Similarity Report ID: oid:30225:461706291

PAPER NAME

Arbitcoin03450_Essay Course - First Draf

t-1.pdf

WORD COUNT CHARACTER COUNT

7948 Words 41945 Characters

PAGE COUNT FILE SIZE

32 Pages 415.7KB

SUBMISSION DATE REPORT DATE

May 23, 2025 4:35 PM UTC May 23, 2025 4:37 PM UTC

9% Overall Similarity
The combined total of all matches, including overlapping sources, for each database.
5% Internet database 7% Publications database
Crossref database Crossref Posted Content database
2% Submitted Works database

Excluded from Similarity Report

Bibliographic material Quoted material

Summary
Logical Possibility and Modal
Structuralism in Mathematics

Section 0: Introduction and Scope of Essay

Mathematical analyses can broadly be seen to fall in two categories: one tries to see
how a given mathematical object behaves within a particular theory, whereas the other
tries to elucidate the nature of the mathematical object in question.

The former analysis, which is more mathematical in nature, deals with providing a
definite characterization of a theory in the form of axioms. However, the specific logical
language employed for these axiomatizations have deep philosophical impact on their
accuracy and usefulness.

Structuralism in the philosophy of mathematics treats mathematical theories as

characterizing abstract structures as opposed to mathematical objects. This follows a
much longer trend in mathematical practice, originating at least with Dedekind’s work on
9
arithmetic. Under a structuralist paradigm, set theory is seen as a generalized theory of
what structures are, and how they relate to one another.

There are significant contemporary debates about the ontological status of these
abstract structures, whether they are objects in their own rights(Shapiro’s ante rem
structuralism), or whether existence claims for these structures can be seen as modal
claims of their instantiations, and the properties which’d be expected therein.

This essay looks at the latter of the two positions.

Section I looks at attempts to axiomatize arithmetic, which is mainly seen as a
mathematical enterprise, but runs into some significant philosophical issues when
considering the underlying logic to be used. After elucidating these concerns, Section II
9
takes a step back and looks at set theory as a unifying foundation for mathematics.
Structuralist arguments for mathematics
-​ Then we will look at similar attempts in axiomatizing set theory, and see
how those differences become acute in this specific case
-​ We’ll look at the beginning of modal structuralism, with Putnam and
Hellman, and see their specific merits and demerits. We’ll lay out the
notion of logical possibility that underpins their theories
1
-​ We’ll dive deeper into the notion of logical possibility as a semantic
primitive, interdefinable with logical validity. Since not a lot of work has
been done directly defending this notion, we’ll primarily look at work
critiquing the reductionist analyses of logical validity, and infer what can be
said for logical possibility.
-​ We’ll discuss Berry’s proposal of potentialist set theory reducible to a
first-order modal logic. Specifically, we’ll look at the innovation of structure
preserving logical possibility as a concept and a formal operator.
-​ We’ll briefly touch upon some other key features of Berry’s proposal and
some further lines of inquiry before concluding.

Keywords: foundationalism, structuralism, logical possibility, modal logic, primitives, …

Section I: Axiomatization of Arithmetic
When we’re axiomatizing arithmetic, we’re treating numbers as a given category, and
characterizing how the system of arithmetic operates. This is to be differentiated with
what may be called an ontological analysis, which looks at the category of numbers and
determines whether they are primitive platonic objects, or if they can be analysed into
more primitive notions.

Dedekind-Peano Arithmetic​ ​ ​ ​ ​ ​
​ ​ ​ ​ ​
Dedekind developed an axiomatization for natural number arithmetic using the idea that
every natural number can be generated by applying a successor relation repeatedly
from an initial number, called zero. This formulation, simplified by Peano, is called the
second-order Peano axiomatization, or Dedekind-Peano Arithmetic. The following is a
recasting of Peano’s axioms in a logical language with a successor function(S) and a
constant predicate for natural numbers(N):

12
1.​ There is a natural number which is not the successor to any natural number. (We can
call it 0 for ease of reference).
6
2.​ For every natural number n, S(n) is a natural number.
3.​ For all natural numbers m and n, if S(m) = S(n), then m = n.

The axioms so far describe the natural numbers, but don’t characterize them fully,
because it admits models of the following type.
Clearly, the intended progression forms a part of the model, but so do closed
disconnected loops of successions, and also disconnected infinite chains. However, it is
also clear that any model of these axioms will always include our desired progression,
so if we can add an axiom to filter out models with the extra bits, which is to say, to
require the smallest model to fulfil the requirements, we’ll have pinned down our
intended progression. This is where the axiom of induction comes in:
​
For all predicates P, if
1.​ P(0)
4
2.​ and for all numbers n, if P(n) then P(S(n)),
then for all numbers n, P(n).

This says that any predicate which applies to 0, and carries over successors, applies
over all the numbers. The quantification over predicates makes this a second-order
theory.

The case with second order theory is seemingly straightforward, because of all the
various predicates quantified over, there must be one which has all and only the desired
natural numbers falling under it, and as a consequence of the axiom must contain all the
numbers. So there cannot be any extra numbers floating in disconnected chains or
loops. And any predicate that contains only a subset of these numbers will fail one of
the original axioms, and hence fail the antecedent of our induction axiom.

So, it can be shown that this axiomatization captures our intended notion categorically,
which is to say, any two models which satisfy the axioms have an isomorphism which
respects the pattern of application of succession. The importance of this feature will
become clearer in the section on structuralism.

