Studies in History and Philosophy of Modern Physics 50 (2015) 43–53

Contents lists available at ScienceDirect

Studies in History and Philosophy

of Modern Physics
journal homepage: www.elsevier.com/locate/shpsb

Dissecting weak discernibility of quanta

Tomasz Bigaj a,b
a
University of Warsaw, Institute of Philosophy, Warsaw, Poland
b
Department of Philosophy, University of California, San Diego, La Jolla, CA, USA

art ic l e i nf o a b s t r a c t

Article history: In this paper I critically examine latest attempts to formalize quantum-mechanical relations that are
Received 8 October 2014 supposed to weakly discern elementary particles. I argue that all of them make illegitimate and
Received in revised form unavoidable reference to numerical identity, and therefore cannot be used as a means to ground (or
6 February 2015
derive) quantitative facts of identity/distinctness in the qualitative characteristics of quantum systems.
Accepted 18 February 2015
I compare my criticism of weak discernibility with the general circularity objection known from the
Available online 17 March 2015
literature, and I show that my argument is more specific, as it is based on a particular criterion which
Keywords: differentiates between legitimate and illegitimate uses of identity. In the end I suggest that we should
Weak discernibility reevaluate the role of permutation invariance in expressing the facts of qualitative differences between
Identity
particles. Taking into account the inevitable symmetrization requirement applied to operators in tensor
Individuality
product spaces, it may be claimed that particles of the same type can be absolutely discerned in some
Elementary particles
accessible states.
& 2015 Elsevier Ltd. All rights reserved.

When citing this paper, please use the full journal title Studies in History and Philosophy of Modern Physics

1. The weak discernibility program: an overview
It has been more than 10 years since the emergence of the idea
to exploit the logical concept of weak discernibility in discussions
on the ontological status of quantum particles. During these years
the original suggestion has grown into a full-blown metaphysical
program aiming at deriving useful lessons regarding the identity
and individuality of fundamental entities from our best physical
theories, and making connections with other far-reaching meta-
physical programs, such as ontic structuralism. In this critical
survey I would like to focus on some technical aspects of weak
discernibility in the context of non-relativistic quantum theory.
One of the main conclusions of my analysis will be that the weak
discernibility (WD) program applied to quanta is actually incap-
able of reaching all of its ambitious goals. The weak discernibility
claims advanced with respect to quantum particles of the same
type (bosons and fermions) are toothless as a tool for establishing
interesting metaphysical conclusions regarding their status
as full-fledged objects, and regarding the validity of non-trivial
metaphysical principles, such as the Principle of the Identity of
Indiscernibles.
E-mail addresses: t.f.bigaj@uw.edu.pl, tfbigaj@gmail.com
Before we plunge into a detailed exegesis of the technical
aspects of the WD program, it may be appropriate to remind the
reader of the historical development of this approach. For the
purpose of bookkeeping I suggest distinguishing four stages of
development of the WD program. The first stage, which may be
called Promising beginnings, is associated with the name of Simon
Saunders. At the time when the dominating view among physicists
and philosophers had been that the PII is violated in quantum
mechanics, and that quanta most probably lack any individuality
due to their indiscernibility, Saunders had the audacity to go
against the flow and argue that there is a largely forgotten sense of
discernibility which may be applicable even to elementary parti-
cles (Saunders, 2003, 2006). Saunders rediscovered the Hilbert–
Bernays method of defining identity using qualitative predicates
only, and argued that Quine’s weak discriminability on which this
method is implicitly based is attainable for fermions of the same
type. Fermions in antisymmetric states can be discerned by the
relation of having opposite spins, and therefore are legitimate
objects. On the other hand, bosons in symmetric product states
remain utterly indiscernible, and as such can be interpreted as
aggregates only.
The second stage of the program, which I refer to as Sharpening
the arguments is marked by the arrival of Fred A. Muller on the
scene. Many critics of Saunders’ early attempts to weakly discern

http://dx.doi.org/10.1016/j.shpsb.2015.02.003
1355-2198/& 2015 Elsevier Ltd. All rights reserved.
44 T. Bigaj / Studies in History and Philosophy of Modern Physics 50 (2015) 43–53
fermions pointed out that it is strictly speaking incorrect to claim
that the relation of having opposite spins can connect two
fermions in the singlet state, since individual particles do not
possess any precise values of spin. In order to repel similar
objections, Muller and Saunders (2008) joined forces and made
an effort to develop a technically rigorous and conceptually sound
method of weakly discerning fermions in all admissible states
using categorical (i.e. non-probabilistic) properties only. In this
approach bosons remained at a slight disadvantage, since their
discernibility was achieved only with the help of probabilistic
relations. However, in a follow-up paper Muller and Seevinck
(2009) strengthened even this claim, showing how to discern
fermions, bosons and any other particles in terms of eigenvalues of
particular quantities, and hence categorically.
After this major victory, the program took a sharp Metaphysical
turn, and therefore entered its third stage, punctuated by two
papers (Muller, 2011, 2015). In them, the technical results achieved
in the previous phase are elevated to the status of sweeping
metaphysical claims about the ultimate nature of physical reality
and its governing principles. Muller abandons the traditional
dichotomy between individuals and non-individuals, introducing
a third category of entities: relationals. Fundamental objects in
physical theories are determined by the relations they participate
in, and this gives support to the broad structuralist stance. On the
other hand, far from being committed to the trash heap of history
by the development of quantum mechanics, PII actually receives a
boost and comes out victorious as one of the most scientifically
justified metaphysical principles.
I suggest that the fourth stage in the development of the WD
program be identified with Friendly criticism done by authors who
do not question the main goal and general results of the program
but notice some gaps in the arguments by Muller, Saunders and
Seevinck that need to be filled out.1 Two recent papers by Adam
Caulton (2013) and Nick Huggett and Josh Norton (2014) are
representative of this stage.2 Both papers aim at improving rather
than rejecting the WD program by selecting new discerning
relations that are better suited for the task. Later in the text we
will discuss the details of these critiques, and we will see that
some of the objections raised in them reveal much more serious
problems with the WD program which cannot be easily remedied
by simply choosing a new discerning quantity.
The ostensible goal of the WD program is to rehabilitate the
Principle of the Identity of the Indiscernibles, which has fallen on
hard times in quantum mechanics.3 This in itself is a worthy
undertaking. However, one can ask why we should be so keen to
save the venerable Leibnizian principle, if not out of the respect for
one of the greatest minds in philosophy and science. The PII tells
us, in rough outline, that all numerically distinct objects have to
differ qualitatively from one another (this is what we usually mean
by “discernibility”). Muller (2015) formulates three reasons why
we should care about such a claim. The first reason given by
Muller is not so much a motivation why we should want PII to be
true, as an explanation of why we should not ignore it if it is
indeed true (“because it teaches us an ontological lesson” is the
submitted answer). The second reason is associated with Muller’s
1
In my current exposition of the WD program I largely ignore the not-so-
friendly criticism of weak discernibility advanced e.g. by Hawley (2006, 2009),
French & Krause (2006), van Fraassen & Peschard (2008), Dieks & Versteegh (2008),
and Ladyman & Bigaj (2010). However, some echoes of these critiques will
reverberate later in the text, mostly in Section 5.
2 7
It should be added here that Caulton no longer supports the WD program
(private communication). See footnote 19 for reference to his latest unpublished
work on the problem of discernibility of quantum particles.
3
Some of the key authors responsible for exposing the apparent plight of PII in
QM are Margenau (1944), French & Redhead (1988), Giuntini & Mittelstaedt (1989),
Redhead & Teller (1992), and Butterfield (1993).
preferred metaphysics of relationals, and as such does not extend
to all forms of PII, only the ones based on relational discernibility.
The third adduced reason is that PII supplies us with clear identity
criteria. This observation is echoed by Saunders in (2003, pp. 289–
291), where he stresses that the truth of the PII enables us to
reduce numerical identity to qualitative facts, and therefore
eliminates the need for identity as a primitive predicate. Saunders
insists that while in general there is nothing wrong with identity
taken as primitive, in the context of physics it is better to see it as
derivative. He points out that physical objects are known to us
through descriptions as objects of predications, and not as some
formal constructs. Thus descriptions in terms of pure physical
properties should enable us to recognize objects of physical
theories as entities numerically distinct from one another.
I hope it would not be a gross distortion of the WD program to
sum up the above ideas as follows. We need the validity of (some
form of) PII, because we want to be able to ground facts regarding
numerical distinctness (numerical identity) of objects in some
qualitative facts pertaining to their arrangements.4 And the pos-
sibility of achieving such grounding (or reduction, to use another
fashionable term) in a principled manner can ensure some
important metaphysical conclusions regarding the status of the
entities in question (that they are objects, or individuals, or at least
relationals). Moreover, specific methods of achieving the required
grounding can give support to independent metaphysical claims,
such as the claim of the ontological primacy of relations (struc-
tures) over objects, commonly known as ontic structuralism.5
The notion of discernibility figuring in the formulation of PII
admits various interpretations.6 The standard version of PII
employs discernibility by properties, typically referred to as
absolute. The two other grades of discernibility are: relative and
weak discernibility. Two objects are relatively discernible if there
is a relation that connects them in one direction but not the other.7
Weakly discerning relations, on the other hand, are such that they
hold between two distinct objects, but do not hold between an
object and itself. Note that from this characterization it trivially
follows that if objects a and b are weakly discerned by relation R,
they must be distinct entities: aa b. This fact may be seen as the
formal basis of the grounding claim: the non-qualitative fact that a
and b are numerically distinct can be reduced to (or inferred from)
the fact that they are connected by the qualitative relation R.
