Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Sonja Smets
Preface v
Introduction 1
2 Underlying Epistemology 13
2.1 Why Realism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1.1 Scientific Realism . . . . . . . . . . . . . . . . . . . . . . . 13
2.1.2 Critical Realism . . . . . . . . . . . . . . . . . . . . . . . 16
2.1.3 Structural Realism . . . . . . . . . . . . . . . . . . . . . . 21
2.2 EPR and The Criterion of Reality . . . . . . . . . . . . . . . . . 22
2.2.1 The Criterion of Reality . . . . . . . . . . . . . . . . . . . 22
2.2.2 EPR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
i
4 Mathematical Framework 42
4.1 The Symbolism . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.2 Questions and their structure . . . . . . . . . . . . . . . . . . . . 42
4.3 Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.3.1 Modeling Properties . . . . . . . . . . . . . . . . . . . . . 43
4.3.2 Properties and their structure . . . . . . . . . . . . . . . . 44
4.4 States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.4.1 Modeling States . . . . . . . . . . . . . . . . . . . . . . . 46
4.4.2 States and their structure . . . . . . . . . . . . . . . . . . 47
4.5 Fundamental Axioms . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.6 Orthocomplements and Questions . . . . . . . . . . . . . . . . . . 52
4.7 Duality of States and Properties . . . . . . . . . . . . . . . . . . 54
4.8 Classical and Non-classical Cases . . . . . . . . . . . . . . . . . . 56
4.9 Compatibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
8 Natural Deduction 93
8.1 Assumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
8.2 Substitution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
8.3 Meet and Join . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
8.4 Weakening and Regularity . . . . . . . . . . . . . . . . . . . . . . 100
8.5 Negation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
8.5.1 Excluded Middle . . . . . . . . . . . . . . . . . . . . . . . 102
8.5.2 Non Contradiction and Ex Falso . . . . . . . . . . . . . . 104
ii
8.5.3 Reductio Ad Absurdum . . . . . . . . . . . . . . . . . . . 105
8.6 Weak Modularity . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
10 Modalities 113
10.1 Modal Language . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
10.2 History of Modality . . . . . . . . . . . . . . . . . . . . . . . . . 116
iii
Afterword 177
References 178
iv
Preface
First of all, I wish to thank Jean Paul Van Bendegem for taking up the task
of supervising this work, knowing his busy schedule at the university I am
very grateful for the time he was able to make free for discussing the subject
matter of this thesis and for commenting on early drafts. Secondly, I thank the
co-supervisor of this work, Dirk Aerts, for the discussions held at the time I
occupied an office at the Center Leo Apostel.
Of course, if it wasn’t for Bob Coecke I probably would never have become
interested in operational quantum logic in the first place. He pointed out the
interesting problems to me which actually led to the writing of this thesis.
Besides for the fact that he encouraged me while I wrote this thesis, I also
wish to thank him for answering the many (!) questions concerning physics and
mathematics which I posed. Also David Moore deserves to be specially thanked
for the many discussions which took place at various conferences and workshops;
and for the pleasure of working together on some joint papers with him and Bob.
Next I want to thank Frank Valckenborgh, Isar Stubbe, Steven Sourbron and
Karin Verelst for discussing topic-related matters and/or the helpful comments
which improved the writing of this thesis. I wish to take this opportunity to
thank also C. Piron for the kind hospitality and discussions held during my visit
in Geneva.
I am very grateful to Rebecca Lee for taking up the role to help me improve
my written English. Finally, I thank the members of the Center for Logic
and Philosophy of Science at the Free University of Brussels, my friends, and of
course my family, in particular my mother, father, sister and her future husband
for their encouragement.
v
Introduction
“If philosophy is interpreted as the quest for the most general and
comprehensive knowledge possible, it obviously becomes the mother
of all scientific inquiry. But it is just as true that the various branches
of science have in their turn exercised a strong influence on the sci-
entists concerned and, beyond that, have affected the philosophical
thinking of each generation.” Albert Einstein [Ein50, p.755]
1
Introduction 2
The present thesis is constructed in four parts. After introducing our object of
investigation against a philosophical background in part I, we give in part II
an overview of operational quantum logic as we mainly find it in the Geneva
School literature [Aer81, Jau68, JauPir63, JauPir69, Moo99, Pir64, Pir72, Pir76,
Pir77, Pir81, Pir85, Pir98, Pir91]. We deal with specific logical aspects in part
III and finally turn in part IV to the recently developed dynamic extension as
put forward in [AmiCoeStu98, Coe00, Coe(nd)b, Coe(nd)c, CoeMooSme(nd)c,
CoeMooStu00, CoeMooStu01, CoeSme00, CoeStu99a, CoeStu99b, Sou00]. In
that last part we elaborate on the construction of a dynamic operational quan-
tum logic.
In chapter 4, rather a lot of technicalities are involved but we believe that they
are important to get a grip on the subject. It is impossible to say anything about
operational quantum logic without knowing the main axioms and theorems,
especially because the Geneva School approach is, from the abstract point of
view, an axiomatic one. This is to say:
“it (the Geneva School approach) is grounded on a system of prim-
itive notions, the choice of a set of specific axioms, and some inter-
pretation rules which translate the primitive notions into concepts
about a physical domain.” [CatNis91, p.1294]
It is further emphasized in [CatNis91], at the basis of such an axiomatic formal-
ized theory, we find a formal language in which all symbols, derived notions and
axioms are defined without reference to any concrete (mathematical) structure
as for instance Hilbert Space. Furthermore, C. Piron proved in [Pir64] that a
certain kind of property lattice satisfying certain axioms can be represented in
the form of the lattice of closed linear subspaces of a generalized Hilbert space.
This theorem proves the generality of the formalism with respect to the Hilbert
Space formalism and justifies operational quantum logic as an important phys-
ical theory. Limiting our exposition to the description of one physical system1
we will present operational quantum logic in a rather new way which differs
1 For an exposition on many physical (sub)systems concentrated on the problems with
separated systems in the non-classical case, we refer to [Aer81]. For a more recent, detailed,
exposition on subsystem recognition we refer to [Val01].
Introduction 4
slightly from the presentation given by C. Piron in [Pir76] where he still con-
centrated on the notion of proposition and propositional system instead of on
the notion of property and property lattice. Where confusion is possible, we will
insert concepts between brackets within citations of Piron’s earlier work. The
same will be done when we talk about questions and the new notion of definite
experimental projects. Concretely, we will model questions and concentrate on
their structure which is obtained by the introduction of a preorder relation with
respect to truth or certainty. Then we will model properties as corresponding
to equivalence classes of questions and concentrate on their structure given by
the partial order relation we obtain with respect to actuality. The collection
of all properties of a physical system will be the “working set” which, as will
be pointed out, forms a complete lattice. The meet of properties corresponds
to the “product of questions” while the join has no “direct” physical meaning.
After explaining how the join differs from a classical disjunction we turn to the
description of a physical system by means of modeling its states. The struc-
ture on states will be supplied by means of an orthogonality relation. Before
we characterize both descriptions of a physical system, by means of states and
by means of properties as dual, we first deal with the three basic axioms of
the Geneva School approach applying to every physical system. Those axioms
will be elaborated upon in a context where we trace some basic physical laws
back to Aristotle. Hereby we concentrate firstly on Aristotle’s conception of
“passing-away” and “coming-to-be” or generation and corruption and secondly
on contraries or opposites. The latter is placed in the discussion on orthocom-
plements of properties. Regarding orthocomplements of properties and their
link to inverse questions we make some specific remarks since that link is not as
straightforward as one would perhaps assume. Later on we deal with classical
and non-classical systems, classical and quantum questions and properties and
try to place all these notions in relation to each other. In this perspective it be-
comes clear which axioms have to be imposed to speak about classical systems
as opposed to pure quantum systems. The road is now open for the notion of
compatibility to be discussed. Here we make clear that the notion of compati-
bility has a mathematical meaning and not the oft-attributed physical meaning
which states that two properties are compatible if their associated questions
can be performed simultaneously without disturbing each other. The section on
compatible properties will prove useful for the discussion of ideal measurements
in the next chapter.
In part III, chapter 7, we enter into the old debate on whether quantum logic
can be regarded as a proper logic, and focus in particular on the question which
concerns us here the most namely, is operational quantum logic a logic? After
making clear what we consider to be a logic, we point out that operational quan-
tum logic gives rise to a logical system of which we provide a natural deduction
presentation in chapter 8. Here we focus especially on its infinitary character
which is linked to the fact that the Geneva School approach works with lattices
which are complete. In chapter 9, we focus on the straightforward algebraic se-
mantical realization for the logical system of operational quantum logic and show
what a Kripke realization looks like. A logical duality between both semantical
realizations can be established based on the underlying duality between states
and properties prevalent in the Geneva School approach. Linked to the idea
of working with possible worlds in a Kripke realization, the notion of modality
naturally emerges. After focusing on the well-known modal B-interpretation of
quantum logic in chapter 10, we approach the possibility of a modal extension
for the logical system of operational quantum logic. The chapter on modalities
will be concluded with a short exposition on the historical suspicion surrounding
modalities. Of course, in a chapter devoted to logical aspects, we cannot avoid
“the problem of an implication connective” in quantum logic . So in chapter
11, we deal with the difference between an implication relation and implication
operation. Focusing on the criteria an implication operation should satisfy, we
handle several candidates which emerge in the context of (orthomodular) quan-
tum logic. Hereby we pay special attention to the so-called Sasaki hook which
by G. Hardegree has been given an interpretation in terms of a Stalnaker condi-
tional. We close chapter 11 with a note on a less known implication-candidate
for quantum logic, called the central implication.
Philosophy of Operational
Quantum Logic
8
Chapter 1
Standard Quantum Logic (SQL) originated with the work of G. Birkhoff and J.
von Neumann in [BirVNeu36]. It has to be mentioned though that the idea of
a logical calculus based on the relation between the properties of a physical sys-
tem and the projections1 defined on a Hilbert space is already present in J. von
Neumann’s earlier work [VNeu32]. In [BirVNeu36], G. Birkhoff and J. von Neu-
mann analyze why the logic underlying the formalism of quantum theory is not
a classical one. They explain that the equivalent experimental propositions or
physical qualities of any “classical mechanical” physical system are represented
by the subsets of its phase-space and that their structure is boolean, while the
mathematical representatives for propositions about quantum systems are not
just subsets of a phase-space. Moreover they conclude in the latter case that
the correct representatives of any experimental proposition are the closed linear
subspaces of a Hilbert space. Obviously such propositions could equivalently
be represented by the projections on a Hilbert space. The structure obtained
from these mathematical representatives is, since the failure of distributivity, not
boolean and called an orthomodular lattice. To be historically exact we have
to note that an orthomodular lattice is an orthocomplemented lattice satisfying
weak modularity2 and not the by G. Birkhoff and J. von Neumann proposed
stronger modular law3 . Anyhow, the failure of the distributivity law is due to
the fact that the logical operation of “quantum disjunction” cannot be repre-
sented in this case by the set-theoretical union. As explained in [DalGiu(nd)],
“for, generally, the union X ∪ Y of two closed subspaces is not a closed sub-
space.” Indeed, what we do have is “the smallest closed subspace including both
X and Y ”. Without going into any detail here about the difference between
classical and quantum structures we would like to stress with [DalGiu(nd)] that
this focus on the Hilbert space model forces some strong requirements on us.
That is to say that the structure of “all” closed subspaces in a Hilbert space,
1 Projections are idempotent and self-adjoint linear operators.
2 For a, b ∈ L, a ≤ b implies that b = a ∨ (b ∧ a! ) (see part II, chapter 4 for more details).
3 In accordance with part II, we can write this modular identity as follows: for a, b, c ∈ L,
if a ≤ c then a ∨ (b ∧ c) = (a ∨ b) ∧ c.
9
CHAPTER 1. OPERATIONAL QUANTUM LOGIC 10
We have to be very careful with the notion of operationality used here. For
philosophers of science, operationality roughly means a collection of views about
how observations and scientific concepts are connected by means of operational
procedures. Such views vary on the ontological status connected to the oper-
ations and the status of the concepts (theoretical terms) which are defined by
them, or on the fact whether different operational procedures can define one
single concept. Operationalism as such was alive in the early thirties among
positivists and behaviorists for whom behavior, experience or operations were
given a leading position in their theories. Since we will be defending the posi-
tion of scientific realism in relation to OQL, operationality here points to the
underlying assumption that with every property of a physical system we can
associate experimental (operational) procedures — which can be performed —
and a well-defined result.6 However contrary to other views on operationalism,
our view differs essentially from those who express that what exists is limited
to what is directly measurable. On the epistemological level our knowledge of
what exists is based on what we could measure or observe — stated counter-
factually — while on the ontological level physical properties exist independent
of our possible observations. This leads us to the fact that our view also differs
essentially from operationalists such as P.W. Bridgman because here physical
properties have an extension in reality and are not just by means of definitions
reducible to sets of procedures. Hence we do not subscribe to [Bri61]:
“The concept of length is therefore fixed when the operations by
which length is measured are fixed: that is, the concept of length
involves as much as and nothing more than the set of operations by
which length is determined. In general, we mean by any concept
nothing more than a set of operations; the concept is synonymous
with the corresponding set of operations.” (p.5)7
So we can now already say that the notion of operationalism within the context
of OQL has to be conceived as an attitude and not as a doctrine. As such it
has nothing to do with empiricism, positivism or behaviorism, but points to the
fact that the role of operations or experiments has changed considerably from
what their function has been in the old quantum theories. It has to do with
the fact that in OQL, one takes the point of view of an engineer and not of an
astronomer, as said by Piron in [Pir91]:
“C’est pourquoi la physique décrite par ce type de propriétés est
plus la physique de l’ingénieur qui agit, que celle de l’astronome
qui ne fait que contempler. Les expériences dont il est question
6 The notion of property will be analyzed in more detail within the framework of OQL in
part II.
7 About this statement, P. W. Bridgman mentions years later in [Bri50] that it should not
be taken out of its context. He stresses that he does not want to set up any “theory” of
meaning and does not maintain that the meaning of a concept involves nothing more than
operations. The statement quoted, then implies that it is sufficient for a physicist to say that
“what a man means by a term is to be found by observing what he does with it, and not by
what he says about it” [Bri50, p.5].
CHAPTER 1. OPERATIONAL QUANTUM LOGIC 12
Underlying Epistemology
attitudes, commitments, ideals or norms. Also A. Fine [Fin86], uses the word “stance” in this
sense, we cite: “For realism is not a matter of words (“external world”, “real state of affairs”,
etc.) but rather a matter of what beliefs and commitments we have in uttering those words.”
(p.107).
13
CHAPTER 2. UNDERLYING EPISTEMOLOGY 14
exist and act, for the most part, quite independently of scientists and their ac-
tivity” [Bha87, p.5]. Scientific realism is a form of what is called transcendental
or contemporary metaphysical realism, according to which there is a world, a
“Ding an Sich”, accessible to our knowledge, not through a special ability or
intuition but through science (see [Oge92]). It is a position within philosophy
of science and as such it concerns scientific theories, activity and explanation.
We believe that the ingredients of this realism involve at least: belief in the
approaching truth of scientific theories, the existence of a mind-independent
reality, belief in reality as discoverable and accurately predictable. In further
analysis and contrary to constructive empiricism2 , scientific realists accept a
theory if they believe that what it (literally) says about the world, is also the
case, or in other words, that it approaches truth.3 When such a belief is held,
it concerns not only the observable part of what a theory claims to be the case,
but also entails that non-observables have an existence in reality. Note that the
distinction between observables and non-observables has to be acknowledged by
the constructive empiricist as well as by the scientific realist, but that for the
latter this distinction is not as crucial because in his theories observables are
not treated any differently than non-observables [VFra99]. We like to think of
observables as described in [VFra80]:
Referring to “scientific realism”, the qualification above clearly does not imply
that our scientific theories are to be based on what we can actually observe
[Pop82]. The oft-drawn conclusion from this namely: “existence independent
from any measurement or observation” holds also in our view. It is however
essential for our position, which may not be shared by every scientific realist,
that our knowledge of what exists is linked to what we could measure, stated
counterfactually. In the next section we will further elaborate on what this ac-
tually means. For now it is useful to note that in our view, the so-called existing
non-observables, as well as the observables, are linked to definite experimental
2 The anti-realist position of B. van Fraassen: “Science aims to give us theories which are
empirically adequate; and acceptance of a theory involves as belief only that it is empirically
adequate.”[VFra80, p.12].
3 The correct statement of realism according to B. van Fraassen in [VFra80] is that “science
aims to give us, in its theories, a literally true story of what the world is like; and acceptance
of a scientific theory involves the belief that it is true.” (p.8).
CHAPTER 2. UNDERLYING EPISTEMOLOGY 15
leads to a well defined, i.e. a positive or negative, result, where we a priori characterized
which possible outcome(s) count as the positive result — the negative result being all the
other possible outcome(s). As such it makes sense to state that: when it is a fact from
in advance that a positive result would obtain should we perform a definite experimental
project, we can say that “it is sure that a positive response would obtain” and “it is not sure”
otherwise. Note that by our use of expressions such as “it is sure” or even “it is certain”, we
do not mean “it is necessary” because indeed “at the very least one would have to assume that
no uncontrolled external agent could act on the system during the course of the experiment
in question” [Moo99, p.66].
CHAPTER 2. UNDERLYING EPISTEMOLOGY 16
Indeed, Rescher adopts here a Kantian distinction between the realm of phe-
nomena and the realm of things-in-themselves. Of the latter he admits that
we can only say that it exists but not what it is like. It is in this light that
we should understand that for him, “the world we know” is “the world as we
perceive it”:
“The pivotal thesis of conceptual idealism is that we standardly
think of reality in mentalesque terms. It rests on two basic theses:
(1) That our world, the world as we know it, is — inevitably — the
world as we conceive it to be, and
(2) That the concepts (thought-instrumentalities) that we standard-
ly use in characterizing and describing the world contain in their
make-up, somewhere along the line, a reference to the operations of
mind.” [Res95, p.124]
CHAPTER 2. UNDERLYING EPISTEMOLOGY 17
The notion of “critical realism” has been used in several senses in the history of
philosophy and is nowadays often associated with the position of R. Bhaskar.
R. Bhaskar conceives of a transcendental realism linked with a critical natural-
ism with respect to social activity (see [ArcBhaColLawNor98, Bha87, Bha91]).
Bhaskar’s critical realism — the name is an elision of the two phrases “transcen-
dental realism” and “critical naturalism” [ArcBhaColLawNor98] — is more a
program in which he intends to combine ontological realism, epistemological rel-
ativism and judgmental rationality. Since Bhaskar’s theory is too general with
respect to the context of this thesis, it is according to us more appropriate to
focus in the remainder of this chapter on I. Niiniluoto’s critical scientific realism.
The stance of critical realism in the first instance means characterizing scien-
tific progress in terms of increasing truthlikeness [Nii99]. This kind of realism
distinguishes itself from others on several points. First it adopts an ontological
realism and a correspondence theory of truth. Contrary to naive realists, it also
adopts the thesis of fallibilism in which truth, as an essential aim of science, can
only be approached. This idea becomes clearer with the following citation of I.
Niiniluoto, [Nii87]:
“Another line of epistemological thought, critical realism, can also
be motivated in Kantian terms: Kant’s phenomena could be inter-
preted as expressions of our partial knowledge of the things as they
are “in themselves” in the mind-independent reality. This idea ac-
cepts “direct” realism and the correspondence theory of truth, but
it is not “naive”, since our knowledge of reality is always laden with
some conceptual framework and is never assumed to be completely
and exactly true. Still, this view rejects relativism and scepticism. If
we recognize that Kant’s categories are not fixed, but rather histor-
ically changing (Hegel) and bound to man-made languages (Peirce),
then each conceptual framework provides us with a perspective on
objective reality. By employing and enriching such frameworks, and
by empirically testing scientific theories formulated in these frame-
CHAPTER 2. UNDERLYING EPISTEMOLOGY 18
type of intelligent being” [Nii99, p.28]. And since conceptual idealism as such
is compatible with ontological realism and even materialism, Niiniluoto says of
it:
“I am inclined to think that it should not be called “idealism” at
all, since it does not attribute any constitutive role to the human or
superhuman mind.” [Nii99, p.28]
We agree with I. Niiniluoto, that the role attached to the mind in conceptual
idealism points mainly to the fact that reality can only be captured in “mind-
correlative terms of reference”. To explain our position, we adopt Niiniluoto’s
critical realism together with a weak form of conceptual idealism. We have
of course to be careful here and say immediately that we do not agree with
Rescher’s strict Kantian distinction between reality out there and reality as we
perceive it, and that we do not adopt a coherence theory of truth. According
to us, the objects of reality exist independent of actual or possible experience
but we know of them via hypothetical measurements. Furthermore, we cannot
escape the fact that in our theory we focus on isolated parts of reality, and that
these isolated parts of reality will be characterized as physical systems and their
physical properties will be identified according to the best of our ability. It is
obvious that hypothetical measurements, characterizations and identifications
are mind-involving activities. Note that Rescher also conceives of identification
as a mind-involving activity [Res95]. To be more explicit, the identification of
the physical properties and the characterization of the system correspond to a
certain belief of the physicist. His belief can be reinforced or it has to be re-
vised according to the availability of corresponding certain definite experimental
projects which assure him of the actual existence of the properties involved. In
[Sme01] we call these mind-involving activities idealizations. Observe that the
word “idealization” in this sense has nothing to do with ontological idealism.
While Rescher may not agree with us, we believe that our knowledge of reality
approaches the ontological world quite closely, at least close enough to omit
some form of Kantian transcendental idealism. More explicitly we wish to avoid
a “Ding an Sich” which is inaccessible to our knowledge and avoid the concep-
tion of a phenomenon as accessible to our knowledge though only perspective
dependent, as merely a reflection of a deeperlying inaccessible reality. In our
opinion, a part of the external world should be called phenomenon. This con-
cept of phenomenon is conceived to be strongly entangled with the concept of
a “Ding an Sich”. If there is such a thing as a “Ding an Sich” it is not ontolog-
ically nor in our perceivability different from a phenomenon. To recapitulate,
what we focus upon in our scientific activity are idealizations of some phenom-
ena conceived as parts of the external world. Reality itself is mind-independent,
although the process of description “involves” some mind-dependent character-
izations. This does not lead us away from our realism, but we have to take care
not to fall into the trap of essentialism. It is not because reality indicates the
most obvious way to isolate phenomena from their environment, that there is
a “unique ready-made ” structure of reality. There are several alternative ways
in which the world can be carved in pieces [Nii87].
CHAPTER 2. UNDERLYING EPISTEMOLOGY 20
The fact that we focus on a world carved into pieces, i.e. isolated parts of reality,
refers in our opinion to the case that knowledge about the world as a whole is
approachable but not immediately attainable. This idea is linked to a critical
realism which expresses that our knowledge of reality is always laden with a
conceptual framework and is never assumed to be completely and exactly true.
In a similar way, once the isolated parts of reality have been “epistemologically”
characterized by means of their actual and potential properties corresponding to
definite experimental projects we discover or approach the “ontological” world
quite closely. Note finally that the fact that scientific activity focuses on the dis-
covery of reality, does not imply that we do not interfere with or act upon reality
as it is. We may want to describe for example the properties of a physical system
before and after we interfered with it. In both cases, before and after interfer-
ence, the properties are described if they correspond to specific (counterfactual)
definite experimental projects. Discovery does not imply that the physical sys-
tems we have in mind have to be necessarily static; we may as well focus on
the change or creation of properties brought about by a specific action upon the
system. To conclude this section we note that the position of the Geneva School,
CHAPTER 2. UNDERLYING EPISTEMOLOGY 21
“The only claim is that ultimately evidence leads the way: if, despite
all efforts, no scientific theory can be constructed which incorporates
our favorite metaphysical assumptions, then no matter how firmly
entrenched those principles might be, and no matter how fruitful
they may have proved in the past, they must ultimately be given
up.” (p.123)
We have given our view of what structural realism stands for and intended to
show that in further research it could be promising to place the causal-dynamic
extension of OQL — cf. part IV — in that perspective. This does not how-
ever overthrow the kind of critical realism against which we placed the original
development of (static) OQL. Indeed, I. Niiniluoto labels structural realism as
a position that might even be characterized as critical “half-realism”; his own
version of critical scientific realism goes actually beyond structural realism:
“If, without in any way disturbing a system, we can predict with cer-
tainty (i.e., with probability equal to unity) the value of a physical
quantity, then there exists an element of physical reality correspond-
ing to this physical quantity” (p.777)
It is now known that although the content of this paper had been discussed by
all three authors, it was B. Podolsky who wrote it and that A. Einstein himself
would probably have expressed things differently. This was indeed admitted
by A. Einstein in a letter to E. Schrödinger (see [Fin86]). For this reason we
cannot be sure that A. Einstein would have formulated the criterion of reality
in the same way. As we may believe A. Fine in his analysis of the letters and
papers in the Einstein Archives, the reality criterion does not appear in any of
Einstein’s own published expositions of the EPR situation [Fin86, p.62]. Fine
also unravels two passages6 where Einstein may have stated something near to
the above version of the reality criterion, though he concludes that Einstein
6 The paper [Ein48] and a letter written to Tany Ehrenfest, October 12, 1938.
CHAPTER 2. UNDERLYING EPISTEMOLOGY 23
seems to have dropped the reality principle choosing to focus on real states
of affairs instead of on values of observables which are the essential ingredient
of the reality criterion. Indeed, Einstein’s own stance does not rely on the
criterion of reality as stated above. It is important to note here that Einstein
shifted away from positivism and empiricism — present in his early research —
to realism later on. We will not enter the discussion of what kind of realism
Einstein adopted, let us instead conclude with the opinion that the Geneva
School adopts the EPR-principle of reality, though from this we cannot and
indeed may not conclude that they rely on Einstein’s realist stance.
2.2.2 EPR
We believe that the Geneva School provides a completion of Quantum Mechanics
which Einstein would have approved of. Before we state exactly how Einstein
thought such a completion should be accomplished, we will first focus on the
conclusion of the EPR-paper and on Einstein’s own (although not put forward
as final) solution to the problem.
The premises used are carefully chosen, the first (1) is derived in the EPR-
paper and we believe it to be a translation of ∼ (∼ q& ∼ p) by the De Morgan
Laws. More precisely, the argument says that for two physical quantities to
have definite values and to be simultaneously real, they have to be described
that way according to the completeness condition. Further the argument goes,
CHAPTER 2. UNDERLYING EPISTEMOLOGY 24
“If then the wave function provides such a complete description of re-
ality, it would contain these values; these would then be predictable.
This not being the case, we are left with the alternatives stated.”
[EinPodRos35, p.139]
In our analysis this argument says: suppose ∼ q holds together with ∼ p. If
this would be the case and the wave function would provide such a description,
we would be able to predict the values (r) of these quantities. Translated, we
arrive at ∼ p & ∼ q → r. This however is not the case: we cannot predict the
values of these quantities, which means that we have ∼ r given and from that
immediately follows ∼ (∼ p & ∼ q). We give this analysis in detail because
we do not agree with A. Fine in [Fin86, p.33] who claims that the argument to
establish the disjunction in the EPR-paper (the first premise above), is shown
immediately via ∼ q → p.
The second premise (2) is shown to hold in the EPR-paper by means of the
famous EPR-example of two systems and the criterion of reality. More precisely
suppose the wave function gives a complete description of the physical reality
of the system: we presume ∼ p. Take as given, two systems of which we know
the states and let these systems interact for a while. After interaction we can
describe the state of the combined systems by means of the Schrödinger equation
but we cannot calculate the state of a single system. To do this, we need to
perform measurements. Suppose we measure a quantity on the first system, as
a consequence the wave function of the second system changes. The conclusion
drawn in the EPR-paper is that:
As they continue to show, it is possible that the two wave functions of the second
system correspond to (are eigenfunctions of) two non-commuting operators rep-
resenting physical quantities. Take for example, momentum and position. By
measuring either one of them on the first system, we can predict with certainty,
and without in any way disturbing the second system, the value of the consid-
ered quantity. Due to the criterion of reality, if we measured the momentum,
then momentum is the quantity considered an element of reality. But, if we
measured the position, this would be the quantity to be an element of reality.