Peano Arithmetic
In modern usage, Peano arithmetic refers almost exclusively to the first-order
restatement of Dedekind-Peano arithmetic. The advantages of such an approach are
obvious, due to the desirable features of completeness and soundness for proof
7
systems in first order logic, which don’t extend to higher order logics.

In the first order axiomatization, the single axiom of induction is replaced with the
following infinite schema of axioms:

For all first-order formulas 𝜑(x), if

1.​ 𝜑(0)
10
2.​ And for all numbers n, if 𝜑(n) then 𝜑(S(n)),
​ Then for all numbers n, 𝜑(n)

The burden of infinitely many axioms is tempered by the existence of a simple decision
process to identify any instance of the axiom, and to enumerate all of them. The
advantageous features of first order logic extend to such systems.

However, these seemingly similar formulations produce very different results, and have
profoundly different philosophical and theoretical implications for all theories formulated
in either types of languages. For now, we need an explicit formulation of one or more
first order formulas which can eliminate all but the desired progression from our models.

To do so with the loops is relatively straightforward, for any closed loop has a finite size.
11
So, for example, to eliminate all 3-membered loops, we can use the following 𝜑(x):

𝜑(x): !(S(S(S(x)))) = x)

In the model given in the example above, we can see that all the numbers in both the
desired progression and the infinite disconnected chains satisfy this requirement, but
not the ones in the loop. Similarly, all n-membered loops can be eliminated.

This may indicate that if we also eliminate the extra chains, we’ll have our desired
model. But we do not have any such formula. Moreover, it can be shown that we cannot
13
have such a formula, because of a result in mathematical logic called the
Löwenheim-Skolem theorem that shows that any first-order theory with at least one
infinite model will have infinite models of all cardinalities. That means, besides the
model we’ve seen, there are many many more, some of them of vast sizes far beyond
the countable natural numbers. These are called the nonstandard models of arithmetic,
as opposed to our desired one which is called the standard model.

5
Difference between First and Second-order languages
5
The key difference lies in the fact that second order logic allows us to quantify over
arbitrary predicates/collection, regardless of whether there is a finite formula which
produces it, because at least theoretically there is a collection which consists only of the
numbers we desire to capture.

Another way to look at this is that second-order logic, by talking about arbitrary
predicates, allows us to define global constraints on our model, while first-order logic
only allows us to talk of local constraints. The feature we require in our model is
connectedness, and there does not seem to be a way to characterize this in a finite
first-order language.
However, there are features of the first-order theory that are highly desirable:
23
1)​ Gödel’s completeness theorem for first order logic ensures that anything which is
true in all models, including the standard one, can be proven in the first order
theory
2)​ First order systems are also sound, so there is no falsehood which can be
proved.

As the second order theory has fewer models than the first order one, we can inevitably
make stronger claims. But there are significant shortcomings in using it to characterize
our theory.
1)​ There is no complete second order proof system. There are proof systems which
are sound, so whatever is provable is true, but there are semantic consequences
that cannot be proved.
2)​ In terms of what can be proved syntactically, there is an assumption of a
25
set-theoretic background in the model theory(Quine famously quipped that
second order logic is “set theory in sheep’s clothing”). Given those assumptions,
15
second order logic turns out to be identical to a many-sorted first order logic.
While this is not a problem in establishing categoricity for arithmetic as long as
we have a robust formulation of set theory, we’ll see later that this becomes an
obstacle in achieving a similar result for set theory.

So there is a tradeoff between provability and expressiveness between first and second
1
order theories. Some of the acuteness of this will come into play when we look at the
axiomatization of set theory in the next section.
Section II: Set Theory and Structuralism
At the beginning of the previous section, I alluded to ontological analyses of numbers.
From Cantor onwards, mathematicians and philosophers looking to unify all of
mathematics on a single foundation provided a variety of ways in which numbers can be
analysed into more primitive notions of collections or sets, via equinumerosity, or some
other set-theoretic construction on which a succession relation can be defined.

4
The naïve conception of set theory, which was famously shown to be incoherent by the
8
discovery of Russell’s paradox, relied on a general comprehension principle, which says
that for any arbitrary property, one may form a set of objects with that property. Without
going into the details of the well-known story, we should note here that there exists a
different conception of sets which is equally intuitive, but uses a restrictive
comprehension principle to form sets in a hierarchy. This conception underlies the
axiomatization developed by Zermelo and expanded by Fraenkel. Zermelo himself
articulated this conception, but it found its most thorough explication in Boolos(1971).

18
The Iterative Conception of Sets
Briefly, the iterative conception goes like this: sets are “formed” in stages. Starting with
some ur-elements, the first stage consists only of sets formed from them. At any stage,
the sets which are formed can only contain as elements whatever has been formed at a
prior stage. The stages are well-ordered, i.e., an ordinal is associated with each stage,
called its “rank”. Sets appear in the hierarchy only after all their elements have
appeared. This arrests the development of any Russell-like paradoxes leveraging
self-containment, and more generally those based on any “closed loops” of
set-membership. The pure sets are those which have no ur-elements.

Hamkins(2020) notes that “it can be shown that every mathematical structure which
arises in set theory with urelements can be constructed isomorphically in the pure sets”.
8
So, for our consideration, we can just consider that part of the hierarchy as the
cumulative hierarchy of the set-theoretic universe.

While Godel’s incompleteness theorem renders any attempt to prove the internal
consistency of ZFC futile, it has nevertheless been widely adopted due to its strength,
and because no contradictions have emerged as yet.