However, we should keep in mind that the grounding is successful
4
In my Bigaj (2015) I point out that there is yet another possible use that PII
can be put to: it can namely help distinguish one object from the rest of the
universe in a way which makes it possible to uniquely refer to it. I argue that weak
discernibility is incapable of reaching this goal; however, there are grades of
discernibility weaker than absolute but stronger than weak discernibility that may
achieve such identification of objects.
5
I admit that some pronouncements of the proponents of the WD program can
be interpreted as claims of a more methodological than metaphysical character.
That is, the existence of weakly discerning relations may be seen as ensuring that
facts regarding numerical identity and distinctness can be derived from empirical,
qualitative statements, without presupposing that the latter facts ground the
former. However, my subsequent criticism of the WD program can easily be
showed to apply to its methodological interpretation as well. If, as I claim it to
be the case, the relations used to weakly discern quantum particles make
inescapable reference to numerical identity, the derivation of the non-qualitative
facts of identity from qualitative facts involving weakly discerning relations is
circular. Thus, in order to argue that particles a and b are numerically distinct, we
have to presuppose the very fact we want to establish.
6
For recent logical analyses of the variety of grades of discernibility see
Ketland (2011), Caulton & Butterfield (2012), Ladyman, Linnebo, & Pettigrew (2012)
and Bigaj (2014).
Huggett & Norton in (2014, p. 40) mistakenly define relative discernibility of
objects a and b by the formula ( ℛ, c ℛ(a, c) 4 ℛ(b, c). This formula can be actually
proven to be equivalent to weak discernibility (under the assumption that variable
ℛ ranges over all two-place formulas definable in given language). I am grateful to
Chris Wüthrich for bringing that error to my attention and for a subsequent
discussion.
 T. Bigaj / Studies in History and Philosophy of Modern Physics 50 (2015) 43–53 45

Pd
only if the used relation R is a legitimate one. In particular, P ijð1Þ P ð2Þ
ij Pj ψ  〉 ¼ i;j ¼ 1 P ð2Þ ð1Þ
ij P ij j ψ  〉 ¼ 2j ψ  〉. From this it can be
reduction fails if R itself contains numerical identity as an essential
concluded that the relation R–2 weakly discerns fermions, since
component. In such a case an inference of the statement of
the following is the case: :R  2 ð1; 1Þ, :R  2 ð2; 2Þ and R  2 ð1; 2Þ,
numerical distinctness a ab from the fact that R(a, b) is clearly
R  2 ð2; 1Þ.
circular.
The result achieved above does not extend to bosons, because
the operator (6) does not act on symmetric states as a multiple of
the identity. Muller and Saunders suggest using a probabilistic
2. Weakly discerning relations in quantum mechanics
weakly discerning relation for bosons, but their theorem to the
effect that weak probabilistic discernibility of bosons is attainable
In this section I will give a brief outline of the details of the
has been superseded by the later and stronger result presented in
weakly discerning methods used by Muller, Saunders and See-
Muller and Seevinck (2009), to which we will turn now.
vinck, starting with Muller and Saunders’ 2008 paper. They begin
Muller and Seevinck make a surprisingly strong claim, announ-
their construction by selecting a complete set of one-dimensional
cing that they can define perfectly legitimate quantum–theoretical
orthogonal projectors Pi in a single-particle Hilbert space ℋ, i.e.
P relations which categorically discern particles in any joint states
such that P i ¼ I and P i P j ¼ δij P i . Using the family of projectors Pi,
whatsoever, not limited to symmetric or antisymmetric ones. They
they introduce operators of the form Pij ¼Pi Pj. In the case when
consider two cases: spinless particles with infinitely-dimensional
the number of identical particles is assumed to be 2, the states of
state spaces, and particles characterized by spin states defined in
composite systems are represented by vectors (or, more generally,
finitely-dimensional Hilbert spaces. In the first case they consider
density operators) in the tensor product ℋ  ℋ. The operator Pij
the position Q and momentum P of each individual particle.
when applied to the first particle will have the form P ijð1Þ ¼ P ij  I,
Limiting ourselves to the case of two particles we can introduce
while P ð2Þ
ij ¼ I  P ij will represent the same property of the second the usual notation:
particle. With the help of these notions we can now define the
following binary relation on the set of particles: P ð1Þ ¼ P  I; P ð2Þ ¼ I  P;
X
d
Q ð1Þ ¼ Q  I; Q ð2Þ ¼ I  Q ;
Rt ðx; yÞ iff P ðxÞ ðyÞ
ij P ij j ψ 〉 ¼ tj ψ 〉 ð1Þ
i;j ¼ 1
and then define relation C:
where tA ℝ. As should be clear from the above formula, relation Rt h i
is supposed to hold between objects x and y iff the system is in an C ðx; yÞ iff P ðxÞ ; Q ðyÞ j ψ 〉 ¼ cj ψ 〉 for some c a 0: ð7Þ
P
eigenstate for the operator di;j ¼ 1 P ijðxÞ P ðyÞ
ij with the corresponding The relation C can be presented informally as “The commutator
eigenvalue equal t. This, given the standard interpretation of the
of x’s momentum and y’s position is a non-zero number”. The
quantum formalism, reflects the fact that the system possesses the
standard commutation relation between the position and momen-
definite value t of the property described by the operator.
tum of a given particle implies that [P(1), Q(1)] ¼[P(2), Q(2)] ¼–iħI  I.
For complete clarity we have to explain how to interpret the
P On the other hand, the position of one particle commutes with the
formula di;j ¼ 1 P ðxÞ ðyÞ
ij P ij . In the most general case when the number momentum of another particle, thus [P(1), Q(2)] ¼[P(2), Q(1)] ¼0.
of particles equals n, the above shorthand expression can be From these facts and definition (7) it follows immediately that
expanded as follows: C(1, 1), C(1, 2), :C(1, 2), :C(2, 1), and consequently :C(x, y) weakly
X
d d 
X   discerns particles 1 and 2. It has to be stressed that this fact is
P ijðxÞ P ðyÞ
ij ¼ I  …  P ðijxÞ  …  I I  …  P ðijyÞ  …  I ; independent of the exact form of the state |ψ〉, therefore weak
i;j ¼ 1 i;j ¼ 1 discernibility holds for any particles in any state.
ð2Þ For particles with spin sa0 there is yet another option
available to weakly discern them. We start with the definition of
where variables x and y indicate that the xth and the yth places in
the spin for a single particle (with a spin number assumed to be s)
the n-fold tensor product are occupied by appropriate one-particle
as S¼ Sx þSy þSz, where Sx, Sy, and Sz are spin components in
operators. In particular, when n ¼ 2, we end up with two distinct
directions x, y, z. As usual, spins characterizing separate particles in
possibilities:
a two-particle system will be represented by appropriate tensor
X
d X
d X
d
products with the identity as one component:
P ð1Þ ð2Þ
ij P ij ¼ P ð2Þ ð1Þ
ij P ij ¼ P ijð1Þ  P ð2Þ
ij ð3Þ
i;j ¼ 1 i;j ¼ 1 i;j ¼ 1 S1 ¼ S  I

X
d X
d S2 ¼ I  S:
P ð1Þ ð1Þ
ij P ij ¼ ðP ðij1Þ Þ2  I ð4Þ
i;j ¼ 1 i;j ¼ 1
The weakly discerning relation is defined as follows:
 2
It is relatively easy to check that the following equalities hold: T ðx; yÞ iff Sx þ Sy  j ψ 〉 ¼ 4sðs þ 1Þħ2 j ψ 〉 ð8Þ
X
d X
d
P ð1Þ ð1Þ
ij P ij ¼ P ð2Þ ð2Þ
ij P ij ¼ 2ðd  1ÞI  I ð5Þ Note that subscripts x and y in (8) indicate variables ranging
i;j ¼ 1 i;j ¼ 1
over particles to be discerned, and not directions in space. Two
! particles can be said to be connected by relation T iff they are in a
X
d X
d X
d
P ð1Þ ð2Þ
ij P ij ¼ P ð2Þ ð1Þ
ij P ij ¼ 2 Pi  Pi  I  I ð6Þ joint state such that their total spin squared takes on a definite
i;j ¼ 1 i;j ¼ 1 i¼1 value equal 4s(sþ1)ħ2. In order to see that it is possible to weakly
discern particles with the help of relation T we should recall first
When the considered state j ψ  〉 of the system belongs to the
that the square of a single’s particle spin is a multiple of the
antisymmetric section ℋ– of the tensor product of Hilbert spaces,
Pd identity: |S|2 ¼s(s þ1) I, and therefore relation T holds between any
the result of the action of the part i ¼ 1 P i  P i of operator (6) particle and itself. On the other hand, when the particles are
equals zero, and hence we have the following equations: distinct, we have |S1 þS2|2 ¼2s(s þ1)ħ2I  I þ2S  S. This implies
Pd ð1Þ ð1Þ Pd ð2Þ ð2Þ Pd
i;j ¼ 1 P ij P ij ψ  〉 ¼ i;j ¼ 1 P ij P ij ψ  〉 ¼ 2ðd  1Þj ψ  〉 and i;j ¼ 1 that the eigenvalues of |S1 þS2|2 never exceed 2s(2s þ1)ħ2, which
46 T. Bigaj / Studies in History and Philosophy of Modern Physics 50 (2015) 43–53
is less than 4s(sþ 1)ħ2. Therefore relation T never holds between a
pair of distinct particles, and T weakly discerns them.