Since the wave functions of both these cases belong to the same reality “also the
quantities have a simultaneous reality”. By presuming non-commutative oper-
ators and arriving at the simultaneous reality of the corresponding quantities,
we obtain the conclusion ∼ q. We should mention that the words in quotation
marks above, namely “also the quantities have a simultaneous reality”, are not
explicitly to be found in the EPR-paper. It appears to follow logically however,
that since both quantities can be elements of reality, and since the connected
wave functions belong to the same reality, their quantities also belong to the
CHAPTER 2. UNDERLYING EPISTEMOLOGY 25
same reality. “Belonging to the same reality” means that they can be simultane-
ously real. How, otherwise, could they have arrived at the following statement:
“Starting then with the assumption that the wave function does give
a complete description of the physical reality, we arrived at the con-
clusion that two physical quantities, with noncommuting operators,
can have simultaneous reality.” (p.141)
For this reason we again do not agree with A. Fine in [Fin86], wherein he states:
in [Ein36] or [Ein48] but based on a similar argument given in Einstein’s letter to Schrödinger
of June 19, 1935.
10 “The standard construal of Einstein’s statistical interpretation, then, is to suppose that
for any quantum system in any state the quantum theoretical probabilities are to be obtained
from an ensemble representation.” [Fin86, p.44].
11 “can one make an ensemble representation for the quantum probabilities for any set of
observables in any state? The answer is “no”, and Bell’s theorem yields that negative answer
...” [Fin86, p.44].
CHAPTER 2. UNDERLYING EPISTEMOLOGY 27
and quantum mechanics. Precisely where the problems emerge with regard to
the separation principle in relation to quantum mechanics, has been analyzed
in the PhD-thesis of D. Aerts, [Aer81], where it is important to note that D.
Aerts works within the context of OQL against the background of Piron’s Rep-
resentation Theorem — cf. chapter 4. He points out in detail that two essential
quantum axioms (weak modularity and the covering law)12 are not fulfilled in
case we want to describe two “separated systems”, in Aerts’ sense, envisioned
as one system. The EPR-paradox is solved, at least if one accepts that A and
B cannot be regarded as separated and describable by quantum mechanics.
Not only the paradox is lost but also the incompleteness of quantum mechanics.
At least if one replaces quantum mechanics with OQL [Aer82]. OQL can de-
scribe classical situations as well as non-classical ones and is exactly constructed
to be a complete theory from the start. The criterion of reality is built into the
definition of counterfactually stated definite experimental projects, which are
elaborated on in part II, chapter 3. And indeed for every possible element of
reality we add the corresponding experimental projects to the theory [Aer82].
Note however that the incompleteness-problem of quantum mechanics disap-
pears with OQL in favor of a new conception of a physical system which throws
ensembles and statistical interpretations overboard. Obviously this is in direct
contrast with the statistical-ensemble interpretation of Einstein. So, what did
Einstein himself think would be a good solution to the incompleteness-problem?
First of all in a letter of November 14, 1954, to Kupperman, he says that he
does not see the possibility of getting rid of the statistical character of quantum
mechanics without changing the fundamental concepts of the whole structure
[Fin86, p.57]. As Fine sees it, Einstein thought that a completion of the theory
had to come from outside and not from within the theory, such as e.g. the so-
lution by means of hidden variables. A new theoretical framework is necessary,
of course the statistical predictions of quantum mechanics should be derivable
within this framework under special assumptions. He was himself not overly
convinced of his own proposed solution using ensembles. Fine refers for this
matter to a letter from Einstein to Born dated 3th of December, 1953. We be-
lieve that OQL provides a solution which obtains completeness from the outside,
proposes a new conceptual framework and no statistics at all. The remainder
of this thesis should make clear if OQL has succeeded in this task.
28
Chapter 3
29
CHAPTER 3. PHYSICAL AND OPERATIONAL FOUNDATION 30
projects, freely in this text. The latter will mainly be dominant when we cite parts of [Moo99].
CHAPTER 3. PHYSICAL AND OPERATIONAL FOUNDATION 31
It has to be stressed that not every physical system — for example, the sun
— can be prepared and therefore not every physical experiment encompasses
a preparation procedure. If preparations are possible, different types of exper-
iments will demand different preparation procedures. An example given by C.
Piron in [Pir76] focuses on a beam of photons which, when dispatched one by
one, pass or are absorbed by a polarizer placed in the beam. A photon can be
prepared in a way that it will pass a polarizer orientated with angle φ, therefore
it is sufficient to consider those photons which already traversed a first polarizer
orientated in the same direction. With the same experiment it is obviously not
possible to prepare photons that will traverse with certainty the polarizer (angle
φ) as well as a polarizer orientated at φ! &= φ (modulo π).
not certain, the statement “β is true” is false, but we do not say “β is false”.”3
Because “β is false” means that “β is impossible”. This issue will be dealt with
in section 3.2.6 and in chapter 6.
Summarizing the above, as in [Aer81], we say that there are three cases: 1) we
can predict the outcome of a question to be “yes” so the question “is” true;
2) we can predict the outcome of a question to be “no”, so the question “is”
impossible and 3) we are not able to predict the outcome to be “yes”, so the
question “is” not true.
Next, consider the absurd question defined in [Moo99] as the following specific
definite experimental project:
Definition 7.
Absurd question:“Do whatever you wish with the system and assign the response
“no”.” (p.67)
The absurd question can never be a true question, since we would never obtain
the response “yes”. Furthermore, we can state that the absurd question is an
impossible question.
In the definitions given by C. Piron it is more explicitly stated that the answer
obtained from effectively performing the arbitrarily chosen question is the an-
swer attributed to the product question. We then say that the product question
is true if it is sure that the answer “yes” would obtain should we perform the
experiment. In other words, we follow [Moo99, p.67],
Proposition 2.
The product of a family of definite experimental projects called A is a “certain
definite experimental project” (the product question is true) for a particular
physical system iff each α ∈ A is certain (true) for the particular physical system.
The product question itself is always defined with respect to a specific family of
questions.
We call the inverse of a specific question, named α, true iff α is impossible and
the inverse of α is impossible iff α is true (see [Moo99]). Given the relationship
between impossible and not true questions, “impossible” is obviously not the
logical negation of “true” — it is not the same as “never certain” or “not true”.
When a question α is not true, this implies nothing with respect to its inverse
which can be either not true or true corresponding to the cases where α is or is
not also impossible.
We could indeed go into further detail about the four causes Aristotle describes
to know the being and becoming of things, but this would lead us too far away
from the subject in hand. In any case, we basically already dealt with three
of the four causes Aristotle considers: (1) the material cause is the material
substrate, (2) formal cause is the form to be realized, (3) the efficient cause
“represents the influences from the outside world that cause the process of mo-
tion towards realization of its true nature” [VerCoe99, p.171] — cf. the above
source of becoming — and (4) the final cause is then the endpoint of this process
or “that for the sake of which”. There still remains however one thing which,
according to us, is essential to understand Aristotle’s reasoning and that is his
introduction of a prime or first mover. This first mover is not to be conceived
of as a God who created the world but rather as the origin of all movement and
change which, in itself unmovable, has to be envisaged as pure actuality. The
main reason for introducing this first mover is firstly, as explained in [Meta-
physics Book Λ,1071b 12-], that if there is something capable of moving things
but which does not perform the act, there will be no movement at all. Secondly,
if everything which moves is itself being moved — called an intermediate mover
— we end up with an infinite causal regression which according to Aristotle has
6 Internal means present “in the very thing which suffers change” [Metaphysics Book Θ,
1049a 13]
CHAPTER 3. PHYSICAL AND OPERATIONAL FOUNDATION 38
to be stopped at some point [Metaphysics Book Λ, 1070a 3-, Physics Book VIII,
267b 1]. Thirdly, eternal movement, for instance the rotation of planets has to
be caused by an unmovable eternal substance [Metaphysics Book Λ, 1073a 27].
We quote [Metaphysics Book Λ, 1069b 36-]:
“For everything that changes is something and is changed by some-
thing and into something. That by which it is changed is the pri-
mary mover; that which is changed, the matter; that into which it
is changed, the form.”
Hereby we have captured the most essential of Aristotle’s theory involving ac-
tuality and potentiality for our needs. This should allow us now to explore the
“differences” of use by the Geneva School. There are some clear points of differ-
ence right from the start: the Geneva School does not adopt Aristotle’s causal
teleology — everything moves towards an end or the realization of its essence
— and neither poses the existence of a first mover.7 We can say however that in
a parallel way, actuality and potentiality are considered by the Geneva School
as modes of being. But we must be careful since in the latter those modes of
being are only assignable to predicable properties, conceived as characteristics
of physical systems. It should be clear that the Geneva School does not consider
substantial change, this reduces the comparison with Aristotle’s use of the above
mentioned notions to the cases of accidental change. Regarding actuality, an
actual property is conceived as an attribute which exists; it is some realization
in reality or in other words; an element of reality. A potential property on the
other hand does not exist in the same way as an actual one, it is conceived
merely as a capability with respect to an actualization since there is always a
chance that it could be realized after the system has been changed. Note that
an actual property is itself also a capability though not in the same sense as
a potential property, i.e. “being breakable” has also been viewed as the ca-
pability of “being broken” in the sense that it is certain that the system will
be broken, i.e. leads to the positive result, should we perform the experiment.
If “being breakable” is a potential property of a particular physical system, it
is not certain that it will be broken should we perform the experiment. In a
sense we may as such say that an actual property is an element of reality and a
“certain” capability, i.e. it is certain that a specific result would obtain should
we perform a well defined experimental project which may change the system
in a specific way; while a potential property is only a “not certain” capability,
i.e. it is not certain that a specific result would obtain should we perform a well
defined experimental project which may change the system in a specific way.
The notions of actuality and potentiality are set out thus by C. Piron in [Pir78]:
“Parmi les propriétés définies pour un système donné, il y a celles
que le système possède actuellement, c’est-à-dire en acte, et celles
7 Although parallels could be established between Aristotle’s notion of substance and the
notion of isolated physical system, this is on the first sight not straightforward and would lead
us to far away from the subject in hand. Henceforth we stick to a comparison with respect to
the notions of actuality and potentiality.
CHAPTER 3. PHYSICAL AND OPERATIONAL FOUNDATION 39
Since the main point of the Geneva School is that every physical system can be
described by means of its properties, obviously any change of a physical system
can also be described by a change of its properties. It is in this sense that no
account is given of Aristotle’s substantial change. Also, contrary to Aristotle,
change should not necessarily be taken as an action in function of some telos. In
chapter 5, section 5.2 we deal with special kinds of measurements and in part IV
we further analyze deterministic and indeterministic changes a physical system
can undergo, but we wish to stress already now that a deterministic change will
not be seen as the realization of a system’s essence, and that an indeterministic
change can for instance be due to any measurement which may be performed
without the a priori aim of actualizing a specific potentiality. It is important
that despite this, we do not give up the idea that every change can be explained
as the passage between what is actual and potential. As C. Piron says, [Pir83]:
“Mais entendons nous bien, je ne désire pas puiser dans les résultats
et les conclusions de la physique d’Aristote qui sont bien sûr com-
plètement dépassés, mais reprendre certains de ses concepts de base
et adapter certains de ses arguments au problème tel que nous le
connaissons aujourd’hui.” (p.170)
Since we have seen that the actual properties are those which are realized for a
physical system, and that it only makes sense to talk about actual or potential
properties if the system is conceived to be in a specific state, we should make
the link between actual properties and the state of a physical system explicit.
Indeed, it is so that the state stands in a correspondence relation to the system’s
actual properties, though contra C. Piron and D. Aerts the state will here not be
CHAPTER 3. PHYSICAL AND OPERATIONAL FOUNDATION 41
We will deal with the relationship between the state property — the notion is
a translation of C. Piron’s propriété-état introduced in [Pir98] — which is the
state-representative and the state set in the next chapter. To return to the
previous section, we now see that when a system’s ontologically realized state
changes, the property which represents it changes as well as its related state
set. And it is because of this that it is possible to conceive of any change as a
change of the attributes of a system.
Mathematical Framework
42
CHAPTER 4. MATHEMATICAL FRAMEWORK 43
in [Pir81] that,
Definition 13.
“A question α is said to be stronger than a question β, i.e. α ≺ β, if every time
α is true, β is true as well.” (p.399)
(1) Reflexive: α ≈ α,
(2) Symmetric: α ≈ β ⇒ β ≈ α,
(3) Transitive: α ≈ β and β ≈ γ ⇒ α ≈ γ.
When a question α is true then so are all questions equivalent to α. From this
it is evident that every true question is equivalent to I, the trivial question.
And every question α is equivalent with αΠI because αΠI is true iff α is true.1
Coming back to some issues concerning the product question we write:
ΠJ αi is true ⇔ αi is true ∀i ∈ J
(ΠJ αi )∼ = ΠJ (αi∼ )
4.3 Properties
4.3.1 Modeling Properties
About a property associated with a particular question we can say the following,
Proposition 3.
There is a one-to-one correspondence between the property a, associated with the
definite experimental project (question) α, and the equivalence class of questions
[α] which is the collection of all definite experimental projects β such that β ≈ α.
This correspondence is realized through the identification of a and [α].
In other words, to each equivalence class of questions there “corresponds” a
property. For our convenience we will introduce the following notion which
expresses that a is associated with α ∈ [α]:
ζ : Q → L : α 1→ a
1 Note that αΠI is another notation for Π(α, I).
CHAPTER 4. MATHEMATICAL FRAMEWORK 44
Here we must add that in the Geneva School literature one originally defined a
as [α]. This however is a definitional coordination which for the sake of our form
of operationalism should be avoided at all costs. Just as the state of a physical
system is not merely a set of actual properties, a property is not merely a set
of equivalent questions. Furthermore we believe that equivalent questions are
not candidate elements of reality. Taking this into account we want to note the
fact that in his earlier work, C. Piron used the notion of “proposition” which
stands in a one-to-one relation to properties. Although we may not generalize
this utterance since in [JauPir69] only a true proposition2 is called a property.
We call the property expressing the existence of a physical system 1 = ζ(I) the
trivial property and 0 = ζ(O) the absurd property.
Theorem 1.
The collection of all properties of a physical system, noted as L, is a complete
lattice.
Quoting C. Piron, [Pir98], this means exactly:
2 “Hence we see that the proposition ā is true if and only if any (and therefore all) of the
α ∈ ā are true. If the proposition ā is true we shall call it a property of the system.” [JauPir69,
p.844]. We introduced the notation ā to distinguish propositions from properties and of course
in [JauPir69] the following holds: ā := [α].
CHAPTER 4. MATHEMATICAL FRAMEWORK 45
The meet3 then defines a new property which satisfies the following relations,
(see [Pir72, p.518]),
∧J ai ≤ aj ∀j ∈ J and b ≤ aj ∀j ∈ J ⇒ b ≤ ∧J ai
because “ΠJ αi is true” iff “αj is true” ∀j ∈ J, and if “β is true” implies “αj is
true” ∀j ∈ J, then “β is true” implies “ΠJ αi is true”. The operationally defined
product question attributes a physical meaning to the meet. This guarantees
that the poset of properties is a complete meet semilattice. The join or least
upper bound, on the other hand, cannot be attributed a direct physical meaning
in the same sense and is defined by means of Birkhoff’s theorem by which every
complete meet semilattice is a complete lattice, (see [Moo99, Pir72, Pir76]),
#
∨J ai = {b ∈ L | ∀i ∈ J : ai ≤ b}
It is written in [Pir76] that ∨J ai always exists because the set L always contains
the trivial property 1. !
When we review the use of meet and join, we see that the meet corresponds to
the classical conjunction whereas the join will not always correspond to the clas-
sical disjunction, due to the possibility of superpositions for quantum systems.
We will come back to this issue in part III.
Proposition 5.
“a ∧ b is actual” ⇔ “a is actual” and “b is actual” ∀a, b ∈ L
“a ∨ b is actual” ⇐ “a is actual” or “b is actual” ∀a, b ∈ L
As is proved in [Pir72], if the latter implication holds in both directions, L is a
distributive lattice.
3 Although ΠJ αi ∈ [ΠJ αi ] corresponds to ∧J ai , any other question in the equivalence class
[ΠJ αi ] need not have the form of ΠJ αi (cf. [FouRan84]).
CHAPTER 4. MATHEMATICAL FRAMEWORK 46
4.4 States
4.4.1 Modeling States
Above we pointed out that the state set is the collection of all properties which
are actual if the system would be realized in a specified state. We now model
this as follows, the state set S(E) = {a | a ∈ L, a actual in E}, where E is
an element of Σ, the state space, which is the collection of all states which are
possible for a particular physical system. In [JauPir69, Pir72] it is stated that
if we have a complete lattice L, S(E) satisfies some specific properties:
(1) If a ∈ S(E), b ∈ L and a ≤ b then b ∈ S(E)
When a is actual and stronger than b, then b is actual and must
therefore be an element of S(E)
Note that when E1 and E2 are represented by the state properties p and q
respectively, we represent E1 ⊥ E2 also by p ⊥ q. If T is a subset of Σ, we use
the notation T ⊥ for the set of all states orthogonal to those in T . Formally,
[Pir98]:
Definition 17.
“T ⊥ = {E ∈ Σ | E ⊥ η, η ∈ T }.” (p.22)
Note that such orthogonality relations were explored first by D. Aerts in [Aer81].
In a way this forms the cornerstone of a full “state space description” of a
physical system, parallel to its property lattice description — for details we
refer to the state-property duality treated in section 4.7.
Axiom 1.
“Let pE1 and pE2 be state properties. If pE1 ≤ pE2 then pE1 = pE2 .” (p.71)
One state property can never imply the actuality of another state property.
This axiom is already insinuated above in 4.4.1, where we pointed out by means
of an example that two different state sets can never be subsets of one another.
Although several properties can be actual in two different states in which the
system can be ontologically realized, there is always at least one property which
is actual in one state but not in the other. More explicitly as said by C. Piron
in [Pir98]:
“C’est là une loi physique déjà formulée par Aristote: si le système
change, passe de l’état E1 à l’état E2 , il s’enrichit de propriétés nou-
velles qui s’actualisent, mais du même coup, il en perd nécessaire-
ment d’autres qui passent en potentialité, ce qui n’est pas possible
si E1 est entièrement contenu dans E2 .” (p.21)
We traced this physical law as C. Piron calls it, back to Aristotle’s [Ari95, On
Generation and Corruption Book I, 318a 23-; Book II, 336b 23-] where it is made
clear that a “passing-away” of this is a “coming-to be” of something else, and
the “coming-to be” of this is a “passing-away” of something else. In a certain
sense, movement is to be conceived as becoming something and being corrupted
of something else, or, when we start with something, it can be seen as obtaining
some qualifications while losing others. It is in this light that Aristotle explains
that there are three principles which underlie any change that takes place: form,
privation and matter [Ari95, Metaphysics Book Λ, 1070b 18-]. As an example
Aristotle deals with colour: here the form is white, the privation is black and the
matter is a surface. In [Ari95, Metaphysics Book Z, 1033a 8-] another example
is given: the matter is a man which becomes healthy — the form — when he is
deprived of being invalid. So we see that becoming healthy goes hand-in-hand
with the passing away of being invalid.
From this first axiom, immediately follows that all state properties of a physical
system are the atoms of the property lattice — for more details see Th.2 below.
Indeed, it is pointed out in [Pir72, Pir98] that each state can be represented
by an atom and every atom represents a state. We can go even further with
the atomistic property lattice as stated in the following theorem, where it then
becomes possible to characterize each property in L by means of its underlying
state properties, [Moo99]:
Theorem 2. "
“Each property lattice L is atomistic: a = {p ≤ a | p an atom } for each
a ∈ L.” (p.71)
Proof
We go over the proof given in [Moo99] in detail.
We have to show that each property lattice is atomistic, or in other words that
the state properties are the atoms of the lattice and that each property can be
CHAPTER 4. MATHEMATICAL FRAMEWORK 49
As we will show, the next axiom allows us to postulate the existence of “oppo-
site” properties.
Axiom 2.
“For each possible state E there exists at least one definite experimental project
(question) α such that α is certain (true) in the possible state E ! if and only if
E ⊥ E ! .” [Moo99, p.72]
Stated otherwise, for every state property p in ΣL , there exists a property p! ∈ L
which is actual iff the state is orthogonal to the one represented by p. As we
will show in the following theorem, the property p! is the opposite of the state
property p. The opposite of any other property of L is then defined by means
of the meet of such opposite properties p! . Following [Moo99], we represent the
opposite of a property by means of the map:
!
!
: L → L : a 1→ {p! | p ≤ a}.
which indicates in [Aer81] the following,
Proposition 8.
a! is actual ⇔ p! is actual for every p ≤ a
a! is actual ⇔ the system is in a state q ⊥ p
We can now look at the properties of ! , which are stated in [Moo99],
Theorem 3.
(1) p! = p!
(2) If a ≤ b then b! ≤ a!
(3) a ∧ a! = 0 for any a
(4) a ≤ a!! and a!!! = a! for any a
(5) (∨J ai )! = ∧J (a!i )
CHAPTER 4. MATHEMATICAL FRAMEWORK 50
Proof
We go over the proof given in [Moo99], although we add comments and details.
!
(1) If we start from p! = {q ! | q ≤ p} where both q and p are state properties
then these state properties have to be equal. ! This means that there is only one
orthogonal state to consider, since p! = {p! } = p! .
(2) Suppose a ≤ b holds, then b is actual in at least all states in which a is
actual. For every possible state represented by p in which b is actual, we have
a property p! which is actual in the orthogonal state. The same holds for a, for
every possible state represented by p in which a is actual we have a property p!
which is actual in the orthogonal state. Since there may be possible states in
which b!is actual and a ! not, we have: {p! | p ≤ a} ⊆ {p! | p ≤ b}. This results
in b! = {p! | p ≤ b} ≤ {p! | p ≤ a} = a! .
(3) We first suppose that a ∧ a! &= 0 holds. This implies that 0 ≤ pE ≤ a ∧ a!
holds. Hence we obtain: pE ≤ a! and pE ≤ a, which implies pE ≤ a! ≤ p!E = p!E
!
due to a! = {p!E | pE ≤ a} ≤ p!E . Note that pE ≤ p!E is just another notation
for E ⊥ E, which expresses an impossibility since the orthogonality relation is
antireflexive. Hence we conclude with a ∧ a! = 0.
(4) From the definition of our orthogonality relation we see that E ⊥ E ! holds
iff also E ! ⊥ E holds, which means for the state properties that pE ⊥ ! pE ! iff
pE ! ⊥ pE . In
! other words: p E ≤ p !
E ! iff p E ! ≤ p !
E . Hence we know that p ≤ {q ! |
p ≤ q ! } = {q ! | q ≤ p! } = p!! — here we used the fact that if p is less than
or equal to any element of a given set, it is less than or equal to the greatest
lower bound of that set. Let us now concentrate on the first part of what we
have to prove. Suppose p ≤ a then, following Th.3.2, a! ≤ p! holds and further
by applying Th.3.2 again we obtain p ≤ p!! ≤ a!! . This results in: a ≤ a!! and
a!!! ≤ a! . However since a! ≤ a! holds and when we apply the first part of this
proof we obtain a! ≤ a!!! which together with a!!! ≤ a! gives the result a!!! = a! .
(5) We have ai ≤ ∨J ai for each i ∈ J and so — applying Th.3.2 — (∨J ai )! ≤ a!i
for all i ∈ J. Since we know that ∧J (a!i ) ≤ a!i for every i ∈ J we have (∨J ai )! ≤
∧J (a!i ). And reversely, if p ≤ ∧J (a!i ) then p ≤ a!i for all i ∈ J and so ai ≤ a!!i ≤ p!
for all i ∈ J. Hence by definition of the join ∨J ai ≤ p! , by Th.3.2 and Th.3.4
p ≤ p!! ≤ (∨J ai )! and so ∧J (a!i ) ≤ (∨J ai )! . !
The final fundamental axiom insures that each property is the opposite of an-
other one:
Axiom 3.
“The map ! : L → L is surjective.” [Moo99, p.72]
Recall that a map f : M1 → M2 is called surjective or onto when ∀y ∈ M2 ∃x ∈
M1 : y = f (x). A map f : M1 → M2 is called injective or one-to-one when
∀x ∈ M1 , ∀y ∈ M1 : x &= y ⇒ f (x) &= f (y). As such we see that due to this
axiom, for each property a ∈ L there exists a property b ∈ L such that a = b!
[Moo99]. Furthermore we can state that ! is an orthocomplementation, which
satisfies [Pir98]:
CHAPTER 4. MATHEMATICAL FRAMEWORK 51
(1) a!! = a
(2) a ∧ a! = 0
(3) a ≤ b ⇒ b! ≤ a!
We concentrate only on the first step as the other steps have been proved above.
In [Moo99] it is stated that if for each a ∈ L there exists a property b ∈ L such
that a = b! , we have a = b! = b!!! = a!! .
Returning to the last two axioms we can say with the words of C. Piron in
[Pir83] that they find their origin in Aristotle’s work:
“Un autre principe enseigné également par Aristote affirme que cha-
que propriété a son contraire. Ceci nous conduit à postuler que le
sous-ensemble des états de Σ orthogonaux à tous ceux d’un sous-
ensemble représentant une propriété, représente toujours également
une propriété, la propriété contraire. Enfin, dans le même ordre
d’idées, nous supposerons aussi que toute propriété peut être con-
sidérée comme le contraire d’une autre.”
The closest we get to this when reading Aristotle, is his saying in [Ari95, Meta-
physics Book Γ, 1005a 3-]:
“For all things are either contraries or composed of contraries, and
unity and plurality are the starting-points of all contraries.”
In [Physics Book I, 188b 25-] this idea is extended in the way that everything
that comes to be by a natural process is said to be a contrary or product of
contraries. We assume that with the product of contraries Aristotle means an
intermediate, derived and composed from contraries such as a colour which is
derived from black and white (see [Ari95, Physics Book I, 188b 21-]). The idea
is then that a thing alters “towards” contrary qualifications [Ari95, Categories,
15b 15-]. With respect to axiom 3, we can even say that Aristotle in [Ari95,
Categories, 11b 34-] explains that a pair of contraries are said to be contraries of
one another:
“Things opposed as contraries, however, are never called just what
they are, in relation to one another, though they are called contraries
of one another. For the good is not called good of the bad, but the
contrary of it; and the white not white of the black, but its contrary.”
Opposites are then characterized as, among other things, contraries or as the
attributes that cannot be present at the same time in that which is receptive of
both [Ari95, Metaphysics Book ∆, 1018a 20-]. Note however that with Aristotle
the idea is given that while two contraries cannot belong at the same moment
to the same thing, it is possible that neither of them belongs when there is an
intermediate, [Ari95, Metaphysics Book I, 1056a 35-]:
“For the combined denial of opposites applies when there is an in-
termediate and a certain natural interval; ...”
CHAPTER 4. MATHEMATICAL FRAMEWORK 52
We have to take into account that there is not an intermediate in all cases,
for instance Aristotle gives the example of a shoe and a hand, there is nothing
intermediate in that case. As such an intermediate is impossible between things
that are not opposite [Ari95, Metaphysics Book I, 1057a 31-]. On the other hand,
for contraries between which there is nothing intermediate, it is necessarily so
that only one of them belongs to the things in which they naturally occur
or are predicated of [Ari95, Categories, 12b 26-]. The latter case comes quite
close to the “classical situation” considered by the Geneva School in which a
property or its opposite will always be actual. The fact that it is possible in
non-classical systems for both a property and its opposite to be potential, in
a way corresponds to the idea of an intermediate situation. We deal in more
detail with classical and non-classical systems in section 4.8. A final remark is
necessary since for Aristotle it is not so that all qualifications (quality’s) admit of
a contrary one. Aristotle mentions in [Ari95, Categories, 10b 15-] that “... there
is no contrary to red or yellow or such colours though they are qualifications.”.
As such we tried to be more careful than C. Piron - cf. his above utterance -
and explained in detail what Aristotle meant by contraries and changement of
things towards their contraries.