Reduction of Arithmetic to Set Theory

One of the uses for ZFC was to reduce numbers as certain set constructs. Zermelo and
von Neumann independently provided formulations for ordinal numbers using pure
sets.These two formulations are not equivalent, but agree on all arithmetic properties
under suitable translations. The following table provides the recursive reinterpretation of
the arithmetic theory using set-theoretic terms:

Arithmetic Von Neumann Zermelo

0 ϕ ϕ
 11
n {0,1,.... n-1} {n-1}

The successor operation can be devised in terms of set membership, very plainly for
Zermelo ordinals, but also for Von Neumann ordinals. This reduction made the status of
numbers ontologically parasitic upon that of sets.

Benacerraf’s identification and access problems

In two highly influential papers, Benacerraf raises significant issues for broadly platonic
understandings of mathematical objects.

Benacerraf(1965) raises what is called the identification problem. He argues that while it
is true that the two formulations above agree on arithmetic assertions interpreted
suitably, that they differ on non-arithmetic statements means that they cannot both be
21
the correct reduction of what a number is. For instance, “0 ∈ 3” is true for von Neumann
ordinals, but false for Zermelo. So we have multiple reductions of numbers, and no
principled way of selecting one as superior to another.

The lesson that Benacerraf draws from this is that “‘Objects’ do not do the job of
numbers singly, the whole system performs the job or nothing does”. This paper marks
the beginning of significant interest in structuralism in philosophy of mathematics.
However, there are still object platonists, who consider natural numbers to be platonic
entities(sui generis, whose properties are all and only those as defined by Peano’s
axioms).

For platonists, Benacerraf(1973) raises another problem, which is called the access
problem. If numbers are platonic entities in a separate realm with no causal interaction
with our perceptual faculties, how do we know of them? This problem does not have a
neat solution as yet, and afflicts structuralists who deem mathematical theories to be
about abstract structures as well. Near the end, we’ll briefly look at Berry’s solution to
this problem.
This move towards structuralism also underscores the importance of the categoricity
7
result, for if we have a categorical axiomatization of a mathematical theory, it separates
the ontological analysis into its own issue, and allows for the employment of the theory
in mathematical practice freely.

7
Axiomatization of Set Theory
6
Similar to the case with arithmetic, there are both first order and second order
axiomatizations for ZFC. But, unlike the former, we do not have a categoricity proof
even using second order axioms. The best substitute is a quasi-categoricity theorem
proved by Zermelo himself, which shows that upto a certain height(corresponding a
largely inaccessible cardinal), the models are isomorphic to each other.

One might argue that we can require the intended model to be the shortest one which
can satisfy the second order ZFC axioms, similar to how it was done for arithmetic, but it
turns out that the shortest model is far smaller than what mathematicians are interested
in. Specifically, some of the results from the larger models(specifically the large cardinal
axioms) prove certain results about natural numbers which we cannot otherwise prove,
but can observe to hold true.

For example, consider the Goodstein theorem, which states that starting with any
number, if we write it in base 2 notation, subtract 1, and continue repeating the process,
increasing the base by 1 each round, we will ultimately reach 0. Performing this
operation initially leads to an explosive growth in the sequence, but in all the cases
we’ve computationally checked, the sequence comes back to rest at 0. This is a result
which cannot be proven with standard arithmetic, and requires large cardinal axioms
which are only available when arithmetic is reinterpreted within a suitably large set
model.

There is also the problem noted before, that we seem to assume a set-theoretic
background in our model theory to prove certain features of our proof systems. But
because we’re dealing with set theory itself, this seems viciously circular. The problem,
5
according to modal structuralists, lies in the assumption of a single hierarchy of sets.

Actualist Interpretation of the Set-Theoretic Hierarchy

The broadly accepted view till the mid-20th century was Platonism(or Actualism) about
the hierarchy of sets. This view treats the talk of “formed in layers” as a manner of
speaking, and considers the entire hierarchy to exist as an abstract mathematical
object. The ordering in the layers is variously interpreted as a logical or ontological
ordering. There are significant problems with such an approach.

First, the hierarchy itself is a mathematical structure, which must submit to a precise
articulation. Specifically, the height of the intended hierarchy, if it were seen as an
abstract object, would be fixed and correspond to an ordinal number. But the stages are
themselves rank-ordered, and the Burali-Forti paradox demonstrates that there is no
ordinal corresponding to the entire sequence of ordinal numbers.

24
Second, it seems that there are inherent limitations on how large of a structure can be
represented in set theory. Clearly, the collection of all sets is not representable as a set
itself, and in common parlance it is called a proper class. But that just moves the
problem upwards, because one can still not talk about the collection of all classes. Note
10
however that Boolos has provided an interpretation of second order quantification as
plural quantification, which he argues is not committed to existence of proper classes.

Third, to require the hierarchy to stop at an arbitrary point, besides being metaphysically
unparsimonious, substantially reduces the appeal of set theory as a foundation to study
arbitrary mathematical structures.

A Potentialist Reading of the Set-Hierarchy

Potentialism about set theory aims to dissolve rather than resolve the question about
the intended height of the hierarchy. It sees the hierarchy as inherently extendable, for
given any hierarchy, one could always add an extra layer of sets on top of it.

Among such readings, two prominent ones are the Parsonian reading, which interprets
set theoretic claims to be about what sets there could be, with the ‘could’ expressing an
interpretational possibility, whereas Putnamians treat set theoretic claims to be about
what hierarchy-like structures are possible, and how they can be extended.

We’ll look only at Putnamian formulations in this essay, but one thing to be noted is that
while both variants look at what sets and set-hierarchies are differently, they have a
fairly overlapping conception of how to understand other mathematical structures as
embedded within this construct.