Weak discernibility done with the help of relations R–2 (Def. 1)
and C (Def. 7) has the nice feature of relying entirely on facts
involving possessing definite values of a particular physical quan-
tity (being in an eigenstate for the corresponding operator). As
explained above, the fact that the relation R–2 holds between two
objects means that the system is in an eigenstate of the operator
(2) with the corresponding eigenvalue equal –2. On the other
hand, what underlies the cases for which R–2 does not hold is the
fact that the system is in an eigenstate for the operator (2) but
with a different eigenvalue (equal 2(d  1)). The situation with the
relation C is analogous, with positive instances of the relation
being associated with the eigenvalue –iħ, and negative instances
corresponding to the eigenvalue 0. Consequently, we can adopt a
minimalistic interpretation of quantum mechanics and still retain
the needed discernibility result. This minimalistic interpretation
assumes only that being in an eigenstate implies possessing the
corresponding eigenvalue (the less controversial half of the eigen-
value/eigenvector link), and that one observable cannot possess
two distinct values simultaneously.
In comparison, relation T (Def. 8) requires stronger assumptions
in order to successfully discern particles. While cases in which a
pair of particles exemplifies T are instances where the assembly is
in an eigenstate of the squared total spin operator, nevertheless
particles that do not stand in relation T do not usually possess a
definite value of the selected quantity. This means that we have to
assume the full e/e link,8 in order to conclude from the fact that a
system is not in an eigenstate that it does not possess the
corresponding eigenvalue.
3. Friendly criticism
Most critics stress that the results presented by Muller &
Saunders and Muller & Seevinck are mathematically unassailable.
However, the exact physical and philosophical meaning of these
technical results remains open to debate. Admittedly, Muller and
Saunders go to great lengths to prove that the relation of weak
discernibility introduced in their paper satisfies the condition of
physicality, and Muller and Seevinck follow suit. To facilitate this
task, Muller and Saunders explicitly introduce two conditions of
physical admissibility: Physical Meaning and Permutation Invar-
iance (2008, pp. 527–528). Physical Meaning requires that all
properties and relations used to discern objects should be “trans-
parently defined in terms of physical states and operators that
correspond to physical magnitudes”. Permutation Invariance, on
the other hand, stipulates that relations used to discern particles
ought to be permutation invariant, which in the case of binary
relations means their symmetricity.9
In their discussion preceding the formulation of the two
conditions of physical admissibility, Muller and Saunders give
some more detailed explanations of how to understand Physical
Meaning and Permutation Invariance. Regarding the “transparent”
method of defining discerning properties and relations in terms of
physical states and operators, their emphasis is placed squarely on
negative instances. That is, they insist that no labeling predicates
(as opposed to labels themselves) should be used in physical
discernment. Similarly, purely mathematical or formal concepts,
such as being a member of a particular set, are to be excluded as
8
Muller & Saunders (2008, p. 513) refer to the full e/e link as the Strong
Property Postulate, and the one-way implication mentioned above as the Weak
Property Postulate.
9
Both conditions are repeated verbatim in Muller & Seevinck (2009, p. 185).
well. But they do not formulate any specific demands regarding
the operators and states used in physically meaningful formulas.
Thus it seems that any Hermitian (self-adjoint) operator defined in
an appropriate Hilbert space should be acceptable as a means to
discern particles.
With respect to the condition of Permutation Invariance, its
official formulation places the restriction of being invariant under
permutations on the discerning relation only. However, the
discussion given by Muller and Saunders may suggest a stronger
interpretation: the operators and states used to construct a
discerning relation should obey the postulate of permutation
invariance as well. Of course the latter condition implies that the
discerning relation will be symmetric, but a symmetric relation
could also be built out of non-permutation-invariant operators.
As we will soon see, the distinction between a weak and a strong
interpretation of Permutation Invariance will play a central role in
the latest friendly critiques of Muller and Saunders and Muller and
Seevinck.
In spite of the apparent care with which Muller, Saunders and
Seevinck treat the issue of the physical meaningfulness of dis-
cernment, it has become the focal point of the critiques delivered
by Caulton (2013), and by Huggett and Norton (2014). Both papers
center their objections around the observation that the relations
used to weakly discern elementary particle fall short of being
physically meaningful despite claims to the contrary. Two reasons
are adduced to support this evaluation: (1) the violation of the
condition of permutation invariance, and (2) the inappropriate use
of multiples of the identity. We will consider these objections
in turn.
Huggett and Norton direct their criticism entirely against the
relation Rt defined in Muller and Saunders (2008). They point out
P
that the operator di;j ¼ 1 P ðxÞ ðyÞ
ij P ij violates the condition of permuta-
tion invariance when the number of particles is greater than or
equal 3. This can be clearly seen when we look at formula (2).
Applying any permutation that exchanges particle x with any
particle other than y will get us a different operator. Huggett and
Norton admit that this problem does not affect cases in which we
limit ourselves to antisymmetric, fermionic states, for in such
cases, thanks to Eqs. (5) and (6), our operators reduce to multiples
of the products of the identity, which are obviously permutation
invariant. However, they point out that operators (2) clearly act in
an asymmetric way in the subspace of bosonic (symmetric) states.
They give a simple example showing that the subspace of sym-
metric states ℋ þ is not invariant under the action of operators of
type (2), and therefore they do not count as operators on ℋ þ . As a
consequence, the admittance of such operators makes it theore-
tically possible to devise measurements which could produce
bosons in non-symmetric states, and that is clearly non-physical.
It may be observed that this does not mean that the relation
Rt(x, y), defined with the help of the non-symmetric operators
Pd ðxÞ ðyÞ
i;j ¼ 1 P ij P ij , is itself non-symmetric. The symmetricity of Rt(x, y)
is ensured by the straightforward fact that the permutations
exchanging x and y preserve the operators in question. Thus
Muller and Saunders’ Permutation Invariance in its weaker form
is still satisfied. But Huggett and Norton have a point in claiming
that the condition of physicality may require a stronger inter-
pretation of Permutation Invariance. Caulton in his critical analysis
makes a similar point regarding the symmetricity of operators
used to weakly discern particles. He uses an analogous argument
not only against the relation Rt, but also against the relations
introduced by Muller and Seevinck, or at least their relation C.
However, his specific remarks on that topic are slightly baffling.
It may be worthwhile to spend some time pondering over their
exact meaning.
At first glance it looks like Caulton favors a super-strong
interpretation of Permutation Invariance, according to which not
 T. Bigaj / Studies in History and Philosophy of Modern Physics 50 (2015) 43–53
only the entire operator used directly to weakly discern particles,
but its components (“building blocks”) as well, should remain
invariant under arbitrary permutations. Here is a fragment from
his paper confirming this stance:
But Rt is defined in terms of nonsymmetric projectors Pi  I, and
so forth. And it is compulsory—that is a necessary condition for
representing a physical quantity—that the quantities obey the
Indistinguishability Postulate (IP), which demands that all
physical quantities be permutation invariant (p. 55—I changed
the original notation for consistency).
Pd ðxÞ ðyÞ
Later in the text Caulton admits that quantities i;j ¼ 1 P ij P ij
are symmetric (provided that the number of particles equals 2, as
we can recall from Huggett and Norton’s analysis), but he still
insists that they are not admissible, because
…the physical significance of these quantities was supposed to
rest on their being constructions out of quantities like Pi I, yet
it is precisely these quantities that run afoul of IP (p. 56, again
my change in notation).10
Caulton forcefully repeats the same charge with respect to the
relation C, pointing out that it is defined in terms of inadmissible,
non-symmetric operators P(x), Q(y). However, in a surprising twist
he refuses to apply the same criticism to the third weakly
discernible relation T (Def. 8). While noting that the definition of
T employs unacceptable non-symmetric operators Sx and Sy, he
nevertheless states that his usual objection does not hold in this
case. The reason for this assessment that he provides is that “it
seems reasonable to take T(x, y) as a natural physical relation”
(p. 59). This is so, because T can be easily parsed in English as the
relation involving the combined total spin of two particles, which
is symmetric and has obvious physical significance.
I cannot say that I am entirely clear on the difference between
relations C and T with respect to the applicability of the super-
strong requirement of permutation invariance. Why cannot we
give an analogous argument to the effect that C should be exempt
from the requirement exactly as T is supposed to be? After all, the
relation C(x, y) can be parsed in English as “the commutator of x’s
momentum and y’s position equals –iħ”, and the commutator is
symmetric. Without further insight into the meaning of the term
“a physically significant quantity” it is hard to say whether the
commutator of position and momentum (which in the case of one
particle reduces to a multiple of the identity) is a physically
significant quantity. (Later we will see that Huggett and Norton
consider all quantities of that type physically inadmissible.).