Proposition 10.
a! is not necessarily tested by questions in the equivalence class [α∼ ], hence a!
is not necessarily equal to ζ(α∼ ).
Given a = ζ(α), it is easy to see that when the property b = ζ(α∼ ) is actual,
a! is not necessarily an actual property and vice versa. So if b is actual then
α∼ is a true question and this is so when α is impossible and when a = ζ(α) is
potential. And when a is potential this does not always imply that its opposite
should be actual, except in the classical case as we will see later on. Conversely,
if a! is actual then a = ζ(α) is potential and α is then not a true question.
However, since we do not know whether α is impossible we cannot claim that
α∼ is true or that b = ζ(α∼ ) is actual.
We could avoid this little inconvenience and follow D. Aerts in [Aer81, Aer82,
Aer83a] who by definition introduces primitive questions. As for example in
[Aer82] a primitive question is described as follows:
“If α is a question testing the property a such that α∼ tests the
property b, then α is a primitive question iff whenever the entity is
in a state orthogonal to a, then α∼ is true and whenever the entity is
in a state orthogonal to b, then α is true. If α is a primitive question
then also α∼ is a primitive question.” (p.1144)
Where we conceive a state to be orthogonal to a property a iff it is orthogonal
to every state E in which a is actual. As such for primitive questions it can then
be proved that if α ≈ β then α∼ ≈ β ∼ [Aer82]. Furthermore it is possible to
define primitive properties as those testable by primitive questions and where it
can be assumed that a! is the property tested by α∼ . As in [CoeMooSme(nd)a]
we think that such primitive questions are maybe a convenient heuristic, but
we are not inclined to accept them as the building blocks of the formalism
— in [Aer82] it are the primitive questions which form the generating set of
questions for a property lattice. We agree however that in axiom 2, it would
be nice to see the mentioned question as the inverse of another question. This
could then even be linked to the definition of the orthogonality relation on
states — see def.16. As such an orthocomplement of a property could then be
associated with the inverse question which is the product of certain component
inverse questions. We will however not adopt this approach since, as mentioned
above, there is no direct physical-operational meaning to be attributed to an
orthocomplement of a property. Coming back to how an orthocomplement is
defined, we draw the following link between orthocomplements of properties and
questions, [CoeMooSme(nd)a]:
!
“... the condition a! = {p! | p ≤ a} implies that a! is generated as
the product of those βp with p majorised by a.”
where βp is then exactly such a question as mentioned in axiom 2. Note also
that in a similar way as we conceive every property of L as the join of under-
lying atoms, we can see every property of L as an orthocomplement of another
property built up by means of coatoms.
CHAPTER 4. MATHEMATICAL FRAMEWORK 54
Let us stress that given the Cartan map, a property is actual in a certain state
if the point representing that state of the system is an element of µ(a); if not
the property is potential [Pir98]. This then allows us to rewrite S(E) = {a |
a ∈ L, a actual in E} as S(E) = {a | a ∈ L, E ∈ µ(a)}. The interpretation of
actual and potential properties by means of the Cartan map remains the same
for classical and non-classical systems. The difference between classical and
non-classical cases lies in the fact that the Cartan map µ is an injective map
in general — for both the non-classical and classical systems — though is said
to be also surjective in the classical case. We will come back to the difference
between non-classical and classical systems in the next section and will for the
moment concentrate on a theorem stating that the Cartan map is also in the
general case a surjective map if we limit the range of µ to the biorthogonal
closed subsets of Σ.7 The theorem is given in [Moo99]:
Theorem 5.
“The property images µ(a) are exactly the biorthogonal subsets of Σ” (p.73)
Proof
To prove this we follow [Moo99] explicitly and add some comments.
First we have to show that µ(a)⊥⊥ = µ(a), and for this it suffices to show that
µ(a! ) = µ(a)⊥ since then µ(a)⊥⊥ = µ(a!! ) = µ(a). According to definition 17,
we know that µ(a)⊥ = {E ! | (∀E)(E ∈ µ(a) ⇒ E ⊥ E ! )}. Hence E ! ∈ µ(a)⊥ iff
(∀E)(pE ≤ a ⇒ pE ≤ p!E ! ). However, since we know that if a, b ∈ L, then a ≤ b
iff whenever p ∈ ΣL , p ≤ a then p ≤ b cf. proposition 6, we obtain a ≤ p!E ! .
But a ≤ p!E ! iff pE ! ≤ a! , and since pE ! is a state property this implies that a! is
actual in state E ! so E ! ∈ µ(a! ). Henceforth µ(a! ) = µ(a)⊥ .
Secondly, we have to show that every biorthogonal subset of Σ is the image of
some property. Note firstly that µ(pE ) = {E ! | pE ! ≤ pE } = {E}, which means
that when pE is the state property representing E, it is actual in the state E and
so pE ! = pE . Similarly we have µ(p!E ) = µ(pE )⊥ = {E}⊥ . Secondly we note that
the states in which ∧J ai is an actual property, is exactly the intersection of all
those states
! in which the ai (i ∈ J) are actual. Now, if T ⊆ Σ, biorthogonal
and a = {pE | E ∈ T ⊥ }, then we see that T is the set of states in which a is
!
actual because:
$
$ =! {µ(pE ) |⊥E ∈ T$ } where µ(a) = µ(∧{p!E | E ∈ T ⊥ }) and
! ⊥
µ(a)
{µ(pE ) | E ∈ T } = {{E}⊥ | E ∈ T ⊥ } = {E ! | (∀E ∈ T ⊥ )E ! ⊥ E} = T ⊥⊥ =
T. !
7 We take subsets of Σ such as T ⊆ Σ which satisfy: T ⊥⊥ = T .
CHAPTER 4. MATHEMATICAL FRAMEWORK 56
We have shown that every biorthogonal subset of Σ is the image of some property
(&= 0). Henceforth, for every property (&= 0) there is always at least one state in
which it is actual, and via the Cartan map we have found a way to indicate the
exact subset of the state space Σ in which that is the case. Summarizing the
above we can say with [Moo99] that:
“each physical system can be characterized by: its property lattice,
a complete, atomistic and orthocomplemented lattice; and its state
space, a set together with an antireflexive, symmetric, and separat-
ing relation.” (p.73)8
The physical duality between states and properties has been extended to a cat-
egorical equivalence [Moo99, Stu99]. Quite recently, category theory in relation
to OQL seemed to increase in importance, in particular with respect to the
dynamic extension on which we will focus in part IV. The reader will probably
agree that the use of another mathematical language at this stage would lead
to unnecessary confusion.
imposes also that for every property an associated question or the inverse is
true, a priori, even before we have decided to perform any experiment (even
when the measuring devices are not given).
Proof
See Th.1.5.15. in [Pir98]. !
Continuing with the state space, we see that classical systems have a maximal
orthogonality relation whereas for quantum systems this relation is minimal
[Moo99]. A maximal orthogonality relation defined on a state space means
that the state space is classical in the sense that any two distinct states are
orthogonal. This leads us to conclude that for classical systems, the given
axiom 2 can be reformulated similarly as in [Aer81, p.35],
For each possible state E there exists at least one definite experi-
mental project α such that α is certain in the possible state E ! iff E !
is different from E.
We will now deal with the notion of a classical property as defined in [Pir98]:
Definition 18.
“Une propriété a ∈ L est appelée une propriété classique si, quel que soit l’état
du système, ou bien a est actuelle ou bien a! est actuelle” (p.24)
Let us now draw the relation between classical properties, classical questions
and classical systems — cf. axiom 4. If α is classical and when ζ(α) = a then
CHAPTER 4. MATHEMATICAL FRAMEWORK 58
Proposition 11.
If the system is classical then all its properties are classical.
But let’s take a look at the converse: if a is classical, the existence of a classical
α such that ζ(α) = a is not guaranteed in general (see [Pir98]). Indeed we can
consider an example where for a given complete boolean property lattice in every
state c ∧ d = ζ(αΠβ) ∈ L is actual or c! ∨ d! ∈ L is actual but where it is not the
case that in every state αΠβ is true or α∼ Πβ ∼ is true. A comparable situation
has — in a somewhat different context — been elaborated on in [Aer81, p.39]
and points to the fact that even when the property lattice is boolean, the system
is not necessarily classical in the sense of axiom 4. Similarly we can say that if
every property of a system is classical it still can be the case that the system is
not classical. This leads us to pose the following in general:
Proposition 12.
The existence of classical properties does not guarantee that the system is clas-
sical.
Of course, a physical system will in general have classical and non-classical
properties [Pir88]. Now we will concentrate on a! . Taking the same case in
which α is a classical question such that ζ(α) = a, we see that ζ(α∼ ) = a! and
since a is classical, also a! is classical. This is obvious since in an arbitrary state
E when a classical question is true or its inverse is true then at the same moment
a classical property has to be actual or its opposite has to be actual.
Note that the set of classical properties of a property lattice L is called the
center ZL of L. As such we can say that classical properties coincide with
central elements, where we characterize central elements as [Moo99]:
“an element c of a bounded lattice X is called central if there exists
c! ∈ X such that for each a ∈ X we have that a = (a ∨ c) ∧ (a ∨ c! ) =
(a ∧ c) ∨ (a ∧ c! ).” (p.75)
The center ZL of central elements of L can be proved to form a complete
boolean sublattice equipped with the same orthocomplementation and with its
own atoms. For more details on this we refer to [Pir98, Th.1.7.6] and [Moo99,
p.75]. Let
! us just note that the atoms of this boolean sublattice have the form
Z(p) = {a ∈ ZL | p ≤ a} [Stu99] — where p itself has been taken as an atom
of L. What we learn from this is that it is possible to give only the classical
description of a physical system. An experimenter can for instance concentrate
only on the classical properties which will lead him to the boolean property
9 For a classical question the property b = ζ(α∼ ) is an orthocomplement of a = ζ(α). The
reason is that for a classical question we must have a ∨ b = 1 and so b = a! (see [FouRan84]).
CHAPTER 4. MATHEMATICAL FRAMEWORK 59
lattice ZL . However, if the system also has non-classical properties, a full char-
acterization of the system should lead to L of which ZL is a sublattice. If the
experimenter insists on concentrating on the classical part, he will identify in
his description two different atoms p, q of L in case those atoms determine the
same atom of ZL — where Z(p) = Z(q). The experimenter cannot distinguish
between states that are macroscopically equivalent as C. Piron calls them in
[Pir98]. From the physical point of view, C. Piron states it as follows, [Pir98]:
“si l’expérimentateur ne distingue pas deux états macroscopique-
ment équivalents, il décrira le système à l’aide d’un treillis classique,
c’est-à-dire un treillis isomorphe aux sous-ensembles d’un ensem-
ble. Nous comprenons alors le rôle de la physique classique comme
ébauche de toute théorie.” (p.25)
When we go a little step further in this reasoning we see that if the only classical
properties of a property lattice are 1 and 0, then we can conclude that any
two
! states will always
! be macroscopically equivalent. Obviously here Z(p) =
{1 | p ≤ 1} = {1 | q ≤ 1} = Z(q) for any p, q in L. In this case we call
the property lattice purely quantum [Pir98]. If the system possesses quantum
and classical properties it is said to be a non-classical system that possesses
superselection rules and can be described by a lattice as follows, [Pir98]:
“Dans le cas général, ..., le treillis du système peut être décrit par
la donnée d’une famille de treillis purement quantiques indexés par
les états macroscopiques du système. Pour des raisons purement
historiques on dit alors que le système possède des règles de super-
sélection.” (p.25)
We have talked already about several kinds of physical systems, and all of them
can be described by means of a property lattice which fulfill axioms 1,2, and
3. The classical axiom 4 is indeed very strong and not always valid — cf. the
example given in [Pir98] of a Stern-Gerlach experiment on a silver atom. There
is now one extra axiom which needs to be satisfied if we want to conceive OQL
as a genuine quantum theory with super selection rules. For this, we need to
drop axiom 4 and introduce another one on top of axioms 1,2 and 3. See for
instance [Aer83a, Pir76],
Axiom 5.
Suppose L is a complete and orthocomplemented lattice, then:
(5.1) L is weak modular iff for a, b ∈ L and a ≤ b we have b = a ∨ (b ∧ a! )
(5.2) L satisfies the covering law iff for L atomic (which means that for every
a ∈ L there exists an atom p of L such that p ≤ a) and a, x ∈ L and p is an
atom of L whenever we have a ∧ p = 0 and a ≤ x ≤ (a ∨ p) then holds: x = a
or x = a ∨ p.
This quantum axiom as we will call it, has a very mathematical nature according
to D. Aerts in [Aer81]. As already mentioned in part I, it is this axiom which
CHAPTER 4. MATHEMATICAL FRAMEWORK 60
4.9 Compatibility
In the literature one encounters the notion of commutativity in the context of
abstract algebraic structures where the notion of compatibility is more often
used in the specific context of the properties of physical systems or propositions
of physical systems as it used to be called earlier (see [BelCas84]). It will become
clear in this section that the existence of noncompatible properties indicates a
difference between classical and non-classical cases. This has been stated more
general in [BelCas84]:
“The occurrence of noncommutative algebraic structure marks a
branching point between classical and quantum mechanics.” (p.125)
10 Note that a property lattice which is complete, atomic, orthomodular and satisfies the
covering law has been called in [Aer81] and [Val00] a Piron lattice.
CHAPTER 4. MATHEMATICAL FRAMEWORK 61
The meaning of this will become apparent if we examine the formal definition
of compatible properties as given in [Pir76]:
Definition 19.
“In a CROC (complete, orthocomplemented and weakly modular property lattice)
two propositions (properties) b and c are said to be compatible if the sublattice
generated by {b, b! , c, c! } is distributive, a property we shall denote by b ↔ c.”
(p.25)11
We stress that the notion of compatibility has this rather mathematical meaning.
As stressed in [Aer83b], the common claim that two properties are compatible
if their associated questions could actually be performed simultaneously with-
out disturbing each other — see for instance also [JauPir63] — does not hold
in general. This is clear when we refer again to the example in [Aer81, p.39]
of a complete boolean lattice in which every property is classical and compat-
ible, but where it is definitely not the case that any pair of properties can be
simultaneously tested without each measurement disturbing the other.
The following definitions occur in [Pir76], and will be used to prove Th.7:
Definition 20.
“A sublattice of L is a subset S of L: b, c ∈ S ⇒ b ∨ c, b ∧ c ∈ S.” (p.23)
Definition 21.
“The sublattice generated by a family of properties is the intersection of all the
sublattices containing this family.” (p.23)
1
then b∧c = b and b∨c = c. If c! ≤ b! then b! ∧c! = c!
and b! ∨ c! = b! . By the two weak modularity rela-
tions, b ∧ c! = b ∧ b! ∧ (b ∨ c! ) = 0 ∧ (b ∨ c! ) = 0
b' c d' and b! ∨ c = [c ∧ (c! ∨ b)]! ∨ c = c! ∨ (c!! ∧ b! ) ∨ c =
c! ∨ (c ∧ b! ) ∨ c = 1 ∨ (c ∧ b! ) = 1 holds. This then
results in the sublattice formed by the set of eight
b
c' d elements: 0, b, b! ∧c, c, c! , b∨c! , b! , 1, which is distribu-
tive in Fig 1 with b ∨ c! = d! and b! ∧ c = d. Note, in
[Kal83] this kind of diagram is given as an example
0
of a distributive lattice. !
Fig 1: Hasse diagram of a distributive
sublattice
This theorem is only a small part of Th.2 in [Kal83,
p.22], due to Foulis, Holland and Nakamura. Given the definition of compatible
properties and Th.7 we reach the following proposition,
Proposition 14.
Given two properties b, c in L, complete and orthomodular, if b ≤ c then b ↔ c
In [JauPir63], proposition 14 is proposed as a replacement axiom for the dis-
tributivity law for quantum systems. Focusing further on certain properties of
the compatibility relation, we see that while we presuppose an orthomodular lat-
tice L when b ↔ c, then also b ↔ c! , b ↔ b! and b ↔ b, [Pir76]. More precisely in
that case any two properties of c, b, c! , b! commute [BelCas84, Th.12.1.4., p.126].
This indicates that the compatibility relation ↔ is reflexive and symmetric13
but, as pointed out in [Pir77], not transitive. In [Pir76], several theorems con-
cerning compatible properties are proved, from which we take the one which
will be of use later on,
Theorem 8.
“In a CROC (complete, orthomodular lattice), if b ↔ ci , ∀i, then b ↔ ∨i ci and
b ↔ ∧i ci ” (provided the index set is not empty). (p.28)
Proof
We refer for the proof to [Pir76, p.28], because it is based on theorems (2.21)
and (2.15) in [Pir76], which we did not treat explicitly in this thesis. !
What we can conclude from this section is that in the case that properties
are represented by projection operators in a Hilbert Space, the compatibility
relation is equivalent with the property that the projections commute with one
another [JauPir63]. Contrary to quantum systems, for classical systems any pair
of properties is compatible [JauPir69]. As is mentioned in [BelCas84], the fact
that we talk about pairs of compatible properties in this case is only essential
when we are working within the weaker structure of an orthomodular poset. For
more details on this matter we refer to [BelCas84, p.127]. Note, this section on
compatible properties will be useful when we treat ideal measurements in the
next chapter.
13 See also Th.2 in [Kal83, p.22].
Chapter 5
63
CHAPTER 5. PROBABILITIES AND IDEAL MEASUREMENTS 64
We intend to follow [AerCoeSme99] quite closely in this section and will show
how quantum probabilities can be explained as being due to a lack of knowledge
of the interaction between the system and its measurement context. Contrary
to classical probabilities, they are not due to a lack of knowledge of the exact
state of the system. That quantum probabilities are of a different nature than
classical probabilities has been shown by the failure of hidden variable theo-
ries indicated by many theorems (e.g. von Neumann’s no-go theorem [VNeu32]
or the elaborations in [Bel64, KocSpe67, JauPir63]). The search for hidden
variables was a search for a (deterministic) classical underlying reality. But if
quantum probabilities are not due to a lack of knowledge of the precise state, are
they then ontologically or objectively present in reality itself, as many physicists
have thought? The answer we propose is no, and to defend our position we will
first analyze what classical probabilities actually are.
In fact, one can prove that for every possible preparation we could perform
when we want to check if a system has a well defined property, there are always
other properties about which we cannot be certain. Think back for instance
on the incompatible properties one encounters in the quantum case but not in
the classical case. When we use the standard quantum formalism to describe
quantum systems, one is almost forced to accept that there do exist probabilities
that do not merely describe a lack of knowledge of the exact situation of the
system, but are also a priori linked to the quantum world. What we have in
mind here is the so-called formula of Born, the standard statistical algorithm
for the quantum state description in case of discrete spectra:
P robΨ (ai : A) =|| Pai Ψ ||2 =| 9αi | Ψ: |2
In other words, the probability according to quantum state Ψ that an observ-
able A obtains the (eigen)value ai is given by the square of the magnitude of
the component of Ψ in the direction of eigenvector αi [Wal84].1 In many inter-
pretations of quantum mechanics, these probability assignments are the most
the theory can say in state Ψ about the system, and exactly this leads to the
position that quantum states are ontologically statistical since only statistical
predictions about the values of quantum observables can be made. We however
do not accept these interpretations since it would imply that the existence of a
system can be asserted only with a certain level of probability. For the remain-
der of this section we will explain in brief our position which does not rely on
objective probabilities.
from OQL. Ψ = Σi &αi | Ψ'αi since Ψ is the linear combination of its basisvectors αi . We
also know that || Ψ ||2 = Σi || Pai Ψ ||2 where Pai is a linear self-adjoint projector. Note that
we used the fact that the components of the vector Ψ in the given basis can be expressed by
means of projections, where &αi | Ψ'αi = Pai Ψ is a component represented by the projection
of Ψ into the subspace associated with ai . Further || Pai Ψ ||2 =|| &αi | Ψ'αi ||2 = &&αi | Ψ'αi |
&αi | Ψ'αi ' = &αi | Ψ'&αi | Ψ'&αi | αi ' = &αi | Ψ'&αi | Ψ' =| &αi | Ψ' |2 . Where the overlined
symbols represent adjoints of complex numbers and &. | .' stands for the inner product.
CHAPTER 5. PROBABILITIES AND IDEAL MEASUREMENTS 66
[Aer93, Aer98, AerCoeSme99]. This view comes down to the fact that during
a non-destructive measurement two different types of processes can take place,
creations and discoveries. Whereas discoveries are mostly associated with classi-
cal observation measurements and do not change the state of a system, creations
do change the state of a system. We will show that creations hold a special po-
sition with respect to quantum probabilities. Let us concentrate on an example
given in [Aer93, AerCoeSme99] which makes this view a bit clearer. It concerns
a survey to determine the opinions of a randomly selected group of persons on
nuclear energy. A few interviewers have to investigate who is ‘for’ and who
‘against’ nuclear energy. The situation can be analyzed as follows. Some mem-
bers of the survey group will already have a fixed opinion (for or against) on the
matter before the interviewer asks any questions, while other members will not
yet have made their mind up. By asking his questions the interviewer will dis-
cover the opinions of those members who had a fixed opinion in advance, and he
will create the situation in which those who were still doubting can give a for or
against answer. This act of creation can manifest itself as e.g. the presentation
of persuasive information by the interviewer on the pro’s and con’s of nuclear
energy. What happens here is that persons are pulled out of a “doubtful”-state
into a “for or against”-state and that this change of state — caused by the
act of creation — depends on the influence of the measurement context (of the
interviewer). When we are only dealing with a discovery, there is almost no
influence of the measurement context. Regarding the probabilities in this ex-
ample, when an interviewer first encounters a member of the survey group, he
does not know his/her opinion and can only make a probabilistic prediction.
We make the following probabilistic assumption characteristic of the interaction
between the interviewer and the doubtful person: the interviewer meeting such
a member of the survey has a 40 % chance of meeting a person who will be
for nuclear energy and a 60 % chance of meeting one against. Suppose that
after he has asked the question, of all the 1000 members he finds 400 for and
600 against. There are now several possible initial distributions which yield this
result, take for example: (i) in advance, 400 are for and 600 are against, (ii) in
advance, all 1000 have no definite opinion, (iii) n are for (we choose n between
0 and 400), 32 n are against, and 1000 − 52 n have no opinion. We want to stress
the fact that the result of the survey depends both on the initial distribution of
the members of the survey and on the probabilistic assumption. Let us analyze
this in more detail. We see that all example cases (i), (ii) and (iii) yield the
same result, but that this result emerged from situations each of a different
nature. In (i) there was no act of creation making this a classical situation. In
the other example cases, the result is partially due to our probabilistic assump-
tion. The probabilities attributed to the change of state of a doubtful person,
cannot be attributed to a lack of knowledge of a more complete specification
of the real state of that person, because a complete specification simply does
not exist. Probabilities about the states in which the other persons with a fixed
opinion will find themselves, can indeed be attributed to a lack of knowledge of
a more complete specification of the real states of those persons. We see that
the creation-discovery view gives a very natural explanation for the existence of
CHAPTER 5. PROBABILITIES AND IDEAL MEASUREMENTS 67
two kinds of probabilities, attributed on the one hand to the lack of knowledge
concerning the measurement process, and on the other to the lack of knowl-
edge of the state of the system. The first kind of probabilities are quantum
probabilities and the motivation for this statement is given below.
quantum probabilities. We see explicitly in our model that the probabilities are
due to a lack of knowledge about where precisely the elastic will break. We
took all possible breakingpoints of the elastic with respect to a specified result
into account. Suppose now that for each possible breakingpoint of the elastic,
there is a corresponding classical experiment. Think of a classical experiment
in the exact way as we presented our model, but replace the elastic with one
that has only one fixed breakingpoint. From this we conclude that when we
perform eu , what happens is that one classical experiment is being performed
but we do not know which one. We called it the hidden measurement, it is the
classical measurement which creates the final state of the system. The presence
of the lack of knowledge of the precise act of creation, lies at the heart of the
so-called indeterministic nature of quantum measurements. And this is exactly
one of the starting-points of a further dynamic elaboration of OQL — see the
example given in part IV, section 12.5. But indeed also in the standard theory
of OQL, C. Piron has concentrated on a special kind of measurements which
enable us to recover quantum probabilities and explain them within this non-
statistical approach. We will concentrate on these special measurements in the
next section.
any more. For this reason we will concentrate on the notion of a measurement
of the first kind as introduced by W. Pauli in [Pau58] as follows:
Clearly such a measurement does not destroy the system and if one carries out
the measurement a second time, one obtains the same result [Pir76]. Following
now the definition of a first kind measurement as given in [Pir76]:
Definition 22.
“A question β is called a measurement of the first kind if, every time the answer
is “yes”, one can state that the proposition (property) b defined by β is true
(actual) immediately after the measurement.” (p.68)
In other words, a question β is a measurement of the first kind if “yes” implies
the actuality of the property b = ζ(β) associated with the measurement. So in
case we perform such a measurement, and the positive response obtains, then
this is a measurement which;
1) did not destroy or severely damage the physical system, and
2) did not alter the resulting property in case this property was
already actual before we performed the experiment
Let us analyze these two points. 1) exactly states that in order to be able to
obtain an actual property afterwards, the system should be somehow preserved.
2), if the question β is a measurement of the first kind and if β is true before we
perform the experimental project, obviously the answer “yes” will obtain when
we perform the experiment. Since in this case, b = ζ(β) was actual before the
performance of the experiment, and if the answer “yes” obtains, it is still actual
immediately afterwards while the question β is true immediately afterwards.
It should be noted that only when we obtain a positive response for such a
measurement, it makes sense to characterize it as first kind. When we perform
a measurement which does not give a positive response, we deal with a situation
in which the measurement;
1) did destroy or severely damage the physical system or;
2) did not destroy or severely damage the physical system. In this
case β was not true before we performed the measurement. The
property β = ζ(β) was definitely not actual beforehand, since if it
were, β would have been true beforehand from which follows that
the result of performing β should have been positive. Given the fact
CHAPTER 5. PROBABILITIES AND IDEAL MEASUREMENTS 70
“An experiment of the first kind must permit one to conclude whe-
ther a certain proposition (property) is true (actual) or not imme-
diately after the experiment.” (p.84)
Indeed, it says literally that a certain property is actual or not afterwards, and
the latter case is only possible if we obtain “a negative response for a first
kind measurement”. Another ambiguity on this point in the Geneva School
literature arises when we think for example of such properties as “not being
breakable” or “not being mortal”. The experiments testing such properties
preserve the system every time the response is positive and destroy it otherwise.
These kinds of measurements are examples of what Piron calls filters. Given
the definition of a filter in section 5.2.4, we see that this are measurements
which are called “first kind” and can have a negative response. We believe that
it would certainly have been less confusing if in the Geneva School literature
one would have attached another name to such “first kind measurements with
a negative response”. Clearly a non-demolition measurement with a negative
response gives us some information about the associated property before the
measurement was performed. This being the case, this situation could have
been linked to the notion of a measurement of the second kind, following W.
Pauli in [Pau58]:
“On the other hand it can also happen that the system is changed
but in a controllable fashion by the measurement — even when, in
the state before the measurement, the quantity measured had with
certainty a definite value. In this method, the result of a repeated
measurement is not the same as that of the first measurement. But
still it may be that from the result of this measurement, an un-
ambiguous conclusion can be drawn regarding the quantity being
measured for the concerned system before the measurement. Such
measurements, we call the measurements of the second kind.” (p.75)
CHAPTER 5. PROBABILITIES AND IDEAL MEASUREMENTS 71
Now we wish to focus on measurements which are such that they perturb the sys-
tem as little as possible. In this sense we want to consider measurements which
leave actual properties as much as possible intact when they are compatible
with the property associated with the measurement. This means immediately
that measurements which are associated with compatible properties but which
cannot be performed together without disturbing each other’s result will not
count as the candidate measurements we have in mind. Still it will be possible
that a system is changed when the properties actual in the initial state of the
system are not compatible with the property associated with the experiment
we want to perform (see also [Pir77]). Besides this natural cause of a system’s
change due to incompatibilities, there is still another possible cause. A system
can also be perturbed due to imperfections of the measuring apparatus. One
could indeed calculate the exact influence of a particular measuring apparatus
on the physical system, but in order to allow a more general description inde-
pendent of any particular apparatus, we need to idealize the situation so that
the system is indeed as little perturbed as possible. In an ideal case we presume
that the apparatus itself does not have those imperfections which lead to per-
turbations and so we demand that all initial actual properties compatible with
the property associated with the measurement remain actual. Note that in real
life such an ideal case is only reachable by approximation. Let us now introduce
3 Of course to see this you have to know that we followed D. Aerts in [Aer81] and have taken
we leave the piece of wood floating on water until it is rotten. Since in such examples the
demolition-nature of β is very weak, at least compared to the demolition-nature of α, we deal
with β as a non-demolition measurement in the above exposition.
