Reconstruction of quantum theory

Alexei Grinbaum

LPHS – Archives Henri Poincaré (CNRS UMR 7117),
23 bd Albert Ier, 54015 Nancy, France

Email alexei.grinbaum@polytechnique.edu

Abstract

What belongs to quantum theory is no more than what is needed for its
derivation. Keeping to this maxim, we record a paradigmatic shift in
the foundations of quantum mechanics, where the focus has recently from
interpreting to reconstructing quantum theory. Several historic and con-
temporary reconstructions are analyzed, including the ones due to Hardy,
Rovelli, and Clifton, Bub and Halvorson. We conclude by discussing the
importance of a novel concept of intentionally incomplete reconstruction.

1 What is wrong with interpreting quantum me-
chanics?

Ever since the first days of quantum mechanics physicists as well as philoso-

phers tried to interpret it, understanding this task as a problem of giving to

the new physical theory a clear meaning. One of the principal reasons why one

has always felt the need for interpretations has to do with the puzzling aspect

of the formalism of quantum mechanics, usually referred to as the measure-

ment problem. Reversible unitary evolution of the wave function, according to

standard quantum mechanics, at the moment of measurement is replaced by an

irreversible transformation known as wavefunction collapse. First and foremost,

interpretations of quantum mechanics aimed at making sense of this surprising

change in the theory’s dynamics, sometimes taking the collapse at face value and

claiming its fundamental irreducible role, or sometimes going to another extreme

and denying the collapse altogether. However, looking globally, the enterprize

of interpreting quantum mechanics failed: today we still have no consensus on

what the meaning of quantum theory is. None of the proposed answers has won

overall acceptance. Perhaps the most remarkable manifestation of the failure to

interpret quantum mechanics is the attitude taught to most young physicists in

1



lecture rooms and research laboratories in the last half century, “Shut up and

calculate!” [37]

Why did attempts at a univocal interpretation fail? Many answers are pos-

sible, and among them we favor two, both showing that there is an intrinsic

deficiency in the idea of interpreting a physical theory with the help of philo-

sophical instruments only.

The first answer is that to a physical theory one would naturally like to give

a physical meaning in the Greek sense of ϕύσις, i.e. we—as part of the physi-

cists’ audience—expect to be told a true story about nature. This is because we

casually tend to apply physical theory to the phenomenal world to learn some-

thing about the latter, and not the world to physical theory in order to invent

a meaning of the theory. Physical theory is above all a tool for predicting the

yet unobserved phenomena; so employing existing knowledge and experience of

the world to interpret physics runs counter to its basic function as a scientific

theory. However, notwithstanding such an against-the-grain direction in which

a philosophical interpretation operates, the former does not necessarily lead to

formal contradiction that would invalidate the interpretation program logically;

more modestly but perhaps no less irritatingly, at the end one is often left with

a feeling of being excluded from mainstream research. Further, as the physics of

today is inseparable from mathematics, a meaning cannot be physical and thus

satisfactory if it is merely heaped over and above the mathematical formalism

of quantum mechanics, instead of coming all the way along with the formalism

as it rises in a derivation of the theory.

The second answer is that we live in a situation where objective truth has

been appropriated by science, and to pass public ratification every increase in

knowledge must confront experimental setups. In this world an interpretation

can only be considered satisfactory when it becomes an integral part of science.

This is not unprecedented in the history of ideas: indeed, many philosophical

questions with the advent of empirical science ceased to be perceived as philo-

sophical and are now treated as scientific. Even such a pronounced critic of

the idea that physics can have implications for philosophy as Steven Weinberg

2



admits that, exceptionally, physics and philosophy can be connected in that

“discoveries in physics sometimes reveal that topics that had been thought to

be proper subjects for philosophical argument actually belong in the province of

ordinary science” [57]. To be able to convince Weinberg and many of those sci-

entists who remain sceptical about the philosophical debate over interpretations

of quantum mechanics, the problem of interpreting quantum mechanics must be

moved into the area of science; only then will the puzzling discord disappear.

2 Reconstruction of physical theory

2.1 Schema

We call reconstruction the following schema adopted to the needs of quantum

theory and different from the notion of rational reconstruction introduced by

Carnap [9]. Theorems and major results of physical theory are formally derived

from simpler mathematical assumptions. These assumptions or axioms, in turn,

appear as a representation in the formal language, of a set of physical princi-

ples. Thus, reconstruction consists of three stages: first give a set of physical

principles, then formulate their mathematical representation, and finally derive

from here the formalism of the theory.

Contrary to the case of interpretation, the three-stage structure of recon-

struction permits the latter to acquire supplementary persuasive power which

arises from the use of mathematical derivation. Established as valid mathemat-

ical results, theorems and equations of the theory become unquestionable and

free of suspicion. ‘Why is it so?’—‘Because we derived it.’ The question of

meaning, previously asked with regard to the formalism, is removed and now

bears, if at all, only on the selection of the principles. No room for mystery

remains in what concerns the meaning of the theory’s mathematical apparatus.

As one of the consequences, in the reconstruction program the measurement

problem loses the central role it has occupied for the success of an interpreta-

tion of quantum mechanics. In reconstruction, one makes sense of all of the

formalism on the basis of first principles, and whatever mathematical element

3



is contained in the formalism that is used in a particular reconstruction, this

element becomes meaningful thanks to the first principles.

Explanatory power of the reconstruction is a power of explanation of where

the structure of the theory comes from; not necessarily a power of explanation

by the theory, of the real world. Traditionally, interpretations focused on the

latter task and gave less attention to the former one. Reconstruction shifts this

focus area: its added value for better understanding quantum theory originates

in the new insights into the structure of the theory, made possible thanks to the

use of mathematical derivation.

2.2 Selection of the first principles

Anyone who wishes to attempt a reconstruction of physical theory must formu-

late the foundational principles which he or she believes plausible and translate

them into mathematical axioms. Then the rest of the theory will be constructed

‘mechanically,’ by means of a formal derivation. The choice of axioms must be

the only allowed freedom in the whole construction. It is commonplace to say

that it is not easy to exhibit an axiomatic system that would stand to such

requirements, especially in the case of quantum theory.