However, it may be pointed out that the requirement of super-
strong permutation invariance is in itself rather implausible. Based
on such a requirement we would have to exclude symmetrized
operators of the form ð1=2ÞðA  I þ I  AÞ, which are clearly built
out of non-symmetric components A I and I  A, and this looks
too extreme, as they have a clear physical interpretation (we will
see that later in the text). So I suggest that we adhere to the
original requirement introduced by Huggett and Norton that the
whole operator (and not its components) used in the definition of
a weakly discerning relation ought to be permutation invariant.
And it has to be admitted that this requirement is violated by all
three relations Rt, C, and T when the number of particles is greater
than 2.
Let us now move on to the second type of criticism. Huggett
and Norton observe that not all operators definable on an appro-
priate state space represent meaningful physical quantities, even if
the condition of symmetricity is satisfied. For instance, they insist
10
A similar thought occurs also in (Huggett & Norton, 2014, p. 48), but they do
not make it a focal point of their criticism.
47
that the identity operator does not represent any physically
significant property due to its “triviality”. Caulton similarly agrees
with the triviality claim, noting that the identity operators corre-
spond to experimental questions that yield the same answer on
every physical state (p. 56). And yet Eqs. (5) and (6) clearly show
P
that the operators di;j ¼ 1 P ðxÞ ðyÞ
ij P ij are multiples of the identity when
applied to antisymmetric states. Thus the relation Rt does not
seem to be physically meaningful.
Remaining on a rather general level of analysis one may
complain that the charge of triviality should be distinguished
from the charge of non-physicality. It does not seem inconsistent
to claim that a trivial concept is nevertheless physically mean-
ingful. I do not see anything unphysical in asking an experimental
question of the sort “What is the probability that the spin of this
electron will be either þ or –?”, as such a question clearly admits
an experimental procedure aiming at answering it, even though
the answer is already well known in advance. This question is
clearly more physical than asking whether this electron bears label
1 or 2. Thus in my opinion Huggett and Norton’s objection should
not be construed as showing that the relation Rt does not satisfy
Muller and Saunders’ requirement of physical meaningfulness, but
rather as suggesting the need for a third requirement (Non-
triviality), separate from the original two. But what can justify
such a need?
Here both Huggett with Norton and Caulton give an interesting
argument, based on the specific method of defining the weakly
discerning relation R–2. They observe that because of Eqs. (5) and
(6) the relation R–2 when applied to fermions can be restated as
follows:
R–2 ðx; yÞ iff ðx ¼ y and 2ðd–1Þj ψ〉 ¼ –2j ψ〉Þ or ðx ay
and –2j ψ〉 ¼ –2j ψ〉Þ; ð9Þ
which, in turn, can be reformulated in an even simpler way as
R–2 ðx; yÞ iff ðx ¼ y and 2 ðd–1Þ ¼ –2Þ or ðx ay and –2 ¼ –2Þ:
ð10Þ
This is clearly a non-physical relation, as there is not a single
term on the rhs of the equivalence which would refer to any
physical property of discerned particles whatsoever. Note that
even reference to the physical state |ψ〉 of the system is missing
from the formula. Essentially, what we have here is a disjunction
whose components contain identities/non-identities and trivial
arithmetic equalities. Discernment by such a relation should
evidently be disallowed.
I entirely agree with this assessment, even though I think the
problem we have just discovered runs much deeper than the mere
triviality issue. But for the time being let us play along. We can
observe that exactly the same problem affects the discerning
relation C, as its definition can be reformulated analogously as
C ðx; yÞ iff ðx ¼ y and –iħ a 0Þ or ðx a y and 0 a 0Þ: ð11Þ
However, with respect to relation T the situation looks differ-
ent. Here the operator defining T (the total spin squared |Sx þSy|2)
is a multiple of the identity only when x ¼y. In the remaining cases
the operator acts differently on different symmetric and antisym-
metric states. This fact is actually responsible for the above-
discussed need to invoke a non-minimalistic interpretation of
quantum mechanics which accepts both implications of the e/e
link. However, it turns out that the very same feature can protect
us now against the threat of triviality.
To sum up, all three proposals of how to weakly discern
elementary particles turn out to be open to some serious objec-
tions. Now it is the time to evaluate the alternative solutions put
forward by the critics themselves.
48 T. Bigaj / Studies in History and Philosophy of Modern Physics 50 (2015) 43–53
4. Alternative proposals
Following their criticism of Muller, Saunders and Seevinck,
both teams of authors wheel out their own preferred methods of
discerning elementary particles. However, they clearly decide to go
in different directions. Huggett and Norton’s strategy can be seen
as more radical, while Caulton essentially repeats the standard
approach of Muller, Saunders and Seevinck with a slightly differ-
ent choice of the main operator on which the discerning relation is
based. In their novel approach, Huggett and Norton opt to ensure
the satisfaction of the strong Permutation Invariance by comple-
tely symmetrizing the appropriate operators. But such a move has
one unwelcome consequence: namely, it eliminates any reference
to specific particles x and y from the definition of such operators.
How, then, can we construct a formula in two variables Φ(x, y) out
of expressions that do not even contain any variables? The
solution submitted by Huggett and Norton is to make use of the
formulas of the kind given in (9)–(11), with the trivial arithmetic
equalities replaced by legitimate physical conditions.
More specifically, Huggett and Norton consider two arbitrary
Hermitian operators representing meaningful physical quantities A
and B. If we wanted to formally represent the fact that A is a
property of particle x while B is a property of particle y, we would
have to use non-symmetric operators of the form I  …  A  …
 B  …  I. But Huggett and Norton insist that we have to
properly symmetrize such operators to satisfy the Invariance
Principle. However, this cannot be done at one stroke—we have
to consider separately cases when x¼y and cases when xay.11 As
a result we obtain the following two fully symmetric operators:
1X
ðAB  I  …  IÞ ð12Þ
np A P
and
1 X
  ðA  B  I  …  IÞ; ð13Þ
n pAP
2 that seem (i.e. from the point of view of the syntax) to treat the
2 x ¼y and x ay cases equally, the fact that a different quantity
where the summation is done over the set of all permutations of n
particles. Now we can define the weakly discerning relation as
follows:
! (1) of Rt at our disposal, there is nothing wrong in reformulating it
P in terms of a disjunction. But note that this condition is not
St ðx; yÞ iff x ¼ y4 1n ðAB  I  …  IÞj ψ 〉Þ ¼ tj ψ 〉 3
pAP
0 1
B C
B 1 X C
Bx a y4   ðA  B  I  …  IÞ ¼ tj ψ 〉C ð14Þ
@ n pAP A objects x and y at all (this is a direct consequence of the
2
2 symmetrization procedure applied beforehand). Thus it looks like,
Relation St weakly discerns two objects in cases when the
eigenequation in the first disjunct of the rhs of (14) is false while
the second eigenequation is true.12 Thus in order to discern
particles occupying a joint state |ψ〉 we have to find two Hermitian
operators A and B such that |ψ〉 is an eigenstate for both operators
(12) and (13) but with different corresponding eigenvalues. Hug-
gett and Norton show how this can be done under the assumption
that |ψ〉 is a fermionic state obtained by antisymmetrizing a direct
product of orthogonal vectors (in their proposal they assume A
and B to be identical). However, their result does not extend to
11
This is a telltale sign of a very significant underlying fact, and I do not want
the reader to miss it. We will return to this issue soon.
12
Note that the reference to the selected particles x and y that was initially
included in the description of the quantities A and B has now been completely lost
due to the symmetrization procedure. This is striking, since as a result the only
remaining occurrence of variables x and y in formula (14) is in the context of the
identity relation. We will return to this problem soon.
arbitrary fermionic states, and they admit that they do not know
whether an alternative procedure of weak discernment by a
relation of the form (14) exists. On the other hand, Huggett and
Norton prove that the relation St defined for any A¼ B does not
discern bosons in symmetric product states. This is a consequence
of the fact that if such a state is an eigenstate of (13), it has to be an
eigenstate for (12) with the same eigenvalue.13
There is a clear trade-off between the scope of Huggett and
Norton’s discernibility result on the one hand and the purported
superiority of their applied method. While the achieved discern-
ibility of quanta is much more limited in scope than that
announced by Muller and Seevinck, supposedly the used method
should be impervious to any possible objections by obstinate
critics of weak discernibility. But is it really? I agree that Huggett
and Norton’s own standards of a proper physically discerning
relation are indeed satisfied by their proposed definition (14).
However, in some important respects the suggested strategy may
be argued to be inferior to the original ones. In all truthfulness,
the only real benefit of Huggett and Norton’s new proposal that I
can think of is that it (unintentionally) draws our attention to a
serious but so far neglected conceptual problem that afflicts to a
certain degree all known attempts to weakly discern quantum
objects. Here is what I mean.
One of the most conspicuous features of definition (14) is that it
characterizes the weakly discerning relation St in terms of two
separate conditions, one of which is applied when the considered
objects are identical, while the other when they are distinct. This
immediately raises the worry that we are defining here not one
relation but two distinct relations only artificially conjoined in a
disjunctive manner. A similar objection is briefly considered by
Caulton in his analysis of the relation Rt, when he discusses its
disjunctive definitions of the form (9) and (10). Recognizing the
possibility of an opposition from a critic he rushes to explain that:
so long as we have a definition of the Rt in terms of quantities
(…) underlies each of these two cases is tolerable (p. 57).