CHAPTER 5. PROBABILITIES AND IDEAL MEASUREMENTS 72
Note in case 3) that of course, even when β was not true beforehand, it can still
give a positive response when we actually perform the measurement.
Let us recapitulate: if β is ideal and the result is “yes”, we know that the state
afterwards consists of actual properties among which we find those properties
which are compatible with ζ(β) = b and actual in the state before we performed
β. In other words, all questions attached to the system which, when performed
together with that ideal measurement, were true beforehand, remain true after-
wards. If on top of this, the property associated with the ideal measurement
was already actual before, we know that the initial state is not perturbed at
all by performance of that measurement [Pir76]. Note that by means of the
compatibility-claim of an ideal measurement we indeed maximally limited the
perturbation of the system as much as it lies in our power to do so in case of a
positive response.
Definition 24.
An ideal measurement of the first kind of a property a is an experimental project
α which satisfies the following three conditions [Moo99, Pir98]:
(i) a = ζ(α) and a! = ζ(α∼ )
(ii) if the positive response is obtained then a is actual immediately after the
measurement;
(iii) if the positive response is obtained then the perturbation suffered by the
system is minimal.
In the case of a classical system, an ideal measurement of the first kind6 is
a classical question which does not change the system, at least when the re-
sponse is positive [Pir98]. This needs some further explanation. We know that
for a change to take place some potential property has to become actual and
vice versa. Since the system is classical it is precisely so that when a poten-
tial property becomes actual, its opposite should become potential. We know
that every pair of properties is compatible in this classical case and that all
actual properties remain actual after a positive response from the ideal, first
kind measurement. This is due to the fact that for a classical system and a
classical question α which in every state (also in the initial state) is true or
impossible, we see that when we perform α and obtain a positive response this
implies that ζ(α) = a was actual beforehand and when we obtain a negative
response this implies that a was potential and a! was actual beforehand. Since
when no actual property has changed into a potential one, no potential prop-
erty can change into an actual one and therefore any change of the system is
excluded. In the case of a quantum system, in general, the ideal measurement
of the first kind perturbs the system [Pir81]. This is due to the possibility of
incompatible properties. Now, in general if a property which is being measured
is not compatible with every property actual in the initial state, then the result
of the measurement is impossible to predict exactly, but for the case where we
deal with an ideal first kind measurement the final state is determined imme-
diately after the measurement if the response is “yes” [Pir72]. More precisely,
in the case we deal with a non-classical system, an ideal measurement of the
first kind allows us to calculate the perturbation suffered by the system in the
course of action [Pir76]. This is exactly what the following theorem in [Pir76]
is about (our brackets):
Theorem 9.
“If a question β is an ideal measurement of the first kind and if the answer is
“yes”, then the state of the system immediately after the experiment is (rep-
resented by) (p ∨ b! ) ∧ b, where p is (represents) the state before, and b is the
proposition (property) defined by β.” (p.68)
6 Because it is quite confusing, we want to note that in some of his work C. Piron calls an
ideal measurement of the first kind just an ideal measurement or sometimes just a measurement
(as done for instance in [Pir98]).
CHAPTER 5. PROBABILITIES AND IDEAL MEASUREMENTS 74
Proof
We extend the proof given in [Pir76, Pir77] with more details. First notice that
p ≤ (p ∨ b) ∧ (p ∨ b! ) because obviously p ≤ (p ∨ b) and p ≤ (p ∨ b! ). Now, we call
c = (p ∨ b) ∧ (p ∨ b! ) actual before the measurement since it is greater than or
equal to p which is the state property before the measurement. Moreover, c is
compatible with b. This follows from the theorems and definition in the section
on compatibility. In more detail, we prove this in 5 steps7 ,
(1) b ≤ p ∨ b ⇒ b ↔ (p ∨ b) (proposition 14)
(2) b ↔ (p ∨ b) ⇒ b ↔ b! ([BelCas84, Th.12.1.4., p.126])
(3) b! ≤ (p ∨ b! ) ⇒ b! ↔ (p ∨ b! ) (proposition 14)
(4) b! ↔ (p ∨ b! ) ⇒ b ↔ (p ∨ b! ) ([BelCas84, Th.12.1.4., p.126])
(5) b ↔ (p ∨ b! ) ∧ (p ∨ b) (theorem 8 applied on (1),(4))
Since β is ideal, any property compatible with b which is actual before the
measurement is again actual afterwards if the answer for β is “yes”, this holds for
c. We state that every other property x, actual beforehand and also compatible
with b, is greater or equal than c. In detail:
(1) x = x ∧ 1 and x = x ∨ 0
(2) Let 1 = b ∨ b! and 0 = b ∧ b! then x = x ∧ (b ∨ b! ) and x = x ∨ (b ∧ b! )
(3) Since b ↔ x, the sublattice generated by {b, b! , x, x! } is distributive.
(4a) Hence x = x ∧ (b ∨ b! ) = (x ∧ b) ∨ (x ∧ b! ) and8
(4b) x = x ∨ (b ∧ b! ) = (x ∨ b) ∧ (x ∨ b! )
(5) With (4b) and p ≤ x, we have:
c = (p ∨ b) ∧ (p ∨ b! ) ≤ (x ∨ b) ∧ (x ∨ b! ) = x
We know that b has to be immediately actual if the performance of β gives
“yes”, since this is a measurement of the first kind. We also know that c is
actual afterwards if the answer is “yes” and in that case, every property greater
then c ∧ b is actual afterwards. Since c ∧ b = (p ∨ b) ∧ (p ∨ b! ) ∧ b and because
obviously (p ∨ b) ∧ b = b, we have (p ∨ b) ∧ (p ∨ b! ) ∧ b = (p ∨ b! ) ∧ b as the strongest
actual property afterwards, it is the final state because the covering law points
out that this is an atom of the property lattice. This completes the proof.
patible in a complete orthocomplemented lattice this relation holds between them and vice
versa.
CHAPTER 5. PROBABILITIES AND IDEAL MEASUREMENTS 75
Coming back to a case we already dealt with before, suppose b, the property
associated with the ideal, first kind measurement β was already actual in the
initial state. Here we see that the initial state will not have been changed by
performing the associated ideal, first kind measurement because: b ∧ (p ∨ b! ) = p
iff p ≤ b (see [Pir76, p.69]). It is obvious that when p represents the initial
state, and p ≤ b then p ≤ (p ∨ b) ∧ (p ∨ b! ) implies p ≤ b ∧ (p ∨ b! ). And because
b ∧ (p ∨ b! ) is an atom it is equal to p.
In the case b! was actual before we perform an ideal, first kind measurement
β, we know that β cannot give a positive response, because if it did, b = ζ(β)
as well as b! would be actual afterwards which is impossible ! Formally this
factually impossible situation comes down to the statement (p ∨ b! ) ∧ b = 0 iff
p ≤ b! , proved in [Pir76, p.69, 70].
The change of state which takes place due to an ideal measurement of the first
kind β can be seen as a map ϕb from the initial state into the final state:
p 1→ (p ∨ b! ) ∧ b. Analyzing the final state we see immediately that b = ζ(β)
is actual and that the initial state has gotten weaker because p ∨ b! is actual
afterwards. This weakening of the initial state is due to the (smallest possi-
ble) perturbation of the system and counts as a loss of information which we
cannot avoid in general when we perform a measurement of this kind. With
Piron’s representation theorem in mind, we see that a purely quantum (com-
plete, atomistic, orthomodular) property lattice has as a possible realization the
set of closed subspaces L(H) of a Hilbert space. As such we can now say with
[Pir76] that it are exactly the projectors such as Pb , which orthogonally project
the vectors of H onto the subspace b, which induce the maps such as ϕb for
quantum systems. And indeed in [Pir76] the map ϕb is proven to have all the
formal properties of a projector. Furthermore, ϕb (p) = (p ∨ b! ) ∧ b is a special
kind of projector namely the so-called Sasaki projection. To the notion of Sasaki
projection we come back in part III and part IV. All this demonstrates that
the idea of ideal first kind measurements in the quantum case has proven its
importance.
will certainly possess a specific property afterwards. What we have in mind are
the following questions, [Pir76]:
Definition 25.
“A question β is called a filter if it is an ideal measurement of the first kind and
if, furthermore, the system is destroyed when the answer is “no”. ” (p.71)
Where the answer “no” in Piron’s terminology means that we do not obtain a
positive response. In our terminology above we already hinted at the fact that
we would like to call situations of demolitions those in which the measurement
gives no response at all. The notion of filter is well chosen, these measurements
filter out systems with a certain actual property (afterwards).
In the literature many critiques, which are in some cases rather clear misun-
derstandings, have been formulated on the OQL of the Geneva School. The
approach has been compared with those of for instance Randall and Foulis,
Ludwig and of Mackie. It is not our intention to go into any detail about
the differences and similarities of those different approaches with respect to
OQL, we instead refer to [CatNis91, CatMarNis94, CatNis93, CoeMooWil00,
FouPirRan83, RanFou83] for some relevant papers. In this chapter we want to
pay specific attention to those points of the theory which easily lead to misin-
terpretations. We will go over remarks and critiques aimed at OQL itself and
try to highlight our own critical points. We conclude the chapter with what we
believe to be some weaknesses of OQL as we have reviewed it above.
paper does not add anything interesting to the discussion. Indeed, we agree with R. Piziak
who reviewed [HadThi84] for the AMS’ MathSciNet and states: “It is right and proper for
any scientific work to be scrutinized and criticized according to its merits. Indeed, this is a
main impetus to progress. But if mathematics is to be used as a tool for criticism, let it be
used properly.”
78
CHAPTER 6. CRITIQUES AND CRITICAL POINTS 79
EssZou81].
First of all, we would like to object to the point of view that the realistic position
of the Geneva School would be some kind of physicalism or materialism — the
view in which reality is reducible to, or emergent from, the realm of physical
things and processes [Nii87].2 We strongly object to any position in which
scientific research in physics would be some kind of bible for practicing any
more empirically oriented science. Our realistic position also is not naive in
the sense that certified truth is believed to be easily accessible [Nii87]. As we
explained in the beginning, the metaphysical stance of the Geneva School lies
in the nature of critical realism. W.K. Essler and G. Zoubek are not the only
ones who thought Piron’s realistic position to be somewhat unclear. In [Bal81]
we see some confusion concerning realism and the EPR-principle of reality:
physical system. We do retain the criterion of reality as put forward in the EPR-
paper, though this does not exclude potential properties. There is a difference
between the lattice of all properties which characterizes the system in general
and the set of actual properties which describes the system in a certain state!
We think this may be the major misconception in [Bal81]. In his analysis of
Piron’s work, he also tries to see properties and questions as sets of physical
systems. This is obviously a picture which will not work, there is no need
to give another interpretation of properties than the one we gave above. A
property corresponds to an equivalence class of questions and a question is
operationally defined. Contrary to what is often thought, a question is not
something vague but rather very-well specified. There is another misconception
concerning properties and questions in [EssZou81], we quote:
“Since he (Piron) presupposes an absurd question 0 he has to assume
an absurd property 0 also which has then to be, according to his
distinction, either an actual or a potential property. But obviously
0 is neither an actual nor a potential property of an object. He
therefore has to distinguish the impossible properties both from the
actual and from the potential properties.” (p.412)
In the sense that potentiality of a property is defined within OQL, 0 is and will
always be a potential property. Especially because “0 is potential” only means:
“not actual”. If one thinks it would improve the theory it is of course possible
to introduce the notion of an impossible property. In that case an impossible
property could correspond to an equivalence class of impossible questions and
an impossible property then implies that the property is potential, but not vice
versa. Concerning the concept of a question, defined in [Pir81] as an experiment
leading to a well defined alternative of which the terms are “yes” or “no”, the
“yes” corresponding to the expected result and the “no” to all the others; there
also exists some confusion. It has maybe not been very clear in [Pir81] but as we
analyzed it above, using [Moo99], it concerns an experimental procedure which
is performable and is such that we can interpret the outcome of it in terms of
yes and no — exactly which outcome corresponds to “yes” is decided upon in
advance. There is even no need to specify which outcome of the experiment we
interpret as “no” since it will be everything not leading to the response “yes”. If
we concentrate on the possible confusion which arises here, we quote [EssZou81]:
“Other concepts of his approach are ambiguous and vague, espe-
cially his central concept of an “experiment leading to an alterna-
tive”. This concept of an experiment may be understood in at least
three different ways:
(1) An experiment is a device that tells the scientist how to handle
things by observing them and especially how to use an apparatus
for this purpose.
(2) An experiment is a single performance of such an experimental
device together with its result.
(3) An experiment is a basic (or minimal covering) law that general-
CHAPTER 6. CRITIQUES AND CRITICAL POINTS 81
Firstly note that there is more to a question than just performing an experi-
ment. In any case, we believe that for C. Piron an experiment is similar to a
procedure which we can perform. OQL clearly distinguishes an experiment and
its performance, this means indeed that something along the lines of (1) is the
case and not (2). However, W.K. Essler and G. Zoubek think that a reconstruc-
tion of Piron’s theory along the lines of (1) would make no sense of Piron’s use
of the terms “actual” and “potential” and therefore they exclude interpretation
(1) as a possibility. This seems to be a mistake, regarding questions we first of
all do not use the terminology of actuality and potentiality which is reserved for
properties. But if Essler and Zoubek have the terminology of “true” and “not
true” questions in mind, we still do not think there is a problem with accepting
interpretation (1). A question is true if in case we would perform it, “yes” is a
sure answer. Since this is a counterfactual definition, it is not in disaccordance
with disconnecting experiments from their performance. However Essler and
Zoubek are convinced that (3) is the case, questions are strict laws and there-
fore properties too are identified with a law or a theory [EssZou81, p.413-414].
One of their arguments concerns the counterfactual definition of a true question,
[EssZou81]:
“The statement “If we performed the question the answer “yes”
would be certain” cannot be justified by referring to some few ex-
periments and its results but only either to some law inductively
based on those results or gained by other laws or to some inductive
method expressing the belief in uniformity of the performances of
this experiment.” (p.414)
We agree with Essler and Zoubek if their statement points to the connection
between the lawlike character of utterances and the decidability of those coun-
terfactual stated utterances. Indeed, when some kind of question is always true
on some kind of physical systems, the question is connected to some law-like
behavior of the system. The same is the case for impossible questions, but not
true questions are then linked to utterances which are definitely not law-like.
Now, we still have to be careful with such generalizations here, for between
utterances with an explicit law-like character and those without there is still a
whole gray-scale left. We do not intend to enter the debate on the status of
laws in connection with counterfactuals within philosophy of science, let us just
mention instead that “experiments” are certainly not laws. We also do not be-
lieve one can so easily generalize and say that any question or property is a law.
Maybe Essler and Zoubek have a small point here but according to us it will
only involve true or impossible questions. Note that this also does not exclude
that experiments can be interpreted as disconnected from their performances.
“In his book [Pir76, p.20] he (Piron) tells us: “... one measures
an arbitrary one of the αi and attributes to Πi αi the answer thus
obtained ... By starting from the definitions it is easy to verify
the following rule: (Πi αi )∼ = Πi (αi∼ )”. The equation “(Πi αi )∼ =
Πi (αi∼ )” holds only in very few degenerate lattices, e.g. if the lattice
contains only the elements I and 0; in every other case we obtain:
“(Πi αi )∼ = Σi (αi∼ )”. So Piron obviously takes “Πi (αi∼ ) = Σi (αi∼ )”
to be true.” (p.415)
It is not evident what Essler and Zoubek are getting at here; perhaps they have
something like the De Morgan Laws in mind which of course are not applicable
here. But since we are talking about operational definitions, what is their
interpretation of the Σ of a family of questions? Especially taking into account
that it has to be operationally defined ! The product questions are operationally
precisely defined and therefore “(ΠJ αi )∼ = ΠJ (αi∼ )” does make sense. Essler
and Zoubek try to analyze their confusion and assume two possible cases. One
in which the members of the family on which a product is defined have to be
questions of a similar kind and one in which they can be different. In the first
case they use their previous law-argumentation which does not take them very
far and in the second case they say that “Piron’s rule of handling a product
is of course wrong since a product may be false even if the arbitrary chosen
member is true, namely, if one of the remaining members is false” [EssZou81,
p.415]. Here we believe they pretty well mix up some fundamental ideas. There
is a difference between obtaining the answer “yes” for a question and a “true”
question. A product question gives the result “yes” if we perform it, this means
exactly that the arbitrarily chosen question gives “yes” if we perform it. So,
even when all questions of a specific family are impossible except for one, the
product can give the result “yes”. It is something different to say that a product
question is true or not true. It is true when every question which we could pick
out of the specific family — relative to which the product is defined — is true.
And a product question is not true when not every question which we could pick
out of the specific family is true; it is impossible when every question which we
could pick is impossible. This leads us to conclude that Essler and Zoubek’s
analysis is simply wrong, a product question can never be “false” (impossible)
when the arbitrary chosen member is true. A product question is “not true” in
case the one chosen member is true and if at least one of the remaining members
is not true. The main point here is that we cannot claim in the latter case that
any “arbitrarily chosen” member is true. Truth and falseness — where false
means impossible — of a product question does not have anything to do with
the truth or falseness of one chosen member, but has to do with the status
of the arbitrarily chosen member, in other words, with every member in the
family. We agree that a product of questions is a confusing concept, being
constructed such that we can still talk about the product of two questions even
if their performances wouldn’t be possible together. Let us now concentrate on
(ΠJ αi )∼ = ΠJ (αi∼ ). A few possible situations indicate this equality:
1) Suppose we have a family of questions {αi ∈ Q | i ∈ J} with J given. If now
CHAPTER 6. CRITIQUES AND CRITICAL POINTS 83
every αi we could select, gives “yes” if we would perform it, then ΠJ αi is true
and (ΠJ αi )∼ impossible (or false). If every arbitrary question we could pick
out of {αi∼ | αi ∈ Q, i ∈ J} gives “no” if we would perform it, then we say that
ΠJ (αi∼ ) is impossible (or false).
2) If on the other hand not every αi we could select gives “yes”, ΠJ αi is not true
and so is (ΠJ αi )∼ not true. If every αi which we could choose gives “no”, ΠJ αi
is impossible and so (ΠJ αi )∼ is true. If any arbitrary αi∼ which we could have
selected gives “yes”, then we say that ΠJ (αi∼ ) is true. If not every arbitrary αi∼
which we could have selected gives “yes”, then we say that ΠJ (αi∼ ) is not true.
Essler and Zoubek clearly did not understand that the product question has an
operational definition and gives rise to a property called the meet. Contrary
to what is claimed in [EssZou81], in order to test the affirmation “a is actual
and b is actual” we do not have to “perform” two experiments on the same
system! We only have to consider the product question αΠβ with ζ(α) = a
and ζ(β) = b. On the other hand it is clear that the join, defined by means of
Birkhoff’s theorem, admits of no direct physical meaning [CoeMooSme(nd)a].
More explicitly,
∨J ai = ζ(Π{β ∈ Q | ∀i ∈ J : αi ≺ β}) with ζ(αi ) = ai .
When we look at the disjunction, the classical “or”, we see that a OR b is a
property of a physical system iff it is possible to construct a question that tests
this property [Aer81] — something which will in general not be the case.
Finally we want to say something with respect to ideal, first kind measurements.
In [Bal81, p.405] W. Balzer thinks that C. Piron in his definition of such kind
of measurements uses basic concepts which are new relative to the underlying
theory he starts with. Especially the notion of a question which is “true af-
terwards” seems to be confusing. Note that C. Piron in [Pir81] talked about
properties which were actual or not afterwards, but we could indeed consider
questions as well. There is however nothing special going on here; what seems
to be forgotten by Balzer is the fact that when we consider a question to be
true, not true or impossible, the physical system is always in a certain singular
realization. And it is of course always with respect to a certain state of the sys-
tem that we can say that the property associated with a question is (potential)
actual. As such, given a particular physical system we can ask whether a certain
question is true or not and it makes sense to ask the question again after the
state of the system has changed. Indeed it also makes sense to ask whether a
question is still true after the system has been submitted to an ideal first kind
measurement. So actually we are puzzled by W. Balzer’s remark where he says:
“...I think the basic vocabulary must be enlarged in a nontrivial way
and new axioms must be adduced. If this is not done I find it difficult
to prove assertions like the following (compare also [Pir76]):
If a system is in state q, if question α is compatible with p and if α is
CHAPTER 6. CRITIQUES AND CRITICAL POINTS 84
performed and yields the answer “yes” then the state of the system
after the experiment is (p ∨ a! ) ∧ a where a is the proposition defined
by α.” (p.406)
Since this “assertion” is quite badly formulated we can only ponder what Balzer
means. Is p actual in state q or how are p and q related? What kind of question
is α and what does it mean for questions to be compatible? We only talked
about compatible properties and reproduced the proof of the well formulated
theorem 4.3 in [Pir76, p.68]!
Finally we may affirm what we have insinuated all along: W.K. Essler and G.
Zoubek did not understand Piron’s work, [EssZou81]:
“As long as Piron does not answer these and similar questions in a
way satisfactory to logicians and especially as long as he does not
determine the logical structure of the elements of his class Q of all
questions his attempt remains as obscure as Heidegger’s approach to
a new ontology: he is able then to immunize his theory against any
critique by pretending that he has been misinterpreted, and no one
will then be able to prove the interpretation’s correctness.” (p.417)
6.2 Questions
First of all we will handle our own critical points of view concerning questions.
It is crucial to understand that questions — contrary to properties — do not
form a complete lattice. Firstly we only have a preorder relation on questions
and secondly, [Aer83a]:
“For a given family αi of questions the infimum (supremum) must
not be unique. However if two questions are both an infimum (supre-
mum) of the same family of questions, then they are equivalent”
(p.74)
And it is for that reason that we think the expression “the set of questions is a
complete lattice” used by D. Aerts in [Aer83a] causes too much confusion.
Concerning equivalent questions, in the sense equivalence has been defined ear-
lier, any question α is equivalent to the product with the trivial question αΠI.
The same holds for inverse questions: α∼ ≈ α∼ ΠI. The situation is however
completely different if we take the inverse questions of α and of αΠI, where
α∼ &≈ (αΠI)∼ . We see that contrary to α∼ , (αΠI)∼ can never give a posi-
tive answer since (αΠI)∼ = α∼ Π0 is equivalent to the absurd question — cf.
proposition 9.
Another point which may cause some confusion concerns the differences between
true, not true and impossible questions. We already addressed the fact that a not
true question differs from an impossible question which in [CatNis91] is called
CHAPTER 6. CRITIQUES AND CRITICAL POINTS 85
“When the physical system has been prepared in such a way that
in the event of an experiment neither the result “yes” nor the result
“no” is certain, we shall say that the question is indeterminate.”
(p.1300)
In order to clear up the possible misunderstandings in relation to the now in-
troduced notion of indeterminate questions, which according to us may be cut
loose from the preparation procedures, we put forward the following relations:
Note that a question and its inverse can never both be true or both be impossible,
but can both be indeterminate. As shown in [Moo99], the product question of
the trivial and absurd questions IΠ0 and its inverse 0ΠI are both not true and
even more, both are indeterminate. We cannot be sure to always obtain “no”,
neither can we be sure to always obtain “yes”, should we perform those questions
because we are allowed to choose arbitrarily between the trivial and absurd
questions. Note also that a question and its inverse are always performable
together, the experimental procedures are for both questions exactly the same,
the only thing which differs between them is the rule interpreting the result in
terms of “yes” and “no”.
6.3 Properties
Concerning properties, there are also some critical points to keep in mind. To
every question on a particular physical system there corresponds a property of
the physical system. By means of a question on a particular physical system,
there is no way of ever claiming that the associated property does not pertain to
that particular physical system. What we mean is that a property, associated
CHAPTER 6. CRITIQUES AND CRITICAL POINTS 86
a is actual ⇔ a! is potential
a! is actual ⇔ a is potential.
In general, for the non-classical and classical cases with a and a! given, we can
only claim the following due to the fact that: a ∧ a! = 0,
a is actual ⇒ a! is potential
a! is actual ⇒ a is potential
We see here that when a is actual, a! has to be potential because if it were
actual then a ∧ a! would also be actual meaning that 0 would be actual, which
leads to a contradiction. For non classical systems we cannot say that α “or”
α∼ is true in every state and therefore, when a! is not actual, it does not imply
anything with respect to a.
Although we limited our exposition of OQL to the overall case of one physical
system described by one property lattice, it should be noted that the same con-
cept of a physical system has in the OQL-literature also been applied to the cases
in which a system consists of subsystems. We believe however that as such the
concept of a physical system together with the related conceptions of separation,
forms an actual limitation of static OQL. As noted in [CoeMooWil00]:
CHAPTER 6. CRITIQUES AND CRITICAL POINTS 87
3 “Separated means that any experiment performed on one system does not alter the other
system (does not change its state)” [Isc(nd)].
Part III
Is Operational Quantum
Logic a Logic?
88
Chapter 7
“Operational Quantum
Logic” and Logic
First of all we want to make explicit that from now on we consider OQL as
satisfying only axioms 1,2,3, unless explicitly stated otherwise. In the tradition
of the Geneva School, J.M. Jauch and C. Piron have repeatedly insisted on not
calling OQL or a property lattice a “logic”. In their opinion, if logic is only
occupied with the study of valid inferences there is no room for any operational
content as explicitly present in OQL. On top of that, they take the well known
problems with implication as an overwhelming reason for why we should not
characterize OQL as a logic [JauPir70]. We believe however that these problems
are not longer a burden to using the label of “logic”, because even besides SQL
— the algebraic structure of closed subspaces in a Hilbert space — also abstract
ortho(modular)logic deals with the same implication problems and is nowadays
conceived as a “real logic” [Dal86, Dal81a]. We quote M.L. Dalla Chiara in
[Dal86]:
“today it is very clear that quantum logic is a “real logic”, at least
in the sense that it possesses a logical status completely similar to
that of other well-accepted logics, which are weaker than classical
logic (like for instance, intuitionistic logic).” (p.427)
It is in this sense that we believe that OQL can give rise to a logical system
which is then of course abstracted from its operational dimension. What we
have in mind is the abstract system of propositions connected to the properties
in a property lattice. Note however that this logical system emerges from OQL
though is in itself quite different from it. To deepen what we mean by the logic
of OQL we will concentrate in the next paragraph on the question “What is
logic?”.
89
CHAPTER 7. “OPERATIONAL QUANTUM LOGIC” AND LOGIC 90
state it in these words, but he recognized “the difficulty of getting any univer-
sally acceptable statement of the general principles that all logicians somehow
seem to recognize” [Put71, p.9]. According to him our approach of focusing on
the definition(s) of logic, would probably not be the right one, since he states
in [Put71]:
“The various extant definitions of “logic” manage to combine circu-
larity with inexactness in one way or another.” (p.3)
This might have been the case in 1971, but today we would not support this
utterance, quite on the contrary, there exist some good definitions that approach
the subject clearly. H. Putnam’s own approach to the question “what is logic?”
consisted of looking at logic itself. Although this can be a good approach,
we doubt it answers the question well enough without taking into account the
whole historical development of logic, which partly means reproducing the work
of e.g. W. Kneale and M. Kneale in [KneKne84].1 Among the several possible
definitions now available, we highlight J. van Benthem’s, [VBen95]:
“Logic is the study of reasoning — encompassing many aspects of
cognition and of rational thinking — both in the sense of description
of reasoning practice and the sense of design of reasoning systems.
And as such, it is a discipline in its own right, without being a
handmaiden to any special employer.” (p.271)
The italics are added to van Benthem’s statement as has been suggested by I.