Moreover, any mathematical structure, no matter how large, can be embedded inside a
larger structure, allowing for the generality requirement for set theory to be able to study
any possible mathematical structure. Categoricity of the embedded structures emerges
as a property of the structure itself, i.e., any sufficiently large structure will contain
elements which relate to each other in a way that carries over to extensions of them via
isomorphism.
Section III: Putnam and Hellman’s Modal
Structuralism

Putnam’s Pencil Points and Arrows

As is evident from the title of Putnam(1967), he does not argue for a foundationalist
programme, where mathematics is reduced to modal logic. He argues, rather, for an
“equivalent descriptions” paradigm, drawing analogy to two predominant theories(or
27
class of theories) in quantum mechanics which treat the world as a system of particles
and a system of waves respectively. To quote,

… they are not merely compatible but equivalent: the primitive terms of each
admit of definition by means of the primitive terms of the other theory, and then
each theory is a deductive consequence of the other. Moreover, there is no
particular advantage of taking one of the two theories as fundamental and
regarding the other one as derived.

The use of set theory to explicate notions of possibility and necessity is well established.
The reverse would be to interpret set-theoretic statements using modal logic, as talking
3
about what would necessarily be the case were there standard models for the set theory
in question. Putnam sketches only an outline of the reverse process, using modal logic
20
to explicate the notion of “all sets”. This was a problem both for naive set theory and for
the Actualist conception of the set hierarchy, and yet there are many legitimate
statements like “For all sets X and Y, there is a set which is their union”.

Concretizing a model of set theory with pencil points and arrows to stand for sets and
membership, he gives a definition of what it is for a model to be standard: there are no
infinitely descending arrow paths, and no more pencil points can be added without
increasing the “rank”. The notion of a standard model Putnam proposes mimics that of
natural numbers given using set-theoretic notions, in a move similar to how the truth
tables for basic logical operators are constructed using simple cases of logical
consequences, and then with that language, one can work in the opposite direction.

Bounded rank set theoretic statements(like “There is a unit set whose sole element is a
unit set”, a rank 2 statement) are now easily analysed as: if a standard concrete model
contains that rank, the statement holds in that model. Unbounded statement, like the
one above, can be translated by talking about extensions of such models: For any
standard concrete model containing X and Y, there is an extension of it such that it
contains the union of X and Y.

Putnam’s proposal is fairly preliminary, and he never expanded on the ideas in his later
work.

Criticism of Putnam
Krisel(1972) criticizes Putnam for taking the number-theoretic truths of any arithmetical
statement to be equivalent to the validity of a first order statement, something which we
can show does not extend to validity in second order axiomatizations. He sees
Putnam’s programme to be reiterating, in a new garb, the formalist programme of
Hilbert.

Putnam is also unclear about the notion of modality, which he takes to be variously an
intuitive logical or mathematical possibility, and physical possibility, but does not sketch
out the details of this notion. Specifically, he says that “the paper does not contain the
remotest hint how the modal concepts, of necessity and possibility in standard models
of set theory, could be used to find new axioms, or at least improve existing arguments
for axioms which are at all problematic”.

Moreover, “The one familiar equivalence between the Schrödinger’s and Heisenberg’s
picture does not extend from elementary quantum mechanics to quantum field theory”.
This belies the non-foundationalist approach, for a lack of apparent priority between
different formulations is not proof that they are in fact equivalent. The latter is a stronger
claim which must be asserted by providing a robust means to translate from one to the
other, which Putnam does not provide. Prima facie, equivocation between actualism and
potentialism seems to vacillate paradoxically between an definite hierarchy and arbitrary
extendibility.

Putnam also doesn’t provide any criteria for what it would take for some physical graph
(formed of pencil points and the like) to form “a standard model for Zermelo set theory”
3
but rather (as we saw above) asks the reader to “accept it on faith that the statement
that a certain graph G is a standard model for Zermelo set theory can be expressed
using no ‘non- nominalistic’ notions except the ‘◊’.

Some other notions that Putnam employs, like “physical space of arbitrarily high
cardinality”, to motivate the idea of a concretized standard model, are also seen to be
controversial, certainly not intuitively graspable.

Hellman’s Modal Structuralism

Hellman(1984) moves away from the notion of equivalent descriptions, and argues
solely for a Potentialist theory. He also fixes modal notion to a primitive logical
possibility, but only briefly describes it as being independent of physical constraints like
the size of the actual world, but beyond this only his application of logical possibility can
clarify the underlying notion. He relies on talk of possible worlds, similar to Berry later
on, but clarifies that it is only a metaphor, not to be taken as a semantic analysis of
logical possibility.

Hellman’s approach sees mathematical statements to have a hypothetical structure,

such that if a mathematical structure of a certain kind exists, then certain properties can
be asserted of it. But of course, this claim would be rendered true vacuously if no
concrete instance of such a structure exists. So a material conditional is not the proper
analysis for such statements. Hellman argues for stronger modal conditionals, which
assert that if there were such structures, certain properties would hold in them. This is
compounded with a categorical component asserting that such structures are possible.

So, for example, arithmetical statements would be a combination of

1)​ If there were ω-sequences, then P would hold in it

2)​ ω-sequences are possible
(An ω-sequence represents an ordering with the same order type as the natural
numbers. As noted before, structuralists view natural numbers as positions within the
abstract structure of an ω-sequence, rather than as individual, self-contained objects)

Without the categorical component, the original problems would recur. But the
commitment to the bare possibility of such structures is smaller than requiring that they
do in fact exist.