In other words, provided that we have the original definition
satisfied in Huggett and Norton’s case. There is no alternative,
uniform way of characterizing the relation St which would not
consider the separate cases x ¼y and xay. Even more curiously,
the conditions corresponding to these cases do not mention
speaking metaphorically, the only formulas that do any discerning
jobs in (14) are the identities x ¼y and x ay. From a logical point of
view, any two conditions T and F such that one is true while the
other is false would do equally well. The expression “(x ¼y and F)
or (x ay and T)” weakly discerns objects, but it does so in a purely
formal and quite trivial way, regardless of what fancy restrictions
we impose on the mathematical and physical meaning of condi-
tions T and F. The only strategy that could rescue such a
disjunctive characterization is proving that T and F are actually
instances of one and the same condition Φ(x, y) applied, respec-
tively, to the cases x ay and x ¼y. But this is not an option for the
relation St.
13
One may note in passing that, logically, Huggett and Norton’s result obtained
for symmetric product states is rather weak, as it shows only that one particular
method fails to achieve weak discernibility in that case. This of course falls short of
proving generally that there is no method of weak discernment for bosons in
product states that satisfies the conditions imposed by Huggett and Norton.
 T. Bigaj / Studies in History and Philosophy of Modern Physics 50 (2015) 43–53
It has to be admitted that Huggett and Norton acknowledge the
existence of the above-mentioned difficulty in the context of the
original relation Rt. But they are strangely dismissive about the
whole issue, noting only that “[we don’t] see any problem with
constructing sensible relations in this way [i.e. disjunctively]” (p.
8). However, they slightly misidentify the problem as being mostly
about creating legitimate properties using disjunctions. In fact, the
problem is, primarily, that in order to introduce relation St we have
to make use of the identity symbol, and in a way which seems to
be essential (indispensable). And this in turn shatters any hope of
using the weak discernibility of quanta in any metaphysically
meaningful way, for instance as the qualitative basis for grounding
their numerical distinctness and their individualities. If we are
happy with weakly discerning relations of the form (14), then why
cannot we go all the way and accept weak discernibility by the
relation x ay? As I argued above, the fact that we conjoin the x ¼y
and xa y cases with meaningful physical conditions is a red
herring—the real job of weak discernment is done entirely by
the identity relation.
One important lesson that can be drawn from the above
analysis of Huggett and Norton’s weakly discerning relation is
that we have to beware of the inappropriate use of the identity
relation while formulating such relations. And under closer scru-
tiny it may be discovered that even the original proposal of a
weakly discerning relation by Muller and Saunders is not entirely
free from a similar problem. But before we get to this point, let us
briefly review Caulton’s alternative strategy of weakly discerning
quantum particles. His idea is to use the well-known concept of
the variance (dispersion) of a given quantity A, which is commonly
defined as the expectation value of the following quantity:
(ΔA)2 ¼(A  〈A〉)2 (see Sakurai & Napolitano, 2011, pp. 33–34).
It can be easily verified that the dispersion of A can be alternatively
presented as 〈(ΔA)2〉¼〈A2〉 〈A〉2. Caulton uses this familiar con-
cept to measure the anticorrelations between eigenvalues of A for
two particles. He introduces the variance of a quantity A as an
operator (q-number, not a c-number):
1
Δ2A ¼ ðA  I  I  AÞ2
4
Defining a new operator A as a measure of the statistical mean of
A:
A ¼ 12 ðA  I þ I  AÞ
2
it can be quickly calculated that Δ ¼ A  A , which justifies the
2 2
A relations, on which their admissibility should depend, is the use of
term “variance”. Moreover, it can be observed that if |φi〉 and |φj〉
are eigenvectors of A with eigenvalues equal, respectively, to ai and
aj, then the product |φi〉  |φj〉 will be an eigenvector of ΔA , and its
2
corresponding eigenvalue will be 1=4(ai  aj)2.
The idea to use the above-introduced notion of the variance of
a quantity A as a means to weakly discern particles is based on the
observation that the anticorrelation between eigenstates for a
single particle is always equal to zero. Hence the discerning
relation can be defined as follows:
ðxÞ ðyÞ 2
V ðx; yÞ iff 4 ðA A Þ j
1
ψ 〉 a 0;
(x)
where A is defined in the usual way as either A  I or I  A,
depending on whether x equals 1 or 2. Note that when x ay, the
rhs of (17) reduces to Δ2A j ψ〉 a 0. On the other hand, with x¼ y we
get the zero operator acting on the vector j ψ 〉. The weak discern-
ment by the relation V is guaranteed if we can find for all joint
states j ψ 〉 a physical quantity A whose dispersion in j ψ 〉 is non-
zero. And, as Caulton argues, this can be done thanks to the fact
that no particle state is dispersion-free with respect to both
position and momentum.
Caulton insists that the use of nonsymmetric components A(x)
and A(y) in (17) is not a problem, since the quantity used in the
49
discerning relation V has clear physical meaning that does not
depend on A(1) and A(2) having physical meaning as separate
entities. This may be so, but it turns out that definition (17) has
one fatal flaw which is closely related to the earlier discussed
problem afflicting Huggett and Norton’s definition (14). To see that
let us reformulate (17) in a disjunctive form as follows:
   
V ðx; yÞ iff x; y and ΔA j ψ 〉 a0 or x ¼ y and 0j ψ 〉 a 0 : ð18Þ
2
The fact that such a reformulation is possible is in itself incon-
sequential, as long as the physical conditions attached to the cases
x¼ y and x ay can be seen as instances of the same physical
property of the system. But here lies the rub: it may be argued that
the trivially false inequality 0 a0 associated with the x ¼y case
does not represent the same type of physical condition as the
inequality ΔA j ψ 〉 a 0. The operator ΔA measures dispersion
2 2
between the values of the same quantity for two particles. But in
order to measure the dispersion of the value of a quantity for a
single particle we do not use the trivial zero operator
2
4 ðA  I  A  IÞ . Instead, we resort to the earlier defined opera-
1
tor (A  〈A〉)2, and it is an elementary fact that if the state of a single
particle is not an eigenstate for A, its dispersion is non-zero in such
a state. Thus the formula V(x, y) does not represent a single,
physical relation, but rather a gerrymandered entity consisting of
two unrelated relations conjoined with the help of the identity
relation.
But what is with the intuition of “measuring anticorrelations”
between eigenstates? I think that the terminology used by Caulton is
misleading. To see this more clearly, let us consider an example of a
symmetric product state of two bosons |φ〉|φ〉. What anticorrela-
tions can be present in such a symmetric state? The only reason that
the operator ΔA can assume a non-zero expectation value in such a
2
state is that |φ〉 may not be an eigenvector for A, and hence the states
of individual particles are not dispersion-free for A. But this is
precisely the same reason why A is not dispersion-free in a single-
particle state |φ〉. If we used the proper concept of dispersion for both
cases x¼y and xay, we would have to come to the conclusion that
ð15Þ the appropriately defined relation does not discern weakly at all.
5. Weak discernibility and identity
ð16Þ
As should be clear now, the critical feature of weakly discerning
the identity relation. But we have to be more precise as to what
uses of identity should be banned from weakly discerning rela-
tions. The key term here is “essential”: all non-essential uses of
identity may be permitted. An obvious example of a non-essential
use is as follows: we can always present a legitimate weakly
discerning relation R(x, y) in the form of the disjunction (x¼ y and
R(x, x)) or (xa y and R(x, y)). But no one should be bothered by
such artificial and easily eliminable reference to identity. However,
the general distinction between essential and non-essential uses
of identity is not easy to draw, and we have to proceed with care in
ð17Þ
order not to exclude otherwise acceptable instances of weak
discernibility. I would like to stress first that, in contradistinction
to some authors who criticize weak discernibility en bloc, I am
willing to grant that there are legitimate, acceptable cases in
which two entities may be said to have acquired their numerical
distinctness (and objecthood) via an appropriately selected weakly
discerning relation connecting the two. The case of Max Black’s
spheres, weakly discerned by the relation “being 5 feet apart” is a
prime example of such a situation. The existence of the relation of
spatial separation ensures that there are two objects and not one,
even though we are still unable to label them individually, or make
reference to one and not the other.
50 T. Bigaj / Studies in History and Philosophy of Modern Physics 50 (2015) 43–53
Some critics of weak discernibility argue that all attempts to
weakly discern two objects inevitably fall victim to the problem of
circularity.14 That is, in order to apply a weakly discerning relation
we have to make sure that there are indeed two objects and not
one, and therefore we have to find a way to discern them before-
hand. Or, to put it differently, we have to include the fact that the
two objects are numerically distinct in our description of the
arrangement to which we will subsequently apply the weak
discernibility strategy. But I do not see why this has to be the
case. Let us consider the Black’s spheres scenario again. We can
describe it as neutrally as possible in the following way: consider a
universe in which there is nothing more than a sphere x with
such-and-such dimensions, weight, chemical structure etc., and a
sphere y qualitatively indistinguishable from x. At this point I do
not decide whether x is distinct from y: they may very well turn
out to be one and the same object, in which case no trouble for PII
arises. But now I introduce one more element to my description: x
and y are 5 feet apart. Bingo! Now we know that there are two
spheres, and the reason we know this is precisely because the
relation of being 5 feet apart is weakly discerning! To put it in
metaphysical terms, the quantitative fact regarding the numerical
distinctness of x and y has emerged out of the qualitative and
relational fact involving the non-zero distance between x and y.15
But a critic may insist that the relation of being separated by a
non-zero distance used in the above example still hides some
inappropriate reference to identity. It may be argued for instance
that the reason why such a relation is capable of weakly discerning
objects in the first place is that ultimately the formal characteristic
of the distance between points contains as one of its axioms the
postulate that the distance between points x and y is zero if and
only if x is identical with y. Without the notion of identity we
would not be able to properly introduce the mathematical notion
of metric, or distance. Does that mean that discerning objects with
the help of their spatial separation is inadmissible?