Németi and H. Andréka in [NemAnd94, p.393]. We follow Németi and Andréka
[NemAnd94] in their opinion that the old definitions screening logic as the sci-
ence or study of valid inferences2 are too narrow to cover the subject matter of
logic nowadays. And as the above definition points out, we cite [NemAnd94]:
“So, there is a subject matter of logic which is “out there in the
world” and, roughly speaking, this subject matter is the phenomenon
of reasoning, cognition and rational thinking (or at least many as-
pects of these).” (p.393)
In this respect Németi and Andréka attribute to logical studies a greater di-
versity, besides the central topic of inference and consequence, logic also deals
with other aspects and mechanisms of cognition and reasoning. The way a logi-
cian approaches the subject matter of logic is by mathematically modeling the
phenomena of reasoning, cognition and rational thinking [NemAnd94]. As such
we are convinced that very specific aspects of the subject matter of logic can
be modeled mathematically in such a way that the result is what we called the
logical system of OQL. To be a bit more explicit, we first have to highlight the
1 This is not a critic to the address of H. Putnam in [Put71] because 1) he never had the
aim in the mentioned essay to give a full survey of opinions and 2) his answer to “what logic
is” existed only in pointing out that “there is a body of “permanent doctrine” in logic; but it
just does not carry one very far, at least when it comes to getting an exact and universally
acceptable statement of the general principles.” [Put71, p.7].
2 “Validity of inferences” being the (semantical) study of consequence relations.
CHAPTER 7. “OPERATIONAL QUANTUM LOGIC” AND LOGIC 91
We will now work out the given scheme for knowledge acquisition for the struc-
ture of physical properties :
0) The objects of thought are the properties of a physical system
3 J. Barwise and E. Hammer defend a modest position towards their thesis in the sense that
they do not claim that it will encompass “everything” called a logical system. They wanted
to provide “a useful way to understand work on logical systems”.
CHAPTER 7. “OPERATIONAL QUANTUM LOGIC” AND LOGIC 92
4 In that sense 0 and 1 will be the only classical propositions/properties we will consider.
Chapter 8
Natural Deduction
Among all possible ways of axiomatizing logical systems, the most popular meth-
ods without doubt are the Gentzen style sequent calculi and natural deduction
systems. In the history and development of ortho(modular)logic we encounter
several kinds of axiomatizations with/without an implication connective. For
natural deduction systems and Gentzen style sequent calculi we refer to the work
done in [CutGib82, Dal86, DalGiu(nd), Gib85, Gib87, Gol74, Nis80, Nis94a,
Nis94b, Tak95, Tam88]. Note that the paper of R.I. Goldblatt appeared before
the others and is seen as a milestone in this development however we want to
note that R.I. Goldblatt himself used the notion of binary logic by which he
conceived logic not as a set of well-formed formulas (wff’s) but as a collection
of ordered pairs of wff’s as indicator for which wff’s are inferable from which
others. In the tradition of Hilbert-style axiomatizations we may place the im-
plication algebra of P.D. Finch and R. Piziak [Piz74] and further in this line
also the work of I.D. Clark in [Cla73] who uses negation and implication as his
basic connectives.
93
CHAPTER 8. NATURAL DEDUCTION 94
There are several ways to deal with so-called infinitary operations or infinite
long conjunctions and disjunctions on a syntactical level. First note the fact
that for classical first-order logic with one-place predicates on a finite domain,
every formula is equivalent to a formula without quantifiers in the following way
[Gri85, p.322]:
(∀x)A(x) = A(a1 ) ∧ ... ∧ A(an );
(∃x)A(x) = A(a1 ) ∨ ... ∨ A(an );
Immediately the idea presents itself to consider an infinite domain of range for
predicates. This then implies crossing the edges of first-order logic which is only
occupied with the study of finite expressions. Indeed the idea of defining the
universal and existential quantifiers in terms of infinite conjunctions and dis-
junctions respectively, can be traced back at least to Charles Saunders Peirce
[Bar81]. This onset of what is now called infinitary logic which began in the
late nineteenth century, did not last very long and fell into disrepute for a while
due to the logic community which was heavily influenced by Hilbert’s Program.
As J. Barwise mentioned it in [Bar81], Gödels incompleteness theorem of 1931
changed things considerably. As is well known, his second incompleteness theo-
rem points out that Hilbert’s idea of searching for a consistency-proof for large
axiomatic systems — such as arithmetic — by finitistic and meta-mathematical
means was quite unprobable.1 Gödel first showed the limitation of the axiomatic
method in the sense that an axiomatic system large enough to encompass arith-
metic is, assuming its consistency, fundamentally incomplete and secondly that
its consistency cannot be proved by means of meta-mathematical propositions
which can be mapped into mathematical ones of this system. More explicitly,
Gödels incompleteness theorem produced a tautology expressible in finite terms
though not finitely provable, namely “..., a statement of the form (∀x)A(x)
about natural numbers such that each of A(0), A(1), A(2), ... is provable from
the first-order axioms about natural numbers, but such that there is no fi-
nite proof of (∀x)A(x) from the same axioms.” [Bar81, p.99]. This indicates
that Gödel actually said that arithmetic is ω-inconsistent, which means that if
1 We want to mention that although Hilbert’s program is limited, it is successful for con-
sistency proofs of arithmetical systems that allow for instance only the addition of cardinal
numbers (but not the multiplication) [NagNew58, p.52].
CHAPTER 8. NATURAL DEDUCTION 95
(∃x)A(x) is provable, also every member of the infinite set of formulas ∼ A(0),
∼ A(1), ∼ A(2), etc. is provable — to derive inconsistency also (∀x) ∼ A(x)
needs to be provable [NagNew58, p.106]. Eventually, we may say that after
1931 infinitary logic very slowly found its way through the logical landscape
with some help in the fifties by A. Tarski’s school of model theory in Berkeley
[Bar81]. As such, we now in the literature encounter several kinds of infini-
tary logics, introduced as extensions of first-order logic which allow infinite long
conjunctions, disjunctions and/or infinite quantifier sequences.
First of all, one should notice that the meet or conjunction obtained its mean-
ing through the operationally defined product question:
! ∧i ai = ζ(Πi αi ) with
ζ(αi ) = ai where we will now also use the notation A for a property which is
actual iff each a ∈ A is actual. This takes us to the definition of the provability
relation for sequents — or the syntactical consequence relation:
Definition 27. !
The provability relation %⊆ P (+) × + is defined by A % b ⇔ A ≤ b, where A % b
means that b is derivable from A or that the sequent A → b is provable. We say
that A % b iff there is a (possible infinite) derivation using the ls-OQL inference
rules such that A → b is the last derived sequent.
CHAPTER 8. NATURAL DEDUCTION 97
The following rules emerge immediately from OQL and have the following form:
A→a1 ...A→an
A→a where the top sequents are the premises from which the bottom
sequent follows as a conclusion. We will write A, a → b instead of A ∪ {a} → b
and a → b instead of {a} → b. In cases where confusion is possible, we will
explicitly use ai for the set {ai | i ∈ J}:
A→a A! ,a→b
(as) A,a→a (sub) A,A! →b
∀i∈J:(A→ai )
(∧E, ∀j ∈ J) A→∧J ai
A→aj (∧I) A→∧J ai
A,ai →b
(w) A→a
A,A! →a (r) A,∧J ai →b
(∼∼ E) A→∼∼a
A→a (∼∼ I) A→a
A→∼∼a
(∼ I) a→b∧∼b
→∼a (tp) a→b
∼b→∼a
The proofs given in the following sections contain sequents of the form A →
which are notational abbreviations of A → e∧ ∼ e. As we already mentioned in
the first chapter, the quantum logical join differs from the classical disjunction
with, as a consequence, the partial failure of the distributivity of meet and join.
The join of two states encompasses also the possibly generated superposition
states while the classical disjunction is nothing more than the settheoretical
union — for an example in Hilbert space see for instance [Gib87, p.94-95]. Note
that this explains why we limit the use of contexts in the rules (∼ I) and (∨E) —
classically those rules have also been called reductio ad absurdum and dilemma
respectively. More specifically for (∨E) we may say that “what follows from a
together with some true statement c, and also follows from b together with c,
need not follow from “a or b” together with c” [Dum78, p.276].
For this example we can draw the following diagram (see Fig 3) where we only
1
consider some properties explicitly: a cor-
responds to αφ , b to αφ! , a! to αφ + π\2
and b! to αφ! + π\2. Clearly this lattice for
a photon is not boolean, more specifically:
a b a' b'
a∧(b∨a! ) = a∧1 = a while (a∧b)∨(a∧a! ) =
0∨0 = 0 (see also [JauPir70]). Note that for
such non-boolean lattices we have (a ∧ b) ∨
0 (a ∧ a! ) ≤ a ∧ (b ∨ a! ) while also the converse
Fig 3: The lattice for a photon a ∧ (b ∨ a! ) ≤ (a ∧ b) ∨ (a ∧ a! ) holds when we
consider boolean property lattices. Return-
ing to ls-OQL, we can now say that indeed a proof of a∧(b∨c) → (a∧b)∨(a∧c)
can be established when we use the classical (∨E) with unrestricted contexts
(the first of following proofs can also be found in [Gib87] where it is given in
the context of general quantum logic and with a different style of notation):
(as)
a, b → b a, c → c(as)
(as) (r) (r)
a, b → a a∧b→b a, c → a(as) a∧c→c
(r) (∨I) (c) (∨I)
a∧b→a a∧b→b∨c a∧c→a a∧c→b∨c
(as) (∧I) (∧I)
(a ∧ b) ∨ (a ∧ c) → a∧b→ a∧c→
(a ∧ b) ∨ (a ∧ c) a ∧ (b ∨ c) a ∧ (b ∨ c)
(∨E)
(a ∧ b) ∨ (a ∧ c) → a ∧ (b ∨ c)
(as) (as)
a, b → a a, b → b a, c → a(as) a, c → c(as)
(∧I) (∧I)
a, b → a ∧ b a, c → a ∧ c
(as) (∨I) (∨I)
a, (b ∨ c) → b ∨ c a, b → (a ∧ b) ∨ (a ∧ c) a, c → (a ∧ b) ∨ (a ∧ c)
(classical ∨E)
a, (b ∨ c) → (a ∧ b) ∨ (a ∧ c)
(r)
a ∧ (b ∨ c) → (a ∧ b) ∨ (a ∧ c)
CHAPTER 8. NATURAL DEDUCTION 99
We will now discuss why exactly we adopt the above inference rules as the most
primitive ones, and will show how other rules can be derived from them.
8.1 Assumption
The assumption rule (as) expresses that A, a % a is always
! derivable!since by the
operational definition underlying the meet we have (A ∪ {a}) = ( A) ∧ a ≤ a.
This rule further express the reflexivity of “%” in analogy with the reflexivity of
“≤” (in case A is empty).
8.2 Substitution
The substitution rule (sub) expresses
! the !transitivity of “%”! which!in OQL
corresponds
! to !
the fact that
! if! A ≤ a and ( A )∧a ≤ b then ( A)∧( A ) ≤ b
! !
— resembles one representation of the cut rule used in Gentzen style sequent
calculi for quantum logic. As we show, the (sub)-rule is derivable from the other
rules in ls-OQL in the case we allow a classical (∨E) with unrestricted contexts.
(prem)
A→a
(w) (as)
A, ∼ a → a A, ∼ a →∼ a
(∧I)
A, ∼ a → a∧ ∼ a
(ex f also) (prem)
A, ∼ a → b A! , a → b
(w) (w)
A! , A, ∼ a → b A! , A, a → b
(classical ∨E)
A! , A → b
Note that to apply the classical rule (∨E), we did not make A! , A → a∨ ∼ a
explicit because it is derivable through weakening and the excluded middle which
can be introduced from an empty set of premises.
If, however, we extend the left sides of the turnstile with contexts we see that
if A, a ∨ b % c then also A, a % c and A, b % c. The converse only holds when the
system consists of classical properties — see subsection 8.5.3 and the algebraical
proof given in [CoeMooSme(nd)b]. It is for precisely this reason that the context
in (∨E) has been restricted. So, it comes as no surprise that the classical rule
known as disjunctive syllogism is not derivable in ls-OQL since its derivation is
based on the classical (∨E)-rule. Even when a ∨ b is a true wff in ls-OQL, we
know that a as well as b can both be false, this implies ∼ a, a ∨ b &% b in ls-OQL.
A,a,b→B A,a∧b→B
A,a∧b→B A,a,b→B
A→B,a,b A→B,a∨b
A→B,a∨b A→B,a,b
A→B,a a,C→D
A,C→B,D
A,∧J ai →b A,ai →b
A,ai →b
(r) A,∧ J ai →b
The first of these rules is derivable within ls-OQL while the last one is an
a →b
explicitly introduced inference rule. Note also that ∧Jjai →b is derivable for any
j ∈ J within ls-OQL while (r) is not. We give the derivations:
(as)
∀i ∈ J : (ai → ai )
(∧I) (prem)
ai → ∧J ai A, ∧J ai → b
(sub)
A, ai → b
(prem, j∈J)
aj → b
(tp, j∈J)
∼ b →∼ aj
(∨I)
∼ b → ∨J ∼ ai
(as) (tp)
∧J ai → ∧J ai ∼ (∨J ∼ ai ) →∼∼ b
(DM 2) (∼∼E)
∧J ai →∼ (∨J ∼ ai ) ∼ (∨J ∼ ai ) → b
(sub)
∧J a i → b
CHAPTER 8. NATURAL DEDUCTION 102
We used (DM 2) which is a De Morgen rule, for more details of which we refer
to the following section. We see that “A, ∧J ai → b implies A, ai → b” ! can be
written as A, ai → ∧J ai for b = ∧J ai which in OQL corresponds to ( A) ∧
(∧ai ) ≤ ∧J ai . And
! for (r), we see that ! “A, ai → b implies A, ∧J ai → b”
corresponds to: “ (A ∪ ai ) ≤ b implies ( A) ∧ (∧J ai ) ≤ b” in OQL. Note that
with respect to the join we can derive within ls-OQL the following rule:
A,∨J ai →b
A,ai →b
ai → aj (as, j∈J)
(∨I) (prem)
ai → ∨J ai A, ∨J ai → b
(sub)
A, ai → b
8.5 Negation
The quantum logical negation, corresponding to orthocomplementation, differs
from the classical negation which corresponds to a settheoretical complement
(see also [BelHal82]). Within our operational approach we may say that the
non-classical feature of the quantum logical negation lies in the fact that when
a property is not actual this does not imply that the orthocomplement should
be actual. The above rules (∼∼ I), (∼∼ E) and (tp) then correspond exactly to
a!! = a and a ≤ b ⇒ b! ≤ a! in the algebraic definition of an orthocomplemen-
tation as we introduced it earlier. The existence of the topelement (maximum)
a ∨ a! = 1 and bottomelement (minimum) a ∧ a! = 0 in any complete lat-
tice corresponds in ls-OQL to the existence of the law of excluded middle and
non-contradiction.
DM1:
(as)
∀j ∈ J : (∼ ai →∼ aj )
(r)
∀j ∈ J : (∧J ∼ ai →∼ aj )
(tp)
∀j ∈ J : (∼∼ aj →∼ (∧J ∼ ai ))
(prem) (n)
A → ∨J ai ∀j ∈ J : (aj →∼ (∧J ∼ ai ))
(∨E)
A →∼ (∧J ∼ ai )
DM2:
(as)
∀j ∈ J : (ai → aj )
(r)
∀j ∈ J : (∧J ai → aj )
(as) (tp)
∨J ∼ ai → ∨J ∼ ai ∀j ∈ J : (∼ aj →∼ (∧J ai ))
(∨E)
∨J ∼ ai →∼ (∧J ai )
(tp)
∼∼ (∧J ai ) →∼ (∨J ∼ ai )
(n) (prem)
∧J ai →∼ (∨J ∼ ai ) A → ∧J a i
(sub)
A →∼ (∨J ∼ ai )
Note that DM3 and DM4 will be derived from the primitive rules and by using
DM1 and DM2:
DM3:
(prem)
A →∼ (∧J ∼ ai )
(DM 2)
A →∼ (∼ (∨J ∼∼ ai ))
(∼∼E)
A → ∨J ∼∼ ai
(∼∼E)
A → ∨J ai
DM4:
(prem)
A →∼ (∨J ∼ ai )
(DM 1)
A →∼ (∼ (∧J ∼∼ ai ))
(∼∼E)
A → ∧J ∼∼ ai
(∼∼E)
A → ∧J ai
Excluded Middle:
(as)
a∧ ∼ a → a∧ ∼ a
(∼I)
→∼ (a∧ ∼ a)
(DM 2)
→∼ (∼ (∼ a∨ ∼∼ a))
(∼∼E)
→∼ a ∨ a
Although the excluded middle holds, it cannot be given the classical semantical
meaning which states that either a is true or ∼ a is true [Dal81a]. Quoting
[BelCas84] we may say that “in a sense, the nonclassical features of the quan-
tum negation and the quantum disjunction compensate each other, giving rise,
formally, to the excluded-middle law”. The nonclassical features are exactly
those which we encounter on the semantical level: as we already mentioned, the
join of two formulas may be true even when none of its components are and
when a is false, this does not imply that ∼ a has necessarily to be true. For the
CHAPTER 8. NATURAL DEDUCTION 104
In most of the papers written on standard quantum logic one makes use of the
De Morgan rules to define the join, given the meet. We will show that within
an orthocomplemented logic indeed (∨E) and (∨I) are immediately derivable
by use of (∧E),(∧I) and the DM-rules.
(prem)
∀i ∈ J : (ai → b)
(tp)
∀i ∈ J : (∼ b →∼ ai )
(∧I)
∼ b → ∧J ∼ ai
(tp) (as)
∼ (∧J ∼ ai ) →∼∼ b ∨J ai → ∨J ai
(∼∼E) (DM 1)
∼ (∧J ∼ ai ) → b ∨J ai →∼ (∧J ∼ ai )
(prem) (sub)
A → ∨J ai ∨J a i → b
(sub)
A→b
(N C):
(as)
a∧ ∼ a → a∧ ∼ a
(∼I)
→∼ (a∧ ∼ a)
(prem) (prem)
a, b → d a, c → d
(tr) (tr)
a, b, ∼ d → a, c, ∼ d →
(classical ∼I) (classical ∼I)
a, ∼ d →∼ b a, ∼ d →∼ c
(∧I) (as)
a, ∼ d →∼ b∧ ∼ c b∨c→b∨c
(tr) (DM 3)
a, ∼ d, ∼ (∼ b∧ ∼ c) → b ∨ c →∼ (∼ b∧ ∼ c)
(sub)
a, ∼ d, b ∨ c →
(classical ∼I)
a, b ∨ c → d
it slightly differently within our natural deduction system. We will first give
some straightforward derivations which will be used in the main proof below:
subproof1:
(as) (as)
(∼ (a ∧ b) ∧ a), (∼ (a ∧ b) ∧ a),
(∼ a ∨ (a ∧ b)) → (∼ a ∨ (a ∧ b)) →
(∼ a ∨ (a ∧ b)) (∼ (a ∧ b) ∧ a)
(∧I) (N C, tp, n)
(∼ (a ∧ b) ∧ a), (∼ a ∨ (a ∧ b)) → (∼ (a ∧ b) ∧ a)∧
(∼ (a ∧ b) ∧ a) ∧ (∼ a ∨ (a ∧ b)) (∼ a ∨ (a ∧ b)) →
(sub)
(∼ (a ∧ b) ∧ a), (∼ a ∨ (a ∧ b)) →
subproof2:
(as) (as)
∼ (a ∧ b), a → a ∼ (a ∧ b), a →∼ (a ∧ b) ...
(subproof 1)
∼ (a ∧ b), a →∼ (a ∧ b) ∧ a
(∧I) (∼ (a ∧ b) ∧ a),
(∼ a ∨ (a ∧ b)) →
(sub)
∼ (a ∧ b), a, (∼ a ∨ (a ∧ b)) →
(r)
∼ (a ∧ b), a ∧ (∼ a ∨ (a ∧ b)) →
subproof3:
(as)
a, ∼ a∨ ∼ b →∼ a∨ ∼ b
(r) (as)
a ∧ (∼ a∨ ∼ b) →∼ a∨ ∼ b a∧b→a∧b
(tp) (DM 2)
∼ (∼ a∨ ∼ b) → a∧b→
∼ (a ∧ (∼ a∨ ∼ b)) ∼ (∼ a∨ ∼ b)
(as) (sub)
a, b → a a ∧ b →∼ (a ∧ (∼ a∨ ∼ b))
(r) (DM 3, DM 4)
a∧b→a a ∧ b →∼ a ∨ (a ∧ b)
(∧I)
a ∧ b → a ∧ (∼ a ∨ (a ∧ b))
proof
... ...
(subproof 3) (subproof 2)
a∧b→ ∼ (a ∧ b),
a ∧ (∼ a ∨ (a ∧ b)) a ∧ (∼ a ∨ (a ∧ b)) →
(OM ) (as, r)
a ∧ (∼ a ∨ (a ∧ b)) → a ∧ b a∧b→b
(sub)
a ∧ (∼ a ∨ (a ∧ b)) → b
Note that subproof 1 could also have been directly written down using the rule
(tr).
Semantical Realizations
In the first chapter of this thesis we mentioned that what nowadays is called
quantum logic is a kind of logic which has in a way abstracted itself from the
original structure of the closed subspaces of a Hilbert Space which represent
quantum-propositions. Such logics have often been characterized by their al-
gebraic semantical structure which is that of a not-necessarily complete or-
tho(modular) lattice. Given such a structure, it is possible to associate to the
formulas in a given formal language, a particular semantical realization or model.
This allows us then to characterize, given a specific model, when a formula is
true or a valid consequence of others. Besides this kind of algebraical semantical
realization, a kind of Kripke semantics was also proposed in the seventies, first
for minimal quantum logic (orthologic), by Dishkant [Dis72]. Both semantical
realizations for minimal quantum logic as well as for orthomodular quantum
logic have shown to be equivalent in the sense that they characterize the same
logic [Dal86, DalGiu(nd)]. It is now our intention to show that for the above
given syntactical language of ls-OQL, we have besides the straightforward alge-
braic semantical realization - based on a property lattice - also a kind of Kripke
semantical realization. Both realizations will be shown to be dual. We want to
point out as well that this duality emerges naturally from the duality of states
and properties mentioned above in 4.7. In the same way the state space Σ and
the property lattice L both describe the same physical system, a semantical
realization based on L and a realization based on Σ will both characterize the
same logic. Finally we will conclude this section with a note on the link between
the previously given syntax and the Kripke semantics. By means of the dual-
ity of the algebraic and Kripke semantics in ls-OQL, a minimal correspondence
proof between the syntactical consequence relation and Kripkian consequence
relation is then straightforward, based on the fact that the provability
! relation
A % a was defined in accordance with the algebraic relation A ≤ a.
107
CHAPTER 9. SEMANTICAL REALIZATIONS 108
Definition 28.
An algebraic realization for the logical language of ls-OQL is a pair A = 9L, v:
consisting of a (atomic, complete, orthocomplemented) property lattice L and
valuation-function v which associates with any formula a of the language an
element in L, and satisfies the following conditions:
v(a ∈ +) = a ∈ L; v(∼ a ∈ +) = a! ∈ L; v(∧J ai ∈ +) = ∧J ai ∈ L and
v(∨J ai ∈ +) = ∨J ai ∈ L
As we mentioned earlier, the use of a, b, c, ... in its context should make clear if
we deal with a wff in + or an element of L. As usual A will be referred to as a
model and the consequence relation will be defined as follows:
Definition 29.
A |= a iff for any A and any element c ∈ L holds that: if for all b ∈ A, c ≤ v(b)
then c ≤ v(a). We say that |= a iff v(a) = 1 for any A.
Definition 30.
A Kripke realization for the language of ls-OQL is a system K = 9Σ, R, Σ⊥⊥ , µ:
where Σ is the state space, R the accessibility relation, Σ⊥⊥ the set of biorthogo-
nal closed subsets of Σ and µ is the Cartan map. µ is a function which associates
with any formula a ∈ +, or equivalently with every property of L, an element of
Σ⊥⊥ and satisfies the following conditions:
µ(∼ a) = µ(a)⊥ ; µ(∧i ai ) = ∩i µ(ai )
The conditions which are satisfied for the Cartan map were proved in chapter
4, section 4.7. We did not explicitly state it before, but Σ⊥⊥ is a complete
ortholattice with inclusion as its order relation, infinitary intersection gives the
meet while the orthogonal is responsible for orthocomplementation (see also
[CoeMooSme(nd)a]). As such it is evident that we can define the orthocomple-
ment T ⊥ of T ∈ Σ⊥⊥ in an orthoframe as T ⊥ = {E | (∀η)(η ∈ T ⇒ η ⊥ E)} -
cf. the definition in chapter 4, subsection 4.4.2.
Also here K is seen as a model and |=⊆ Σ × +, the semantical consequence
relation, will be defined as follows:
Definition 31.
E |= a iff for any K: E ∈ µ(a).
A |= b iff for any K and state E, if ∀a ∈ A E |= a, then E |= b.
We write |= a iff µ(a) = Σ for any K.
Similarly to the algebraic realization truth of a formula corresponds to the
actuality of the property regardless the state of the system. As such we claim,
2 The notion of an orthoframe as introduced in the relational semantics given by R.I. Gold-
blatt in [Gol74] and further used in [Nis80], is slightly different in the sense that it is defined
by &Σ, ⊥'.
CHAPTER 9. SEMANTICAL REALIZATIONS 110
Theorem 11.
(i) E |= ∧J ai ⇔ ∀i ∈ J : E |= ai
(ii) E |= ∨J ai ⇐ ∃i ∈ J : E |= ai
(iii) E |=∼ a ⇔ (∀E ! ) (E ! RE ⇒ E ! &|= a)
Proof
(i) E |= ∧J ai iff E ∈ µ(∧J ai ) iff E ∈ ∩J µ(ai ) iff ∀i ∈ J : E ∈ µ(ai ) iff ∀i ∈ J :
E |= ai .
(ii) ∃i ∈ J : E |= ai iff ∃i ∈ J : E ∈ µ(ai ) iff E ∈ ∪J µ(ai ). Since it is proved in
[Aer81, p.27]3 that ∪J µ(ai ) ⊆ µ(∨J ai ) we obtain E ∈ µ(∨J ai ) or equivalently
E |= ∨J ai .
(iii) E |=∼ a iff E ∈ µ(∼ a) iff (∀E ! )(E ! RE ⇒ E ! &∈ µ(∼ a)⊥ ) iff (∀E ! )(E ! RE ⇒
E ! &∈ (µ(a)⊥⊥ = µ(a))) iff (∀E ! )(E ! RE ⇒ E ! &|= a). !
Note that the accessibility and orthogonality relations between states are also
important when we want to analyze the difference between two distinct states.
More explicitly we can prove the following theorem in OQL, [Moo99]:
Theorem 12.
“If E1 &= E2 then there exists E3 such that E1 ⊥ E3 and E2 RE3 .” (p.73)
Proof
We add details to the proof in [Moo99],
First we write E1 ⊥ E3 as pE3 ≤ p!E1 and E2 RE3 as pE3 &≤ p!E2 . Now we will
try to prove that if there is no E3 such that pE3 ≤ p!E1 and pE3 &≤ p!E2 then it
is not so that E1 &= E2 . The case in which we cannot find such an E3 is when
p!E1 ≤ p!E2 (since this would imply that pE3 ≤ p!E2 ). So given p!E1 ≤ p!E2 , we have
pE2 = p!!E2 ≤ p!!E1 = pE1 and also pE2 = pE1 since pE1 is an atom and pE2 &= 0. We
then obtained the converse of E1 &= E2 and this completes the proof. !
Based on the work done in [DalGiu(nd), Dal86], it is now clear for ls-OQL that
given any Kripke realization K in which |= a holds, we can construct an algebraic
realization in which |= a holds and vice versa. Furthermore, with respect to the
semantical consequence in a Kripke realization, A |= b, we can construct an
algebraic realization in which also A |= b holds and vice versa.
Theorem 13.