First, where do candidate axioms for quantum theory come from and how

does one judge which statements that can plausibly be taken as axioms? Prior

to pronouncing such a judgment, one must develop an intuition of what is plau-

sible about quantum theory and what is not. This can only be achieved by

practicing the theory, i.e. by taking its prescriptions at face value, applying

them to systems under consideration in particular tasks, and obtaining results.

In short, one needs to acquire a real ‘know-how’ above and beyond the theo-

retical knowledge that quantum mechanics could solve such and such problems.

Researcher’s intuition develops from experience; it cannot arise from abstract

knowledge ‘in principle.’

However, taking the prescriptions of quantum theory at face value, apply-

ing them and obtaining results will not yet make things clear about quantum

mechanics. Indeed, one can possess the knowledge about how to apply a cer-

4



tain tool without caring about the structure of the tool nor its meaning. The

quantum mechanical know-how serves purely as such a tool for developing one’s

intuition about which candidate idea is a plausible foundational principle and

which other candidate idea will not pass the test. Candidate foundational prin-

ciples need not even be theorems of the already existing quantum mechanics:

one’s judgment may be such that a new statement—false or only conditionally

true in quantum theory—will be taken as axiom in reconstruction of a new

theory. Examples of such principles will be investigated in Section 3.5.

Second, what shall we require from first principles? They must be simple

physical statements, i.e. assertions whose meaning is immediately, easily acces-

sible to a scientist’s understanding. They must also be such as to permit a clear

and unambiguous translation of themselves into mathematically formulated ax-

ioms. A derivation of quantum theory will then rely on these axioms.

2.3 Status of the first principles

A reconstruction program includes a derivation of quantum theory, but in the

previous section one was told to apply and use it in order to motivate the

derivation. Is there a vicious circle here? We submit that there is none, and this

thanks to the status of the first principles. Namely, they should not necessarily

be viewed as ultimate truths about nature. Independently of one’s ontological

commitments, the first principles have only a minimal epistemic status of being

postulated for the purpose of reconstructing the theory in question. As with

the 19th-century mathematics, in theoretical physics the axiomatic method is

to be separated from the attitude which the Greeks had toward axioms: that

they represent the truth about reality. Much of the progress of mathematics is

due to understanding that an axiom can no longer be considered an ultimate

truth, but merely a basic structural element, i.e. an assumption that lies at the

foundation of a certain theoretical structure. In mathematics, after departing

from the Greek concept of axiom, “not only geometry, but many other, even very

abstract, theories have been axiomatized, and the axiomatic method has become

a powerful tool for mathematical research, as well as a means of organizing the

5



immense field of mathematical knowledge which thereby can be made more

surveyable” [26]. A similar attitude is to be taken with respect to axioms used

for the formal derivation of a physical theory. To give a concise formula, a

methodological prescription that gives the minimal status of the first principles

in a reconstruction program, runs as follows:

• If the theory itself does not tell you that the states of the system, or any
other variables, are ontic, then do not take them to be ontic.

To explain the above prescription, return first to the idea that, in developing

an intuition with respect to the plausibility of the foundational principles used

to derive a theory, one takes this theory as a given and applies it practically, so

as to acquire a know-how that would justify the choice of principles. Now, when

one is working with several physical theories, ideas that have previously been

used as foundational in theory I, may turn out to be derivative (i.e., theorems) in

theory II. Examples include the case of thermodynamics and statistical physics

or the relation between macroscale hydrodynamics and the low-level molecular

theory of liquids. Such considerations show the limits of philosophical assump-

tions that one can make about the status of the first principles used in the

reconstruction of a given physical theory. Indeed, generically nothing can be

said about their ontological content or the ontic commitments that arise from

the principles. It is more economical and would amount to a certain episte-

mological modesty to treat the foundational principles as axioms hic et nunc,

i.e. in a given theoretical description. Epistemological modesty requires that

one brackets his or her personal motivation for the choice of first principles and

reconstructs the theory based on the principles themselves. Reconstruction be-

comes meaningful solely thanks to the sense of the first principles on which it

is based.

Reconstruction of a physical theory has its main advantage compared to

philosophical interpretation of the theory in the fact that it moves a number of

questions, previously thought of as philosophical, to the realm of science, and

this in virtue of the mathematical derivation which the reconstruction program

6



operates. However, philosophical problems do not altogether disappear; they

still apply to the first principles and take the form of a problem of their jus-

tification. Evidently, it is a minimal logical condition that such a justification

should not be seen as a mathematical deduction of the principles from the theory

in whose very foundations they lie. Once one has obtained a full formalism of

the theory in an epistemologically modest reconstruction, it is then possible to

ask the reconstructed theory itself if it allows a realist interpretation of the first

principles from which it has been derived or, perhaps, it imposes constraints

on possible ontological commitments. And while in general the status of the

first principles as ultimate truths about reality is not a necessity, certain recon-

struction programs are such that this status can be safely, or almost, attributed

to the principles within a particular reconstruction in question. The task of

justification is therefore external to the reconstruction program and must be

executed by one with a different set of presuppositions, i.e. by taking the the-

ory as a given and motivating from there why the principles that were involved

in the reconstruction are simple, physical, and plausible. Therefore, philosophy

is not fully chased out of physics. On the contrary, by demarcating the fron-

tier between what can be treated as a scientific question and what belongs to

metatheory, one contributes to a better understanding of the structure of the

theory and of those of its foundational postulates which require a metatheoretic

interpretation and justification.

It is often claimed that applying to theoretical physics the same methodology

as in axiomatizations in mathematics leads to a problem, exposed by Einstein.

While supporting the axiomatic move in mathematics in that it “dispels the ob-

scurity which formerly surrounded the principles,” Einstein argues that if one

wants to apply a similar move in physics, then one has to face the difficulty of

connecting “conceptual schemata” with “real objects” [13]. Applied to an epis-

temologically modest reconstruction, this problem is no more than apparent, if

the status of the first principles is properly freed from ontological commitments.