Hellman extends this to second order ZF axioms to give a categorical specification of

9
the intended models of set theory, with the view that the theory spells out what holds for
all structures of the relevant type. One significant difference between his analysis for
1
other mathematical theories and set theory is that he avoids quantification into the
scope of the modal operator for the former, due to controversy about how to interpret
such statements of “cross-world identities”. For the latter, he allows for quantification in
the case of set theory by relying on a principle of extendibility, which argues that models
of sets can always be interpreted to have possible extensions.

Criticisms of Hellman
Shapiro(1993) argues that the categoricity result that Hellman is able to prove is weaker
than what is required. To quote, “Hellman only shows that any two models of
second-order arithmetic that appear in the same world are isomorphic”, and not that
“any two models of second-order arithmetic are isomorphic.” This is because he does
not permit quantification of bound variables into the modal scope(in his modal
structuralist translations of arithmetical theories), which Shapiro argues could mean that
different “worlds” have a different range of quantification for such variables.

The key issue with Hellman is that he uses second order quantification, which may be
problematic for our enterprise as we noted earlier. His later work swaps out second
order quantification with plural quantification and laws of mereology(using a Lewisian
framework), but it’s not intuitively clear that the laws of mereology are logically and not
just metaphysically necessary. Berry also argues that there is a duplication of concepts
here, for there is a way that second order quantification can be read modally(over all
possible collections).

1
For his paraphrase of set theory, Hellman allows quantification into the scope of the
modal operator. There are Quinean worries about essentialism, but more importantly,
most sufficiently strong modal systems(like the ones Hellman uses) lead to seemingly
contradictory statements like “Everything necessarily exists”. And the ones that allow for
contingent objects have their own paradoxical results.

Berry’s approach is to sidestep the controversy, and attempt a foundation which doesn’t
require quantifying into the modal scope at all. This also has the benefit of a theory
which can be regimented into its first order portion and its modal portion. In this, she
follows closely with Putnam as opposed to Hellman, while rejecting Actualism.

Berry’s notion of Possibility

Like Hellman, Berry also requires the modal notion to be that of logical possibility.
Before examining this in the next section, I will just contrast this to other notions which
she considers and reject prima facie as too restrictive:

Logical possibility is a less demanding notion than metaphysical possibility(for example,

it is metaphysically impossible but logically possible for someone to be a man and a
bird), and also conceptual possibility, which requires analytic truths to be held fixed(for
example, it is conceptually impossible for someone to be a married bachelor, but
logically possible). She also rejects Tarskian interpretability, which will be discussed
below.

She also notes that what is actual should be possible, but in an actualist interpretation
of set theory, the real world contains all the sets, so its domain is strictly larger than that
of any set-theoretic model.. So it is plausible to think that not every logically possible
state of affairs has a set theoretic model.
Section IV: Primitive Logical Possibility
Both Putnam and Hellman argue for an intuitive notion of logical possibility which we
should include in our language as a primitive operator. Berry provides some ancillary
evidence for the plausibility of this notion, but only indicates that it is a domain of
evolving discourse in philosophy of logic.

Logical Possibility and Logical Validity

The key argument relies on the observation that logical possibility is interdefinable with
16
logical validity, as follows:

22
An argument is valid if and only if it is not logically possible for all its premises to
be true and its conclusion to be false.

This is similar to the notion of a semantical predicate that Quine(1953) argues to be the
only satisfactory notion of possibility or necessity. Although not all the discussion here
deals with possibility in a Quinean sense, we’ll note later how Berry’s use is
commensurable with this.

7 19
In model-theoretic analysis however, a sentence of first-order logic is called valid, if it is
true in all models, and the domains of such models are defined set-theoretically. So,
when we’re dealing with logical truths concerning set theoretic statements themselves,
17
we have prima facie reason to assume that we cannot capture the notion of all models
defined as sets, which would itself not be a set.

There are two ways out: we either rely on the set theoretic reflection principle, which
assures us that at least some model exists for every true statement. So, for a logically
true statement, its converse can be shown to not have a model. Or to put it another way,
4
we cannot have a valid yet false statement. The second is from the completeness
theorem, which assures us that any valid statement has a proof in the standard first
order axiomatic system.
But Boolos notes that “it is rather strange that appeal must apparently be made to one
or another non-trivial result in order to establish what ought to be obvious, viz., that a
sentence is true if it is valid”.

Boolos argues for a notion called supervalidity, according to which “a sentence of the
language of set theory is supervalid if it is true, no matter what sets its variables range
over(as long as there is at least one set over which they range) and no matter what
4
pairs of sets the relation of membership is taken to apply to.” Validity, supervalidity and
provability coincide for first order formulae, but such a claim cannot be made for second
and higher order ones.

Etchemendy’s critique of Tarskian semantics

2
One of the most widespread conceptions of logical truth and validity follow from Tarski’s
semantic analysis of logical truth and consequence, and the later model-theoretic
28
developments on the basis of Tarski’s analysis. Tarski notes that in lieu of a general
2
notion of truth, formal semantics tries to explain a relative notion of truth as: “x is true in
y”. The choice of y, between meaning or fact, leads to the two variants: interpretational
and representational semantics.

2
Representational semantics develops the notion of truth in a world(or state of affairs),
5
and uses that to provide truth-conditions for the terms of a language. A logical truth is
2
what’s “true in all genuinely possible configurations of the world”. But such a
configuration is conceived as possible only with respect to the laws of logical possibility,
precisely the thing that it is supposed to explicate.