A |= b in any K iff A |= b in any A.
Proof
A |= b in K!iff (∀a ∈$A)(E |= a) ⇒ (E |= b) iff (∀a ∈ A)(E ∈ µ(a)) ⇒ (E ∈ µ(b))
iff (E ∈ (µ(
! A)) = µ(A)) ⇒ (E ∈ µ(b)) iff there is a Kripke realization
! K in
which µ( A) ⊆ µ(b) iff there is an algebraic realization A in which A ≤ b iff
A |= b in A. !
The relationship between syntax and semantics — between provability and va-
lidity — can be established by means of what is known as soundness and com-
pleteness. In particular the soundness and completeness theorems state: A % a
iff A |= a. As we already mentioned, due to the infinitary operations, complete-
ness causes difficulties for ls-OQL. If we would restrict ourselves to the countable
part of this logic, we obtain a similar calculus as the one for ortho(modular)
logic in [DalGiu(nd)] for which a completeness proof has been given. However,
since we defined our probability relation as exactly corresponding to the alge-
braic semantics, the following is stated as a theorem in [CoeMooSme(nd)b] but
will here just be a corollary,
Corollary 1.
A % b iff A |= b
Proof !
A % b iff A ≤ b iff A |= b. !
Chapter 10
Modalities
Given the above Kripkian style semantics where we interpreted possible states
as possible worlds, the question of a modal ls-OQL immediately arises. We
know that quantum logic can been given a so-called modal B-interpretation as
is shown in [Dal81b, Dal86, DalGiu(nd), Gol74]. The advantages of a modal
interpretation, as M.L. Dalla Chiara stresses in [Dal86], lie in the fact that it
“permits us to reinterpret with “classical eyes” the behavior and the meaning
of a system of non-classical logical constants”. The reason why a modal in-
terpretation of quantum logic remained for a long while in the marginal area
of quantum logic research has two probable origins: firstly, quantum logicians
have concentrated mostly on the standard algebraic semantics and even more
on the algebra itself and secondly, maybe we are not yet fully freed from the
suspicion surrounding modalities which we encounter in the history of the de-
velopment of logic. In the ensuing paragraphs we will elaborate on the existing
modal interpretation of quantum logic as well as on a modal ls-OQL which is
to be conceived not as a modal re-interpretation but as an extension of ls-OQL.
We then want to close this section with a slight elaboration on the mentioned
“historical” suspicion surrounding modalities.
113
CHAPTER 10. MODALITIES 114
and has any join preserving map between complete lattices a unique
meet preserving right adjoint.”
Let us return now to the modal operators. More explicitly — referring back
to chapter 4, section 4.8 where ZL stands for the set of classical properties of
L — in [CoeMooSme(nd)b], the modal operators ! and ♦ have obtained their
meaning in the following way:
Since each property lattice then has a ZL , it can be equipped with modal
operators ! and ♦ on L which can be given the usual meaning of necessity
and possibility respectively. This allows us to define the following syntactical
hierarchy3 :
Definition 33.
p % !a ⇔ p ≤ !(a) and p % ♦a ⇔ p ≤ ♦(a).
The semantic realization for the language of ls-OQL extended with modal oper-
ators such that ♦a =∼ ! ∼ a, is then an extension of the above given Kripkian-
style realization, where Σ⊥⊥ is also closed under the modal operation !. The
operation ! on sets of states is defined as follows, [DalGiu(nd)]:
Definition 34.
“For any A ⊆ Σ, !A := {E | ∀η : E &⊥ η ⇒ η ∈ A}”
Where we add that ♦A := (!A⊥ )⊥ and µ(!a) = !(µ(a)). As such we can
prove the following semantical hierarchy:
3 In[CoeMooSme(nd)b] the hierarchy is defined by focusing explicitly on the necessity
and possibility of provability instead of on the provability of the necessity or possibility of a
formula: p -#" a ⇔ p ≤ !(a) and p -' a ⇔ p ≤ ♦(a).
CHAPTER 10. MODALITIES 116
Theorem 14.
1) E1 |= !a ⇔ ∀E2 : E1 RE2 implies E2 |= a
2) E1 |= ♦a ⇔ ∃E2 : E1 RE2 and E2 &|=∼ a
Proof
1) E1 |= !a ⇔ E1 ∈ µ(!a) ⇔ ∀E2 : E1 RE2 implies E2 ∈ µ(a) ⇔ ∀E2 : E1 RE2
implies E2 |= a.
2) E1 |= ♦a ⇔ E1 |=∼ ! ∼ a ⇔ E1 &|= ! ∼ a ⇔ ∃E2 : E2 RE1 and E2 |& =∼ a. !
study of modal concepts was abandoned for some time in the history of modern
logic and had to wait for C.I. Lewis to pick up the lead again. Perhaps this is
not so surprising, considering a saying from medieval times “De modalibus non
gustabit asinus” which according to W. Kneale in [Kne62] means that modal
logic is difficult and unappetizing. Even when we look back at Aristotle’s work,
traditionally viewed as a milestone in theories on modal conception, there are
quite a few confusions and inconsistencies which gave “modal logic” not such a
great onset.
“What is, necessarily is, when it is; and what is not, necessarily
is not, when it is not. But not everything that is, necessarily is;
and not everything that is not, necessarily is not. For to say that
everything that is, is of necessity, when it is, is not the same as
saying unconditionally that it is of necessity.”
We should actually place this thought in the wide picture of chapter 9 in [De
Interpretatione] where Aristotle questions the principle of bivalence. Indeed,
future contingents form an exception to that rule. The well known example
Aristotle uses to support his point, is a future sea-battle, [Ari95, De Interpre-
tatione, 19a 30-, our brackets and italics]:
“... no one deliberates about the past, but about what is future and
contingent, while what is past is not capable of not having taken
place; ...”
Of course not only contingency is a mode of attribution — the notion is used
in [Pie80] — considered by Aristotle. First the distinctions are drawn between
possibility, contingency, impossibility and necessity before he then in [Ari95,
Interpretatione, chapter 12] starts the analyses of the relations between their
affirmations and denials. Besides the theory of modal statements, also in his the-
ory of modal syllogisms he distinguishes between premises which are apodeictic
(assert necessity), assertoric (assert actuality), or problematic (assert possibil-
ity). More explicitly, [Ari95, Prior Analytics, 29b 29-]:
“Since there is a difference according as something belongs, nec-
essarily belongs, or may belong (for many things belong, but not
necessarily, others neither necessarily nor indeed at all, but it is
possible for them to belong), it is clear that there will be different
deductions for each of these, and deductions with differently related
terms, one concluding from what is necessary, another from what is,
a third from what is possible”
The notion of deduction is here used for syllogism. As such it is now clear
that Aristotle included his theory of modality in his theory of arguments; the
notions of possibility and necessity have the generality that any statement on
any theme may be qualified by it while also the distinction between necessary
and non-necessary predication is basic in his theory of predicables [KneKne84,
p.82]. Of course we cannot avoid the fact that Aristotle’s theory of modal
statements is difficult and confusing — cf. [De Interpretatione, chapter 13] —
while his theory of modal syllogisms contains some explicit mistakes5 — cf.
[KneKne84, p.87-91] and [Luk51] where J. Lukasiewicz for that reason attaches
a greater importance to Aristotle’s modal logic of statements. As is explained
in [CoeMooSme(nd)b], some of the difficulties arise from the definition of the
notion of possibility:
“while it is probably inevitable that Aristotle should focus on a bi-
valent notion of the possible which is opposed to both necessity and
impossibility, the distinction between the necessary and the merely
factual being central to his metaphysics, strict adherence to a uni-
valent notion would undoubtedly have rendered his system more
elegant and less prone to confusion.”
Similarly, some difficulties may emerge from his use of modality sometimes as
de dicto and at other times as de re. Indeed, attempts were made to explain the
inconsistencies in Aristotle’s modal syllogistics by means of these two distinct
methods of interpreting modalities — we trace this distinction back at least
5 We quote W. Kneale and M. Kneale in [KneKne84, p.92],“... he develops his theory
of syllogism with two apodeictic premises, the only undoubtedly correct part of his modal
syllogistic.”.
CHAPTER 10. MODALITIES 119
to Abelard and later to Peter of Spain and St. Thomas Aquinas (see [Kne62]).
This distinction involves the scope of the modalities, either the modality applies
to the dictum (proposition itself) or to the thing to which is referred. J.M. Loux
in [Lou79] gives the following explanation of de dicto and de re modalities:
“Whereas de dicto modality attaches to propositions, an ascription
of de re modality specifies the modal status of an object’s exempli-
fication of an attribute”
Loux notes that the two kinds of modalities are not wholly unconnected (see
[Lou79, p.32]). This distinction has indeed been valuable — as is noted in
[Pie80], changing the scope of a modality changes its meaning — though perhaps
did not really serve to explain Aristotle’s inconsistencies because he used both
kinds of modalities in an inconsistent way (see [Pat95, p.8]). What may be
said of de dicto and de re modalities is that Frege made a distinction in the
same line where de dicto modalities correspond to the senses of signs and de re
modalities correspond to their references (see [Kne62]). Whatever the cause of
these inconsistencies, we agree with W. Kneale and M. Kneale in [KneKne84]
that we shouldn’t be too harsh on Aristotle who, with his modal theory, broke
completely fresh ground and we cannot ignore his great merits, one being the
first use of propositional variables6 in connection with modal logic.
We already mentioned that contemporary modal logic found its axiomatic form
in the beginning of the twentieth century with Lewis, though it took some
decades before any decent semantical model theory could be given for it. Lewis
is well known for his contribution to propositional modal logic, he introduced
a new strict implication — vs. the classical material implication —, though
it has to be noted that a definition of a strict implication in Lewis’ sense had
already been introduced by Sextus Empiricus 7 . It was when one picked up
the old Leibnizian strategy of considering more than one world, that our topic
was raised to a new perspective. The strategies of Hintikka, Kanger, Kripke
and some earlier attempts of Carnap merged into a possible world semantics for
modal logic and other intensional logics such as e.g. tense logic, epistemic logic,
deontic logic. If we look further into this possible world semantics for modal
logic, we enter into the philosophical debate on possible worlds. This debate
is of course connected to the debate on counterfactual discourse of which it is
claimed that by analyzing it we refer to some possible worlds namely those as
closely as possible to the actual one. Following Loux in [Lou79, p.33], we stress
that the difference with modal discourse lies in the fact that it involves quan-
tification over all possible worlds. We will not handle the details of the kinds
of discourses which appeal to possible worlds, in the remainder of this chapter
it is more interesting to draw some lines within the whole area of debate. Us-
ing the analysis of [Lou79], we can discern two main topics, namely transworld
6 As explained in [KneKne84, p.91], propositional variables are understood as “letters which
identity and possible nonactual objects. In the first case one concentrates on
the question of whether the inhabitants of the possible and actual worlds are
world-bound or transworld individuals. By following the line of transworld in-
dividuals and talking about identical individuals in different worlds, we end up
in the discussion on the properties of identity — especially the principle of in-
discernability and the property of transitivity. On the other hand, following
the path of world-bound individuals leads us to extreme essentialism especially
when we assume that things could not have been different in some world. A
possible escape from this essentialism is offered by David Lewis’ counterpart
theory in which individuals are world-bounded but have counterparts in other
worlds. By accepting this we obviously also escape the problems of the proper-
ties of identity. The objection here is that it maybe does not stick to our modal
intuitions.
Turning to the second topic, the ontology of nonactual objects. We distinguish
some main parties in the debate, the possibilists, the modal or nonmodal actu-
alists, conceptualists and combinatorialists. As opposed to the actualists, David
Lewis — an extreme possibilist — assigns to all possible objects (counterparts
or not) a real ontology. For an actualist, existing objects are present in the
actual world, as such they associate to possible worlds an actual existence from
their identification with — or construction from — actual objects. As Loux
puts it: “the contents of the actual world cannot be characterized in nonmodal
terms”. Those sceptical about modal actualism as Loux calls it, tried to pro-
pose a nonmodal approach. Conceptualism then gives the answer that possible
worlds may exist but only as a result of a mental construction — we think here
of Nicholas Rescher. Combinatorialists would answer that possible worlds are
just different combinations of the basic entities out of which the actual world is
constructed, whatever those entities might be. All the pro’s and con’s of these
approaches are in detail explained in [Lou79] where Loux himself favors a modal
actualist approach.
Chapter 11
We already mentioned that there are certain problems with the implication in
ls-OQL or quantum logic in general. We fail to define a philonian conditional
or material implication as it has been called by Whitehead and Russell and
is now generally used in classical proposition or predicate logic. Note that the
partial order relation ≤ of OQL is an implication relation and not an implication
operation or logical connective [Har79, JauPir70, Pir64]. Although not stated
in exactly those terms, this point was already noted by G. Birkhoff and J. von
Neumann in 1936, [BirVNeu36]:
“The following point should be mentioned in order to avoid misun-
derstanding: If a, b are two physical qualities, then a ∪ b, a ∩ b and a!
... are physical qualities too ... But a ⊂ b is not a physical quality;
it is a relation between physical qualities.”1 (p.25)
Indeed, the difference between an operation and a relation consists of the fact
that, contrary to what we would expect from an operation, if a and b are ele-
ments of an arbitrary property lattice; a ≤ b does not express a new property.
This of course depends on how we defined ≤, i.e. as a natural consequence of
the preorder on questions and on that level an experimental arrangement cor-
responding to a material implication is in general not so straightforward. Note
that the problem reduces itself to that of a disjunction by the definition of a
material implication. It is however intriguing to note that G. Emch and J.M.
Jauch argued in 1965 that a ≤ b is what they called a literally material impli-
cation — encoding an empirical fact — and not a logical implication expressing
a linguistic meaning [EmcJau65]. Bearing this in mind, the notion of material
used by Emch and Jauch, is according to us, perhaps more suited as label for
1 Note that in [BirVNeu36], physical qualities stand in a one-one relation with the closed
linear subspaces of a Hilbert space — which are, as we understand it, their mathematical rep-
resentatives — and can be interpreted as well as sets of equivalent experimental propositions.
121
CHAPTER 11. LOGICAL OPERATIONS VS. RELATIONS 122
the logical implication relation than for the logical implication operation.2 A
second reason might be that a nested implication relation in the sense of for
example a ≤ (b ≤ c) does not make much sense [Pir64]. Besides such difficulties
in OQL, which can maybe be avoided by not using the label of “logic” for prop-
erty lattices, naturally the problem of finding a material implication operation
also is present in quantum logic (=ortho-(modular)-logic) conceived as an ab-
stract logical theory. Since for the latter quantum logic, the problem was quite
intriguing and led to a search for adequate implication operations, we feel that
it is necessary to broaden our spectrum and analyze this problem in our context
of ls-OQL — the abstract study of operational quantum logic.
a ∧ x ≤ b ⇔ x ≤ a −→ b (11.1)
ditional.
3 The law of exportation holds in the reverse direction by applying modus ponens x ∧ a ≤
a ∧ (a −→ b) ≤ b.
4 First we want to note that in [Cur77] there is theorem stating that a classical implicative
lattice with zero and negation such that A! = A −→ 0 is a boolean algebra. Further, we find
it interesting to note that the dual of “b −→ a”, as used in any implicative lattice, is “a − b”,
where the operation “−” called subtraction is basic in the characterization of the so-called
subtractive lattices. The conditions a ≤ b ∨ (a − b) and a ≤ b ∨ c ⇒ a − c ≤ b, together then
characterize these subtractive lattices and merge into the condition a ≤ b ∨ c ⇔ a − c ≤ b by
means of a ≤ c ∨ (a − c) ≤ b ∨ c. In [Cur77] subtractive and implicative lattices taken together
are called Skolem lattices.
CHAPTER 11. LOGICAL OPERATIONS VS. RELATIONS 123
In other words, the implicative condition is seen as a special case of the adjunc-
tive implication condition. The most common example of an adjunctive im-
plication definable on an orthomodular lattice is the Sasaki arrow, also named
Sasaki hook and called quasi-implication by P. Mittelstaedt. Formally,
Proposition 18.
The Sasaki hook is defined by ϕ∗a : L+ → L : x+ 1→ a! ∨ (a ∧ x+ ) and is the right
adjoint of the so-called Sasaki projection defined by ϕa : L → L+ : x 1→ a∧(a! ∨x).
More explicitly, ϕa (x) ≤ x+ iff x ≤ ϕ∗a (x+ ).
Proof
For the proof of a ∧ (a! ∨ x) ≤ x+ ⇔ x ≤ a! ∨ (a ∧ x+ ) we concentrate first
on ⇐: since x ≤ a! ∨ (a ∧ x+ ) and we know that a! ≤ a! ∨ (a ∧ x+ ) we obtain
x ∨ a! ≤ a! ∨ (a ∧ x+ ) from which follows a ∧ (x ∨ a! ) ≤ a ∧ (a! ∨ (a ∧ x+ )). Since
a ∧ x+ ≤ a we obtain by orthomodularity that a ∧ x+ = a ∧ (a! ∨ (a ∧ x+ )) which
then implies that a ∧ (x ∨ a! ) ≤ a ∧ x+ ≤ x+ .
For ⇒ we start from a ∧ (a! ∨ x) ≤ x+ and a ∧ (a! ∨ x) ≤ a which implies that
a ∧ (a! ∨ x) ≤ x+ ∧ a from which we can obtain a! ∨ (a ∧ (a! ∨ x)) ≤ a! ∨ (x+ ∧ a).
Since a! ≤ x ∨ a! and by orthomodularity we obtain a! ∨ x = a! ∨ (a ∧ (a! ∨ x))
it follows x ≤ a! ∨ x ≤ a! ∨ (x+ ∧ a). !
To discover exactly how the Sasaki projections emerge in OQL we refer back to
chapter 5, subsection 5.2.3. The Sasaki hook however is an operation defined as
a two-variable lattice polynomial. A full proof is given in [Kal83] of the fact that
among the distinct two-variable polynomials which emerge from the elements of
the free orthomodular lattices generated by a two-element set, the Sasaki hook
is one of the following five polynomials:
p1 (a, b) = (a! ∧ b) ∨ (a! ∧ b! ) ∨ (a ∧ (a! ∨ b))
p2 (a, b) = (a! ∧ b) ∨ (a ∧ b) ∨ ((a! ∨ b) ∧ b! )
p3 (a, b) = a! ∨ (a ∧ b)
p4 (a, b) = b ∨ (a! ∧ b! )
p5 (a, b) = (a! ∧ b) ∨ (a ∧ b) ∨ (a! ∧ b! ),
which satisfy the following strengthened law of entailment:
p(a, b) = 1 if f a ≤ b. (11.2)
implication operation to that relation. While this is exactly what 11.2 stands
for, we are hesitant about calling this condition the “Birkhoff-von Neumann ’36
requirement” as is done in [Kal83]. More explicitly, for experimental proposi-
tions P and Q the following conditions are said to be equivalent in [BirVNeu36]:
“(a) The mathematical representative of P is a subset of the math-
ematical representative of Q.
(b) P implies Q — that is, whenever one can predict P with cer-
tainty, one can predict Q with certainty.
...” (p.5)
The classical material implication is not one of the above given five orthomodular
1
lattice polynomials satisfying condition 11.2, it
is well-known that there are counterexamples of
orthomodular lattices for which we have a! ∨ b =
1 but not a ≤ b, see for instance (Fig 4) in
a' a b b'
which we present the orthomodular lattice called
M02 or Chinese Lantern with atoms a, a! , b, b!
[Kal74, Kal83]. We want to stress as well that
0 the five polynomials we mention here are not
Fig 4: Hasse diagram of MO2 implication-connectives for the non-weakly mod-
ular case. A proof in [Dal86, DalGiu(nd)] shows
that for the case of the Sasaki hook, if a ≤ b then a! ∨(a∧b) = 1 iff a∧(a∨b)! = 0
which holds only in case of orthomodularity, the proofs for the other four poly-
nomials are similar. Coming back to the adjunctive implication condition, we
can say with [CoeMooSme(nd)b] that −→f is only correctly called an implica-
tion in the sense of condition 11.2, in case a −→f b = 1 iff fa (1) ≤ b which
translates into the condition fa (1) = a. Note that fa (1) = a is obviously sat-
isfied in the case of the Sasaki projection by ϕa (1) = a ∧ (a! ∨ 1) = a ∧ 1 = a.
CHAPTER 11. LOGICAL OPERATIONS VS. RELATIONS 126
The reason we paid more attention to the Sasaki hook then to the other four
polynomials, is because it approaches the material implication of classical logic
more closely than the other candidates. It is made explicit in [DalGiu(nd)] that
for all five polynomials (1 ≤ i ≤ 5) we have pi (a, b) = a −→ b in condition
11.1 iff the lattice is boolean. However the Sasaki hook also satisfies a weaker
condition namely,
p(a, b) = a! ∨ b if a ↔ b (11.4)
His main point in [Har75] is built up by means of his notion of a filter, where
an a-filter is to be interpreted as a yes-no question which gives a positive result
in case the system possesses the property a or, as Hardegree puts it, in case
the system passes the a-filter. These filters are used to express the conditional
property c(a, b) as follows:
“If s were to pass the a-filter, would it then (with certainty) pass the
b-filter? A system s in state E for which the answer to this question
is affirmative is said to have property c(a, b), ...” (p.416)8
symbolism.
CHAPTER 11. LOGICAL OPERATIONS VS. RELATIONS 127
“Thus if (and only if) a system s is in state E, which satisfies p3 (a, b),
then s has property c(a, b) to the effect that if s were to pass an a-
filter, then it would also pass a b-filter. Whether a given system has
this property usually depends on its state;...” (p.416)
And we go further with, [Har79]:
“In certain well specified circumstances, however, s will have prop-
erty c(a, b) regardless of its state. This is to say that no matter what
initial state s is prepared in, if s passes an a-filter, it is certain to
pass any immediately subsequent b-filter. This corresponds to the
customary operational characterization of the relation of implication
among binary experiments pertaining to system s; ...” (p.74)
We will now investigate exactly what these quotes mean in the context of OQL.
Referring back to part II, we saw that α ≺ β stands for the declaration “if α
is true then β is true”. Since the truth of a question is inherently linked to a
certain state of the system we can be more precise and say that α ≺ β means
the following:
It should be clear that when we mention here the notion of “transforming one
state into another” we mean this in a hypothetical sense, i.e. we make an
abstraction from the fact that actually one state can only be transformed into
another state when the latter is accessible (not orthogonal) to it. However if
we look at p3 (a, b) this kind of abstraction is not sustainable anymore. Indeed,
p3 (a, b) is valid in at least those states Ei which are such that they can actually
be transformed in a state E2 in which a is actual and b is actual. Otherwise
worded, the states in which p3 (a, b) is actual are at least those in which b is actual
after a positive response is obtained in case the system would be submitted to
an ideal first kind measurement α with ζ(α) = a. The idea of taking an ideal
first kind measurement comes from the use of the words “minimally different”
by Hardegree in the following, [Har79]:
“The basic intuition is fairly clear: in deciding the truth of a coun-
terfactual conditional (A > B) in a given situation x, we consider
(envisage) a particular situation y which makes A true and which is
in some relevant sense minimally different from x. The conditional
(A > B) is then true in situation x exactly if B is true in situation
y.” (p.68)9
9 Note that Hardegree works here with the algebra of subsets of a state space, where x and
y denote states or worlds and A and B denote elementary sentences. An elementary sentence
has the form “magnitude ... has value in (Borel set) ...”. (A > B) denotes the so-called
Stalnaker conditional which is: “the sentence — provided it exists — which is true at any
state x exactly if B is true at the nearest A-state to x.” [Har79, p.68].
CHAPTER 11. LOGICAL OPERATIONS VS. RELATIONS 128
Against the background of condition 11.2, we see that p3 (a, b) is valid in every
possible state of the system when a ≤ b. In other words, when a ≤ b then
µ(p3 (a, b)) = Σ and this is exactly the case Hardegree refers to when he says
that “in certain circumstances, s will have property c(a, b) regardless of its
state”. This however leads to a small problem when, given pE ≤ a ≤ b, E is
not accessible to every other state of the system. Unless we solve this problem,
it renders the given operational interpretation of p3 (a, b) inadequate. Of course
the problem of inaccessible states can also occur when a &≤ b. Note that in
the latter case it can still be so that p3 (a, b) is actual in certain states though
in general this will not be so in all possible states of a system. Let us now
concentrate on solving the problem of inaccessible states. Take the example of
p3 (a, b) in a complete orthomodular lattice, the system will not give a positive
response to an ideal first kind measurement of α with ζ(α) = a if a! is actual
in the initial state and, as we saw earlier in section 5.2.3, the state afterwards
will then be (p ∨ a! ) ∧ a = 0. But since it can well be that µ(a! ) ⊆ µ(p3 (a, b)),
Hardegree defines for the consistency of his counterfactual interpretation of the
Sasaki hook an absurd world (state) 0. He then treats the absurd world as the
nearest world in which sentence A is true if and only if there are no worlds
accessible in which sentence A is true. Following his definition in [Har75]:
“We also define the absurd world 0 to be the world which satisfies
every sentence, including contradictions.” (p.413)
Indeed if we see 0 as an absurd state in Hardegree’s sense, we have found a
way to avoid the problem which may arise due to inaccessible states. And as a
consequence we can say that for a complete orthomodular lattice: if a! is actual
in the initial state p, then p3 (a, .) is actual in p, or more generally, [Har79]:
“As an immediate consequence of these considerations, if a sentence
A is impossible relative to x, then x satisfies every conditional of
the form (A > B). A fortiori, if A is contravalid (true nowhere
except 0), then every conditional of the form (A > B) is valid (true
everywhere).” (p.69)
Furthermore, Hardegree gave the formal analysis in the lattice of closed sub-
spaces of a Hilbert space of what it means for p3 (a, b) to be valid in a state E1
[Har75, Har79]. While we will avoid the technicalities, we follow [DalGiu(nd),
Dal86] in the interpretation of E1 |= p3 (a, b):
We want to finish this chapter with a note on the analysis of L. Herman, E.L.
Marsden and R. Piziak in [HerMar75] which shows that contrary to the Sasaki
CHAPTER 11. LOGICAL OPERATIONS VS. RELATIONS 129
Dynamic Operational
Quantum Logic
130
Chapter 12
12.1 Inductions
Recall that a question or a definite experimental project can be regarded as
an experimental procedure with an evaluation-rule allowing us to interpret the
results. Moreover, we can distinguish among all experimental procedures per-
formable on a physical system, those which can lead to the destruction of the
system and those which cannot. It is now our aim to focus first on experimental
procedures independently from the questions to which they might pertain, i.e.
by abstracting over the evaluation. We will concentrate on the notion of an
induction, originally introduced in [AmiCoeStu98] though we refine it here as
follows,
Definition 37.
An induction e ∈ εs is a physical procedure that can be effectuated on a particular
physical system s. This procedure, when carried out, might change s, modify the
collection of its actual properties and its realized state1 or even destroy s. In
case an induction cannot lead to the destruction of s, we call the induction soft
and otherwise hard.
In [AmiCoeStu98] the notion of induction has been introduced as “an externally
imposed change of a particular physical system”. We find this characterization
a little too ambiguous since it does not refer explicitly to the notion of a physical
procedure.
131
CHAPTER 12. OPERATIONAL FOUNDATION & MATH. FRAME 132
true from in advance. We see here that for the case of soft inductions on which
we focus, β is an element of Q1 as well as of Q2 . It is important that only in
e
the case of soft inductions do we immediately obtain I1 # I2 with I1 ∈ Q1 and
I2 ∈ Q2 since when e cannot change the nature of the system we have I1 ≺ e.I2 .
Let us now focus on a specific induction denoted by 0 and called “freeze the
system”. An act of freezing the system implies that the system’s properties
remain as they were before this action was performed.4 In particular we obtain
the following proposition:
Proposition 20.
(1) 0.α ≈ α
%
(2) α # β ⇔ α ≺ 0.β ≈ β
%
In proposition 20.(2) we see that for # ⊆ (Q1 × Q2 ), where 0 leaves the system
intact, Q1 is equal to Q2 with respect to the truth of the questions.
