Metaphysical necessity, numerical identity, and the a priori

PHIL UA-78 Metaphysics

Prof. Cian Dorr
10th September 2024

1
Metaphysical necessity and
possibility
Different interpretations of ‘could have’ in English

‘Could have’ is flexible: on different occasions, we can use it to mean different things.

▶ The white ball couldn’t have gone into the pocket without hitting the red ball
(since the red ball was right in front of the pocket)
▶ Someone could have picked up the red ball and made the white ball go straight
into the pocket without hitting any other balls.

2
Different interpretations of ‘could have’ in English

‘Could have’ is flexible: on different occasions, we can use it to mean different things.

▶ The white ball couldn’t have gone into the pocket without hitting the red ball
(since the red ball was right in front of the pocket)
▶ Someone could have picked up the red ball and made the white ball go straight
into the pocket without hitting any other balls.

Some of these meanings are broader than others.

▶ The claim that interpretation 1 of ‘could have’ is broader than interpretation 2

means, roughly, that when a sentence ‘It could have been that …’ is true on
interpretation 2, it is also true on interpretation 1 (holding fixed the interpretation
of whatever goes in the dots).

2
‘Could have’ and ‘had to’

‘Could have’ goes hand in hand with ‘had to’.

▶ ‘It could have not been that P’ ≈ ‘It didn’t have to be that P’
▶ ‘It had to not be that P’ ≈ ‘It couldn’t have been that P’

3
‘Could have’ and ‘had to’

‘Could have’ goes hand in hand with ‘had to’.

▶ ‘It could have not been that P’ ≈ ‘It didn’t have to be that P’
▶ ‘It had to not be that P’ ≈ ‘It couldn’t have been that P’

And hence…

▶ ‘It could have been that P’ ≈ ‘It didn’t have to not be that P’
▶ ‘It had to be that P’ ≈ ‘It couldn’t have not been that P’

3
We say that one interpretation of ‘had to’ is broader than another when the
interpretation of ‘could have’ corresponding to the former is broader than the
interpretation corresponding to the latter.

▶ The claim that interpretation 1 of ‘had to’ is broader than interpretation 2 means,
roughly, that when a sentence ‘It had to be that …’ is true on interpretation 1, it is
also true on interpretation 2 (holding fixed the interpretation of whatever goes in
the dots).

4
Necessity and possibility

‘Could have’ is one of many ways we have in English for talking about (a wide range of
notions of) possibility, and ‘had to’ is one of many ways of talking about (the
corresponding range of notions of) necessity.
▶ ‘It could have been that P’ ≈ ‘It could be that P’ ≈ ‘It is/was possible for it to be
that P’ ≈ ‘Possibly, P ’…
▶ ‘It had to be that P’ ≈ ‘It must be that P’ ≈ ‘It is/was necessary that P’ ≈
‘Necessarily, P’…

5
Necessity and possibility

‘Could have’ is one of many ways we have in English for talking about (a wide range of
notions of) possibility, and ‘had to’ is one of many ways of talking about (the
corresponding range of notions of) necessity.
▶ ‘It could have been that P’ ≈ ‘It could be that P’ ≈ ‘It is/was possible for it to be
that P’ ≈ ‘Possibly, P ’…
▶ ‘It had to be that P’ ≈ ‘It must be that P’ ≈ ‘It is/was necessary that P’ ≈
‘Necessarily, P’…
The general pattern:
▶ Possibly ≈ not necessarily not
▶ Necessarily ≈ not possibly not

This is called the logical duality of ‘possibly’ and ‘necessarily’.

5
Nomic and metaphysical modality

In this course we will be especially interested in two of the many available

interpretations of modal words like ‘possible’ and ‘necessary’.

1 Nomic (aka physical) necessity and possibility. Here, ‘It could have been that P’
can be restated less ambiguously as ‘The laws of nature are compatible with its
being that P’, and ‘It had to be that P’ can be restated as ‘The laws of nature
guarantee that P’.

6
Nomic and metaphysical modality

2 Metaphysical (aka absolute) ne-

cessity and possibility. This is what Kripke calls ‘necessity and possibility tout court’.
[I will argue] that characteristic theoretical identifications like ’Heat is the
motion of molecules’, are not contingent truths but necessary truths, and
here of course I don’t mean just physically necessary, but necessary in the
highest degree, whatever that means. (Physical necessity, might turn out to
be necessity in the highest degree. But that’s a question which I don’t wish to
prejudge. At least for this sort of example, it might be that when something’s
physically necessary, it always is necessary tout court.) (N&N p. 99).