The latter do have a bearing indeed, but only on justifying the choice of postu-

lates. Starting from a particular set of principles (stage 1) represented formally

7



(stage 2), mathematical derivation (stage 3) proceeds in exactly the same way as

in mathematics. Reconstruction understood as stages 1-3 is therefore analogous

to an axiomatization in mathematics. However, unlike mathematical axiomati-

zation, the reconstruction program also invokes the problem of justification of

the choice of first principles. But when the first principles are formulated in an

epistemologically modest way, Einstein’s “conceptual schemata,” or structural

elements of physical theory, become the only building blocks of the latter. Un-

ambiguous derivation of the theory’s formalism is detached from the question

of reality of the world that the theory describes, with respect to which one is

free to take different viewpoints. For quantum theory, this detachment amounts

to operating a reconstruction of quantum theory from a set of first principles

devoid of the necessity of being justified on the ontological grounds.

3 Examples of reconstruction

3.1 Early examples of reconstruction

In the last decade reconstruction has become a major trend in the foundations

of quantum mechanics. Before describing this recent work, let us first look

further back in the history of quantum mechanics: there too axiomatic deriva-

tions occupy an eminent place. The first paper where quantum mechanics was

treated axiomatically appeared shortly after the creation of quantum mechanics

itself: in 1927 Hilbert, von Neumann and Nordheim stated their view of quan-

tum mechanics as one in which “. . . [the theory’s] analytical apparatus, and

the arithmetic quantities occurring in it, receives on the basis of the physical

postulates a physical interpretation. Here, the aim is to formulate the physical

requirements so completely that the analytical apparatus is just uniquely de-

termined. Thus the route is of axiomatization” [27, our emphasis]. It is on this

route of axiomatization that von Neumann in collaboration with Birkhoff was

led to study the logic of quantum mechanics [6]. Following their work, many ax-

iomatic systems were proposed, e.g. by Zieler [58], Varadarajan [54, 55], Piron

[39, 40], Kochen and Specker [31], Guenin [19], Gunson [20], Jauch [29], Pool

8



[44, 45], Plymen [43], Marlow [36], Beltrametti and Casinelli [4], Holland [28],

or Ludwig [33]. Another branch of axiomatic quantum theory, the algebraic ap-

proach was first conceived by Jordan, von Neumann and Wigner [30] and later

developed by Segal [48, 49], Haag and Kastler [21], Plymen [42], Emch [14] and

others; for a recent review, see [8].

However, a vast majority of these axiomatic developments do not fall under

our notion of reconstruction, as they were based on highly abstract mathemati-

cal assumptions and not, as we require, on simple physical principles. Consider

for instance the exemplary work by Mackey [34, 35]. He develops quantum me-

chanics as follows. Take a set B of all Borel subsets of the real line and suppose

we are given two abstract sets O (a to-be space of observables) and S (a to-be
space of states) and a (to-be probability) function p which assigns a real number

0 ≤ p(x, f, M ) ≤ 1 to each triple x, f, M , where x is in O, f is in S, and M is
in B. Assume certain properties of p listed in axioms M1-M9:

M1 Function p is a probability measure. Mathematically, we have p(x, f, ∅) =
0, p(x, f, R) = 1, and p(x, f, M1 ∪M2 ∪M3 . . .) =

∑∞
n=1 p(x, f, Mn) when-

ever the Mn are Borel sets that are disjoint in pairs.

M2 Two states, in order to be different, must assign different probability dis-

tributions to at least one observable; and two observables, in order to be

different, must have different probability distributions in at least one state.

Mathematically, if p(x, f, M ) = p(x′, f, M ) for all f in S and all M in B
then x = x′; and if p(x, f, M ) = p(x, f′, M ) for all x in O and all M in B
then f = f′.

M3 Let x be any member of O and let u be any real bounded Borel func-
tion on the real line. Then there exists y in O such that p(y, f, M ) =
p(x, f, u−1(M )) for all f in S and all M in B.

M4 If f1, f2,. . . are members of S and λ1 + λ2 + . . . = 1 where 0 ≤ λn ≤ 1, then
there exists f in S such that p(x, f, M ) = ∑∞n=1 λnp(x, fn, M ) for all x in
O and M in B.

9



M5 Call question an observable e in O such that p(e, f, {0, 1}) = 1 for all f in
S. Questions e and e′ are disjoint if e ≤ 1 − e′. Then a question ∑∞n=1 en
exists for any sequence (en) of questions such that em and en are disjoint

whenever n 6= m.

M6 If E is any compact, question-valued measure then there exists an observ-

able x in O such that χM (E) = E(M ) for all M in B, where χM is a
characteristic function of M.

M7 The partially ordered set of all questions in quantum mechanics is isomor-

phic to the partially ordered set of all closed subspaces of a separable,

infinite-dimensional Hilbert space.

M8 If e is any question different from 0 then there exists a state f in S such
that mf (e) = 1.

M9 For each sequence (fn) of members of S and each sequence (λn) of non-
negative real numbers whose sum is 1, one-parameter time evolution group

Vt : S 7→ S acts as follows: Vt (
∑∞

n=1 λnfn) =
∑∞

n=1 λnVt(fn) for all t ≥ 0;
and for all x in O, f in S, and M in B, t → p(x, Vt(f ), M ) is continuous.

In Mackey’s nine axioms all essential features of the quantum formalism

are directly postulated in their mathematical form: the Hilbert space structure

in M5-M8, the state space and probabilistic interpretation in M1-M4, and the

time evolution in M9. The list of axioms is long and their meaning is far

from transparent; indeed, at no point is one given an intuition as to where

these mathematical definitions come from or how one justifies them on physical

rather than formal grounds. In fact, Mackey’s concern in the early 1950s was

with a precise mathematical axiomatization of quantum mechanics rather than

with the question of what quantum mechanics tells us about the world or with

reconstructing its formalism from the set of such fundamental ideas. Thus,

the first stage of the reconstruction schema, at which one formulates physical

principles, is absent from Mackey’s work, and instead one starts directly at the

second stage where the first principles appear in mathematical form.