Interpretational semantics is closer to Tarski’s own analysis. Tarski’s analysis sought to

2
capture two important features of logical consequence: valid arguments are
truth-preserving, and truth-preservation is independent of empirical aspects of the
2
objects concerned. So, a necessary condition is that validity of an argument be immune
 2
to arbitrary substitutions for the non-logical constants. However, this is not a sufficient
condition, as it may be satisfied due to a lack of expressiveness in the language, i.e.,
limits to the non-logical vocabulary in a language, leading to a lack of falsifying
substitutions. The remedy is the notion of satisfaction for sentential functions. This
2
theory interprets logical truth to derive from the non-logical expressions of a given
sentence

2
Logical truth of a sentence P is equated with the ordinary truth of a universal
generalisation of a sentential function created from P by replacing all non-logical
2
expressions with variables. So what is a logical truth depends only on one’s choice of
logical constants.

But Etchemendy notes that “the problem of the logical constants is a red herring”. That
2
is because there is an assumption that the truth of a universal generalisation warrants
the logical truth of all its instances, which is not acceptable. While Tarski recognizes the
modal feature in truth-preservation, none of the features of his analysis account for it.
This leads to cases of overgeneration and undergeneration as follows:

2
-​ Undergeneration: a logically valid statement may be declared invalid due to an
inappropriate choice of logical constants
-​ Overgeneration: ‘substantive’ extralogical influences can lead to seeing some
argument as valid, when it is infact not. (Chapter 7)

Etchemendy ultimately argues that “Tarski’s account of logical truth and consequence
does not capture, or even come close to capturing, the pretheoretic conception of the
logical properties.” But what explains its success then?

Field on set-theoretic surrogates for necessity

First, on model-theoretic notions, Field(2008) argues that an argument given by
Kreisel(1972) called the squeezing argument shows that as far as first order languages
are concerned, the extensional adequacy of ‘intuitive validity’ and “truth in all models”
follows from the soundness and completeness of first order logic. But again, this does
not extend to higher order logics. Since we lack a completeness proof for second and
14
higher order logics, the intuitive notion of logical truth differs from the model-theoretic
one.

Field notes that the adequacy is a result of working in restricted languages, where we
are working with only a part of the actual world in our quantifier domains. So the
set-theoretic surrogate is fine for such cases. But he notes that it seems we do in fact
employ an unrestricted langage, and have a notion of valid argument for it.

However, both Etchemendy and Field lacks a positive proposal for how to remedy this
situation. Tarski himself was aware of the intuitive notions of logical concepts, but was
skeptical due to their vagueness. Etchemendy suggests what would be required to
correct Tarski's account to avoid extensional errors: incorporating a guarantee that
assessments are not influenced by substantive, extralogical claims. However, achieving
this would essentially require replacing the problematic reduction principle(truth of
universal quantification implies logical truth of instances) with one that presupposes an
independent understanding of logical truth itself, thus making the correction project futile
in terms of providing a novel definition. Correcting the defect is "precisely equivalent to
solving the original problem de novo"

Analogy with Quantifiers

An analogy may be sought from the notion of quantifiers. Some philosophers have
sought to analyse the notion of a quantifier into chains of conjunction for the universal
quantifier and disjunction for the existential one. It is clear that such an account is
adequate as long as we’re dealing with a language such that, at least theoretically,
every object in the universe is distinctly nameable, so the variable of quantification may
be seen as a stand-in for the potentially infinitely many constants which name the
objects. But often we’re dealing with a universe with vastly larger number of objects(for
example, assuming space to be continuous, there are uncountably many spatial points),
and in such cases, what a universally quantified statement asserts is more than even an
infinite conjunction, which can at least capture a denumerable number of spatial points.
As an area of ongoing debates, there is no consensus on what the intuitive logical
concepts are. But assuming that more work will shed light on this, we’ll move ahead
with how Berry utilizes this concept.
Section V: Berry’s Potentialist Set Theory

Conditional Logical Possibility

As noted above, the structure is more relevant in mathematics than the particular
objects which instantiate it. Potentialist set theory can be formulated as talk of what
segment of a hierarchy of sets extends another, so comparing their structures. These
structures are patterns of how some relations apply. To say that a certain truth follows
from a structure implies that as long as we fix the pattern of application of the relations,
we can swap out some or all of the objects with others without affecting the truth value.

An illustrative example can be of the pigeonhole principle. In its casual formulation, it

says that if there are more pigeons than there are holes that they occupy, then at least
one hole is multiply occupied. The relevant relations here are: there are things of type
A(“pigeons”), some other things of type B(“holes”), and each of the things of type A
relates uniquely to a thing of type B(“occupies”). Even if one pigeon dies and is
immediately replaced by a hatchling, the truth of the overall statement holds, as the
pattern of application of A and B remain unchanged.

Berry suggests adding a generalized expansion of the logical possibility operator in our
language. There are three key distinctions of this conditional logical possibility:
1)​ In logical possibility simpliciter, the domain varies freely. In conditional logical
possibility, that portion of the domain which is relevant to the structural facts is
fixed, the rest may vary freely as before.
2)​ To keep the domain fixed is not to keep the extensions fixed, but rather the
pattern of application fixed. Formally, this is to say that the extensions of these
relations together are fixed up to isomorphism.
3)​ As the rest of the universe unaffected by these relations is left to vary freely, the
size of the universe does not have to be held fixed.

To preserve only the structural facts avoids any controversy about de re possibility.
Of the many advantages for this notion as listed by Berry, two seem the most appealing:
1)​ Cleaves reasoning about logical possibility in two parts, first order logic and
Modal Structural Principles, each of which can be independently analysed.
2)​ Eliminates the use of second order logic or plural quantification.