This argument may be extended for any arbitrary relation that
purports to weakly discern all objects in a given domain. It may be
pointed out that any attempt to formally characterize a relation R
in a way which would guarantee that all distinct objects are
indeed weakly discerned by R is bound to contain reference to the
identity relation. This is so, because otherwise we would be able to
define purely qualitatively an absolute notion of identity, and this
is impossible, at least when we restrict ourselves to first-order
14
The circularity charge is leveled e.g. by Hawley (2006, 2009), French &
Krause (2006, pp. 169–171), and, in the context of the metaphysical debate on
spacetime, by Wüthrich (2009). An extensive response to this charge can be found
in Muller (2013). I have no room here for a detailed analysis of Muller’s defense
against the circularity charge. However, I believe that his argument is ineffective
against my criticism of the weakly discerning relations used in the context of
quantum mechanics. As I understand it, Muller effectively argues that the
proponents of WD do not commit a petitio principii, since they simply conclude
that two numerically distinct objects are qualitatively discerned by a permissible
relation. I, on the other hand, question precisely this conclusion: in the case of
purported weakly discerning relations of QM particles are not qualitatively
discerned, since the relations themselves make unavoidable reference to numerical
distinctness, ergo are not purely qualitative. Incidentally, this shows that my
criticism of WD, while related to the general charge of circularity, is nevertheless
different in character. I do not argue that in order to apply any weakly discerning
relation we have to know in advance that the objects in question are numerically
diverse; only that the weakly discerning relations used in quantum mechanics are
inappropriate, because they themselves are not purely qualitative.
15
Of course the derivation of the fact that there are two spheres from the fact
that they are weakly discerned by the relation of being 5 feet apart is a matter of
pure logic (I assume here that the relation of spatial separation is truly irreflexive,
and thus I ignore the possibility that the spheres could be identified with bundles
of universals for which it is possible to be multiply localized). However, this does
not mean that the extralogical, metaphysical assumption of PII does not play a role
here. The universal truth of PII ensures that in every case when two distinct objects
are present their distinctness can be reduced to (or derived from) some pure
qualitative facts.
languages. But I believe that it would be unfair to hold this fact
against the proponents of weak discernibility. The use of the
identity symbol in general, abstract characterizations of the con-
cepts we apply to weakly discern seems to me rather innocuous.
What is problematic is when we resort to the facts of identity/
distinctness when applying a weakly discerning relation to a
particular, concrete case. If we have to know that a specific object
x is distinct from an object y before we can apply a weakly
discerning relation to x and y, the procedure becomes unaccep-
table. Of course we have to keep in mind that the notion of
acceptability used here is a relative one. What we mean in this
context is that weak discernment achieved in such an “unaccep-
table” way does not have any interesting metaphysical conse-
quences regarding the status of the objects in question. Yet from a
formal perspective even discerning objects directly with the help
of the relation x ay is as good as any other discernment. But I take
it that the proponents of the WD program are interested not in
merely “formal” discernment but something metaphysically more
substantial.
6. Some generalizations
I will now try to spell out the procedure of weak physical
discernment in more general terms capable of covering the
specific cases presented so far. The aim of this exercise is to draw
a reasonably clear line between acceptable and unacceptable
forms of weakly discerning relations, following the general dis-
cussion in the previous section.
The procedures described in earlier sections can be subsumed
under the following general pattern. The starting point is a
definition of a mapping σ : P  P-ℳ from ordered pairs of
physical entities to mathematical objects of a particular kind.
The object σ(i, j)A ℳ assigned to entities i and j is supposed to
represent a legitimate, meaningful physical property of the pair.
Then a statement E[σ(i, j)] is formulated, which involves the
selected mathematical object. With only a slight loss of generality
we may assume that E[σ(i, j)] is a mathematical equation expres-
sing the fact that the property σ(i, j) takes on a particular
admissible value (categorical discernment), or a particular prob-
abilistic distribution over all admissible values (probabilistic dis-
cernment). Finally, a binary relation is defined as follows: R(i, j) iff
E[σ(i, j)] is true. This relation can be said to weakly discern objects
a and b iff R(a, b) but :R(a, a) and :R(b, b).
One obvious condition which has to be satisfied is that the
equation E[σ(i, j)] is not supposed to include the identity symbol
connecting object-variables i, j (for obvious reasons identities
connecting terms referring to mathematical objects are not only
permitted but even indispensable). But this is not sufficient to
ensure the acceptability of the relation R. After all, illegitimate
reference to identity may be already included in the description of
the mapping σ(i, j). One method of revealing such a forbidden use
of identity may be suggested as follows. If the mapping σ(i, j) is
defined in a way which treats differently cases when i¼j from
cases when iaj, it may be suspected that illegitimate hidden
reference to identity is made. But how are we supposed to tell
when this happens? And what does it precisely mean that cases
i¼j and i aj are “treated differently”? Looking at the explicit
formalization of the mapping σ is of little help here, because a
notation can always be craftily devised to unify distinct cases into
a seemingly uniform formula which appears not to contain
identity. A marginally better solution could be to stipulate that
objects σ(i, j) should not belong to separate mathematical cate-
gories depending on whether i¼ j or ia j. For instance, we would
not intuitively accept as uniform a mapping which would assign a
vector in ℂ3 to pairs of identical objects, and a 4  4 matrix of reals
 T. Bigaj / Studies in History and Philosophy of Modern Physics 50 (2015) 43–53
to pairs of distinct objects. Unfortunately, cases that we have
already encountered are more subtle. While it may be claimed that
operators (3) and (4) differ in an important respect, they never-
theless belong to the same broad category of linear operators
acting on the tensor product of n identical Hilbert spaces. We need
a more refined approach.
The key to the new requirement introduced below is the observa-
tion that the object σ(i, j) is typically the result of some mathematical
operations performed on components related to objects i and j taken
separately. In the case of Muller and Saunders’ relation Rt(i, j), these
operations include multiplication and addition of Hermitian operators
acting in the tensor product of spaces. But it has to be noted that in
order to properly execute an appropriate operation we have to know
in advance exactly which place in the product is occupied by which
single-particle operator. To illustrate this, let us consider again the
familiar example:
AðiÞ ¼ I  …  Ai  …  I;
BðjÞ ¼ I  …  Bj  …  I;
where A and B are Hermitian operators in the single-particle space.
What is the result of the multiplication of these two operators? We
can write it down simply as A(i)B(j), but we should not forget that the
actual result depends in a non-trivial way on whether i is the same
number as j. If i¼j, then the result of the multiplication will be given as
I … AiBi  … I, i.e. the product of n–1 identities and one multiple
operator AB in the ith place. However, when iaj, the multiplication
algorithm is quite different: I … Ai  … Bj  … I (assuming that
ioj), and thus A and B are not multiplied, as they do not act in the
same single-particle space. Without prior knowledge regarding the
identity of i and j we cannot even properly execute the required
operation. I would like to stress that the problem is not merely that
the result of the prescribed operation is different, depending on
whether i¼j or iaj. After all, this is to be expected from a weakly
discerning relation. I make a stronger claim: the operation itself is
meaningfully defined through an executable algorithm only when we
know which case obtains.16
In order to make this crucial point clearer, let us contrast the
above-mentioned case with the already mentioned case of a
perfectly legitimate weakly discerning relation: spatial separation.
We start with defining for each physical object i its corresponding
position encoded as a 3-vector ri (relative to a particular coordi-
nate system). The mapping σ will be naturally characterized as
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
follows: σ(i, j)¼|ri  rj|¼ ðxi  xj Þ2 þ ðyi  yj Þ2 þ ðzi  zj Þ2 , and the
corresponding weakly discerning relation will receive the follow-
ing characterization: D(i, j) iff |ri  rj| 40. It is easy to observe that
the definition of the operation leading to the object σ(i, j) (i.e. the
subtraction of one vector from another, followed by taking its
modulus) does not depend on the assumption regarding the
numerical identity of i and j. I do not have to know whether i is
the same object as j in order to execute the operation of subtract-
ing the vector associated with j from the vector associated with i.
Thus no use of identity seems to be essential in order to mean-
ingfully introduce the relation D.17
16
For those who are not fully convinced by this claim I have the following
question to consider. Suppose that you are satisfied with the universal character-
ization of the product A(i) B(j) which is neutral with respect to the identity of i and j.
What would you do if someone insisted that they still do not understand the
procedure involved in this product? I am sure that you would ask them to consider
separate cases and come up with the two results discussed above. This shows to me
that our understanding of the universal multiplication formula is actually depen-
dent on the understanding of individual cases.
