosophy, 94(3), 533–548.
Weber, Z., & Cotnoir, A. (2015). Inconsistent boundaries. Synthese, 192(5),
1267–1294.
Weir, A. (2004). There are no true contradictions. In G. Priest, J. C. Beall, & B.
Armour-Garb (Eds.), The law of non-contradiction (pp. 385–417). Oxford:
Clarendon Press.
Whitehead, A. N., & Russell, B. (1910). Principia Mathematica. Cambridge:
Cambridge University Press. (in three volumes, 1910–1913).

Acknowledgments
Thanks to Stewart Shapiro and Penelope Rush for the opportunity to write this.
Thanks to Guillermo Badia, Thomas Ferguson, Patrick Girard, Franci Man-
graviti, Ed Mares, Hitoshi Omori, and Graham Priest for helpful comments,
corrections, and discussions.

The Philosophy of Mathematics
Penelope Rush
University of Tasmania
From the time Penny Rush completed her thesis in the philosophy of mathematics
(2005), she has worked continuously on themes around the realism/anti-realism divide
and the nature of mathematics. Her edited collection The Metaphysics of Logic
(Cambridge University Press, 2014), and forthcoming essay “Metaphysical Optimism”
(Philosophy Supplement), highlight a particular interest in the idea of reality itself and
curiosity and respect as important philosophical methodologies.
Stewart Shapiro
The Ohio State University
Stewart Shapiro is the O’Donnell Professor of Philosophy at The Ohio State University, a
Distinguished Visiting Professor at the University of Connecticut, and a Professorial
Fellow at the University of Oslo. His major works include Foundations without
Foundationalism (1991), Philosophy of Mathematics: Structure and Ontology (1997),
Vagueness in Context (2006), and Varieties of Logic (2014). He has taught courses in logic,
philosophy of mathematics, metaphysics, epistemology, philosophy of religion, Jewish
philosophy, social and political philosophy, and medical ethics.
About the Series

This Cambridge Elements series provides an extensive overview of the philosophy of
mathematics in its many and varied forms. Distinguished authors will provide an
up-to-date summary of the results of current research in their fields and give their own
take on what they believe are the most significant debates influencing research, drawing
original conclusions.

The Philosophy of Mathematics
Elements in the Series

Mathematical Structuralism
Geoffrey Hellman and Stewart Shapiro
A Concise History of Mathematics for Philosophers
John Stillwell
Mathematical Intuitionism
Carl J. Posy
Innovation and Certainty
Mark Wilson
Semantics and the Ontology of Number
Eric Snyder
Wittgenstein’s Philosophy of Mathematics
Juliet Floyd
Ontology and the Foundations of Mathematics: Talking Past Each Other
Penelope Rush
Mathematics and Metaphilosophy
Justin Clarke-Doane
Zach Weber
A full series listing is available at: www.cambridge.org/EPM

Zach Weber-Paraconsistency - in - Mathematics

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Zach Weber-Paraconsistency - in - Mathematics

Uploaded by

Copyright:

Available Formats

Weber

Paraconsistent logic makes it possible to study inconsistent

Cover image: CTRd / Getty Images ISSN 2399-2883 (online)

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

Cambridge University Press is part of the University of Cambridge.

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

Elements in the Philosophy of Mathematics

Abstract: Paraconsistent logic makes it possible to study inconsistent

Keywords: nonclassical mathematics, inconsistent mathematics,

© Zach Weber 2022

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

1 Invitation to Paraconsistency in Mathematics:

4 Calculus, Topology, and Geometry 53

5 Whither Paraconsistency in Mathematics? 65

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

1 Invitation to Paraconsistency in Mathematics:

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

So we seem to have shown that

as an apparently natural consequence of some apparently true premises. There

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

paraconsistency. Whether or not we should think of paraconsistent theories as

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

nontrivial, it must, it seems, be embedded in a paraconsistent logic, where the

ν(¬A) is at least t iff ν(A) is at least f

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

ν(A ∨ B) is at least t iff ν(A) is at least t or ν(B) is at least t

On this account of negation, a sentence is evaluated as true iff its negation is

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

a space, then, for different approaches to paraconsistency in mathematics,

1.2.2 A Very Brief History of Paraconsistency in Mathematics

14 Interested readers may consult Gomes and D’Ottaviano (forthcoming).

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

of what is now sometimes called inconsistent mathematics. These included

1.2.3 Goals: Recapture, Expansion, Revision

16 Interested readers may consult Brady and Mortensen (2014).

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

In order to sustain the [paraconsistent] challenge to classical logic it will have

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

Insofar as mathematics relies on valid argument, its proper formalization

A more mainstream view is that standard mathematics is couched in classi-

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

there is no point in advocating the replacement of classical logic with para-

Such a take on paraconsistency is clearly not mounting any “challenge” to clas-

The properties of inconsistent theories tend to be invariant over a large class

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

So a more radical way still is to say that paraconsistency in mathematics may

1.2.4 Into the Paraconsistent

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

26 See “Concluding Self-Referential Postscript” to Priest (1979).

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

one does have “gentle” explosion,

p, ¬p, ◦(p) ⊢LFI q.

An explosive negation is definable: it holds when ¬p and ◦(p) both hold.

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

More widely, the idea of assuming or “recapturing” various bits of classical-

• A “lower limit logic” that is the stable, static base logic

https://doi.org/10.1017/9781108993968 Published online by Cambridge University Press

or are otherwise consistent relative to some classical structure (see Appendix

1.3 Realisms and Anti-realisms About (Inconsistent)

33 See Burgess (2005) and Weir (2004).