CLP version of Peano’s Induction Axiom

Berry replaces the prior first order infinite axiom schema with the following modal one:

Induct: □N;S If 0 is 𝜑 and the successor of every 𝜑 number is 𝜑 then every

number is 𝜑.

Berry argues that this schema has the same force as the second-order quantifier, which
can pick out a counterinductive collection of natural numbers only if it was logically
possible for 𝜑 to apply to exactly those numbers, in violation of the above axiom.

As we noted earlier, it is the possibility of a predicate that applies only to the standard
numbers which allows us to lay down a categorical characterization. But here, that work
has been offloaded to the modal operator, whose application is to be interpreted as
keeping the subscripted relations fixed upto an isomorphism, which eliminates the
nonstandard models due to lack of such isomorphisms across domains of different
cardinalities.

However, it is not certain that this can be considered a proof of categoricity yet, because
the crucial step is to ensure that an isomorphism can be created. Berry herself does not
provide the proof directly, but Scambler in his review(2023) of Berry’s work points out
what may be a weakness. What Berry needs is the following:

1
PA(N, S) ∧ PA(M, R) → ◇N,S,M,Riso(G, N, S, M, R)
Where PA stands for the first order peano axioms plus the modal induction axiom
1
schema, and iso is a formula that expresses that the predicate G constitutes an
1
isomorphism between N and M.​ ​ ​ ​ ​ ​ ​ ​ ​

1
Now the obvious route to such a proof is to show that for each n in N, it is possible
(given the structures of N, S, M, R) that there is a partial isomorphism between n and
some m in M, and then argue that these possible isomorphisms can be ‘pasted
together’ into a full isomorphism in some possible circumstances (using principles of
Berry’s logic). But the former statement clearly uses quantification into the scope of the
modal operator (‘for each n in N it is possible . . . n . . . ’), and consequently this cannot
even be stated in the obvious way in Berry’s language.

It may be the case that Berry’s axioms allow for a proof without quantification, but
having not provided it, we have to take her at her word that such a proof exists.

Truth-Conditions for CLP

8
Berry argues that we can use the familiar background of Actualist set theory to mimic
5 26
the intended truth conditions for statements in a language containing the logical
possibility operator ◊ alongside usual first-order logical vocabulary. But ultimately the
truth conditions and axioms are to be learned via immersion, which Berry takes to be a
form of abductive generalization from observations in various cases.

This also responds to the access problem as given by Benacerraf. Because the laws of
logic are subject-matter neutral, she argues that observations from accessible domains
like physical entities can be abstracted away and expanded. This is similar to a solution
presented by Maddy(quotation from Hamkins(2020)): “we can gain knowledge of
abstract objects through experience with concrete instantiations of them.”

Berry argues for something like the following:

 1)​ First, knowledge of non-mathematical objects, obtained through ordinary
perceptual and inferential methods, can give some initial data about what is
actual, is logically possible.
2)​ Abduction from regularities in what’s actual to what’s necessary. One example is
that inference to the best explanation for certain facts never happening is that it
would be logically impossible for them to happen. This provides a potential
source of knowledge of ¬♢φ facts.
3)​ One can always increase the accuracy of our methods by reflecting on whether
the laws we accept at any point lead to internal inconsistencies.

She does not argue that all of these are conscious processes. They may be realized at
an evolutionary level in our dispositions. Nor does this commit us to empiricism about
mathematics, for the methods of obtaining laws does not disqualify knowledge gained a
priori from using those laws.

I’ve only alluded to some of the strategies that Berry identifies to bolster her argument
towards a pretheoretic access to laws of logical possibility, but a significant engagement
with all of her strategies vis-a-vis their relative merits and demerits against other
strategies involves us in issues of comparing modal structuralism with non-modal
variants, which is beyond the remit of this essay(cf. Berry (2018b) for details).

Brief Remarks on other Features

So far, we’ve been concerned with the key philosophical issue that modal strcuturalism
hinges on, which is the notion of intuitive logical possibility. In this section, I will just
make brief summary remarks on some other features of Berry’s paraphrase of set
theory:

1)​ Berry has provided a translation schema for regular set theoretic statements into
their modal variants. While she shows that every theorem of set theory is a
theorem of her system, a consistency proof to ensure that no non-theorems get
translated is lacking.
2)​ While Hellman’s programme is purely nominalistic in its vein, Berry allows for
both a nominalist interpretation(using the same modal if-then strategy as
Hellman), as well as a neo-Carnapian realist interpretation, where a weak
1
quantifier variance thesis allows for a realist interpretation of quantificational
claims in restricted mathematical domains, as a way of soothing Quinean
indispensability concerns.
Section VI: Conclusion
This essay has only scratched the surface of the large amount of literature already
available on modal structuralism. It is clear that a significant treatment of the intuitive
notion of logical possibility independent of the set theoretic analysis needs to be
proffered. But it may perhaps be illuminating to continue the formal work here, in the
spirit of Putnam’s equivalent description paradigm, as a way to illuminate the way
forward.

One key observation is the modal element operative in providing categoricity results in
second order languages. Both Hellman and Berry’s solutions seem incomplete as of
now, but the identification of this crucial element is suggestive of some future directions
of work. Specifically, an attempt to provide it in a modal framework while bracketing
concerns about quantification might be illustrative, because none of the philosophical
issues related to modal quantification seem prima facie relevant to categoricity.