10



Mackey’s axioms M5-M8 were consequently reformulated in the language of

quantum logic, thereby rephrasing the assumptions that underlie the Hilbert

space structure. This has been the case, most prominently, in [29, 39, 40] and

also in an important state-of-the-art book [4]. Quantum logical assumptions are

simple enough to be accessible for direct comprehension, in contrast to Mackey’s

mathematically formulated axioms, but they tend to be linguistic rather than

physical. This means that one typically argues that it makes no sense to speak

about certain concepts unless some suitable ‘trivial’ properties of these con-

cepts had been postulated, e.g., the notion of proposition is only meaningful

if, as in Ref. [12], negation or partial order, or, as in Ref. [4], implication, are

defined. Although we fully acknowledge that linguistic a priori arguments can

be interesting and powerful, we however separate them from the reconstruction

program as introduced above: in the latter, first principles from which the the-

ory is derived should have a physical meaning, i.e. tell us something directly

and intuitively apprehensible about the world. Such principles, ideally, should

be independent of a particular formalism in which one then derives quantum

theory, and therefore should not rely on the language of quantum logic as just

one among many such formalisms.

3.2 Hardy’s reconstruction

It is startling how Mackey’s and similar axiomatic sets for quantum mechanics

differ from systems of first principles proposed by several contemporary authors.

Although some of these latter ones remain very much in the spirit of earlier pro-

posals of sets of abstract mathematical postulates (e.g., [41]), even in such cases

the author typically feels the need to give a non-technical, physical motivation

for the choice of axioms. Still further on the way to foundational physical prin-

ciples rather than purely mathematical axioms, one finds an interesting example

of reconstruction coming from Hardy’s instrumentalist derivation of quantum

theory [23]. Unlike Mackey who starts with two large abstract sets and an

abstract real-valued function, Hardy’s “five reasonable axioms” set up a link

between two initially introduced natural numbers, K and N . K is the number

11



of degrees of freedom of the system and is defined as the minimum number of

probability measurements needed to determine the state. Dimension N is de-

fined as the maximum number of states that can be reliably distinguished from

one another in a single measurement. The axioms then are:

H1 Probabilities. In the limit as n becomes infinite, relative frequencies (mea-

sured by taking the proportion of times a particular outcome is observed)

tend to the same value for any case where a given measurement is per-

formed on an ensemble of n systems prepared by some given preparation.

H2 Simplicity. K is determined by a function of N where N = 1, 2, . . . and

where, for each given N , K takes the minimum value consistent with the

axioms.

H3 Subspaces. A system whose state is constrained to belong to an M dimen-

sional subspace behaves like a system of dimension M .

H4 Composite systems. A composite system consisting of subsystems A and B

satisfies N = NANB and K = KAKB.

H5 Continuity. There exists a continuous reversible transformation on a system

between any two pure states of that system.

Hardy’s list of axioms is considerably shorter and simpler than Mackey’s,

and although four of H1-H5 still use mathematical language in their formulation,

the meaning of the axioms in Hardy’s instrumentalist setting can be grasped

easier than the meaning of Mackey’s M1-M9. In fact, this meaning is already

suggested by the names given to the axioms by Hardy. One can rephrase H1-

H4 into physical principles from which one derives the formalism of the theory.

This would provide for the missing first stage of the reconstruction schema and

thus amount to a complete example of a reconstruction. Thus, at the first stage

of reconstruction, the following physical principles are postulated that rephrase

Hardy’s axioms; they are simple but non-trivial: for H1, assume that probability

can be introduced as relative frequency and it is a well-defined concept obeying

12



the laws of probability theory; for H2, assume that the number of parameters

needed to characterize a state is linked in a minimal way to the number of

states that can be distinguished in one measurement, i.e., information carrying

capacity of the system; for H3, that systems that have the same information

carrying capacity have the same properties; for H4, assume multiplicability of

the information carrying capacity. At the second stage of the reconstruction

schema, one formulates these principles mathematically (in Hardy’s way); and,

at the third stage, one uses Hardy’s theorems to derive the full-blown formalism

of quantum mechanics.

The particular instrumental philosophy does not play a crucial role in the

derivation: Hardy himself acknowledges that his axioms can be adopted by a

realist as well as a hidden variable theorist or a partisan of collapse interpreta-

tions. Thus, choice of the underlying philosophy is not critical to the success of

the derivation, and Hardy’s reconstruction advances our understanding of quan-

tum theory irrespectively of the justification which one may have for the axioms.

What matters are the simple physical principles formulated as axioms H1-H4.

This is exactly what one would expect given the status of the first principles.

We shall however see below an opposite example, in which a justification used

for the fundamental principles will limit the area where mathematical derivation

is both applicable and meaningful.

Still, in Hardy’s case it is not so clear whether axiom H5 has a physical

meaning. Because it is this axiom that makes the theory quantum rather than

classical, the reconstruction program cannot be said to be completely imple-

mented. To illustrate this point, we distinguish two types of continuity assump-

tions that are made in axiomatic derivations of quantum theory. Continuity

assumptions of type 1 select the correct type of numeric field which is used in

the construction of the Hilbert space of the theory; namely, of the field C of

complex numbers. Solèr’s theorem [51] or Zieler’s axioms [58] are examples of

type 1 continuity assumptions. Hardy’s case is different and is an example of

a continuity assumption of type 2, which is ultimately responsible for the ap-

pearance of the superposition principle. Examples of other type 2 assumptions

13



include Gleason’s non-contextuality [16], Brukner’s and Zeilinger’s homogene-

ity of parameter space [7], Landsman’s two-sphere property [32], or Holland’s

axioms C and D [28] which bear a particular resemblance to Hardy’s H5:

(C) Superposition principle for pure states:

1. Given two different pure states (atoms) a and b, there is at

least one other pure state c, c 6= a and c 6= b that is a
superposition of a and b.

2. If the pure state c is a superposition of the distinct pure

states a and b, then a is a superposition of b and c.

(D) Ample unitary group: Given any two orthogonal pure states

a, b ∈ L, there is a unitary operator U such that U (a) = b.

We see that various axiomatic systems for quantum theory contain, under one

form or another, the assumption of continuity, and it is this assumption which

is largely responsible for making things quantum. Whatever the framework of

the reconstruction program, bringing in topological considerations is essential.