17
I would like to stress that the theory in which the locations of individual
points are introduced may very well make an indispensable use of identity. For
instance, it seems necessary that this theory be able to express the fact that no two
distinct entities can have the same location function. But such a use of identity
51
To sum up, the condition that we have introduced here is that all
mathematical operations involved in creating the object σ(i, j) should
be insensitive to the facts of numerical identity/distinctness regarding
objects i and j, or in other words that the execution of all the required
operations should be essentially the same regardless of whether i¼ j or
iaj. I am willing to admit that these characterizations do not exactly
live up to the standards of precision found in mathematical logic, but it
seems to me that the intended meaning should be sufficiently clear for
anybody familiar with mathematical practice. And now it can be easily
verified that the variety of relations proposed in order to weakly
discern quantum particles of the same type unambiguously violate the
aforementioned condition. That is, the operators introduced to weakly
discern quanta clearly arise as a result of different procedures
depending on whether we are dealing with one object or two distinct
objects. This problem may be the most conspicuous in the case of
Huggett and Norton’s proposal, where the quantum-mechanical
conditions corresponding to the x¼y and xay cases are openly
different, but it affects all the other strategies equally strongly.
At the risk of being repetitious, I would like to emphasize that the
formulas (1), (7), (8) and (17) are not built with the help of a single
operator, as the notation might suggest. Rather, the equations defining
appropriate relations cover two separate instances involving two
distinct operators. It has to be underscored that my criticism does not
rely on the additional premise that these distinct operators necessarily
correspond to distinct physical properties. In fact, I have to admit that I
do not know how to decide whether for instance operators (5) and (6)
represent distinct properties or the same property of the system of
many particles, and I do not care. But what is important is that in order
to decide whether relation R–2 holds between two quantum particles x
and y, I have to perform a mathematical operation which requires from
me that I already know in advance whether x is identical with or
distinct from y. Thus the successful application of R–2 to the pair (x, y)
cannot ensure that they are distinct objects, since I have to know that
even before I start the whole procedure. Again, there is nothing wrong
in admitting that R–2 does discern x and y weakly, as there is nothing
wrong in admitting that the relation xay discerns them too. But the
metaphysical moral from the discernment achieved in such a way is
meager. If we wanted to find a way to reduce facts of numerical identity
and distinctness to some qualitative, genuinely physical states of affairs,
we should look elsewhere.
And I submit that this “elsewhere” would have to be a theory other
than the standard quantum theory of many particles. The way the
standard formalism represents properties of systems of many particles
makes it virtually impossible to find a way of encoding relations that
would not be prone to the problem of the sensitivity to the facts of
identity. This is so, because in the tensor product formalism each
particle is assigned its own unique Hilbert space, and operators acting
in the total product consist of components each of which acts in its
individual space. Thus operators representing properties of the system
when x¼ y are bound to look and behave differently than operators
representing “the same” properties in the case when xay—the former
will have their components limited to one Hilbert space of states for
particle x¼ y, whereas the latter will contain two types of components,
(footnote continued)
should be permitted, as I explained previously, otherwise no meaningful weak
discernibility would ever be attainable. Another possible objection to this example
could be that it actually makes the weak discernibility of distinct spatial points a
consequence of their absolute discernibility by position vectors ri. But this is
incorrect. As is well known, vectors ri do not represent legitimate physical
properties of objects, since they are not invariant under the symmetries of the
underlying theory, i.e. the Galilean transformations. Only differences between
vectors taken at the same time are invariant, and therefore can be interpreted as
representing genuine properties of systems. In order to avoid this confusion, we
could resort to an alternative, and more general method of defining the distance
between two points as the length of the shortest of all continuous curves
connecting these points (cf. Maudlin, 2007, p. 87ff).
52 T. Bigaj / Studies in History and Philosophy of Modern Physics 50 (2015) 43–53
one acting in x’s space, the other in y’s space. And to decide which
“global” operator to choose, we have to rely on the previous knowl-
edge regarding which is the case: x¼y or xay.
Interestingly enough, this problem may be avoided when we move
to the Fock space formalism. Here the states of multiparticle systems
are represented by occupation numbers associated with certain
properties of individual particles (see Sakurai & Napolitano, 2011, p.
460ff). If |ki〉 are eigenvectors for a selected single-particle quantity K,
then the state of a system can be given as the vector |n1, n2, …, ni, …〉,
where ni is the number of particles occupying state |ki〉. Supposing that
K is an additive quantity, we can define a multiparticle operator K
representing the corresponding property of the entire system as
P
follows: K ¼ i ki N i , where ki is an eigenvalue of K and Ni is the
number operator associated with state |ki〉: Ni ¼ a†i ai. It is characteristic
that the form of the operator K remains the same regardless of the
actual number and distribution of particles among the available states.
Thus we do not have to know in advance how many particles occupy
which states in order to be able to apply K to a given state. That way
we can hope to be able to find a property K such that it will produce a
different value for a state in which some selected particles a and b are
identical, and different when these particles are distinct (this differ-
ence can be of course reflected only in different occupation numbers,
since the Fock space formalism does not admit individual labels). For
instance, a state in which a¼b may be given as |0, 0, …,1, …〉, whereas
states for aab can take the form of either |0, 0, …, 2, …〉 (available
only for bosons) or |0, 0, …,1, …,1,…〉 (for bosons and fermions). The
fact that K looks and acts the same regardless of which state we are
applying it to (of course the results of its actions will generally be
different!) makes it possible to dodge the problem of an essential use
of identity in definitions of weakly discerning formulas.
7. Further reflections
So far the results of the analysis of the numerous attempts to
implement the WD program have been rather negative. Is there not
anything positive that can be learned from the sustained effort of so
many proponents of weak discernibility? I believe there is. To see
that let us go back to the requirement of permutation invariance that
was so vigorously argued for by Huggett and Norton. This is an
important take-home lesson: the only physically meaningful state-
ments about systems of particles of the same type are those which
can be formulated with the help of fully symmetric operators and
states. Even though everybody pays lip service to this principle, not
all of its consequences are properly acknowledged and accepted. And
one of the crucial but often neglected consequences is that the very
question of discernibility of particles has to be cashed out in a
language that obeys the principle of permutation invariance.18 But
can the question “Is particle a discernible from particle b?” even be
properly formulated without violating the fundamental principle on
which the whole theory of quantum particles is based?
To see more clearly what the source of the problem is, let us
start with the standard way of parsing the question of absolute
discernibility. As absolute discernibility regards the possession of
properties by individual objects, we have to be able to express
statements of the form “Particle a possesses property P”. But such
statements blatantly violate the principle of permutation invariance.
And if we try to apply to them the procedure of symmetrization,
suggested by Huggett and Norton as a way to accommodate PI, the
result will be that all ascriptions of a given property to distinct
18
Huggett (2003) acknowledges the fact that the well-known proof of the
Indiscernibility Thesis given in French & Redhead (1988) relies on an inappropriate
formalization of the properties of individual particles which violates PI. He
proposes an alternative method using the so-called conjugacy condition, which
unfortunately trivializes the problem of discernibility, as I explain in the main text.
individual particles will be translated into one and the same state-
ment “Some particle possesses property P”. Under such an analysis
the thesis that no particles of the same type can be absolutely
discerned becomes a trivial logical truth, since the above-mentioned
statement cannot be both false and true.
But can we not do any better than that? I submit we can. There are
clear and simple methods of expressing the thought that some objects
are discernible in a way that does not infringe upon the principle of
permutation invariance and yet does not trivialize the problem of
discernibility. Instead of using labels, or unbound variables, we can
resort to the well-known logical procedure of Ramsification, which
can get us for instance from the statement “a is absolutely discernible
from b” to its Ramsey version “There are objects x and y such that x is
absolutely discernible from y” (or, equivalently, “Some objects are
absolutely discerned from some objects”). And the last Ramsified
sentence can in turn be explicated as “There are objects x and y and a
property P such that x possesses P while y does not possess P”. So
there is hope that the last sentence could be expressed in the
permutation-invariant language of the quantum theory of many
particles. But can this hope be realized? The answer is “yes”, if only
we resist the temptation to find a direct quantum-mechanical
translation of the separate components of the above-stated sentence.
As we already know, there is no quantum-mechanical formalization of
the expression “x possesses property P” satisfying the principle of
permutation invariance. However, the whole sentence has a clear PI-
preserving representation in the quantum formalism.
Let P be a projector in a single-particle space representing a given
quantum-mechanical property P. For the sake of simplicity let us
consider a system consisting of two particles of the same type. If we
wanted to find an operator representing the joint property of the
system “particle 1 possesses P and particle 2 does not possess P”, this
could be easily done with the help of the following tensor product:
P  ðI–PÞ. However, in order to satisfy PI we have to symmetrize this
product, thus arriving at the following projector acting in
ℋ  ℋ : P  ðI–PÞ þ ðI–PÞ  P. It should be evident that the last
operator encodes the property of the entire system expressed in the
statement “Some particle possesses property P and some particle does
not possess property P”. And if the system is in a state which makes
this whole sentence true, this can mean only one thing: the objects in
question are absolutely discernible by property P.
It is an elementary fact that there are some antisymmetric/
symmetric states which are eigenstates for appropriately selected
operators of the form P  ðI–PÞ þðI–PÞ  P. Thus bosons and
fermions can be absolutely discerned in some states, contrary to the
commonly accepted Indiscernibility Thesis. It is true that this absolute
discernibility result does not extend to all bosonic and fermionic
states. It is also true that there are some thorny issues pertaining to the
fermionic states. It turns out for instance that in some states fermions
can be simultaneously discerned by sets of incompatible properties,
which raises the question of whether the well-known no-go theorems,
such as the Bell or Kochen–Specker theorems, are infringed upon.