14
One of the charges against second order logic was that it settles the truth value of
certain things which are considered independent of ZFC, like the Continuum
Hypothesis. The same is true for Berry’s system. Berry’s answer to this is an appeal to
truth-values for these statements independent of our ability to formally prove it, but
set-theoretic pluralists like Hamkins might argue that the significant interest shown by
mathematicians with both types of systems, i.e., those where CH is assumed to be false
and where it is assumed true, and some of the paradoxical results that appear on either
side, seems to suggest that some of the axioms that Berry considers as a part of her
system might not be as general as we’d want them to be.
References

Benacerraf, P. (1965). What Numbers Could not Be. The Philosophical Review, 74(1), 47–73.
https://doi.org/10.2307/2183530

Benacerraf, P. (1973). Mathematical Truth. The Journal of Philosophy, 70(19), 661–679.

https://doi.org/10.2307/2025075

Berry, Sharon (2018). Modal Structuralism Simplified. Canadian Journal of Philosophy 48

(2):200-222.

Berry, S. (2022). A Logical Foundation for Potentialist Set Theory. Cambridge: Cambridge
University Press.

Boolos, G. (1971). The Iterative Conception of Set. The Journal of Philosophy, 68(8), 215–231.
https://doi.org/10.2307/2025204

Boolos, G. (1985). Nominalist Platonism. The Philosophical Review, 94(3), 327–344.

https://doi.org/10.2307/2185003

Etchemendy, John (1990). The concept of logical consequence. Cambridge: Harvard University
Press.

Field, Hartry (2008). Saving truth from paradox. New York: Oxford University Press.

Hamkins, Joel David (2020). Lectures on the philosophy of mathematics. Cambridge,

Massachusetts: The MIT Press.

Hellman, Geoffrey (1989). Mathematics Without Numbers: Towards a Modal-Structural

Interpretation. Oxford, England: Oxford University Press.

Hellman, G. (1990). Toward a Modal-Structural Interpretation of Set Theory. Synthese, 84(3),

409–443. http://www.jstor.org/stable/20116825
Kreisel, Georg (1972). Hilary Putnam. Mathematics without foundations. The journal of
philosophy, vol. 64, pp. 5–22. Journal of Symbolic Logic 37 (2):402-404.

Metschl, U. (1994). [Review of The Concept of Logical Consequence, by J. Etchemendy].

Erkenntnis (1975-), 41(2), 281–284. http://www.jstor.org/stable/20012584

Putnam, H. (1967). Mathematics without Foundations. The Journal of Philosophy, 64(1), 5–22.
https://doi.org/10.2307/2024603

Quine, W. V. (1953). Three Grades of Modal Involvement. Proceedings of the XIth International
Congress of Philosophy 14:65-81.

Scambler, Chris (2023). Sharon Berry.*A Logical Foundation for Potentialist Set Theory.
Philosophia Mathematica 31 (2):277-282.

Shapiro, S. (1993). [Review of Mathematics Without Numbers., by G. Hellman]. Noûs, 27(4),

522–525. https://doi.org/10.2307/2215793
 Similarity Report ID: oid:30225:461706291

9% Overall Similarity
Top sources found in the following databases:
5% Internet database 7% Publications database
Crossref database Crossref Posted Content database
2% Submitted Works database

TOP SOURCES
The sources with the highest number of matches within the submission. Overlapping sources will not be
displayed.

academic.oup.com
1 2%
Internet

Ulrich Metschl. "Book review", Erkenntnis, 1994

2 2%
Crossref

docplayer.net
3 <1%
Internet

ebin.pub
4 <1%
Internet

Quantifiers Quantifiers and Quantifiers Themes in Logic Metaphysics a...

5 <1%
Crossref

University of Birmingham on 2016-03-09

6 <1%
Submitted works

Language Truth and Logic in Mathematics, 1998.

7 <1%
Crossref

"The Palgrave Companion to the Philosophy of Set Theory", Springer S...

8 <1%
Crossref

Sources overview
 Similarity Report ID: oid:30225:461706291

"Reflections on the Foundations of Mathematics", Springer Science an...

9 <1%
Crossref

documents.mx
10 <1%
Internet

John Srdjan Petrovic. "Advanced Calculus - Theory and Practice", Routl...

11 <1%
Publication

Manchester Metropolitan University on 2024-05-17

12 <1%
Submitted works

University of Queensland on 2020-06-10

13 <1%
Submitted works

Guillermo E. Rosado Haddock. "Against the Current", Walter de Gruyter ...

14 <1%
Crossref

Jaakko Hintikka. "Truth, Negation and Other Basic Notions of Logic", T...
15 <1%
Crossref

Harrie de Swart. "Philosophical and Mathematical Logic", Springer Scie...

16 <1%
Crossref

Chris Tillman, Adam Russell Murray. "The Routledge Handbook of Prop...

17 <1%
Publication

STEWART SHAPIRO. "FREGE MEETS ZERMELO: A PERSPECTIVE ON I...

18 <1%
Crossref

Shashi Mohan Srivastava. "A Course on Mathematical Logic", Springer ...

19 <1%
Crossref

University of Edinburgh on 2021-04-11

20 <1%
Submitted works

Sources overview
 Similarity Report ID: oid:30225:461706291

University of Sydney on 2018-04-07

21 <1%
Submitted works

arxiv.org
22 <1%
Internet

plato.stanford.edu
23 <1%
Internet

services.phaidra.univie.ac.at
24 <1%
Internet

ciul.ul.pt
25 <1%
Internet

Sharon Berry. "Modal structuralism simplified", Canadian Journal of Ph...

26 <1%
Crossref

hdl.handle.net
27 <1%
Internet

Alexius Meinong The Shepherd of Non-Being, 2015.

28 <1%
Crossref

Sources overview