7
Nomic and metaphysical modality

Many philosophers–including me–are happy to characterize metaphysical modality as

the broadest modality. That is:

▶ If there is any sense in which things ‘could have been’ a certain way, then it’s
metaphysically possible for them to be that way.
▶ If it’s metaphysically necessary that things are a certain way, then it’s true in
every sense that they ‘had to be’ that way.

8
Nomic and metaphysical modality

You may be more familiar with the expressions ‘logically possible’ and ‘logically
necessary’, which some philosophers used to use for what’s now called ‘metaphysical’
possibility and necessity. But this can be misleading.
▶ It is (plausibly) metaphysically necessary that no bachelor is married. But the
sentence ‘No bachelor is married’ is not provable in any known system of
logic—in logic, any two nouns or adjectives are interchangeable.
One would standardly say that ‘No bachelor is married’ isn’t a logical truth. But this
remark just concerns sentence, not about what it says about the world.

9
Nomic and metaphysical modality

You may be more familiar with the expressions ‘logically possible’ and ‘logically
necessary’, which some philosophers used to use for what’s now called ‘metaphysical’
possibility and necessity. But this can be misleading.
▶ It is (plausibly) metaphysically necessary that no bachelor is married. But the
sentence ‘No bachelor is married’ is not provable in any known system of
logic—in logic, any two nouns or adjectives are interchangeable.
One would standardly say that ‘No bachelor is married’ isn’t a logical truth. But this
remark just concerns sentence, not about what it says about the world.
Philosophers sometimes confuse things by saying ‘It is not logically necessary that no
bachelor is married’ when they mean “‘No bachelor is married” is not a logical truth’.
But this is bad practice—in philosophical writing, one needs to be a stickler about
using quotation marks.
9
One reason to care about this topic

Recall that last week we were talking about “comprehensive world-pictures” like
Democrateanism, which include a hard-to-make-sense-of claim ‘completeness’ (which I
wrote as ‘that’s all’).
The Democratean’s idea is that if you say the truth about what atoms there are, what
shapes they are, and how far apart they are (at every time), you have in some sense
given a ‘complete’ description of the world. But what could this mean (if it’s not
obviously false)?

10
One reason to care about this topic

Recall that last week we were talking about “comprehensive world-pictures” like
Democrateanism, which include a hard-to-make-sense-of claim ‘completeness’ (which I
wrote as ‘that’s all’).
The Democratean’s idea is that if you say the truth about what atoms there are, what
shapes they are, and how far apart they are (at every time), you have in some sense
given a ‘complete’ description of the world. But what could this mean (if it’s not
obviously false)?
A plausible idea is that at least part of what it means is a supervenience claim. Every
fact is metaphysically necessitated by the facts about atom-arrangement: that is, every
fact is such that it’s metaphysically necessary that if the atom-arrangement facts are
as they are, that fact obtains.

10
Numerical identity and necessity
The concept of numerical identity

‘This is numerically identical to that’ ≈ ‘This and that are one and the same’ ≈ ‘This
is the very same thing as that’ ≈ ‘This is that’.
There’s another use of ‘identical’ in play when we talk about two twins, or two peas in
a pod, or two copies of a newspaper, or two electrons being ‘identical’. For two things
to be ‘identical’ in this other sense—‘qualitatively identical’—is for them to be exactly
similar, at least in some relevant range of respects.
By contrast, it simply can’t happen that two things are numerically identical! If x is
numerically identical to y, then x and y are not two things, but one thing.

11
Numerical identity and numerals

Sometimes, a claim about numerical identity can be equivalently expressed using

numerals:

▶ ‘I own a cat x and a cat y such that x is not numerically identical to y’ ≈ ‘I own
at least two cats’
▶ ‘For any cats x and y such that if x is owned by me and y is owned by me, then x
is numerically identical to y’ ≈ ‘I own at most one cat’.