Why we need Lop — which is L with a reversed order so that all meets become
joins and vice versa — will be explained below and we should also admit that it
is through a slight abuse of notation that the property ζ(e.α) equals to e.ζ(α).
Against the background of definition 38 we can say that
e.a is an actual property for a system in a certain realization if it
is sure that a would be an actual property of the system should we
perform induction e.
This explains that if e.a is actual it indeed “guarantees the actuality of a” with
respect to e, while if e.a is potential it does not. Also here we can say that “e.a is
actual” implies that “a is actual after performing e”. Again it is this expression
e
which forms the underlying idea of introducing a formal causal relation # with
respect to an induction e on properties. Here we can say the following:
e e
The relation # allows us to express for every e.a that e.a # a
holds, with the attached meaning that “the actuality of e.a, before
performing e guarantees the actuality of a after performing e”.
e
In the same way we can define # ⊆ (L1 × L2 ) as follows:
Definition 40.
e
a # b := the actuality of a before e induces the actuality of b after e.
This definition gives us the following proposition:
Proposition 23.
e
a # b ⇔ a ≤ e.b
Obviously we now obtain the same kind of propositions as we have stated before,
although we change the equivalences into equalities since we can imagine the
situation in which ζ(0.α) = ζ(α) = a:
Proposition 24.
(1) ∀a ∈ L : 0.a = a
%
(2) ∀a, b ∈ L : a # b iff a ≤ b
(3) ∀e1 , e2 ∈ ε, a ∈ L : (e1 &e2 ).a = e1 .(e2 .a)
(4) ∀e ∈ ε, ai ∈ L : e.(∧J ai ) = ∧J (e.ai )
(5) ∀ei ∈ ε, a ∈ L : (∨J ei ).a = ∧J (ei .a)
Proposition 24.(5) explains exactly why we used Lop earlier. Since we know
that to each equivalence class of questions there corresponds a property, we
obtain against the background of proposition 22.(2) the situation in which
ζ((∨J ei ).α) = (∨J ei ).a corresponds to ζ(ΠJ (ei .α)) = ∧J (ei .a). This implies
that “∀a, the map −.a : ε → L turns ε − joins5 into L-meets” [CoeMooStu00].
And here lies the inspiration for turning L-meets into L-joins by the introduc-
tion of Lop in order to obtain a formal preservation of the join rather than an
antitone map.
5 When ε is conceived as a set, the arbitrary choice of inductions corresponds to the join of
inductions.
CHAPTER 12. OPERATIONAL FOUNDATION & MATH. FRAME 137
Definition 41.
e
T # T ! := if T contains the realized state before e, T induces that T ! contains
the realized state after e.
And thus we put forward:
Proposition 25.
e
T # T ! ⇔ T ⊆ e.T !
Also here we obtain the same kind of propositions as we have stated before in
the case of inductions as actions on properties:
Proposition 26.
(1) ∀T ∈ P (ΣL ) : 0.T = T
%
(2) ∀T, T ! ∈ P (ΣL ) : T # T ! iff T ⊆ T !
(3) ∀e1 , e2 ∈ ε, T ∈ P (ΣL ) : (e1 &e2 ).T = e1 .(e2 .T )
(4) ∀e ∈ ε, Ti ∈ P (ΣL ) : e.(∩J Ti ) = ∩J (e.Ti )
(5) ∀ei ∈ ε, T ∈ P (ΣL ) : (∨J ei ).T = ∩J (ei .T )
CHAPTER 12. OPERATIONAL FOUNDATION & MATH. FRAME 138
where ẽ! : ΣL → P (ΣL ) : p 1→ ẽ! (p) is “the set of possible outcome-states when
performing the induction e on s in state p” [AmiCoeStu98, p.556]. Note that
ẽ(T ) is the union of the sets of possible outcome-states which we obtain when
we consider performing e on s in every possible initial state in T . The fact
that we may consider a “set” T of possible initial states will become relevant in
section 12.4.1. Further we should note that while considering soft inductions,
an induction preserves the system which implies that ∀T ∈ P (ΣL ) : ẽ(T ) = ∅ iff
T = ∅ [AmiCoeStu98].
2) State Propagation:
& e
ẽ : P (ΣL )1 → P (ΣL )2 : T1 1→ {T2 ∈ P (ΣL )2 | T1 # T2 }
From the way in which we conceived the maps expressing state causation and
state propagation follows that we obtain the following proposition:
Proposition 27.
e
ẽ(T ) ⊆ T ! ⇔ T # T ! ⇔ T ⊆ ẽ∗ (T ! )
This proposition points out that ẽ = ẽ∗ . More precisely we see that ẽ∗ preserves
intersections — see also 26.(4) — while ẽ preserves unions.
We will close this subsection with a note on the structure which arises by con-
sidering such state propagation and state causation maps. We introduce the no-
tation JL(P (ΣL )1 , P (ΣL )2 ) for the collection of all union-preserving maps from
P (ΣL )1 to P (ΣL )2 and the notation of M L(P (ΣL )2 , P (ΣL )1 ) for the collection
of all intersection-preserving maps from P (ΣL )2 to P (ΣL )1 . By considering
P (ΣL )1 as equal to P (ΣL )2 and abstracting over the indices 1 and 2, we can
follow the idea in [AmiCoeStu98] and introduce the following sets:
Definition 42.
ẽ1 ◦ ẽ2 = e!
2 &e1 (12.1)
∨J ẽi = ∨'J ei (12.2)
∗
e˜1 ∗ ◦ e˜2 ∗ = e!
1 &e2 (12.3)
∗
∧J e˜i ∗
= ∨'J ei (12.4)
1) Property Causation:
" e
ē∗ : L2 → L1 : a2 1→ e.a2 = {a1 ∈ L1 | a1 # a2 }
2) Property Propagation:
! e
ē : L1 → L2 : a1 1→ {a2 ∈ L2 | a1 # a2 }
Given those mappings, ē∗ (a2 ) is the weakest property whose actuality guaran-
tees the actuality of a2 and ē(a1 ) is the strongest property whose actuality is
induced by that of a1 . As such, we have ē(ē∗ (a2 )) ≤ a2 while a1 ≤ ē∗ (ē(a1 )).
Consequently, since this defines an adjunction [CoeMoo00], we immediately ob-
tain the following propositions proved in [CoeMooStu01]:
Proposition 28.
(1) a ≤ b ⇒ ē∗ (a) ≤ ē∗ (b)
(2) a ≤ b ⇒ ē(a) ≤ ē(b)
e
(3) a ≤ ē∗ (b) ⇔ a # b ⇔ ē(a) ≤ b
Given the definition of a unital quantale above, we will show that (ε̄, ∨, ◦) is a
unital quantale and that (ε̄∗ , ∧, ◦) is a unital co-quantale.
Firstly, we show that (ε̄, ∨) is a complete lattice. We define a pointwise ordering
of the maps ē as follows:
It has been proved in [AmiCoeStu98] that this pointwise computable join cor-
responds exactly to the one defined above as ∨J ei . The join semi-lattice we as
such obtain is a complete lattice.
Secondly, we show that (ē, ∨, ◦) is a unital quantale. Since ε is closed under finite
concatenation, & induces a quantale product ◦ by means of definition (12.5).
Note that ◦ by its construction as standard “composition” is associative. The
following distribution laws are straightforward proved:
∀ēi , ē ∈ ε̄ : ē ◦ (∨J ēi ) = ∨J (ē ◦ ēi ) and (∨J ēi ) ◦ ē = ∨J (ēi ◦ ē)
Note that 0̄ can then be conceived as a unit element for the quantale (ε̄, ∨, ◦) ⊆
JL(L, L).
Thirdly, by means of the notion of a Galois adjunction it is well known, see
for example [CoeMoo00], that we obtain the following “so-called” quantaloidal
isomorphism: JL(L, L) ∼ = M L(L, L)coop . Given this quantaloidal isomorphism
we obtain the causal duality: ε̄ ∼
= (ε̄∗ )coop . Henceforth it follows that (ε̄∗ )coop ⊆
M L(L, L) coop
. Given the fact that pointwise joins translate into pointwise
meets, we can declare that indeed ε̄∗ ⊆ M L(L, L) gives (ε̄∗ , ∧, ◦) as a co-
quantale. It is a unital co-quantale since ◦ defined in (12.7) distributes over
arbitrary ∧, defined in (12.8), as follows:
∀e¯i ∗ , ē∗ ∈ ε̄∗ : (ē∗ ) ◦ [∧J (e¯i ∗ )] = ∧J [(ē∗ ) ◦ (e¯i ∗ )] and
[∧J (e¯i ∗ )] ◦ (ē∗ ) = ∧J [(e¯i ∗ ) ◦ (ē∗ )]
and the unit element is given by 0̄∗ .
We have reached the level on which the duality between property causation and
property propagation translates itself in a causal duality between ε̄ and (ε̄∗ )coop ,
for more details we refer to [Coe00, CoeMooStu00].
have one initial property given, is it possible to consider the strongest propaga-
tion with respect to actuality; and only when we have a final property given, is
it possible to consider the cause, i.e. the weakest property that guarantees actu-
ality. In this section we will introduce the notion of an actuality set of properties
and the notion of a property set which allow us to extend our reasoning from the
level of states and properties to the level of sets of properties. As such we will
be able to introduce the maps expressing property set propagation and prop-
erty set causation. Further on, we will firstly explain the relation between these
maps expressing state propagation and property set propagation. Secondly, we
will focus on the kind of property propagations which can be linked to any state
propagation. Note that when we consider a state propagation it is possible that
we are dealing with a situation in which we have a lack of knowledge of the pre-
cise “initial” and “final” realized state, which makes it impossible to pinpoint
the corresponding property propagations immediately. We therefore first need
to introduce the notion of an operational resolution.
Proposition 29. !
Cr : P (Σ) → L : T 1→ Cr (T ) = {a | pE ≤ a (∀E ∈ T )} = ∨(E∈T ) pE
Proof
We start from the fact that we have the following join: for any {Ti }i ⊆ P (Σ) :
∨i Cr (Ti ) = Cr (∪i Ti ) [CoeStu99a, CoeStu99b, Stu99]. From this follows that:
∨(E∈T ) Cr ({E}) = Cr (∪(E∈T ) {E}). Since Cr (∪(E∈T ) {E}) = Cr (T ), also ∀T ∈
P (Σ) : ∨(E∈T ) Cr ({E}) = Cr (T ). !When L is atomistic we have Cr ({E}) = pE
and obtain Cr (T ) = ∨(E∈T ) pE = {a | pE ≤ a (∀E ∈ T )}.8 !
Note here that ∨(E∈T ) pE has to be interpreted as the join of atoms corresponding
to all possible states in T . Since ∀E ∈ T : pE ≤ ∨(E∈T ) pE , we reach the point
that ∨(E∈T ) pE is always actual in any realized state of T . In the above proof
of proposition 29, we expressed that if there is only one possible state such that
T = {E} then Cr (T ) = pE , which means that when we have a physical system
realized in a state E, the strongest actual property!which we can attach to it is
the corresponding state property in L since pE = {a | pE ≤ a}.
The above exposition is only a special case within a more general approach to
operational resolutions and corresponds to the given example in [CoeStu99a,
CoeStu99b] where Σ is considered to be a subset of L without the bottom
element, L is atomistic and Σ is the full set of states. In general, the resolution-
map Cr still attaches to a set of possible states the strongest property which
is definitely actual but this strongest actual property is not necessarily the
join of corresponding state-properties. For a detailed exposition we refer to
[CoeStu99a], but let us just note that in this general case “it is not true that for
any subset T of Σ there is a property that is actual if and only if the state of the
system is in T ” [Stu99] unless every subset T of Σ is closed. For that reason there
exists a theorem which states that any operational resolution factors uniquely
into a closure operator9 on Σ, C : P (Σ) → F(Σ) ⊆ P (Σ), for which C(∅) =
∅ and an order-preserving embedding θ : F(Σ) → L — where F(Σ) is the
collection of C-closed subsets of Σ — such that the image of the embedding θ is
isomorphic to F(Σ) as a complete lattice [CoeStu99a, CoeStu99b, Stu99]. It is
not our intention to handle all technicalities involved, though referring back to
proposition 29 we can say the following. As is explicitly shown in [Stu99], when
Σ ⊂ L without bottom element we have: C : P (Σ) → F(Σ) : T 1→ {E | E ≤ ∨T }
and θ : F(Σ) → L : F 1→ ∨F where θ−1 : L → F(Σ) : a 1→ {E | a ∈
S(E) and E ∈ Σ} is exactly a Cartan map — cf. chapter 4, section 4.7. In
partial order of L, (imCr , ≤) is a join semilattice where the join is given by: ∀{Ti }i ⊆ P (Σ):
∨i Cr (Ti ) := Cr (∪i Ti ) and has as top element Cr (Σ) and as bottom element Cr (∅). Since any
join semilattice is a complete lattice by means of Birkhoffs’ theorem the claim is valid.
8 The last part of this proof will become clear by the end of this section.
9 C is a closure operator on Σ if for all T, T ! ∈ P (Σ): T ⊆ C(T ), T ⊆ T ! ⇒ C(T ) ⊆ C(T ! ),
C(C(T )) = C(T ). This closure is called T0 if in addition the following conditions are satisfied:
C(∅) = ∅ and C({E1 }) = C({E2 }) ⇒ E1 = E2 for any E1 , E2 ∈ Σ; or is called T1 if in addition
the following conditions are satisfied C(∅) = ∅ and C({E}) = {E} for any E ∈ Σ. As such
T1 stands for “all points are closed”. A set T ⊆ Σ is closed if C(T ) = T . See for instance
[CoeStu99a, CoeStu99b] for more details on the use of the closure in relation to operational
resolutions.
CHAPTER 12. OPERATIONAL FOUNDATION & MATH. FRAME 145
Definition 46.
An actuality set is a set of properties of which we are certain that at least one
property is actual.
In the same sense that we may have a lack of knowledge of the precise state
of the system so that the only thing we can say with certainty is that the
system is realized in “a state” contained in T ⊆ Σ; we might also have a lack
of knowledge of the precise properties of a physical system which are actual so
that the only thing we can say with certainty is that the system has at least
one actual property contained in A ⊆ L. Such kinds of lacking knowledge of
the actuality of properties are not expressible on the level of L since we lack the
notion of a classical disjunction in L. On the level of P (L) we have the union
which fulfills the roll of classical disjunction. Since the join map assigns to each
actuality set A the strongest property whose actuality is implied by all elements
of A, it can also be"interpreted as the map"which translates unions " into joins.
More explicitly, if : P (L) → L : A 1→ A = ∨(a∈A) a, then A is always
actual, since as we saw in chapter 4, section 4.3.2, the following implication is
secured in one direction:
∃aj ∈ {ai | i ∈ J} : aj is actual ⇒ ∨J ai is actual.
"
We will be more explicit and call this “join” of an actuality set, A, or in other
words “the strongest property of which A implies actuality” a definite actual
property of an actuality set [CoeSme00].
and codomain to the collection of certain ideals taken in L. Following here the
recent work of B. Coecke in [Coe(nd)a]10 , a motivation for this restriction results
also from the fact that the intersection of two sets of properties on the level of
P (L) gives in the case when a ≤ b the following {a} ∧P (L) {b} = {a} ∩ {b} = ∅.
As explicitly noted in [Coe(nd)a] this clearly does not encode the kind of con-
junction we would like in order to develop a logic of actuality sets. For this
reason we will work with the collection DI(L) of distributive ideals taken in L.
While we will explain below what distributive ideals are, we now introduce our
notion of property set:
Definition 47. The elements of DI(L) ⊆ P (L), i.e. the distributive ideals of
L, are named “property sets”.
An ideal is distributive if it’s an order ideal, i.e. a ≤ b ∈ I ⇒ a ∈ I and
I &= ∅,
" closed under distributive joins, i.e. if A ⊆ I has a distributive join
then A ∈ I [BruLak70, Coe(nd)a]. Referring to " the latter, a subset A of
a (meet-semi) lattice L has a distributive join or A is called distributive if,
[BruLak70, Coe(nd)a],
1) Its supremum" exists,
"
2) ∀b ∈ L : b ∧ A = {b ∧ a | a ∈ A}
Let us now be a little more specific. In the case we work with L atomistic and
ΣL , i.e. the set of atoms of L, we can give the following operational motivation
for the mentioned restriction to DI(L):
a ∈ I ⇒ ∀p ≤ a :"p ∈ I ⇒ µ(a) ⊆ I
µ(a) ⊆ I ⇒ a = L µ(a) ∈ I
10 In his construction B. Coecke uses a purely mathematical result on injective hulls for meet
semi-lattices [BruLak70].
11 A map is a bijection or a one-to-one correspondence if it is injective and surjective. For
f , a bijection, there is an inverse f −1 defined such that f −1 (b) = a iff f (a) = b, which is also
a bijection. With a slight abuse of concepts we can say that f and f −1 define “a bijection”.
CHAPTER 12. OPERATIONAL FOUNDATION & MATH. FRAME 147
a ∈ I ⇔ µ(a) ⊆ I ⇔ µ(a) ⊆ I ∩ ΣL
So we obtain:
And reversely:
As such we showed that ρ and ρ−1 are inverse, so they are bijective, and since
they are isotone12 , they are isomorphisms of complete lattices.
Furthermore, DI(L) ∼ = P (ΣL ) indicates that for every property set in DI(L)
there exists a corresponding T ∈ P (ΣL ). Of course this implies that for every
union and intersection of elements of P (ΣL ) there is a corresponding element
in DI(L). On a more abstract level — not restricted to the L-atomistic case —
the language of DI(L) is in [Coe(nd)a] equipped with logical connectives such
that the logic of actuality sets is obtained. In chapter 13 we will deal with this
logic in detail. For now it is our intention to finish this chapter with a section
on property set propagation and property set causation and an example of what
we consider to be perfect measurement inductions.
called property set causation assigning causes, i.e. largest property sets which
guarantee an actuality set.
Given the propagation-causation duality ẽ = ẽ∗ for states, and our construction
of property set propagation and causation given above, we also have % $e =%$e ∗.
e
As such we can now be more specific about # on DI(L), and say that:
Proposition 30.
e
%
$
e (A1 ) ⊆ A2 ⇔ A1 # A2 ⇔ A1 ⊆ %
$
e ∗ (A2 )
e
Here A1 # A2 expresses that “if the property set A1 is an actuality set before
e, A1 induces that A2 is an actuality set after e”.
The following scheme will explain how each map expressing an induced property
set propagation gives us by means of an operational resolution an associated map
expressing a property propagation of actual properties.
ē " ē " "
L1 −→ L2 : A1 1−→ ē( A1 ) = ($ %e (A1 ))
↑
" ↑
" ↑
" ↑
"
| | ⊥ ⊥
$
#
e $
#
e
DI(L)1 −→ DI(L)2 : A1 1−→ %
$
e (A1 )
CHAPTER 12. OPERATIONAL FOUNDATION & MATH. FRAME 149
Without being very specific we can of course use the fact that in general the
operational resolution Cr and the Cartan map µ are adjoint, to obtain a trans-
lation of this scheme into the case of causation-maps. For this reason we will
use the Cartan map as follows — for specific details we refer to [Coe(nd)c]:
ē∗
L2 −→ L1
| |
µ µ
↓ ↓
ẽ∗
P (ΣL )2 −→ P (ΣL )1
↑ ↑
∼
= ∼
=
↓ ↓
$
#
e∗
DI(L)2 −→ DI(L)1
As is proved in [CoeMooSme(nd)c], these constructions allow us to state the
following proposition:
Proposition 31.
1) [ẽ(µ(a1 )) ⊆ µ(a2 )] ⇔ [µ(ē(a1 )) ⊆ µ(a2 )]
%
2) [ẽ(ρ−1 (A1 )) ⊆ ρ−1 (A2 )] ⇔ [($e (A1 )) ⊆ A2 ]
Finally we can say something of the structure of maps % $e . Note firstly that
for DI(L) ⊆ P (L), % $
e preserves disjunctions. Secondly we can say that the
preservation of the join for ē imposes a continuity condition on %
$
e:
) ) ) )
A= B⇒ %$
e (A) = %
$e (B) (12.12)
Since ε̃ ∼
=%$ε, also ($
%ε, ∨, ◦) is a unital quantale. Similarly as in the case of ε̃,
we see that %$ε is a join-complete lattice with the order pointwise obtained as
follows:
%
$
e1 ≤ %
$
e 2 ⇔ ∀I ∈ DI(L) : %
$
e 1 (I) ⊆ %
$
e 2 (I)
This yields a pointwise computable join as follows,
"
∨J %
$ %
e i : DI(L) → DI(L) : I 1→ DI(L),J ($
e i (I))
Again here ∨J % $
e i expresses the propagation map on DI(L) corresponding to the
induction expressing an arbitrary choice. The quantale product is here similarly
definable as follows:
%
$
e 1◦ %
$
e 2 :=|e2 &e1|
CHAPTER 12. OPERATIONAL FOUNDATION & MATH. FRAME 150
From the above follows that we obtain the following map indicating a quantale
morphism, it preserves ◦ and ∨:
)
∨[−] : %
$
ε → ε̄ : %
$
e 1→ [L → L : a 1→ %
$
e (↓ a)]
Remember that the join-preserving maps ε̄ ⊆ JL(L1 , L2 ) and that their Ga-
lois adjoints ε̄∗ ⊆ M L(L2 , L1 ). In the same line of thought we could lift the
above discussion from the operational to a more theoretical level and discuss
the relations between JL(L1 , L2 ), M L(L2 , L1 ) and the collection of “all” maps
on DI(L) satisfying the continuity condition and disjunction preservance, for
this we refer to [CoeMooStu00].
Where we take
ϕ{b,b! } (T ) = {ϕb (p) !
{ϕb! (p) | p ∈ T, p &≤ b}
" | p ∈ T, p &≤ b } ∪ −1
ρ[ϕ{b,b! } ](A) = { L" B | B ⊆ ϕ{b,b! } (ρ (A))}
∨[ρ[ϕ{b,b! } ]](∨A) = ρ[ϕ{b,b! } ](A)
Construction of Dynamic
Operational Quantum Logic
In this chapter we present recent work under development for which we refer
to [Coe(nd)c, CoeMooSme(nd)c, CoeMooSme(nd)d]. In the first instance we
pay attention to the logic of actuality sets as a static basis which can then be
extended by dynamic implications and conjunctions allowing us explicitly to deal
with notions such as propagation and causation. We will conclude this chapter
with a reinterpretation of the logical description for perfect measurements as
presented in [CoeSme00].
more clear in this context. Furthermore we have to mention that Coecke’s notion of proposition
is exactly our notion of property set.
152
CHAPTER 13. CONSTRUCTION OF DYNAMIC OQL 153
Among the connectives with which the logic of actuality sets is equipped, we
consider first the disjunction and conjunction as follows, [Coe(nd)a],
! $
DI(L) : P (DI(L)) → DI(L) : A 1→ A
" (
DI(L) : P (DI(L)) → DI(L) : A 1→ C( A)
Here C indicates the closure under intersections which ( appoints the smallest
distributive ideal in which a set such as for " example A is embedded. Note
that it has been proved in [Coe(nd)a] that DI(L) is indeed distributive, i.e.
" !
DI(L) (A) is an ! actuality set iff ∃A ∈ A is an actuality set; while DI(L) is
conjunctive, i.e. DI(L) (A) is an actuality set iff ∀A ∈ A is an actuality set:
" (
“For A ⊆ DI(L) we then have that DI(L) (A) = C( A) is an
(
actuality set "if and only if at least one a ∈ A is actual, this since
all elements L B in the disjunctive closure are distributive, and as
such
" if and only if at least!one A ∈ A is an actuality set. Thus,
DI(L) is$disjunctive ... If DI(L) A is an actuality set then at least
one a ∈ A is actual so all A ∈ A are actuality sets. Conversely,
if all A ∈ A are actuality sets ! then for all A ∈ A at least one
aA !∈ A is actual such $ that L {aA | A ∈ A} $ is actual
! and thus
↓ ( L {aA | A ∈ A}) = {↓ aA | A ∈ A} ⊆ A = DI(L) A is an
!
actuality set. Thus, DI(L) is conjunctive ... ”
It is now our intention to take a closer look into this static implication and
to analyze some candidates for a dynamic implication. In the following sec-
tions on implication-connectives, we will follow the ideas presented in [Coe(nd)c,
CoeMooSme(nd)c, CoeMooSme(nd)d].
CHAPTER 13. CONSTRUCTION OF DYNAMIC OQL 154
Here we exchanged E (
∈ Σ for p ∈ ΣL and extended µ to DI(L) — where µ = ρ−1
— to allow µ(A) := µ[A] ⊆ ΣL . And given the fact that p |= (A →DI(L) B)
iff p ∈ µ(A →DI(L) B), we can now state that:
With this information we can reformulate the given static implication →DI(L)
in the following way:
A →DI(L) B = {c ∈ L | (c |= A) ⇒ (c |= B)}
!
This naturally leads to the following proposition:
Proposition 33.
%
µ(A →DI(L) B) = ΣL ⇔ A ⊆ B ⇔ A # B
The proof of this adjunction is implied by µ(A ∩ −) = µ(A →DI(L) −); explicitly
i.e. (µ(A) ∩ µ(B)) ⊆ µ(C) ⇔ µ(B) ⊆ µ(A →DI(L) C) and has been worked out
in [Coe(nd)c].
e
The natural candidate for → to consider is defined as follows:
e
µ(A → B) := {p ∈ ΣL | (p |= A) ⇒ (ẽ({p}) |= B)}
e
We will now prove that →, as defined above, is indeed a good candidate to
e
satisfy proposition 34. First note that A # B iff %
$ %
e (A) ⊆ B iff µ($
e (A)) ⊆ µ(B)
since µ = ρ−1 is on DI(L) even as on L injective and order-preserving, i.e.
isotone.
Proposition 35.
e
%
µ(A → B) = ΣL iff µ($
e (A)) ⊆ µ(B)
Proof:
e
µ(A → B) = ΣL ⇔ c (µ(A)) ∪ ẽ∗ (µ(B))
⇔ µ(A) ⊆ ẽ∗ (µ(B)) = {p ∈ ΣL | ẽ({p}) ⊆ µ(B)}
⇔ ∀p ∈ µ(A) : ẽ({p}) ⊆ µ(B)
⇔ ẽ(µ(A)) ⊆ µ(B)
%
It remains to show now that ẽ(µ(A)) ⊆ µ(B) iff µ($
e (A)) ⊆ µ(B), for which we
refer to proposition 31. !
e
Given the definition of µ(A → B), it is obvious that in case e = 0 this dynamic
implication is reduced to the static implication. Using our notations, we express
this as follows:
Proposition 36.
%
(A → B) = (A →DI(L) B)
And against the background of proposition 34, we can also prove:
Proposition 37.
e
(A →$%
e (A)) = L
CHAPTER 13. CONSTRUCTION OF DYNAMIC OQL 157
Proof
e e
Since (A →$%
e (A)) = L ⇔ µ(A →$%
e (A)) = ΣL , a proof can be established based
∗
%
on µ(A) ⊆ ẽ (µ($
e (A))):
e
%
e (A))
µ(A →$ %
= {p ∈ ΣL |(p |= A) ⇒ (ẽ({p}) |=$
e (A))}
%
= c (µ(A)) ∪ ẽ∗ (µ($
e (A))
= ΣL
!
The last mentioned implication will always be valid. As such we can say that
e
when A → B is valid — i.e. equal to L — it explains that if A is an actuality
set it guarantees that B is an actuality set after we perform induction e. In
e
other words, the “causal relation” # extends into a propagation-implication on
property sets.