As it is exceedingly difficult to formulate a physical principle which may provide

a meaning for the continuity assumptions of type 2, all reconstruction programs

that employ them, suffer from intrusion of an element of mathematical abstrac-

tion.

In a more recent reformulation of his axioms [24, 25] Hardy suggests a way

to avoid the mathematically abstract type 2 continuity assumption that he has

previously made in axiom H5. In the new version Hardy maintains H5 without

the word “continuous”:

H5′ Reversibility. There exists a reversible transformation on a system between

any two pure states of that system.

It is then hypothesized that this new reversibility axiom, which is strictly weaker

than the former H5, if combined together with axiom H3 about the structure

of subspaces, will allow one to derive continuity in any finite-dimensional space.

14



The idea is to build on the fact that a limited number of finite groups is avail-

able in low-dimensional spaces. If the precision of a reversible transformation

between pure states in H5′ is sufficiently small, then this transformation, by

virtue of being an element of group of reversible transformations, will necessar-

ily be continuous, because no finite group will be available for this transforma-

tion to belong to, with a number of elements required for the carefully adjusted

precision. If further research shows that this conjecture can be carried through

and turned into a theorem, then for the first time one will have a set of algebraic

assumptions H1-H5′ that can be used to reconstruct either classical or quantum

theory, and the latter will be selected by fixing a particular value of the precision

parameter in reversible transformations between pure states. Solèr has proved

that the topology of quantum theory can be obtained from a set of assumptions

formulated algebraically, in the case of infinite-dimensional space [51]. Hardy’s

conjecture, if proved, will show how to achieve this deduction of topology from

algebra in the case of finite-dimensional spaces. Astonishingly, in this recon-

struction program the choice between classical and quantum theory will only

depend on the value of one numeric parameter.

3.3 Rovelli’s reconstruction

The critique expressed in Section 3.2 with respect to Hardy’s continuity axiom

applies to the example of reconstruction initially proposed by Rovelli [47], that

I have developed elsewhere [17, 18]. Here, the reconstruction starts from two

information-theoretic axioms:

R1 There exists a maximum amount of relevant information that can be ex-

tracted from a system.

R2 It is always possible to obtain new information about the system.

At the first sight it seems that R1 and R2 contradict each other. Indeed,

R1 says that the quantity of information is finite, while from R2 it follows that

this quantity must be infinite, because it is always possible to obtain some new

information. The reason why there is no contradiction lies in the use of the

15



term ‘relevant’ in R1, which does not appear in R2. In Rovelli’s reconstruction

relevance of new information must be judged with respect to information that

is already possessed by the observer. Bringing about new information can, not

only increase the amount of information currently available to the observer,

but also reduce it, due to the fact that some previously relevant information

may become irrelevant. Therefore, what is relevant depends on the particular

sequence of questions asked by the observer, and for a different observer a change

in the amount of information brought in by the same new piece of information

may vary. In accordance with his general relational approach, information in

Rovelli’s reconstruction should be taken as an observer-dependent, rather than

objective, notion. It is information in Shannon’s sense, that is indexed by

two indices: first related to the observed system, about which this information

has been obtained, second related to the observing system, that has obtained

information about the first system. It is impossible, in Rovelli’s view, to separate

the notion of information from its second index and to speak about objective

information independently of the observing system.

From R1 and R2, with the help of a few quantum logical assumptions, one

derives the formalism of quantum mechanics. In particular, if one postulates

that information is obtained through answers to yes-no questions that can be

asked about the system, then supplementary assumptions are that the set of such

yes-no questions forms a complete atomic orthocomplemented lattice. From

axiom R1 and a formal definition of relevance of yes-no questions with respect to

each other one derives that this lattice is orthomodular. If a further assumption

is made about the lattice of questions being isomorphic to the lattice of all

closed subspaces of a Banach space constructed over a numeric field (i.e. real

or complex numbers or quaternions), one then obtains that this Banach space

is indeed a Hilbert space. Its quantum rather than classical character follows

from axiom R2.

While the supplementary assumptions cast a shadow on the conceptual clar-

ity of the reconstruction much in the same fashion as does axiom H5 for Hardy’s

approach, the whole program presents itself differently from Hardy’s instrumen-

16



talism. Mathematical derivation being still devoid of ontological commitments,

the proposed justification of the first principles does not refer to an ontology.

Rather, by reconstructing quantum theory from information-theoretic princi-

ples, we point at its epistemological character and at its role as a theory of (a

certain kind of) knowledge, one with certain limits on the kind of information

one may be dealing with. The most general theory of this kind of information

takes the form of quantum theory. Here again reconstruction appears more ap-

pealing than a mere interpretation as it leaves room for any justification of first

principles, some such justifications being possibly different from ours. Indeed,

one may equally well choose to adopt a specific ontological picture to justify

R1-R2. At the same time, regardless of a concrete philosophical justification

for first principles, the meaning of quantum theory stands clear: it is a general

theory of information constrained by several information-theoretic principles.

3.4 The CBH reconstruction

Clifton, Bub and Halvorson (CBH) propose a set of quantum informational

constraints from which one derives the basic elements of quantum theory [10].

They postulate three fundamental principles:

CBH1 No superluminal information transfer via measurement.

CBH2 No broadcasting.

CBH3 No bit commitment.

CBH argue that the meaning of axiom CBH1 is that when Alice and Bob

perform local measurements, Alice’s measurements can have no influence on

the statistics for the outcomes of Bob’s measurements, and vice versa. They

also submit that “otherwise this would mean instantaneous information transfer

between Alice and Bob” and “the mere performance of a local measurement

(in the nonselective sense) cannot, in and of itself, transfer information to a

physically distinct system.” Upon reading these statements, we feel strongly

that CBH take distinct and distant to be synonyms. Such a terminological

17



identification might indeed be a tacit assumption among quantum information

theorists, whose analysis is limited to finite-dimensional Hilbert spaces; but in

the full-blown C∗-algebraic framework, which CBH also employ, the meaning of

the two terms is quite different. We have here an example of the way in which

the initial quantum informational departure point of the CBH1-CBH3 principles

constrains the use of the C∗-algebraic formalism to only the situations where

these principles make sense from the point of view of quantum information; in

fact, the formalism is at the same time routinely applied to other settings as well.