12
Some terminological notes

1: In this course, when we just say ‘identity’/‘identical’ without modification, we

always will mean numerical, never qualitative.
2: ‘Distinct’ means ‘not (numerically) identical’
3. At least in philosophy (and arguably also in mathematics), the ‘equals’ sign ‘=’ is
almost always used to mean ‘is numerically identical to’. ‘�’ means ‘is not numerically
identical to’.

13
Some terminological notes

4. In this course we will always be using the English word ‘identity’ interchangeably
with ‘being identical’—compare ‘adjacency’/‘being adjacent’. ‘Identity’ also has a
different use as a count noun, as in ‘Eating eight apples a day is an important part of
Mike’s identity’. This use will play no role in the course. If you want to use the word
‘identity’ in your writing in a way that couldn’t be replaced with ‘being identical’,
probably something has gone wrong.

14
The logic of numerical identity

There are two basic logical principles about numerical identity.

Reflexivity Everything is identical to itself.

Leibniz’s Law If x is identical to y, y has every property that x has.

15
The logic of numerical identity

There are two basic logical principles about numerical identity.

Reflexivity Everything is identical to itself.

Leibniz’s Law If x is identical to y, y has every property that x has.

We can deduce two other useful principles from these

Symmetry If x is identical to y, y is identical to x.

Transitivity If y is identical to x and z is identical to y, z is identical to x.

15
The logic of numerical identity

Proof of Symmetry: Suppose x is identical to y. By Reflexivity, x is identical to x; or in

other words, x has the property of being identical to x. But by Leibniz’s Law, y has
every property that x has. So, y also has the property of being identical to x, or in
other words, y is identical to x.

16
The logic of numerical identity

Proof of Symmetry: Suppose x is identical to y. By Reflexivity, x is identical to x; or in

other words, x has the property of being identical to x. But by Leibniz’s Law, y has
every property that x has. So, y also has the property of being identical to x, or in
other words, y is identical to x. Proof of Transitivity: Suppose y is identical to x and z
is identical to y. By Reflexivity, x is identical to x; or in other words, x has the property
of being identical to x. By Leibniz’s Law, y has every property that x has and z has
every property that y has, which implies that z has every property that x has. So, z has
the property of being identical to x; or in other words, z is identical to x.

16
An obviously atrocious argument against Transitivity

(a) and (b) are both true, but (c) is false:

(a) Kamala Harris is numerically identical to some human being

(b) Some human being is numerically identical to Donald Trump
(c) Kamala Harris is numerically identical to Donald Trump

17
An obviously atrocious argument against Transitivity

(a) and (b) are both true, but (c) is false:

(a) Kamala Harris is numerically identical to some human being

(b) Some human being is numerically identical to Donald Trump
(c) Kamala Harris is numerically identical to Donald Trump

Obviously this not a good objection to Transitivity. We can’t infer from (a) and (b)
that there is something such that Kamala Harris is identical to it and it is identical to
Donald Trump.

17
An obviously atrocious argument against Transitivity

There’s a general moral here. In English, determiner phrases like ‘some human being’
and ‘a human being’ look syntactically like proper names (‘Kamala Harris’) and
pronouns (‘she’, ‘it’). But logically, they behave in some rather special ways that mean
you can’t take for granted that ‘For any object x, x does such and such’ rules out ‘[DP]
does not do such-and-such’.

18
An obviously atrocious argument against Transitivity

There’s a general moral here. In English, determiner phrases like ‘some human being’
and ‘a human being’ look syntactically like proper names (‘Kamala Harris’) and
pronouns (‘she’, ‘it’). But logically, they behave in some rather special ways that mean
you can’t take for granted that ‘For any object x, x does such and such’ rules out ‘[DP]
does not do such-and-such’.
This also applies (although less obviously) to determiner phrases using ‘the’, of the
form ‘the F’ (e.g. ‘the president of the US’).

18
Bogus counterexamples to Leibniz’s Law

(Pointing to a photo showing a new car on the dealer lot.) This car is shiny. (Pointing
to a photo showing a beat up old car.) That car is not shiny. So, this car is not
numerically identical to that car.