Indeed, for µ(A ⊗e B) := ẽ(µ(A) ∩ µ(B)), the following has been proved in
[Coe(nd)c],
e
ẽ(µ(A) ∩ µ(B)) ⊆ µ(C) ⇔ µ(B) ⊆ c (µ(A)) ∪ ẽ∗ (µ(C)) = µ(A → C)
which obviously implies the existence of a unique negation: ¬e (−) = ¬DI(L) (−).
e
As is mentioned in [Coe(nd)c, CoeMooSme(nd)c] if we equip DI(L) with →
and ⊗e for each e ∈ ε, we obtain a collection of commutative
" quantales. More
explicitly, for each induction e, we obtain that (DI(L), DI(L) , ⊗e ) is a commu-
tative quantale but in general not a locale2 . Following [Coe(nd)c], this is shown
as follows:
) )
A ⊗e ( B) = {A ⊗e B | B ∈ B}
DI(L) DI(L)
Proof
" "
µ(A ⊗e ( DI(L) B)) = ẽ(µ(A)) ∩ µ( DI(L) B)
(
= ( B∈B (µ(A) ∩ µ(B)))
ẽ(
= " ẽ(µ(A) ∩ µ(B))
B∈B
= µ( DI(L) {A ⊗e B | B ∈ B})
! " ! !
Note that in case e = 0 we obtain a locale (DI(L), DI(L) , DI(L) ) since DI(L)
"
and DI(L) are distributive [CoeMooSme(nd)c].
To conclude this subsection we say that the dynamic algebra we obtain is the
complete boolean algebra of property sets DI(L) which functions as a left quan-
tale module over %$
ε equipped with extra “propagation”-related connectives. In
the next section we will first analyze the implication for causations and then
look at the structure which is obtained.
13.3.2 Causation-implications
Besides the propagation-implications which, when valid, express that the (for-
ward) causal relation holds between property sets, it is reasonable to look at
causation-implications which, when valid, express that a so-called backward re-
lation holds between property sets. This backward relation is defined on DI(L)
as follows, [Coe(nd)c, CoeMooSme(nd)c, CoeMooSme(nd)d]
Definition
e
50.
A & B := if B is an actuality set after e then was A an actuality set before e.
This leads us to the following proposition:
Proposition 40.
e
A & B ⇔$%e ∗ (B) ⊆ A
e
Let us now introduce the so-called causation-implication ← as follows:
e
µ(A ← B) := {p ∈ ΣL | (p |= A) ⇐ (ẽ({p}) |= B)}
e
(A ← B) %
:= {c ∈ L | (c |= A) ⇐ ($
e (↓ c) |= B)}
!
CHAPTER 13. CONSTRUCTION OF DYNAMIC OQL 160
e
As such we can say that when A ← B is valid — i.e. equal to L — it explains
that if B is an actuality set after e then was A an actuality set before we
e
perform induction e. In other words, the “backward relation” & extends into a
causation-implication on property sets. In particular this suggests that we can
prove the following proposition:
Proposition 42.
e
%
($
e ∗ (A) ← A) = L
Proof
e e
%
Since ($ %
e ∗ (A) ← A) = L ⇔ µ($
e ∗ (A) ← A) = ΣL , we can establish a proof:
e
%
e ∗ (A) ← A)
µ($ %
= {p ∈ ΣL | (ẽ({p}) |= A) ⇒ (p |=$
e ∗ (A))}
%
= µ($e ∗ A) ∪ c (ẽ∗ (µ(A)))
= ΣL
!
%
When we define A ← B := A ←DI(L) B, we obtain:
Proposition 43.
(A →DI(L) B) = (B ←DI(L) A)
Similar as we saw earlier, it has been proved in [Coe(nd)c] that we can find a
e
conjunction-adjoint for ←,
e
(−e ⊗B) = (− ← B)
against the background of the following definition: µ(Ae ⊗B) := µ(A)∩ẽ∗ (µ(B)).
Again we can now interpret (Le ⊗−) as the largest property set which, if it is an
actuality set, guarantees that (−) is an actuality set. In other words, (Le ⊗−)
assigns causes with respect to induction e:
Proposition 44.
(Le ⊗−) = ẽ∗ (µ(−))
Proof
!
It is in this case important to note that while µ(A ⊗e B) = µ(B ⊗e A) we have
µ(Ae ⊗B) &= µ(B e ⊗A) unless e = 0.
which clearly does not imply a unique negation. As such ¬e is quite different
e
from ¬.
e
If we equip DI(L) with ← and e ⊗ for each e ∈ ε, we obtain a collection of
non-commutative co-quantales — ! cf. [Coe(nd)c]. More explicitly, for each in-
duction e, we have that (DI(L), DI(L) , e ⊗) is a non-commutative co-quantale.
Note that
! we obtain a co-quantale since e ⊗ distributes at both sides over ar-
bitrary DI(L) . Just as before, in case e = 0 we obtain a “commutative”
" ! ! "
frame (DI(L), DI(L) , DI(L) ) since then DI(L) and DI(L) are distributive
[CoeMooSme(nd)c].
We see here that µ(A → B) = c (f˜∗ (µ(A))) ∪ ẽ∗ (µ(B)) and as such we can
e|f
e %|e e e|%
recover A ← B as B → A and A → B as A → B. As expected, we find a
non-commutative conjunction operator which is defined on DI(L) as follows in
[CoeMooSme(nd)c]:
Also here we then see that (A ⊗e B) can be recovered as (A ⊗e◦% B). A variation
on this second kind of bi-labeled implications will prove its importance later on
for which we now introduce some useful definitions:
e
µ(A →
→ B) := {p ∈ ΣL | (p |= ẽ(µ(A))) ⇒ (p |= B)}
%◦e
= (A → B)
= c
(ẽ(µ(A))) ∪ µ(B)
e
← B) :=
µ(A ← {p ∈ ΣL | (p |= ẽ(µ(A))) ⇐ (p |= B)}
= ẽ(µ(A)) ∪ c (µ(B))
Proof
!
As explained in [CoeMooSme(nd)b], we can consider the subjunctive condition-
als as encoding the implicative process where we first modify the state to render
it compatible with the antecedent before testing the subsequent. Referring back
to chapters 5 and 11, the Sasaki projections which are the maps corresponding
to ideal measurements of the first kind are specific examples of what fa can
be. Hence we are now in the position to show more explicitly what Hardegree’s
interpretation of the Sasaki hook as a Stalnaker conditional means. We will
show that µ(a −→sub b), under the conditions 1) and 2) for a Sasaki projection
ϕa , yields the set of states in which the Sasaki hook ϕ∗a (b) is an actual property.
ϕa
p |= (a −→sub b) ⇔ p ∈ Ka or [ϕa (p) |= (a −→ b)]
⇔ p ∈ Ka or [ϕa (p) |= a ⇒ ϕa (p) |= b]
⇔ p ∈ Ka or ϕa (p) &|= a or ϕa (p) ≤ b
⇔ p ∈ Ka or ϕa (p) &|= a or p ≤ ϕ∗a (b)
⇔ p ∈ Ka or ϕa (p) &|= a or p ≤ a! ∨ (a ∧ b)
Note that now we will use the notation e{b,b! } to express that we induce b or b! on
a physical system, and as before we will use ϕ{b,b! } to express the corresponding
map on P (ΣL ). It has to be noted that K{b,b! } expresses in this case the set of
states p for which ϕb (p) or ϕb! (p) would yield 0.
e{b,b! }
µ(A −→ sub B) = {p ∈ ΣL | p ∈ K{b,b! } or (ϕ{b,b! } ({p}) |= A −→ B)}
= {p ∈ ΣL | p ∈ K{b,b! } or
(ϕ{b,b! } ({p}) |= A ⇒ ϕ{b,b! } ({p}) |= B)}
= K{b,b! } ∪ c (ϕ∗{b,b! } (µ(A))) ∪ ϕ∗{b,b! } (µ(B))
And where ϕ{b,b! } (T ) yields the union of sets of Sasaki projections, ϕ∗{b,b! } (T )
yields the intersection of sets of Sasaki hooks. Let us explain this in more detail.
Since we know that the sasaki hook ϕ∗b is the adjoint of the sasaki projection
ϕb , i.e.,
∀p1 ∈ T1 : p1 &≤ b! , ∀p2 ∈ T2 : b ∧ (b! ∨ p1 ) ≤ p2 ⇔ p1 ≤ b! ∨ (b ∧ p2 )
we have by means of the covering law:
∀p1 ∈ T1 , ∀p2 ∈ T2 : ϕb (p1 ) = p2 or ϕb (p1 ) = 0 ⇔ ϕ∗b (p2 ) = p1
Now we can prove that ϕ{b,b⊥ } = ϕ∗{b,b! } . In order to do so we first need the
following proposition:
Proposition 46.
ϕb (T1 ) = {b ∧ (b! ∨ p1 )|p1 ∈ T1 }\{0} ⊆ T2 ⇔ T1 ⊆ {b! ∨ (b ∧ p2 )|p2 ∈ T2 } = ϕ∗b (T2 )
Proof
⇒: If ϕb (T1 ) ⊆ T2 then ∀p1 ∈ T1 ∃p2 ∈ T2 such that ϕb (p1 ) = p2 or ϕb (p1 ) = 0
hence ϕ∗b (p2 ) = p1 and as such T1 ⊆ ϕb (T2 ).
⇐: If T1 ⊆ ϕ∗b (T2 ) then ∀p1 ∈ T1 ∃p2 ∈ T2 such that p1 = ϕ∗b (p2 ) hence
ϕb (p1 ) = p2 or ϕb (p1 ) = 0 and as such ϕb (T1 ) ⊆ T2
!
Proof
{b ∧ (b! ∨ p1 )|p1 ∈ T1 }\{0} ∪ {b! ∧ (b ∨ p1 )|p1 ∈ T1 }\{0} ⊆ T2
⇔
{b ∧ (b! ∨ p1 )|p1 ∈ T1 }\{0} ⊆ T2 and {b! ∧ (b ∨ p1 )|p1 ∈ T1 }\{0} ⊆ T2
⇔
T1 ⊆ {b! ∨ (b ∧ p2 )|p2 ∈ T2 } and T1 ⊆ {b ∨ (b! ∧ p2 )|p2 ∈ T2 }
⇔
T1 ⊆ {b! ∨ (b ∧ p2 )|p2 ∈ T2 } ∩ {b ∨ (b! ∧ p2 )|p2 ∈ T2 }
CHAPTER 13. CONSTRUCTION OF DYNAMIC OQL 165
!
Of course we can also express now what ρ[ϕ{b,b! } ]∗ (A) ∈ DI(L) means:
#
ρ[ϕ{b,b! } ]∗ (A) = {ϕ∗b (a) | a ∈ A} {ϕ∗b! (a) | a ∈ A}
DI(L)
Given the fact that we also have: ϕb ∨ ϕb! = ϕ∗b ∧ ϕ∗b! on the level of L; we
immediately obtain the “adjoint” of the scheme represented in section 12.5.
Analyzing this axiom, we see that there are two different kinds of predicates,
IN D(−) has induction terms in its range while In(−) has property terms in its
range. Exactly by using these predicates we can distinguish different levels in
our logic. The property terms themselves are wff’s of ls-OQL, which has been
analyzed in chapter 8, and as such they allow us to express for instance In(∨J ai )
or In(∧J ai ). Similarly, it is also possible for the induction terms to express for
instance the arbitrary choice of two inductions IN D(e ∨ f ) or the concatenation
of two inductions IN D(e&f ). Since of course an actuality set A ∈ DI(L) can
contain more than one property, they are represented by means of ⊕a∈A In(a)
which expresses the disjunction with respect to the actuality of a specific set of
properties. Note that ⊕a∈A In(a) differs from the formula In(∨A) in that the
latter expresses a definite actual property. Contrary to this way of dealing with
actuality sets, we could also have introduced a predicate ranging over actuality
sets but that would divert us from using the explicit level of ls-OQL. Note that
⊕ expresses a classical disjunction. For ⊗ and ' things are more complicated
than first thought and will be analyzed in detail below.
4 Assuming that we work with one physical system we will drop the index s.
CHAPTER 13. CONSTRUCTION OF DYNAMIC OQL 166
In [CoeSme00] we used ' as a linear logical connective and can show now that
its usage was correct to express the content of a map ρ[ϕ{b,b! } ], but not correct
to express any propagation of “actuality sets” in case of a perfect measurement
e
induction. If we want to express the latter we should of course use → with for
instance e = ϕ{b,b! } . Let us now analyze the usage of ' in the former case. The
content of the map ρ[ϕ{b,b! } ] expresses that ρ[ϕ{b,b! } ] as an action on an actu-
ality set A yields ρ[ϕ{b,b! } ](A). Based on ideas used in linear logic — see next
chapter — we need a causal implication which exactly expresses the consump-
tion of initial resources and simultaneously the production of final resources.
Analyzing this statement precisely, it says that the initially given actuality set
must after performing an induction have been transformed into a final actuality
set. Concretely this means that we cannot work with an implication based on
e
the causal relation # as introduced earlier. Instead we need a much stronger
relation which attaches to a “final” property set, if it is an actuality set after e,
exactly the needed “initial” actuality set before e. Let us now introduce such a
relation:
Definition 51.
e
A## B ⇔ ($%e (A) = B)
%
Having the connective ⊗e in mind and since we know that (A ⊗e L) =$
e (A) we
can now immediately say the following:
e
A#
# B ⇔ (A ⊗e L) = B
Where IN D(ϕ{b,b! } ) ⊗ ⊕a∈A In(a) is seen as: “the induction ϕ{b,b! } forms an
action on actuality set A”. Next we then see that axiom 13.1 expresses the
following:
Or in other words:
ϕ{b,b! }
A #
# ϕ{b,b! } (A)
168
CHAPTER 14. COMPARISON WITH LINEAR LOGIC 169
(weakeningL) A→B
A,a→B (weakeningR) A→B
A→B,a
A,a,a→B A→a,a,B
(contractionL) A,a→B (contractionR) A→a,B
Indeed, dropping weakening and contraction implies that linear formula cannot
be duplicated or contracted at random, in other words, our resources are re-
stricted. As an example we can say that we do not eat the same apple twice or
demolish the same piece of chalk twice. Though there is another consequence
of importance, dropping those two structural rules leads to two kinds of dis-
junctions and conjunctions. We will obtain a so-called additive disjunction ⊕
and additive conjunction & and a so-called multiplicative disjunction ℘ and
multiplicative conjunction ⊗. The following left and right rules will make their
differences clear. Although we further on prefer introduction and elimination
rules, the left and right-rules are used here because they give a more explicit view
on the symmetry of ⊗ with ℘ and ⊕ with & — for more details on symmetry
we refer to [Sam00].
A,a,b→B A→a,b,B
(⊗ L) A,a⊗b→B (℘ R) A→a℘b,B
We will show the well known result that by allowing the structural rules and by
using (⊗) we can express the (&)-rules and vice versa:
For & R:
(prem) (prem)
A → a, B A → b, B
(⊗R)
A, A → a ⊗ b, B, B
(exchanges R and contractions L,R)
A → a ⊗ b, B
For ⊗ R:
(prem)
A2 → b, B2
(prem) (weakenings L,R)
A1 → a, B1 A2 , A1 → b, B2 , B1
(weakenings L,R) (exchanges L,R)
A1 , A2 → a, B1 , B2 A1 , A2 → b, B1 , B2
(&R)
A1 , A2 → a&b, B1 , B2
CHAPTER 14. COMPARISON WITH LINEAR LOGIC 170
For ⊗ L:
(prem)
A, a, b → B
(&L)
A, a&b, b → B
(&L)
A, a&b, a&b → B
(contraction L)
A, a&b → B
For & L:
(prem)
A, a → B
(weakening L)
A, a, b → B
(⊗L)
A, a ⊗ b → B
For ⊕ R:
(prem)
A → a, B
(weakening R, exchanges R)
A → a, b, B
(℘R)
A → a℘b, B
For ℘ R:
(prem)
A → a, b, B
(⊕R)
A → a ⊕ b, b, B
(⊕R)
A → a ⊕ b, a ⊕ b, B
(contraction R)
A → a ⊕ b, B
For ⊕ L:
(prem) (prem)
A, a → B A, b → B
(℘L)
A, A, a℘b → B, B
(exchanges L and contractions L,R)
A, a℘b → B
For ℘ L:
(prem) (prem)
A1 , a → B1 A2 , b → B2
(weakenings L,R) (weakenings L,R)
A1 , a, A2 → B1 , B2 A2 , b, A1 → B2 , B1
(exchanges L) (exchanges L,R)
A1 , A2 , a → B1 , B2 A1 , A2 , b → B1 , B2
(⊕L)
A1 , A2 , a ⊕ b → B1 , B2
The meaning of ℘ is not that clear in the linear logical literature, but will become
clearer when we follow J.Y. Girard in his construction that every atomic formula
of his linear logical language has by definition a negation (−)⊥ . Running a bit
ahead of our story, the meaning of a℘b will now come down to the situation
where “if” not a is given “then” b is given and “if” not b is given “then” a is
given. Of course this explanation is linked to the commutative case where a
linear logical implication is defined as a⊥ ℘b := a ' b which by transposition
equals b⊥ ' a⊥ . In the same sense, only in the commutative case where a ⊗ b
equals b ⊗ a, does it make sense to say that a ⊗ b comes down to simultaneous
given resources.
We will be a little more specific on the linear logical implication and analyze
the underlying philosophical ideas as presented in [Gir89] where of course ' is
defined by means of ℘. What is important is that the linear implication should
mimic exactly what happens when a non-iteratable action is being performed.
Where we conceive a non-iteratable action to be such that after its performance
the initial resources are not available any more as initial resources. The linear
implication should as such express the consumption of initial resources and
simultaneously the production of final resources. Indeed, as stated in [Gir95],
the linear implication ' expresses a form of causality: a ' b is to be conceived
as “from a get b”. More explicitly, [Gir89]:
“A causal implication cannot be iterated since the conditions are
modified after its use; this process of modification of the premises
(conditions) is known in physics as reaction.” (p.72)
If we understand this correctly, the act of consumption and production is called
a non-iteratable action while the process of modification of initial conditions,
the deprivation of resources, is called reaction. The idea of relating action
and reaction is in a sense metaphorically based on Newton’s action-reaction
principle in physics. Girard uses this metaphor also when he explains why
every formula has by definition a linear negation which expresses a duality or
change of standpoint, [Gir89]:
“action of type A = reaction of type A⊥ .” (p.77)
CHAPTER 14. COMPARISON WITH LINEAR LOGIC 172
Leaving this action-reaction debate aside, we can now explain why our given
interpretation of non-iterability is quite subtle. First note that in Girard’s
standard linear logic, % a ' a is provable from an empty set of premises. As
such ' represents in a ' a the identity-action which does not really change
resources, but only translates initial ones into final ones. And although we could
perform the action twice in the following sense: a1 ' a2 ' a3 — for convenience
we labeled the resources —, this still does not count as an iteratable action since
a1 is to be conceived as an initial resource, different from a2 , the final resource
of the first action. But we can go further in this discussion and follow Girard in
stipulating the fact that we may still encounter situations in which the picture of
“consuming all initial resources”, does not hold. Linear logic henceforth allows
also the expression of those actions which deal with stable situations and which
are iteratable. In the latter case the use of exponentials is necessary. As such it
becomes possible to define the intuitionistic implication ⇒ as follows a ⇒ b =!a
' b [Gir89], where ! expresses the linear exponential called of course which
gives !a the meaning that a’s use as a resource is unlimited.
In [Gir89], Girard reveals the link between states, transitions and linear im-
plication. In particular it is the following statement which places the linear
implication in a for us interesting context, thinking of course about the above
discussed dynamic extension of OQL, [Gir89]:
“In fact, we would like to represent states by formulas, and transi-
tions by means of implications of states, in such a way that the state
S ! is accessible from S exactly when S ' S ! is provable from the
transitions, taken as axioms” (p.74)
In [Gir89] this statement applies to for instance systems such as Petri nets, Tur-
ing machines, chessboard games, lists of molecules, lists of beliefs, etc. Focusing
“in this sense” on physical systems, and using ' for transitions of states, we
will further on investigate how ' can be conceived in the context of our logic of
actuality sets. Before we enter into that discussion we will first say something
CHAPTER 14. COMPARISON WITH LINEAR LOGIC 173
about non-commutative linear logic. Note as well that there is much more to
standard linear logic than our short exposition of its syntax and some of its
underlying dynamic ideas.
Given a unital quantale (Q, ∨, ⊗), with 1 as the neutral element with respect to
⊗, we can state the following straightforward proposition, [ConMir96, Ros90a,
Yet90]:
Proposition 47.
The endomorphisms a ⊗ −, − ⊗ a : Q → Q have right adjoints, a ' − and
− ◦− a respectively:
a ⊗ c ≤ b"iff c ≤ a ' b c ⊗ a ≤ b iff c ≤ b ◦− a
a ' b =" {c ∈ Q : a ⊗ c ≤ b}
b ◦− a = {c ∈ Q : c ⊗ a ≤ b}
We know that in the standard linear logic as presented by Girard, the following
holds a ' b = a⊥ ℘b = (a ⊗ b⊥ )⊥ = (b⊥ ⊗ a)⊥ = b℘a⊥ = b⊥ ' a⊥ . In cyclic
linear logic where we have now two implications, things change in the sense that
we obtain now:
(a ⊗ b)⊥ = b⊥ ℘a⊥ (a℘b)⊥ = b⊥ ⊗ a⊥
a ' b = a⊥ ℘b b ◦− a = b℘a⊥
CHAPTER 14. COMPARISON WITH LINEAR LOGIC 174
To be more explicit, the linear negation can be interpreted in the unital quantale
(Q, ∨, ⊗) by means of a cyclic dualizing element, which can be defined as follows,
[ConMir96, Ros90a, Ros90b, Yet90]
Definition 52.
An element ⊥∈ Q is dualizing if:
⊥ ◦− (a '⊥) = a = (⊥ ◦− a) '⊥, for every a ∈ Q
Definition 53.
An element ⊥∈ Q is cyclic if:
a '⊥=⊥ ◦− a, for every a ∈ Q
Here the operation − '⊥ or similar ⊥ ◦− − is called the linear negation and
can be written as (−)⊥ . Note that a unital quantale with a cyclic dualizing
element ⊥ is called a Girard quantale, the notion has been introduced by Yetter
in [Yet90]. These Girard quantales can be equipped with modal operators to
interpret the linear logical exponentials and form as such a straightforward
semantics for the cyclic as well as standard linear logical syntax. In the latter
case a ' b = b ◦− a. The disadvantage cyclic linear logic offers is that it “is
still not non-commutative enough to properly express time’s arrow” [Yet90].
Indeed, where ' is conceived as a causal implication it would have been nice
to conceive ◦− as expressing past causality, though this interpretation is too
misleading according to [Gir89]. In a way we agree with him since in cyclic
linear logic a ' b and a⊥ ◦− b⊥ are the same — in the sense that they are both
equal to a⊥ ℘b.
There is also here much more to say about non-commutative linear logic, espe-
cially since research on it is now in full development and expansion, though we
feel that our overview suffices for our needs.
or in other words:
e "
A '· B := %
DI(L) {C ∈ DI(L) | (C |=$
e (A)) ⇔ (C |= B)}
e ! e
:= (A →
→ B) DI(L) (A ← ← B)
e
In particular we see that µ(A '· B) = [c (ẽ(µ(A)) ∪ µ(B)] ∩ [c (µ(B)) ∪ ẽ(µ(A))];
this allows us to prove the following proposition:
e e
Proposition 48. A '· B = L ⇔ A #
#B
Proof
e e
A '· B = L ⇔ µ(A '· B) = ΣL
⇔ [c (ẽ(µ(A)) ∪ µ(B)] ∩ [c (µ(B)) ∪ ẽ(µ(A))] = ΣL
⇔ ẽ(µ(A)) = µ(B)
⇔ %
$e (A) = B
e
⇔ A# #B
e
Note that we constructed A '· B in such a way that it can express the same
e e ! e
thing as A⊗e L = B. We see as well that if B ◦−
· A := (A ←
← B) DI(L) (A →
→ B),
we obtain:
e e
A '· B = B ◦−
· A
CHAPTER 14. COMPARISON WITH LINEAR LOGIC 176
%
Indeed, if no action is performed then A '· B = L iff A = B.
e
We explicitly constructed this implication '· against the underlying idea of
e
the relation ##. We could of course also have followed another direction. What
we mean is that instead of concentrating on A ⊗e L = B we could also have
concentrated on Le ⊗B = A. While in the first case one focuses on a condition
with respect to propagation, the latter yields a condition with respect to causa-
tion. Since those two approaches can yield different implications and since we
think that the ideas behind the standard linear logical implication touch on a
propagation-condition more closely than on a causation-condition, we worked
out the first possibility.
Another point which deserves our attention is the difference between the ap-
proach of dynamic OQL and the linear logical approach. Contrary to dynamic
OQL, it should be well understood that linear logic is not a temporal logic, no
preconceptions of time or processes is built into it. More explicitly, [Gir89]:
“Linear logic is eventually about time, space and communication,
but is not a temporal logic, or a kind of parallel language: such
approaches try to develop preexisting conceptions about time, pro-
cesses, etc.. In those matters, the general understanding is so low
that one has good chances to produce systems whose aim is to avoid
the study of their objects ... The main methodological commitment
is to refuse any a priori intuition about these objects of study, and
to assume that (at least part of) the temporal, the parallel features
of computation are already in Gentzen’s approach, but are simply
hidden by taxonomy.” (p.104)
As such dynamic OQL started out with a different methodology. The objects
of study are well-known as physical systems and their properties just as the in-
ductions performable on physical systems. In a sense this information has been
encrypted in the formulas we used. All dynamic propagation- and causation-
implications, have been labeled by inductions, and this is different from the
linear logical implications which are used to express any (non-specified) transi-
tion. Exactly this forms quite an argument against applying the linear logical
implications in a context of dynamic OQL. Even though it is formally possible
to construct a kind of implication which expresses the intuitive ideas behind
the linear logical implication, the natural “dynamic operational quantum logic”
e
which needs to be further developed, should in the first instance not use '· or
e e e
· but rather the more naturally arising → and ←.
◦−
Afterword
177
References
178
REFERENCES 179
[JauPir69] J.M. Jauch and C. Piron, “On the Structure of Quantal Proposition
Systems”, Helvetica Physica Acta, 42, 842-848 (1969)
[NagNew58] E. Nagel and J.R. Newman, Gödels Proof, New York University
Press, New York (1958); Dutch translation by J.M. Debrot: De Stelling
van Gödel, Het Spectrum, Utrecht/Antwerpen (1975)
[NemAnd94] I. Németi and H. Andréka, “General Algebraic Logic: A Perspec-
tive on “What is Logic””, in D.M. Gabbay (ed.), What is a logical System?,
Clarendon Press, Oxford, 393-443 (1994)
[Nii87] I. Niiniluoto, “Varieties of Realism”, in P. Lahti and P. Mittelstaedt
(eds.), Symposium on the Foundations of Modern Physics, 1987, The
Copenhagen Interpretation 60 Years after the Como Lecture, Joensuu, Fin-
land, 6-8 August 1987, World Scientific, Singapore (1987)
[Nii99] I. Niiniluoto, Critical Scientific Realism, Oxford University Press, UK
(1999)
[Nis80] H. Nishimura, “Sequential Method in Quantum Logic”, Journal of Sym-
bolic Logic, 45, 339-352 (1980)
[Nis94a] H. Nishimura, “Proof Theory for Minimal Quantum Logic I”, Interna-
tional Journal of Theoretical Physics, 33, 103-113 (1994)
[Nis94b] H. Nishimura, “Proof Theory for Minimal Quantum Logic II”, Inter-
national Journal of Theoretical Physics, 33, 1427-1443 (1994)
[Oge92] E. Oger, “Putnams kritiek van het metafysische realisme”, in E. Oger
and F. Beukens (eds.), Denken in alle staten, negen profielen van heden-
daagse Amerikaanse filosofen, Pelckmans, Kapellen (1992)
[Pat95] R. Patterson, Aristotle’s Modal Logic, Essence and Entailment in the
Organon, Cambridge University Press, UK (1995)
[Pau58] W. Pauli, Handbuch der Physik, Vol.5, Part 1: Prinzipien der Quan-
tentheorie I (1958); English translation by P. Achuthan and K. Venkatesan:
General Principles of Quantum Mechanics, Springer Verlag, Berlin (1980)
[Pie80] G. Piéraut-Le Bonniec, The Development of Modal Reasoning, Genesis
of Necessity and Possibility Notions, Academic Press, New York (1980)
REFERENCES 188
[Put71] H. Putnam, Philosophy of Logic, George Allen & Unwin LTD., London
(1971)