Unlike Hardy’s derivation, which is independent of the particular instrumental

justification of its first principles, the CBH reconstruction cannot be carried

through outside the field of quantum information, because its mathematics,

while still valid outside this field, requires a new justificatory language. Besides

the problem of synonymy of ‘distant’ and ‘distinct,’ the quantum informational

departure point also restricts the question of time evolution. The latter is tacitly

taken by the CBH to be the usual quantum mechanical time evolution, while in

the general C∗-algebraic framework this is typically not the case and a variety

of different ‘temporal’ evolutions are available [11]. This and other problems

arising from the generality of the C∗-algebraic framework are avoided by the

CBH reconstruction at the price of confining itself to the quantum informational

setting.

Axiom CBH2 is used to establish that the C∗-algebras of Alice and Bob, A
and B, taken separately, are non-Abelian. It is interesting to note that non-
Abelianness of A and B is proved by assuming that they are kinematically
independent. This means that quantumness, of which non-Abelianness is a

necessary ingredient, is not a property of any given system taken separately, as

if it were the only physical system in the Universe; on the contrary, to be able

to derive the quantum character of the theory, one must consider the system in

the context of at least one other system that is physically distinct from the first

one. As a consequence, for example, this forbids treating the whole Universe as

a quantum system if one reconstructs quantum theory along the CBH lines.

18



Axiom CBH3 entails nonlocality: spacelike separated systems must at least

sometimes occupy entangled states. It is not proved, however, that actually

instantiated states fill the space of all entangled states. CBH show that if Alice

and Bob have spacelike separated quantum systems, but cannot prepare any

entangled state, then Alice and Bob can devise an unconditionally secure bit

commitment protocol. From this theorem the authors deduce that the impos-

sibility of unconditionally secure bit commitment entails that “if each of the

pair of separated physical systems A and B has a non-uniquely decomposable
mixed state, so that A ∨ B has a pair {ρ0, ρ1} of distinct classically correlated
states whose marginals relative to A and B are identical, then A and B must be
able to occupy an entangled state that can be transformed to ρ0 or ρ1 at will

by a local operation.” The term ‘separated’ is essential and, nevertheless, its

precise meaning is not given. Once again, this can be compared to the confusion

between distinct and distant. When CBH claim that Alice and Bob represent

“spacelike separated systems,” while formally Alice and Bob are just two C∗-

algebras, one sees how the way in which CBH apply the algebraic formalism is

constrained by the context of quantum information theory. We witness here an

interesting situation in which the language and the context used to formulate

and to justify the fundamental principles set a limit on the applicability of the

mathematical formalism in which these principles are represented. Even if the

formalism can be understood more generally than within the initially chosen

disciplinary setting, one still cannot make his way out of this linguistic and

contextual prison; without the sense of the axioms being lost. If one persists

and escapes, and then obtains a new mathematical result, this result will be

void of physical meaning until a new, broader justification of the fundamental

principles has been given. Philosophical and linguistic justification, and math-

ematical derivation, play here a game of mutual onslaught and retreat which,

ultimately, leads to the advance of science, in the way described in Section 1.

Notwithstanding the difficulties with justification, the CBH result would be

a perfect example of reconstruction were it not for a great deal of mathematical

structure which is implied by the choice of the C∗-algebraic framework. The

19



assumptions of the algebraic formalism include at the very least, the relations

between operators satisfying linearity, the number field being C, and the states

giving rise to the Hilbert space representation via the GNS construction. Once

one lists all such tacit supplementary assumptions, the CBH reconstruction

appears once again to suffer from the defect of incorporating a serious mathe-

matical abstraction, similar to the derivations from axioms H1-H5 or R1-R2.

3.5 Intentionally incomplete reconstructions

All reconstructions that we have discussed until now shared the goal of deriving,

at the final end, the full-blown structure of quantum theory. Recently, a new

type of information-theoretic reconstructions appeared, intentionally not aimed

at deriving the whole quantum theoretic structure [1, 2, 3, 22, 46, 50, 52]. Chris-

tened pejoratively by their own authors, these “toy models”, “fantasy quantum

mechanics,” or “quantum mechanics lite” employ a methodology of reconstruc-

tion that has been overlooked by the previous generations of researchers: it is

now claimed as helpful to reconstruct, not the full version but only a certain part

of quantum theory. One builds, therefore, a new theory which is, from the very

beginning, not intended to be the quantum theory; but this new theory allows

nevertheless to better understand the structure of the ‘true’ quantum theory.

To the same old ultimate aim of better understanding quantum mechanics toy

models provide a variety of new promising insights.

The idea of modifying usual quantum theory is not new. Nonlinear ex-

tensions of the Schrödinger equation have been explored by various authors

[5, 15, 56], and comprehensive reviews of these attempts can be found in [38, 53].

Nonlinear models are also sometimes analyzed in the context of quantum the-

ories more general than quantum mechanics, e.g., in quantum gravity. Inten-

tionally incomplete reconstructions, however, differ substantively from these

attempts to modify quantum theory. While the latter take standard linear

quantum mechanics to be incomplete and purport to replace it by a nonlin-

ear, more complete theory, toy models see themselves as incomplete. They do

not question the validity of quantum mechanics and do not compete with it

20



in explaining empirical phenomena. Methodologically, as this is routinely em-

phasized in the opening paragraphs of articles introducing toy models, e.g., in

[3], toy models focus on important physics principles, which are upheld while

other possibilities are modified. This way the consequences of such and such

principle are investigated, independently of other principles. Thus, intentionally

incomplete reconstructions are not aimless, but allow one to achieve a better un-

derstanding of the structure of quantum theory. Incompleteness, then, becomes

a feature rather than a flaw of toy models.

Most of the existing examples of toy models are based on information-

theoretic principles. To compare the toy model with standard quantum theory,

one then asks whether the former reproduces quantum computational phenom-

ena available in the latter. Among others, such questions may include:

• Does the toy model allow superluminal signalling?