19
Bogus counterexamples to Leibniz’s Law

(Pointing to a photo showing a new car on the dealer lot.) This car is shiny. (Pointing
to a photo showing a beat up old car.) That car is not shiny. So, this car is not
numerically identical to that car. Obviously something has gone wrong here! I could
just as well have said ‘This car is shiny and was once owned by me’ and ‘That car is
not shiny and was once owned by me’. Given ‘This car is not numerically identical to
that car’, I could infer ‘There are x and y such that x is not numerically identical to
that car and x was once owned by me and y was once owned by me’, and hence ‘there
are at least two cars that were once owned by me’.

19
Bogus counterexamples to Leibniz’s Law

Diagnosis: when I say ‘This car is shiny’ and ‘that car is not shiny’ I mean something
like ‘This car was shiny at the time this photo was taken’ and ‘That car was not shiny
at the time that photo was taken’. Or maybe I mean ‘This car is depicted as shiny in
this picture’ and ‘That car is depicted as not shiny in that picture’. Anyhow, my two
uses of ‘is shiny’ express two different properties.
So, I cannot infer from ‘This is shiny and that is not shiny’ to ‘There is a property that
this has and that does not have’, for exactly the same reason that I can’t infer from
‘That animal (pointing at one animal) is a cat and that animal (pointing at a different
animal) is a dog’ to ‘There is an animal that is a cat and is also a dog’.

20
The necessity of identity

Two important principles:

Necessity of Self-identity (NSI) For any object x, it is necessary that x is identical
to x.
Necessity of Identity (NI) For any objects x and y, if x is identical to y, it is
necessary that x is identical to y.

21
The necessity of identity

Two important principles:

Necessity of Self-identity (NSI) For any object x, it is necessary that x is identical
to x.
Necessity of Identity (NI) For any objects x and y, if x is identical to y, it is
necessary that x is identical to y.
There is a really important proof of NI, assuming NSI.
Suppose x is identical to y. By NSI, it is necessary that x is identical to x; or in
other words, x has the property of being something such that it is necessary that x
is identical to it. But by Leibniz’s Law, y has every property that x has; so y also
has the property of being something such that it it is necessary that x is identical
to it. In other words, it is necessary that x is identical to y.
This applies to any interpretation of ‘necessary’, including metaphysical necessity. 21
The necessity of identity

Although NSI is widely accepted, there are some philosophers who think it’s false on
the grounds that (e.g.) if your parents had never met, you would not even have been
self-identical. But there is a weaker claim whose proof does not rely on NSI:

Conditional Necessity of Identity (CNI) For any objects x and y, if x is identical to

y, it is necessary that if x is identical to x, then x is identical to y.

We can prove this in exactly the same way as NI, except that instead of relying on NSI,
we now merely rely on the (entirely uncontroversial) claim that for any object x, it is
necessary that if x is identical to x, then x is identical to x.

22
A bad objection to the Necessity of Identity

The inventor of bifocals is identical to the first Postmaster-General. But it is not

necessary that the inventor of bifocals is identical to the first Postmaster-General.
So there exist objects x and y such that x is identical to y but it is not necessary
that x is identical to y.

23
A bad objection to the Necessity of Identity

The inventor of bifocals is identical to the first Postmaster-General. But it is not

necessary that the inventor of bifocals is identical to the first Postmaster-General.
So there exist objects x and y such that x is identical to y but it is not necessary
that x is identical to y.

Response: Remember the earlier warning about determiner phrases!

23
A bad objection to the Necessity of Identity

Often, when determiner phrases occur in the same sentence as modal words (like
‘necessary’ and ‘possible’), we get a scope ambiguity.
▶ ‘Some red ball could have been selected’ can mean either ‘For some red ball x, x
could have been selected’ [wide scope] or ‘It could have been that there was a ball
that was both red and selected’ [narrow scope]
▶ ‘The inventor of bifocals could have been German’ can mean either ‘The inventor
of bifocals is an x such that x could have been German’ [wide scope] or ‘It could
have been that a German invented bifocals before anyone else’ [narrow scope].
▶ If we give both descriptions wide scope ‘It is necessary that the inventor of
bifocals is identical to the first Postmaster-General’ is equivalent to ‘The inventor
of bifocals and the first Postmaster-General are an x and y such that it is
necessary that x is identical to y’. On this reading it is plausibly true! And this is
the only reading on which it follows from NI. 24
Proper names

According to an old theory that Kripke very influentially argued against, proper names
like ‘Kamala Harris’ are short for definite descriptions—maybe ‘the Democratic
candidate for president’ or ‘the famous person called “Kamala Harris”’.
But proper names do not seem to give rise to scope ambiguities when they combine
with modals. ‘Kamala Harris could have not been Kamala Harris’ seems to be
unambiguously false, whereas ‘the Democratic candidate could have not been the
Democratic candidate’ has a true (narrow scope) reading.
Kripke puts this by saying that proper names are rigid designators.