However, the prospect of even partially rehabilitating PII in its
strongest, most traditional guise, seems worth the effort put in an
attempt at a solution. The WD program was an ambitious undertaking
to reverse the dominating trend in the philosophical analysis of the
problem of identity and individuality in quantum mechanics. I suggest
that now is the right time to embark on a new and even more
ambitious program: the Absolute Discernibility (AD) program.19
19
There are good reasons to believe that the AD program is already under way.
Statements to the effect that quantum particles of the same type can differ with
respect to their properties are occasionally made by physicists and philosophers
alike (see e.g. Ghirardi, Marinatto, & Weber, 2002; Pooley, 2006). Recently the idea
that the permutation-invariant states of similar fermions and similar bosons admit
the possibility of their absolute discernibility has been noted and analyzed from a
philosophical point of view in Dieks & Lubberdink (2011), Friebe (2014), Saunders
 T. Bigaj / Studies in History and Philosophy of Modern Physics 50 (2015) 43–53
The AD program, as I see it, should primarily aim to clarify the
metaphysical consequences of the fact that some states of “identical”
fermions and bosons make it possible to discern them by individual
properties, where their discernibility is encoded in an appropriate
symmetric operator for which the state is an eigenstate. The first
observation that has to be made here is that the particles discerned in
such a way cannot be associated with the indices used in the
symmetric/antisymmetric mathematical representation of the state.
Instead, they should be identified by descriptions involving precisely
those properties that were used to discern the particles in question.20
However, it would be premature to conclude that in the current
approach particles should simply be considered bundles of properties.
For one, if in a given state there are two (or more) distinct properties P
and Q discerning two particles (as is certainly the case for fermions),
we cannot say for sure whether P and Q are bundled together or not,
as we can only state that one (but not the other) particle possesses P,
and one (but not the other) particle possesses Q, without identifying
which is which. Thus the question of the metaphysical nature of
quanta in the AD approach remains open and should be subject to
further investigations.
Another issue which merits further investigation is whether the
limited validity of the strong property-based variant of PII can restore
the individuality of quantum particles. For it is still true that in some
states (for instance in symmetric product states for bosons and in
some genuinely entangled states for fermions and bosons21) particles
possess the exact same quantum-mechanical properties. It is unclear
to me whether we need the unconditional and universal validity of PII
in order to speak about individual objects. For instance Dieks and
Lubberdink (2011) speak about individual particles as “emerging” the
moment the joint state shifts from a PII-violating to a PII-preserving
one. An alternative solution (which, under closer scrutiny, may turn
out to be just a variant of Dieks and Lubberdink’s emergence
hypothesis) is to apply one of the well-known strategies to defend
the validity PII in the troublesome cases. For instance Friebe (2014)
uses the summing defense, arguing that in the non-discerning states
there are no numerically distinct particles but an undivided (or even
undividable) whole. These proposals are at a very early stage of
development, and much has to be done in order to turn them into a
viable metaphysical conception of quantum particles.
Acknowledgments
I would like to thank the members of the Southern California
Philosophy of Physics Group at Irvine, and two anonymous referees,
for their useful comments to earlier versions of this paper. The work
on this paper was supported by the Marie Curie Grant FP7-PEOPLE-
2012-IOF-328285.
References
Bigaj, T. (2014). On discernibility and symmetry, Erkenntnis, http://dx.doi.org/10.
1007/s10670-014-9616-y.
(footnote continued)
(2013), and Caulton (2014). I have given my own “metaphysical” arguments for the
absolute discernibility thesis based on the assumption of essentialism in Bigaj
(2015).
20
This contention is explicitly formulated and endorsed in Dieks & Lubberdink
(2011) and Friebe (2014).
21
An exhaustive analysis of the difference between genuinely and non-
genuinely entangled states of many particles of the same type is the main topic
of the extensive paper (Ghirardi et al., 2002). For a different approach to the same
problem see Ladyman, Linnebo, & Bigaj (2013) and (Caulton 2014). It should be
mentioned here that Caulton’s unorthodox approach to the problem of permuta-
tion invariance leads to an even stronger conclusion than the one presented above;
namely that fermions of the same type are discernible in all states (even
“genuinely” entangled ones).
53
Bigaj, T. (2015). Exchanging quantum particles. In: P. E. Bour, G. Heinzmann,
W. Hodges, & P. Schroeder-Heister (Eds.), 14th CLMPS 2011 Proceedings, 19
(pp. 185–198). Philosophia Scientiae.
Butterfield, J. (1993). Interpretation and identity in quantum theory. Studies in
History and Philosophy of Science, 24, 443–476.
Caulton, A. (2013). Discerning “indistinguishable” quantum systems. Philosophy of
Science, 80, 49–72.
Caulton, A. (2014), Qualitative individuation in permutation-invariant quantum
mechanics. arXiv:1409.0247v1 [quant-ph].
Caulton, A., & Butterfield, J. (2012). On kinds of indiscernibility in logic and
metaphysics. British Journal for the Philosophy of Science, 63, 27–84.
Dieks, D., & Lubberdink, A. (2011). How classical particles emerge from the
quantum world. Foundations of Physics, 41, 1051–1064.
Dieks, D., & Versteegh, M. (2008). Identical quantum particles and weak discern-
ibility. Foundations of Physics, 38, 923–934.
French, S., & Krause, D. (2006). Identity and physics: A historical, philosophical and
formal analysis. Oxford: Clarendon Press.
French, S., & Redhead, M. (1988). Quantum physics and the identity of indiscern-
ibles. British Journal for the Philosophy of Science, 39, 233–246.
Friebe, C. (2014). Individuality, distinguishability and (non-)entanglement: A
defense of Lebiniz’s principle. Studies in History and Philosophy of Modern
Physics
Ghirardi, G., Marinatto, L., & Weber, T. (2002). Entanglement and properties of
composite quantum systems: a conceptual and mathematical analysis. Journal
of Statistical Physics, 108(112), 49–122.
Giuntini, R., & Mittelstaedt, P. (1989). The Leibniz Principle in quantum logic.
International Journal of Theoretical Physics, 28, 159–168.
Hawley, K. (2006). Weak discernibility. Analysis, 66, 300–303.
Hawley, K. (2009). Identity and indiscernibility. Mind, 118, 101–119.
Huggett, N. (2003). Quarticles and the identity of indiscernibles. In: W. K. Brading, &
E. Castellani (Eds.), Symmetries in physics (pp. 239–249). Cambridge: Cambridge
University Press.
Huggett, N., & Norton, J. (2014). Weak discernibility for quanta, the right way.
British Journal for the Philosophy of Science, 65, 39–58.
Ketland, J. (2011). Identity and indiscernibility. The Review of Symbolic Logic, 4(2),
171–185.
Ladyman, J., & Bigaj, T. (2010). The Principle of the Identity of Indiscernibles and
quantum mechanics. Philosophy of Science, 77, 117–136.
Ladyman, J., Linnebo, O., & Bigaj, T. (2013). Entanglement and non-factorizability.
Studies in History and Philosophy of Modern Physics, 43(3), 215–221.
Ladyman, J., Linnebo, O., & Pettigrew, R. (2012). Identity and discernibility in
philosophy and logic. The Review of Symbolic Logic, 5, 162–186.
Margenau, H. (1944). The Exclusion Principle and its philosophical importance.
Philosophy of Science, 11, 187–208.
Maudlin, T. (2007). The metaphysics within physics. Oxford: Oxford University Press.
Muller, F. A. (2011). Withering away, weakly. Synthese, 180, 223–233.
Muller, F. A. (2015). The rise of relationals. Mind, 124, 201–237.
Muller, F. A., & Saunders, S. (2008). Discerning fermions. British Journal for the
Philosophy of Science, 59, 499–548.
Muller, F. A., & Seevinck, M. P. (2009). Discerning elementary particles. Philosophy of
Science, 76, 179–200.
Pooley, O. (2006). Points, particles, and structural realism. In: D. Rickles, S. French, &
J. Saatsi (Eds.), The structural foundations of quantum gravity (pp. 83–120).
Oxford: OUP.
Redhead, M., & Teller, P. (1992). Quantum physics and the identity of the
indiscernibles. British Journal for the Philosophy of Science, 43, 201–218.
Sakurai, J. J., & Napolitano, J. (2011). Modern quantum mechanics (2nd ed). San
Francisco: Adison-Wesley.
Saunders, S. (2003). Physics and Leibniz’s principles. In: K. Brading, & E. Castellani
(Eds.), Symmetries in physics: Philosophical reflections (pp. 289–307). Cambridge:
Cambridge University Press.
Saunders, S. (2006). Are quantum particles objects? Analysis, 66, 52–63.
Saunders, S. (2013). Indistinguishability. In: R. Batterman (Ed.), Oxford handbook of
philosophy of physics (pp. 340–380). Oxford: Oxford University Press.
van Fraassen, B., & Peschard, I. (2008). Identity over time: Objectively, subjectively.
Philosophical Quarterly, 58, 15–35.
Wüthrich, C. (2009). Challenging the spacetime structuralist. Philosophy of Science,
76, 1039–1051.