• Does the toy model allow bit commitment?

• Does the toy model allow teleportation, dense coding, or remote steering?

• Does the toy model allow exponential speed-up relative to classical com-
putation or solving NP-complete problems in polynomial time?

Other toy models are inspired by information-theoretic principles, but they are

compared with standard quantum theory in the aspects that do not necessarily

relate to computation:

• Does the toy model allow nonlocality and to what extent?

• Is the toy model contextual?

• Does the toy model possess a continuum of states, measurements and
transformations between states?

Answers given to all of these questions can be yes or no depending on the model.

Investigating then the difference between the first principles of a particular toy

model, and of standard quantum mechanics, one learns with precision which

21



fundamental principle is responsible for which element of the quantum theoretic

structure. Spekkens’s toy model, for example, accommodates such quantum

phenomena as noncommutativity, interference, the multiplicity of convex de-

compositions of a mixed state, no cloning, teleportation, and others [52]. We

learn that the continuous state space, the existence of a Bell theorem, or con-

textuality, all of which are absent from this toy model, go unconnected with

the appearance of the phenomena that are reproduced. Analogously, the toy

model known as non-local, or PR, boxes [3, 46] allows non-local correlations

that are strictly stronger than those allowed by quantum mechanics, while it

only slightly modifies the quantum mechanical state space. One then sees that

non-locality is not an exclusively quantum feature and, further, that the amount

of non-locality in quantum theory is smaller than in some other theories. We

can then conjecture, for instance, that the ‘true’ quantum theory has as much

non-locality as it does, and not more or less, due to the continuum of states and

to reversible transformations between pure states, both of these being left out

of the PR boxes toy model.

These examples show how one obtains the deeper insight into the structure

of quantum theory and therefore achieves what was initially promised by the

program of intentionally incomplete reconstruction. Toy models, although in-

complete, form a very fertile class of reconstructions. Their recent advent in

the area of the foundations of quantum theory manifests the fact that the shift

from interpretation of quantum theory to its reconstruction gave birth, in this

area, to many a new, previously non-existent idea.

4 Conclusion

We have argued that reconstruction is the exclusive way to make things clear

about quantum mechanics. As such, this idea is not novel but has been in the

air for some time, and a concise statement can for example be found in Rovelli

[47],

Quantum mechanics will cease to look puzzling only when we will

22



be able to derive the formalism of the theory from a set of sim-

ple physical assertions (“postulates,” “principles”) about the world.

Therefore, we should not try to append a reasonable interpretation

to the quantum mechanical formalism, but rather to derive the for-

malism from a set of experimentally motivated postulates.

What is novel, however, is that an increasing number of researchers work nowa-

days on reconstructing quantum theory, and the time is ripe to promote this

general framework to the status of a widely accepted paradigmatic shift in the

body of work in the foundations of physics.

Reconstruction brings in clarity to where interpretation was struggling to

make sense of a physical theory. What belongs to physical theory is no more

than what is needed for its derivation. All other questions belong to metathe-

ory and are related to the metatheoretic justification task for the choice of first

principles. However, completely reconstructing quantum theory remains only a

partially solved problem. Notwithstanding, reconstruction has been successfully

competing with more traditional interpretations, due to its appealing concep-

tual transparency and to the clarity that it brings into the structure of the

theory. It would be too ambitious to expect that all of modern quantum theory,

including field theory and quantum gravity, could be derived from a few axioms.

In other words, none of the reconstructions that have been proposed until now

is complete, and some are intentionally incomplete. Although to a varying de-

gree, mathematical abstraction is a necessity for each of the currently existing

reconstructions. However, if we want to understand the meaning of even most

advanced parts of quantum theory, and to reach a consensus in this understand-

ing, it is then inevitable that simple physical principles be formulated and put

in the very foundation of quantum theory.

Acknowledgements

Parts of this paper were written during visit to the Perimeter Institute for

Theoretical Physics. Thanks to Lucien Hardy, Jeff Bub and Bill Demopoulos

23



for their comments.

References

[1] S. Aaronson. Is quantum mechanics an island in theoryspace? In A. Khren-
nikov, editor, Proceedings of the Växjö Conference “Quantum Theory:
Reconsideration of Foundations”. Växjö University Press, 2004, quant-
ph/0402095.

[2] S. Aaronson. Quantum computing, postselection, and probabilistic
polynomial-time. Proc. Roy. Soc. Lond., A461:3473–3482, 2005, quant-
ph/0412187.

[3] J. Barrett. Information processing in non-signalling theories. 2005, quant-
ph/0508211.

[4] E.G. Beltrametti and G. Cassinelli. The logic of quantum mechanics.
Addison-Wesley, Reading, 1981.

[5] I. Bialynicki-Birula and J. Mycielski. Nonlinear wave mechanics. Annals
of Physics, 100:62–93, 1976.

[6] G. Birkhoff and J. von Neumann. The logic of quantum mechanics. Ann.
Math. Phys., 37:823–843, 1936. Reprinted in: J. von Neumann Collected
Works Pergamon Press, Oxford, 1961, Vol. IV, pp. 105–125.

[7] C. Brukner and A. Zeilinger. Information and fundamental elements of
the structure of quantum theory. In L. Castell and O. Ischebeck, edi-
tors, Time, Quantum, Information, pages 323–356. Springer-Verlag, 2003,
quant-ph/0212084.

[8] R. Brunetti and K. Fredenhagen. Algebraic approach to quantum field
theory. 2004, math-ph/0411072.

[9] R. Carnap. Der Logische Aufbau der Welt. Berlin, 1928.

[10] R. Clifton, J. Bub, and H. Halvorson. Characterizing quantum theory
in terms of information-theoretic constraints. Found. Phys., 33(11):1561–
1591, 2003.

[11] A. Connes and C. Rovelli. Von Neumann algebra automorphisms and time-
thermodynamics relation in general covariant quantum theories. Class.
Quant. Grav., 11:2899–2918, 1994.

[12] W. Demopoulos. Elementary propositions and essentially incomplete
knowledge: A framework for the interpretation of quantum mechanics.
Noûs, 38(1):86–109, 2004.