25
Proper names

According to an old theory that Kripke very influentially argued against, proper names
like ‘Kamala Harris’ are short for definite descriptions—maybe ‘the Democratic
candidate for president’ or ‘the famous person called “Kamala Harris”’.
But proper names do not seem to give rise to scope ambiguities when they combine
with modals. ‘Kamala Harris could have not been Kamala Harris’ seems to be
unambiguously false, whereas ‘the Democratic candidate could have not been the
Democratic candidate’ has a true (narrow scope) reading.
Kripke puts this by saying that proper names are rigid designators. This means that we
don’t have to worry about the inference from the Necessity of Identity to, e.g., ‘If
Superman is Clark Kent, then it is necessary that Superman is Clark Kent’.

25
Why the necessity of identity matters

We can use the necessity of identity to argue from identity claims to claims of
metaphysical necessity.
For example: we can argue from the premise that Superman is Clark Kent to the
conclusion that it is metaphysically necessary that everyone who has met Superman
has met Clark Kent.
(The argument relies also on the uncontroversial premise that ‘It is metaphysically
necessary that everyone who has met Superman has met Superman’.)

26
Why the necessity of identity matters

This form of argument becomes much more powerful and interesting when we turn
from identity of objects to identity of properties.
For example: we can argue from the premise that the property of being golden is the
property of consisting of atoms with 79 protons to the conclusion that it’s
metaphysically necessary that every gold thing consists of atoms with 79 protons.

27
Why the necessity of identity matters

This form of argument becomes much more powerful and interesting when we turn
from identity of objects to identity of properties.
For example: we can argue from the premise that the property of being golden is the
property of consisting of atoms with 79 protons to the conclusion that it’s
metaphysically necessary that every gold thing consists of atoms with 79 protons. And
the premise seems plausible! When chemists analysed a bunch of golden things and
found that they all consisted of atoms with 79 protons, they got strong evidence for a
claim about what it is to be golden—namely, that to be golden is to consist of atoms
with 79 protons. (In my view, Kripke’s most important achievement is to make the
world safe for this sort of argument by clearing away various bad objections to it.)

27
Necessity and a priority
The concept of a priori knowability

Some philosophers writing in the earlier part of the twentieth century seem to equate
claims of metaphysical necessity with claims about a priori knowability—things that
can be known “without relying on sense-experience”.

▶ The question what it even means to know something ‘without relying on

sense-experience’, and whether it’s even possible to know anything in this way, is
an interesting one, but one for epistemology rather than metaphysics—we are only
mentioning the idea because we want to set it aside.

A basic point from Lecture 1 of N&N is that these are different notions: the claim that
something is metaphysically necessary has nothing to do with knowledge or people.

28
The concept of a priori knowability

Some philosophers writing in the earlier part of the twentieth century seem to equate
claims of metaphysical necessity with claims about a priori knowability—things that
can be known “without relying on sense-experience”.

▶ The question what it even means to know something ‘without relying on

sense-experience’, and whether it’s even possible to know anything in this way, is
an interesting one, but one for epistemology rather than metaphysics—we are only
mentioning the idea because we want to set it aside.

A basic point from Lecture 1 of N&N is that these are different notions: the claim that
something is metaphysically necessary has nothing to do with knowledge or people. A
less basic point is that that either of ‘It is metaphysically necessary that P’ and ‘It is
knowable a priori that P’ can be true without the other being true.

28
The necessary a posteriori and contingent a priori

▶ It is metaphysically necessary that Hesperus is Phosphorus TRUE BY NI

▶ It is knowable a priori that Hesperus is Phosphorus SEEMS FALSE
▶ It is knowable a priori that if anything is one metre long, the standard metre is
one metre long SEEMS TRUE
▶ It is metaphysically necessary that if anything is one metre long, the standard
metre is one metre long SEEMS FALSE

29