[13] A. Einstein. Geometry and experience. Address to the Prussian Academy
of Sciences, 27 January 1921.

24



[14] G.G. Emch. Algebraic methods in statistical mechanics and quantum field
theory. John Wiley, New York, 1972.

[15] N. Gisin. Weinberg’s non-linear quantum mechanics and superluminal com-
munication. Phys. Lett., A143:1–2, 1990.

[16] A. Gleason. Measures on the closed subspaces of a Hilbert space. Journal
of Mathematics and Mechanics, 6:885–894, 1967.

[17] A. Grinbaum. Elements of information-theoretic derivation of the formal-
ism of quantum theory. International Journal of Quantum Information,
1(3):289–300, 2003.

[18] A. Grinbaum. Information-theoretic principle entails orthomodularity of a
lattice. Foundations of Physics Letters, 18(6):573–592, 2005.

[19] J. Guenin. Axiomatic formulations of quantum theories. J. Math. Phys.,
7:271–282, 1966.

[20] J. Gunson. On the algebraic structure of quantum mechanics. Comm.
Math. Phys., 6:262–285, 1967.

[21] R. Haag and D. Kastler. An algebraic approach to quantum field theory.
J. Math. Phys., 5:848–861, 1964.

[22] L. Hardy. Disentangling nonlocality and teleportation. 1999, quant-
ph/9906123.

[23] L. Hardy. Quantum theory from five reasonable axioms. 2001, quant-
ph/00101012.

[24] L. Hardy. Probability theories with dynamic causal structure: A new frame-
work for quantum gravity. 2005, gr-qc/0509120.

[25] L. Hardy. Private communication. April 2006.

[26] A. Heyting. Axiomatic projective geometry. North-Holland, Amsterdam,
1963.

[27] D. Hilbert, J. von Neumann, and L. Nordheim. Über die Grundlagen
der Quantenmechanik. Math. Ann., 98:1–30, 1927. (Reprinted in J. von
Neumann Collected Works Pergamon Press, Oxford, 1961, Vol. I, pp. 104–
133).

[28] S.S. Holland Jr. Orthomodularity in infinite dimensions; a theorem of M.
Solèr. Bull. Amer. Math. Soc., 32(2):205–234, 1995.

[29] J.M. Jauch. Foundations of Quantum Mechanics. Addison-Wesley, 1968.

[30] P. Jordan, J. von Neumann, and E. Wigner. On an algebraic generalization
of the quantum mechanical formalism. Ann. Math., 35:29–34, 1934.

25



[31] S. Kochen and E.P. Specker. Logical structures arising in quantum the-
ory. In Addison J. et al., editors, The Theory of Models. North-Holland,
Amsterdam, 1965.

[32] N.P. Landsman. Mathematical Topics Between Classical and Quantum Me-
chanics. Spinger, New York, 1998.

[33] G. Ludwig. An Axiomatic Basis for Quantum Mechanics. Springer, 1985.

[34] G.W. Mackey. Quantum mechanics and Hilbert space. Amer. Math.
Monthly, 64:45–57, 1957.

[35] G.W. Mackey. Mathematical Foundations of Quantum Mechanics. Ben-
jamin, New York, 1963.

[36] A.R. Marlow. Orthomodular structures and physical theory. In A.R. Mar-
low, editor, Mathematical Foundations of Quantum Theory, pages 59–70.
Academic Press, 1978.

[37] N.D. Mermin. Could Feynman have said this? Physics Today, page 10,
May 2004.

[38] P. Nattermann. On (non)linear quantum mechanics. Symmetry in Nonlin-
ear Mathematical Physics, 2, 1997.

[39] C. Piron. Axiomatique quantique. Helvetia Physica Acta, 36:439–468, 1964.

[40] C. Piron. Survey of general quantum physics. Found. Phys., 2:287–314,
1972.

[41] I. Pitowsky. Quantum mechanics as a theory of probability. 2005, quant-
ph/0510095.

[42] R.J. Plymen. C∗-algebras and Mackey’s axioms. Comm. Math. Phys.,
8:132–146, 1968.

[43] R.J. Plymen. A modification of Piron’s axioms. Helvetia Physica Acta,
41:69–74, 1968.

[44] J.C.T. Pool. Baer ∗-semigroups and the logic of quantum mechanics.
Comm. Math. Phys., 9:118–141, 1968.

[45] J.C.T. Pool. Semimodularity and the logic of quantum mechanics. Comm.
Math. Phys., 9:212–228, 1968.

[46] S. Popescu and D. Rohrlich. Nonlocality as an axiom for quantum theory.
Foundations of Physics, 24:379, 1994, quant-ph/9508009.

[47] C. Rovelli. Relational quantum mechanics. Int. J. of Theor. Phys., 35:1637,
1996.

26



[48] I. Segal. Postulates of general quantum mechanics. Ann. Math., 48:930–
948, 1947.

[49] I. Segal. Mathematical Problems of Relativistic Physics. American Mathe-
matical Society, Providence, 1963.

[50] J. Smolin. Can quantum cryptography imply quantum mechanics? Quan-
tum Information and Computation, 5:161–169, 2005, quant-ph/0310067.

[51] M.P. Solèr. Characterization of Hilbert spaces with orthomodular spaces.
Comm. Algebra, 23:219–243, 1995.

[52] R. Spekkens. In defense of the epistemic view of quantum states: a toy
theory. 2004, quant-ph/0401052.

[53] G. Svetlichny. Nonlinear quantum mechanics. quant-ph/0410036.

[54] V.S. Varadarajan. Probability in physics and a theorem on simultaneous
observability. Comm. Pure and Appl. Math., 15:189–217, 1962.

[55] V.S. Varadarajan. Geometry of quantum theory. Van Norstand, Princeton,
1968.

[56] S. Weinberg. Precision tests of quantum mechanics. Phys. Rev. Lett.,
62:485–488, 1989.

[57] S. Weinberg. Sokal’s hoax. The New York Review of Books, XLIII(13):11–
15, 1996.

[58] N. Zieler. Axioms for non-relativistic quantum mechanics. Pacific J. Math.,
11:1151–1169, 1961.

27